AN EXAMINATION OF CONCEPTUAL KNOWLEDGE USING NEAR-INFRARED
SPECTROSCOPY AND ELECTROENCEPHALOGRAPHY
by
Brian Duffels
B.Sc. Honours University of Northern British Columbia, 2004
M.Sc. University of Alberta, 2010

THESIS SUBMITTED IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
IN
PSYCHOLOGY

UNIVERSITY OF NORTHERN BRITISH COLUMBIA
June 2022
© Brian Duffels, 2022

Abstract
Traditionally, models of conceptual knowledge have relied upon amodal theories that
largely overlook how environmental stimuli are converted into amodal representations and how
perceptions reactivate these representations and translate them back into subjective modal
experiences. Developed more recently, grounded cognition theories propose that physical
experiences and conceptual knowledge rely, at least in part, on the same brain regions. Thus,
conceptual knowledge is hypothesized to be experienced through the reactivation of the same
brain regions that are activated during physical experiences with the environment. Furthermore,
if these grounded hypotheses are correct, researchers should be able to observe predictable
influences of grounded information on brain activity as well as participant response latencies and
accuracy in experimental conditions.
To this end, three experiments were conducted testing these hypotheses using semantic
categorization tasks while simultaneous recordings were taken using functional near infrared
spectroscopy (NIRS) and electroencephalography. It was hypothesized that the influence of
automatically reactivated grounded information would be facilitatory (i.e., resulting in faster and
more accurate responses for semantically richer words) when it was task congruent, but would be
inhibitory (i.e., resulting in slower and less accurate responses for semantically richer words)
when it was task incongruent, thus illustrating the automatic simulation of grounded information
in the processing and retrieval of conceptual knowledge.
NIRS was employed to monitor event-related patterns of prefrontal cortex (PFC)
hemodynamics associated with these tasks. It was hypothesized that trials with high levels of
task-relevant semantic information would be discernably different than those with low levels or
those trials high in task-incongruent information. That is, given the high levels of task-relevant

~ ii ~

semantic information, these trials should be comparatively easier, thus requiring less activity in
the PFC, resulting in less pronounced hemodynamic responses. Electroencephalography was
employed to monitor the full-scalp event-related patterns of brain activity associated with the
experimental tasks. It was hypothesized that event-related potential deflections and scalp
topography would be able to discern qualitatively and quantitatively different patterns of activity
as a function of the amount and relevance of grounded information obtained through physical
and emotional experiences with the word stimuli’s referents.
The behavioural, accuracy, and electroencephalography data generally support these
hypotheses. When a stimulus’s grounded information is high and task relevant, participants
responded more quickly and accurately, and had discernably different patterns of brain activity
than when a stimulus’s grounded information was low and task relevant. When a stimulus’s
grounded information was high and task-irrelevant, participants were slower and less accurate,
and exhibited patterns of brain activity that reflected both the additional semantic information
and the additional processing necessary to reconcile the task incongruence. Unfortunately, data
obtained from NIRS failed to illustrate meaningful condition differences. Possible reasons for
this are discussed in detail in Chapter 3.
Collectively, the data presented in this dissertation serve to advance and extend the
claims made by grounded cognition theorists by illustrating the automatic simulation of
information obtained through interactions with the environment. Further research is required to
extend this work to other brain regions and to develop NIRS methods that can address these
research questions.

~ iii ~

TABLE OF CONTENTS
Abstract…………………………………………………………………………………………... ii
Table of Contents……………………………………………………………………………….... iv
List of Tables.…………………………………………………………………………………… vii
List of Figures…………………………………………………………………………............... viii
List of Abbreviations……………………………………………………………………...…….. xii
Dedication……………………………………………………………………………………… .xiv
Chapter 1: Introduction…………………………………………………………………………… 1
References…...…………………………………………………………………………… 6
Chapter 2: Investigating the Neurological Correlates of Grounded Conceptual Knowledge……. 7
Program of Research……………………………………………………………………. 25
Near Infrared Spectroscopy……………………………………………………... 25
NIRS in Cognition………………………………………………………………. 30
Electroencephalography………………………………………………………… 34
Experimental Procedure………………………………………………………… 41
Independent Variables…………………………………………………... 42
Dependent Variables……………………………………………………. 43
Summary………………………………………………………………………………... 44
References………………………………………………………………………………. 47
Chapter 3: Using Functional Near Infrared Spectroscopy to Investigate Grounded Semantics…. 68
Prefrontal Cortex Functions……………………………………………………………... 70
Cognitive Investigations of the PFC using NIRS………………………………………... 77
Experimental Predictions……………………………………………………………….. 84

~ iv ~

Method………………………………………………………………………………….. 87
Participants……………………………………………………………………… 87
Apparatus……………………………………………………………………….. 88
Lexical Decision Tasks………………………………………………………….. 88
Semantic Categorization Tasks…………………………………………………. 89
Analyses………………………………………………………………………… 90
Results…………………………………………………………………………………... 92
Behavioural……………………………………………………………………... 92
Near Infrared Spectroscopy……………………………………………………... 96
Discussion……………………………………………………………………………... 100
Behavioural Results……………………………………………………………. 101
NIRS Results…………………………………………………………………... 102
Conclusion……………………………………………………………………... 108
Figures…………………………………………………………………………………. 110
References……………………………………………………………………………... 122
Tables………………………………………………………………………………….. 138
Chapter 4: Electrophysiological Evidence of Automatically Activated Grounded Semantic
Knowledge…………………………………………………………………………….. 142
Hypotheses…………………………………………………………………………….. 148
Summary of Hypotheses……………………………………………………….. 151
Method……………………………………………………………………...…………. 152
Participants…………………………………………………………………….. 152
Apparatus…………………………………………………………………….... 152

~v~

Analyses……………………………………………………………………...... 151
Results……………………………………………………………………..................... 154
Behavioural……………………………………………………………………. 155
Electroencephalography……………………………………………………….. 156
Discussion……………………………………………………………………............... 164
Electroencephalography Results………………………………………………. 165
Figures……………………………………………………………………..................... 180
References……………………………………………………………………............... 203
Chapter 5: Conclusion…………………………………………………..................................... 216
Behavioural Results…………………………………………………............................. 216
Near Infrared Spectroscopy…………………………………………………................. 219
Electroencephalography…………………………………………………...................... 222
Conclusions…………………………………………………......................................... 224

~ vi ~

LIST OF TABLES
TABLE 1: Stimuli for Experiment 1…………………………………………………………… 138
TABLE 2: Stimuli for Experiment 2……………………………………………….................... 140
TABLE 3: Stimuli for Experiments 3 & 4……………………………………………………… 141

~ vii ~

LIST OF FIGURES
Figure 1: Clockwise, from top-left: NIRS time course data for left PFC [O2Hb] (μM), [HHb] (μM),
[THb] (μM), and Sat% (%) in Experiment 1. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation……………………...…………… 110
Figure 2: Clockwise, from top-left: NIRS time course data for right PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 1. Data were cleaned up with a 200 ms
average moving window for ease of display and interpretation. ………………………… 111
Figure 3: Plotted left hemispheric [THb] difference between high- and low-frequency words for
the first 14 seconds following stimulus presentation In Experiment 1. Error bars are Bayesian
99% credible intervals. …………………………………………..……………………… 112
Figure 4: Clockwise, from top-left: NIRS time course data for left PFC [O2Hb] (μM), [HHb] (μM),
[THb] (μM), and Sat% (%) in Experiment 2. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation. ………………….……………… 113
Figure 5: Clockwise, from top-left: NIRS time course data for right PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 2. Data were cleaned up with a 200 ms
average moving window for ease of display and interpretation. ………………………… 114
Figure 6: Plotted right hemispheric [O2Hb] differences between low-BOI words and no-go trials
for the first 14 seconds following stimulus presentation In Experiment 2. Error bars are
Bayesian 99% credible intervals. …………………………...…………………………… 115
Figure 7: Clockwise, from top-left: NIRS time course data for left PFC [O2Hb] (μM), [HHb] (μM),
[THb] (μM), and Sat% (%) in Experiment 3. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation. ……………………….………… 116

~ viii ~

Figure 8: Clockwise, from top-left: NIRS time course data for right PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 3. Data were cleaned up with a 200 ms
average moving window for ease of display and interpretation. ………………………… 117
Figure 9: Plotted right hemispheric [THb] differences between high- and low-EE words for the
first 14 seconds following stimulus presentation in Experiment 3. Error bars are Bayesian
99% credible intervals…………….…………………………..……………….………… 118
Figure 10: Clockwise, from top-left: NIRS time course data for left PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 4. Data were cleaned up with a 200 ms
average moving window for ease of display and interpretation. ………………………… 119
Figure 11: Clockwise, from top-left: NIRS time course data for right PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 4. Data were cleaned up with a 200 ms
average moving window for ease of display and interpretation. ………………………… 120
Figure 12: Plotted left hemispheric [O2Hb] differences between high-EE words and no-go trials
for the first 14 seconds following stimulus presentation in Experiment 4. Error bars are
Bayesian 99% credible intervals…………………….…………………...……………… 121
Figure 13: Electrode placement reference for all experiments. View is from above, with a person
facing towards the top of the screen. Image obtained from Geodesic Sensor Net Technical
Manual (2007). ………………………………………………...…………...…………… 180
Figure 14: Topographic display of all ERP waveforms for Experiment 2 with all electrodes aligned
as per Figure 1………………………………………………...….……………………… 181
Figure 15: Topographic scalp plot for Experiment 2. Times (ms) are time since stimulus
presentation. Red denotes positivity and blue denotes negativity. Scale is -10 to 10 µV.
………………………………….………………………..……………………………… 182

~ ix ~

Figure 16: Difference waveform (High BOI – Low BOI) for electrode 8 in Experiment 2. Error
bars are 99% Bayesian credible intervals. ……………………………….……………... 183
Figure 17: Topographic display of all ERP waveforms for Experiment 3 with all electrodes
aligned as per Figure 1………………...……………...………………………………… 184
Figure 18: Topographic scalp plot for Experiment 3. Times (ms) are time since stimulus
presentation. Red denotes positivity and blue denotes negativity. Scale is -10 to 10 µV.
…………………….……………………….……..………………….………………….. 185
Figure 19: Difference waveform (High EE – Low EE) for electrode 1 in Experiment 3. Error bars
are 99% Bayesian credible intervals………………………..…………….……………... 186
Figure 20: Difference waveform (High EE – Low EE) for electrode 3 in Experiment 3. Error bars
are 99% Bayesian credible intervals……………………………..……….……………... 187
Figure 21: Difference waveform (High EE – Low EE) for electrode 5 in Experiment 3. Error bars
are 99% Bayesian credible intervals………………………………..…….……………... 188
Figure 22: Difference waveform (High EE – Low EE) for electrode 11 in Experiment 3. Error
bars are 99% Bayesian credible intervals..……………………………….……………... 189
Figure 23: Difference waveform (High EE – Low EE) for electrode 13 in Experiment 3. Error
bars are 99% Bayesian credible intervals..……………………………….……………... 190
Figure 24: Difference waveform (High EE – Low EE) for electrode 14 in Experiment 3. Error
bars are 99% Bayesian credible intervals..……………………………….……………... 191
Figure 25: Difference waveform (High EE – Low EE) for electrode 16 in Experiment 3. Error
bars are 99% Bayesian credible intervals………………………….……………………. 192

~x~

Figure 26: Difference waveform (High EE – Low EE) for electrode 24 in Experiment 3. Error
bars are 99% Bayesian credible intervals……………….………………………………. 193
Figure 27: Difference waveform (High EE – Low EE) for electrode 25 in Experiment 3. Error
bars are 99% Bayesian credible intervals..……………………………….……………... 194
Figure 28: Difference waveform (High EE – Low EE) for electrode 26 in Experiment 3. Error
bars are 99% Bayesian credible intervals..……………………………….……………... 195
Figure 29: Difference waveform (High EE – Low EE) for electrode 29 in Experiment 3. Error
bars are 99% Bayesian credible intervals..……………………………….……………... 196
Figure 30: Topographic display of all ERP waveforms for Experiment 4 with all electrodes
aligned as per Figure 1..…………………….…………...……………………………… 197
Figure 31: Topographic scalp plot for Experiment 4. Times (ms) are time since stimulus
presentation. Red denotes positivity and blue denotes negativity. Scale is -10 to 10 µV.
………………………………………………...........…………………….……………... 198
Figure 32: Difference waveform (High EE – Low EE) for electrode 3 in Experiment 4. Error bars
are 99% Bayesian credible intervals……....………………………..…….……………... 199
Figure 33: Difference waveform (High EE –Low EE) for electrode 5 in Experiment 4. Error bars
are 99% Bayesian credible intervals. ...…………………………..….………………….. 200
Figure 34: Difference waveform (High EE – Low EE) for electrode 11 in Experiment 4. Error
bars are 99% Bayesian credible intervals...……………….…………………………….. 201
Figure 35: Difference waveform (High EE – Low EE) for electrode 29 in Experiment 4. Error
bars are 99% Bayesian credible intervals...……………………………….…………….. 202

~ xi ~

LIST OF ABBREVIATIONS
[x]

concentration of x

ANOVA

analysis of variance

BOI

body-object interaction

BOLD

blood-oxygen level dependent

CW

continuous wave

DL-PFC

dorsolateral prefrontal cortex

EE

emotion experience

EEG

electroencephalography

EMG

electromyograph

ERP

event related potential

fMRI

functional magnetic resonance imaging

fNIRS

functional near infrared spectroscopy

FPC

frontopolar cortex

HHb

deoxygenated hemoglobin

IFC

inferior frontal cortex

ITI

inter-trial interval

LDT

lexical decision task

LPC

late positive component

O2Hb

oxygenated hemoglobin

OF-PFC

orbitofrontal prefrontal cortex

PFC

prefrontal cortex

PSS

perceptual symbol systems

~ xii ~

Sat%

regional tissue oxygen saturation

SCT

semantic categorization task

SE

standard error

THb

total hemoglobin

VM-PFC

ventromedial prefrontal cortex

~ xiii ~

DEDICATION
I have such an amazing network of loving and supportive friends, family, and colleagues
without whom this work would never have been completed. My mother, who forgives my every
trespass and thinks the world of me regardless of my antics. My brother, who sees the best in me
despite knowing better. My beautiful wife, who believes in me even when I am at my worst and
supports me in every way. This work is dedicated to you. To the rest of my friends and family
who saw me through this process: you may never know what your support meant to me. To all of
you: thank you for believing in me.

~ xiv ~

1
CHAPTER 1: INTRODUCTION
The four word categorization experiments presented here were conducted in an effort to
address questions surrounding the nature of human conceptual knowledge. Two neuroimaging
technologies (electroencephalography and near-infrared spectroscopy; EEG and NIRS) were
employed concurrently with behavioural (response latency and response accuracy)
measurements to contribute to this growing body of research. A grounded cognition perspective
was adopted in the formation and interpretation of the hypotheses in each experiment.
Collectively, the data presented in Chapters 3 and 4 provide evidence that the processing of
conceptual information involves the automatic reactivation of semantic information obtained
from physical and emotional experiences with one’s environment. Importantly, these data (see
Chapter 3, Experiment 3) demonstrate that reactivation of grounded knowledge is automatic
even when it is not congruent with task demands. Additionally, the data acquisition and analysis
techniques employed in the research presented demonstrate a new and comparatively accessible
methodology to address fundamental questions in cognitive psychology. In sum, the data
presented in the following chapters serve as a contribution to and extension of grounded
cognition hypotheses as well as a novel method for the analysis and presentation of
psychophysiological data.
One of the key predictions made by grounded cognition theory is that the neurological
basis of conceptual knowledge should vary as a function of the nature of the knowledge
(Barsalou, 1999; Pecher et al., 2011; Pulvermuller, 2013). This prediction is the result of the
grounded hypothesis that conceptual knowledge is experienced through the partial reactivation of
at least some of the same sensory, motor, and emotion brain regions that were active during the
acquisition of that knowledge (Barsalou, 1999). This hypothesis is based largely on Barsalou’s

2
(1999) Perceptual Symbol Systems (PSS) framework. The PSS proposes that conceptual
knowledge is formed and represented symbolically in sensory, motor, and emotion brain regions
during experiences with one’s environment. Importantly, different concepts should be realized in
the brain differentially in both qualitative and quantitative manners. For example, while the
concepts of flashlight and fork may be primarily represented neurologically in visual and motor
brain regions, one may predict that individuals would have more experience with forks than with
flashlights. As a result, one may predict that the neurological activation of concepts that have
greater amounts of semantic information obtained through physical experiences with the words’
referents would have greater levels of activity in the implicated motor regions. Conversely, while
the concepts of fork and sunset may both be represented neurologically in visual brain regions,
sunset would not have any motor representation due to individuals having no bodily interactions
with sunsets. As a result, one may predict that the characteristic patterns of neuronal activity
associated with processing such concepts would be qualitatively different. A more thorough
exploration of the theoretical basis for this project can be found in Chapter 2.
Across four experiments, participants’ lexical semantic knowledge was activated through
one lexical decision task (LDT) and three semantic categorization tasks (SCT). In the LDT,
participants are presented with a letter string and are to identify if the string spells a real word. In
principle, one need only recognize the correct spelling of a word to complete this task
successfully. In a SCT, participants are presented with a real word and are asked to make a
categorical decision about that word (e.g., Is this word a concrete noun?: Newcombe et al., 2012;
Is this word easily imageable?: Hargreaves et al., 2012; Is this a verb?: Noguchi et al., 2002). To
successfully complete a SCT trial, participants must activate their conceptual representations of a
presented word and compare representation against current task demands. By manipulating the

3
semantic dimensions along which stimuli vary, researchers can experimentally observe the role
of that dimension on task completion. In the experiments presented in Chapters 3 and 4, the
stimuli varied along the dimensions of body-object interaction (BOI) ratings (i.e., a measure of
the ease of bodily interacting with a word’s referent) and emotional experience (EE; i.e., a
measure of the ease by which a word evokes emotion information; Newcombe et al., 2012). The
effect of these variables on conceptual processing were measured behaviourally through analyses
of response latencies and accuracy in the LDT and SCTs. Detailed predictions and more
thorough reviews of these semantic variables are found in Chapters 2, 3, and 4.
To measure the qualitative and quantitative processing differences of various concepts,
two neuroimaging techniques were employed concurrently. First, near-infrared spectroscopy
(NIRS) was used to measure hemodynamic activity in the prefrontal cortex (PFC). This
technology works by emitting light in the near-infrared spectrum into target tissues and recording
the amount and wavelength of light that exits nearby tissue (Sato et al., 2013). This technique is
effective due to the fact that oxygenated and deoxygenated hemoglobin preferentially absorb
different wavelengths of near-infrared light. Thus, by emitting two or more wavelengths of light
into tissue of interest and observing the comparative amounts of each wavelength that escape,
inferences can be made about the composition of regional blood flow (Elwell & Hebden, 1999).
Such data is informative in cognitive research due to regional perfusion overshoot (Verner et al.,
2013). That is, when neurons fire, they consume resources such as glucose from the regional
blood supply. When compensating for this resource depletion, oxygen-rich arterial blood fresh
blood is delivered to the area. As such, researchers can make inferences about task-dependent
neuronal processing as a function of changes in regional hemoglobin oxygenation. A more
thorough review of this technology and hypotheses specific to NIRS are presented in Chapter 3.

4
When neurons fire, there is a small voltage discharge that is associated with the action
potential. Electroencephalography (EEG) was used in the experiments in Chapter 4 to detect
these polarity changes at the scalp. EEG, as a technology has been long validated for its ability to
show task- and condition-dependent regional and amplitude differences in electrical brain
activity (Khrizman, 1973). EEG devices can record upwards of 1000 times per second, making it
an ideal functional neuroimaging technique for observing brain activity that is uniquely
associated with a single trial’s brain activity (event-related potentials; ERP). In the four
experiments presented in the subsequent chapters, full-scalp ERPs were recorded concurrently
with NIRS recordings because the two technologies are compatible; NIRS uses fiber-optic cables
to carry signals and, as a result, does not interfere with the sensitive electrical recordings made
by EEG devices. Particular ERP components of interest and specific hypotheses related to the
ERP data are found in Chapter 4.
Broadly speaking, the general hypotheses across all experiments are as follows. First,
response latencies will be shorter and more accurate for stimuli with high levels of grounded
semantic richness (see Chapter 2) except when the nature of the grounded semantic richness is
incongruent with task demands (i.e., Chapter 3 – Experiment 3). Second, NIRS will reveal
greater amounts of cerebral hemoglobin oxygenation and total saturation accompanied with
lower amounts of deoxygenated hemoglobin (see Chapter 3) for more effortful trials. That is, for
those trials that are low in task-relevant grounded semantic richness or for those trials high in
task-irrelevant grounded semantic richness. Finally, EEG will reveal a) quantitative differences
insofar as various ERP components directly reflect cognitive effort (i.e., low grounded semantic
richness trials will be harder to process) and b) qualitative differences insofar as the scalp
distribution of various ERP components will differ as a function of the nature of task demands

5
and the manipulated dimension of grounded semantic richness. More thorough and specific
predictions can be found in Chapter 3 (behavioural and NIRS) and Chapter 4 (EEG).
To demonstrate the degree to which the physiological hypotheses are supported by the
data, a novel Bayesian approach is introduced in Chapters 3 and 4. Specifically, difference waves
were calculated across time between high- and low-grounded semantic richness trials and then
displayed with Bayesian 99% credible intervals. This is an application of Nathoo et al.’s (2018)
validation of this technique to calculate variance across time and space (the authors originally
used this technique to plot reach and grasp trajectories). This technique is applicable to highdensity data arrays without incurring the risk of a Type I error as is typical in frequentist
analyses. By displaying difference waves with Bayesian credible intervals, a viewer is able to
simply look at the data and observe when, across time, non-zero condition differences occur. The
data can then be interpreted in relation to the original physiological waveform components. This
is particularly well suited to the current program of research due to the many different
dimensions along which psychophysiological data can vary (see Chapter 2 for further
discussion).
In sum, the data presented in the chapters that follow stand as an application and
extension of grounded cognition hypotheses with regard to semantic processing of conceptual
knowledge. Further, neurological predictions are tested through concurrent applications of two
different neuroimaging technologies and analyzed with a novel methodology. Collectively, these
experiments will contribute to and strengthen grounded interpretations of human conceptual
knowledge while illustrating the neurological correlates thereof using non-invasive technologies
and introduce a sophisticated and accessible method for interpreting the data.

6
References
Barsalou, L. W. (1999). Perceptual symbol systems. Behavioral and Brain Sciences, 22(4), 577660.
Elwell, C. & Hebden, J. (1999). Near infrared spectroscopy. Retrieved from:
www.ucl.ac.uk/medphys/research/borl/intro/nirs
Khrizman, T. P. (1973). Characteristics of intercentral relationships in electrical processes of the
brain in 2-to 3-year old children during voluntary motor acts. Voprosy Psychologii.
Nathoo, F. S., Kilshaw, R. E., & Masson, M. E. (2018). A better (Bayesian) interval estimate for
within-subject designs. Journal of Mathematical Psychology, 89¸ 1-9.
Noguchi, Y., Takeuchi, T., & Sakai, K. L. (2002). Lateralized activation in the inferior frontal
cortex during syntactic processing: Event related optical topography study. Human Brain
Mapping, 17, 89-99.
Pecher, D., Boot, I., & Van Dantzig, S. (2011). Abstract concepts: Sensory-motor ground,
metaphors, and beyond. In Brian Ross (Ed.) The Psychology of Learning and Motivation,
54, 217-248.
Pulvermuller, F. (2013) Semantic embodiment, disembodiment or misembodiment? In search of
meaning in modules and neuron circuits. Brain & Language, 127, 86-103.
Sato, H., Yahata, N., Funane, T., Takizawa, R. Katura, T., Atsumori, H., … Kasai, K. (2013). A
NIRS-fMRI investigation of prefrontal cortex activity during a working memory task.
NeuroImage, 83, 158-173.
Verner, M., Herrmann, M. J., Troche, S. J., Roebers, C. M., & Rammsayer, T. H. (2013).
Cortical oxygen consumption in mental arithmetic as a function of task difficulty: A nearinfrared spectroscopy approach. Frontiers in Human Neuroscience, 7, 217.

7
CHAPTER 2:
Investigating the Neurological Correlates of Grounded Conceptual Knowledge
Classical accounts of cognition claim that knowledge gained through sensory, motor, and
emotional experience is translated into cognitive representations that are amodal, and hence
stripped of the associated features of the original lived experience (e.g., the shape or colour of
what was visually observed) (Fodor, 1983). In other words, conceptual knowledge involves the
transformation of sensorimotor and emotion information into a form of abstract cognitive
representational code comparable to a computer language (Niedenthal, et al., 2005). Amodal
accounts of cognition have, however, been challenged as being more descriptive than predictive
(see Barsalou, 1999). Amodal theories are likely sufficient for describing patterns of observed
data from conceptual processing research and have guided the development of cognitive models
that identify and quantify key factors in semantic processing, such as a word’s age of acquisition
(e.g., Dehaene, 2011; Machery, 2016). However, it has been argued that amodal theories of
cognition have failed to explain how sensorimotor and emotion information is transformed into
amodal cognitive representations, how external stimuli passively prompt retrieval of these
representations, and how they are transformed back to and experienced as subjective experiences
(Glenberg, et al., 2000); and that amodal theories do not offer an explanation of how or why
humans evolved amodal representation processes. This issue is frequently discussed as the
grounding problem (e.g., Harnad, 2003; Pecher & Zwaan, 2005), which is the debate of how
symbolic representations of conceptual knowledge are mapped from the environment, into
mental symbolic representation, and then from these representations back into a state with
references to the real world. Without verifiable cognitive mechanisms capable of bridging one’s
multimodal interactions with the physical environment (arising through sensory, motor, and

8
emotion experience) and the cognitive processing thereof, it is argued that classical cognitive
theories have implicit shortcomings (Johnson, 2017).
Barsalou (1999) proposed an influential grounded cognition framework that he called
Perceptual Symbol Systems (PSS). According to this framework, partially reactivated neural
networks in sensorimotor and emotion regions, or “simulations”, are thought of as perceptual
symbols. From this perspective, unconscious automatic processing (e.g., sensation and
perception) and conscious, effortful processing (e.g., decision making and selective attention)
both rely, at least partially, on the same brain regions. By accepting this proposition, proponents
of grounded cognition have alleviated the need to account for the transformation of sensorimotor
and emotion representations to amodal cognitive representations that putatively represent
previous experiences; the brain forms, stores, reactivates, and combines these perceptual symbols
to accomplish dynamic cognitive processing that allows for efficient interactions between
humans and their environments.
Barsalou’s (1999) PSS framework further proposes that the sensory, motor, and emotion
networks formed through one’s physical interactions with the environment can serve
symbolically in cognition. Take, for example, one’s conceptual knowledge of dog. Every time
one perceives a dog, there are corresponding patterns of neural activation (e.g., brain regions
responsible for the sight of the dog’s shape and colour, the smell of a dog, the feeling of a dog’s
fur, etc.). A subset of these experiences is stored as the perceptual symbol for dog. Langacker
(1987) described this process as superimposing transparencies atop one another: anything
common to all or most transparencies will remain clear, whereas anything unique to a single
transparency will become blurred and indistinct over time. Across one’s life, there will be
commonalities between these separate dog experiences. As a result, the neural activations that

9
are common across all such experiences will develop stronger interconnections, thereby
producing a reliable sense of meaning for one’s concept of dog.
Furthermore, Barsalou's PSS framework allows for perceptual symbols to be recombined
dynamically. For example, in addition to dog one may develop a concept of purple resulting
from all of one’s perceptual and physical experiences with purple stimuli. Although one may not
ever actually perceive a purple dog, by activating both dog and purple perceptual symbols
concurrently one can conceptualize a purple dog. By experiencing the activation of both purple
and dog perceptual symbols together, one may form a connection between the two concepts,
creating a new perceptual symbol without direct perceptual or physical experience with its
referent. Cognitive activity, according to the PSS framework, can form new perceptual symbols
during offline cognitive processing.
The PSS framework is compatible with other grounded frameworks of cognition. For
example, Anderson et al. (2013) propose that higher cognitive operations are softly assembled.
That is, cognitive operations are not discrete entities. Rather, cognitive operations such as
memory are realized simultaneously across multiple brain regions with no discrete beginning,
end, or location. Cognition from this perspective is a dynamic process that must always be
understood contextually as an interaction between one’s history, context, and current task
demands. If correct, this framework would imply that, while there may be neurological
commonalities across multiple instances of the same process, there is no a priori reason for
neuropsychological researchers to study discrete temporal epochs or scalp topographies
associated with particular processes. Together, these frameworks suggest that conceptual
knowledge is assembled contextually and dynamically. The PSS framework is, however,
comparatively specific in its predictions. A perceptual symbol cannot truly exist if it is

10
spontaneously, dynamically, and differentially formed for each activation. Neuroimaging
researchers could dissociate between these predictions by examining the stability in patterns of
activation across trials and across experimental replications.
Pulvermuller (2013) has asserted that any theory that relies on amodal cognitive
representations is fundamentally incomplete and exists in defiance of modern evidence.
Recently, theorists have begun to accept that cognition’s primary purpose is to guide interactions
between one’s body and environment (Anderson et al., 2013). In this sense, cognition is best
understood as the product of evolution. Higher cognition must be thought of as the product of
simpler systems that evolved earlier (Pecher et al., 2011). Evolution does not lead to the creation
new brain structures without building on pre-existing structures, so it seems reasonable to
conclude that higher level cognitive processes are built upon operations of sensation and action
(Pecher et al., 2011). The most accurate cognitive theories are, as a result, likely to be those that
are heavily reliant on sensation, perception, and action to explain cognitive processes.
Specifically, human cognition would be the product of evolutionary adaptations that
improved reproductive success by increasing the efficiency of environmental processing.
Sensation and action would evolve first, providing the basis for higher cognition (e.g., selective
attention, memory, decision making, etc.). All subsequent cognitive processes would rely on this
foundation as they developed. This idea is in direct contrast with the classical cognition idea that
there is no such relationship between cognition and the environment (Meteyard et al., 2012). The
theoretical position that many, if not all, cognitive processes are dependent upon and resultant of
the interactions between one's body and environment is known variously as embodied or
grounded cognition (Barsalou, 1999).

11
From a grounded cognition perspective, many cognitive processes rely on sensory and
motor brain regions (e.g., Barrett, & Lindquist, 2008; Barsalou, 1999; Caligiore & Fischer,
2012); and some theorists even argue that all cognitive processes rely on such brain regions (e.g.,
Chimero, 2013; Favela, 2014). The underlying idea is that one’s conceptual knowledge is largely
due to one’s previous sensorimotor interactions with the environment (Barsalou, 1999). While
humans interact with the environment, a subset of the sensory and motor cortices that are active
during perception and action begin to form networks due to co-occurrence effects (Hebb, 1949).
Offline cognition (i.e., mental processing of information that is not immediately necessary for
action, like reading written language), it is argued, is possible through subsequent partial
reactivations of these networks (Siakaluk et al., 2008; Barsalou, 2016; Barsalou, 1999, calls these
networks ‘simulators’, and partial reactivation ‘simulations’). That is, one’s thoughts and
memories of an experience are, at least partially, represented neurologically in the same brain
regions that are active during that experience (for example, see Hargreaves et al., 2012).
Research supporting the automatic use of information gathered through physical
interactions with the environment can be found in many different areas of psychology. Such
examples include the following:
•

While standing on a balance board and making estimates of perceived objects
(e.g., how tall is the Eiffel Tower?), participants leaning to the left routinely gave
lower estimates than those standing upright or leaning to the right (Eerland et al.,
2011). By leaning to the left, concepts associated with ‘left’ were presumably
activated, such as the lower end of a number line. By having low numbers
activated, participants were primed to make lower estimates.

12
•

People thinking of the future tend to lean forward, while those thinking of the past
lean backwards (Miles et al, 2010). People conceptualize ‘future’ as being ahead
of them, and ‘past’ as being behind them.

•

The perception of a cup’s handle has been found to prime related motor responses
in unrelated tasks (Tucker & Ellis, 2001).

•

Pianists have shown activation in motor regions of their brain while recognizing
recordings of their own performances (Repp & Knoblich, 2004).

•

Wheeler (2000) presented participants with words either visually or auditorily,
and when asked to remember these words in an fMRI scanner, activation was
observed in the visual and auditory cortices in each condition respectively, which
lends support that context-dependent learning effects may be partially explained
with a grounded cognition perspective (see Conway, 2002).

Collectively, the evidence presented above makes a strong argument that information
gained through a history of physical interactions with the environment is fundamental to
cognitive processing. However, thus far the evidence supports online cognition: mental
operations designed for immediate action utilizing information gained through similar
experiences
The findings summarized above illustrate how bodily information, accrued over one’s
lifetime, influences people’s perception of and interaction with their current environment.
However, one of the strong claims of grounded cognition is that the same underlying bodily
processes are responsible for offline processing as well. Conceptual information, knowledge, and
memories are fundamentally embodied. Research in the area of language processing more
generally, and lexical or semantic decision making more specifically, provides further evidence

13
that supports this claim. From a grounded cognition perspective, people understand written
language if they have had some degree of bodily interaction with or reaction to a word’s referent.
People’s reliance on grounded information to comprehend sentences is well documented
in the literature. To understand a sentence’s meaning, people are thought to use words and
grammar as cues to activate and combine the perceptual symbols to simulate the described
referents (Zwaan & Madden, 2005). Evidence for this assertion is seen in patterns of data that are
not explainable from a classical cognition theory of language comprehension. Rather, from an
embodied cognition perspective, people represent a sentence’s subject differently based on
context (Stanfield & Zwaan, 2001; Zwaan et al., 2002). For example, when reading the sentences
“the eagle is in the nest” or “the eagle is in the sky”, classical accounts would suggest that the
‘eagle’ in both sentences is the same amodal code. However, people identify images of the
sentences’ subject more quickly if the image depicts the subject in a sentence-congruent
orientation, suggesting that readers prime different representations of eagles in both instances.
For instance, after reading “the eagle is in the nest”, people identify a picture of an eagle sitting
more quickly than they identify a picture of an eagle flying. This finding is congruent with an
embodied explanation in that people would be simulating words’ referents rather than activating
amodal codes. Similarly, people represent the distance between two objects described in a
sentence differently based on context (Morrow & Clark, 1988). Specifically, when asked to read
“the mouse approached the fence" or “the tractor approached the fence”, people were more likely
to judge the distance between the two objects to be greater if the sentence depicted a tractor than
if it did a mouse. From a grounded cognitive perspective, this finding can be interpreted as
people simulating both referents, and activating knowledge of the size of the referents, their
likely speed of locomotion, and the distances from which each referent has typically been

14
viewed. These findings are not predicted by amodal cognitive theories in that each word should
activate the same amodal code regardless of context.
Further evidence of grounded effects in sentence comprehension is provided by Kaschak
and Glenberg (2000), who presented readers with sentences containing made-up verbs derived
from nouns (e.g., crutched, paper clipped, etc.). Their findings suggested that people would
accept a novel verb as sensible if the physical object that constituted the verb was large enough
to interact with the other referents in the sentence. For example, “Dennis paper clipped the potato
to Janice” would be viewed as nonsensical, whereas “Janice crutched the potato to Dennis” was
considered to be legitimate. As the verb in both instances is made up, the sentences should be
unintelligible if people are simply using words as cues to access stored information. Similarly,
people reading sentences describing two tasks being done concurrently are more likely to find
the sentences as sensible if the tasks can physically be carried out together, and insensible if they
cannot (De Vega et al., 2004). For example, “Morgan chewed gum while painting a fence” is
sensible, whereas “Cameron played the drums while painting a fence” was deemed insensible.
Both sentences, grammatically, are sensible and understandable. However, if people are
simulating the words’ referents, then there would be a conflict in the effort to simulate painting a
fence and playing drums concurrently. Finally, it was also found that altering one’s current
physical state can alter the speed of sentence processing (Glenberg et al., 2005). Readers were
asked to hold a pen with their teeth without lip contact (engaging the same muscles as a smile
would), or with their lips, without touching their teeth (engaging the same muscles as a frown
would). All readers processed sentences more quickly when the sentence contained information
that was congruent with their simulated mood (e.g., processing sentences that contained ‘happy’
information while they were holding a pen that engaged muscles involved in smiling).

15
Abstract metaphor comprehension is another example of where applying a grounded
cognition perspective has aided in understanding language processing. Metaphors, by their
nature, are concepts that are understood in terms of other concepts. Phrases such as “it went over
my head” or “it is beneath me” suggest that abstract concepts are often understood in reference to
one’s body. Lakoff and Johnson (1980, 1999) have proposed a framework wherein conceptual
knowledge of abstract metaphors is understood through previous grounded experiences with
one’s body and environment. The concept ‘up’, for example, is commonly used to convey many
different concepts. For example, up can convey more (“prices are going up”), control (“she
works above me”), goodness (“things are looking up”), or rationality (“she raised the level of
discourse”). The concept of ‘idea’ is another example. Specifically, ideas are conceptualized as
physical objects when they are discussed; ideas get into one’s head or are difficult to get across
to other people; ideas need to be captured. By relating abstract concepts to grounded concepts,
people engage in a simulation process that enables comprehension.
Wilson and Gibbs (2007) empirically examined embodiment effects in abstract metaphor
comprehension. Their data supported two central hypotheses. First, people who process
metaphorical language prime the physical actions implied by the metaphor. Second, engaging in
behaviours prime people’s processing of behaviour-congruent metaphors. Participants who had
performed or had imagined performing an action were faster at identifying a metaphor that
implied that action. For example, participants who imagined squeezing an object were faster to
recognize the metaphor “grasp the concept” than any metaphor containing a different action
(e.g., “put two and two together”), or no action at all (e.g., “you are in over your head”). Even
though the metaphors did not require one to actually imagine physical actions to understand their
meaning, the physical event underlying the metaphor was still primed, and vice versa. This is

16
further evidence that information obtained from bodily experiences is important in language
comprehension.
It could be argued that the evidence presented above regarding abstract metaphor
comprehension reflects an automatic mimicry process underpinning cognition. Rather than
activating symbolic representations, people may be parroting certain stimuli to activate
information obtained through bodily experiences. Halberstadt et al. (2009) demonstrated a topdown embodied effect that challenges this assertion. Participants in their study were shown facial
expressions of emotion that were computerized midpoint morphs of happy and angry faces.
These faces, as a result, did not look like authentic expression of emotion. Each computergenerated face was, however, labelled as being happy or angry. Participants were to learn which
faces corresponded to which emotion, despite the faces’ lack of correlation with the labelled
emotion. The dependent variable was electromyography (EMG) measurements of facial muscles.
The researchers tested whether participants would automatically mimic the computer faces. At
test, when the participants had to look at a face and identify its assigned emotion without a label
present, the EMG recordings showed that people were activating muscles responsible for smiling
when seeing a “happy” face morph, and the muscles for angry expressions when seeing an
“angry” face morph. Participants were not mimicking. Rather, when seeing a stimulus that they
knew to be happy, they were simulating their previous bodily experiences with happiness. This is
precisely in accordance with a perceptual symbol system account of grounded cognition.
The studies reported in Chapters 3 and 4 have examined the behavioural implications of
grounded cognition on the influence of meaning, as measured by grounded semantic richness, in
visual word processing. Pexman (2012) defined semantic richness, generally, as the varying
amounts of information, along any given semantic dimension, that are elicited when processing

17
visually presented words. The idea of semantic richness may be used fruitfully in better
understanding the observed effects of various grounded semantic dimensions in visual word
recognition, including the dimensions of sensory experience (e.g., Juhasz, et al., 2011),
imageability (e.g., Bennet et al., 2011, Cortese & Fugett, 2004; Yap, et al., 2015), body-object
interaction (BOI; e.g., Siakaluk, et al., 2008a; Siakaluk, et al., 2008b), concreteness (e.g., Kanske
& Kotz, 2007), and emotion information (Newcombe et al., 2012; see also Muraki, et al., 2019).
These observed effects can thus be understood as arising because words that contain more
grounded semantic richness due to many body-environment interactions are processed more
quickly than words that contain less grounded semantic richness, due to fewer body-environment
interactions. According to the PSS framework, facilitation of grounded semantic processing
would be due to more efficient simulations of grounded semantic knowledge associated with
greater levels of sensorimotor and emotional experiences with the environment contained in the
semantically richer words. To continue the example given above, one may predict that, in visual
word recognition tasks (such as semantic categorization tasks; SCT), the concept dog would be
more easily processed and be responded to more quickly than the concept purple, because there
are many more multimodal interactions a human can have with dogs (e.g., visual, tactile,
olfactory, emotional) than with the colour purple (e.g., visual, but perhaps in special cases,
emotional).
As noted, response latency advantages for words with greater levels of semantic richness
have been reliably found in the literature (e.g., Strain, et al., 1995; Wellsby, et al., 2011). The
influence of imageability effects have been found to be reliably facilitatory in lexical processing.
For example, Cortese and Fugett (2004) had participants rate the imageability of 3,000
monosyllabic words. The participants provided the imageability ratings more quickly for words

18
to which they assigned high imageability ratings than for words to which assigned lower
imageability ratings (see also Binder, 2007; Li, et al., 2020; and Plaut et al., 1996 for other
observations of facilitatory imageability effects). Juhasz et al. (2011) found that sensory
experience ratings (the degree to which a sensory or perceptual experience is elicited by a
presented word) account for a significant amount of variance in lexical decision task (LDT)
response latencies. BOI has also been found to facilitate lexical processing (e.g., Siakaluk et al.,
2008a; Tillotson, et al., 2008; Tousignant & Pexman, 2012; Van Havermaet & Wurm, 2014;
Yap, et al., 2012). For example, Siakaluk et al. (2008b) observed shorter response latencies to
words rated high in BOI in both visual (e.g., “Is this letter string a real word?”) and phonological
LDTs (e.g., “Does this sound like a real word?”). Yap and Seow (2014) found that words rated
high in both positive and negative valence were responded to more quickly in a visual LDT.
However, other research has found variability in the effects of valence ratings on semantic
processing response latencies (e.g., Estes & Adelman, 2008; Kuperman et al., 2014; Vinson et
al., 2014; Yap & Seow, 2014). Emotion information has alternatively been measured through
emotional experience ratings (the degree to which a word evokes emotion information; EE).
Higher levels of EE have been found to facilitate response latencies and accuracy rates of
abstract noun categorization (Newcombe et al, 2012; see also Moffat, et al., 2015; Siakaluk, et
al., 2014; Siakaluk, et al., 2016). Furthermore, the facilitatory influence of grounded semantic
information has been observed outside of visual word recognition (e.g., tool identification: Witt,
et al., 2010; mood-congruent sentence comprehension: Glenberg, et al., 2005). These findings
suggest that grounded semantic richness is multimodal and includes multiple sensory (e.g.,
imageability, sensory experience), motor (e.g., BOI), and emotion (e.g., valence, EE) dimensions
of grounded bodily experience with the environment.

