Neural Correlates of Concreteness
in Semantic Categorization
Penny M. Pexman, Ian S. Hargreaves, Jodi D. Edwards, Luke C. Henry,
and Bradley G. Goodyear
Abstract
& In some contexts, concrete words (CARROT) are recognized
and remembered more readily than abstract words (TRUTH).
This concreteness effect has historically been explained by two
theories of semantic representation: dual-coding [Paivio, A. Dual
coding theory: Retrospect and current status. Canadian Jour-
nal of Psychology, 45, 255–287, 1991] and context-availability
[Schwanenflugel, P. J. Why are abstract concepts hard to un-
derstand? In P. J. Schwanenflugel (Ed.), The psychology of word
meanings (pp. 223–250). Hillsdale, NJ: Erlbaum, 1991]. Past ef-
forts to adjudicate between these theories using functional mag-
netic resonance imaging have produced mixed results. Using
event-related functional magnetic resonance imaging, we re-
examined this issue with a semantic categorization task that
allowed for uniform semantic judgments of concrete and ab-
stract words. The participants were 20 healthy adults. Functional
analyses contrasted activation associated with concrete and ab-
INTRODUCTION
A central issue in cognitive science concerns the manner
in which word meanings are represented in the human
cognitive system. Important insight on this issue is pro-
vided by studies of concreteness effects: concrete words
(e.g., CARROT) are, in certain tasks, recognized and
remembered more readily than abstract words (e.g.,
TRUTH) (e.g., Kroll & Merves, 1986; Marschark & Paivio,
1977; James, 1975). This phenomenon has been ex-
plained by at least two competing theories of seman-
tic representation. According to the dual-coding theory
(Paivio, 1991), concrete words are represented by both
verbal and visual codes, whereas abstract words are rep-
resented by only verbal codes. As a consequence, con-
crete words are better remembered and more quickly
recognized. The dual-coding theory posits that the pro-
cessing of concrete words will be associated with ac-
tivation both in left-lateralized language regions and
in image-processing regions that are typically localized
to the right hemisphere (although see, for instance,
Fiebach & Friederici, 2003, for arguments that right hem-
University of Calgary, Canada
D 2007 Massachusetts Institute of Technology
stract meanings of ambiguous and unambiguous words. Results
showed that for both ambiguous and unambiguous words, ab-
stract meanings were associated with more widespread cortical
activation than concrete meanings in numerous regions associ-
ated with semantic processing, including temporal, parietal, and
frontal cortices. These results are inconsistent with both dual-
coding and context-availability theories, as these theories pro-
pose that the representations of abstract concepts are relatively
impoverished. Our results suggest, instead, that semantic re-
trieval of abstract concepts involves a network of association
areas. We argue that this finding is compatible with a theory
of semantic representation such as Barsalou’s [Barsalou, L. W.
Perceptual symbol systems. Behavioral & Brain Sciences, 22,
577–660, 1999] perceptual symbol systems, whereby concrete
and abstract concepts are represented by similar mechanisms
but with differences in focal content. &
isphere activation is not necessary to support the dual-
coding theory). Processing of abstract words, however,
will be associated predominantly with activation in left
hemisphere language-processing regions, with no re-
cruitment of regions in the right hemisphere.
In contrast, the context-availability theory suggests that
concrete words are more easily recognized and remem-
bered because more contextual information is available in
memory for concrete concepts (Schwanenflugel, 1991).
This contextual information includes the situations and
settings in which concepts are encountered. For ab-
stract concepts, associated contextual information is less
readily available. Support for the context-availability the-
ory is provided by studies showing that the usual con-
creteness effect is eliminated when concrete and abstract
words are presented in supportive sentence contexts
(Schwanenflugel, Harnishfeger, & Stowe, 1988).
Both the dual-coding and context-availability theories
assume that the representations for concrete words in-
volve extra features (image-based coding, available con-
text) that are not present in the representations for
abstract words. As such, neither theory is compatible
with studies reporting category-specific semantic impair-
ments involving abstract and concrete words. That is, for
Journal of Cognitive Neuroscience 19:8, pp. 1407–1419
instance, patients with deep dyslexia tend to show an
enhanced concreteness effect: greater impairment read-
ing abstract words than concrete words (e.g., Katz &
Goodglass, 1990; Coltheart, 1980). In contrast, other pa-
tients exhibit a reverse concreteness effect, with greater
impairment for reading and defining concrete words
than abstract words (Breedin, Saffran, & Coslett, 1994;
Sirigu, Duhamel, & Poncet, 1991; Warrington & Shallice,
1984; Warrington, 1981). This double dissociation sug-
gests that there may be functional and structural dif-
ferences in the neural representation of concrete and
abstract concepts ( Warrington, 1981).
Further, past efforts to determine the neural corre-
lates of the concreteness effect in the healthy brain and
to resolve the theoretical debate between dual-coding
and context-availability theories have produced remark-