19
In the visual word recognition studies summarized above, response patterns can be
accounted for by the feedback activation framework of Hino and Lupker (1996). According to
the feedback activation framework, visually recognizing words involves the activation of three
types of units: orthographic, phonological, and semantic. Orthographic units process the
spellings of words, phonological units process the sounds of words, and semantic units process
the meanings of words. This framework further proposes that the activation of one set of units
may influence the activation of the other sets of units through feedforward and feedback
activation. The feedback activation framework may be used to interpret SCT response latency
patterns. In a SCT, all stimuli are real words that vary along a particular semantic dimension
(e.g., imageability, BOI, or EE), and a specific categorization decision is made (e.g., “Is the word
concrete or not?”). According to the feedback activation framework, words with greater levels of
semantic richness should activate more semantic units than words with lower levels of semantic
richness, thereby allowing for faster settling of these units and thus facilitating categorization
responses. Response latency differences are therefore hypothesized to be the result of varying
levels of grounded semantic richness. For example, words referring to objects typically
associated with increased physical interactions, as operationalized by higher BOI ratings, should
activate a greater number of semantic units (i.e., the richness of one’s sensorimotor simulation)
related to such interactions, thereby facilitating responses to such words. In contrast, words
referring to objects that are not typically associated with physical interactions should activate
fewer semantic units, thereby leading to slower response latencies to such words.
One possible interpretation of the facilitatory effects described above is that increased
levels of semantic richness should always automatically facilitate response latencies. However,
results presented by Newcombe, et al. (2012) challenge this strict interpretation, such that

20
facilitatory grounded semantic richness effects occur when higher levels of grounded semantic
information are congruent with the decision criterion but are inhibitory when they are
incongruent with the decision criterion. For example, in a SCT using the decision criterion, “Is
this a concrete noun?”, faster response latencies and more accurate categorizations were made
for concrete nouns rated higher in imageability and BOI than concrete nouns rated lower on both
dimensions. These facilitatory effects are likely related to the idea that imageability and BOI
ratings are diagnostic of concrete nouns, and hence congruent with the specific decision criterion
used in that experiment. Importantly, however, concrete nouns rated high in EE were categorized
less accurately as being concrete nouns than those words rated low in EE. This inhibitory effect
is likely related to the idea that emotion information (as measured in this study using the EE
dimension) is diagnostic of abstract nouns (Vigliocco, et al., 2009), and thus incongruent with
the decision criterion. As such, the automatic activation of more emotion information for
concrete words rated high in EE would signal that an abstract noun was being processed, thereby
increasing the incongruency between what EE is diagnostic of (abstract concepts) and what the
word actually is (a concrete noun). This incongruence leads to slower response latencies and/or
increases the probability that participants make a wrong SCT response.
The inhibitory influence of task-incongruent semantic richness is difficult to reconcile
with feedback models such as that proposed by Hino and Lupker (1996). In a feedback model of
lexical processing, stimuli possessing more semantic richness should generate more feedback to
other units, thus facilitating the resolution of contextual decision criteria. That is, it should be
easier to recognize whether a stimulus satisfies a decision criterion if it has more semantic
richness. This prediction is not supported by the data presented above. Barsalou’s PSS
framework, however, can account for these data. Specifically, a visually presented stimulus will

21
activate the perceptual symbol(s) that have been associated and experienced with the stimulus,
and it is that symbolic representation that is compared against the decision criterion. Therefore, it
can be hypothesized that there should be a gradient of congruence between automatically
activated perceptual symbol(s) and the contextual decision criterion. However, grounded
frameworks such as the PSS suggest that greater amounts of semantic richness should be
reflected neurologically with a greater number of perceptual symbols’ activation in perceptual,
motor, and emotional cortices (see also Binder et al., 2016 for proposed primary features that
comprise conceptual knowledge), with frontal brain regions recruited to integrate perceptual
symbols into contextual task demands. The difference between these two frameworks can be
dissociated by analyzing behavioural and neuroimaging data concurrently.
The findings presented above suggest that the effects of grounded semantic richness can
be either facilitatory or inhibitory, depending on the congruency or incongruency between task
demands and the grounded semantic dimension under examination. The specific semantic
knowledge that is automatically activated may be considered as diagnostic information regarding
a presented word’s categorical membership in relation to current decision criteria. Importantly,
this automatic activation of semantic information influences responding through being congruent
or incongruent with top-down expectations imposed by task demands.
The finding that semantic information influences task performance in visual and
phonological LDTs, where semantic information is not explicitly required to successfully
identify a stimulus as a real word (orthographic information would suffice for visual LDTs and
phonological information would suffice for phonological LDTs) further supports the conclusion
that semantic knowledge is automatically activated when processing a word. According to the
feedback activation framework (Hino & Lupker, 1996), the activation of semantic units provides

22
feedback to orthographic units and to phonological units. However, whether the feedback from
additional semantic unit activation is facilitatory or inhibitory depends on whether there is
congruency between the nature of the semantic knowledge and the contextual task demands. For
example, Yap et al. (2011) found that the number of senses associated with a visually presented
word facilitated word recognition response latencies and accuracy in a LDT. Similarly, Siakaluk
et al. (2008b) observed facilitatory effects for words rated high in BOI in a phonological LDT.
As a result, it can be concluded that semantic information is automatically activated and
subsequently influences lexical decision processing despite its contextual irrelevance.
Conversely, inhibitory response latency effects have been observed in LDTs to words with
synonyms or homophones (Hino et al., 2013; Pecher, 2001). The number of synonyms that a
word has should be irrelevant in a LDT because orthographic information is the primary type of
information that is used to successfully complete the task. In this instance, however, it is clear
that one’s knowledge about a word is being automatically accessed and that greater ambiguity in
that word’s processing is detrimental to response execution even when the processing was not
contextually necessary.
To account for the effects described above, it is assumed that participants are
automatically activating conceptual knowledge when presented with a word stimulus. The
neurological nature of conceptual knowledge has been investigated, and is largely congruent
with a grounded cognition framework. Importantly, conceptual knowledge has been found to be
dynamic, flexible, and distributed across many brain areas including modality-specific regions
(Fernandino et al., 2015; Hoenig et al., 2008; Kiefer & Pulvermuller, 2012). In an LDT,
differential ERP patterns have been observed for action- and sound-related verbs (Popp et al.,
2016). Experience is critical to conceptual representation. That is, repeated meaningful

23
interactions with a concept leads to greater grounding of that concept’s representation in motor
and sensory brain regions (Kiefer & Trumpp, 2012). The specific conceptual features that are
represented in sensory and motor brain regions are a function of an individuals’ sensory and
motor experiences during concept acquisition (Kiefer & Pulvermuller, 2012). Kiefer et al. (2007)
trained participants with novel objects in different training conditions. In one condition, the
shape of the object was diagnostic of its category membership, while in the other condition, a
feature detail was diagnostic. During a categorization task of these novel objects, differential
patterns of brain activity in sensory-motor brain regions was observed. These data suggest that
conceptual knowledge is plastic and experience-dependent. It should be noted, however, that
some researchers (e.g., Fernanidino et al., 2016a, 2016b) have used computer learning models to
predict patterns of brain activity in heteromodal brain regions associated with sensory-motor
experience (e.g., sound, colour, shape, manipulability, and visual motion). These findings would
suggest that there are other processes involved in the realization of conceptual knowledge than
reactivation of sensorimotor brain regions alone.
Abstract concepts have been observed to activate visual and action brain regions as early
as 178 ms after stimulus presentation in ERP analyses of LDT experiments and 22 ms after
stimulus presentation in conceptual decision tasks (Harpaintner et al., 2020a). The early effects
of category-attribute interactions suggest that these effects reflect very rapid access of grounded
conceptual information (Hoenig et al, 2008). When comparing fMRI patterns of activity between
LDT processing and looking/pointing tasks, significant overlap was observed between looking
and visual abstract words (temporal-occipital regions) and between pointing and motor abstract
words (frontal and parietal motor areas) (Harpaintner et al., 2020b). Fernandino et al. (2021)
found affective information represented in heteromodal cortices (e.g., frontal, parietal, and

24
temporal cortex). These data were interpreted as reflecting experiential information insofar as
lived experiences are multimodal in nature, and thus will recruit heteromodal brain regions to
integrate the activity that experiences elicit in sensorimotor brain regions. Differential patterns of
brain activity during the processing of the same emotion have been observed as a function of the
context in which the emotion was being experienced (Wilson-Mendenhall et al., 2011).
With evidence steadily accumulating that grounded semantic richness influences
responding in visual word recognition tasks, the discussion turns to the neurological implications
of these findings. Specifically, if the propositions of grounded cognition are found to be
theoretically useful, then there are two primary predictions that would stand out in contrast
against purely amodal cognitive predictions. First, the regional brain correlates of conceptual
knowledge processing should include sensory and motor regions. Some researchers, however,
support the grounded by interaction hypothesis, which proposes that sensory and motor regions
activate in conjunction with amodal knowledge centres to represent conceptual knowledge (e.g.,
Tomasino & Rumiati, 2013). For example, Ralph et al. (2017) propose the controlled semantic
cognition framework that is a hub and spoke model of semantic processing. In this model,
sensorimotor experiences are stored in modality-specific brain regions and comprise the spokes
while a single transmodal hub is located in the bilateral anterior temporal lobes. Second, and
more importantly, there should be measurably different patterns of neural activation as a function
of the specific sensory and motor experiences that originally contributed to and thereby
constitute the original experiences that led to the acquisition of that conceptual knowledge. For
example, the neural activation of the concept of coffee could include visual centres for colour,
tactile centres for temperature, olfactory centres for smell, and gustatory centres for taste,
whereas the activation of the concept of horizon should only include visual centres.

25
To this end, research findings have supported these two grounded cognition hypotheses
by observing activation in sensory and motor brain regions during cognitive tasks that do not
explicitly rely upon grounded knowledge for their completion (e.g. Buccino, et al., 2001; Chao &
Martin, 2000; Fadiga et al., 1997; Gallese et al., 1996; Hargreaves et al., 2012), and that the
activated sensory and motor regions are a function of the original experiences (e.g., auditory
regions for sound-related concepts: Kiefer, et al., 2008; motor regions for action-related
sentences: Tettamanti et al., 2005; olfactory regions for smell-related words: González, et al.,
2006).
Program of Research
To contribute to the growing literature that investigates the biological underpinnings of
grounded conceptual knowledge, four experiments were conducted. These experiments serve to
first, replicate and extend previous research in the field and second, to adapt non-invasive, novel,
and comparatively accessible technologies to the field. In doing so, grounded cognition theories
will be strengthened while simultaneously extending the methodological and statistical tools
available to researchers investigating the patterns of brain activity evoked by the processing of
grounded conceptual knowledge. To this end, the four experiments presented in the following
chapters employed near-infrared spectroscopy (NIRS) and electroencephalography (EEG) to
monitor the time course of brain activity elicited by an LDT and three SCTs, analysed with
Bayesian credible intervals (Nathoo et al., 2018).
Near-Infrared Spectroscopy. In the experiments described in Chapter 3, near-infrared
spectroscopy was used to investigate prefrontal cortex hemodynamic activity. NIRS technology
is based on how near infrared light interacts with biological tissue (Bahnmueller, et al., 2014;
Elwell & Hebden, 1999; Sato et al., 2013). Specifically, biological tissue is largely transparent to

26
near infrared light from ~650-1000 nm yet is predictably absorbed by pigmented compounds
such a hemoglobin (Amyot et al., 2012; Ferrari & Quaresima, 2012). By shining near infrared
light into tissue, such as the cerebral cortex (through the scalp) and monitoring how much light
emerges from the tissue after having been reflected and scattered by interactions with intervening
tissues and fluids, (typically ~2-4 cm away from the light emitter), the hemodynamics (e.g., local
blood volume) in tissue of interest can be measured. Critically, near infrared light is absorbed
differentially by oxygenated and deoxygenated hemoglobin (O2Hb and HHb respectively). By
selecting the wavelength of emitted light and measuring how these wavelengths differentially
escape the tissue(s) being measured, one can directly measure the relative concentrations of
oxygenated hemoglobin ([O2Hb]) and deoxygenated hemoglobin ([HHb]). Based on these
values, blood volume and tissue metabolism (e.g., oxygen consumption) can be calculated.
NIRS overcomes many of the limitations of more popular functional magnetic resonance
imaging (fMRI) techniques (Cui et al., 2011). For instance, one limitation when employing fMRI
is that researchers must often oversample due to predictable difficulties in correcting for
movement artifacts (Power et al., 2012; Yerys et al., 2009). NIRS measurements are taken using
small detectors—called optodes—that are affixed to one’s scalp and therefore move with one’s
head, thereby reducing movement artifacts and increasing the applicability of NIRS to grounded
programs of research (Bahnmueller et al., 2014; Visani et al., 2015). A second limitation of fMRI
is that measurements require placing participants in an enclosed and noisy environment that
many people find claustrophobic and fear- or anxiety-inducing, or that is otherwise unsuitable for
some participants (Doi et al., 2013; Furusho et al., 2002). As a result, NIRS is gaining popularity
for investigating cerebral hemodynamics in children and adults who might otherwise find it

27
difficult to participate in an fMRI study (Fishburn et al., 2014). These advantages collectively
enable researchers to collect larger samples more easily (Doi et al., 2013).
Additionally, NIRS technology employs fibre optic cables to carry information between
the measurement device and the optode, as opposed to electrical wiring, thereby enabling NIRS
to be employed concurrently with technologies such as EEG (because there is no signal
interference between the optical fibres and the EEG wires; Doi et al., 2013; Sawan et al., 2012;
Zama et al., 2019). Recording NIRS signals concurrently with EEG has been validated as
indexing the same neurological activity through neurovascular coupling; cortical regions that are
more active produce more electrical activity and also consume more resources that need to be
replenished with greater regional blood volume and oxygenation, despite the fact that both
signals occur at significantly different speeds (milliseconds in EEG and seconds in NIRS;
Balconi et al., 2015; Petzold & Murthy, 2011). Regional cerebral oxygenation increases
following cognitive activity due to regional perfusion overshoot, wherein regional blood supply
is replaced with fresh oxygen-rich blood following periods of activity that lead to depletion of
resources, such as glucose or oxygen (Hermann et al., 2008; Rossi et al., 2012; Verner et al.,
2013). Regional hyperoxygnation has been taken as evidence of functional activation (Hofmann
et al., 2008), while the associated decrease in [HHb] has been interpreted as being due to faster
wash out of deoxygenated hemoglobin due to the event-related increase of oxygenated
hemoglobin (Rossi et al., 2011). Further, NIRS devices can sample at sufficiently high
frequencies so as to be sensitive to neuronal activation in response to cognitive tasks
(Bahnmueller et al., 2014), though the hemodynamic behaviours of interest often occur from 3-8
seconds after stimulus presentation (Moosmann et al., 2003). Finally, NIRS is dramatically more
affordable, relatively portable (or at least more easily transportable), and has superior application

28
potential for participants who have difficulty staying still (Arenth et al., 2007; Ye et al., 2009).
Collectively, these attributes suggest that NIRS is a promising technology for investigating the
neurological correlates of cognitive activity in a non-invasive manner.
Most NIRS devices are continuous wave (CW) devices (Kontos et al., 2014), which emit
light continuously, varying between wavelengths that are absorbed differentially by O2Hb and
HHb. The measured data reflect relative changes in the concentrations of the two compounds.
Alternatively, with additional mathematical processing, other devices provide additional
information, such as the absolute values of total blood volume, the absolute concentrations of
O2Hb and HHb, and local tissue oxygen saturation (Calderon-Arnulphi et al., 2009). Devices that
measure tissue oxygen saturation in addition to hemodynamic variables (i.e., hemoglobin
concentration change) are called spatially resolved instruments, and those that measure absolute
values are referred to as frequency domain devices (Strangman et al., 2002). These devices
operate in a manner similar to a CW device, but during the emission of light at a particular
wavelength, the light is modulated in intensity at approximately 100 MHz. This modulation
produces a measurable emission intensity waveform that can be compared against the waveform
of light that is detected at the scalp after it has passed through the tissue of interest. As it is
known that photons slow down as they pass through tissue, there is a predictable phase shift in
this waveform that can be detected by frequency domain devices (Elwell & Hebden, 1999). This
phase shift information can be used to infer the actual distance that detected photons travelled
before exiting the tissue, providing an index of penetration depth (i.e., how deeply the light
penetrated into the tissue of interest before returning to the detector; this is roughly half the
distance between the emitter and detector; Sato et al., 2013). This path length measurement can

29
be used with a modified version of the Beer-Lambert law to calculate absolute values, as a
function of the known absorption properties of O2Hb and HHb (Elwell & Hebden, 1999).
It is important to note that the emitted light will pass through superficial tissue such as the
scalp and skull. Such tissue is thought to be unresponsive to cognitive demands in most
circumstances, and as such the absorption and scattering coefficients of this tissue can be viewed
as constants in NIRS observations (though, it is possible for blood flow to the scalp to increase if
the sensor begins to irritate the skin, or under particularly straining mental tasks).
The data obtained from NIRS experiments have been found to correlate strongly with the
blood-oxygen level dependent (BOLD) signal from fMRI analyses, thereby validating the use of
NIRS to detect cerebral hemodynamics as a measure of spatially sensitive brain activity
occurring at most ~0.8 cm into the cerebral cortex; despite, like EEG, not being able to
specifically detect subcortical activation (Amyot et al., 2012; Sato et al., 2013). Further evidence
that NIRS is a reliable tool for cognitive research is found in experiments that explicitly examine
its test-retest reliability by comparing participants’ recordings across multiple experimental
sessions (6 months: Sato et al., 2006; one year: Schecklmann et al., 2008, Watanabe et al., 2003).
It should be noted, however, that significant variability between participants has been noted in
NIRS signals, thus making it vital that researchers record a substantial number of observations
for statistical conclusions regarding cerebral hemodynamics (Amiri et al., 2014; Banville et al.,
2017).
Some researchers have taken to reporting changes in [O2Hb] as an index of cerebral
activity, given its correlation with fMRI’s BOLD signal (Sato et al., 2013), though there is still
some debate about whether [O2Hb], total hemoglobin concentration ([THb]), or [HHb] is the best
measure. Changes in [THb] would reflect the brain drawing on larger volumes of blood to meet

30
increased task demands, whereas changes in [O2Hb] would reflect a change the composition of
regional blood, and a change in [HHb] would suggest the brain’s increased use of oxygen during
task demands. It has been concluded, however, that due to neurovascular coupling the increase in
cerebral [THb] cannot be met with a proportional amount of oxygen consumption, and that this
will lead to an increase in [THb] and [O2Hb] and a resulting proportional decrease in [HHb]
(Schroeter et al., 2002).
NIRS in Cognition. The application of NIRS to cognitive research is quickly growing in
popularity and has been found to detect patterns of cerebral hemodynamics that are unique to
specific tasks that a participant is completing (e.g., Banville et al., 2017; Herold et al., 2018;
Pinti et al., 2020). These unique patterns demonstrate that the cerebral activity detected by NIRS
reflects task-dependent cognitive processing rather than a generic state of activity associated with
cognitive effort (Banville et al., 2017). To this end, non-invasive brain-computer interfaces have
been developed that employ NIRS technology to recognize intentional brain activity, often with
the goal of enabling individuals with disabilities to control devices such as wheelchairs (e.g.,
Banville et al., 2017; Khan et al., 2014; Morioka et al., 2014). Many researchers elect to monitor
the prefrontal cortex (PFC) because it is the brain region behind the forehead, meaning that there
is no hair on the skin to interfere with light emission or detection, the bone is thinner in this
region than it is in other areas of the skull, and the brain-scalp distance is smaller in this region
compared to others (Balconi et al., 2015). Nakadoi et al. (2012) demonstrated increased
concentrations of oxygenated hemoglobin [O2Hb] in the left PFC when participants evaluated the
emotional significance of presented stimuli. In an n-back working memory task, participants are
presented with a series of stimuli and are to respond when the current stimulus matches one that
was presented n-back, where n is a number that varies from condition to condition. Fishburn, et

31
al. (2014) observed increased [O2Hb] in the PFC during an n-back task, though increases were
observed bilaterally. This increase could be interpreted as a general component of greater
cognitive effort given that a similar increase in [O2Hb] was observed bilaterally in the PFC
during random number generation and another n-back task (Hoshi et al., 2013). Increases in PFC
[O2Hb] and corresponding decreases in [HHb] have also been found during go/no-go animal
detection tasks (Medvedev et al., 2011), and the Wisconsin Card Sorting Task (Fallgatter &
Strik, 1998). Left dorsolateral PFC increases of [O2Hb] have been observed during personal
moral judgements, as compared to impersonal dilemmas (Dashtestani et al., 2018). Increases in
[O2Hb] and [THb] in the left PFC have been observed during confrontational naming tasks
(making a forced choice of a label for an observed picture; Sakatani, et al. 1998). While actively
reading aloud, participants have exhibited a decrease in [O2Hb] and an increase in HHb
(Fallgatter et al., 1998). The PFC has also shown increased [O2Hb] and decreased [HHb] in
verbal fluency tasks where a participant is given an alphabet constraint (e.g., ‘words that begin
with the letter R’) and asked to generate as many exemplars as possible (Chaudhary, et al.,
2011). During mental rotation tasks, decreases of [O2Hb] and [THb] along with characteristic
increases of [HHb] have been observed in the PFC (Banville et al., 2017). Collectively, these
data suggest that NIRS is a reliable instrument for monitoring PFC activity and that it is sensitive
to the demands of many types of cognitive tasks, including the PFC’s role in mental
manipulation and monitoring processes (Hoshi et al., 2003).
Examining brain regions other than the PFC is possible, though it can be more difficult
given the presence of hair, and that the brain-scalp distance is often larger (Cui et al., 2011). For
example, increased [O2Hb] has been observed in the inferior frontal cortex (IFC) in languageimpaired individuals compared to controls in a picture-pointing sentence comprehension task

32
(i.e., which picture corresponds to a presented sentence; Fu et al., 2016). Increased [O2Hb] has
been observed in frontotemporal regions during an oddball paradigm (Tong et al., 2005).
Additionally, in a Stroop paradigm, increases in [O2Hb] and [THb] were observed in the left IFC
on incongruent trials (Ehlis et al., 2005; Schroeter et al., 2002). Similar increases have been
observed during a Stroop task in the superior parietal and frontal cortices of the left hemisphere
(Hock et al., 1997). In a verbal fluency task, Hall et al. (2013) observed increased [O2Hb] in the
fronto-temporal areas corresponding to Broca’s and Wernicke’s areas. Banville et al. (2017)
reported increased left temporal [O2Hb] during word generation tasks, and bilateral temporal
[O2Hb] and [THb] increases during mental subtraction and motor imagery tasks. Increased
[O2Hb] has been observed in the occipital lobe when one is viewing pleasant stimuli and
increased [HHb] when processing emotional stimuli (Herrmann et al., 2008). During voluntary
motor planning, increased [O2Hb] has been observed in premotor areas (Zama et al., 2015).
Though more difficult to obtain, NIRS data from non-frontal brain regions have been obtained,
and it is in line with other brain imagining studies.
A common convention in NIRS research is to employ a blocked experimental design,
wherein a series of trials are presented rapidly, and mean amplitudes recorded. However, in light
of pilot data collected for this program of research, it was deemed important to record more
information than means alone, and thus a single-trial design was employed. In an event-related
single trial design, there are more physiological metrics that can be recorded such as gain,
dispersion, and lag (e.g., Schroeter et al., 2002).
In single-trial (event-related) go/no-go NIRS experiments, data is averaged across trial
types and reported for the seconds following stimulus presentation. By doing so, experimenters
can report the gain (peak amplitude), dispersion (duration of hemodynamic change), and the lag

33
(time from stimulus presentation to peak hemodynamic amplitude) (Schroeter et al., 2002).
These additional metrics stand to characterize the typical hemodynamic response more
accurately to a variety of cognitive tasks. For example, Schroeter et al. (2002) conducted an
event-related NIRS experiment using a colour-matching Stroop experiment. It was reported that
[HHb] and [O2Hb] gains and lag were greater for colour-incongruent trials. This difference in
timing would have been missed had the researchers adopted a blocked design.
To validate a single-trial design for the current program of research, pilot data were
collected that directly compared the data from a blocked design and a single-trial design with a
25 second inter-trial interval. The task for both experiments was a SCT with a decision criterion
of “Is this word easily imageable?”, with the go-trials varied on BOI ratings. BOI was chosen
because it is one of the strongest facilitatory effects in SCTs (e.g., Siakaluk et al., 2008a; Sidhu
et al., 2014). In the blocked design, stimuli were presented for 2500 ms or until a participant
responded, and the intertrial interval was jittered between 1000 and 3000 ms to control for
anticipatory effects. The data from this experiment were analyzed with a 2 x 2 analysis of
variance (ANOVA: 2 hemisphere (left/right) x 2 condition (high/low BOI)). Disappointingly, no
significant effects were observed; the effects of BOI on SCT performance were not evident in
either PFC hemisphere’s hemodynamics. In the single-trial design, however, stimuli were
presented identically, though the intertrial interval was set at 25 seconds so as to be confident
that cerebral hemodynamics had returned to baseline between trials (see Strangman et al., 2002;
Tak et al., 2015). Rather than averaging the hemodynamic response over time and several trials,
as in the blocked design, an event-related time course for each trial was obtained, and an average
time course per participant per condition type was calculated. In this experiment, bilateral
condition differences were evident, with high BOI trials showing no significant hemodynamic

34
response, while low BOI trials showed a marked amplitude increase peaking approximately five
seconds after stimulus presentation and returning to baseline approximately 10-12 seconds after
stimulus presentation. In light of these data, the decision was made to adopt a 15 second intertrial interval for the current program of research so as to give participants’ hemodynamic
response to each trial enough time to return to baseline between each trial.
In blocked go/no-go NIRS experiments, the mean hemodynamic amplitude for no-go
trials would be obtained by averaging across a block of many sequential no-go trials, without any
go trials. The reported mean hemodynamic amplitudes for go trials, however, would be reported
as the average of a block of many sequential trials, the majority of which being go trials, yet
containing some no-go trials. This is to say that the peak amplitudes reported for critical
experimental trials in blocked designs includes data from some incongruent no-go trials as well.
Furthermore, by taking an average across multiple trials completed within 20-30 seconds, mean
amplitude is the only hemodynamic metric that can be meaningfully reported.
NIRS was thus chosen for the current research to examine this technology’s application
to the study of grounded cognition due to its high sampling rate and compatibility with EEG. By
targeting the prefrontal cortex with NIRS measurements, the experiments presented in Chapter 3
will be able to address questions about the comparative effort to process semantically rich trials
of varying task relevance for this cortical region.
Collectively, the findings summarized above serve to justify the application of NIRS
technology in the current program of research. Explicit predictions for the NIRS data can be
found in Chapter 3.
Electroencephalography. Many cognitive neuroimaging studies have taken to using
technologies such as EEG to investigate the neural correlates of cognitive processing due to

35
EEG’s strengths in monitoring time-related aspects of semantic processing. EEG was used in the
experiments presented in Chapter 4. It is a non-invasive neuroimaging technique wherein
electrodes are placed on the scalp to record electric polarity changes produced by spontaneous
brain activity (i.e., the summed electrical activity produced by many neurons firing at once). This
technology is used because of its non-invasiveness, relative affordability, and its high rate of
sampling (e.g., up to 1000 Hz). EEG recordings have been found to be consistent within an
individual over time, suggesting that people’s characteristic brain wave patterns are reliable
indices of cerebral activity (Tomarken, et al., 1992). By reporting EEG data as a relative change
across conditions for each participant rather than absolute mean values, both ERPs and brain
wave pattern changes can be compared across individuals (Antonenko et al., 2010). EEG does
not accurately pinpoint specific brain regions that may be responsible for observed patterns of
activation because EEG takes measurements from the scalp, when the source of the electric
current is separated from an electrode by cerebrospinal fluid, the skull, and the scalp (Srinivasan,
1999). However, as the number of electrodes on the scalp is increased, the relative spatial
accuracy improves. Though a region of activity can be deduced with some certainty, EEG cannot
report whether the observed activity arose from local cortical tissue or from deeper structures
(Andreassi, 2013). However, results have shown differential patterns of activation depending on
the task that a participant is performing. For example, Khrizman (1973) showed that EEG can
detect activity in the motor cortex and lower parietal lobe when participants are tapping their
finger, yet activity is seen in the motor cortex and frontal lobe when performing a sorting task.
Such observations support and validate EEG measurements in research.
EEG techniques rely on the accurate placement of electrodes (units ~6-8 mm in size,
including a detecting wire and an electrolyte-soaked sponge) over known brain regions

36
(Andreassi, 2013). This is most commonly done by employing the 10-20 System (Jasper, 1958;
Trans-Cranial Technologies Ltd, 2012). This system works by measuring the distance between
predetermined physical landmarks on a person’s head and placing electrodes in locations that are
at 10% and 20% increments of those distances. The major landmarks used are the nasion (nose
bridge), the inion (protuberance in the back of one’s head over the occipital lobe), and the divot
that can be felt immediately in front of one’s ears. The skull’s midpoint (top) is identified by
marking the convergence point at 50% of the distance between the nasion and inion and also
50% of the distance between both ear landmarks. Then, using these landmarks and midpoint,
electrodes are placed according to Jasper’s (1958) validated distances. For EEG devices that
employ more electrodes, there are guidelines put forth by the American EEG Society (1991) to
guide electrode placement that ensure reliability between experiments. EEG software relies on a
single electrode placed on a reference point (cheek, chin, shoulder) that will detect regular
rhythmic patterns of activation in a participant's body to filter the EEG signal, thereby isolating
the electrical activity that is unique to the brain (Andreassi, 2013). Many companies are
beginning to market EEG caps that hold all electrodes in a fixed relative distance from one
another, thereby adhering to the 10-20 System without need for single-electrode measurements
(Trans-Cranial Technologies Ltd., 2013). This enables all electrodes to be placed in their correct
positions simultaneously. By using such a cap, individuals’ total time requirement for
participation is reduced while also improving data quality and reliability across experiments and
researchers.
EEG may also be used to detect ERPs, which are characteristic patterns of electrical
activity that follow immediately after stimulus presentation that represent the immediate and
automatic processes elicited by the stimuli by cortical neurons (Andreassi, 2013; Tomarken, et

37
al., 2007). ERP data are often reported as distinct components in the ERP signal. Specifically,
these components are typically labeled with a P or N, denoting whether the deflection is positive
or negative compared to baseline (Luck, 2014). Further, each ERP deflection is given a
numerical identifier using one of two typical naming conventions: either a single digit denoting
the chronological ordering of the deflection (e.g., P2 would be the second positive deflection in a
given ERP), or with a larger numerical identifier signifying the temporal epoch in which the
deflection occurred (e.g., P200 would be a positive deflection that peaked between 200 and 300
ms after stimulus presentation, regardless of whether it was the first or second positive
deflection) (Luck, 2014). The latter convention will be used here and in Chapter 4.
The deflections of interest for the experiments presented in Chapter 4 are the P200, N400,
P600, and N600. The P200 has been interpreted as being reflective of and increased by automatic
attention capture, task difficulty, and/or semantic richness (Kounios et al., 2009; Taylor & Khan,
2000). The variability of factors that influence the P200 suggests that this deflection broadly
captures early automatic processing. The N400 is one of the most characteristic deflections in
ERPs produced by semantic processing. Increased N400 amplitude has been typically associated
with the processing of semantic or contextual irregularities such as when a sentence concludes
unexpectedly (Heinze et al., 1998; Xue, et al., 2015). The P600 is a late deflection that may not
contribute directly to response execution but has been interpreted as one’s automatic reanalysis
of words that require additional processing (Coulson, et al., 1998; Gouvea, et al., 2010). By
contrast, a negative deflection at the same time (the N600) is thought to be reflective of
additional processing due to semantic incongruence, particularly in the processing of humour
(Du et al., 2012).

38
Many ERP effects have been observed in semantic processing experiments. The
amplitude of the N400 has been shown to decrease when processing words that occur frequently
in written text (Assadollahi & Pulvermüller, 2001; Van Petten & Kutas, 1990), and increase for
words with larger orthographic neighbourhoods (the number of words that differ from a word by
one letter; Holcomb et al., 2002). Hauk & Pulvermüller (2004) demonstrated that the effects of
word length and frequency on N400 amplitude are independent and additive. Concrete words
(e.g., chair, dog) elicit a stronger N400 than do abstract words when other lexical variables are
controlled (Barber et al., 2013). This concreteness effect is further enhanced when the concrete
words have fewer associates (i.e., they have less semantic richness; Kounios et al., 2009). N400
amplitude is greater for nouns than verbs in a primed LDT (Rösler et al., 2001). When making a
congruency decision, the N400 is lower for congruent words than to incongruent words (Neville,
et al., 1986). In a sentence verification task, a pronounced N400 was observed when people were
presented with the final sentence term in mismatch trials (i.e., reply “yes” to ‘[x] is not a [y]’ or
“no” to ‘[x] is a [y]’) than to congruent trials (Fischler et al., 1983). Taken together, the data
reliably show an N400 waveform deflection when semantic information is processed, and that
the N400 is affected by the amount of semantic information required to successfully complete
task demands.
Some ERP research has been done specifically on semantic richness and semantic
categorization. In an SCT where word pairs were presented, and participants were to identify
whether the second word was an exemplar of the first, N400 attenuation was observed for typical
exemplars, “yes” responses, and when the word pairs were semantically related, regardless of
category membership (Heinze et al., 1998). This demonstrates that the amount of processing
necessary for an SCT is reflected in N400 amplitude. The N400 has been directly tied to

39
embodied semantic processing and language comprehension (Willems et al., 2008). Xue et al.
(2015) conducted an ERP study on the role of BOI in sentence processing in non-native
speakers. ERPs were taken when participants read the last word in a visually presented sentence
acceptability task. The target words were either high or low in BOI and could have been
presented after a semantically rich or semantically poor sentence. A P200 was observed in
central parietal regions (electrode P4) for all high BOI words. The N400 was most pronounced
when high BOI words were presented in poor contexts versus rich contexts and was larger across
all high BOI trials than in low BOI trials. When embodied information (a form of grounded
semantic richness) is task-relevant, the N400 is attenuated. However, when a word activates an
unexpected amount of type of information, the N400 is exaggerated, suggesting a higher amount
of semantic processing is required to resolve task demands. Collectively, the data reviewed
suggests that the N400 is the ERP deflection more frequently associated with semantic
processing, and that congruence with task demands predicts its amplitude.
Although deflections occurring after the N400 may not directly reflect the processing of a
word stimulus’s semantic content (typically completed within the first 500 ms), they may reflect
additional elaborative processing from anomalous stimuli (Kurchinke & Mueller, 2019; Wang et
al., 2019). That is to say that the semantic content of a word stimulus may capture a degree of
automatic attention that elicits additional processing due to its perceived irregularities. The P600
is a positive deflection that typically peaks approximately 600 ms after stimulus presentation and
has been observed to be particularly sensitive to a stimulus’s emotional content (Coulson et al.,
1998; Gouvea et al., 2010). The N600 is another late component that has been interpreted as
reflecting additional processing elicited by a stimulus’s contextual semantic incongruence and
humour (Du et al, 2012). It is possible that, by 600 ms after stimulus presentation, a word will

40
have already been categorized in the context of current task demands. As such, ERP condition
differences in the P600 and N600 can be interpreted as automatically elicited additional
processing of stimuli that are perceived as atypical category exemplars (e.g., a concrete noun that
evokes high levels of emotion information, which is typically associated with abstract nouns).
ERP analyses in Chapter 4 extend to 800 ms after stimulus presentation so as to capture these
indices of semantic processing. Full ERP predictions can be found in Chapter 4.
Experimental Tasks
In the LDT participants were presented with letter strings that were either real words or
non-words and were to respond only to the real words. Real words varied in printed word
frequency. The other three experiments were SCTs with real words as stimuli. In one SCT,
participants were asked to respond only to words that were easily imageable, with all high
imageability go trials varied in BOI ratings, with half of the words rated significantly higher in
BOI than the other half. The other two SCTs used the same real word stimuli, with half the
stimuli being abstract nouns, and the other half being concrete nouns. In one experiment
participants were asked to respond to abstract nouns, and in the other experiment participants
were asked to respond to concrete nouns. Both abstract and concrete nouns were manipulated for
EE ratings with half of each word type being high in EE, and the other half significantly lower.
In all four experiments, participants were presented with a letter string, and were asked
(1) to make a response if that letter string met a decision criterion for that experiment or (2) to
make no response if the word did not meet the decision criterion. This procedure was selected
because participant responses to target trials have been found to be shorter, less variable, and
more accurate than when a two-choice procedure is employed (Gomez et al., 2007; Gordon &
Caramazza, 1982). The response time differences between the two procedures have been

41
interpreted as an additional process being required by participants to choose between two
responses in addition to the cognitive process(es) associated with the experimental task. It has
been found that participant discriminability (d’) and decision biases (log β) are comparable
between the two procedures despite the shorter response times in go/no-go designs (Chiarello et
al., 1988; Measso & Zaidel, 1990). In a two-choice design, participants who have been instructed
to make a speeded response will feel pressured to respond to all stimuli, thus increasing the
probability of making a no response if a stimulus is less familiar or less characteristic of the
decision criterion. However, in a go/no-go procedure, if a stimulus does not quickly provoke a go
response, processing will continue because participants do not have the option to make a no
response, and this may result in a longer response latency that reflects the additional processing
of less typical category members.
Experimental Procedure
While detailed explanations of the experimental methods are presented in Chapters 3 and
4, the commonalities are presented briefly here. For any single trial, participants were seated in
front of a computer, instructed to attend to a fixation point (+). Following the inter-trial interval,
a letter string was presented in capital letters, centred on the screen. If that letter string was
congruent with that experiment’s decision criterion (e.g., “Is this a concrete word?”), participants
were to press the spacebar. If the letter string did not satisfy the current decision criterion, they
were to do nothing. Go/no-go designs, as compared to more standard forced two-choice designs,
are thought to be less cognitively demanding during non-lexical processing, resulting in fewer
false positives, shorter response latencies, and higher accuracy on average (Lee, 2019). On
average, the experiments lasted for 60-80 minutes. An inter-trial-interval of 15 seconds was used

42
in all four experiments so as to allow hemodynamic activity to return to baseline following
stimulus presentation.
Full behavioural predictions for the influence of these variables on participant response
latencies are detailed in Chapter 3.
Independent Variables
The grounded semantic richness dimensions that were manipulated were BOI and EE.
BOI was chosen because it has reliably been shown to facilitate response latencies to the
presentation of easily imageable words in SCTs and, consequently, is thought to be diagnostic of
how concrete concepts are represented (e.g., Hargreaves et al., 2012). Concrete objects are, by
their nature, physical entities that allow for varying degrees of bodily interactions. For example,
spoon refers to a concrete class of objects that are easily imageable and can be easily physically
interacted with (and thus would have a high BOI rating), whereas cloud refers to a less concrete
class of objects that are also easily imageable but cannot be easily physically interacted with (and
thus would have a low BOI rating). EE was chosen because it is thought to be diagnostic of how
abstract words are represented (Newcombe et al., 2012; Vigliocco, et al., 2009). Importantly, EE
captures a dimension of emotion knowledge that represents information that is gained through
one’s emotion experiences across a variety of environmental contexts (Newcombe et al., 2012).
As such, it was decided to operationalize emotion information as EE ratings. Abstract concepts
are, by their nature, intangible concepts that refer to ideas rather than concrete objects that can be
physically interacted with. For example, envy is an abstract concept with no physical referent.
However, when experiencing or thinking about envy, one will have emotional associations with
the experience of being envious. Full behavioural predictions for the influence of these variables
on participant response latencies are detailed in Chapter 3.

43
By manipulating these dimensions of grounded semantic richness, target trials vary on
comparative difficulty within an experiment such that some trials will be more easily recognized
as satisfying the decision criterion. By analyzing easy and hard trials within an experiment, a
greater degree of experimental control is possible than if the data were analyzed between
different dimensions of grounded semantic richness. For example, in any given experiment, the
decision criteria for a go-response is the same for all trials. This means that every single trial will
involve the automatic activation of a word stimulus being integrated into the same contextual
demands. As all other variables are controlled between trials, the behavioural, EEG, and NIRS
differences between trials will reflect differential amounts of processing elicited by varying
degrees of the same dimension of grounded semantic richness. Therefore, to the extent that a
manipulated dimension is diagnostic of experimental decision criteria, the data that are produced
will reflect the varying levels of a single dimension of grounded semantic richness and the
executive efforts to apply that information to the current situation. Without such a difficulty
manipulation, the patterns of activation between target trials would be predicted a priori to be
qualitatively different according to the PSS framework, and thus increasingly challenging to
interpret meaningfully.
Dependent Variables
The dependent variables were response latencies, response accuracy, full-scalp eventrelated potentials (ERPs) and prefrontal cortex (PFC) hemodynamics. The manipulated variables
were word frequency in the LDT and grounded semantic richness in the SCTs. The specific
dimensions of grounded semantic richness that were manipulated were BOI and EE. Specific
details for the NIRS and EEG devices used in each experiment can be found in Chapters 3 and 4
respectively.

44
Summary
Several predictions are implicit in the philosophical perspective that conceptual
knowledge is obtained primarily through one’s sensory, motor, and emotion experiences with the
environment. When processing any given concept, regardless of contextual demands, one may
predict that the concept’s sensory, motor, and emotion information is automatically activated.
This automatic activation can then be hypothesized to influence experimental response latencies.
Specifically, with greater amounts of sensory, motor, and/or emotion information associated with
a given concept, that concept’s processing may be able to reach a critical decision threshold more
quickly than a comparable concept with fewer such associated information when that
information is congruent with current task demands. However, if the fundamental nature of that
concept’s associated information is incongruent with current task demands, one can predict a
response latency deficit when responding under such circumstances. These observations have
been largely confirmed in the research summarized above. The current program of research
stands to strengthen these behavioural observations through replication and extension.
Additionally, by using a 15-second inter-trial interval, the data presented in Chapters 3 and 4
stand to demonstrate that these behavioural effects are not due to contextual priming/momentum
created by responding to similar stimuli every few seconds and are actually fundamental to how
conceptual information is processed.
By testing the neurological predictions of grounded cognition, the current program of
research stands to contribute to the growing body of literature that endeavours to elucidate the
nature of conceptual knowledge representation. While many ERP experiments analyze targeted
electrode sites and report averaged or maximum amplitude within arbitrary boundaries, the
current program of research provides a much more thorough contribution to the literature. From

45
a grounded cognition perspective, many different brain regions are hypothesized to automatically
contribute to the reactivation of conceptual knowledge. To investigate this dynamic process more
fully, data were recorded and analyzed across the whole scalp. This difference allows for a more
complete description of the cortical processing of conceptual knowledge. Furthermore, rather
than average across time to report ERP metrics, the full ERP time course was computed at each
electrode for each condition, and difference waves reported. This approach allows the current
research to contribute a more thorough understanding of both when and where different
dimensions of conceptual knowledge are processed. Moreover, a novel Bayesian analysis was
employed so as to identify periods of meaningful condition differences confidently and
accurately. This analytical technique stands to be a valuable contribution to future studies.
Finally, the current research is one of few to apply NIRS to the study of grounded
cognition, particularly given the concurrent application of EEG. By monitoring PFC
hemodynamic responses to single trial presentations, a novel and comparatively accessible
neurological marker of cognitive processing is presented in Chapter 4. In a go/no-go experiment,
such as those presented in Chapter 4, a participant must respond to a presented stimulus or
inhibit the automatically generated response that is elicited. It is hypothesized that word stimuli
that are poor category exemplars (e.g., imageable nouns that are rated low in BOI) will be more
difficult to inhibit because their processing takes longer to reach a decision threshold. This
additional processing may be reflected in the PFC due to the increased effort to identify such
trials as category exemplars. To best capture and report these effects, a full time course Bayesian
condition difference analysis was conducted as described above. As such, Chapter 4 stands to
contribute to the field a novel technological index of grounded cognition processing as well as a
novel analytical technique for reporting such data.

46
Together, the data presented in the following chapters examine, analyze, and describe the
grounded nature of conceptual knowledge across four experiments, involving two neuroimaging
techniques accompanied by behavioural analyses. Collectively, these data will contribute to the
growing body of research that demonstrates the fundamentally grounded nature of human
cognition.

47
References
American Electroencephalographic Society (1991). Guidelines for standard electrode position
nomenclature. Journal of Clinical Neurophysiology, 8, 200-202.
Amiri, M., Poulliot, P., Bonnéry, C., Leclerc, P-O., Desjardins, M., Lesage, F., & Joanette, Y.
(2014). An exploration of the effect of hemodynamic changes due to normal aging on the
fNIRS response to semantic processing of words. Frontiers in Neurology, 5(249), 1-11
Amyot, F., Zimmermann, T., Riley, J., Kainerstorfer, J. M., Chernomordik, V., Mooshagian, E.,
… Wassermann, E. M. (2012). Normative database of judgment of complexity task with
functional near infrared spectroscopy – Application for TBI, Neuroimage, 60(2), 879883.
Anderson, M. L., Richardson, M. J. & Chemero, A. (2013). Eroding the boundaries of cognition:
Implications of embodiment. Topics in Cognitive Science, 4, 717-730.
Andreassi, J. L. (2013). Psychophysiology: Human behavior & physiological response.
Psychology Press.
Antonenko, P., Paas, F., Grabner, R., & van Gog, T. (2010). Using electroencephalography to
measure cognitive load. Educational Psychology Review, 22, 425-438.
Arenth, P. M., Ricker, J. H., & Schultheis, M. T. (2007). Applications of functional near-infrared
spectroscopy (NIRS) to neurorehabilitation of cognitive disabilities. Clinical
Neuropsychology, 21, 38-57.
Assadollahi, R. & Pulvermuller, F. (2001). Neuromagnetic evidence for early access to cognitive
representations. Neuroreport, 12(2), 207-213.

48
Bahnmueller, J., Dresler, T., Ehlis, A., Cress, U., & Nuerk, H. (2014). NIRS in motion –
unraveling the neurocognitive underpinnings of embodied numerical cognition. Frontiers
in Psychology, 5(743), 1-4.
Balconi, M., Grippa, E., & Vanutelli, M.E. (2015). What hemodynamic (fNIRS),
electrophysiological (EEG) and autonomic integrated measures can tell us about
emotional processing. Brain and Cognition, 95, 67-76.
Banville, H., Gupta, R., & Falk, T. H. (2017). Mental task evaluation for hybrid NIRS-EEG
brain-computer interfaces. Computational Intelligence and Neuroscience, 2017.
Barber, H. A., Otten, L. J., Kousta, S., & Vigliocco, G. (2013). Concreteness in word processing:
ERP and behavioural effects in a lexical decision task. Brain and Language, 125(1), 4753.
Barrett, L., & Lindquist, K. A. (2008). The embodiment of emotion. In G. R. Semin, E. R. Smith
(eds.), Embodied grounding: Social, cognitive, affective, and neuroscientific approaches
(pp. 237-262). New York, NY, US: Cambridge University Press.
Barsalou, L. W. (1999). Perceptual symbol systems. Behavioral and Brain Sciences, 22(4), 577660.
Barsalou, L. W. (2016). Can cognition be reduced to action? Processes that mediate stimuli and
responses make human action possible. In The Pragmatic Turn: Toward Action-Oriented
Views in Cognitive Science. Engle, A., K., Friston, K. J., & Kragic, D. (Eds.).
Cambridge, MA: MIT Press.
Bennet, S. D. R., Burnett, A. N., Siakaluk, P. D., & Pexman, P. M. (2011). Imageability and
body-object interaction ratings for 599 multisyllabic nouns. Behavior Research Methods,
43(4), 1100-1109.

49
Binder, J. R. (2007). Effects of word imageability on semantic access: Neuroimaging studies. In
Kraut, M. A., Hard J. (Eds.), Neural Basis of Semantic Memory. Cambridge University
Press, pp. 149-181.
Binder, J. R., Conant, L. L., Humphries, C. J., Fenandino, L., Simons, S. B., Aguliar, M., &
Desai, R. H. (2016). Toward a brain-based componential semantic representation.
Cognitive Neuropsychology, 33(3-4), 130-174.
Buccino, G., Binkofski, F., Fink, G. R., Fadiga, L., Fogassi, L., & Gallese, V., et al. (2001).
Action observation activates premotor and parietal areas in a somatoptopic manner: An
fMRI study. European Journal of Neuroscience, 13, 400-404.
Calderon-Arnulphi, M., Alaraj, A., & Slavin, K. V. (2009). Near infrared technology in
neuroscience: Past, present and future. Neurological Research, 31, 605-614.
Caligiore, D. & Fischer, M. H. (2012). Vision, action, and language unified through
embodiment. Psychological Research, 77, 1-6.
Chao, L. L., & Martin, A. (2000). Representation of manipulable man-made objects in the dorsal
stream. Neuroimage, 12, 478-484.
Chaudhary, U., Hall, M., DeCerce, Rey, G., & Godavarty, A. (2011). Frontal activation and
connectivity using near-infrared spectroscopy: Verbal fluency language study. Brain
Research Bulletin, 84, 197-205.
Chiarello, C., Nuding, S., & Pollock, A. (1988). Lexical decision and naming asymmetries:
Influence of response selection and response bias. Brain and Language, 34(2), 302-314.
Chimero, A. (2009). Radical Embodied Cognitive Science. Cambridge, MA, MIT Press.

50
Conway, M. A. (2002). Sensory-perceptual episodic memory and its context: Autobiographical
memory. In Episodic Memory: New Directions in Research, ed. A. Baddeley, J. P.
Aggleton, M. A. Conway, pp. 53-70, Oxford: Oxford University Press.
Cortese, M. J., & Fugget, A. (2004). Imageability ratings for 3,000 monosyllabic words.
Behaviour Research Methods, Instruments, and Computers, 36, 384-387.
Coulson, S., King, J. W., & Kutas, M. (1998). Expect the unexpected: Event-related brain
response to morphosyntactic violations. Language and Cognitive Processes, 13(1), 2158.
Cui, X., Bray, S., Bryant, D. M., Glover, G. H., & Reiss, A. L. (2011). A quantitative
comparison of NIRS and fMRI across multiple cognitive tasks. NeuroImage, 54(4), 28082821.
Dashtestani, H, Zaragoza, R., Kermanian, R., Knutson, K. M., Halem, M., Casey, A.,
Karamzadeh, N. S., Anderson, A. A., Boccara, A. C., & Gandjbakhche, A. (2018). The
role of the prefrontal cortex in a moral judgment task using functional near-infrared
spectroscopy. Brain & Behavior, 8(11), e01116.
Dehaene, S. (2011). The number sense: How the mind creates mathematics. Oxford: Oxford
University Press.
De Vega, M., Robertson, D. A., Glenberg, A. M., Kaschak, M. P., & Rinck, M. (2004). On doing
two things at once: Temporal constraints on actions in language
comprehension. Memory and Cognition, 32(7), 1033-1043.
Doi, H., Nishitani, S., & Shinohara, K. (2013). NIRS as a tool for assaying emotional function in
the prefrontal cortex. Frontiers in Human Neuroscience, 7(770), 1-6.

51
Du, X., Qin, Y., Tu, S., Yin, H., Ting, W., Yu, C., & Qiu, J. (2012). Differentiation of stages in
joke comprehension: Evidence from an ERP study. International Journal of Psychology,
48(2), 149-157.
Eerland, A., Guadalupe, T. M., & Zwaan, R. A. (2011). Leaning to the left makes the Eiffel
Tower seem smaller: Posture-modulated estimation. Psychological Science, 22(2), 15111514.
Ehlis, A., Herrmann, M. J., Wagener, A., & Fallgatter, A. J. (2005). Multi-channel near-infrared
spectroscopy detects specific inferior-frontal activation during incongruent Stroop trials.
Biological Psychology, 69, 315-331.
Elwell, C. & Hebden, J. (1999). Near infrared spectroscopy. Retrieved from:
www.ucl.ac.uk/medphys/research/borl/intro/nirs
Estes, Z., & Adelman, J. S. (2008). Automatic vigilance for negative words is categorical and
general. Emotion, 8(4), 453-457.
Fadiga, L., Arbib, M. A., & Rizzolatti, G. (1997). Premotor cortex activation during observation
and naming of familiar tools. Neuroimage, 6, 231-236.
Fallgatter, A. J., Müller, T. J., & Strik, W. K. (1998). Prefrontal hypo oxygenation during
language processing assessed with near-infrared spectroscopy. Neuropsychobiology,
37(4), 215-218.
Fallgatter, A. J. & Strik, W. K. (1998). Frontal brain activation during the Wisconsin Card
Sorting Test assessed with two-channel near-infrared spectroscopy. European Archives of
Psychiatry and Clinical Neuroscience, 248, 245-249.
Favela, L. H. (2014). Radical embodied cognitive neuroscience: Addressing "grand challenges"
of the mind sciences, Frontiers in Human Neuroscience, 8, 1-10.

52
Fernandino, L., Binder, J. R., Desai, R. H., Pendl, S. L., Humprines, C. J., Gross, W. L., Conant,
L. L., & Seidenberg, M. S. (2016). Concept representation reflects multimodal
abstraction: A framework for embodied semantics. Cerebral Cortex, 26(5), 2018-2034.
Fernandino, L., Humphries, C. J., Conant, L. L., Seidenberg, M. S., & Binder, J. R. (2016).
Heteromodal cortical areas encode sensory-motor features of word meaning. The Journal
of Neuroscience, 36(38), 9763-9769.
Fernandino, L., Humphries, C. J., Seidenberg, M. S., Gross, W. L., Connant, L. L., & Binder, J.
R. (2015). Predicting brain activation patterns associated with individual lexical concepts
based on five sensory-motor attributes. Neurophychologia, 76, 17-26.
Fernandino, L., Tong, J-Q., Conant, L. L., Humphries, C. J. & Binder, J. R. (2021). Decoding the
information structure underlying the neural representation of concepts. Neuroscience,
119(6).
Ferrari, M. & Quaresima, V. (2012). A brief review on the history of human functional nearinfrared spectroscopy (fNIRS) development and fields of application. NeuroImage, 63,
921-935.
Fischler, I., Bloom, P. A., Childers, D. G., Roucos, S. E., & Perry Jr, N. W. (1983). Brain
potentials related to stages of sentence verification. Psychophysiology, 20(4), 400-409.
Fishburn, F. A., Norr, M. E., Medvedev, A. V., & Vaidya, C. J. (2014). Sensitivity of fNIRS to
cognitive state and load. Frontiers in Human Neuroscience, 8(76), 1-11.
Fodor, J. A. (1983). The Modularity of Mind, Cambridge: MIT Press.
Fu, G., Wan, N. J. A., Baker, J. M., Montgomery, J. W. Evans, J. L., & Gillam, R. B. (2016). A
proof of concept study of function-based statistical analysis of fNIRS data: Syntax

53
comprehension in children with specific language impairment compared to typicallydeveloping controls. Frontiers in Behavioral Neuroscience, 10(108), 1-15.
Furusho, J., Suzuki, A., Takakusa, Y., Kawaguchi, F., Ichikwa, N., & Kato, T. (2002).
Simultaneous study of interictal EEG and near-infrared spectroscopy in a boy with
epilepsy. International Congress Series, 1232, 673-676.
Gallese, V., Fadiga, L., Fogassi, L., & Rizzolatti, G. (1996). Action recognition in the premotor
cortex. Brain, 119, 593-609.
Glenberg, A. M., Robertson, D. A., David, A., 2000. Symbol grounding and meaning: A
comparison of high-dimensional and embodied theories of meaning. Journal of Memory
and Language, 43(3), 379-401.
Glenberg, A., M., Havas, D., Becker, R., & Rinck, M. (2005). Grounding Language in Bodily
States: The Case for Emotion. In Grounding Cognition: The Role of Perception and
Action in Memory, Language, and Thinking, D. Pecher & R. A. Zwaan (Eds.), pp 224245. Cambridge, UK: Cambridge University Press.
Gomez, P., Ratcliff, R., & Perea, M. (2007). A model of the go/no-go task. Journal of
Experimental Psychology: General, 136(3), 389-413.
González, J., Barros-Loscertales, A., Pulvermüller, F., Meseguer, V., Sanuán, A., Belloch, V., &
Ávila, C. (2006). Reading cinnamon activates olfactory brain regions. NeuroImage,
32(2), 906-912.
Gordon, B. & Caramazza, A. (1982). Lexical decision for open- and closed-class words: Failure
to replicate differential frequency sensitivity. Brain and Language, 15, 143-160.
Gouvea, A. C., Phillips, C., Kazanina, N., & Poeppel, D. (2010). The linguistic processes
underlying the P600. Language and Cognitive Processes, 25(2), 149-188.

54
Halberstadt, J., Winkelmann, P., Niednthal, P. M., & Dalle, N. (2009). Embodied conception:
How embodied emotion concepts guide perception and facial action. Psychological
Science, 20(10). 1254-1261.
Hall, M., Chaudhary, U., Rey, G., & Godavarty, A. (2013). Fronto-temporal mapping and
connectivity using NIRS for language-related paradigms. Journal of Neurolinguistics,
26(1), 178-194.
Hargreaves, I. S., Leonard, G. A. Pexman, P. M., Pittman, D. J., Siakaluk, P. D., & Goodyear, B.
G. (2012). The neural correlates of the body-object interaction effect in semantic
processing. Frontiers in Human Neuroscience, 6(22), 1-8.
Harnad, S. (2003). Symbol grounding problem. In, Encyclopedia of Cognitive Science. Nature
Publishing Group: Macmillan.
Harpaintner, M., Sim, E-J., Trumpp, N. M., Ulrich, M., & Kiefer, M. (2020). The grounding of
abstract concepts in the motor and visual system: An fMRI study. Cortex, 124, 1-22.
Hauk, O. & Pulvermüller, F. (2004). Effects of word length and frequency on the human eventrelated potential. Clinical Neurophysiology, 115, 1090-1103.
Hebb, D. O. (1949). The Organization of Behavior: A Neuropsychological Theory. New York:
Wiley.
Heinze, H., Muente, T., & Kutas, M. (1998). Context effects in a category verification task as
assessed by event-related brain potential (ERP) measures. Biological Psychology, 47,
121-135.
Herold, F., Wiegel, P., Scholkmann, F., & Müller, N. G. (2018). Applications of functional nearinfrared spectroscopy (fNIRS) neuroimaging in exercise–cognition science: A systematic,
methodology-focused review. Journal of Clinical Medicine, 7(12), 466.

55
Herrmann, M. J., Huter, T., Plichta, M. M., Ehlis, A-C., Alpers, G. W., Mühlberger, A., &
Fallgatter, A. J. (2008). Enhancement of activity of the primary visual cortex during
processing of emotional stimuli as measured with event-related functional near-infrared
spectroscopy and event-related potentials. Human Brain Mapping, 29, 28-35.
Hino, Y., Kusunose, Y., Lupker, S.J., & Jared, D. (2013). The processing advantage and
disadvantage for homophones in lexical decision tasks. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 39(2), 529-551.
Hino, Y., & Lupker, S. J. (1996). Effects of polysemy in lexical decision and naming: An
alternative to lexical access accounts. Journal of Experimental Psychology: Human
Perception and Performance, 22(6), 1331–1356.
Hock, C., Villringer, K., Müller-Spahn, F., Wenzel, R., Heekeren, H., … Villringer, A. (1997).
Decrease in parietal cerebral hemoglobin oxygenation during performance of a verbal
fluency task in patients with Alzheimer’s disease monitored by means of near-infrared
spectroscopy (NIRS) - correlation with simultaneous rCBF-PET measurements. Brain
Research, 755, 293-303.
Hoenig, K., Sim, E-J., Bochev, V., Herrnberger, B., & Kiefer, M. (2008). Journal of Cognitive
Neuroscience, 20(10), 1799-1814.
Hofmann, M. J., Herrmann, M. J., Dan, I., Obrig, H., Conrad, M., Kuchinke, L., Jacobs, A. M. &
Fallgatter, A. J. (2008). Differential activation of frontal and parietal regions during
visual word recognition: An optical topography study. NeuroImage, 40(3), 1340-1349.
Holcomb, P. J., Grainger, J., & O’Rourke, T. (2002). An electrophysiology study of the effects
of orthographic neighborhood she on printed word perception. Journal of Cognitive
Neuroscience, 14(6), 938-950.

56
Hoshi, Y., Tsou, B., Villock, V., Tanosaki, M., Iguchi, Y., … Oda, I. (2013). Spatiotemporal
characteristics of hemodynamic changes in the human lateral prefrontal cortex during
working memory tasks. NeuroImage, 20, 1493-1504.
Jasper, H. H. (1958). Report of the committee on methods of clinical examination in
electroencephalography. Electroencephalography and Clinical Neurophysiology, 10,
370-375.
Johnson, M. (2017). Embodied mind, meaning, and reason: How our bodies give rise to
understanding. The University of Chicago Press: Chicago.
Juhasz, B. J., Yap, M. J., Dicke, J., Taylor, S. C., & Gullick, M. M. (2011). Tangible words are
recognized faster: The grounding of meaning in sensory and perceptual systems.
Quarterly Journal of Experimental Psychology, 64, 1683-1691.
Kanske, P. & Kotz, S.A. (2007). Concreteness in emotional words: ERP evidence from a
hemifield study. Brain Research, 1148, 138-148.
Kaschak, M. P., & Glenberg, A. M. (2000). Constructing meaning: The role of affordances and
grammatical constructions in sentence comprehension. Journal of
Memory and Language 43, 508–529.
Khan, M. J., Hong, M. J., & Hong, K-S. (2014). Decoding of four movement directions using
hybrid NIRS-EEG brain-computer interface. Frontiers in Human Neuroscience, 8, 244.
Kiefer, M. & Pulvermuller, F. (2012). Conceptual representations in mind and brain: Theoretical
developments, current evidence, and future directions. Cortex, 48(7), 805-825.
Kiefer, M., Sim, E-J., Herrnberger, B., Grothe, J., & Hoenig, K. (2008). The sound of concepts:
Four markers for a link between auditory and conceptual brain systems. The Journal of
Neuroscience, 28(47), 12224-12230.

57
Kiefer, M., Sim, E-J., Liebich, S., Hauk, Olaf, & Tanaka, J. (2007). Experience-dependent
plasticity of conceptual representations in human sensory-motor areas. Journal of
Cognitive Neuroscience, 19(3), 525-542.
Kiefer, M. & Trumpp, N. M. (2012). Embodiment theory and education: The foundations of
cognition in perception and action. Trends in Neuroscience and Education, 1(1), 15-20.
Kontos, A. P., Huppert, T. J., Beluk, N. H., Elbin, R. J., Henry, L. C., French, J., … Collins, M.
W. (2014). Brain activation during neurocognitive testing using functional near-infrared
spectroscopy in patients following concussion compared to healthy controls. Brain
Imaging and Behavior, 8(4), 621-634.
Kounios, J., Green, D. L., Payne, L., Fleck, J. I., Grondin, R., & McCrae, K. (2009). Semantic
richness and the activation of concepts in semantic memory: Evidence from event-related
potentials. Brain Research, 1282, 95-102.
Khrizman, T. P. (1973). Characteristics of interdental relationships in electrical processes of the
brain in 2-3 year old children during voluntary motor acts. Voprosy Psikologii, 19, 107117.
Kuchinke, L. & Mueller, C.J. (2019). Are there similarities between emotional and familiaritybased processing in visual word recognition? Journal of Neurolinguistics, 49, 84-92.
Kuperman, V., Estes, Z., Brysbaert, M., & Warriner, A. B. (2014). Emotion and language:
Valence and arousal affect word recognition. Journal of Experimental Psychology:
General, 143(3), 1065-1081.
Lakoff, G., & Johnson, M. (1980). The metaphorical structure of the human conceptual system.
Cognitive Science, 4, 195-208.

58
Lakoff, G., & Johnson, M. (1999). Philosophy in the flesh: The embodied mind and its challenge
to western thought. Basic books.
Langacker, R. (1987). Foundations of cognitive grammar, Theoretical prerequisites. Vol. 1,
Stanford, CA: Stanford University Press.
Lee, H., Lee, Y., Tae, J., & Kwon, Y. (2019). Advantage of the go/no-go task over the yes/no
lexical decision task: ERP indexes of parameters in the diffusion model. PLOS ONE,
14(7), e0218451.
Li, D., Song, D. & Wang, T. (2020). Concreteness and imageability and their influences on
Chinese two-character word recognition. Reading & Writing, 33, 1443-1476.
Luck, S. J. (2014). An introduction to the event-related potential technique. (2nd Ed). Cambridge:
MIT Press.
Machery, E. (2016). The amodal brain and the offloading hypothesis. Psychonomic Bulletin &
Review, 23, 1090-1095.
Measso, G. & Zaidel, E. (1990). Effect of response programming on hemispheric differences in
lexical decision. Neuropsychologia, 28(7), 635-646.
Medvedev, A. V., Kainerstorfer, J. M., Borisov, S. V., & VanMeter, J. (2011). Functional
connectivity in the prefrontal cortex measured by near-infrared spectroscopy during ultra
rapid object recognition. Journal of Biomedical Optics, 16(1), 1-10.
Meteyard, L., Cuadrado, S. R., Bahrami, B., & Vigliocco, G. (2012). Coming of age: A review
of embodiment and the neuroscience of semantics. Cortex, 48, 788-804.
Miles, L., Nind, L., & Macrae, C. (2010). Moving through time. Psychological Science, 21(2),
222-223.

59
Moffat, M., Siakaluk, P. D., Sidhu, D. M., & Pexman, P. M. (2015) Situated conceptualization
and semantic processing: Effects of emotional experience and context availability in
semantic categorization and naming tasks. Psychonomic Bulletin & Review, 22(2), 408419.
Moosmann, M., Ritter, P., Krastel, I., Brink, A., Thees, S., Blankeburg, F., … Villringer, A.
(2003). Correlates of alpha rhythm in functional magnetic resonance imaging and near
infrared spectroscopy. NeuroImage, 20, 145-158.
Morioka, H., Kanemura, A., Morimoto, S., Yoshioka, T., Oba, S., Kawanabe, M., & Ishii, S.
(2014). Decoding spatial attention by using cortical currents estimated from
electroencephalography with near-infrared spectroscopy prior information. NeuroImage,
90, 128-139.
Morrow, D. G., & Clark, H. H. (1988). Interpreting words in spatial descriptions. Language and
Cognitive Processes, 3, 275-291.
Muraki, E.J., Sidhu, D.M., & Pexman, P.M. (2019). Mapping semantic space: property norms
and semantic richness. Cognitive Processing, 21, 637-649.
Nakadoi, Y., Sumitani, S., Watanabe, Y., Akiyama, M., Yamashita, N., & Ohmori, T. (2012).
Multi-channel near-infrared spectroscopy shows reduced activation in the prefrontal
cortex during facial expression processing in pervasive developmental disorder.
Psychiatry and Clinical Neurosciences, 66, 26-33.
Nathoo, F. S., Kilshaw, R. E., & Masson, M. E. (2018). A better (Bayesian) interval estimate for
within-subject designs. Journal of Mathematical Psychology, 89¸ 1-9.

60
Neville, H. J., Kutas, M., Chesney, G., & Schmidt, Al. L. (1986). Event-related brain potentials
during initial encoding and recognition memory of congruous and incongruous words.
Journal of Memory and Language, 25(1). 75-92.
Newcombe, P. I., Campbell, C., Siakaluk, P. D., & Pexman, P. M. (2012). Effects of emotional
and sensorimotor knowledge in semantic processing of concrete and abstract nouns.
Frontiers in Human Neuroscience, 6(275), 1-15.
Niedenthal, P.M., Barsalou, L.W., Ric, F., & Krauth-Gruber, S. (2005). Embodiment in the
acquisition and use of emotion knowledge. In L. Feldman Barrett, P.M. Niedenthal, & P.
Winkielman (Eds.), Emotion and consciousness (pp. 21-50). New York: Guilford.
Pecher, D. (2001). Perception is a two-way junction: Feedback semantics in word recognition.
Psychonomic Bulletin & Review, 8, 545-551.
Pecher, D., Boot, I., & Van Dantzig, S. (2011). Abstract concepts: Sensory-motor ground,
metaphors, and beyond. In Brian Ross (Ed.) The Psychology of Learning and Motivation,
54, 217-248.
Pecher, D., & Zwaan, R. A. (2005). Introduction to grounding cognition. Grounding Cognition:
The Role of Perception and Action in Memory, Language, and Thinking, 1-7.
Petzold, G. C. & Murthy, V. N. (2011). Role of astrocytes in neurovascular coupling. Neuron,
71, 782-797.
Pexman, P. M. (2012). Meaning based influences on visual word recognition. In J. S. Adelman
(Ed.), Visual word recognition. Vol. 2 (pp. 24-43). New York: Psychology Press.
Pinti, P., Tachtsidis, I., Hamilton, A., Hirsch, J., Aichelburg, C., Gilbert, S., & Burgess, P. W.
(2020). The present and future use of functional near‐infrared spectroscopy (fNIRS) for
cognitive neuroscience. Annals of the New York Academy of Sciences, 1464(1), 5.

61
Plaut, D., McClelland, J., Seidenberg, M., & Patterson, K. (1996). Understanding normal and
impaired word reading: Computational principles in quasi-regular domains.
Psychological Review, 10.3, 56-115.
Popp, M., Trumpp, N. M., & Kiefer, M. (2016). Feature-specific event-related potential effects
to action- and sound-related verbs during visual word recognition. Frontiers in Human
Neuroscience, 10, 637.
Power, J.D., Barnes, K.A., Snyder, A.Z., Schlaggar, B.L., & Petersen, S. E. (2012). Spurious but
systematic correlations in functional connectivity MRI networks arise from subject
motion. Neuroimage, 59(3), 2142-2154.
Pulvermuller, F. (2013) Semantic embodiment, disembodiment or misembodiment? In search of
meaning in modules and neuron circuits. Brain & Language, 127, 86-103.
Ralph, M. A. L., Jefferies, E., Patterson, K., & Rogers, T. T. (2017). The neural and
computational basis of semantic cognition. Nature Reviews Neuroscience, 18(1), 42-55.
Repp, B. H., & Knoblich, G. (2004). Perceiving action identity: How pianists recognize their
own performances. Psychological Science, 15, 604-609.
Rösler, F., Streb, J., & Haan, H. (2001). Event-related potentials evoked by verbs and nouns in a
primed lexical decision task. Psychophysiology, 38(4), 694-703.
Rossi, S., Telkemeyer, S., Wartenburger, I., & Obrig, H. (2012). Shedding light on words and
sentences: Near-infrared spectroscopy in language research. Brain & Language, 121,
152-163.
Sakatani, K., Xie, Y., Lichty, W., Li, S., & Zuo, H. (1998). Language-activated cerebral blood
oxygenation and hemodynamic changes of the left prefrontal cortex in poststroke aphasic
patients: A near-infrared spectroscopy study. Stroke, 29, 1299-1304.

62
Sato, H., Kiguchi, M., Maki, A., Fuchino, Y., Obata, A., Yoro, T., & Koizumi, H. (2006).
Within-subject reproducibility of near-infrared spectroscopy signals in sensorimotor
activation after 6 months. Journal of Biomedical Optics, 11(1), 014021.
Sato, H., Yahata, N., Funane, T., Takizawa, R. Katura, T., Atsumori, H., … Kasai, K. (2013). A
NIRS-fMRI investigation of prefrontal cortex activity during a working memory task.
NeuroImage, 83, 158-173.
Sawan, M., Salam, M., Gelinas, S., Le Lan, J., Lesage, F., & Nguyen, D. K. (2012). Combined
NIRS-EEG remote recordings for epilepsy and stroke real-time monitoring. In Circuits
and Systems (ISCAS), 2012 IEEE International Symposium on Circuits and Systems. (pp.
13-16). IEEE.
Schecklmann, M., Elis, A-C., Plichta, M. M., & Fallgatter, A. J. (2008). Functional near-infrared
spectroscopy: A long-term reliable tool for measuring brain activity during verbal
fluency. NeuroImage, 43(1), 147-155.
Schroeter, M. L., Zysset, S., Kupka, T., Kruggel, F., & von Cramon, D. Y. (2002). Near-infrared
spectroscopy can detect brain activity during a color-word matching Stroop task in an
event-related design. Human Brain Mapping, 17, 61-71.
Siakaluk, P. D., Knol, N., & Pexman, P. M. (2014). Effects of emotional experience for abstract
words in the Stroop task. Cognitive Science, 38(8), 1698-1717.
Siakaluk, P.D. Newcombe, P.I., Duffels, B., Li, E., Sidhu, D.M., Yap, M.J., & Pexman, P.M.
(2016). Effects of emotional experience in lexical decision. Frontiers in Psychology:
Cognitive Science, 7(1157).

63
Siakaluk, P. D., Pexman, P. M., Aguilera, L., Owen, W. J., & Sears, C. R. (2008a). Evidence for
the activation of sensorimotor information during visual word recognition: The bodyobject interaction effect. Cognition, 106, 433-443.
Siakaluk, P. D., Pexman, P. M., Sears, C. R., Wilson, K., Locheed, K., & Owen, W. J. (2008b).
The benefits of sensorimotor knowledge: Body-object interaction facilitates semantic
processing. Cognitive Science, 32(3), 591-605.
Sidhu, D. M., Kwan, R., Pexman, P. M., & Siakaluk, P. D. (2014). Effects of relative
embodiment in lexical and semantic processing of verbs. Acta Psychologica, 149, 32-39.
Srinivasan, R. (1999). Methods to improve the spatial resolution of EEG. International Journal
of Bioelectromagnetism, 1(1), 102-111.
Stanfield, R.A. & Zwaan, R.A. (2001). The effect of implied orientation derived from verbal
context on picture recognition. Psychological Science, 12, 153-156.
Strain, E., Patterson, K., & Seidenberg, M. S. (1995). Semantic effects in single-word naming.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 1140-1154.
Strangman, G., Culver, J. P., Thompson, J. H., & Boas, D. A. (2002). A quantitative comparison
of simultaneous BOLD fMRI and NIRS recordings during functional brain activation.
NeuroImage, 17, 719-731.
Taylor, M. J., & Khan, S. C. (2000). Top-down modulation of early selective attention processes
in children. International Journal of Psychophysiology, 37, 135-147.
Tettamanti, M., Buccino, G., Saccuman, M. C., Gallese, V., Danna, M., Scifo, P., Fazio, F.,
Rizzolatti, G., Cappa, S. F., & Perani, D. (2006). Listening to action-related sentences
activates fronto-parietal motor circuits. Journal of Cognitive Neuroscience, 17(2), 273281.

64
Tillotson, S. M., Siakaluk, P. D., & Pexman, P. M. (2008). Body-object interaction ratings for
1,618 monosyllabic nouns. Behavior Research Methods, 40, 1075-1075.
Tomarken, A., Davidson, R. J., Wheeler, R. E., & Kinney, L. (1992). Psychometric properties of
resting anterior EEG asymmetry: Temporal stability and internal consistency.
Psychophysiology, 29, 576-592.
Tomasino, B., & Rumiati, R. I. (2013). Introducing the special topic “the when and why of
sensorimotor processes in conceptual knowledge and abstract concepts.” Frontiers in
Human Neuroscience, 7, 498.
Tong, Y., Rooney, E. J., Bergethon, P. R., Martin, J.M., Sassaroli, A., Ehrenberg, B. L., Toi,
V.V., Aggarwal, P., Ambady, N., & Fantini, S. (2005). Studying brain function with
near-infrared spectroscopy concurrently with electroencephalography. In Optical
Tomography and Spectroscopy of Tissue VI (Vol. 5693, pp 444-449). International
Society for Optics and Photonics.
Tousignant, C. & Pexman, P. M. (2012). Flexible recruitment of semantic richness: Context
modulates body-object interaction effects in lexical-semantic processing. Frontiers in
Human Neuroscience, 6, 53.
Trans Cranial Technologies Ltd. (2012). 10/20 System Positioning Manual, retrieved from
www.trans-cranial.com/local/manuals/10_20_man_v1_0_pdf
Tucker, M., & Ellis, R. (2001). The potentiation of grasp types during visual object
categorization. Visual Cognition, 8, 769-800.
Van Havermaet, L. R. & Wurm, L. H. (2014). Semantic effects in word recognition are
moderated by body-object interaction. Mental Lexicon, 9, 1-22.

65
Van Petten, C. & Kutas, M. (1990). Interactions between sentence context and word frequency in
event-related brain potentials. Memory and Cognition, 18(4),

380-393.

Verner, M., Herrmann, M. J., Troche, S. J., Roebers, C. M., & Rammsayer, T. H. (2013).
Cortical oxygen consumption in mental arithmetic as a function of task difficulty: A nearinfrared spectroscopy approach. Frontiers in Human Neuroscience, 7, 217.
Vigliocco, G., Meteyard, L., Andrews, M., & Kousta, S. (2009). Toward a theory of semantic
representation. Language & Cognition, 1, 219-248.
Vinson, D., Ponari, M., & Vigliocco, G. (2014). How does emotional content affect lexical
processing? Cognition & Emotion, 28(4), 737-746.
Visani, E., Canafoglia, L. Giloli, I., Rossi Sebastiano, D., Contarino, V. E., Duran, D., Panzica,
F. … (2015). Hemodynamic and EEG time-courses during unilateral hand movement in
patients with cortical mylclonus. An EEG-fMRI and EEG-TD-fNIRS study. Brain
Topography, 28, 915-925.
Wang, X., Shangguan, C., & Liu, J. (2019). Time course of emotion effects during emotion-label
and emotion-laden word processing. Neuroscience Letters, 699, 1-7.
Watanabe, A., Matsuo, K., Kato, N., & Kato, T. (2003). Cerebrovascular response to cognitive
tasks and hyperventilation measured by multi-channel near-infrared spectroscopy.
Journal of Neuropsychiatric Clinical Neuroscience, 15, 442-449.
Wellsby, M., Siakaluk, P. D., Owen, W. J. & Pexman, P. M. (2011). Embodied semantic
processing: The body-object interaction effect in a non-manual task. Language &
Cognition, 3(1), 1-14.

66
Wheeler, M. E., Petersen, S. E., & Buckner, R. L. (2000). Memory’s echo: Vivid remembering
reactivates sensory-specific cortex. Proceedings of the National Academy of Sciences,
11125-11129.
Willems, R. M., Özyürek, A., Hagoort, P. (2008). Seeing and hearing meaning: ERP and fMRI
evidence of word versus picture integration into a sentence context. Journal of Cognitive
Neuroscience, 20(7), 1235-1245.
Wilson, N. L., & Gibbs, R. W. (2007). Real and imagined body movement primes metaphor
comprehension. Cognitive Science, 31(4), 721-731.
Wilson-Mendenhall, C. D., Barrett, L. F., Simmons, W. K., & Barslou, L. W. (2011). Grounding
emotion in situated conceptualization. Neuropsychologia, 49(5), 1105-1127.
Witt, J. J., Kemmerer, D., Linkenauger, S. A., & Fulham, J. (2010). A functional role for motor
simulation in identifying tools. Psychological Science, 21(9), 1215-1219.
Xue, J., Marmolejo-Ramos, F., & Pei, X. (2015). The linguistic context effects on processing of
body-object interaction words: An ERP study on second language learners. Brain
Research, 1613.
Yap, M. J., Pexman, P. M., Wellsby, M., Hargreaves, I. S., & Huff, M. J. (2012). An abundance
of riches: Cross-task comparisons of semantic richness effects in visual word recognition.
Frontiers in Human Neuroscience, 6, 72.
Yap, M. J. & Seow, C. S. (2014). The influence of emotion on lexical processing: Insights from
RT distributional analysis. Psychonomic Bulletin & Review, 21, 526-533.
Yap, M. J., Tan, S. E., Pexman, P. M., & Hargreaves, I. S. (2011). Is more always better? Effects
of semantic richness on lexical decision, speeded pronunciation, and semantic
classification. Psychonomic Bulletin & Review, 18, 742-750.