ably inconsistent results (see Table 1 for a summary of
the literature that compares activation for concrete and
abstract concepts). A small number of these studies have
claimed that their results are consistent with the dual-
coding theory and are inconsistent with the context-
availability theory. Binder, Westbury, McKiernan, Possing,
and Medler (2005) argued that a left-lateralized network
of brain regions is associated with processing of abstract
concepts, whereas a bilateral network is associated with
processing of concrete concepts. Similarly, Sabsevitz,
Medler, Seidenberg, and Binder (2005) reported more
extensive right hemisphere activation for concrete con-
cepts than for abstract concepts. Sabsevitz et al. (2005)
argued that their results provided evidence against the
context-availability theory because they observed greater
activation for abstract words than for concrete words in
several left-lateralized regions.
The results of other studies have been taken as sup-
port for both dual-coding and context-availability theo-
ries. Fiebach and Friederici (2003) reported that concrete
words generated more activation in the left basal tem-
poral lobe than did abstract words. They argued that this
region has been associated with mental imagery (e.g.,
Farah, 2000), and thus, activation here supports the dual-
coding theory. Further, Fiebach and Friederici reported
that abstract words generated more activation in the left
inferior frontal gyrus, an area associated with strategic
selection of semantic knowledge (e.g., Thompson-Schill,
D’Esposito, Aguirre, & Farah, 1997). This finding was
taken as support for the context-availability theory.
The results of several other studies involve a pattern
whereby no neural regions showed greater activation
for concrete than for abstract words (e.g., Noppeney
& Price, 2004; Grossman et al., 2002; Kiehl et al., 1999;
Perani et al., 1999). Instead, abstract words were asso-
ciated with greater activation in several regions. These
results are not consistent with either dual-coding or
context-availability theories because neither of these
predict that abstract words should be associated with
greater activation than concrete words. Indeed, both
dual-coding and context-availability theories assume that
1408 Journal of Cognitive Neuroscience
semantic representations for abstract words are, in some
way, impoverished (lacking image-based information,
lacking available context). Yet the finding that abstract
words generate greater activation in certain regions sug-
gests that a different approach may be required to
capture activity associated with the semantic represen-
tations for abstract words. Kiehl et al. suggested that
abstract words might be processed by a neural pathway
in the right hemisphere and concrete words primarily
in the left hemisphere (this is essentially opposite to
the claims made by Binder, Westbury, et al., 2005).
Noppeney and Price reasoned that their results could
support the view that concrete and abstract concepts
are associated with separate neural substrates for repre-
sentation and/or for retrieval. Grossman et al. noted that
abstract concepts differ from concrete concepts in that
abstract concepts are not associated with a stable set of
perceptual features. Also, the meanings of abstract con-
cepts are captured through complex associations with
other concepts (see also Crutch & Warrington, 2005).
As such, Grossman et al. argued that the features of
abstract concepts ‘‘have a distributed neural representa-
tion and these features must be gathered together and
integrated in order to establish a coherent concept un-
derlying a word’’ (p. 945). Grossman et al. further sug-
gested that this neural integration process is associated
with activity predominantly in association cortices.
One source of the variability observed in previous
studies could be the tasks used (see Table 1 for exam-
ples). Although a certain degree of semantic coding
must be involved in each of these tasks, the processing
required to make a response varies considerably. This
may account for some of the differences in the patterns
of activation observed across these studies. It has been
suggested that the lexical decision task (LDT) can be
performed on the basis of visual (orthographic) familiar-
ity of the stimuli, with only superficial semantic pro-
cessing (Balota, Paul, & Spieler, 1999; Balota, Ferraro, &
Connor, 1991). As such, it may not be appropriate to
draw strong conclusions about semantic representation
on the basis of LDT performance. Further, in many of
the semantic decision tasks that have been used, two
or three words are presented on each trial, making it
difficult to determine what these tasks tell us about
semantic processing of single-word stimuli. In the pres-
ent study, we designed a task that required uniform
semantic judgments of single words. We presented both
concrete and abstract words as stimuli in this task.
A second variable that was uncontrolled in all of the
previous studies on the representation of concrete and
abstract words is semantic ambiguity. Most words are am-
biguous. Truly unambiguous words, or words that have
only one referent, are actually quite rare, yet the stimu-
li used in some of the previous studies (e.g., Binder,
Westbury, et al., 2005) appear to be largely unambigu-
ous. If there are differences in the neural correlates of
concreteness for semantically ambiguous words and
Volume 19, Number 8
 Table 1. Authors, Tasks, and Peak Coordinates for Reviewed Studies