67
Yap, M. J., Yim, G. Y., & Pexman, P. M. (2015). Semantic richness effects in lexical decision:
The role of feedback. Memory & Cognition, 43, 1148-1167.
Ye, J. C., Tak, S., Jang, K. E., Jung., J., & Jang, J. (2009). NIRS-SPM: Statistical parametric
mapping for near-infrared spectroscopy. Neuroimage, 44, 428-447.
Yerys, B. E., Jankowski, K. F., Shook, D., Rosenberger, L. R., Barnes, K. A., Berl, M. M., … &
Gaillard, W. D. (2009). The fMRI success rate of children and adolescents: Typical
development, epilepsy, attention deficit/hyperactivity disorder, and austism spectrum
disorders. Human Brain Mapping, 30(10), 3426-3435.
Zama, T. Takahashi, Y., & Shimada, S. (2015). Simultaneous measurement of
electroencephalography and near-infrared spectroscopy during voluntary motor
preparation. Scientific Reports, 5(1), 1-9.
Zama, T. Takahashi, Y., & Shimada, S. (2019). Simultaneous EEG-NIRS measurement of the
inferior parietal lobule during a reaching task with delayed visual feedback. Frontiers in
Neuroscience, 13, 301.
Zwaan, R. A. & Madden, C. J. (2005). Embodied sentence comprehension. In Grounding
Cognition: The Role of Perception and Action in Memory, Language, and Thinking, D.
Pecher & R. A. Zwaan (Eds.), pp 224-245. Cambridge, UK: Cambridge University Press

68
CHAPTER 3:
Using Functional Near Infrared Spectroscopy to Investigate Grounded Semantics
Surprisingly little conceptual knowledge research has been completed using functional
near-infrared spectroscopy (fNIRS). This is surprising considering the close relationship between
the data that NIRS provides, and the more familiar BOLD signal produced by fMRI. Being a
comparatively affordable functional neuroimaging technique with applicable temporal and
spatial resolution capable of monitoring cortical activity, fNIRS is well suited to cognitive
research.
In lexical decision tasks (LDT), where participants are presented with a letter string and
must decide whether the letter string is a real word or not, increases in oxygenated hemoglobin
concentration ([O2Hb]) have been observed during nonword trials in the left superior frontal
gyrus and left inferior parietal gyrus alongside decreases in deoxygenated hemoglobin
concentration ([HHb]) in the left inferior frontal gyrus (Hofmann et al., 2008). Similarly, during
nonword trials, increased [O2Hb] has been observed in the left frontal lobe in children (Sela et
al., 2011), in older adults (Amiri et al., 2014), and in individuals with dyslexia (Sela, et al.,
2014). Employing a go/no-go procedure in a LDT, Chen et al. (2010) recorded an increase in
[O2Hb] in the left temporal lobe for words with high phonographic neighbourhood ratings
(words that have many similar sounding word neighbours when replacing a single phoneme with
another in the same location). In an auditory lexical decision task, greater increases of [O2Hb]
have been recorded in fronto-temporal regions when participants are processing phonetically
legal nonwords than for illegal nonwords, while greater increases were observed for all
nonwords than for real words (Rossi et al., 2011).

69
In an auditory sentence verification variation of a semantic categorization task (SCT),
where participants are typically presented with only real words and asked to decide whether the
presented word matches a decision criterion (e.g., Is this a verb?), bilateral increases in [O2Hb] in
the inferior frontal gyrus were observed in response to syntactically anomalous sentences
(Noguchi et al., 2002). When listening to spoken language, increased [THb] has been observed in
the left temporal lobe when the heard language changes (Rossi et al., 2012). In another test of
semantic processing, participants were presented with two words sequentially that comprised
two-word phrases and were asked to read the words silently. Increased [O2Hb] was observed in
Broca’s and Wernicke’s areas on trials where the phrase was anomalous as compared to trials
when the phrase was semantically correct (Horovitz & Gore, 2004). Similarly, increased [O2Hb]
has been observed in Broca’s area during metaphor comprehension (Schneider et al., 2014).
These few NIRS studies of semantic and conceptual knowledge processing demonstrate that the
brain is hemodynamically responsive to semantic task demands.
Given the fact that NIRS is particularly suited to monitoring the prefrontal cortex (PFC)
through the forehead, it has become increasingly applied to studying emotional stimuli as the
PFC is among the brain regions implicated in emotional processing, experience, and regulation
(Balconi et al., 2011; Balconi et al, 2015; Damasio, 1996). The orbitofrontal PFC has been
implicated in the process of assigning emotional value and saliency to exogeneous stimuli (Doi
et al., 2013). Activation of the PFC has been associated with activation of the autonomic nervous
system when one is presented with emotional stimulation (Tanida et al., 2007). When processing
emotional stimuli, increases in [O2Hb] have been observed in the right PFC for negative or
avoidance stimuli and in the left PFC for positive or approach stimuli, while bilateral [O2Hb]
PFC increases have been observed when participants are reinterpreting emotional stimuli

70
(Balconi et al., 2015; Herrmann et al., 2003; Kida & Shinohara, 2013a; Minagawa-Kawai et al.,
2009). These PFC hemodynamic patterns to emotional stimuli have been interpreted as being
either due to the emotional experience itself or with increased activity due to attending to and
processing emotional stimuli (Hermann et al., 2003). When anticipating an incoming electric
shock in a conditioning experiment, greater [O2Hb] has been observed in the right PFC,
supporting the observation that the right PFC processes stimuli with negative valence more than
does the left PFC (Morinaga et al., 2007). Collectively, these data support the use of NIRS to
investigate lexical cognition.
Prefrontal Cortex Functions
The PFC has been attributed with a broad range of flexible behaviour, the scope of which
exceeds that of other animals (Miller & Cohen, 2001). In humans, the PFC is the most elaborated
brain region, likely due to the increased sensory and cognitive processing abilities that humans
possess (Miller & Cohen, 2001). The functional role attributed to the PFC has been described as
exceeding the summed processes of its parts (Wilson et al., 2010). Although data suggest that the
PFC is responsible for executive functioning (e.g., inhibition (Aron et al., 2003; Aron et al.,
2004; Pardo et al., 1990), response preparation (Picton et al., 2006), cognitive flexibility (Miller
& Cohen, 2001), and general intelligence (Duncan, 2001)), demonstrations of functional double
dissociations are rare (Wilson et al., 2010). Experimental observations of PFC activity tend to be
too task-specific to be interpreted as the actual function of the PFC, or so general (e.g., problem
solving) as to lack theoretical usefulness (see Ramnani & Owen, 2004). As a result, it may be
misleading to claim that any observed activation in the PFC reflects any one specific process.
Rather, given the pattern of activation in the PFC across various experiments, PFC activity may
reflect integrative processing of information generated and processed elsewhere in the brain

71
(Bechara et al., 2000). Given its key role in integrative functioning, the PFC is an ideal candidate
for neuroimaging studies of cognitive processes such decision making in semantic categorization
tasks.
Broadly speaking, lesions to the PFC frequently disrupt efficient organized behaviour,
resulting in either a failure to initiate appropriate behaviour, or disinhibited and distracted
behaviour (Duncan, 2001). PFC lesions often produce difficulty in following instructions when
those instructions vary (Cohen et al., 1999). These disruptions suggest that the PFC ultimately
controls human behaviour insofar as it decides when to initiate or withhold a particular behaviour
(Picton et al., 2006). Increased bilateral activity has been observed during go/no-go experiments
as compared to experiments that only require a go response, further supporting the interpretation
that the PFC plays a role in organizing behaviour (Casey et al., 1997). It has been suggested that
signals generated in the PFC serve to bias activity in other brain regions so as to efficiently meet
current task demands (Miller & Cohen, 2001). An example of this is in the Stroop task when
participants are presented with a coloured word stimulus that evokes automatic word and colour
recognition activity – the PFC is thought to be responsible for biasing the colour recognition in
favour of the stronger automatic semantic activation so as to facilitate task completion. Though it
is possible that the PFC stores some information, the breadth of stimuli that have been observed
to elicit activity in the PFC suggests that it may, instead, store algorithms so as to flexibly
process information stored elsewhere (Duncan, 2001; Wood & Grafman, 2003). The PFC’s
responsibility for the flexible application of processing algorithms would imply that the PFC
directs human behaviour by processing representations of stimuli, generated elsewhere in the
brain, rather than by stimuli themselves (Goldman-Rakic, 1987). With a behavioural goal
presented by one’s current context, the PFC would then be responsible for integrating the

72
product of more posterior brain regions so as to adapt to and achieve particular outcomes
(Ramnani & Owen, 2004). Furthermore, it could be concluded that the PFC is therefore
particularly involved in controlling attentional systems towards comparatively novel situations
that do not already have an established behavioural template (Shallice & Burgess, 1998).
Despite the broad and flexible range of executive functions attributed to the PFC, some
hemispheric differences have been observed, and some conclusions have been proposed
regarding localized processing. For example, the left PFC has been associated with the
processing of verbal information more so than the right PFC, though some activation in the right
hemisphere is also evident (Duncan, 2001). The right hemisphere, however, has been associated
with the processing of episodic memories when cognitive demands are low, though bilateral
activation has been observed in high demand contexts (Nolde et al., 1998). In go/no-go
experiments, individuals with lesions to the left PFC have frequently shown difficulty in
preparing an appropriate response, while individuals with lesions to the right PFC have shown
difficulty inhibiting responses (e.g., Garavan et al., 2006; Kawashima et al., 1996; Picton et al.,
2006; Verfaellie & Heilmann, 1987). Similar behaviour has been observed in a stop-signal
paradigm insofar as individuals with right PFC lesions had difficulty inhibiting their response,
with the amount of difficulty positively correlated with the degree of impairment (Aron et al.,
2003). This difficulty with inhibition in individuals with right PFC damage often results in a
more generalized impulsive pattern of behaviour (Miller, 1992). It is possible, however, that the
hemispheric differences that have been observed reflect interruptions to a more dynamic and
integrative function of the PFC than would be concluded from such findings. For example, in
macaques learning object-reward associations, regional lesions to the dorsal or ventral PFC of
either hemisphere did not prevent the ability to learn the relationships, yet bilateral PFC lesions

73
that included the premotor cortex did prevent the learning of object-reward relationships (Parker
& Gaffan, 1998). This observation would suggest that the bilateral PFC is flexible in its
processing of environmental information, with the ability to be recruited in multiple ways when
solving a variety of problems (Wilson et al., 2010; also see Duncan’s (2001) Adaptive Coding
Model). Broadly speaking, then, the PFC can be described as playing a role in executive
functioning, specifically in selective attention, filtering of perception, and biasing of behavioural
responses in light of integrated information from more posterior brain regions (Bechara et al.,
2000; Duncan, 2001; Koechlin et al., 2000; Wilson et al., 2010).
The PFC is frequently subdivided into smaller regions that have been investigated for
localized function. The dorsolateral prefrontal cortex (DL-PFC) evolved from motor regions of
the brain, has connections with the basal ganglia, and has a role in motor control (Wood &
Grafman, 2003). As a result, this region is thought to play a role in the representation of action
(Wood & Grafman, 2003). The DL-PFC has been observed to be active when an individual is
processing exogeneous information (Christoff & Gabrieli, 2000). The dorsomedial PFC, by
contrast, has been observed to activate during the processing of social information (Fletcher et
al., 1995; Goel et al., 1995), and activity in this region has been interpreted to reflect the degree
of relational complexity amongst environmental stimuli (Kroger et al., 2002). For example, to
assess whether an encountered animal is dangerous, one must have awareness of one’s efficacy,
the distance to and size and species of the animal, as well as other factors including one’s
proximity to shelter or the presence of peers. The ventromedial prefrontal cortex (VM-PFC),
with its connections to emotion and memory regions, has been implicated in the integration of
bodily signals that are indicative of emotional experience with perceptual signals and memories
(Doi et al., 2013; Wood & Grafman, 2003). The role of the VM-PFC has been described as

74
processing incoming information about the environment and the emotion information previously
categorized with familiar stimuli or to reactivate a learned emotion experience by signalling the
appropriate neural structures (Bechara et al., 2000). Lesions to the VM-PFC prevent one from
experiencing avoidance-related emotions such as that associated with taking a risk (e.g., the
Wisconsin Card Sorting Task; Berman et al., 1995; Buchsbaum et al., 2005). In not experiencing
risk-associated emotion states, individuals with VM-PFC lesions fail to learn to regulate their
behaviour in accordance with real-world consequences (Bechara et al., 2000). Lesions to the
ventrolateral PFC, by contrast, have been found to increase variability in one’s responses, which
has been interpreted as interference to one’s ability to update and monitor incoming information
(Picton et al., 2006).
Other localizations have been identified towards the rostral portions of the PFC. The
orbitofrontal prefrontal cortex (OF-PFC) is located immediately above the eye sockets and is
predominantly situated along the ventral surface, though some portion of it extends to directly
behind the forehead. The OF-PFC has been implicated in the processing of social and emotional
stimuli (Drevets, 2001; Koechlin et al., 2000), as well as in response inhibition (Siddiqui et al.,
2008). Lesions to the OF-PFC result in individuals who act irresponsibly and generally lack
affect (Hornak et al., 2003). In macaques in a go/no-go experiment with some rewarding stimuli,
individuals with lesioned OF-PFC were more likely to respond to no-go trials, and their
performance was worse still when the stimulus-reward pairings would change mid-experiment,
suggesting the OF-PFC plays a role in learning and cognitive flexibility (Iversen & Mishkin,
1970). Activity in this region has been observed when individuals are touched with stimuli that
produce pleasant or painful experiences more so than when touched by stimuli that produce
neutral experiences (Rolls et al., 2003). Interestingly, this activity was not found with

75
comparatively more intense pressure that produced neutral experiences, suggesting that activity
in the OF-PFC is not reflective of stimulus intensity, but rather, of the valence of sensory
experiences. The OF-PFC, as a result, has been interpreted as being involved in stimulusreinforcement learning (Rolls, 2004). Within the OF-PFC itself, activity in the medial region has
been observed when an individual is winning money, while activity in the lateral region has been
observed when money is lost (O’Doherty et al., 2001). That activity is observed in the OF-PFC
when money is used as a behavioural consequence highlights that the OF-PFC is responsive to
learned relationships as primary reinforcers and punishments (Rolls, 2004). Collectively, these
data suggest that the OF-PFC plays a role in assessing the emotional value of perceived stimuli
as well as processing the representations of the relationships between stimuli and their
behavioural consequences (Ramnani & Owen, 2004). As a result, the OF-PFC is one of the key
regions implicated in the somatic marker hypothesis (Bechara et al., 2000). This hypothesis
proposes that emotion experiences produce bodily states that affect cognition, and thus
behaviour, through a ‘body loop’ (Poppa & Bechara, 2018). Evidence for this hypothesis comes
from observations such as individuals with lesions to the OF-PFC failing to develop anticipatory
anxious skin responses in risk-taking scenarios (Bechara et al., 2000). Thus, the OF-PFC would
be implicated as a key region to investigate when one is perceiving stimuli with learned emotion
information (Barret & Simons, 2015).
The frontopolar cortex (FPC) is the region of cortex directly behind one’s forehead and is
the most anterior brain region. Activity in the FPC has been reliably observed during the retrieval
of episodic and recognition memory, suggesting its role in the processing of previously learned
information (Duncan & Owen, 2000; Wagner et al., 1998). This region has also been found to
activate during the processing of introspective states, further implicating the FPC as a brain

76
region responsible for higher order cognitive processing (Christoff & Gabrieli, 2000; Ramnani &
Owen, 2004). While the right FPC has been implicated in the retrieval of more familiar source
memory (Tulving, 1983), the amount of activity in the left FPC has been shown to scale with the
amount of perceptually detailed information within a particular memory retrieval (Ranganath et
al., 2000). These activity patterns have been interpreted as evidence that the FPC is at least partly
responsible for the binding of new information previously acquired knowledge, such as thinking
about a known concept in a new context (Kroger et al., 2002). Experimental data has shown
bilateral activation of the FPC during card sorting tasks (Berman et al., 1995), right hemisphere
activation when responding a target stimulus feature (Rao et al., 1997), and left hemisphere
activity during the comprehension of sentence semantics (Goel et al., 1997).
Despite the breadth of localized PFC functions, it has been observed that a wide variety
of tasks recruit similar regions of the PFC, thereby suggesting that, as mentioned above, the
function of the PFC is broad, and that its total capabilities exceed the sum of observed
localizations (Duncan, 2001). Broadly speaking, the PFC is recruited when current task demands
require top-down processing to guide behaviour in accordance with one’s intentions, or perhaps
to guide task-related behaviour when one’s context requires learned behaviour that generates a
less powerful neural signal than is automatically produced by a stimulus (Miller & Cohen, 2001;
Ozawa et al., 2014). The PFC, then, can be thought to employ contextually relevant processes
that are situationally adaptable so as to be reflective of previously learned associations (Wood &
Grafman, 2003). With well learned associations being stored in posterior cortical regions, the
PFC is recruited predominantly when one is performing a less familiar task or attending to less
familiar relationship amongst stimuli (Koechlin et al., 2000). That is, activity in the PFC is likely
not automatically elicited by a stimulus itself, but by the contextual relationships in which a

77
stimulus is involved (Ramnani & Owen, 2004). As a result, the PFC is a region of interest for
evidence of activity elicited by novel experimental tasks such as those in the experiments
presented here.
Cognitive Investigations of the PFC using NIRS
To date, there has been no widely agreed upon method to investigate or analyze NIRS
data in psychology (Bendall et al., 2016; Doi, 2013; Masataka et al., 2015). Various researchers
employ different devices, each with a variable number of channels as well as different tasks,
each with a variety of stimuli, and the data are neither processed nor analyzed consistently across
experiments (Bonetti et al., 2018). As a result, a degree of subjective analysis is required to draw
inferences from across NIRS studies. A recent review of NIRS experiments investigating PFC
activity found that approximately 90% of such studies were research into verbal fluency or
working memory, while the remaining 10% were loosely classified as “other” topics, sharing no
particular theme or focus (Bonetti et al., 2018). This is particularly interesting given that NIRS
has been noted as particularly well suited to the study of cognitive activity in the cortex,
especially amongst populations and in contexts that are otherwise difficult for other
neuroimaging technologies (Fishburn et al., 2014). A selection of the working memory and
verbal fluency research will be reviewed below so as to provide context for the predictions made
in the current program of research.
The most straightforward NIRS observation is that, during cognitive effort, cortical
hemodynamic responses ([THC], [O2Hb]) amplitudes scale linearly with task difficulty (Ayaz et
al., 2012; Fishburn et al., 2014). Simply employing one’s working memory has been found to
activate one’s PFC as evidenced by increased [O2Hb] (Hetu et al., 2013). Moriya & Sakatani
(2017) found that mental imagery, interpreted as the placing and maintenance of visual

78
information into working memory storage, improves working memory performance and is
associated with increased PFC [O2Hb] during working memory tasks. It has also been observed
that, as task demands increase in difficulty, neural activity becomes more anterior, independently
of successful task completion, suggesting that this anterior activity reflects mental effort alone
(Causse et al., 2017). An implication of this finding is that fNIRS is suitable for monitoring PFC
activity during cognitive tasks as effort itself is observable. For example, in a n-back working
memory experiment, PFC [O2Hb] levels increase and [HHb] levels decrease as n increases while
response latencies and accuracy worsen (Ang et al., 2014; Li et al., 2005). Furthermore, the time
course of the hemodynamic response has been found to be sufficient in identifying the difficulty
of a task that a participant is completing with an 84% accuracy for identifying hard blocks versus
idle activity, 68% accuracy for identifying easy blocks versus idle activity, and 73% accuracy for
identifying hard block from easy blocks (Ang et al., 2014). Similar data have been observed in
mental arithmetic tasks in that the difficulty of the operation is correlated with the time course
and amplitude of the PFC hemodynamic response (Ang et al, 2010; Tanida et al, 2004).
McKendrick et al. (2014) found that working memory training increased [O2Hb] levels in the left
dorsolateral PFC and right ventrolateral PFC; and that bilateral [O2Hb] levels in the ventrolateral
PFC were negatively correlated with working memory performance. Of particular relevance to
the research presented here is the observation that, in a verbal working memory task, the word
frequency of a given stimulus was found to influence participants’ behaviour (Berglund-Barraza
et al., 2019). Specifically, participants were slower to respond yet more accurate in their
responses, while the PFC hemodynamic response was lower when the target stimulus had low
print frequency. These data were interpreted as indicating that low frequency words are unique
compared to their high frequency counterparts, and were therefore easier to hold in working

79
memory, thereby facilitating accuracy and requiring less effort from the PFC. So, for PFC NIRS
data from working memory experiments, observed activity can be interpreted as the amount of
effort required to execute a task-relevant response rather than the amount of processing elicited
by a particular stimulus.
Within the verbal fluency experiments, it is generally found that simply engaging in the
task of spontaneously producing words that meet a certain criterion (e.g., words that start with
the letter “L”) produces bilateral increases of [O2Hb] and decreases [HHb], with this effect more
pronounced in the left hemisphere than the right (e.g., Kahlaoui et al., 2012). This effect has been
found to be even more pronounced in individuals with panic disorder insofar as the amount of
activity in the left inferior PFC is positively correlated with the severity of their panic disorder
symptoms (Nishimura et al., 2009). Similarly, it has been found that individuals with depression,
in addition to having less bilateral PFC activity in general, show less pronounced [O2Hb]
increases during verbal fluency tasks than do individuals without (Fu et al., 2018). Similar results
were reported by Noda et al. (2012) who observed a negative correlation between right frontal
and temporal [O2Hb] levels during verbal fluency tasks and participants’ severity of depressive
symptoms. Collectively, these data suggest that NIRS is able to detect mental effort in the PFC
as well as to discern between individual condition differences.
One area of PFC NIRS research that has received attention outside of working memory
and verbal fluency is emotion experience and control. Although emotional experiences are
thought to originate in subcortical limbic regions, it has been proposed that they are elaborated
on, processed in the context of incoming information, and regulated contextually in more
anterior brain regions, including the prefrontal cortex (Doi et al., 2013). As discussed above,
there is a lack of consistency in NIRS research, and this issue extends to emotion research as

80
well, with variability in the induction methods and behavioural tasks employed, with some
researchers calling for greater rigor and highlighting a need for more replication due to
inconsistencies in reported findings (Bendall et al, 2016). Regardless, the PFC is a region of
interest for cognitive research into emotion experiences because emotion is thought to influence
behaviour in both top-down and bottom-up manners (Bendall et al., 2016). That is, emotional
experience will influence behaviour automatically due to altered brain activity that originates in
more posterior limbic regions, as well as by causing an individual to employ cognitive strategies
to regulate emotion experiences so as to complete contextual task demands. It is the latter effect
that is likely to be captured by experiments monitoring PFC activity (Ozawa et al., 2014).
When presented with rewarding visual stimuli, [O2Hb] levels in the orbitofrontal cortex
have been found to increase (Rolls, 1996), and gentle pleasing touch has been found to increase
[O2Hb] levels in the bilateral anterior PFC (Kida & Shinohara, 2013a; 2013b). Emotion-laden
music has been found to produce [O2Hb] level changes in the PFC (Hermann et al., 2003; Tai &
Chau, 2009, Yang et al., 2007) with NIRS signals uniquely characterizing the valence of the
music with 70% accuracy (Moghimi et al., 2012a; 2012b), with negative valence and intensity
ratings producing larger peak [O2Hb] amplitudes. When presented with photos of their
grandchildren, grandmothers have increased [O2Hb] in the inferior and medial anterior PFC
regardless of the child’s expression, though this effect was more pronounced when the child was
smiling (Kida et al., 2014). Minagawa-Kawai et al. (2009) observed similar patterns in mothers
viewing photographs of their children. Increases in [O2Hb] have been observed bilaterally when
one is looking at negative images as compared to neutral images (Glotzbach et al., 2011). Hoshi
et al. (2011), however, observed this effect for negative images to be predominantly lateralized
in the right hemisphere, while also reporting bilateral [O2Hb] decreases when participants were

81
viewing positive images. Similarly, Ozawa et al. (2014) observed increased [O2Hb] in the right
anterior ventromedial PFC when participants were viewing negative stimuli. Herrmann et al.
(2003) found that the degree of self-monitoring during a mood induction was positively
correlated with the [O2Hb] increase in the PFC, whereas this correlation was not found in mood
inductions that were more self-reflective, suggesting that data obtained by NIRS from the PFC
does not reflect an emotional experience itself, but rather the amount of additional cognitive
activity associated with the experience.
In a depressed population, left dorsolateral PFC [O2Hb] levels were found to be higher at
rest in those individuals with comparatively greater frequencies of spontaneously occurring
positive thoughts (Koseki et al., 2013). Tanida et al. (2007) have reported that sympathetic
nervous system activity is positively correlated with right hemisphere PFC activity. However, it
is important to consider whether the PFC activity is a component of one’s emotional experiences
or whether it reflects one’s efforts to regulate one’s emotional experiences. When recording from
depressed individuals, it has been observed that the hemodynamic response in the right
dorsolateral PFC is negatively correlated with the severity of individuals’ anxiety symptoms,
whereas bilateral antero-medial PFC activity is similarly negatively correlated with overall
depressive symptom severity (Liu et al., 2014). These data were interpreted by the authors as the
result of reduced cognitive control that ultimately results in depressive symptoms. Support for
this interpretation comes from patterns of hemodynamic responses to distractions during the
processing of emotional stimuli. Van Dillen et al. (2009) found that distractor tasks increased
PFC hemodynamic response, while simultaneously reducing subjective reports of negative mood
states induced by negatively valenced images. Similarly, Ozawa & Hiraki (2017) extended this
finding by demonstrating that the speed of a participants’ finger tapping (the distractor task)

82
modulated the pattern of [O2Hb] levels in the dorsolateral PFC, and negative emotion ratings
with faster tapping leading to greater PFC activity and lower negative emotion ratings. These
data support the interpretation that the intensity of the PFC hemodynamic response
predominantly reflects cognitive control during emotion experiences.
In contrast to the data presented above, Hermann et al. (2003) did not observe any PFC
hemodynamic differences to uniquely identify the processing of positive or negative stimuli. It
has been suggested, however, that these differences in experimental results may stem from
differences in approach to data analysis (Ozawa et al., 2014). More specifically, Ozawa et al.
(2014) noted that, when positive and negative hemodynamic responses were compared to
baseline data, the perception of both positive and negative stimuli could be interpreted as
increasing PFC activity, whereas comparing the same hemodynamic responses to neutral stimuli,
no statistically significant effects were observable. The variability amongst the findings
presented above makes a coherent interpretation of NIRS technology’s ability to detect the
processing of emotional experience in the PFC rather difficult. Ozawa et al. (2014), however,
found that task-irrelevant emotional images presented prior to task completion will modulate the
hemodynamic response in the PFC elicited by an n-back working memory task, with greater
hemodynamic responses following the perception of negative stimuli, suggesting that the effects
of emotional control are evident in the PFC. In light of the working memory data presented
above, it can be concluded that any stimulus that requires additional contextual processing will
produce activity that is evident in the PFC with NIRS. Additionally, to the extent that the PFC is
involved in the simulation of learned emotion associations, these effects may be detectable as
well.

83
Although NIRS research has, to date, focused predominantly on verbal fluency and
working memory, some research has been done in other domains of cognitive science. A central
hypothesis to such research is that, as described above, brain regions that are more active have
greater glucose metabolism, and that this consumption of resources will result in an observable
increase in regional blood flow as homeostasis is restored (Bonetti et al, 2018). An extension of
this neurovascular coupling is that, when participants’ cognitive efforts increase, there should be
more brain activity, and thus more pronounced hemodynamic changes that are detectable by
fNIRS.
When presented with taste stimuli and tasked with committing the flavours to memory
for later recognition, bilateral PFC [O2Hb] increases have been observed, thus supporting the
interpretation that PFC activity reflects contextual integration and cognitive effort (Okamoto et
al., 2011). During the recognition phase of the taste memory experiment, PFC activity was
predominantly lateralized to the right hemisphere. Children diagnosed with ADHD performing a
Stroop task show less hemodynamic activity (typically right lateralized PFC) than do control
participants (Inoue et al., 2012; Jourdan Moser et al., 2009; Negoro et al., 2010). When
participants play video games for at least five minutes, lower levels of dorsolateral PFC [O2Hb]
are recorded (Matsuda & Hiraki, 2006). This observation likely reflects participants learning and
adapting to the video game’s demands and performing the five-minute block more easily. During
the Wisconsin card sorting task and a verbal fluency task, bilateral increases of PFC [O2Hb] were
evident in 60% of participants (Sumitani et al., 2015). Although this effect was not evident in
10% of participants, lateralized increases were observed in the remaining 30%. However, there
were individual differences in the hemispheric lateralization of this effect. These findings
suggest that hemodynamic responses to task demands are highly individualized. However, it is

84
also possible that individuals approach and complete the Wisconsin card sorting task in a variety
of ways, thereby adding variability to the role that one’s PFC plays in successful task
completion.
In a go/no-go experiment, healthy control participants exhibit bilateral inferior PFC
[O2Hb] increases and [HHb] decreases during trial blocks with no-go trials as compared to
blocks comprised exclusively of go trials (Herrmann et al., 2005). Similar observations have
been made during stop-signal experiments when participants must inhibit an initiated response
following a stop signal, with [O2Hb] increases most pronounced in the right hemisphere
(Boecker et al., 2007). Xiao et al. (2012) also observed inhibition-evoked right-hemisphere
dominant [O2Hb] increases during no-go trials in healthy control participants and found this
effect to be lessened in participants diagnosed with ADHD or autism. The lessened
hemodynamic response to inhibition efforts in ADHD and autism populations is correlated with
these individuals’ comparatively poorer task performances and is thought to reflect difficulty
with inhibition processes. Nguyen et al. (2020; 2021) reported increased functional connectivity,
as measured with fNIRS, in the right PFC as a function of successful go/no-go task completion
such that participants with higher accuracy had greater functional connectivity in the PFC.
Collectively, these data support the interpretations made above that NIRS technology can detect
effort in the PFC, and that this activity will likely reflect cognitive effort with a right hemisphere
lateralization if that effort is inhibitory in nature.
Experimental Predictions
Across the four experiments detailed below, lexical and semantic variables were
manipulated in one lexical decision task (LDT) and three semantic categorization tasks (SCTs) to
test predictions based in grounded cognition theory concerning the hemodynamic response in the

85
anterior prefrontal cortex using fNIRS. Broadly speaking, it was predicted that some
experimental conditions, having lower levels of lexical or semantic information, would be
comparatively more difficult regardless of whether that information was explicitly necessary for
successful task completion. fNIRS was employed to measure the hypothesized additional
cognitive effort needed to respond to stimuli in these conditions as operationalized by greater
[O2Hb] and lower [HHb] in the anterior PFC. Additionally, it was predicted that these same trials
would have slower response latencies and/or lower accuracy rates compared to those trials with
stimuli rated higher in lexical and semantic information, as detailed below.
Experiment 1 was a go/no-go LDT with go trials having either high or low word
frequency ratings. Behaviourally, it was predicted that words rated high in word frequency
would be responded to more quickly and accurately than words rated low in word frequency.
Word frequency was chosen as a lexical variable as a proof of concept to test whether cognitive
effort could be detected with fNIRS in single-trial conditions. It was predicted that, because high
frequency trials would be relatively easier trials for participants to recognize as real words,
would produce the fastest response latencies, highest accuracy, lowest PFC [O2Hb], and highest
[HHb]. No-go trials, requiring no response, cannot have response latency predictions made about
them in the current experimental design, though it is predicted that more errors will be made for
low frequency words than for high frequency words. However, hemodynamically, it was
predicted that no-go trials would require higher amounts of cognitive effort so as to inhibit
participants’ automatically generated response potential that is elicited by stimulus presentation,
and that this inhibition would be reflected in greater PFC [O2Hb] and lower [HHb] as a result,
and furthermore that these changes would be evident predominantly in the right hemisphere.
Two possible hemodynamic responses to the condition in which ‘yes’ responses were to be made

86
to words rated low in word frequency were predicted. The first possibility was that the pattern of
PFC hemodynamic response to low frequency go trials would look similar to responses to high
frequency trials, but with more cognitive effort required, as reflected in greater amplitudes for
[O2Hb] and lower amplitudes for [HHb]. The alternative hypothesis was predicated on the
assumption that stimulus presentation would automatically prime participants to respond and that
this automatically primed response is fully inhibited in no-go trials for nonwords. Low-frequency
words would be more difficult to recognize as a real word, and therefore inhibitory processes
would remain engaged for longer than they would for high frequency words. This would explain
the comparatively longer response latencies for low frequency words and would thus be
predicted to produce a right-lateralized PFC response, similar to, but of lower intensity than, the
fully inhibited no-go trials.
Experiment 2 was a SCT where go trials were either high or low in body-object
interaction (BOI) ratings. This experiment aimed to replicate the behavioural observation that
concrete nouns are responded to more quickly when rated high in BOI due to these nouns having
greater levels of grounded semantic richness obtained through bodily interactions with the
nouns’ referents, thus allowing them to be processed more efficiently. Although the decision
criterion for this SCT was “Is this an easily imageable word?”, it was predicted that concrete
nouns rated high in BOI would be processed more efficiently due to their higher levels of
grounded semantic richness and thereby produce faster response latencies and higher levels of
accuracy, despite BOI not being explicitly required for successful task completion. This
cognitive efficiency was predicted to produce the comparatively lowest amounts of PFC [O2Hb]
and the comparatively highest amounts of PFC [HHb]. The pattern of hemodynamic response to

87
no-go and low BOI word trials was predicted to be the same as for Experiment 1, with low BOI
word trials being examined for their similarity to high BOI word trials or inhibited no-go trials.
Experiments 3 and 4 were also SCTs, but with half of the word stimuli being concrete
nouns and half being abstract nouns. The decision criterion for Experiment 3 was “Is this an
abstract noun?”, and in Experiment 4 was “Is this a concrete noun?”. Half of each word type was
rated high in emotional experience (EE), and half low in EE. As detailed in Chapter 2, abstract
nouns are thought to be grounded in one’s associated emotion experiences with the nouns’
referents. As a result, the behavioural and hemodynamic predictions for Experiment 3 are the
same as for Experiments 1 and 2 with words rated high in EE being the easiest to process as
abstract nouns, and thus producing the fastest response times, highest accuracy, and least
pronounced hemodynamic responses. In Experiment 4, however, high EE concrete nouns are
predicted to be the more difficult trials because concrete nouns are not thought to be as grounded
in emotional experiences (see Chapter 2 for more discussion). As such, high EE concrete nouns
in Experiment 4 are predicted to have the slowest response latencies, least accurate responses,
and most pronounced hemodynamic responses. High EE trials will also be examined for their
similarity to successfully inhibited no-go trials.
Method
Participants. In each of the four experiments, participants were University of Northern
British Columbia students enrolled in psychology classes. 26 individuals participated in
Experiment 1, and 25 individuals participated in Experiments 2-4. No participants were involved
in more than one experiment. There were 20, 19, 18, and 19 female participants in Experiments
1-4 respectively. All participants asserted that they were native English speakers and had normal

88
or corrected-to-normal vision. All participants received bonus course credit for their
participation.
Apparatus.
Behavioural. For all four experiments, response latencies and accuracy were collected on
a Dell Optiplex 7010 desktop computer running Superlab (version 5.0.3) software. Word stimuli
were presented on a 21.5” Dell P2212H monitor in capitalized white Times New Roman 24point font on a black background. Participants were positioned at a table and instructed to ensure
that they could comfortably reach the spacebar of a keyboard for the duration of their
experiment.
NIRS. NIRS data were acquired by OxiTS software (version 3.1.4.1) running on a Dell
Optiplex 7010 desktop computer connected to a two-channel frequency domain device
(OxiplexTS, ISS, Champaign, IL). Each channel consisted of a single optode with four pairs of
emitters and a single detector, and with each emitter pair at a different distance from the detector.
Within each emitter pair, one emitter broadcast near-infrared light at 684 nm and the other at
830nm. Each optode was placed above the brow ridge, below electrodes 1 and 2 (left and right
hemisphere, respectively; see Figure 1, Chapter 4 for details), covering the anterior portion of the
dorsolateral prefrontal cortex.
Lexical Decision Tasks. Experiment 1 consisted of an LDT with a go/no-go procedure,
whereby the stimuli were letter strings (half being real words and the other half being nonwords)
and participants were instructed to press the spacebar when the words were presented and to not
press the spacebar when the nonwords were presented. Stimuli were the same as employed by
Sears et al. (2008; Table 1). High frequency and low frequency words were selected such that
there are no significant differences on orthographic neighbourhood, number of letters, number of

89
morphemes, number of syllables, and phonological neighbourhood size. There were 66 high
frequency words, 66 low frequency words, and 132 non-words, for a total of 264 trials in this
experiment. An intertrial interval of 15,000 ms was used to allow prefrontal hemodynamics to
return to baseline after each trial before a new letter string was presented.
Semantic Categorization Tasks. For Experiments 2, 3, and 4, participants completed a
go/no-go SCT, wherein all the stimuli were words (half being exemplars of a particular decision
criterion and the other half not being exemplars of that decision criterion; see below for the
decision criteria used in these SCTs), and participants were instructed to press the spacebar when
the exemplar words were presented and to not press the spacebar when the non-exemplar words
were presented. For Experiment 2, the decision criterion was ‘easily imageable’; thus,
participants were instructed to press the spacebar when easily imageable words were presented
(e.g., cage and prince), and not to press the spacebar when less imageable words were presented
(e.g., dose and node). For Experiment 3, the decision criterion was ‘abstract noun’; thus,
participants were instructed to press the spacebar when abstract nouns were presented (e.g.,
advance and discipline), and not to press the spacebar when concrete nouns were presented (e.g.,
basement and elbow). For Experiment 4, the decision criterion was ‘concrete noun’; thus,
participants were instructed to press the spacebar when concrete nouns were presented, and not
to press the spacebar when abstract nouns were presented. As in Experiment 1, a 15,000 ms
inter-trial interval was used in all three SCT experiments. As noted in Chapter 2,
electroencephalography data were recorded concurrently, and those data are described in detail
in Chapter 4.
As noted, Experiment 2 used the decision criterion of, “Is this word easily imageable”.
The stimuli were the same as those used in Hargreaves et al. (2012), such that 36 words were

90
rated high in BOI, 36 words were rated low in BOI, and these words were intermixed with 60
filler words that were rated low in imageability. Words rated high in BOI were matched with
words rated low in BOI for word length, number of features, familiarity ratings, concreteness,
imageability, print frequency, contextual diversity, orthographic neighbourhood size, and
phonological neighbourhood size (Table 2).
For Experiments 3 and 4, the stimuli were the same as those used in Newcombe et al.
(2012), such that the stimuli consisted of 70 concrete words and 70 abstract words. Half of each
word type were rated high in EE, and the other half were rated low in EE. The high-EE words
and low EE words were matched on HAL log frequency, age of acquisition, orthographic
neighbourhood size, number of letters, number of syllables, number of morphemes, concreteness,
and imageability (Table 3). As noted, for Experiment 3, participants were instructed to press the
spacebar only to the abstract words, whereas for Experiment 4, participants were instructed to
press the spacebar only to the concrete words.
Analyses.
Behavioural Data Processing. The behavioural data were processed in the following
ways. First, participants were excluded from all analysis if their LDT or SCT accuracy rate was
under 70%. Considering accuracy, no participants were removed from Experiments 1 and 4, one
participant was removed from Experiment 2, and two participants were removed from
Experiment 3. Next, words that were correctly responded to less frequently than 70% of the time
were excluded from all analyses. No words were removed from Experiment 1. Eight words were
removed from Experiment 2 (1 high BOI and 7 low BOI; 6.1% of the data), five words were
removed from Experiment 3 (all were low EE; 3.5% of the data), and eight words were removed
from Experiment 4 (7 high EE and 1 low EE; 5.7% of the data). Next, all trials with incorrect

91
responses were removed from all analyses. One hundred forty-four responses were removed
from Experiment 1 (2.1% of the data), 392 responses were removed from Experiment 2 (12.4%
of the data), 364 responses were removed from Experiment 3 (11.3% of the data), and 332
responses were removed from Experiment 4 (10.2% of the data). Extreme response latencies
were also removed from all analyses. Extreme response latencies were identified as faster than
250 ms and those slower than 2000 ms. No trials were removed for being faster than 250 ms in
any experiment. Nine trials from Experiment 1 (<.01% of the data), 22 trials from Experiment 2
(<.01% of the data), 71 trials from Experiment 3 (.02% of the data), and 26 trials from
Experiment 4 (<.01% of the data) were removed for being slower than 2000 ms. Finally, for each
participant, response latency outliers were identified as those greater than 2.5 standard deviations
from the cell mean of each condition and then removed from all analyses. Using this procedure,
69 trials from Experiment 1 (.01% of the data), 17 trials from Experiment 2 (<.01% of the data),
nine trials from Experiment 3 (<.01% of the data), and 18 trials from Experiment 4 (<.01% of the
data) were removed. From the remaining trials, average condition means (high- and lowfrequency, high- and low-BOI, and high- and low-EE) were calculated for each participant and
then averaged across participants. Separately, item-level analyses were conducted on the same
remaining data by averaging response time latencies for each single word in each experiment.
Averages were tested using Bayes Factors (ttestBF; R, version 3.5.3).
NIRS. To control for movement artifacts (that take the form of impossibly extreme values
produced by an optode lifting from the skin), all Sat% scores greater than 100% were replaced
with a value of 100% - a highly implausible value, but a possible one. All concentrations (µM) of
THb, O2Hb, and HHb greater than 200 µM were similarly replaced with the possible score of
200 µM. All NIRS data for the entire experiment were then filtered with a zero-phase, forward-

92
reverse Butterworth filter with a low pass cut-off frequency of 0.5 Hz, high pass 1.5 Hz, ring and
roll of 40 dB and 3 dB, respectively. Using each participant’s entire data, standard deviations
were then calculated for each condition in each hemisphere. Any participant whose data
contained 30% or more data points that exceeded 2.5 standard deviations from the mean was
excluded from the final NIRS analyses (2 participants were removed in Experiment 1, none from
Experiments 2-4).
To control for drift within hemodynamic values across the entire experiment, the data
was baseline corrected by subtracting the one-second average from each trials entire time course,
thereby ensuring a stable baseline of comparison for each trial. Within-participant condition
means were then calculated for each participant by taking the average of each 20 ms time point
across all trials of each condition type for the 12 seconds following each stimulus presentation.
These within-participant means were then averaged across participants, creating a single 12second averaged global hemodynamic response for each hemisphere of each participant in each
condition in each experiment. A 12-second difference wave was then calculated by subtracting
the value of the Low Frequency/BOI/EE condition from the value of the High condition at each
20 ms timepoint.
As per Nathoo et al. (2018), within-participant Bayesian credible intervals were then
calculated at each time point across the entire 12-second difference wave. This allows, with just
a glance, an evaluation of the strength of the evidence across time, of a model that allows for the
hemodynamic difference at each time point to be different than zero.
Results
Behavioural.