Concrete > Abstract Abstract > Concrete

Authors Task Region Coordinates (x, y, z) Region Coordinates (x, y, z)

Binder, Westbury, et al. (2005) Lexical Decision L angular gyrus 37, 74, 26 L precentral gyrus 48, 9, 25
R angular gyrus 52, 58, 22 48, 7, 40
54, 48, 33 L inferior frontal gyrus 46, 18, 4
R middle temporal gyrus 49, 49, 14 39, 15, 14
L middle frontal gyrus 28, 25, 48 35, 27, 7
38, 19, 42 L inferior frontal sulcus 48, 22, 17
L posterior cingulate gyrus 5, 35, 38 47, 33, 8
9, 45, 13 L superior temporal gyrus 44, 12, 16
7, 37, 36 54, 4, 9
R posterior cingulate gyrus 5, 35, 38
L precuneus 12, 62, 24
3, 74, 31
R precuneus 6, 68, 30
11, 54, 35
3, 62, 41
Sabsevitz et al. (2005) Semantic Similarity Decision L parahippocampal gyrus 27, 22, 20 L superior temporal gyrus 49, 6, 14
25, 37, 12 L superior temporal sulcus 46, 29, 3
L inferior temporal gyrus 57, 49, 14 63, 51, 12
L fusiform gyrus 45, 52, 15 L middle temporal gyrus 59, 47, 3
R hippocampus/amygdala 21, 5, 14 R middle temporal gyrus/sulcus 46, 9, 17
R hippocampus 25, 14, 18 47, 13, 9
R parahippocampal gyrus 26, 26, 16 47, 1, 16
L angular/superior occipital gyrus 28, 79, 36 L inferior frontal gyrus 43, 22, 5
31, 66, 31 50, 15, 9
38, 73, 41 L superior frontal gyrus 9, 49, 33
Pexman et al.

R angular gyrus 36, 60, 32

42, 69, 31
R angular/supramarginal gyrus 45, 47, 42
L inferior/orbital frontal gyrus 22, 26, 9
1409

L inferior/middle frontal gyrus 42, 39, 12

 Table 1. (continued )
1410

Concrete > Abstract Abstract > Concrete

Authors Task Region Coordinates (x, y, z) Region Coordinates (x, y, z)

Journal of Cognitive Neuroscience

L middle frontal gyrus 37, 27, 19

L superior frontal gyrus 19, 8, 52
L subcallosal gyrus 11, 18, 12
R inferior/orbital frontal gyrus 27, 26, 7
37, 35, 8
R middle frontal gyrus 28, 10, 46
45, 31, 14
43, 45, 6
L posterior cingulate/isthmus 14, 54, 15
R posterior cingulate/isthmus 6, 52, 9
1, 62, 26
Fiebach and Friederici (2003) Lexical Decision L basal temporal 27, 41, 4 L inferior frontal gyrus 46, 23, 7
Grossman et al. (2002) Pleasantness Decision – L posterolateral temporal 60, 32, 12
52, 68, 4
L prefrontal 24, 44, 12
R medial frontal 16, 36, 4
R posterolateral temporal 56, 32, 16
Kiehl et al. (1999) Lexical Decision – R superior temporal gyrus 56, 11, 0
Noppeney and Price (2004) Synonymy Decision – L inferior frontal gyrus 54, 21, 6
L anterior temporal pole 51, 18, 27
51, 9, 24
L middle temporal gyrus/ 60, 42, 6
superior temporal gyrus
Perani et al. (1999) Lexical Decision – L inferior frontal gyrus 44, 14, 4
L superior temporal gyrus 58, 8, 16
Volume 19, Number 8