93
Statistical Analysis. Response latencies were analyzed in R (R Core Team, 2013) using
the ttestBF function (Morey & Rouder, 2011). This function calculates a statistic (Bayes factor)
reflecting the likelihood of the experimental data being produced by either a null hypothesis
model wherein both conditions’ data come from the same normal distribution (i.e., no treatment
effect), or by a model reflecting a non-zero difference between conditions (i.e., a treatment
effect). Although ‘significance’ is not a metric used in Bayes factor analyses, conventional
benchmarks have been proposed: a Bayes factor of 3 is considered to be weak but statistically
significant evidence, and a Bayes factor of 10 is considered strong evidence (Aczel, et al., 2017).
Experiment 1. The mean within-participant response latencies for low-frequency words
and high-frequency words were respectively 814.33 ms (SE 32.92) and 786.40 ms (SE 28.88).
For the item-level analysis, the mean response latencies for low-frequency words and highfrequency words were respectively 819.02 ms (SE 11.39) and 788.24 ms (SE 7.36).
The Bayes factor was 6.51 for the within-participant analysis, and 4.86 for the item-level
analysis. These results should be considered statistically significant evidence that response
latencies are better predicted by a model that includes word frequency than one that does not,
and that high frequency words are responded to more quickly than are low frequency words.
There were 11 errors made in response to high frequency trials, 19 errors made in
response to low frequency trials, and 111 errors made in response to no-go trials. Because of the
low number of errors in Experiment 1, no analysis was conducted.
Experiment 2. In the within-participant analysis, the mean response latency for words
rated low in BOI was 1045.66 ms (SE 38.21), and 932.37 ms (SE 36.27) for words rated high in
BOI. In the item-level analysis, words rated low in BOI had a mean response latency of 1038.09

94
ms (SE 20.00), and a mean response latency of 934.40 ms (SE 11.55) for words rated high in
BOI.
The within subject analysis produced a Bayes factor of 1,586,393 and the item-level
analysis produced a Bayes factor of 937.31. Both analyses provide very convincing evidence that
SCT response time latencies using the decision criterion of ‘is the word easily imageable’ are
better predicted by a model that accounts for BOI than one that does not. More specifically, the
results show that words rated high in BOI are responded to more quickly than are words rated
low in BOI.
Accuracy data for the within-subjects analysis were analyzed with a two-tailed paired
sample t-test. There were 32 errors made in response to high BOI trials, 150 made in response to
low BOI trials, and 358 false alarms. The accuracies to high and low BOI trials were
significantly different (t(24) = 6.22, p < .01). The item-level accuracy analysis was conducted
with a two-tailed t-test with the number of correct responses to each high- and low-BOI word
compared. The accuracies for high- and low-BOI words were significantly different (t(34) =
3.94, p < .01)
Experiment 3. In the within-subject analysis, the mean response latency was 1255.00 ms
(SE 45.71) for abstract nouns rated low in EE and 1139.56 ms (SE 40.16) for abstract nouns
rated high in EE. In the item-level analysis, abstract nouns rated low in EE had a mean response
latency of 1227.56 (SE 17.75) and 1126.21 ms (SE 13.43) for abstract nouns rated high in EE.
The within-subjects analysis of response time difference produced a Bayes factor of
17,444.18, and the items-level analysis produced a Bayes factor of 949.76. Both analyses
produce very convincing evidence that SCT response latencies to abstract nouns are better
predicted by a model that accounts for EE than one that does not. More specifically, the data

95
show that abstract nouns rated high in EE are responded to more quickly than abstract nouns
rated low in EE.
Accuracy data for the within-subjects analysis were analyzed with a two-tailed paired
sample t-test. There were 64 errors made in response to high EE trials, 185 made in response to
low EE trials, and 224 made in response to no-go trials. The accuracies to high and low EE trials
were significantly different (t(24) = 8.22, p < .01). The item-level accuracy analysis was
conducted with a two-tailed t-test with the number of correct responses to each high- and low-EE
word compared. The accuracies for high- and low-EE words were significantly different (t(34) =
4.34, p < .01)
Experiment 4. In the within-subject analysis, the mean response latency for concrete
nouns rated low in EE was 977.72 ms (SE 21.36), and 1040.11 ms (SE 28.33) for concrete nouns
rated high in EE. In the item-level analysis, concrete nouns rated low in EE had a mean response
latency of 980.98 ms (SE 13.54), and 1040.55 ms (SE 25.83) for concrete nouns rated high in
EE.
The within-subject analysis of response time difference produced a Bayes factor of 226,
and the items-level analysis produced a Bayes factor of 4.16. The within-subjects analysis is
convincing evidence that participants’ SCT response latencies to concrete nouns are better
predicted by a model that accounts for EE than one that does not, while the item-level analysis
provides weaker (but still statistically significant) evidence that response latencies to concrete
nouns can be better predicted by a model that accounts for EE than one that does not. More
specifically, the data show that concrete nouns rated low in EE are responded to more quickly
than concrete nouns rated high in EE.

96
Accuracy data for the within-subjects analysis were analyzed with a two-tailed paired
sample t-test. There were 128 errors made in response to high EE trials, 31 in response to low EE
trials, and 175 made in response to no-go trials. The accuracies to high and low EE trials were
significantly different (t(24) = -5.73, p < .01). The item-level accuracy analysis was conducted
with a two-tailed t-test with the number of correct responses to each high- and low-EE word
compared. The accuracies for high- and low-EE words were significantly different (t(34) = -3.68,
p < .01)
Near-Infrared Spectroscopy.
Experiment 1. NIRS data for the 12 seconds following stimulus presentation are
presented for [O2Hb], [HHb], [THb], and the regional tissue oxygen saturation (Sat%) in the left
hemisphere in Figure 1, and for the right hemisphere in Figure 2. In the first four seconds after
stimulus presentation, the immediate processing of presented stimuli is evident in the left
hemisphere’s characteristic decrease in [O2Hb] and Sat% and increase in [Hb]. Left hemisphere
[THb] decreases during the first 2 seconds and then slowly returns to baseline during the intertrial interval. This pattern is nearly identical in the right hemisphere except for high frequency
words, which do not show a meaningful pattern of decreasing or increasing. This
interhemispheric difference suggests that letter strings are processed in the left hemisphere
automatically, reflecting efforts to ascertain the linguistic properties of the stimulus. In the right
hemisphere, however, the data reveal an inhibitory process in which no-go trials and low
frequency words require more processing to inhibit response execution, while high frequency
words are more easily recognized as real words, and therefore do not recruit the right
hemisphere’s inhibitory processes. Additionally, right hemispheric [THb] and [O2Hb] decrease

97
in the final four seconds of the intertrial interval, perhaps reflecting anticipatory inhibitory efforts
as participants, having just inhibited responding, prepare to do so again.
Unfortunately, analysis of hemodynamic responses failed to reveal reliable condition
differences. As described above, difference waves were calculated by subtracting the average
hemodynamic response for one condition from another for the 12 seconds following stimulus
presentation. These difference waves were then fit with Bayesian credible intervals. For a
significant condition difference to be inferred, these credible intervals needed to not include zero.
For example, in the left hemisphere, a model that includes effects of word frequency is
statistically superior to a model of the null hypothesis that does not include effects of word
frequency for approximately 100 ms beginning four seconds after stimulus presentation, with
high frequency words having greater [THb] (Figure 3). Similarly, in the left hemisphere, the data
show that at nine and nearly 12 seconds after stimulus presentation, for approximately 100 ms at
a time, a model that includes effects of word frequency is statistically superior to a model of the
null hypothesis that does not include effects of word frequency, with low-frequency word trials
having greater [THb]. Given the brief duration, small magnitude, and erratic timing of these
differences, these data are best interpreted as being the product of noise and not meaningful
condition differences. As a result, no other hemodynamic condition differences will be reported
here.
The data for Experiment 1 suggest that general patterns of lexical processing can be
captured by fNIRS, but that these data lack statistical power to make conclusions about condition
differences, if any exist.
Experiment 2. NIRS data for the 12 seconds following stimulus presentation are
presented for [O2Hb], [HHb], Sat%, and [THb] in the left hemisphere in Figure 4, and for the

98
right hemisphere in Figure 5. As in Experiment 1, in the first 2-4 seconds after stimulation,
evidence of stimulus processing is evident in the left hemisphere in the form of decreased
[O2Hb], Sat%, and [THb], accompanied by a slight increase in [HHb]. However, this pattern
does not hold for words rated low in BOI. [O2Hb] and [THb] increase shortly after stimulus
presentation of words rated low in BOI, though these differences are small and likely attributable
to noise, as discussed below. Sat% in the left hemisphere increases above baseline beginning 6
seconds after stimulus presentation for both no-go trials and for words rated low in BOI, while
Sat% decreases for words rated High in BOI during the same timeframe. This difference could
be indicative of additional semantic processing of High BOI words. In the right hemisphere,
[O2Hb] and Sat% increase above baseline approximately 4 seconds after stimulus presentation.
[HHb] in the right hemisphere decreases during this time following the presentation of a no-go
trial, unlike go trials where the levels stay near baseline. Collectively, these right hemisphere
differences may reflect participants’ efforts to inhibit their responses. However, [THb] in the
right hemisphere during this time increases for both no-go trials and for words rated high in BOI,
but not for words rated low in BOI. These differences challenge the interpretation of inhibitory
effects in the right hemisphere from these data.
As with Experiment 1, analyses of the difference waves did not provide meaningful or
reliable condition differences. For illustrative purposes, Figure 6 shows the difference wave
between low-BOI and no-go trials of [O2Hb] in the right hemisphere, including Bayesian
credible intervals. Although there are several brief periods in this difference wave that suggest
that a model that accounts for the effects of BOI is statistically superior to a model that does not,
these periods align with bouts of particularly noisy data, and therefore are most likely best
interpreted as false positives.

99
Experiment 3. NIRS data for the 12 seconds following stimulus presentation are
presented for left hemisphere [O2Hb], [HHb], Sat%, and [THb] in Figure 7, and for the right
hemisphere in Figure 8. Bilateral [THb] increases are evident in participant responses to abstract
nouns rated low in EE beginning approximately two seconds after stimulus presentation and
persisting for the entire measured interval. The same pattern is observed bilaterally with
increased [O2Hb] in participant responses to words rated low in EE during the same period,
though the effect does not begin until approximately four seconds after stimulus presentation in
the left hemisphere. In the left hemisphere, [O2Hb] increases in participant responses to words
rated high in EE shortly after stimulus presentation but return to baseline during the intertrial
interval. These data cannot be interpreted as increased effort to inhibit responses because
increased concentrations were not observed during no-go trials. The increased [THb] and [O2Hb]
in the right hemisphere in participant responses to low EE trials may therefore reflect additional
processing of atypical abstract noun exemplars.
Analyses of the difference waves did not reveal meaningful or reliable condition
differences. For illustrative purposes, Figure 9 shows the difference wave comparing [THb]
during high- and low-EE trials in the right hemisphere, with Bayesian credible intervals, where
visual inspection of the data suggests sizeable differences. Although there are two brief periods
in this difference wave that would statistically support the interpretations presented above, these
periods align with spikes in the data, and therefore are most likely false positives.
Experiment 4. NIRS data for the 12 seconds following stimulus presentation are
presented for [O2Hb], [HHb], Sat%, and [THb] in the left hemisphere in Figure 10, and for the
right hemisphere in Figure 11. No-go trials show increased [O2Hb] and decreased [HHb] in the
left hemisphere beginning shortly after stimulus presentation while [THb] increases for all three

100
conditions. In the right hemisphere, [O2Hb] increases for all three condition types, but does so
more rapidly for no-go and low EE trials. [HHb] decreases are most evident for high EE trials
starting four seconds after stimulus presentation and appears to increase for low EE trials
approximately seven seconds after stimulus presentation.
As with the previous experiments, analysis of the difference waves did not reveal
meaningful periods of hemodynamic differences. Figure 12 shows the difference wave of [O2Hb]
no-go trials subtracted from low EE trials as this effect looked to be significant in Figure 10.
Although there is a marked period extending from approximately five to seven seconds after
stimulus presentation where no-go trials appear to have significantly higher [O2Hb] than do lowEE trials, comparison to Figure 2 shows that low-EE trials during this window are near baseline
while high EE trials and no-go trials are above baseline. This finding is hard to reconcile with
previous data as no-go trial hemodynamics, interpreted as inhibition processes, would be
predicted to occur in the right hemisphere and should not require more effort in the left
hemisphere than do no-go trials. As a result, given the conclusions of the previous three
experiments, caution is urged in interpreting the data in Figure 12.
Discussion
The behavioural predictions of the four experiments detailed above were supported. In a
LDT (Experiment 1), high frequency words were responded to more quickly. In Experiments 2
and 3, nouns rated high in grounded semantic richness (BOI and EE, respectively) were
responded to more quickly and more accurately. These latter two findings were predicted
because it is believed that greater levels of bodily and emotion experience with a word’s referent
leads to greater levels of grounded semantic richness in the conceptual representation of that
word, thereby leading to more efficient processing and faster recognition of that word in the

101
SCTs used in these two experiments. In Experiment 4, concrete nouns rated high in EE were
responded to more slowly and less accurately, thereby supporting the hypothesis that information
obtained through emotion experiences are more indicative of abstract concepts and thereby
increase the difficulty for participants to identify a concrete noun when EE ratings are high.
Behavioural Results.
The response latency data for all three SCTs support the hypotheses that one’s
sensorimotor and emotion experiences are integral to one’s conceptual knowledge
representation. Additionally, these experiences produced the predicted response latency
differences. Experiment 1 replicated the well-established word frequency effect, with high
frequency words being responded to more quickly and accurately than low frequency words.
From a grounded cognition perspective, Experiments 2 and 3 examined the behavioural evidence
that word stimuli containing more task-congruent grounded semantic richness (BOI and EE,
respectively) would elicit faster response latencies. Experiment 4 tested the fundamental nature
of emotion experiences in knowledge representation by recording response latency data when
grounded semantic richness (EE) was incongruent with task demands.
Consistent with prior research, Experiments 2 and 3 demonstrated that participants
responded more quickly to words rated high in grounded semantic richness (BOI and EE),
whereas in Experiment 4, participants responded more quickly to words low in grounded
semantic richness (e.g., congruency: Pexman, et al., 2003; Taler et al., 2013; incongruency:
Newcombe et al., 2012). This differential influence of grounded semantic richness suggests that
semantic richness obtained through bodily and emotional experiences with the environment is
activated automatically when reading visually presented words. Newcombe et al. (2012)
observed lower categorization accuracy rates for concrete nouns rated high in EE but did not

102
observe a statistically significant response latency effect, as was the case in Experiment 4. The
response latency and accuracy data of Experiment 4, taken together with Newcombe et al.’s
(2012) observations, suggest that high levels of EE interfere with the recognition and processing
of concrete nouns in SCTs. That is, high levels of EE are diagnostic of abstract concepts and, as
a result, concrete nouns rated high in EE are more difficult to process as being concrete. These
data support the hypothesis that conceptual knowledge is obtained through bodily and emotion
sensations and behaviours, and that the activation of that knowledge automatically involves the
reactivation of those very experiences regardless of whether doing so is contextually
advantageous.
NIRS Results.
Although visual inspection of the NIRS waveforms for the 12 seconds following stimulus
presentation suggested the presence of some event-related activity that supported the
hemodynamic predictions made, a Bayesian analysis of difference waves did not reveal
convincing evidence of condition differences. Although it is possible that these data suggest that
the anterior PFC is not involved in lexical or semantic decisions, the literature review presented
above makes this interpretation improbable. As a result, it is necessary to discuss possible
reasons for why the results occurred across the four experiments presented above.
The first possibility regarding the lack of significant PFC activity detected by NIRS in
these experiments is that, perhaps, the difficulty manipulation was not substantial enough to
elicit measurably different hemodynamic responses in the anterior PFC on single trials. That is,
despite the initial hypothesis that, for example, an abstract concept rated low in EE would be
more difficult to identify as an abstract noun, and despite the many observations summarized
above that PFC hemodynamics can reflect task difficulty, it is possible that, in the experiments

103
presented here, manipulations of grounded semantic richness alone were insufficient to
differentially activate the anterior PFC. This possibility could be investigated by using fMRI or a
NIRS device with a greater number of channels. With a greater number of channels, it is possible
that a more posterior effect would be evident because, as task difficulty increases, the associated
brain activity becomes increasingly anterior (Causse et al., 2017). Thus, it is possible that the
difficulty manipulation in these experiments was not reflected in the PFC. With fMRI, it is
possible that the activity associated with condition differences in the experiments presented here
would be detectable at a depth that NIRS cannot readily detect.
Another possibility is that the current experiments lacked statistical power to properly
capture small cerebral hemodynamic changes. The majority of cognitive experiments employing
NIRS technology use blocked designs. That is, even in go/no-go experiments, the data are
collected over the course of several trials. For example, Inoue et al. (2012) employed a go/no-go
design where “go trial” data were operationalized as the average hemodynamic amplitude across
2-minutes of go trials and “no go” data were operationalized as the average hemodynamic
amplitude of 2-minutes of mixed trials, half of which were go trials and half of which were no-go
trials. Inoue et al. (2012) used a 16-channel NIRS device covering a 14x3.5 cm area of the
forehead, using wavelengths 730 and 850 nm, sampled at 3 Hz. Hemodynamic differences in
experiments such as these would therefore reflect periods of time where response inhibition
occurred, rather than reflect the inhibition process itself, as was attempted in the experiments
reported here. Ang et al. (2014) reported ‘single trial’ classification of task difficulty using 32channel NIRS technology using wavelengths 760 and 830 nm, sampled at 1.81 Hz. However, an
assessment of their research design shows that a trial in this experiment lasted several minutes. It
can be concluded, then, that most researchers measure cerebral hemodynamics over the course of

104
several minutes of sustained cognitive effort so as to capture recordable effects in the brain. In
conducting the literature review for the experiments presented here, no researcher explicitly
stated this justification for their research design, though blocked designs certainly appear to be
the norm. In designing the experiments presented here, the decision to employ a single-trial
design with long intertrial intervals was driven by the fact that hemodynamic responses are slow
to return to baseline and that the sampling rate of NIRS is sufficiently high as to fully capture the
shape of the hemodynamic response (Pinti et al., 2018). Earlier pilot data related to this research
suggested that single-trial PFC condition differences were obtainable in SCTs, with the
hemodynamic response primarily located in the right hemisphere. Future replications of these
experiments would do well to contrast the hemodynamic response from blocked and single-trial
designs so as to investigate the full range of cerebral hemodynamics elicited by lexical and
semantic decisions.
Another possible factor that may have contributed to the absence of clear hemodynamic
data in these experiments is the equipment itself. More specifically, the OxiplexTS is a twochannel device with flexible optodes that allow it to be affixed to the scalp for cerebral
monitoring. The optodes on the OxiplexTS are approximately two inches wide and three inches
long, with fixed emitter-detector distances. The first implication of employing the OxiplexTS is
that only a very limited region of the cortex can be monitored and, if the event-related
hemodynamic activity is not occurring in precisely that location, it will be difficult to detect. The
pilot data mentioned above suggested that the anterior PFC was a viable location for the
observation of hemodynamic responses elicited by semantic categorization. However, those pilot
data were collected with NIRS alone; EEG was not collected concurrently. As a result, it is
possible that, in accommodating the EEG caps, the optodes were placed in a more anterior

105
location than they were during pilot data collection so as to allow for accurate electrode
placement. With a different NIRS device designed to accommodate EEG electrodes, or that
allowed for customized emitter-detector placement, data collection would have been possible at
more lateral locations. Using sizeable optodes with the ability to adjust emitter-detector distances
is normally not a concern when monitoring cerebral activity. However, it does limit the positions
that are available when simultaneously recording EEG data.
As detailed above, both the dorsolateral and ventrolateral PFC have been implicated in
cognitive processes beyond integrative contextual processing, which would have expanded the
breadth of conclusions that could be drawn from the experiments presented here. Additionally, it
would have been ideal to be able to monitor multiple brain regions. A device with additional
channels would have allowed for recording from frontotemporal cortex along with the anterior
and lateral PFC. Doing so would have increased the number of possible PFC functions that could
be inferred from the data and would increase both spatial acuity and statistical power. Finally, for
Experiment 2 in particular, it would have been optimal to measure from the motor,
somatosensory, and parietal cortexes given the grounded cognition hypothesis that conceptual
knowledge involves reactivation of the neural regions involved in forming that conceptual
knowledge. As such, NIRS should be able to detect activity in these cortexes during the
processing of referents that one has bodily experience with (see Hargreaves et al., 2012 for fMRI
data supporting this hypothesis). Unfortunately, pilot data associated with this research suggested
that the OptiplexTS was unable to obtain interpretable data from this region. This is unsurprising
given that the brain-scalp distance is comparatively greater here than at the PFC.
In Experiments 3 and 4, there are two possible reasons why grounded semantic richness
effects were not observed in the PFC. The first is spatial in nature. Specifically, it is possible that

106
the effect of emotion information in the simulation and processing of conceptual knowledge
happens predominantly in the ventromedial PFC (e.g., Ozawa et al. 2019). Future extensions of
this research would do well to record from as many locations as possible so as to test this
hypothesis. The second reason concerns how grounded semantic richness attributable to emotion
experiences was operationalized in these experiments. Specifically, as discussed in Chapter 2,
emotion information was operationalized as EE ratings. That is, positive and negative valence
were collapsed into a single dimension that captures the ease by which a concept elicits an
emotion experience. Previous research has shown that EE accounts for response latency
variability above and beyond that of positive and negative valence (e.g., Moffat, et al., 2015;
Newcombe et al., 2012; Siakaluk, et al., 2014; Siakaluk, et al., 2016). However, as detailed
above, hemispheric differences have been observed during the processing of positive and
negative stimuli, with negative stimuli primarily activating the right PFC and positive stimuli
primarily activating the left PFC (Balconi et al., 2015; Herrmann et al., 2003; Kida & Shinohara,
2013a; Minagawa-Kawai et al., 2009). As a result, it is possible that the PFC was involved in the
automatic processing of emotion information during the abstract and concrete noun SCTs in
Experiments 3 and 4, but due to positively and negatively valenced words being averaged
together during data analysis in these experiments, the associated PFC hemodynamic activity
may not have been captured here. That is, if the right hemisphere was active and the left
hemisphere was inactive during the processing of negatively valenced words, and this pattern
reversed for the processing of positively valenced words, then the act of averaging the signals
across positive and negative word trials could produce an artificial null result. Unfortunately,
testing this particular hypothesis will be difficult.

107
Future replications of Experiments 3 and 4 will have to be careful when trying to
ascertain whether there are hemispheric differences when participants are processing concepts
with high positive and negative valence. One approach to testing this would be to replicate the
current design with word stimuli that were exclusively positively or negatively valanced.
However, participants may quickly learn to respond to a word’s valence rather than its
abstractness or concreteness. To prevent this possibility, the abstract and concrete nouns would
need to be balanced on their valence, and separate experiments may need to be done regarding
positive and negative valence as a result. Though possible, this would be a challenging task.
Another option would be to replicate the current design but with at least twice the number of
trials to increase statistical power sufficiently so as to allow for the analysis of valence effects in
addition to the effects of EE. A consequence of this design would be that the experiment would
require nearly three hours to complete without greatly reducing the inter-trial interval which
would have the unfortunate effect of obscuring the time course data for individual trials. These
concerns would need to be taken into consideration prior to future investigations of the
hemodynamic response to the processing of emotion information in the PFC during SCTs.
A final consideration is that the data were too noisy to detect real effects in these
experiments. One reason for this could be due to the fact that NIRS signals were recorded
concurrently with EEG data for an average of 80 minutes per session. Once participants were
seated comfortably and an EEG cap fitted to their head, the optodes were placed through and
under the EEG cap. To keep them in place, cohesive tape was wrapped around the participants’
heads several times. However, in analyzing the EEG data for Chapter 4, it was clear that data
integrity suffered over the duration of the experiments. That is, the electrodes lifted over the
course of the experiments. Due to the constant pull of the elastic straps that comprise the EEG

108
caps, upward pressure created by the participants’ hair, or participant movement throughout the
lengthy session lifting can occur, causing the EEG cap or cohesive tape to be repositioned.
Regardless of the reason, evidence of data degradation over time in Chapter 4 highlights the
possibility that the optodes may have changed position as well during these experiments, thereby
adding variance to the NIRS signal. Future replications would do well to employ NIRS devices
that are more appropriately designed to integrate with EEG or to maintain contact on the scalp
for extended periods of time.
Conclusions.
Behaviourally, the data support the conclusion that grounded semantic information is an
automatic component of conceptual processing, even when it is not conducive to successful task
completion. It is interesting and important that the dimensions of grounded semantic richness
manipulated in these experiments were not explicitly relevant to task completion because the
results therefore support the ideas that, when categorizing nouns’ concreteness or abstractness, a
person need not (a) be explicitly aware of their bodily interactions with imageable objects nor (b)
be aware of their emotional experiences with concepts.
The addition of NIRS monitoring during LDT and SCT performance was intended to
capture differential prefrontal activity due to the increased effort required to process or inhibit
specific trials as a function of their task congruency. These data stood to serve as an indirect
physiological correlate of conceptual processing or of inhibitory effort. The NIRS data did not
reveal meaningful condition differences in the PFC during task completion. Participant response
latencies illustrated that grounded semantic dimensions of conceptual knowledge were activated
automatically when completing an SCT trial. Unfortunately, however, these effects were not
evident in the PFC data obtained via NIRS. Future studies would do well to continue this line of

109
research while incorporating a greater number of optodes and measuring from multiple brain
regions. By doing so, the clarity of the obtained data may be enhanced by being able to filter
NIRS data using nearby optode locations as a covariate, and thereby increasing the power of
NIRS to lead to observations of the semantic effects evident in the behavioural data reported
here.

110

Figure 1. Clockwise, from top-left: NIRS time course data for left PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 1. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation.

111

Figure 2. Clockwise, from top-left: NIRS time course data for right PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 1. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation.

112

Figure 3. Plotted left hemispheric [THb] difference between high- and low-frequency words for
the first 14 seconds following stimulus presentation In Experiment 1. Error bars are Bayesian
99% credible intervals.

113

Figure 4. Clockwise, from top-left: NIRS time course data for left PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 2. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation.

114

Figure 5. Clockwise, from top-left: NIRS time course data for right PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 2. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation.

115

Figure 6. Plotted right hemispheric [O2Hb] differences between low-frequency words and no-go
trials for the first 14 seconds following stimulus presentation In Experiment 2. Error bars are
Bayesian 99% credible intervals.

116

Figure 7. Clockwise, from top-left: NIRS time course data for left PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 3. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation.

117

Figure 8. Clockwise, from top-left: NIRS time course data for right PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 3. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation.

118

Figure 9. Plotted right hemispheric [THb] differences between high- and low-EE words for the
first 14 seconds following stimulus presentation in Experiment 3. Error bars are Bayesian 99%
credible intervals.

119

Figure 10. Clockwise, from top-left: NIRS time course data for left PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 4. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation.

120

Figure 11. Clockwise, from top-left: NIRS time course data for right PFC [O2Hb] (μM), [HHb]
(μM), [THb] (μM), and Sat% (%) in Experiment 4. Data were cleaned up with a 200 ms average
moving window for ease of display and interpretation.

121

Figure 12. Plotted left hemispheric [O2Hb] differences between high-EE words and no-go trials
for the first 14 seconds following stimulus presentation in Experiment 4. Error bars are Bayesian
99% credible intervals.

122
References
Aczel, B., Palfi, B., & Szaszi, B. (2017). Estimating the evidential value of significant results in
psychological science. PLoS ONE, 12(8), 1-8.
Amiri, M., Poulliot, P., Bonnéry, C., Leclerc, P-O., Desjardins, M., Lesage, F., & Joanette, Y.
(2014). An exploration of the effect of hemodynamic changes due to normal aging on the
fNIRS response to semantic processing of words. Frontiers in Neurology, 5(249), 1-11
Ang, K. K., Guan, C., Lee, K., Lee, J. Q. , Nioka, S., & Chance, B (2010). A brain-computer
interface for mental arithmetic task from single-trial near-infrared spectroscopy brain
signals, Proceedings of the 20th International Conference on Pattern Recognition., pp.
3764-3767.
Ang, K. K., Yu, J., & Guan, C. (2014). Single-trial classification of NIRS data from prefrontal
cortex during working memory tasks. 36th Annual International Conference of the IEEE
Engineering in Medicine and Biology Society, 2008-2011.
Aron, A.R., Fletcher, P.C., Bullmore, E.T., Sahakian, B.J., Robbins, T.W. (2003). Stop-signal
inhibition disrupted by damage to right inferior frontal gyrus in humans. Nature
Neuroscience, 6(2)., 115-116.
Aron, A. R., Robbins, T. W., & Poldrack, R. A. (2004). Inhibition and the right interior frontal
cortex. Trends in Cognitive Sciences, 8(4), 170-177.
Ayaz, H., Shewokis, P. A., Bunce, S., Uzzetoglu, K., Willems, B., & Onaral, B. (2012). Optical
brain monitoring for operator training and mental workload assessment. NeuroImage,
59(1), 36-47.

123
Balconi, M., Bortolotti, A., & Gonzaga, L. (2011). Emotional face recognition, EMG response,
and medial prefrontal activity in empathic behavior. Neuroscience Research, 71(3), 251259.
Balconi, M., Grippa, E., & Vanutelli, M.E. (2015). What hemodynamic (fNIRS),
electrophysiological (EEG) and autonomic integrated measures can tell us about
emotional processing. Brain and Cognition, 95, 67-76.
Barret, L. F., & Simmons, W. K. (2015). Interoceptive predictions in the brain. Nature Reviews
Neuroscience, 16, 419-429.
Bechara, A., Damasio, H. & Damasio, A. R. (2000). Emotion, decision making and the
orbitofrontal cortex. Cerebral Cortex 10, 295-307.
Bendall, R. C. A., Eachus, P., & Thompson, C. (2016). A brief review of research using nearinfrared spectroscopy to measure activation of the prefrontal cortex during emotional
processing: The importance of experimental design. Frontiers in Human Neuroscience,
10(529), 1-7.
Berglund-Barraza, A., Tian, F., Basak, C., & Evans, J. L. (2019). Word frequency is associated
with cognitive effort during verbal working memory: A functional near infrared
spectroscopy (fNIRS) study. Frontiers in Human Neuroscience, 13, 433.
Berman, K. F., Ostrem, J. L., Randolph, C., Gold, J., Goldberg, T. E., Coppola, R., Carson, R. E.,
Herscovitch, P., & Weinberger, D. R. (1995). Physiological activation of a cortical network
during performance of the Wisconsin Card Sorting Test: A positron emission tomography
study. Neuropsychologia, 33, 1027-1046.

124
Boecker, M., Buecheler, M. M., Schroeter, M. L., & Gauggel, S. (2007). Prefrontal brain
activation during stop-signal response inhibition: An event-related functional near-infrared
spectroscopy study. Behavioural Brain Research, 176, 259-266.
Bonetti, L. V., Hassan, S. A., Lau, S-T., Melo, L. T., Tanaka, T., Patterson, K. K., & Reid, W.
D.(2018). Oxyhemoglobin changes in the prefrontal cortex in response to cognitive tasks:
A systematic review. International Journal of Neuroscience, 129(2), 195-203.
Buchsbaum, B. R., Greer, S., Chang, W-L., & Berman, K. F. (2005). Meta-analysis of
neuroimaging studies of the Wisconsin card-sorting task and component processes. Human
Brain Mapping, 25(1), 35-45.
Casey, B.J., Trainor, R.J., Orendi, J.L., Schubert, A. B., Nystrom, L. E., Giedd, J.N., … &
Rapoport, J. L., (1997). A developmental functional MRI study of prefrontal activation
during performance of a go-no-go task., Journal of Cognitive Neuroscience, 9, 835-847.
Causse, M., Chua, Z., Peyasakhovich, V., Del Campo, N., & Matton, N. (2017). Mental
workload and neural efficiency quantified in the prefrontal cortex using fNIRS. Scientific
Reports, 7(5222).
Chen, H., Vaid, J., Boas, D., & Bortfeld, H. (2010). Examining the phonological
neighborhood density effect using near infrared spectroscopy. Human Brain
Mapping, 32(9), 1363-1370.
Christoff, K. & Gabrieli, J. D. E. (2000). The frontpolar cortex and human cognition: Evidence
for a rostrocaudal hierarchical organization within the human prefrontal cortex.
Psychobiology, 28(2), 168-186.

125
Cohen, J. D., Barch, D.M., Carter, C.S., & Servan-Schreiber, D. (1999). Schizophrenic deficits in
the processing of context: Converging evidence from three theoretically motivated
cognitive tasks. Journal of Abnormal Psychology, 108, 120-133.
Damasio, A.R. (1996). The somatic marker hypothesis and the possible functions of the
prefrontal cortex. Philosophical Transactions of the Royal Society B: Biological Sciences,
351, 1413-1420.
Doi, H., Nishitani, S., & Shinohara, K. (2013). NIRS as a tool for assaying emotional function in
the prefrontal cortex. Frontiers in Human Neuroscience, 7(770), 1-6.
Drevets, W. C. (2001) Neuroimaging and neuropathological studies of depression: Implications
for the cognitive-emotional features of mood disorders. Current Opinion in Neurobiology,
11, 240–249.
Duncan, J. (2001). An adaptive coding model of neural function in prefrontal cortex. Nature
Reviews Neuroscience, 2(11), 820-829.
Duncan, J. & Owen, A. M. (2000) Common regions of the human frontal lobe recruited by
diverse cognitive demands. Trends in Neuroscience, 23, 475–483.
Fishburn, F. A., Norr, M. E., Medvedev, A. V., & Vaidya, C. J. (2014). Sensitivity of
fNIRS to cognitive state and load. Frontiers in Human Neuroscience, 8(76),
1-11.
Fletcher, P. C., Happe, F., Frith, U., Baker, S. C., Dolan, R. J., Frackowiak., R. S. J., & Frith, C.
D. (1995). Other minds in the brain: A functional imaging study of ‘theory of mind’ in
story comprehension. Cognition 57, 109–128.
Fu, L., Xiang, D., Xiao, J., Yao, L., Wang, Y., Xiao, L., Wang, H., Wang, G., & Lieu, Z. (2018).
Reduced prefrontal activation during the Tower of London and verbal fluency task in

126
patients with bipolar depression: A multi-channel NIRS study. Frontiers in Psychiatry, 9,
214.
Garavan, H., Hester, R., Murphy, K. Fassbender, C., & Kelly, C. (2006). Individual differences
in the neuroanatomy of inhibitory control, Brain Research, 1105(1), 130-142.
Glotzbach, E., Mühlberger, A., Gschwendtner, K.,Fallgatter, A. J., Pauli, P., & Herrmann, M.
J.(2011). Prefrontal brain activation during emotional processing: A functional near
infrared spectroscopy study (fNIRS). The Open Neuroimaging Journal, 5, 33.
Goel, V., Gold, B., Kapur, S., & Houle, S. (1997). The seat of reason? An imaging study of
deductive and inductive reasoning. NeuroReport, 8, 1305-1310.
Goel, V., Grafman, J., Sadato, N. & Hallett, M. (1995). Modeling other minds. Neuroreport 6,
1741–1746.
Goldman-Rakic, P. S. (1987). Motor control function of the prefrontal cortex. Ciba Foundation
Symposium, 132, 187-200.
Hargreaves, I. S., Leonard, G.A., Pexman, P. M., Pittman, D. J., Siakaluk, P. D., & Goodyear, B.
G. (2012). The neural correlates of the body-object interaction effect in semantic
processing. Frontiers in Human Neuroscience, 6(22), 1-8.
Hermann, M.J., Ehlis, A.-C., & Fallgatter, A.J. (2003). Prefrontal activation through task
requirements of emotional induction measured with NIRS. Biological Psychology, 64, 255263.
Hetu, S., Gregoire, M.,, Saimpont, A., Coll, M. P., Eugene, F., Michon, P. E., & Jackson, P. L.
(2013). The neural netowrk of motor imagery: An ALE meta-analysis. Neuroscience &
Biobehavioral Reviews, 37(5), 930-949.

127
Hofmann, M. J., Herrmann, M. J., Dan, I., Obrig, H., Conrad, M., Kuchinke, L., Jacobs, A. M.
& Fallgatter, A. J. (2008). Differential activation of frontal and parietal regions during
visual word recognition: An optical topography study. NeuroImage, 40(3), 1340-1349.
Hornak, J., Bramham, J., Rolls, E. T., Morris, R. G., O’Doherty, J., Bullock, P. R., & Polkey, C.
E. (2003). Changes in emotion after circumscribed surgical lesions of the orbitofrontal
and cingulate cortices. Brain, 126, 1691–1712.
Horovitz, S. G. & Gore, J. C. (2004). Simultaneous event-related potential and nearinfrared spectroscopic studies of semantic processing. Human Brain Mapping,
22(2), 110-115.
Hoshi, Y., Huang, J., Iguchi, Y., Naya, M., Okamoto, T., & Ono, S. (2011). Recognition of
human emotions from cerebral blood flow changes in the frontal region: A study with
event-related near-infrared spectroscopy. Journal of Neuroimaging, 21, 94–101.
Inoue, Y., Sakihara, K., Gunji, A., Ozawa, H., Kimiya, S., Shinoda, H., Kaga, M., & Inagaki, M.
(2012). Reduced prefrontal hemodynamic response in children with ADHD during the
Go/NoGo task: A NIRS study. NeuroReport, 23(2), 55-60.
Iversen, S. D., & Mishkin, M. (1970). Perseverative interference in monkey following selective
lesions of the inferior prefrontal convexity. Experimental Brain Research, 11, 376–386.
Jourdan Moser S., Cutini S., Weber P., & Schroeter, M. L. (2009). Right prefrontal brain
activation due to Stroop interference is altered in attention-deficit hyperactivity disorder:
a functional near-infrared spectroscopy study. Psychiatry Research, 173, 190–195.
Kahlaoui, K., Di Sante, G., Barbeau, K., Maheux, M., Lesage, F., Ska, B., & Joanette, Y. (2012).
Contribution of NIRS to the study of prefrontal cortex for verbal fluency in aging. Brain
and Language, 121(2), 164-173.