R inferior frontal gyrus 52, 20, 12

R temporal pole 42, 16, 36
R parieto-occipital junction 40, 70, 36
R anterior cingulate gyrus 6, 16, 40
R amygdala 30, 4, 8

L = left hemisphere; R = right hemisphere.

semantically unambiguous words, this could be another
source of variability in the results of previous studies
investigating concreteness effects. In the present study,
our goal was to control this variability, in order to pro-
vide new insight about the representation of concrete
and abstract word meanings, by contrasting patterns of
neural activation produced when concrete and abstract
meanings are generated by ambiguous and unambigu-
ous words.
The task used in the present study was a semantic
categorization task and the decision category we se-
lected was ‘‘consumable’’ (i.e., is it edible/drinkable or
not?). Our interest was in activation associated with
semantic processing of concrete and abstract meanings
and not in the activation associated with complicated
decision-making strategies. The consumable decision
category seemed an appropriate choice because it was
not so novel or obscure that participants would need to
consider each stimulus word in an analytic way or reason
extensively about the stimuli in order to respond (Hino,
Pexman, & Lupker, 2006; Medin, Goldstone, & Gentner,
1993). Also, this decision category was broad enough to
allow a large number of stimuli to be presented, thus
increasing the power of our design.
METHODS
Participants
Participants were 20 healthy adults (12 women and 8
men, mean age 26.5 years, SD = 4.5) who were paid for
participation. This study was reviewed and approved by
the University of Calgary Research Ethics Board and all
participants gave written informed consent prior to im-
aging. All participants were right-handed, had normal or
corrected-to-normal vision, and were monolingual native
English speakers. None of the participants reported any
history of psychological or developmental disorders, neu-
rological impairments or trauma, and none were taking
any prescription medications.
Materials
Our goal was to select four categories of word stimuli to
be presented in the semantic categorization task: (1) am-
biguous words with two concrete meanings (concrete–
concrete ambiguous), (2) ambiguous words with two
abstract meanings (abstract–abstract ambiguous), (3) un-
ambiguous words with one concrete meaning (concrete
unambiguous), and (4) unambiguous words with one
abstract meaning (abstract unambiguous). The stimulus
selection process is described next.
In order to develop categories of ambiguous words
whose two dominant meanings varied in concreteness,
two raters independently categorized all 387 ambiguous
words from the Gilhooly and Logie (1980) norms. Each
word was rated in terms whether it was: (1) a true homo-
nym (i.e., two distinct, unrelated meanings) or a poly-
semous word (two related meanings); (2) inadmissible
due to being a proper name (e.g., JACK), a heterophone
(e.g., LEAD), or a colloquialism; and crucially, (3) wheth-
er the two dominant meanings of the ambiguous words
could be categorized as either concrete–concrete (e.g.,
EARTH), concrete–abstract (e.g., GRAVE), or abstract–
abstract (e.g., REASON). In cases where an ambiguous
word had more than two meanings, the raters made
their categorization decision based upon the two dom-
inant meanings of the concept (as indexed by the pro-
duction frequency of each meaning in the Gilhooly and
Logie norms). The results of these independent cate-
gorizations were compared. In cases of disagreement, a
third rater provided arbitration.
The rating process thus yielded three categories of
ambiguous stimuli (concrete–concrete items, concrete–
abstract items, and abstract–abstract items). Membership
within these categories was then further constrained by
incorporating more stringent categorical definitions using
the mean concreteness ratings provided in Gilhooly and
Logie (1980). Each concept was assigned a concrete-
ness difference score reflecting the absolute difference
in mean rated concreteness between the dominant and
subordinate meanings. For the concrete–concrete and
abstract–abstract categories, it was important that domi-
nant and subordinate meanings be relatively close in con-
creteness. Therefore, all items with an absolute difference
score greater than or equal to 1 standard deviation (M =
0.68, SD = 0.48 and M = 0.62, SD = 0.48, respectively)
were excluded from the concrete–concrete ambiguous
and abstract–abstract ambiguous categories. These ex-
cluded items had one concrete and one abstract mean-
ing, and we included a set of 36 of these items as fillers
in the experiment.
In selecting stimuli, the objective was a stimulus set
that was large enough to allow an event-related design
(as many words in each category as possible), while at
the same time controlling as many lexical and semantic