128
Kawashima, R., Satoh, K. Itoh, H., Ono, S., Furumoto, S., Gåtor, R., … & Fukuda, H. (1996).
Functional anatomy of go/no-go discrimination and response selection - a PET study in
man., Brain Research, 728, 79-89.
Kida, T., Nishitanni, S., Tanaka, M., Takamura, T., Sugawara, M., & Shinohara, K. (2014). I
love my grandkid! An NIRS study of grand maternal love in Japan. Brain Research,
1154, 131-137.
Kida, T. & Shinohara, K. (2013a). Gentle touch activates the anterior prefrontal cortex: An NIRS
study. Neuroscience Research, 76, (1-2), 76-82.
Kida, T. & Shinohara, K. (2013b). Gentle touch activates the prefrontal cortex in infancy: An
NIRS study. Neuroscience Letters, 541, 63-66.
Koechlin, E., Corrado, G., Pietrini, P. & Grafman, J. (2000). Dissociating the role of the medial
and lateral anterior prefrontal cortex in human planning. Proceedings of the National
Academy of Science USA, 97, 7651-7656.
Koseki, S., Noda, T., Yokoyama, S., Kunisato, Y., Ito, D., Suyama, H., Matsuda, T., Sugimura,
Y., Ishihara, N., Shimizu, Y., Nakazawa, K., Yoshida, S., Arima, K., & Suzuki, S.
(2013). The relationship between positive and negative automatic thought and activity in
the prefrontal and temporal cortices: A multi-channel near infrared spectroscopy (NIRS)
study. Journal of Affective Disorders, 151(1), 352-359.
Kroger, J. K., Sabb, F. W., Fales, C. L., Bookheimer, S. Y., Cohen, M. S. & Holyoak, K. J.
(2002). Recruitment of anterior dorsolateral prefrontal cortex in human reasoning: A
parametric study of relational complexity. Cerebral Cortex 12(5), 477–485.
Li, C., Gong, H., Zeng, S., & Luo., Q. (2005). Verbal working memory load affects prefrontal
cortices activation: Evidence from a functional NIRS study in humans. In Complex

129
Dynamics and Fluctuations in Biomedical Photonics II (Vol. 5696, pp. 33-40).
International Society for Optics and Photonics.
Liu, X., Sun, G., Szhang, X., Xu, B., Shen, C., Shi, L., Ma, X., Ren, X., Feng, K., & Liu, P.
(2014). Relationship between the prefrontal function and the severity of the emotional
symptoms during a verbal fluency task in patients with major depressive disorder: A multichannel NIRS study. Progress in Neuro-Psychopharmacology and Biological Psychiatry,
54(3), 114-121.
Masataka, N., Perlovsky, L., Hiraki, K. (2015). Near-infrared spectroscopy (NIRS) in functional
research of prefrontal cortex. Frontiers Media SA.
Matsuda, G. & Hiraki, K. (2006). Sustained decrease in oxygenated hemoglobin during video
games in the dorsal prefrontal cortex: A NIRS study of children. NeuroImage, 29(3), 706711.
McKendrick, R., Ayaz, H., Olstead, R., & Parasuraman, R. (2014). Enhancing dual-task
perfromance with verbal and spatial working memory training: Continuous monitoring of
cerebral hemodynamics with NIRS. NeuroImage, 85(3), 1014-1026.
Miller, L. A. (1992). Impulsivity, risk-taking, and the ability to synthesize fragmented
information after frontal lobectomy. Neuropsychologia, 30, 69-79.
Miller, E. K. & Cohen, J. D. (2001). An integrative theory of prefrontal cortex function. Annual
Review of Neuroscience, 24, 167-202.
Minagawa-Kawai, Y., Matsuoka, S., Dan, I., Naoi, N., Nakamura, K., & Kojima, S. (2009).
Prefrontal activation associated with social attachment: Facial-emotion recognition in
mothers and infants. Cerebral Cortex, 19, 284-292.

130
Moffat, M., Siakaluk, P. D., Sidhu, D. M., & Pexman, P. M. (2015) Situated conceptualization
and semantic processing: Effects of emotional experience and context availability in
semantic categorization and naming tasks. Psychonomic Bulletin & Review, 22(2), 408419.
Moghimi, S., Kushki, A., Guerguerian, A. M., & Chau, T. (2012a). Characterizing emotional
response to music in the prefrontal cortex using near infrared spectroscopy. Neuroscience
Letters, 525, 7-11.
Moghimi, S., Kushki, A., Power, S., Guerguerian, A. M., & Chau, T. (2012b). Automatic
detection of a prefrontal cortical response to emotionally rated music using multi-channel
near-infrared spectroscopy. Journal of Neural Engineering, 9(2), 026022.
Morey, R.D. & Rouder, J.N. (2011). Bayes factor approaches for testing interval null hypotheses.
Scientific Reports, 3, 19-28.
Morinaga, K., Akiyoshi, J., Matsushita, H., Ichioka, S., Tanaka, Y., Tsuru, J., & Hanada, H.
(2007). Anticipatory anxiety-induced changes in human lateral prefrontal cortex activity.
Biological Psychology, 74(1), 34-38.
Moriya, M. & Sakatani, K. (2017). Effects of motor imagery on cognitive function and prefrontal
cortex activity in normal adults evaluated by NIRS. Advances in Experimental Medical
Biology, 977, 227-231.
Nathoo, F. S., Kilshaw, R. E., & Masson, M. E. (2018). A better (Bayesian) interval estimate for
within-subject designs. Journal of Mathematical Psychology, 89¸ 1-9.
Negoro H, Sawada M, Iida J, Ota T, Tanaka S, Kishimoto T. (2010). Prefrontal dysfunction in
attention-deficit/hyperactivity disorder as measured by near-infrared spectroscopy. Child
Psychiatry & Human Development, 41(2), 193–203.

131
Newcombe, P. I., Campbell, C., Siakaluk, P. D., & Pexman, P. M. (2012). Effects of
emotional and sensorimotor knowledge in semantic processing of concrete and
abstract nouns. Frontiers in Human Neuroscience, 6(275), 1-15.
Nguyen, T., Condy, E., & Gandjbakhche, A. (2020). Functional connectivity in the prefrontal
cortex during a simple versus an emotional Go/No-Go task: an fNIRS study. In
Microscopy Histopathology and Analytics (pp. JTu3A-38). Optical Society of America.
Nguyen, T., Condy, E. E., Park, Soongho, Friedman, B. H., & Gandjbakhche, A. (2021).
Comparison of functional connectivity in the prefrontal cortex during a simple and an
emotional Go/No-Go task in female versus male groups: an fNIRS study. Brain Science,
11(17), 909-920.
Nishimura, Y., Tanii, H., Hara, N., Inoue, K., Kaiya, H., Nishida, A., Okada, M., & Okazaki, Y.
(2009). Relationship between the prefrontal function during a cognitive task and the
severity of the symptoms in patients with panic disorder: A multi-channel NIRS study.
Psychiatry Research: Neuroimaging, 172(2), 168-172.
Noda, T., Yoshida, S., Matsuda, T., Okamoto, N., Sakamoto, K., Koseki, S.., Numachi, Y.,
Matsumshima, E., Kunugi, H., & Higuchi, T. (2012). Frontal and right temporal
activations correlate negatively with depression severity during verbal fluency task: A
multi-channel near-infrared spectroscopy study. Journal of Psychiatric Research, 46, 1-8.
Noguchi, Y., Takeuchi, T., & Sakai, K. L. Lateralized activation in the inferior frontal
cortex during syntactic processing: Event-related optical topography study.
Human Brian Mapping, 17(2), 89-99.
Nolde, S. F., Johnson, M. K. & Raye, C. L. (1998). The role of prefrontal cortex during tests of
episodic memory. Trends in Cognitive Science, 2, 399–406.

132
O’Doherty, J., Kringelbach, M. L., Rolls, E. T., Hornak, J., & Andrews, C. (2001). Abstract
reward and punishment representations in the human orbitofrontal cortex. Nature
Neuroscience, 4, 95–102.
Okamoto, M., Wada, Y., Yamaguchi, Y., Kyutoku, Y., Clowney, L., Singh, A. K., & Dan, I.
(2011). Process-specific prefrontal contributions to episodic encoding and retrieval of tests:
A functional NIRS study. NeuroImage, 54(2), 1578-1588.
Ozawa, S. & Hiraki, K. (2017). Distraction decreases prefrontal oxygenation: A NIRS study.
Brain and Cognition, 113, 155-163.
Ozawa, S., Matsuda, G., & Haraki, K. (2014). Negative emotion modulates prefrontal cortex
activity during a working memory task: A NIRS study. Frontiers in Human Neuroscience,
7(770), 159-169.
Pardo, J. V., Pardo, P. J., Janer, K. W. & Raichle, M. E. (1990). The anterior cingulate cortex
mediates processing selection in the Stroop attentional conflict paradigm. Proceedings of
the National Academy of Science USA 87, 256–259.
Parker, A. & Gaffan, D. (1998). Memory after frontal/temporal disconnection in monkeys:
Conditional and non-conditional tasks, unilateral and bilateral frontal lesions.
Neuropsychologia, 36, 259-271.
Pexman, P. M. (2012). Meaning based influences on visual word recognition. In J. S. Adelman
(Ed.), Visual word recognition. Vol. 2 (pp. 24-43). New York: Psychology Press.
Pexman, P. M., Lupker, S. J., & Hino, Y. (2002). The impact of feedback semantics in visual
word recognition: Number-of-features effects in lexical decision and naming tasks.
Psychonomic Bulletin & Review, 9(3), 542-549.

133
Picton, T.W., Stuss, D.T., Alexander, M.P., Shallice, T., Binns, M.A., & Gillingham, S. (2006).
Effects of focal frontal lesions on response inhibition. Cerebral Cortex, 17(4), 826-838.
Pinti, P., Tachtsidis, I., Hamilton, A., Hirsch, J., Aichelburg, C., Gilbert, S., & Burgess, P. W.
(2018). The present and future use of functional near-infrared spectroscopy (fNIRS) for
cognitive neuroscience. Annals of the New York Academy of Sciences, 1464(1), 5-29.
Poppa, T. & Bechara, A. (2018). The somatic marker hypothesis: Revisiting the role of the ‘body
loop’ in decision-making. Current Opinion in Behavioral Science, 19, 61-66.
Rao, S. M., Bobholz, J. A., Hammeke, T. A., Rosen, A. C., Woodley, S. J., Cunningham, J. M.,
Cox, R. W., Stein, E. A., & Binder, J. R. (1997). Functional MRI evidence for subcortical
participation in conceptual reasoning skills. NeruoReport, 8, 1987-1993.
Ramnani, N. & Owen, A. M. (2004). Anterior prefrontal cortex: Insights into function from
anatomy and neuroimaging. Nature Reviews Neuroscience, 5, 184-194.
Ranganath, C., Johnson, M. K., & D’Esposito, M. (2000). Left anterior prefrontal activation
increases with demands to recall specific perceptual information. Journal of Neuroscience,
20, RC108
R Core Team (2013). R: A language and environment for statistical computing. R Foundation for
Statistical Computing. Vienna, Austria.
Rolls, E. T., (1996). The orbitofrontal cortex. Philosophical Transactions of the Royal Society B,
351, 1433-1444.
Rolls, E.T. (2004). The functions of the orbiofrontal cortex. Brain and Cognition, 55, 11-29.
Rolls, E. T., O’Doherty, J., Kringelbach, M. L., Francis, S., Bowtell, R., & McGlone, F. (2003).
Representations of pleasant and painful touch in the human orbitofrontal and cingulate
cortices. Cerebral Cortex, 13, 308–317.

134
Rossi, S., Jürgenson, I.B., Hanulíková, A., Telkemeyer, S., Wartenburger, I., Obrig, H. (2011).
Implicit processing of phonotactic cues: Evidence from electrophysiological and vascular
responses. Journal of Cognitive Neuroscience, 23(7), 1752-1764.
Rossi, S., Telkemeyer, S., Wartenburger, I., & Obrig, H. (2012). Shedding light on words and
sentences: Near-infrared spectroscopy in language research. Brain & Language, 121,
152-163.
Schneider, S., Rapp, A. M., Haeußinger, F. B., Ernst, L. H., Hamm, F., Fallgatter, A. J., & Ehlis,
A-C. (2014). Beyond the N400: Complementary access to early neural correlates of novel
metaphor comprehension using combined electrophysiological and hemodynamic
measurements. Cortex, 53, 45-59.
Sears, C.R., Siakaluk, P.D., Chow, V.C., & Buchanan, L. (2008). Is there an effect of print
exposure on the word frequency effect and the neighborhood size effect? Journal of
Psycholinguistic Research, 37, 269-291.
Sela, I., Horowitz–Kraus, T., Izzetoglu, M., Shewokis, P., Izzetoglu, K., Onaral, B., &
Breznitz, Z. (2011). Brain activity of young and adult Hebrew speakers during
lexical decision task: FNIR application to language. In D. Schmorrow & C.
Fidopiastis (Eds.), Foundations of augmented cognition: Directing the future of
adaptive systems (pp. 231–239). Berlin, Germany: Springer.
Sela, I., Izzetoglu, M., Izzetoglu, K., & Onaral, B. (2014). A function near-infrared
spectroscopy study of lexical decision task supports the dual route model and the
phonological deficit theory of dyslexia. Journal of Learning Disabilities, 47(3),
279-288.

135
Shallice, T. & Burgess, P. (1998). The domain of supervisory processes and the temporal
organization of behaviour. in The Prefrontal Cortex: Executive and Cognitive Functions
(eds Roberts, A. C., Robbins, T. W. & Weiskrantz, L.) 22-35. Oxford University Press,
Oxford, UK.
Siakaluk, P. D., Knol, N., & Pexman, P. M. (2014). Effects of emotional experience for abstract
words in the Stroop task. Cognitive Science, 38(8), 1698-1717.
Siakaluk, P.D. Newcombe, P.I., Duffels, B., Li, E., Sidhu, D.M., Yap, M.J., & Pexman, P.M.
(2016). Effects of emotional experience in lexical decision. Frontiers in Psychology:
Cognitive Science, 7(1157).
Siddiqui, S. V., Chatterjee, U., Kumar, D., Siddiqui, A., & Goyal, N. (2008). Neuropsychology
of prefrontal cortex. Indian Journal of Psychiatry, 50(3), 202-208.
Sumitani, S., Tanaka, T., Tayoshi, S., Ota, K., Kameoka, N., Morimune, M., Shibuya-Tayoshi,
S., Kinouchi, S., Ueno, S., & Ohmori, T. (2015). Hemodynamic changes in the prefrontal
cortex during mental works as measured by multi channel near-infrared spectroscopy
(NIRS). Journal of Medical Investigation, 52, 302-303.
Tai, K. & Chau, T. (2009). Single-trial classification of NIRS signals during emotional induction
tasks: Towards a corporeal machine interface. Journal of Neuroengineering and
Rehabilitation, 6, 39.
Taler, V., Kousaie, S., & López Zunini, R. (2013). ERP measures of semantic richness: The case
of multiple senses. Frontiers in Human Neuroscience, 7(5), 1-6.
Tanida, M., Katsuyama, M., & Sakatani, K. (2007). Relation between mental stress-induced
prefrontal cortex activity and skin conditions: A near-infrared spectroscopy study. Brain
Research, 1184, 210-216.

136
Tanida, M., Sakatani, K., Takano, R., & Tagai, K. (2004). Relation between asymmetry of
prefrontal cortex activities and the autonomic nervous system during a mental arithmetic
task: Near infrared spectroscopy study, Neuroscience Letters, 369(1), 69-74.
Tulving, E. (1983). Elements of Episodic Memory, Clarendon, Oxford.
Verfaellie, M., Heilmann, K.M. (1987). Response preparation and response inhibition after
lesions of the medial frontal lobe. Archives of Neurology, 44(12), 1262-1271.
Van Dillen, L. F., Heslenfeld, D. J., & Koole, S. L. (2009). Tuning down the emotional brain: An
fMRI study of the effects of cognitive load on the processing of affective images.
Neuroimage, 45(4), 1212–1219
Wagner, A. D., Desmond, J. E., Glover, G. H. & Gabrieli, J. D. E. (1998). Prefrontal cortex and
recognition memory. Functional- MRI evidence for context-dependent retrieval processes.
Brain 121, 1985–2002.
Wilson, C. R. E., Gaffan, D., Browning, P. G. F., & Baxter, M.G. (2010). Functional localization
within the prefrontal cortex: Missing the forest for the trees? Trends in Neuroscience,
33(12), 533-540.
Wood, J. N. & Grafman, J. (2003). Human prefrontal cortex: Processing and representational
perspectives. Nature Reviews Neuroscience, 4(2), 139-147.
Xiao, T. Ziao, Z, Ke, X., Hong, S., Yang, H., Su., Y. chu, K., Xiao, X., Shen, J. & Liu, Y.
(2012). Response inhibition impairment in high functioning autism and attention deficit
hyperactivity disorder: Evidence form near-infrared spectroscopy data. PLoS One, 7(10),
e46569.

137
Yang, H., Zhou, Z., Liu, Y., Ruan, Z., Gong, H., Luo, Q., & Lu, Z. (2007). Gender difference in
hemodynamic responses of prefrontal area to emotional stress by near-infrared
spectroscopy. Behavioural Brain Research, 178, 172-176.

138
Table 1
Stimuli for Experiment 1
High Frequency
ARMY
AWAY
BEGIN
BOOK
BORN
BROWN
CLASS
CLAY
CLEAN
CLEAR
DATA
DATE
DEAL
DOWN
EDGE
EYES
FACT
FAIR
FILM
FIRM
FIXED
FOUND
FULL
GIVEN
HEART
KNOW
LADY
LARGE
LATER
LIGHT
LOWER
MEAN
MEANT
MONTH
NIGHT
NORTH
NOTE
ORDER
PASS
PEACE
POWER
RACE

Low Frequency
ATOM
BACON
BAKER
BITE
BLANK
BLOND
BOMB
BORED
BOWL
CALF
CLUE
CORK
CRIB
CUTS
DASH
DEBT
DEER
FIRED
FISH
FOAM
FUEL
GLAD
GLOOM
GRADE
GRAPE
HUMOUR
INCH
JUMP
KNEE
KNOT
LEMON
LINK
LION
LODGE
LOYAL
MILD
MOUSE
NURSE
PATH
PEACH
PINK
PITY

Non-Words
ANID
BIVER
BLAVE
BOKY
BOUS
BRAB
BULE
BULP
CACK
CEED
CINK
CLANE
CLEAK
CLEE
CLIDE
COOR
CUDDY
DOWED
FANK
FRACE
FREG
FRINT
FUMP
GADE
GAKED
GARED
GICE
GINUS
GLODE
GOUTH
GREE
GRIKE
GUNG
HEARY
HEEM
HEVER
HOAL
HORRY
JAME
JELT
JITE
JOOT

PANI
PEAF
PIGE
PIPEG
PLOSS
POAM
POOG
PORSE
POWN
PROKE
PRONT
PUNG
RALT
RASTY
RECK
RISH
RODGE
RONCH
ROUTH
SCOCK
SEAX
SENG
SERE
SHEEL
SHILE
SHYME
SILP
SLAMP
SLANK
SLARE
SLEED
SLET
SLORE
SLOVE
SLOY
SLUNT
SMAT
SNAN
SOAN
SOIN
SOLLY
SOUN

139
REACH
RISE
RIVER
ROAD
ROLE
SHARE
SIGN
SIZE
SORT
SOUTH
STAGE
STAY
STOP
STORE
SURE
TALK
TRUE
UNIT
VOTE
WALK
WATCH
WHOLE
WORTH
WRITE

PLOT
PUMP
PUSH
REACT
RIDE
SCARE
SHADE
SHIPS
SKIN
SPADE
SPAN
SPARE
SPRAY
STACK
STEEL
STEM
SUITE
TIRED
TOWER
TREAT
TWIN
WARN
WEAVE
ZERO

JUNA
KEED
KILA
LENY
LINDER
LOLD
LOMP
LUSK
MALED
MARER
METT
MOATH
MODEO
MOEL
MOOST
MOUL
NAZE
NEEP
NINCH
NING
NOAST
NOIL
NOVE
PALI

SPOW
SPUT
STORN
TANNY
TELON
TIDDY
TIVID
TOACH
TOAR
TRAKE
TREAK
TREST
TRIAR
TROP
TROSS
TROW
VACED
VACK
VASH
VIRE
VUFF
WAVEG
WEAG
YINK

Note: High- and low-frequency words are balanced on orthographic neighbourhood, number of
letters, number of morphemes, number of syllables, and phonological neighbourhood size, and
the non-words are matched with all words on length and do not differ significantly in the number
of orthographic neighbours.

140
Table 2
Stimuli for Experiment 2
High BOI
Low BOI
BELT
ASH
BOOT
BACK
CAGE
BAND
CARD
BAY
CHILD
BIRCH
CORD
BRAIN
COUCH
BRASS
DRILL
CASE
FEET
COAST
FRIEND
FROST
GATE
GAME
HAT
GANG
HOOK
HEART
LAKE
JAIL
LOCK
KING
MAIL
KNIGHT
MAN
LANE
MAT
LINT
MATE
LOOT
NECK
LUNG
PRIEST
PLACE
PURSE
PRINCE
ROOM
PUMP
SEAT
ROOF
SILK
SHOP
STAIR
SLIT
STRING
SONG
SUIT
SPOT
SWORD
STRIPE
THING
SUN
TOOL
TRAIL
TOY
TAR
TUBE
WAR
VEST
WELL
WHEEL
WITCH

Low Imageability Fillers
WISP
EVE
NODE
NEED
GALE
ROLE
PART
WAY
SPAN
SPRIG
CLANK
FRAUD
FLUX
FIEND
GUISE
WISH
REALM
THING
GAIT
BLOKE
STANCE
MODE
ZEAL
LODE
LYE
THYME
HALF
BRIG
HOAX
DOSE
RIFT
SCRUFF
PEAT
VERB
EBB
BANE
NAPE
TERM
DEED
MYTH
CHANCE
SHAME
STILE
CLINK
FLAW
ZING
CUSP
FUNK
WRIT
LUCK
CHOICE
BULK
STEPPE
HYPE
MOOD
QUIP
TREND
GROUCH
FUGUE
PIP
PLOY
TOT
WIT
HUNCH
TACT
SLAG

Note: High- and low-BOI words are balanced on length, number of features, familiarity ratings,
concreteness, imageability, print frequency, contextual diversity, and Levenshtein orthographic
and phonological distances.

141
Table 3
Stimuli for Experiments 3 & 4
High EE Abstract
Low EE Abstract
ADVERSITY
ADVANCE
AFTERLIFE
AMOUNT
ATROCITY
APPROACH
ATTITUDE
APTITUDE
BELIEF
ARRAY
CONCERN
BLANDNESS
CONTROL
CAPACITY
COURAGE
COMPOUND
CRISIS
CUSTOM
DEMOCRACY
DECREASE
DESIRE
DEDUCTION
DISCIPLINE
DISPOSITION
DISTRESS
EFFECT
EMANCIPATION FACILITY
ENVY
FEATURE
EQUITY
FIGMENT
FAILURE
FORETHOUGHT
FEELING
HEREDITY
FUTURE
HOUR
JUDGEMENT
HYPOTHESIS
JUSTICE
ILLUSION
MIRACLE
IMAGE
NEGLECT
INCLINE
OFFENCE
INSTANT
OPPORTUNITY
MANNER
PATIENCE
METHOD
RELIEF
NONSENSE
REVENGE
NOTHING
SOBRIETY
PARDON
SOUL
RATING
SUPPORT
REVIEW
SUPPRESSION
TENDENCY
TRUTH
TRIFLE
VALUE
UPKEEP
WEAKNESS
VENTURE

High EE Concrete
ARTIST
BABY
BEDROOM
BULLET
CHICKENPOX
COLLEGE
COMRADE
CONCERT
COUSIN
DAUGHTER
DAYLIGHT
DESTROYER
DIAMOND
DISEASE
HURRICANE
HUSBAND
JEWEL
LAUGHTER
LEADER
LION
MISTRESS
MONSTER
MOUNTAIN
NAVY
PARTNER
PENICILLIN
PICTURE
POLICE
SCHOLAR
SERVANT
SHERIFF
STUDENT
SUMMER
TUBERCULOSIS
WINTER

Low EE Concrete
ALUMINIUM
AUDITORIUM
BARREL
BASEMENT
BASIN
BLACKSMITH
BUTTER
CARPET
CHINCHILLA
COLLAR
COLUMN
CORNER
COTTON
DISHWASHER
DOORWAY
ELBOW
ENGINE
ENVELOPE
FABRIC
FINGER
FRECKLES
HELMET
HIGHWAY
HOTEL
JACKET
KITCHEN
MERCHANT
NUMBER
PITCHER
PROJECTOR
QUARTER
RECTANGLE
SANDWICH
SCREWDRIVER
STATION

Note: High- and low-EE words are balanced within concrete and abstract word categories on log
HAL frequency, age of acquisition, orthographic neighbourhood, number of letters, number of
syllables, number of morphemes, concreteness, and imageability.

142
CHAPTER 4:
Electrophysiological Evidence of Automatically Activated Grounded Semantic Knowledge
Experimental findings show differential event-related potential (ERP) patterns (either
quantitatively in a deflection’s magnitude at a single electrode, or qualitatively with regard to the
topology of brain activity across multiple electrodes) depending on task demands, illustrating its
usefulness in identifying and testing the neurological markers of human cognition (e.g., Barber,
et al., 2013; Kutas & Hillyard, 1980). As mentioned in Chapter 2, electroencephalography (EEG)
is a technology with excellent temporal resolution which makes it ideal for observing fast
cognitive processes such as language processing (Hauk, 2016). More specifically, responding to
word stimuli requires rapid identification of the orthography of the presented letters and
activation of associated semantic information before making a response, which is typically done
in less than one second (Hargreaves & Pexman, 2014; Laszlo & Federmeier, 2014). Hauk (2016)
outlined the necessary cognitive processes that precede responding as follows: visual processing,
lexical and semantic long-term memory retrieval, working memory processing, and response
selection. Eye-tracking data suggest that words are recognized as unique entities within 100-150
ms after presentation (Laszlo & Federmeier, 2014). With so many cognitive processes happening
in such a short period of time, investigating the time course of grounded cognition in semantics
requires a technology that is capable of monitoring these processes as they happen in real time.
Although there is some debate regarding how to discuss ERP components as discrete entities
(Luck, 2014), there are characteristic patterns of ERP deflections in semantic processing
experiments that can be investigated.
Previous research has attempted to illustrate the time course of semantic processing,
though the results have yet to produce a coherent picture of what dimensions are processed in

143
what order and under which conditions. When single words are used as experimental stimuli, the
earliest reliable effects have been observed within 100-150 ms after stimulus presentation. Using
a signal-to-respond methodology, Hargreaves and Pexman (2014) observed behavioural
facilitation for the number of senses evoked by a word after 100 ms of word processing. Using
ERP methods, Kuchinke and Mueller (2019) observed differences in the P1 and N1 (peaking
100-130 ms after stimulus presentation) for emotional words. Similarly, higher P100 magnitudes
have been observed in parieto-occipital regions for emotion-laden words, suggesting a very early
stage in word processing that is sensitive to emotional information, particularly when that
information is negative (Wang, et al., 2019; Zhang et al., 2014). Using regression analyses,
Laszlo and Federmeier (2014) showed superior statistical prediction of the first 50 ms of ERP
data in visual word recognition with models that include the influence of lexicality and of the
first 120 ms with models that include the influence of semantics. Furthermore, the topography of
brain responses has been observed to differ between word categories as early as 100 ms after
stimulus presentation (Mosely et al., 2013). Collectively, these findings suggest that the brain’s
processing of word stimuli is well underway by 100-150 milliseconds after stimulus
presentation, with some evidence of semantic sensitivity at that time.
By 200 ms after stimulus presentation, the influence of semantic information on ERP
patterns is reliably observed (Clarke, 2020). A word’s imageability has been found to modulate
ERP components prior to 250 ms after stimulus presentation (Hauk, 2016). The P200 peaks
between 150 and 275 ms after stimulus presentation, and greater amplitudes have been observed
for words with greater levels of emotion information, especially with negatively valanced words
(González, et al., 2014; Kanske & Kotz, 2007). P200 amplitude is also greater during poor
memory performances (i.e., recalling fewer items), and when processing sentences with

144
semantically constrained endings (e.g., Federmeier & Kutas, 2002; Freunberger, et al., 2007).
The P200 has also been shown to be sensitive to noun/verb distinctions and to be enhanced for
emotion-laden words (Yudes, et al, 2016). Additionally, the P200 has been shown to be
enhanced when processing words with more associated concepts (Stuellein et al., 2016). The
P200 reflects both the amount of experience with a word’s referent and the amount of effort
required to process stimulus ambiguity (Wang et al., 2019).
Many of the most well-known semantically-associated ERP components peak between
approximately 300 ms and 500 ms after stimulus presentation, the most common of which is the
N400 component. The N400 was first observed to be amplified in response to the last word in a
sentence if that word was unexpected (e.g., “he takes his coffee with cream and bananas”; Kutas
& Hillyard, 1980; Van Petten, et al., 1999). The N400 has since been found to be amplified by
unexpected faces and pictures (Lau, et al., 2008). Greater N400 amplitudes have been observed
to individual words possessing many semantic features or associated concepts (Vegara-Martínez
et al., 2017). N400 amplitudes are typically decreased in response to familiar or contextually
supported stimuli, and even in response to repeated non-word or pseudoword stimuli (Lau et al.,
2008; Laszlo & Federmeier, 2014, Stuellein et al., 2016; Vegara-Martínezet et al., 2017). These
data suggest that the N400 does not exclusively reflect semantic processes. Rather, it could be
interpreted as reflecting either contextual integration between stimuli and task demands or the
effort to retrieve contextually relevant memories (Lau et al., 2008).
The N400 may not be a single component, making it difficult to interpret. Rather, the
N400 may reflect many processes that occur at the same time (an interpretation that is supported
by difficulty to localize a single N400 generator in the brain; Lau et al., 2008). N400 amplitude is
inconsistently influenced by semantic richness. For example, the N400 is attenuated when

145
processing emotion words (e.g., Kanske & Kotz, 2007; Kotz & Paulmann, 2007; Wang et al.,
2019), yet is larger in response to action than to non-action words (Yudes et al., 2016).
Conversely, more extreme N400 amplitudes have been observed by 370 ms for emotion words in
LDTs, but as early as 170 ms in semantic contexts (Schact & Sommer, 2009). The differential
ERP time course in different tasks using emotion information as the manipulated semantic
dimension suggests feed-forward processing in which an individual is primed to process or
attend to particular types of information depending on context.
Beyond the N400 window, semantic effects have been observed in what is called the late
positive component (LPC) and the P600 (Coulson et al., 1998; Gouvea et al., 2010). The LPC
has been found to be largely sensitive to emotion information, leading to an interpretation that it
reflects additional sustained evaluative processing (Kurchinke & Mueller, 2019; Wang et al.,
2019). The P600 has been referred to as the ‘syntactic positive shift’ and is thought to be elicited
by word stimuli that require reanalysis (Coulson et al., 1998; Gouvea, et al., 2010). ERP
components this late after stimulus presentation are likely reflective of post-lexical processing
that does not contribute directly to the contextual identification of a presented stimulus. Some
stimuli, being more ambiguous due to having lower levels of semantic richness than others,
require extended processing even after lexical processes related to a word’s meaning are
completed. All other information about a presented word stimulus has been activated and
processed previously in the first ~500 ms after stimulus processing.
Some ERP research has been done on semantic richness, specifically. For example, in a
SCT using word pairs as stimuli, and decision criterion of, “Is the second word an example of the
first?”, smaller N400 deflections were observed for typical exemplars (i.e., “yes” responses), and
when the word pairs were semantically related, regardless of category membership (Heinze et al.,

146
1998). This result demonstrates that N400 amplitudes are associated with the activation of a
word’s meaning.
The N400 has also been tied to embodied semantic processing and language
comprehension (e.g., Blomberg et al., 2020; Xue et al., 2015). In an LDT, Blomberg et al. (2020)
measured ERP data for four different noun categories: specific concrete (e.g., hammer), general
concrete (e.g., tool), abstract (e.g., reason), and emotional abstract (e.g., love). N400 amplitudes
were most negative for specific concrete nouns, followed by general concrete nouns, abstract
nouns, and finally by emotional abstract nouns. These data show that the N400 is modulated by
many variables including the specificity and amount of emotion information associated with a
concept’s representation. Xue et al. (2015) measured ERPs to the last word of a visually
presented sentence using stimuli that varied in body-object interaction ratings (BOI) in a
sentence acceptability task. The target words were either rated high or low in BOI and were
presented after either a semantically rich sentence (i.e., possessing at least one associated
sensorimotor property) or a semantically poor sentence (i.e., possessing only a description of the
word). Regardless of sensorimotor context, high BOI words were responded to more quickly and
sentences ending in high BOI words were judged as more acceptable than those sentences ending
in low BOI words. High BOI words produced smaller N400s than did low BOI words at central
and parietal locations, particularly when high BOI words were presented in semantically poor
contexts (e.g., “you brush pieces of a baked crumb”) versus semantically rich contexts (e.g., “you
brush the small sticky crumb”). In an object recognition experiment, less pronounced N400
deflections were observed in the anterior temporal lobe when the target object was primed by the
presentation of a word or picture depicting an object that could be manipulated in the same way
as the target object (Kiefer et al., 2011). This finding suggests that N400 amplitude does, in fact,

147
reflect semantic processing rather than lexical processing. In sum, higher levels of semantic
richness interacted with context and produced ERPs that reflected the amount of effort required
to interpret the words contextually, such that N400 amplitude was greater for words with higher
levels of semantic richness, yet comparatively lower when semantically rich words were
contextually integrated more easily.
Given the findings summarized above, ERP data may allow researchers to observe when
lexical and semantic information is processed. Such an approach is capable of demonstrating
distinctly different patterns of neural activity produced by the manipulation of grounded
semantic richness variables. As such, regardless of where in the brain a particular ERP
component originates, if the manipulation of grounded semantic variables (e.g., BOI and EE)
produces ERP differences (e.g., magnitude, scalp distribution, and/or timing) conclusions may be
made regarding the relative automaticity and contribution of grounded cognition dimensions to
semantic processing. Amodal theories of cognition, by comparison, struggle to adequately
explain such observations, particularly if the effects were evident very early in the course of
processing or in tasks where grounded semantic information is not explicitly required for
successful task completion. By analyzing the time course of ERPs after single word presentation,
it is possible to elucidate the intrinsic and automatic nature of grounded semantic information in
conceptual processing.
In the research presented below, several behavioural and neurological predictions
regarding the effects of grounded semantics in three SCTs were tested. EEG was employed in
three experiments to investigate whether there are differential patterns of response latencies and
neurological activity during SCTs wherein grounded semantic information is not explicitly
relevant for task completion.

148
In Experiments 2 and 3, high levels of the dimensions of grounded semantic richness and
emotional experience (BOI and EE) were congruent with the decision criteria, “Is this word
easily imageable?” and, “Is this an abstract noun?” respectively. As such, higher levels of
grounded semantic richness are predicted to be facilitatory and thus produce shorter response
latencies and attenuated ERP amplitudes due to the congruence between the nature of the
semantic information and the decision criteria. In Experiment 4, however, high levels of the
dimension of grounded semantic richness (EE) were incongruent with the decision criterion, “Is
this a concrete noun?” As such, high levels of grounded semantic richness are predicted to be
inhibitory and to produce longer response latencies and greater ERP magnitudes due to the
incongruence between the decision criterion and the nature of the semantic information (EE).
Hypotheses
Experiment 2 tested the grounded cognition hypothesis that semantically rich words
(operationalized here as words with referents associated with greater amounts of bodily
experiences and interactions) are processed more efficiently than less semantically rich words
(operationalized here as those with referents associated with comparatively fewer bodily
experiences and interactions). This experiment employed a modified version of the go/no-go
SCT used by Hargreaves et al. (2012), with the decision criterion, “Is this word easily
imageable?” All stimuli requiring a button press (i.e., the “go” trials) were highly imageable, and
half of the highly imageable words were rated high in BOI, whereas the other half were rated low
in BOI. It was predicted that, compared to words rated low in BOI, words rated high in BOI
would be responded to more quickly, more accurately, and would produce discernibly different
ERPs. These differences were predicted due to the relative ease of processing and the increased
number of sensory and motor brain regions thought to be responsible for the processing of words

149
with greater amounts of contextually congruent grounded motor semantic richness. Specifically,
the P200 component was predicted to be larger for words rated low in BOI due to their lower
levels of grounded motor semantic richness, which would lead to their being more difficult to
process due to their contextual ambiguity as imageable nouns. Similarly, the N400 was predicted
to have lower amplitudes for words rated high in BOI due to their contextual relevance (words
that one has interacted with frequently are also likely to be seen frequently, thus facilitating an
‘imageable’ response). The later components (P600/LPC) were predicted to have higher
amplitudes for words rated low in BOI due to the increased probability that these stimuli warrant
more processing as atypical exemplars of imageable nouns. In sum, an imageable concept with
low BOI ratings will be more difficult to identify as imageable and thus require additional
processing in the later stages.
Experiment 3 tested the grounded cognition hypothesis that abstract nouns with referents
with greater levels of associated emotion experiences are processed more efficiently than abstract
nouns with lower levels of associated emotion experiences. Specifically, abstract concepts are
hypothesized to be understood more quickly and accurately when they are associated with more
emotion experiences. Experiment 3 used a SCT with the same decision criterion as Newcombe et
al. (2012), with the addition of EEG observations to investigate the neural activity correlated
with processing varying levels of emotion information while identifying abstract nouns. The
decision criterion for the SCT was, “Is this an abstract noun?” As such, participants were to
respond to abstract nouns and to make no response to concrete nouns. Half of the abstract nouns
were rated high in EE and thus thought to be more ‘typical’ exemplars of abstract nouns. It was
predicted that words rated high in EE (greater grounded semantic richness obtained through
emotion experiences with the words’ referents) would be responded to more quickly than those

150
rated low in EE, while also producing discernibly different ERPs reflecting the congruency
between of emotion information and the decision criterion. The other half of the abstract nouns,
rated low in EE, were hypothesized to be more difficult to process as abstract nouns due to their
comparative lack of diagnostic EE information. Similar to Experiment 2, it was predicted that the
P200 would be smaller for words rated high in EE because they are typical exemplars of abstract
nouns, and thus comparatively easier to process. The N400 was also predicted to be smaller for
words rated high in EE due to comparative ease to integrate these stimuli into the context of the
decision criterion due to congruence between EE and task demands, and the P600/LPC would be
larger for words rated low in EE because these words are atypical exemplars of abstract nouns.
Experiment 4 tested the hypothesis that the activation of grounded information is an
automatic part of conceptual knowledge, regardless of its task relevance. Specifically, by
employing the same stimuli as Experiment 3 but with the SCT decision criterion, “Is this a
concrete noun?” participants would be responding to concrete nouns with high and low EE.
Concrete nouns have, by their nature, physical referents that are easily perceived and many may
be relatively easily physically interacted with, but typically have fewer associated emotion
experiences (see Table 1 in Newcombe et al., 2012 to compare EE ratings of concrete and
abstract nouns, noting that concrete nouns have significantly lower EE ratings). As such, it was
hypothesized that concrete nouns rated high in EE would be atypical concrete nouns, because the
additional grounded semantic richness obtained through emotion experiences would be more
diagnostic of abstractness than of concreteness. It is predicted that response latencies to concrete
nouns rated high in EE will be slower compared to more typical concrete nouns rated low in EE.
This hypothesis is congruent with previous research (e.g., Newcombe et al, 2012) that observed
lower accuracy rates for concrete nouns rated high in EE and is important for the following

151
reasons. If supported, the hypothesis that concrete nouns rated high in EE would be atypical
would suggest that grounded semantic richness need not always be facilitatory to successful task
completion. Moreover, if participants respond more slowly to concrete nouns rated high in EE, it
would suggest that grounded semantic richness is activated automatically when reading a word,
regardless of whether the specific semantic information is congruent (and thus advantageous) to
current task demands. Observing slower response latencies to concrete nouns rated high in EE
would contribute to addressing the question of how conceptual knowledge is processed in the
brain. If the grounded PSS framework is correct, then this processing of conceptual knowledge
should entail an automatic reactivation of the sensory, motor, and emotion experiences that are
commonly associated with that concept’s real-world referents. As a result, it is predicted that the
P200 for concrete nouns rated high in EE will be larger due to the additional effort needed to
process atypical concrete nouns. The N400 is predicted to be larger for concrete nouns rated high
in EE due to the contextual incongruence between task demands and the diagnostic nature of the
semantic information being activated. Finally, the P600 is predicted to be larger for concrete
nouns rated high in EE due to the reanalysis of atypical concrete noun exemplars.
Summary of Hypotheses. In sum, participant response latencies were predicted to be
shorter and accuracy rates higher when responding to word stimuli that were rated high in taskcongruent grounded semantic richness (Experiments 2 and 3). Conversely, participant response
latencies were predicted to be longer and accuracy rates lower when responding to word stimuli
that were rated high in task-incongruent grounded semantic richness (Experiment 4). In
Experiments 2 and 3, lower P200, N400, and P600 amplitudes were predicted for word trials
with high grounded semantic richness because of the comparative ease of processing due to the
grounded semantic richness’s task congruence. In Experiment 4, however, larger P200, N400,

152
and P600 amplitudes were predicted for word trials with high grounded semantic richness
because, in this experiment, the grounded semantic richness (EE) was not congruent with task
demands.
Method
Participants. Three separate groups of 25 University of Northern British Columbia
undergraduate students enrolled in psychology classes participated in each experiment (i.e., no
individual participated in more than one experiment). There were 19, 18, and 19 female
participants in Experiments 2, 3, and 4, respectively. Participants indicated that they were native
English speakers and had normal or corrected-to-normal vision. Participants received bonus
course credit for their participation.
Apparatus.
Behavioural. Behavioural data were collected with Superlab (version 5.0.3). Stimuli were
presented in white 24-point capitalized Times New Roman font centred on a 21.5” Dell P2212H
monitor on a black background. Participants were seated in front of a table so that they could
reach the keyboard, see the monitor, and remain comfortable for the duration of the experiment.
EEG. Electrical brain activity was recorded at 250 Hz with a 32-channel EGI HydroCel
Geodesic sensor net with silver-silver chloride electrodes (Magstim EGI, Eden Prairie, MN).
Two cap sizes were available (54-56 cm and 56-58 cm). Participants whose heads did not fit
either cap were given course credit but did not participate further in the experiment. EEG data
were captured with EGI’s Net Station Acquisition software (version 5.2.0.1).
Semantic Categorization Task. The SCTs for all three experiments are fully detailed in
Chapter 2.
Analyses.