factors as possible across word categories. We decided
that the most important variable to be equated across
word categories was consumability. The stimulus cate-
gory items were to be presented on the ‘‘no’’ trials in
the semantic categorization task and filler items were to
be presented on the ‘‘yes’’ trials in the semantic cate-
gorization task.1 The ‘‘no’’ trial items should have low
consumability ratings and the ‘‘yes’’ trial items should
have high consumability ratings. Also, across categories,
the stimuli should be equally typical nonmembers of the
consumable decision category; that is, there might be
a tendency for concrete items to be perceived as some-
what more consumable than abstract items (one can
always try to eat concrete things, but cannot even try to
eat abstract things), and by collecting a priori ratings,
we were able to select items in a way that limited this
bias. As such, a separate group of 16 participants was
asked to complete a ratings task using a 6-point scale
Pexman et al. 1411
(where 1 = ‘‘not at all representative of the category’’
and 6 = ‘‘highly representative of the category’’) and
were directed to rate all stimuli in terms of the degree to
which each concept was consumable. These participants
made ratings for the stimulus category items and also for
the filler items. We also considered it important to en-
sure that ambiguous and unambiguous words were, in
fact, perceived as such. An additional group of 20 par-
ticipants was asked to rate all items (the ‘‘no’’ trials
stimuli, plus 27 nonwords) in terms of how many mean-
ings each stimulus had (using a scale from 0 to 2, where
0 = no meanings, 1 = one meaning, and 2 = two or
more meanings; Hino & Lupker, 1996). The results of
these ratings tasks are reported by category in Table 2.
In the stimulus selection process, we next considered
the relative dominance of the ambiguous words’ mean-
ings. That is, ambiguous words were selected such that
the categories of ambiguous stimuli had equivalent
meaning dominance, as measured by the production
frequency of the words’ dominant meanings. The mean
production frequency of the dominant meaning of items
in the concrete–concrete ambiguous category was 18.13
(SD = 9.54) and for items in the abstract–abstract am-
biguous category was 17.38 (SD = 8.08) (Gilhooly &
Logie, 1980). Finally, we matched the stimulus catego-
ries as much as possible for printed frequency (Kučera &
Francis, 1967). This matching was possible for the ambig-
uous word categories but was more difficult to achieve
for the unambiguous word categories. We wanted to use
items from the Binder, Westbury, et al. (2005) study in
our unambiguous stimulus categories, as this would fa-
cilitate comparison to the present study. The abstract
and concrete items used in the Binder, Westbury, et al.
study differed somewhat in mean frequency, according
to the frequency measure we used. We replaced some of
the Binder, Westbury, et al. items in order to mitigate
this frequency mismatch. As a result, 29 of the 36 con-
crete unambiguous items were from Binder, Westbury,
et al. and 7 were from the MRC Database (Wilson, 1988;
Coltheart, 1981), 32 of the 36 abstract unambiguous
items were from Binder, Westbury, et al. and 4 were
from the MRC Database. The small frequency differences
that remained were not significant, that is, the mean fre-
quencies of the concrete–concrete ambiguous category
and the abstract–abstract ambiguous category were not
significantly different [t(70) = 0.31, p = .76], and the
mean frequencies of the concrete unambiguous cate-
gory and the abstract unambiguous category were not
significantly different [t(70) = 1.27, p = .20]. Following
the stimulus selection process, there were a total of 180
‘‘no’’ trial stimuli (nonconsumables; see Appendix A).
There were also 150 ‘‘yes’’ trial stimuli (consumables).
Acquisition of Behavioral Data
All stimuli were presented to participants in the MR scan-
ner using a rear-mounted LED projector display system
(Avotec, Stuart, FL). Trial sequences were generated and
presented using Presentation (Neurobehavioral Systems,
Albany, CA), running on an Intel Pentium III 700-MHz per-
sonal computer located outside the magnet room. Seman-
tic categorization responses and response latencies were
recorded from a two-button MR-compatible StimSelect
response pad (Compumedics Neuroscan, El Paso, TX).
Participants first completed 16 practice trials and were
given verbal feedback if they responded incorrectly to
any of the practice items. Using an event-related func-
tional magnetic resonance imaging (fMRI) study design,
stimuli were presented for 1000 msec with a randomized
intertrial interval (ITI) of 4000 ± 2000 msec. This vari-
able ITI was used in order to sample different points
along the hemodynamic response function during image

Table 2. Mean Characteristics (Standard Deviations in Parentheses) for Stimulus Categories

Concrete–Concrete Abstract–Abstract Concrete Abstract
Concrete Abstract Ambiguous Ambiguous Unambiguous Unambiguous
Example EARTH, CIRCUS REASON, GUESS EARTH REASON CIRCUS GUESS
Number of stimuli 72 72 36 36 36 36
Mean concreteness 5.56 (0.43) 3.42 (0.41) 5.39 (0.46) 3.82 (0.43) 5.73 (0.47) 3.02 (0.49)
(Gilhooly &
Logie, 1980)
Rated consumability 1.41 (0.18) 1.09 (0.12) 1.39 (0.20) 1.06 (0.11) 1.44 (0.22) 1.11 (0.17)
Rated ambiguity 1.39 (0.16) 1.30 (0.20) 1.64 (0.17) 1.48 (0.24) 1.14 (0.20) 1.12 (0.18)
Printed frequency 49.01 (60.30) 57.38 (66.07) 69.91 (69.36) 73.19 (80.19) 28.11 (39.77) 39.39 (40.05)
(Kučera &
Francis, 1967)
Syllabic length 1.63 (0.68) 1.92 (0.85) 1.33 (0.47) 1.81 (0.92) 1.91 (0.73) 2.02 (0.77)
Printed length 5.72 (1.54) 5.97 (1.38) 5.28 (1.28) 6.25 (1.57) 6.17 (1.58) 5.69 (1.14)