153
Data Preprocessing. Data preprocessing occurred in the following manner. First, any
participant with an error rate exceeding 30% was removed from the data sets (one participant
from Experiment 2, two participants from Experiment 3, and no participants from Experiment 4).
Second, any word that was responded to incorrectly in more than 30% of the instances of its
presentation across participants was removed from all analyses (eight items from Experiment 2:
one high-BOI word and seven low-BOI words, 6.1% of the data; five items from Experiment 3:
five low-EE words, 3.5% of the data; and eight items from Experiment 4: seven high-EE words
and one low-EE word, 5.7% of the data). Incorrect responses were then removed from all further
behavioural or EEG analyses (392 trials from Experiment 2 (12.4% of the data); 364 trials from
Experiment 3 (11.3% of the data); and 332 trials from Experiment 4 (10.2% of the data).
Response latency outliers were identified and removed from all analyses if they were faster than
250 ms (no trials from any experiment were removed in this way) or slower than 2000 ms (22
trials from Experiment 2: < 0.01% of the data; 71 trials from Experiment 3: 0.02% of the data;
and 26 trials in Experiment 4: <0.01% of the data). Then, for each participant, response latencies
greater than 2.5 standard deviations from the cell mean of each condition were removed from all
analyses (17 trials from Experiment 2: <0.01% of the data; nine trials from Experiment 3: <.01%
of the data; and 18 trials from Experiment 4: <0.01% of the data. From the remaining data,
average response latencies for high- and low-BOI stimuli (Experiment 2), and high- and low-EE
stimuli (Experiments 3 and 4) were calculated for each participant and then averaged across
participants. Separately, average response latencies were calculated for each single word for
item-level analyses. Averages were tested using Bayes Factors (BayesFactor; R, version 3.5.3).
EEG. EEG data were analyzed with the EEGLAB toolbox for MATLAB (Delorme &
Makeig, 2004). First, the data were digitally band-pass filtered between 0.2-50 Hz. Next, bad

154
channels were identified by visual inspection and then interpolated using the spherical
interpolation method. Then time course data were separated into single-trial epochs, baseline
corrected to the mean of the 100 ms prior to stimulus presentation, for 800 ms after stimulus
presentation. Trials were then flagged if they contained artifacts (identified as deflections that
exceed a voltage threshold of 100 µV). Participants with more than 50% of trials removed this
way were omitted from EEG analyses (11 participants were removed from Experiment 2, six
from Experiment 3, and nine from Experiment 4).
Within-subject condition means (high- and low-BOI, Experiment 2; high- and low-EE,
Experiments 3 and 4) were calculated for each participant by taking the average of each 4 ms
time point across all trials of each condition type for the 800 ms following each stimulus
presentation. Then these within-subject means were averaged across participants, creating a
single 800 ms averaged ERP for each condition. Next, an 800 ms difference wave was calculated
by subtracting the value of the low condition from the value of the high condition, for each
experiment, at each 4 ms time point.
As per Nathoo, Kilshaw, & Masson (2018), within-subject 99% Bayesian credible
intervals were then calculated for each 4 ms time point across the entire 800 ms difference wave.
This procedure evaluates the strength of the evidence, across time, for a model that allows for the
cerebral activity difference between conditions at each 4 ms time point to be different from zero.
At any point in time, if the credible interval range does not include zero, then it can be assumed
that a model that includes the independent variable (in the present studies, either BOI or EE) is
better than a model that does not include the independent variable. The timing and direction of
the observed difference can then be interpreted in the context of the original ERP components.
Results

155
Behavioural.
Statistical Analysis. Comparisons of response latency differences were analyzed using the
ttestBF function from the BayesFactor package in R (Morey & Rouder, 2011; R Core Team,
2013). This function produces a statistic (Bayes factor) representing the strength of a model
where the two groups’ means are assumed not to be zero compared to a model representing a
null hypothesis that both groups’ data come from the same normal population. Decision-making
criteria for ‘significance’ in Bayes factor analyses are often set between 3 (deemed ‘anecdotal
evidence’) and 10 (deemed ‘strong evidence’; Aczel et al., 2017).
Experiment 2. The average within-participant response latencies were 1045.66 ms (SE
38.21 ms) for words rated low in BOI and 932.37 ms (SE 36.27 ms) for words rated high in BOI.
The average between-item response latencies were 1038.09 ms (SE 20.00 ms) for words rated
low in BOI and 931.40 ms (SE 11.55 ms) for words rated high in BOI.
The Bayes factor was 1,586,393 for the within-participant analysis, and 937.31 for the
between-item analysis. The within-participant analysis provides very convincing evidence that
response latencies are best predicted by a model that accounts for BOI than one that does not,
with high-BOI words being responded to more quickly than low-BOI words. The between-words
item analysis provides very convincing evidence that response latencies are best predicted by a
model that accounts for BOI, with high-BOI stimuli producing faster response latencies across
participants than did low-BOI stimuli.
Experiment 3. The average within-participant response latency was 1255.00 ms (SE
45.71 ms) for words rated low in EE, and 1139.56 ms (SE 40.16 ms) for words rated high in EE.
The average between-item response latencies were 1227.56 ms (SE 17.75 ms) for words rated
low in EE, and 1126.21 ms (SE 13.43 ms) for words rated high in EE. The Bayes factors were

156
17,444.18 for the within-participants analysis of response latencies differences, and 949.76 for
the between-item analysis. The within-participant analysis provides very convincing evidence
that response latencies are best predicted by a model that accounts for EE than one that does not,
with high-EE words being responded to more quickly than were low-EE words. The betweenwords item analysis provides very convincing evidence that response latencies are best predicted
by a model that accounts for EE, with high-EE stimuli producing faster response latencies across
participants than did low-EE stimuli.
Experiment 4. The average within-participant response time was 977.72 ms (SE 21.36
ms) for concrete words rated low in EE, and 1040.11 ms (SE 28.33 ms) for words rated high in
EE. The average between-item response latencies were 980.98 ms (SE 13.54 ms) for words rated
low in EE, and 1040.55 ms (SE 25.83 ms) for words rated high in EE.
The Bayes factor was 151.34 for the within-participant analysis, and 4.75 for the
between-item analysis. The within-participant analysis provides convincing evidence that
response latencies are best predicted by a model that accounts for EE than one that does not, with
low-EE concrete words being responded to more quickly than were high-EE concrete words. The
between-words item analysis provides some evidence that response latencies to concrete nouns
are best predicted by a model that accounts for EE, with low-EE concrete nouns producing faster
response latencies across participants than did low-BOI stimuli.
Electroencephalography.
Experiment 2. A full-scalp view of the electrode placements used across all three
experiments is presented in Figure 13. Individual ERP waveforms for all electrodes are presented
in Figure 14. Figure 15 depicts a full-scalp topography of ERP polarity across time. The ERP
waveform resulting from this task is characterized by four distinct components. In chronological

157
order, there is first an early positive deflection evident approximately 100-200 ms after stimulus
presentation that appears in all but the centre-most electrodes. This is most likely a P200
component reflecting initial task processing, with condition differences at this time being
illustrative of task difficulty. Second, a negative component is evident between 200 and 420 ms
after stimulus presentation that is visible in all but the centre-most electrodes, but most
pronounced in parieto-occipital locations. This activity is evident earlier for words rated low in
BOI and is more pronounced in the left hemisphere (Figure 15). This is most likely a N400
component reflecting the integration of stimuli with current task demands with greater
amplitudes at this time reflecting increased integration difficulty. Third, a long positive
deflection is evident beginning as early as 300 ms after stimulus presentation and extending past
the 800 ms window in some electrodes. This deflection is primarily observable bilaterally in
electrodes in temporal and occipital regions, though it is more prominent in the right hemisphere.
This is likely a P600 or LPC and is reflective of re-evaluation of stimuli, with greater amplitudes
suggesting increased processing for atypical exemplars. Finally, a small number of more central
electrodes show a relatively flat, if not negative, component from approximately 420-800 ms
after stimulus presentation. This negativity is evident primarily in the left fronto-temporal
regions and is more pronounced for words rated high in BOI.
To examine the effects of grounded semantic richness in the production of these four
ERP components, difference waves were calculated by subtracting the value of the averaged low
BOI ERP from the averaged high BOI ERP at each 4 ms time point. Then, as described above,
99% Bayesian credible intervals were calculated and plotted at each time point. If these intervals
extend past, and include zero, then it can be concluded that a model allowing for the difference
between conditions to be not-zero at that time point is not compellingly stronger than a model

158
representing the null hypothesis. For those time points where zero is not included within the
credible intervals, it can be concluded that a model describing a non-zero difference better
predicts the observed data than does a model reflecting a null hypothesis. Full interpretation of
these difference waves and credible intervals requires an examination of the timing and direction
of the non-zero differences within the context of the original averaged ERPs. Only electrodes in
which notable differences were observed will be presented here.
Figure 6 shows the difference wave for electrode 8 (close to the midline, right
hemisphere, parietal region). Approximately 300 ms after stimulus presentation, there is a brief
instance where the non-zero model appears more probable than the null hypothesis. However,
given that the difference wave is reliably negative at this time, this non-zero effect may be
indicative of meaningful condition differences, especially given the conservative thresholds used
in these analyses. This difference reflects either the N400 or the early portions of the pronounced
P600 evident in Figure 14 (see Electrode 8). As the N400 is sensitive to semantically incongruent
information, and High BOI words are thought to be more congruent with imageability, this
information is suggestive that these differences are due to increased positivity during the early
moments of the P600, reflecting the participants’ reanalysis of low BOI words because lower
levels of motor knowledge led to more extensive decision-making processing regarding whether
a word is easily imageable.
Experiment 3. The same four ERP components are evident in this experiment as are in
Experiment 2. ERP waveforms for individual electrodes are presented in Figure 17 while Figure
18 depicts a full scalp time course of ERP polarity. The P200 is evident in all but the centralfrontal electrodes and is more pronounced in the left hemisphere. The N400 is seen in all but the
central-parietal regions and is most pronounced in the fronto-temporal regions. For words rated

159
low in EE, the N400 is more pronounced in the right hemisphere, whereas it is more pronounced
in the left hemisphere for words rated high in EE. The N600 is most evident for words rated high
in EE and is most pronounced in frontal regions of the left hemisphere. Through visual
inspection, it appears that the left-frontal negativity is more extreme than was observed in
Experiment 2. The P600 is more positive and persists longer for words rated low in EE. Visual
inspection suggests that this positivity is less extreme than was observed in Experiment 2.
As in Experiment 2, difference waves were calculated by subtracting the averaged ERP
for words rated low in EE from the averaged ERP for words rated high in EE every 4 ms after
stimulus presentation, and these figures can be interpreted as described above in Experiment 2
(Figures 19-29). Only those difference waves that illustrate non-zero differences are presented.
Negative differences in the P200 are evident in Figures 19-21 (electrodes 1, 3, 5; left
hemisphere, frontal and parietal regions) suggesting that low EE abstract nouns required greater
amounts of processing at this stage. It appears that there is no P200 evident for words rated high
in EE in frontal regions (Figure 17; electrodes 1, 2, 3; 1 & 3 left hemisphere), suggesting that
words rated high in EE required very little initial processing in this SCT given their congruence
with the decision criterion. However, it is also possible that this difference reflects an early onset
of the N400 for these trials, though this may suggest that high EE abstract nouns require more
effort to integrate contextually or that early processing at this time reflects the processing of
emotion information itself. Words rated high in EE consistently produced more negative
deflections between 200 and 800 ms after stimulus presentation. While the later windows appear
to be trending positive and can be interpreted as low EE words producing a greater P600
reflecting additional processing due to task incongruence, the 200-400 ms window is less easily
interpreted. These data suggest that high EE words, which are hypothesized to be the most

160
typical exemplars of abstract concepts in this experiment, may be either more difficult to process
contextually, or are more difficult to retrieve from long term memory, despite task demands
priming abstractness.
Bilateral positive P200 condition differences are evident, suggesting greater early
processing of words rated high in EE (Figures 24-28; electrodes 14, 16, and 24-26 respectively;
see Figure 13 for spatial references). These differences occur predominantly in the right
hemisphere and encompass primarily temporal and occipital regions with some parietal and
inferior temporal lobe effects evident. As a result, for these electrodes, it can be concluded that a
model that accounts for the effects of EE is a better predictor of these data than is a model
representing a null hypothesis without this effect. Unlike Experiment 2, these P200 differences
cannot be interpreted as additional processing of atypical task exemplars. However, as per Wang
et al. (2019), these P200 differences may reflect increased processing of emotion information
itself.
Evidence for the superiority of a model including the effects of EE in the N400 deflection
is apparent in left hemisphere fronto-parietal regions, and in the right fronto-parietal (Figures 1922; electrodes 1, 3, 5, and 11 respectively; see Figure 13 for spatial references) ranging from 290
to 450 ms after stimulus presentation. In the left hemisphere, a negative deflection occurs during
this time, illustrating that the N400 was more pronounced for words rated high in EE.
Consequently, these data reveal increased activity during this window for emotion information
itself as it cannot be reasonably concluded that high EE abstract concepts are more difficult to
retrieve or to contextualize in Experiment 3, especially when the behavioural data show that
these words were responded to more quickly and accurately. Unlike the electrodes that showed
increased P200 activity for high-EE words, the electrodes showing increased processing for

161
high-EE words during the N400 are located on the left hemisphere, predominantly located over
frontal and parietal regions. These data may have been produced in part by more extreme
negativity during a N170, which has been found to be influenced by emotion information (Zhang
et al., 2014).
Condition differences are evident at several left-hemisphere electrodes between 500-800
ms after stimulus presentation (Figures 19, 21, 22, and 29; electrodes 1, 5, 11, 13, and 29). In the
left fronto-temporal regions, positive differences occur late in the ERP and may reflect an N800,
which suggests greater amounts of activity for words with more emotion information, which is
characteristic of this deflection (De Pascalis et al., 1996). Negative differences between 500 and
750 ms after stimulus presentation in the temporal lobe (Figure 13) occurs during a P600
window with words rated low in EE showing greater positive deflections, capturing extended
effort for processing atypical abstract nouns. However, the positive deflection at this time
quickly returns to baseline, where the later negative differences appear at 500 and 750 ms after
stimulus presentation. It is possible that this reflects both a P600 and a N600/800 in the left
temporal lobe that are counteracting one another as their signals are summed. If this is true, then
the later component of the ERPs should be viewed as competing signals reflecting additional
attention to low-EE nouns due to task incongruence being expressed simultaneously with
increased emotion processing for the high-EE abstract nouns. Electrode 29 also shows a P600
component with words rated low in EE having a more pronounced deflection occurring in the
anterior fronto-temporal area (Figure 29; electrode 29), further suggesting additional processing
of low-EE abstract concepts.

162
Experiment 4. The same four ERP components are evident in Experiment 4 as observed
in Experiments 2 and 3 (Figure 30). A full-scalp topographic time course of ERP activity is
depicted in Figure 31.
The P200 component is evident bilaterally in temporal and occipital regions, though
visual inspection suggests the P200 is less pronounced than in Experiment 3. This result likely
reflects the comparative ease of this task compared to that of Experiment 3 (i.e., concrete nouns
are generally responded to more quickly and accurately than are abstract nouns; Newcombe et
al., 2012). The N400 is also evident bilaterally in all electrodes except for the most central
electrodes. The N400 is most pronounced in left frontal-temporal areas. The P600 appears in left
frontal regions and visual inspection shows the N600 to be more negative than in Experiments 2
and 3 and is most evident in temporal, parietal, and occipital regions with visual inspection
revealing greater magnitude in the right hemisphere.
As in Experiment 3, difference waves calculated by subtracting the averaged ERP for
concrete nouns rated low in EE from the averaged ERP for those rated high in EE and can be
interpreted as in Experiments 2 and 3 (Figures 32-35; electrodes 3, 5, 11, and 29 respectively).
Only those difference waves that illustrate non-zero differences are presented.
All meaningful differences were observed in frontal and temporal regions in the left
hemisphere. A negative difference between 150 and 200 ms after stimulus presentation was
recorded in electrode 3 (Figure 32). At this location, there was no marked P200 component for
words rated high in EE, whereas the ERP for words rated low in EE shows a small positive
deflection. The difference in the P200 between these conditions could imply that low-EE
concrete nouns require additional processing. However, this interpretation does not seem
plausible given that low-EE concrete nouns are typical exemplars of concrete concepts.

163
Alternatively, this difference could be interpreted as an early onset of the N400 for both
conditions, but more so for concrete nouns rated high in EE. This alternative would be
interpreted as additional processing to integrate high levels of emotion information with the SCT
decision criterion in this experiment because emotion information is indicative of abstract
concepts, thereby making concrete nouns rated high in EE diagnostically incongruent with the
decision criteria to respond concrete nouns.
A negative difference between 300 and 450 ms after stimulus presentation is evident in
electrodes 3 and 5 (superior fronto-parietal, left hemisphere; Figures 32 and 33). It appears that
the N400 at this time is reflecting additional processing for high-EE concrete concepts. However,
in Experiment 3, the same pattern was observed for abstract words rated high in EE. The greater
negativity to both concrete and abstract nouns rated high in EE may suggest that the greater
N400 in this experiment reflects the processing of emotion information and/or the additional
effort to integrate high levels of EE within the context of this experiment’s decision criterion.
Concrete nouns rated high in EE produced a more negative deflection across the entire ERP time
course at electrodes 3 and 5 (Figures 32 and 33). However, there are no truly pronounced ERP
components evident at these locations (Figure 30). This comparatively flat ERP suggests that, at
these locations, the total activation of the surrounding areas may have summed to a
predominantly null ERP. Although this makes the interpretation of these data more difficult, the
reliably more pronounced negativity for concrete concepts rated high in EE is in line with the
observations in Experiment 3 that high EE nouns require additional processing due to their
grounded semantic richness obtained through emotion experiences.
Lastly, late (600-800 ms after stimulus presentation) negative differences are evident in
electrodes 11 and 29 (inferior fronto-temporal regions; Figures 34 and 35). The ERP component

164
at this time for electrode 11 (Figure 34) is relatively flat for words rated low in EE but appears to
reflect a N600 for concrete nouns rated high in EE, again showing additional processing for
either stimuli with greater levels of grounded semantic richness that are incongruent with current
task demands or for emotion information itself.
Discussion
Collectively, the data presented above largely support the initial behavioural and ERP
hypotheses. Behaviourally, response latencies were hypothesized to be faster to stimuli with
greater amounts of grounded semantic richness when that grounded semantic richness is
congruent with task demands (Experiments 2 and 3) and to be slower when that grounded
semantic richness is incongruent with task demands (Experiment 4). An extended discussion of
these data’s implications for the grounding of conceptual knowledge is detailed in Chapter 3. To
summarize: grounded information is automatically activated as part of a simulation during the
processing of conceptual knowledge regardless of whether that knowledge is task relevant or
beneficial to successful task completion.
Greater amplitudes during the P200, N400, and P600 windows were predicted for words
rated low in BOI and EE in Experiments 2 and 3 due to the additional processing required for
stimuli with less diagnostic grounded semantic richness, while greater amplitudes to words rated
high in EE were predicted during the same windows in Experiment 4 due to the incongruence
between grounded semantic richness and the decision criterion.
The evidence suggests that, when accessing conceptual knowledge, grounded semantic
richness information that was obtained through bodily and emotion experiences with the
environment is automatically activated or simulated. Further, the ERP data suggest that grounded
semantic richness effects are not singular in nature. Rather, ERP components reflecting the

165
influence of grounded semantic richness are produced from several neural processes, in both
hemispheres, and come online at varying times after stimulus presentation, as evidenced by
condition differences across four ERP components. This supports the soft assembly perspective
of conceptual knowledge (Anderson et al., 2013). Anderson et al. (2013) propose that cognitive
processes are unlikely to occur in discrete brain regions; it is more likely that cognitive processes
are “softly assembled” across several brain regions at once. Softly assembled objects are only
identifiable as discrete entities at a specific moment in time. However, there is never a particular
number or configuration of constituents that defines the object. For example, a flock of birds is
softly assembled; a flock can be recognized as such, but there is no distinct number or
configuration of birds that qualifies the birds as a flock. Similarly, cognitive processes like
memory, attention, or language comprehension are likely performed across several brain regions
concurrently, with no discrete combination of activity distinctly qualifying the activation as a
particular process. Any observable neural response to a stimulus will, as a result, be observed as
a shift within an ever-changing dynamic system rather than a discretely observable event. Data,
such as those presented here, would not be observed if conceptual knowledge was stored
together as a discrete abstract representational code (Niedenthal et al., 2005). These data
contribute to the growing body of work that is demonstrating how cognitive processes occur the
way they do.
Electroencephalography Results.
An understanding of the ERP data is important because they provide a time course
analysis that complements the interpretations provided by an analysis of response latencies and
accuracy (Kounios & Holcomb, 1994). Although a substantial amount of EEG data was
discarded due to the presence of artifacts, one can be confident in the data because the ERP

166
components observed in the experiments described above are consistent with what is expected in
SCT experiments. There were four primary ERP components evident for all three experiments.
The first, the P200, has previously been observed in anterior and central regions, and greater
magnitudes have been interpreted as reflecting task difficulty and automatic evaluation of stimuli
(e.g., Taylor & Khan, 2000), as well as semantic richness effects (Kounios et al., 2009). Second,
the N400 has been observed most frequently in central and parietal regions and has been
interpreted, in part, as reflecting the amount of effort needed to integrate stimuli into current task
demands (e.g., Heinze et al., 1998). Third, the P600 has been observed in centro-parietal regions
and is thought to reflect automatic reanalysis of stimuli. Finally, the N600 has been observed
over fronto-central regions and has been interpreted as reflecting additional processing due to
perceived incongruities (e.g., Du et al., 2013). Each component will be discussed in turn below.
P200 Observations. In the first 300 ms after stimulus presentation, visual inspection of
scalp topography (Figures 15, 18) shows that, in Experiments 2 and 3, early positivity is most
evident and diffuse to words rated low in grounded semantic richness, regardless of its type (BOI
or EE), and its congruence with decision criteria. In Experiment 4 (Figure 39), however, the
same pattern is observed for words rated high in EE. These observations support the hypothesis
that greater P200 amplitudes would be produced by stimuli with comparatively low levels of
task-congruent grounded semantic richness.
Interestingly, visual inspection suggests that the scalp topography of the early positivity
differs between experiments. In Experiment 2, when the primary source of grounded semantic
richness was BOI, early positivity was predominantly posterior for low BOI trials. In
Experiments 3 and 4, however, when the source of grounded semantic richness was EE, early
positivity is more frontal and more focused in the left hemisphere. These data are similar to those

167
of Pauligk et al. (2019) who reported more pronounced central and posterior concreteness
effects, whereas emotion effects were observed more frontally and in the left hemisphere. This
difference in ERP distribution also supports the hypothesis that, from a grounded cognition
perspective, qualitatively different ERP scalp distributions should be observed as a function of
the nature of a word’s grounded semantic richness.
Kounios et al. (2009) observed that the P200 increased in magnitude for semantically
rich trials (as observed in Experiment 4), but Kounios’s (2009) interpretation does not easily
align with the current data. Kounios et al. proposed that semantically rich trials produce higher
P200 magnitudes; however, more difficult trials are thought to also produce higher P200
magnitudes. Given the widely observed effects of semantic richness correlating with faster
response latencies, these two interpretations are not compatible. In other words, as in
Experiments 2 and 3, SCT trials with stimuli rated high in task-congruent grounded semantic
richness would be considered relatively easy, and subsequently predicted to have both a more
pronounced P200, due to greater levels of grounded semantic richness, and a lower P200 due to
the comparative ease of processing. In Experiment 4, however, one might predict that the
increased amounts of emotion information for concrete nouns would make the decision, “Is this a
concrete noun?” more difficult due to the incongruity between increased levels of EE and the
decision criterion of “is the word concrete?”, thereby producing longer response latencies and
greater P200 magnitudes. The P200 deflections to stimuli rated high in EE in Experiment 4
would also be predicted to be greater in amplitude due to the comparatively greater levels of
grounded semantic richness. Although the response latencies were longer for High EE trials, the
P200 in Experiment 4 was of lower magnitude for words than in Experiments 2 and 3. More
research is needed to confirm whether the P200 deflection reflects activity from multiple

168
locations (reflecting multiple streams of automatic processing), thus producing dissociable
effects that can be produced independently by either grounded semantic richness, task relevance,
or trial difficulty.
Interestingly, in Experiment 2, the effects of BOI during the P200 were statistically
evident in the right hemisphere at a parietal electrode, whereas in Experiment 3, EE effects in the
P200 were observed in the left hemisphere in frontal and parietal electrodes, and in Experiment
4, the effect of EE on the P200 was also evident in the left hemisphere, at a frontal electrode.
These differences are perhaps due to the source of the grounded semantic richness in the
experiments; BOI and EE are processed differently very quickly after stimulus presentation. In
Experiment 3, when the influence of EE was congruent with task demands, EE improved the
strength of a predictive model at frontal sites bilaterally. Specifically, abstract nouns rated low in
EE elicited lower P200 amplitudes at these electrodes, which suggests that words with less
grounded semantic richness did not require as much processing at this stage. As such, it is
possible that the increased P200 amplitudes at frontal sites for words rated high in EE may
reflect the processing of emotion information itself rather than of task difficulty.
A caveat here is that the P200 component, despite showing condition differences, occurs
at the same time across experiments and conditions. This consistent timing across experiments
and conditions suggests that the processing differences at this time do not directly reflect or
produce the observed response latency differences. Rather, processing differences, as captured in
P200 differences in the current experiments, likely influence response times through feedforward facilitation of later ERP processes. Given the absence of a visible P200 for high EE
trials in Experiment 4, it could be concluded that the type of processing that this component
captures does not reflect task difficulty incurred by task relevance per se. Rather, it appears as

169
though what is being observed in the current experiments, in addition to the reported P200
effects of additional processing for atypical task exemplars, is evidence of very early, automatic
processing of grounded semantic richness itself. Further, the fact that the dimensions of
grounded semantic richness that were manipulated in these experiments were grounded in bodily
experience lends credibility to the claim that conceptual knowledge is inherently grounded in
sensory, motor, and emotion experiences.
Across all experiments, whenever grounded semantic richness was high, regardless of
congruence with task demands, diminished P200 components were observed in frontal regions,
though there were some hemispheric differences concerning the specific electrodes that showed
the effects. However, when the source of grounded semantic richness was EE, and that
information was task congruent, greater P200 amplitudes were observed in more posterior
electrodes, primarily in the right hemisphere. When high levels of EE were incongruent with task
demands, these effects were not seen, though the P200 was still diminished in frontal sites for
high-EE trials.
In summary, the P200 reported here does not appear to be a singular component that
solely captures task difficulty. A visual comparison among experiments suggests that the P200 is
more pronounced for concrete than for abstract nouns, and yet further pronounced on trials that
contextually require deeper processing. Furthermore, when the source of grounded semantic
richness is BOI, condition differences are observed in the right hemisphere; yet differences are
observed primarily in the left hemisphere when EE is the critical source of grounded semantic
richness, with these differences more evident when EE is task-relevant in facilitating responding
to abstract nouns. These data suggest that the activation of grounded semantic units is inherent to

170
conceptual processing in single word presentation, but that the extent to which that information
is processed varies as a function of task congruence.
N400 Observations. The topographical plots (Figures 14, 18, and 30) for all three
experiments highlight some differences in the negative activity occurring 200-500 ms after
stimulus presentation, which may be best interpreted as a N400 component. The earliest negative
deflections are evident in Experiments 3 and 4 in both high- and low-EE conditions, and
primarily appear in left posterior regions. This early posterior negativity is less evident in
Experiment 4 (responding to concrete nouns) than in Experiment 3 (responding to abstract
nouns). Although stimuli in both experiments varied along the EE dimension, EE was not
diagnostic of a go-trial in Experiment 4, and thus did not exert facilitatory response latency
effects. However, the early N400 components between 200 and 300 ms after stimulus
presentation appear more negative to low-EE trials in both experiments. As such, this early N400
activity could be interpreted as reflecting the automatic processing of EE, regardless of task
relevance, with greater negativity when EE was diagnostic of a go-trial, and additionally
reflective that more processing was required, across experiments when EE information is less
available, regardless of task relevance. In Experiment 2, however, where participants were
responding to concrete nouns and BOI was the manipulated dimension of grounded semantic
richness, left posterior N400 activity is also evident, though not pronounced until approximately
300-400 ms after stimulus presentation (Figure 18), and is most pronounced on High BOI (task
congruent) trials. These data collectively suggest that semantic information obtained through
emotion experiences is processed automatically between 200 and 400 ms after stimulus
presentation, whereas semantic information obtained through physical interactions with the
environment is processed between 300 and 400 ms after stimulus presentation in the same

171
posterior brain regions. As such, there is no single interpretation of the left posterior N400 across
these experiments, supporting the assertion made above that this component is not a single entity,
as neither the nature of grounded semantic richness being processed nor the relevance of high
levels of grounded semantic richness are sufficient alone to predict the timing or magnitude of
the N400.
Although visual inspection of the N400 in Experiment 2 suggests increased intensity for
High BOI trials, analysis of the difference waves show no evidence of condition differences
during the N400 window. This result was unexpected given the common interpretations that the
N400 captures contextual integration or memory retrieval. Further, the result is especially
unexpected given the marked response latency differences observed in Experiment 2, and that
concrete nouns have been found to produce more pronounced N400s than do abstract nouns
(Holcomb et al., 1999; Kounios & Holcomb, 1994). One possible explanation of this observation
is that all Experiment 2 stimuli were comparably easy to contextualize as imageable given the
extended intertrial interval, the relative simplicity of the task, and that all go-trials were concrete
nouns. As a result, although an N400 is evident in Experiment 2, these data do not support the
conclusion that BOI is differentially processed during this time.
Clearly, the observed N400 does not reflect the same semantic processes in all three
experiments. N400 activity was observed in all three experiments, but meaningful condition
differences were only observed during the N400 window in Experiments 3 and 4 when EE was
the manipulated dimension of grounded semantic richness. It can be concluded, then, that the
negative ERP activity during the N400 window in these three experiments reflects heterogeneous
semantic processes, and that this activity is particularly sensitive to grounded semantic richness
obtained from bodily emotion experiences. Future research should look at components before

172
and after the N400 when looking for ERP evidence of grounded conceptual processing of other
grounded semantic richness dimensions.
It is interesting that there were condition differences in Experiment 4 during the N400
window. Even though emotion information did not strictly need to be processed to successfully
recognize a concrete noun, it is important to remember that the no-go trials were abstract nouns.
If, as hypothesized, abstract concepts are intrinsically represented through their emotion
information, it is possible that participants responding under an “Is this a concrete noun”
decision criterion may have automatically activated EE information so as to recognize a no-go
trial. That is, when emotion information is even implicitly relevant, N400 magnitudes
automatically reflect the amount of emotion information on that trial. If participants were
inhibiting their responses when high levels of EE were activated (i.e., if the trial initially led
toward a decision that the stimulus was an abstract noun), one would predict the response latency
differences that were observed (high-EE concrete nouns being initially recognized as abstract
concepts, thereby requiring inhibition), and the observed N400 condition differences even
though EE was not diagnostic of a go trial. This sort of interpretation is congruent with previous
research that has suggested that the N400 can capture the inhibition of automatically activated
incompatible knowledge (e.g., Debruille, 1998; Kounios, 1996).
The topographical plots for all three experiments (Figures 15, 18, & 31) are consistent
with prior research demonstrating that the scalp distributions of the N400 to abstract and
concrete nouns are known to differ, suggesting that the cognitive and neural demands for
processing both word types are fundamentally different (Holcomb et al., 1999; Kounios &
Holcomb, 1994; West & Holcomb, 2000). Such differences have been interpreted as being the
product of at least two separate neural contributors to the N400 including an imagine system and

173
a verbal processing centre (Kounios & Holcomb, 1994). Holcomb et al. (1999) showed that
concrete nouns are more likely than abstract nouns to produce an N400 in frontal regions. This
observation is incongruent with the current findings. However, given the go/no-go design
employed in Experiments 2-4, there can be no direct comparison made between the experiments
presented here and Holcomb et al.’s (1999) experiments as data for responses to concrete and
abstract nouns are not available under the same task demands in the experiments presented here.
That is, in the current go/no-go design, data are only available for concrete or abstract trials in
any one experiment, whereas in a yes-no design, data are available for both trial types in the
same experiment with the same decision criteria. However, these data still support a modal
perspective of conceptual knowledge processing in that amodal theories would not intuitively
predict separate neural contributors for the processing of different dimensions of grounded
semantic richness.
In Experiment 2, the only statistical evidence of a N400 ERP condition difference was
observed in the frontal region of the right hemisphere, with words rated high in BOI producing a
more negative ERP beginning as early as 300 ms after stimulus presentation. This is in line with
previous research, described above, suggesting that concrete nouns should activate a frontal
imagine system, and that the most imageable words produced the strongest effect in these
regions (Holcomb et al., 1999). In Experiments 3 and 4, however, statistical evidence of N400
differences was found almost exclusively in frontal and medial regions of the left hemisphere. In
all such cases, the N400 was most pronounced for words rated high in EE, regardless of whether
participants were responding to abstract or concrete nouns. The only exception to this was a
statistical difference showing more pronounced N400 negativity for words rated low in EE in the
right hemisphere in Experiment 3. It is possible that this difference reflects, for atypical abstract

174
nouns, a unique processing difference that occurs in the right hemisphere, whereas grounded
semantic richness derived from EE is processed predominantly in the left hemisphere. These data
support the grounded cognition hypothesis that emotional experiences with words’ referents
should be automatically activated whilst being neurologically discernible in semantically
influenced ERP components.
Late ERP Observations. Late ERP differences were observed in all three experiments.
N600 condition differences were observed in the left hemisphere in both Experiments 3 and 4,
with more pronounced negativity observed for high-EE trials. Due to the different decision
criteria in each experiment, this additional negativity can be interpreted as additional processing
of emotion information, regardless of its task relevance. Given how late in processing these
effects were observed, it is unlikely that these differences are directly contributing to response
latencies. This interpretation is supported by the differences in response latencies for high-EE
trials in each experiment, with high-EE words being responded to more quickly in Experiment 3,
but more slowly in Experiment 4. Therefore, what was observed in the N600 in Experiments 3
and 4 would be best explained as sustained elaborative processing of emotion information itself
independent of task demands. The production of additional neural activity by emotion
information, regardless of its facilitatory or inhibitory contribution to response execution,
suggests that it is an automatic element of conceptual processing when grounded semantic
richness obtained from bodily emotion experiences is available.
Late positive condition differences were observed in the P600/LPC in Experiment 2 and
3 with less semantically rich trials producing greater positivity. That these differences were
observed in low grounded semantic richness trials is congruent with the interpretation that the
P600 captures the amount of effort needed to process a concept and is reflective of trials that

175
require further processing (Hagoort, 2003). The right hemisphere difference for Experiment 2
was found over the superior parietal area, whereas the left hemisphere difference for Experiment
3 was found in inferior temporal regions. These results suggest that, although P600 differences
are being evoked in both experiments by trials that are incongruent with task criteria due to
having less grounded semantic richness, the neural generators responsible for the additional
processing required for these experiments is different due to the specific type of grounded
semantic information (BOI and EE) being processed. Similar to the P200 and N400 discussions
above, these data suggest that, much like how semantic richness is not derived from a single
source of information, the P600 component is not a singular event with a single underlying
process (e.g., Kounios et al., 2009).
Visual inspection of the topographic plots for Experiments 2, 3, and 4 (Figures 15, 18,
and 31) shows that more pronounced negativity was observed in the left hemisphere’s frontal and
temporal regions for words rated high in EE in Experiments 3 and 4. That this result was not
observed in Experiment 2 suggests that this observation is unique to EE and not indicative of
grounded semantic richness in general, but supports the observation that different sources of
grounded semantic richness account for unique variance in SCT and LDT performance (Pexman
et al., 2008). Again, regardless of task relevance and the associated behavioural consequences,
evidence of grounded semantic richness is observable with ERP analyses.
Conclusions. The use of EEG to monitor SCT performance shows that ERP waveforms
capture condition differences elicited by the manipulation of grounded semantic variables.
Specifically, in the experiments presented above, the manipulated dimensions of grounded
semantic richness did not need to be activated to successfully complete the experimental tasks.
That ERP waveforms reflect varying levels of automatically activated grounded semantic

176
richness that is not being explicitly processed by participants is compelling evidence that
grounded semantic richness is an automatic and integral part of conceptual knowledge.
The observation that grounded semantic richness obtained through emotion experiences
was incongruent with concrete noun recognition in Experiment 4 further supports the
interpretation of these data as reflecting the grounded nature of conceptual processing
(Vigliocco, et al., 2009). If accessing grounded information was somehow strategic in nature,
rather than a necessary and automatic part of conceptual activation, the behavioural and ERP
data in Experiment 4 ought not to have demonstrated an influence of emotion information on
concrete noun recognition. Rather, given that the task was to identify concrete nouns,
participants would have created response decision rules that pertain exclusively to concreteness
(e.g., BOI, imageability, etc.), and emotion information, being indicative of abstract concepts,
and not concrete concepts, would not influence behaviour or neural activity. As such, these data
strongly indicate that grounded semantic information is an integral element in the processing of
conceptual knowledge. The addition of EEG measurements helped support these conclusions.
Early processing differences were observed as a function of manipulated grounded
semantic richness, suggesting that semantic information is activated within the first 150-200 ms
after stimulus presentation. Further, this activation is differentially sensitive to the amounts of
grounded semantic information, regardless of task relevance. These observations support the
conclusion that modal information is inherently part of conceptual knowledge.
The N400 component was sensitive to the processing of emotion information but was not
sensitive to BOI. This observation will be important to researchers choosing the time windows
for more typical ANOVA-style analyses and for researchers selecting semantic dimensions for
future study. Additionally, the N400 emotion differences in Experiments 3 and 4 varied

177
according to the hemisphere in which they were observed, despite being produced by
manipulations of the same semantic dimension (EE). These data help to illustrate when emotion
information contributes to response execution in single word presentation SCTs and supports a
more full-scalp approach to observing semantic effects, rather than targeting regions of interest,
as is more commonly practiced.
Beyond the N400 deflection, care must be taken to dissociate between positive and
negative deflections, as they were found, here, to be differentially sensitive to the experimental
manipulations. The late positive deflections showed condition differences only in Experiments 2
and 3 and, in both experiments, increased positivity was observed for low grounded semantic
richness trials. As mentioned above, this is in line with the interpretation that these trials required
additional processing due to their task incongruence. This effect was not observed in Experiment
4, perhaps due to the relatively weaker effects observed when participants were responding to
concrete nouns. The Bayes Factors for the behavioural effects in Experiment 4 were an order of
magnitude lower than in the other two experiments, despite still providing convincing evidence
for statistical models that include EE. Perhaps this indicates that the influence of EE in
Experiment 4 was less pronounced than were the manipulated dimensions of grounded semantic
richness in the other two experiments. That is, BOI may be integral to concrete nouns even when
assessing imageability, and EE may be integral to judging abstractness, while EE may indirectly
signal “not concrete” when judging concreteness. This would explain the lower Bayes Factors
and the absence of pronounced condition differences in the later positive ERP components in
Experiment 4. The N600, however, appears to have been sensitive primarily to the processing of
additional emotion information as it only showed condition differences in Experiments 3 and 4.
In both experiments, the N600 differences were observed in fronto-temporal regions of the left

178
hemisphere. The similar timing and location support the conclusion that these N600 differences
reflect similar neural generators and emotion-information-processing functions.
While some researchers employ jittered stimulus presentation, the experiments presented
here did not. Using a variable inter-trial interval is commonly done to avoid the possibility that
each stimulus appears at the same phase of a brain wave (e.g., alpha waves). For example, if all
go-trials’ presentation was phase locked to the peak of the alpha wave, then the data would be
systematically biased across all ERP deflections. However, most cognitive experiments employ
inter-trial intervals of ~500-3000 ms, whereas the experiments presented here employed at
15,000 ms interval. As such, the probability of such a long interval becoming systematically
synchronized with an underlying brain wave was highly unlikely. Additionally, the influence of
alpha wave activity is generally removed during filtering. Another reason researchers may
employ presentation jittering is to avoid recording anticipatory activity as when participants can
reliably predict the presentation of each stimulus. This is important as EEG activity prior to
stimulus presentation is frequently used for baseline corrections, and any systematic pattern of
activity during that time would thus bias the correction calculation. However, as the delay
between stimulus presentations increases, the ability of participants to accurately predict the
moment of presentation decreases, thus reducing any predictive activity to random noise in the
final analyses.
Collectively, the ERP data presented here suggest that grounded semantic richness effects
can be observed early in processing, that grounded emotion effects are evident during the N400
window, and that trials low in grounded semantic richness produce P600 differences when
grounded semantic richness is more relevant to task completion. It is exciting to see the
effectiveness of full-scalp time course analyses. Though the behavioural effects of grounded

179
semantic richness have been observed for some time, observations of the neural correlates of the
differing time course for the activation of semantic knowledge is needed (Amsel, 2011).
However, the three experiments reported here represent a first step in testing these hypotheses.
That is, to definitively state that the neural regions producing these data are motor and perceptual
regions, techniques that allow for greater spatial resolution will need to be employed. Although it
is tempting to emphasize the number of ERP condition differences evident in the left
hemisphere, where language processes are thought to primarily reside, it needs to be noted that
ERP components can be produced by lateralized neural activity, or by contralateral generators
whose measurable electrical field is observed in the other hemisphere (Kounios & Holcomb,
1994). Future studies with denser EEG arrays employing techniques designed to increase spatial
resolution could clarify the neural regions producing condition differences in ERP analyses (e.g.,
Current Source Density estimates and Surface Laplacian computation; Burle et al., 2015).
Regardless, these data demonstrate that the nature of semantic knowledge activation is
performed across the brain in different regions, and that information obtained from bodily
experiences with the environment is activated automatically, regardless of task relevance.
Conceptual knowledge is, ultimately, grounded in nature.