1412 Journal of Cognitive Neuroscience Volume 19, Number 8

acquisition and allow greater detectability of responses
in the signal (Birn, Cox, & Bandettini, 2002). All partic-
ipants completed four runs of the semantic categoriza-
tion task. Three of the four runs contained 82 trials (45
‘‘no’’ trials and 37 ‘‘yes’’ trials), with a total scan dura-
tion of 7 min 4 sec, and one run contained 84 trials (45
‘‘no’’ trials and 39 ‘‘yes’’ trials), with a total scan dura-
tion of 7 min 13 sec. The order of the runs was random-
ized across participants.
A trial was considered an error and was excluded from
the latency analysis if the categorization latency was lon-
ger than 2500 msec or shorter than 250 msec, or if the
categorization response was incorrect. Mean latencies
and error percentages for the behavioral data are pre-
sented in Table 3.
Acquisition of fMRI Data
Images were acquired using a 3-Tesla General Electric
MR scanner, equipped with an eight-channel phased ar-
ray head coil (Signa Excite; GE Healthcare, Waukesha,
WI). The MR sequence for functional imaging was a two-
shot gradient-recalled echo-planar imaging (EPI) T2*-
weighted sequence, with whole head coverage (96 
96 matrix, zero-filled to 128  128, FOV = 24 cm, TE =
30 msec, TR = 1500 msec, flip angle = 60, 24 oblique/
axial slices, 5 mm thick). High-resolution T1-weighted
images (0.94  0.94  2.00 mm) were collected using
a 3-D inversion recovery-prepped anatomical MRI se-
quence to anatomically register the functional data.
Image analyses were carried out using FEAT (FMRI Ex-
pert Analysis Tool) Version 5.4, part of FSL (FMRIB’s
Software Library, www.fmrib.ox.ac.uk/fsl). Images were cor-
rected for head motion during postprocessing using the
intramodal motion correction tool MCFLIRT ( Jenkinson,
Bannister, Brady, & Smith, 2002; Jenkinson & Smith,
2001). Prior to image analysis, image data were recon-
structed and subjected to high-pass temporal filtering
(Gaussian-weighted LSF straight line fitting, with sigma—
25.0 sec). The following prestatistics processing was ap-
plied: slice-timing correction using Fourier-space time-
series phase-shifting; nonbrain removal using BET (Brain
Table 3. Mean Semantic Categorization Response Latencies
and Error Percentages (Standard Deviations in Parentheses)
for Stimulus Categories
Stimulus Category Latency (msec) Error (%)
Concrete 674.19 (52.82) 0.84 (1.25)
Abstract 651.56 (40.40) 0.35 (1.30)
Concrete–concrete ambiguous 672.27 (82.73) 0.42 (1.02)
Abstract–abstract ambiguous 650.28 (73.76) 0.00 (0.00)
Concrete unambiguous 676.17 (70.67) 1.25 (1.68)
Abstract unambiguous 652.05 (89.03) 0.70 (1.53)
Extraction Tool; Smith, 2002); spatial smoothing using a
Gaussian kernel of full-width half-maximum 6 mm; and
mean-based intensity normalization of all volumes by the
same factor. For all runs and participants, time-series
statistical analyses for each stimulus category were then
carried out using FILM (FMRIB’s Improved Linear Model)
with local autocorrelation correction (Woolrich, Ripley,
Brady, & Smith, 2001). Registration to high-resolution
images was carried out using FLIRT (FMRIB’s Linear Im-
age Registration Tool; Jenkinson et al., 2002; Jenkinson
& Smith, 2001). Higher-level contrasts of planned com-
parisons of the stimuli in the critical categories (‘‘no’’
trials) across participants were carried out using FLAME
(FMRIB’s Local Analysis of Mixed Effects) Stage 1 only
(i.e., without the final MCMC-based stage) (Woolrich,
Behrens, Beckmann, Jenkinson, & Smith, 2004; Beckmann,
Jenkinson, & Smith, 2003). These comparisons included:
(1) both categories of concrete stimuli (concrete–concrete
ambiguous and concrete unambiguous) versus both cate-
gories of abstract stimuli (abstract–abstract ambiguous and
abstract unambiguous); (2) concrete–concrete ambiguous
versus abstract–abstract ambiguous; and (3) concrete un-
ambiguous versus abstract unambiguous. In each of these
comparisons, response latencies were included as cova-
riates in order to statistically remove the effects of latency
differences from comparisons. Z (Gaussianized T/F) sta-
tistic images were thresholded at p = .05 with a cluster
size of at least 12 contiguous voxels, resulting in a cor-
rected p value of .01 (Forman et al., 1995). Registration to
standard space (Talairach & Tournoux, 1988) was carried
out using FLIRT ( Jenkinson et al., 2002; Jenkinson &
Smith, 2001).
RESULTS AND DISCUSSION
Behavioral Data
Response latencies for the semantic categorization task
were examined to test for effects of concreteness. There
were too few response errors to warrant analyses of er-
ror data. Response latencies were analyzed in compar-
isons with subjects (t1) and, separately, items (t2) treated
as random factors. Overall, semantic categorization la-
tencies were faster to abstract words than to concrete
words [t1(19) = 4.23, p < .001; t2(142) = 2.88, p <
.005].2 When ambiguous and unambiguous stimuli were
considered separately, the same effects were observ-
ed. Response latencies were significantly faster for the
abstract–abstract ambiguous stimulus category than
for the concrete–concrete ambiguous stimulus category
[t1(19) = 2.77, p < .05; t2(70) = 2.23, p < .05]. Similarly,
response latencies were significantly faster for the ab-
stract unambiguous stimulus category than for the con-
crete unambiguous stimulus category [t1(19) = 3.35,
p < .005; t2(70) = 1.96, p = .05]. In the context of this
task, these reverse concreteness effects likely reflect
a modest decision bias. That is, despite our efforts to
Pexman et al. 1413
match concrete and abstract word categories for rated
consumability, faster response latencies to words with
abstract meanings suggest that these words were still
perceived as somewhat less consumable than were
words with concrete meanings. As mentioned, in the
analyses of imaging data (reported next), we statistically
controlled for these latency differences.
fMRI Data
For all contrasts, significant regions of activation, z scores,
and corresponding Talairach coordinates are presented in
Table 4.
Planned contrasts testing the overall effect of concrete-
ness (collapsing across ambiguous and unambiguous
word categories) revealed that the stimuli with abstract
meanings were associated with significantly greater ac-
tivity compared to stimuli with concrete meanings in an
extensive network of regions including the left middle
and superior temporal gyri, the left inferior frontal gyrus
and left inferior frontal lobule, the left medial frontal
gyrus, the left precentral gyrus, the left superior parietal
lobule, the left inferior occipital gyrus, the posterior cin-
gulate cortex, the right inferior temporal gyrus, the right
supramarginal gyrus, the right inferior parietal lobule,
the right postcentral gyrus, and bilaterally in the middle
frontal gyrus and middle temporal gyrus (see Figure 1).
The reverse contrast showed no significant activity.
Similarly, planned contrasts testing the effect of
concreteness for ambiguous stimuli (abstract–abstract