180

Figure 13. Electrode placement reference for all experiments. View is from above, with a person
facing towards the top of the screen. Image obtained from Geodesic Sensor Net Technical
Manual (2007).

181

Figure 14. Topographic display of all ERP waveforms for Experiment 2 with all electrodes
aligned as per Figure 13.

182

Figure 15. Topographic scalp plot for Experiment 2. Times (ms) are time since stimulus
presentation. Red denotes positivity and blue denotes negativity. Scale is -10 to 10 µV.

183

Figure 16. Difference waveform (High BOI – Low BOI) for electrode 8 in Experiment 2. Error
bars are 99% Bayesian credible intervals.

184

Figure 17. Topographic display of all ERP waveforms for Experiment 3 with all electrodes
aligned as per Figure 13.

185

Figure 18. Topographic scalp plot for Experiment 3. Times (ms) are time since stimulus
presentation. Red denotes positivity and blue denotes negativity. Scale is -10 to 10 µV.

186

Figure 19. Difference waveform (High EE – Low EE) for electrode 1 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

187

Figure 20. Difference waveform (High EE – Low EE) for electrode 3 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

188

Figure 21. Difference waveform (High EE – Low EE) for electrode 5 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

189

Figure 22. Difference waveform (High EE – Low EE) for electrode 11 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

190

Figure 23. Difference waveform (High EE – Low EE) for electrode 13 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

191

Figure 24. Difference waveform (High EE – Low EE) for electrode 14 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

192

Figure 25. Difference waveform (High EE – Low EE) for electrode 16 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

193

Figure 26. Difference waveform (High EE – Low EE) for electrode 24 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

194

Figure 27. Difference waveform (High EE – Low EE) for electrode 25 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

195

Figure 28. Difference waveform (High EE – Low EE) for electrode 26 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

196

Figure 29. Difference waveform (High EE – Low EE) for electrode 29 in Experiment 3. Error
bars are 99% Bayesian credible intervals.

197

Figure 30. Topographic display of all ERP waveforms for Experiment 4 with all electrodes
aligned as per Figure 13.

198

Figure 31. Topographic scalp plot for Experiment 4. Times (ms) are time since stimulus
presentation. Red denotes positivity and blue denotes negativity. Scale is -10 to 10 µV.

199

Figure 32. Difference waveform (High EE – Low EE) for electrode 3 in Experiment 4. Error
bars are 99% Bayesian credible intervals.

200

Figure 33. Difference waveform (High EE – Low EE) for electrode 5 in Experiment 4. Error
bars are 99% Bayesian credible intervals.

201

Figure 34. Difference waveform (High EE – Low EE) for electrode 11 in Experiment 4. Error
bars are 99% Bayesian credible intervals.

202

Figure 35. Difference waveform (High EE – Low EE) for electrode 29 in Experiment 4. Error
bars are 99% Bayesian credible intervals.

203

References
Aczel, B., Palfi, B., & Szaszi, B. (2017). Estimating the evidential value of significant results in
psychological science. PLoS ONE 12(8), 1-8.
Amsel, B. D. (2011). Tracking real-time neural activation of conceptual knowledge using singletrial event-related potentials. Neuropsychologia, 49, 970-983.
Anderson, M. L., Richardson, M. J. & Chemero, A. (2013). Eroding the boundaries of cognition:
Implications of embodiment. Topics in Cognitive Science, 4, 717-730.
Andreassi, J. L. (2013). Psychophysiology: Human behavior & physiological response.
Psychology Press.
Barber, H. A., Otten, L. J., Kousta, S., & Vigliocco, G. (2013). Concreteness in word
processing: ERP and behavioural effects in a lexical decision task. Brain and
Language, 125(1), 47-53.
Barsalou, L. W. (1999). Perceptual symbol systems. Behavioral and Brain Sciences,
22(4), 577-660.
Benau, E. M., Gregersen, S.C., Siakaluk, P.D., O’Hare, A.J., Johnson, E.K., & Atchley, R.A.
(2018). Sweet-cheeks vs. pea-brain: Embodiment, valence, and task all influence the
emotional salience of language. Cognition and Emotion, 32, 691-708.
Bennet, S. D. R., Burnett, A. N., Siakaluk, P. D., & Pexman, P. M. (2011). Imageability
and body-object interaction ratings for 599 multisyllabic nouns. Behavior
Research Methods, 43(4), 1100-1109.

204
Binder, J. R. (2007). Effects of word imageability on semantic access: Neuroimaging studies. In
Kraut, M. A., Hard J. (Eds.), Neural Basis of Semantic Memory. Cambridge University
Press, pp. 149-181.
Blomberg, F., Roll, M., Frid, J., Lindgren, M., & Horne, M. (2020). The role of affective
meaning, semantic associates, and orthographic neighbours in modulating the N400 in
single words. The Mental Lexicon, 15(2), 161-188.
Buccino, G., Binkofski, F., Fink, G. R., Fadiga, L., Fogassi, L., & Gallese, V., et al.
(2001). Action observation activates premotor and parietal areas in a
somatoptopic manner: An fMRI study. European Journal of Neuroscience, 13,
400-404.
Burle, B., Spieser, L., Roger, C., Casini, L., Hasbroucq, T., & Vidal, F. (2015). Spatial and
temporal resolutions of EEG: Is it really black and white? A scalp current density view.
International Journal of Psychophysiology, 97(3), 210-220.
Chao, L. L., & Martin, A. (2000). Representation of manipulable man-made objects in
the dorsal stream. Neuroimage, 12, 478-484.
Clarke, A. (2020). Dynamic activity patterns in the anterior temporal lobe represents object
semantics. Cognitive Neuroscience, 11(3), 111-121.
Cortese, M. J., & Fugget, A. (2004). Imageability ratings for 3,000 monosyllabic words.
Behaviour Research Methods, Instruments, and Computers, 36, 384-387.
Coulson, S., King, J. W., & Kutas, M. (1998). Expect the unexpected: Event-related brain
response to morphosyntactic violations. Language and Cognitive Processes, 13(1), 2158.

205
Debruille, J. B. (1998). Knowledge inhibitor and N400: A study with words that look like
common words. Brain & Language, 62, 202-220.
Dehaene, S. (2011). The number sense: How the mind creates mathematics. Oxford: Oxford
University Press.
Delorme, A., & Makeig, S. (2004). EEGLAB: an open source toolbox for analysis of single-trial
EEG dynamics including independent component analysis. Journal of Neuroscience
Methods, 134(1), 9-21.
De Pascalis, V., Fiore, A. D., & Sparita, A. (1996). Personality, event-related potential (ERP)
and heart rate (HR): An investigation of Gray’s theory. Personality and Individual
Differences, 20(6), 733-746.
Duris, J., Kumpan, T., Duffels, B., Matheson, H.E., Pexman, P.M., & Siakaluk, P.D. (2017).
Effects of emotion information on processing pain-related words in visual word
recognition. The Mental Lexicon, 12(3), 283-308.
Estes, Z., & Adelman, J. S. (2008). Automatic vigilance for negative words is categorical
and general. Emotion, 8(4), 453-457.
Fadiga, L., Arbib, M. A., & Rizzolatti, G. (1997). Premotor cortex activation during
observation and naming of familiar tools. Neuroimage, 6, 231-236.
Federmeier, K. D., & Kutas, M. (2002). Picture the difference: Electrophysiological
investigations of picture processing the two cerebral hemispheres. Neurophychologia, 40,
730-747.
Fodor, J. A. (1983). The Modularity of Mind, Cambridge: MIT Press.
Freunberger, R., Klimesch, W., Doppelmayer, M., & Höller, Y. (2007). Visual P2 component is
related to theta phase-locking. Neuroscience Letters, 3, 181-186.

206
Gallese, V., Fadiga, L., Fogassi, L., & Rizzolatti, G. (1996). Action recognition in the
premotor cortex. Brain, 119, 593-609.
Glenberg, A., M., Havas, D., Becker, R., & Rinck, M. (2005). Grounding Language in Bodily
States: The Case for Emotion. In Grounding Cognition: The Role of Perception and
Action in Memory, Language, and Thinking, D. Pecher & R. A. Zwaan (Eds.), pp 224245. Cambridge, UK: Cambridge University Press.
Glenberg, A. M., Robertson, D. A., David, A., 2000. Symbol grounding and meaning: A
comparison of high-dimensional and embodied theories of meaning. Journal of Memory
and Language, 43(3), 379-401.
González, A.J., Triñanes, Y., Zurrón, M., & Carrillo-de-la-Peña, M.T. (2014). Brain processing
of task-relevant and task-irrelevant emotional words. Cognitive, Affective, & Behavioral
Neuroscience, 14, 939-950.
González, J., Barros-Loscertales, A., Pulvermüller, F., Meseguer, V., Sanuán, A., Belloch, V., &
Ávila, C. (2006). Reading cinnamon activates olfactory brain regions. NeuroImage,
32(2), 906-912.
Gouvea, A. C., Phillips, C., Kazanina, N., & Poeppel, D. (2010). The linguistic processes
underlying the P600. Language and Cognitive Processes, 25(2), 149-188.
Hagoort, P. (2003). How the brain solves the binding problem for language: A
neurocomputational model of syntactic processing. NeuroImage, 20(1), S18-S29.
Hargreaves, I. S., Leonard, G. A. Pexman, P. M., Pittman, D. J., Siakaluk, P. D., & Goodyear, B.
G. (2012). The neural correlates of the body-object interaction effect in semantic
processing. Frontiers in Human Neuroscience, 6(22), 1-8.

207
Hargreaves, I.S. & Pexman, P.M. (2014). Get rich quick: The signal to respond procedure
reveals the time course of semantic richness effects during visual word recognition.
Cognition, 131(2), 216-242.
Harnad, S. (2003). Symbol grounding problem. In, Encyclopedia of Cognitive Science. Nature
Publishing Group: Macmillan.
Hauk, O. (2016). Only time will tell – why temporal information is essential for our
neuroscientific understanding of semantics. Psychonomic Bulletin & Review, 23, 10721079.
Heinze, H., Muente, T., & Kutas, M. (1998). Context effects in a category verification
task as assessed by event-related brain potential (ERP) measures. Biological
Psychology, 47, 121-135.
Hino, Y., Kusunose, Y., Lupker, S.J., & Jared, D. (2013). The processing advantage and
disadvantage for homophones in lexical decision tasks. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 39(2), 529-551.
Holcomb, P. J., Kounios, J., Anderson, J. E., & West, W. C. (1999). Dual-coding, contextavailability, and concreteness effects in sentence comprehension: An electrophysiological
investigation. Journal of Experimental Psychology: Learning, Memory, and Cognition,
25(3), 721-742.
Johnson, M. (2017). Embodied mind, meaning, and reason: How our bodies give rise to
understanding. The University of Chicago Press: Chicago.
Juhasz, B. J., Yap, M. J., Dicke, J., Taylor, S. C., & Gullick, M. M. (2011). Tangible words are
recognized faster: The grounding of meaning in sensory and perceptual systems.
Quarterly Journal of Experimental Psychology, 64, 1683-1691.

208
Kanske, P. & Kotz, S.A. (2007). Concreteness in emotional words: ERP evidence from a
hemifield study. Brain Research, 1148, 138-148.
Kiefer, M., Eun-Jin, S., Helbig, H., & Graf, M. (2011). Tracking the time course of action
priming on object recognition: Evidence for fast and slow influence of action on
perception. Journal of Cognitive Neuroscience, 23(8), 1864-1874.
Kiefer, M., Sim, E-J., Herrnberger, B., Grothe, J., & Hoenig, K. (2008). The sound of concepts:
Four markers for a link between auditory and conceptual brain systems. The Journal of
Neuroscience, 28(47), 12224-12230.
Kotz, S.A. & Paulmann, S. (2007). When emotional prosody and semantic dance cheek to cheek:
ERP evidence. Brain Research, 1151, 107-118.
Kounios, J. (1996). On the continuity of thought and the representation of knowledge:
Electrophysiological and behavioural time-course measures reveal the levels of structure
in semantic memory. Psychonomic Bulletin & Review, 3(3), 265-286.
Kousta, S. T., Vinson, D. P., & Vigliocco, G. (2009). Emotion words, regardless of polarity,
have a processing advantage over neutral words. Cognition, 112(3), 473-481.
Kounios, J., Green, D. L., Payne, L., Fleck, J. I., Grondin, R., & McCrae, K. (2009). Semantic
richness and the activation of concepts in semantic memory: Evidence from event-related
potentials. Brain Research, 1282, 95-102.
Kuchinke, L. & Mueller, C.J. (2019). Are there similarities between emotional and familiaritybased processing in visual word recognition? Journal of Neurolinguistics, 49, 84-92.
Kuperman, V., Estes, Z., Brysbaert, M., & Warriner, A. B. (2014). Emotion and
language: Valence and arousal affect word recognition. Journal of Experimental
Psychology: General, 143(3), 1065-1081.

209
Kuonios, J., & Holcomb, P. J. (1994). Concreteness effects in semantic processing: ERP
evidence supporting dual-coding theory. Journal of Experimental Psychology: Learning,
Memory and Cognition, 20(4), 804-823.
Kutas, M., & Hillyard, S. A. (1980). Reading senseless sentences: Brain potentials reflect
semantic incongruity. Science, 207, 203-205.
Langacker, R. (1987). Foundations of cognitive grammar, Theoretical prerequisites. Vol. 1,
Stanford, CA: Stanford University Press.
Laszlo, S. & Federmeier, K.D. (2014). Never seem to find the time: Evaluating the physiological
time course of visual word recognition with regression analysis of single item ERPs.
Language, Cognition, and Neuroscience, 29(5), 642-661.
Lau, E.F., Phillips, C., & Poeppel, D. (2008). A cortical network for semantic: (de)constructing
the N400. Nature Reviews Neuroscience, 9, 920-933.
Li, D., Song, D. & Wang, T. (2020). Concreteness and imageability and their influences on
Chinese two-character word recognition. Reading & Writing, 33, 1443-1476.
Luck, S. J. (2014). An introduction to the event-related potential technique. (2nd Ed). Cambridge:
MIT Press.
Machery, E. (2016). The amodal brain and the offloading hypothesis. Psychonomic Bulletin &
Review, 23, 1090-1095.
Moffat, M., Siakaluk, P. D., Sidhu, D. M., & Pexman, P. M. (2015) Situated conceptualization
and semantic processing: Effects of emotional experience and context availability in
semantic categorization and naming tasks. Psychonomic Bulletin & Review, 22(2), 408419.

210
Morey, R. D., & Rouder, J. N. (2011). Bayes factor approaches for testing interval null
hypotheses. Psychological Methods, 16, 406-419.
Mosely, R.L., Pulvermüller, F., & Shtyrov, Y. (2013). Sensorimotor semantics on the spot: Brain
activity dissociates between conceptual categories within 150 ms. Scientific Reports, 3,
1928.
Müller, O., & Hagoort, P. (2006). Access to lexical information in language comprehension:
Semantics before syntax. Journal of Cognitive Neuroscience, 18(1), 84-96.
Muraki, E.J., Sidhu, D.M., & Pexman, P.M. (2019). Mapping semantic space: property norms
and semantic richness. Cognitive Processing, 21, 637-649.
Nathoo, F. S., Kilshaw, R. E., & Masson, M. E. (2018). A better (Bayesian) interval estimate for
within-subject designs. Journal of Mathematical Psychology, 89¸ 1-9.
Newcombe, P. I., Campbell, C., Siakaluk, P. D., & Pexman, P. M. (2012). Effects of
emotional and sensorimotor knowledge in semantic processing of concrete and
abstract nouns. Frontiers in Human Neuroscience, 6(275), 1-15.
Niedenthal, P.M., Barsalou, L.W., Ric, F., & Krauth-Gruber, S. (2005). Embodiment in
the acquisition and use of emotion knowledge. In L. Feldman Barrett, P.M.
Niedenthal, & P. Winkielman (Eds.), Emotion and consciousness (pp. 21-50).
New York: Guilford.
Pauligk, S., Kotz, S. A., & Kanske, P. (2019). Differential impact of emotion on semantic
processing of abstract and concrete words: ERP and fMRI evidence. Scientific Reports,
9(14439).
Pecher, D. Perception is a two-way junction: Feedback semantics in word recognition.
Psychonomic Bulletin & Review, 8, 545-551.

211
Pecher, D., Boot, I., & Van Dantzig, S. (2011). Abstract concepts: Sensory-motor ground,
metaphors, and beyond. In Brian Ross (Ed.) The Psychology of Learning and
Motivation, 54, 217-248.
Pecher, D., & Zwaan, R. A. (2005). Introduction to grounding cognition. Grounding Cognition:
The Role of Perception and Action in Memory, Language, and Thinking, 1-7.
Pexman, P. M. (2012). Meaning based influences on visual word recognition. In J. S. Adelman
(Ed.), Visual word recognition. Vol. 2 (pp. 24-43). New York: Psychology Press.
Pexman, P. M., Lupker, S. J., & Hino, Y. (2002). The impact of feedback semantics in visual
word recognition: Number-of-features effects in lexical decision and naming tasks.
Psychonomic Bulletin & Review, 9(3), 542-549.
Pexman, P. M., Holyk, G. G., & Monfils, M-H. (2003). Number-of-features effects and semantic
processing. Memory & Cognition, 31(6), 842-855.
Pexman, P. M., Siakaluk, P. D., Bodner, G. E., & Pope, J. (2008). There are many ways to be
rich: Effects of three measures of semantic richness on visual word recognition.
Psychonomic Bulletin & Review, 15(1), 161-167.
Plaut, D., McClelland, J., Seidenberg, M., & Patterson, K. (1996). Understanding normal and
impaired word reading: Computational principles in quasi-regular domains. Psychological
Review, 10.3, 56-115.
R Core Team (2013). R: A language and environment for statistical computing. R Foundation for
Statistical Computing. Vienna, Austria.
Ralph, M. A. L., Jefferies, E., Patterson, K., & Rogers, T. T. (2017). The neural and
computational basis of semantic cognition. Nature Reviews Neuroscience, 18(1), 42-55.

212
Schact, A. & Sommer, W. (2009). Time course and task dependence of emotion effects in word
processing. Cognitive, Affective, & Behavioral Neuroscience, 9, 28-43
Siakaluk, P. D., Knol, N., & Pexman, P. M. (2014). Effects of emotional experience for abstract
words in the Stroop task. Cognitive Science, 38(8), 1698-1717.
Siakaluk, P.D. Newcombe, P.I., Duffels, B., Li, E., Sidhu, D.M., Yap, M.J., & Pexman, P.M.
(2016). Effects of emotional experience in lexical decision. Frontiers in Psychology:
Cognitive Science, 7(1157).
Siakaluk, P. D., Pexman, P. M., Aguilera, L., Owen, W. J., & Sears, C. R. (2008a). Evidence for
the activation of sensorimotor information during visual word recognition: The bodyobject interaction effect. Cognition, 106, 433-443.
Siakaluk, P. D., Pexman, P. M., Sears, C. R., Wilson, K., Locheed, K., & Owen, W. J.
(2008b). The benefits of sensorimotor knowledge: Body-object interaction
facilitates semantic processing. Cognitive Science, 32(3), 591-605.
Strain, E., Patterson, K., & Seidenberg, M. S. (1995). Semantic effects in single-word naming.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 1140-1154.
Stuellein, N., Radach, R.R., Jacobs, A.M., & Hoffman, M.J. (2016). No one way ticket from
orthography to semantics in recognition memory: N400 and P200 effects of associations.
Brain Research, 1639, 88-98
Taler, V., Kousaie, S., & López Zunini, R. (2013). ERP measures of semantic richness: The case
of multiple senses. Frontiers in Human Neuroscience, 7(5), 1-6.
Tettamanti, M., Buccino, G., Saccuman, M. C., Gallese, V., Danna, M., Scifo, P., Fazio, F.,
Rizzolatti, G., Cappa, S. F., & Perani, D. (2006). Listening to action-related sentences

213
activates fronto-parietal motor circuits. Journal of Cognitive Neuroscience, 17(2), 273281.
Tillotson, S. M., Siakaluk, P. D., & Pexman, P. M. (2008). Body-object interaction ratings for
1,618 monosyllabic nouns. Behavior Research Methods, 40, 1075-1075.
Tomarken, A., Davidson, R. J., Wheeler, R. E., & Kinney, L. (1992). Psychometric
properties of resting anterior EEG asymmetry: Temporal stability and internal
consistency. Psychophysiology, 29, 576-592.
Tomasino, B., & Rumiati, R. I. (2013). Introducing the special topic “the when and why of
sensorimotor processes in conceptual knowledge and abstract concepts.” Frontiers in
Human Neuroscience, 7, 498.
Tousignant, C. & Pexman, P. M. (2012). Flexible recruitment of semantic richness: Context
modulates body-object interaction effects in lexical-semantic processing. Frontiers in
Human Neuroscience, 6, 53.
Taylor, M. J., & Khan, S. C. (2000). Top-down modulation of early selective attention processes
in children. International Journal of Psychophysiology, 37, 135-147.
Van Havermaet, L. R. & Wurm, L. H. (2014). Semantic effects in word recognition are
moderated by body-object interaction. Mental Lexicon, 9, 1-22.
Van Petten, C., Coulson, S., Rubin, S., Plante, E., & Parks, M. (1999). Time course of word
identification and semantic integration in spoken language. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 25(2), 394-417.
Vegara-Martínez, M., Comesaña, M., & Perea, M. (2017). The ERP signature of the contextual
diversity effect in visual word recognition. Cognitive, Affective, & Behavioral
Neuroscience, 17, 461-474.

214
Vigliocco, G., Meteyard, L., Andrews, M., & Kousta, S. (2009). Toward a theory of semantic
representation. Language & Cognition, 1, 219-248.
Vinson, D., Ponari, M., & Vigliocco, G. (2014). How does emotional content affect
lexical processing? Cognition & Emotion, 28(4), 737-746.
Wang, X., Shangguan, C., & Liu, J. (2019). Time course of emotion effects during emotion-label
and emotion-laden word processing. Neuroscience Letters, 699, 1-7.
Wellsby, M., Siakaluk, P. D., Owen, W. J. & Pexman, P. M. (2011). Embodied semantic
processing: The body-object interaction effect in a non-manual task. Language &
Cognition, 3(1), 1-14.
West, W. C., & Holcomb, P. J. (2000). Imaginal, semantic, and surface-level processing of
concrete and abstract words: An electrophysiological investigation. Journal of Cognitive
Neuroscience, 12(6), 1024-1037.
Williams, L. E., & Bargh, J. A. (2008). Experiencing physical warmth promotes interpersonal
warmth. Science, 322(5901), 606-607.
Witt, J. J., Kemmerer, D., Linkenauger, S. A., & Fulham, J. (2010). A functional role for motor
simulation in identifying tools. Psychological Science, 21(9), 1215-1219.
Xu, D., Qin, Y., Tu, S., Yin, H., Wang, T., Yu, C., & Qiu, J. (2013). Differentiation of stages in
joke comprehension: Evidence from an ERP study. International Journal of Psychology,
48(2), 149-157.
Xue, J., Marmolejo-Ramos, F., & Pei, X. (2015). The linguistic context effects on processing of
body-object interaction words: An ERP study on second language learners. Brain
Research, 1613.

215
Yap, M. J., Pexman, P. M., Wellsby, M., Hargreaves, I. S., & Huff, M. J. (2012). An abundance
of riches: Cross-task comparisons of semantic richness effects in visual word recognition.
Frontiers in Human Neuroscience, 6, 72.
Yap, M. J. & Seow, C. S. (2014). The influence of emotion on lexical processing:
Insights from RT distributional analysis. Psychonomic Bulletin & Review, 21,
526-533.
Yap, M. J., Tan, S. E., Pexman, P. M., & Hargreaves, I. S. (2011). Is more always better? Effects
of semantic richness on lexical decision, speeded pronunciation, and semantic
classification. Psychonomic Bulletin & Review, 18, 742-750.
Yap, M. J., Yim, G. Y., & Pexman, P. M. (2015). Semantic richness effects in lexical decision:
The role of feedback. Memory & Cognition, 43, 1148-1167.
Yudes, C., Domínguez, A., Cuetos, F., & de Vega, M. (2016). The time-course of processing of
grammatical class and semantic attributes of words: Dissociation by means of ERP.
Psicológica, 37, 105-126.
Zhang, D., He., W., Wang, T., Luo, W., Zhu, X., Gu, R., Li, H., & Luo, Y. (2014). Three stages
of emotional word processing: An ERP study with rapid serial visual presentation. Social,
Cognitive, and Affective Neuroscience, 12(12), 1897-1903.

216
CHAPTER 5:
Conclusion
In the preceding chapters, grounded cognition hypotheses of conceptual processing were
tested by analyzing behavioural, functional near-infrared spectroscopy (fNIRS), and
electroencephalography (EEG) data. Participants completed a lexical decision task (LDT) and
three semantic categorization tasks (SCTs). All experiments used a go/no-go procedure wherein
participants, when presented with a letter string, pressed the spacebar if the letter string satisfied
that experiment’s decision criterion and did not respond if the letter string did not satisfy the
decision criterion. This procedure was used to allow for the examination of the brain and
behaviour patterns associated with the processing of grounded semantic dimensions of
conceptual knowledge. An inter-trial interval of 15 seconds was used for all experiments. Doing
so ensured that task-related hemodynamic changes could return to baseline between trials and to
strengthen confidence that any behavioural data were not the product of experimentally induced
priming effects. Collectively, these data replicate and extend previous observations in grounded
semantic research, while also applying novel experimental, technological, and statistical methods
to the field.
Behavioural Results
In the LDT it was hypothesized that high frequency words would be responded to more
quickly and accurately than low frequency words. In the imageability and abstract noun SCTs it
was predicted that participants would respond more quickly and accurately to words rated high in
grounded semantic richness (body-object interaction (BOI) and emotion experiences (EE),
respectively) than words rated low on the same dimensions. This is because, for imageable
nouns, it was hypothesized that one’s physical interactions with the words’ referents would be

217
automatically retrieved through grounded activation and simulation during task performance,
because such bodily experiences play an important role in how concrete nouns are learned and
retrieved from memory. For abstract nouns, it was hypothesized that one’s emotion experiences
with the words’ referents would be similarly automatically activated during task performance,
because such emotion experiences play an important role in how abstract nouns are learned and
retrieved from memory. In summary, when reading a word rated high in grounded semantic
richness, it was anticipated that the additional activation of bodily and emotion information
should, in theory, facilitate participants’ efforts to recognize the target word as an imageable or
abstract noun, thereby resulting in shorter response latencies and higher accuracy.
In the concrete noun SCT, it was hypothesized that concrete nouns rated high in EE
would be responded to less quickly and less accurately than those rated low in EE. Grounded
emotion information was hypothesized to be inhibitory in the concrete noun SCT because, unlike
in the abstract noun SCT described above, concrete nouns are thought to be learned and
experienced during memory retrieval as a function of bodily experiences, not emotion
experiences. Therefore, it was anticipated that during the concrete noun SCT, when the grounded
emotion experiences associated with a concrete noun were automatically and implicitly
activated, high levels of EE would be experienced as an abstract noun more than as a concrete
noun. Despite having higher levels of grounded semantic richness, concrete nouns with high
levels of EE would therefore require additional processing to resolve the discrepant emotion
information (i.e., as compared to information indicating it is a concrete word). Consequently, the
additional processing should, in theory, slow participant responses and increase the likelihood of
incorrect responses.

218
The behavioural findings from the three SCT experiments described above support the
hypothesis that grounded dimensions of semantic richness are automatically and implicitly
activated during SCT performance regardless of their task relevance. They also supported the
response latency and accuracy hypotheses and extended previous observations regarding the
relationship between task-relevance and the influence of grounded semantic richness on SCT
performance. The dimensions of grounded semantic richness that were manipulated in these
experiments were not explicitly necessary to successfully complete the experimental tasks. As
such, observing latency and accuracy differences that vary as a function of the manipulated
variable suggests that grounded semantic information is implicitly and automatically analyzed
when retrieving conceptual knowledge. For example, in the imageability SCT, where go-trials
varied in BOI ratings, participants were only required to assess whether they could easily
imagine a presented word’s referent. There was no need to consider whether they had previous
physical interactions with that referent. Participants responded more quickly and accurately to
words rated high in BOI, which strongly suggests that participants were automatically accessing
task-irrelevant bodily information. These data provide evidence that information obtained
through one’s sensory, motor, and emotional experiences with the environment is part of how
people represent and access conceptual knowledge.
In addition to the response latency and accuracy data, these four experiments offer two
further contributions to the field. First, a 15-second inter-trial interval (ITI) is atypically long
compared to conventional SCT experiments. Such a long ITI was chosen to allow for
hemodynamic changes to return to baseline between trials, and it also allows for a unique
observation of SCT performance. In conventional SCTs with comparatively rapid stimulus
presentation (e.g., every 1-3 seconds) it is possible that response time differences may be due, at

219
least in part, to a priming effect. For example, in an abstract noun SCT, participants may be
primed for abstractness by a single trial, and still benefit from such a priming effect when the
next trial starts. As a result, response time differences may be amplified accordingly. However,
by waiting 15 seconds between each trial, it is less likely that this sort of priming would
influence participant responses. Furthermore, the response latency effect found in the concrete
noun SCT is a novel finding. These new patterns of data observed with slow ITIs suggest that
this experimental design is a valid method that provides insight into the nature of human
conceptual processing.
Finally, the analytical methods adapted to these experiments are worthy of note.
Oftentimes, neurophysiological data are segmented, averaged, and subsequently analyzed using
frequentist statistics. In the experiments presented here, high-density Bayesian credible intervals
were used to show when condition differences exist across time, with no data lost to averaging
across pre-established temporal epochs. This technique provides a powerful tool for the analysis
of time course data as it does not suffer the risk of committing a Type I error when calculating a
great number of difference scores. Presenting the full time course of difference waves fit with
high density Bayesian credible intervals allows people to identify the presence, timing, and
strength of condition differences—typically with just a glance at each interval. Rather than
making a binary decision regarding the rejection of a null hypothesis, the analytical methods
adapted in the experiments presented here allow for meaningful discussion of the presence and
nature of condition differences. Future research would increase the validity and accessibility of
neuroimaging analyses by adopting such techniques, particularly in light of the growing evidence
that ERP components do not each reflect distinct singular brain processes.
Near Infrared Spectroscopy

220
fNIRS was employed in this research to investigate activity in the prefrontal cortex (PFC)
due to conceptual processing as operationalized by comparative changes in the concentrations of
oxygenated hemoglobin ([O2Hb]), deoxygenated hemoglobin ([HHb]), total hemoglobin
([THb]), and regional tissue oxygen saturation (Sat%). In the LDT, where the decision criterion
was “Is this a real word?” and all go-trials varied in printed word frequency, visual inspection of
the fNIRS data suggested that task-related activity primarily occurred in the left PFC.
Specifically, approximately 4 seconds after stimulus presentation, decreases of [O2Hb] and Sat%
were evident along with an increase of [HHb] for no-go and low-frequency trials. This
observation was interpreted as a general pattern of effort associated with lexical processing and
served to guide the interpretation of the SCT data.
In the imageable noun SCT, visual inspection of fNIRS data revealed two different
patterns. First, in the left hemisphere, increased Sat% was observed for low-BOI trials, and
decreased Sat% was observed for high-BOI trials. This pattern was interpreted similarly as for
the LDT, namely, that greater levels of semantic processing are evident in the left PFC. Second,
increased [O2Hb] was observed in the right hemisphere for low-BOI and no-go trials. This
pattern was interpreted as increased processing to inhibit automatically generated responses. That
is, after a long ITI, the presentation of a stimulus may automatically generate a behavioural
response code in the brain, and no-go trials would require that response code to be fully
inhibited. Concrete nouns rated low in BOI have less grounded semantic richness, or diagnostic
information (i.e., less experience of physically interacting with the word’s referent), and likely
were at least partially inhibited before participants fully recognized the word as an imageable
concrete noun. This interpretation is supported by the response latency data discussed above.

221
In the abstract noun SCT, visual inspection of the fNIRS data suggested the same two
patterns of data as were observed for the imageable noun SCT. First, bilateral [THb] and [O2Hb]
increases were observed for low-EE trials compared to high-EE trials. This pattern was
interpreted as above: additional processing required for atypical task exemplars evident in the
left PFC, and inhibitory effects no-go and low-EE trials in the right PFC.
However, in the concrete noun SCT, PFC hemodynamic activity was located primarily in
the right hemisphere. Specifically, [O2Hb] increases were observed for all three conditions,
suggesting that processing of concrete nouns may occur in the right hemisphere more than the
left. Additionally, [HHb] decreases were most evident for high EE trials, suggesting additional
inhibitory processing for these atypical concrete nouns. The comparative lack of left PFC activity
may have been due to the comparative ease of this experiment. That is, due to their grounded
nature, concrete nouns may be easier to identify than are abstract nouns.
Across these four experiments, the general pattern of the fNIRS findings appeared to be
the following. The processing of lexical stimuli produces [O2Hb] increases in the left hemisphere
of the PFC. The processing of abstract concepts appears to activate the left hemisphere more than
the right hemisphere whereas, in contrast, the processing of concrete concepts appears to activate
the right more than the left. Right hemisphere increases in [O2Hb] may reflect the inhibition of
automatically generated responses as seen in no-go trials and atypical category exemplar trials.
These patterns are congruent with previous research, thus strengthening confidence in the
interpretations of the patterns described above.
Unfortunately, when condition differences were analyzed via 12-second time course
difference waves plotted with Bayesian high density credible intervals, no statistically significant
differences were evident. That is, these data could not be conclusively interpreted as reflecting

222
differential PFC hemodynamic activity elicited by the variables manipulated in these
experiments. While this was disappointing and unexpected given pilot data collected prior to this
research program, the general pattern of hemodynamic activity in each experiment described
above suggests that fNIRS is sensitive to cognitive processing in the PFC. However, as discussed
more fully in Chapter 3, several reasons exist as to why condition differences were not observed
here. First, it is possible that the experimental manipulations were not challenging enough to
elicit activity in the anterior PFC. Second, given the small size of the effects, it is possible that
these experiments lacked statistical power to capture the effects. Third, it is possible that
operationalizing emotion information as EE instead of valence resulted in an averaging-out of
the effects, as other researchers have reported hemispheric PFC differences for positively and
negatively valenced stimuli. Finally, it is possible that, in collecting fNIRS data concurrently
with EEG data, employing an elastic electrode cap and cohesive tape for ~90-minute sessions led
to optode shifting and subsequent data degradation that interfered with analyses. In light of the
behavioural data discussed above, it can be concluded that the experimental manipulations (i.e.,
manipulations of grounded semantic richness dimensions) were successful. As such, it appears to
be the case that fNIRS recordings from the anterior PFC in these experiments did not capture
these processing differences.
Electroencephalography
EEG was employed in the SCT experiments to test the hypotheses that electrical brain
activity will vary qualitatively and quantitatively as a function of the grounded nature of
conceptual knowledge that an individual is processing. That is, rather than a single amodal
knowledge centre activating to retrieve all general knowledge, different patterns of brain regions
were hypothesized to contribute to conceptual processing as a function of the lived experiences

223
associated with those concepts. Scalp topography was inspected along with the P200, N400, and
P600/N600 deflections.
The amplitude of the P200 component is generally associated with task difficulty (greater
amplitudes for harder trials) and the automatic evaluation of stimuli. Greater amplitudes were
observed in the imageability SCT for low BOI trials, in the abstract noun SCT for low EE trials,
and in the concrete noun SCT for high EE trials. This pattern suggests that these trial types
automatically and immediately required more effort to process. When examining the
topographical pattern of activity across these experiments, however, the P200 was observed
primarily in posterior regions in the imageability SCT, while occurring in left frontal regions in
the abstract and concrete noun SCTs. That this component occurs in different locations in
different experiments but occurs at the same time—and is of greater magnitude compared to
‘hard’ trials in all experiments—suggests that the P200 component is not a singular entity.
Rather, electrical brain activity during this time reflects an automatic level of effort being exerted
to process a presented stimulus in the context of current task demands and that grounded
dimensions of grounded semantic richness are automatically activated and influence task
difficulty.
The N400’s amplitude is often interpreted as reflecting the amount of effort necessary to
integrate stimulus information into current task demands. In the abstract and concrete noun
SCTs, the N400 occurred primarily in left frontal medial regions, with more extreme negativity
for low EE trials. This suggests that the N400 in these experiments reflects the processing of EE
itself, regardless of task relevance, with higher levels of EE producing greater levels of
processing. In the imageability SCT, the N400 was most evident in the frontal right hemisphere
and appeared to occur later than it did in the concrete and abstract noun SCTs. This suggests that

224
the nature of information being contextually analyzed during the N400 varies as a function of
semantic variables being analyzed and that, similar to the P200, the N400 is not a discrete
singular component. Rather, negative deflections occurring around this time appear to reflect the
actual processing of semantic variables themselves.
The later deflections showed condition differences as well. In the imageability SCT,
condition differences were observed in the right parietal region, while occurring in the left
temporal region in the abstract and concrete noun SCTs. Additionally, more pronounced
deflections were observed for irregular stimuli (low BOI nouns, low EE abstract nouns, and high
EE concrete nouns). As the P600 is generally interpreted as reflecting additional processing of
incongruities between a stimulus and current task demands, these data support previous
observations of these deflections in semantics research.
The EEG data recorded in these experiments supports the hypotheses that information
obtained from bodily experiences is part of one’s representation of conceptual knowledge and
that this information is activated automatically during conceptual processing as evidenced in
qualitative and quantitative ERP differences. Although the semantic dimensions of BOI and EE
did not necessarily need to be activated in the SCTs for successful task completion, differences
between high- and low-richness trials were evident in ERP deflections regardless of task
relevance. Amplitude and scalp topography differences in event-related electrical brain activity
support the hypotheses that grounded information is integral to one’s representation and
processing of conceptual knowledge.
Conclusions
The EEG data presented in Chapter 4 demonstrate differing patterns of brain activity as a
function of the automatic activation of task-irrelevant dimensions of grounded semantic richness

225
associated with grounded experience, even when those dimensions are incongruent with current
task demands. Although no statistically significant condition differences were obtained via the
fNIRS measurements reported in Chapter 3, the patterns of activity observed were congruent
with grounded hypotheses. Further research will be needed to validate an event-related fNIRS
methodology when studying semantic processing in the PFC.
Collectively, the four experiments presented in this dissertation support the claim that
conceptual knowledge is at least partly grounded in sensorimotor and emotion experience.
Conceptual knowledge is processed more quickly and accurately when current task demands are
congruent with the automatic reactivation of brain regions generally associated with the
experiences of typical categorical exemplars. Extending beyond this, however, the data presented
here convincingly demonstrate that grounded information obtained through physical, sensory,
and emotional experiences with categorical members is automatically reactivated during
conceptual processing even when doing so interferes with successful and timely task completion.