Table 4. Areas of Significant Activation in Contrasts between Stimulus Categories

Contrast Region(s) of Activation Brodmann’s Area (BA) Z Score Talairach Coordinates (x, y, z)
Abstract > Concrete Left Superior Parietal Lobule 7 3.6 (3.3) 30, 58, 58
Right Inferior Parietal Lobule 40 3.5 (3.0) 38, 44, 50
Right Postcentral Gyrus 40 3.1 (2.7) 54, 30, 48
Left Inferior Frontal Lobule 7 2.9 (2.7) 42, 66, 46
Left Superior Frontal Gyrus 8 3.1 20, 26, 44
Left Middle Frontal Gyrus 6 2.7 (2.7) 44, 6, 42
Right Middle Frontal Gyrus 6 2.7 (2.7) 40, 14, 42
Right Supramarginal Gyrus 40 3.5 (2.9) 46, 42, 36
Left Precuneus 7 [2.7] 22, 64, 30
Left Middle Temporal Gyrus 39 3.5 (3.2) 52, 66, 24
Right Middle Temporal Gyrus 39 3.2 (3.1) 40, 72, 26
Posterior Cingulate 31 3.4 0, 52, 26
Medial Frontal Gyrus 10 (3.2) 6, 48, 16
Left Precentral Gyrus 43 3.2 58, 6, 10
Left Inferior Frontal Gyrus 45 3.6 56, 20, 6
Left Superior Temporal Gyrus 39 [2.6] 48, 52, 6
Medial Frontal Gyrus 10 (3.0) 2, 54, 4
Left Superior Temporal Gyrus 13 3.2 (3.1) 56, 36, 6
Left Superior Temporal Gyrus 38 3.7 (3.4) 48, 16, 6
Left Middle Temporal Gyrus 21 4.4 (3.7) 56, 36, 6
Left Inferior Occipital Gyrus 19 3.3 [2.9] 46, 86, 6
Right Inferior Temporal Gyrus 37 3.2 (2.9) 66, 46, 8
Right Inferior Temporal Gyrus 21 2.7 60, 10, 14
Left Medial Frontal Gyrus 11 2.8 [2.7] 2, 54, 14
Left Middle Temporal Gyrus 21 3.3 52, 2, 32
Concrete > Abstract Not significant

Z scores in parentheses are those for contrasts involving ambiguous word categories (abstract–abstract ambiguous vs. concrete–concrete ambig-
uous) and Z scores in brackets are those for contrasts involving unambiguous word categories (abstract unambiguous vs. concrete unambiguous).

1414 Journal of Cognitive Neuroscience Volume 19, Number 8

Figure 1. Contrast map
showing areas of significantly
greater activity for the
abstract stimulus conditions
compared to the concrete
stimulus conditions.

Figure 2. Contrast map

showing areas of significantly
greater activity for the
abstract–abstract ambiguous
stimulus condition compared
to the concrete–concrete
ambiguous stimulus condition.

Figure 3. Contrast map

showing areas of significantly
greater activity for the abstract
unambiguous stimulus
condition compared to the
concrete unambiguous
stimulus condition.

Pexman et al. 1415

ambiguous stimuli vs. concrete–concrete ambiguous stimu-
li) indicated that the ambiguous stimuli with abstract
meanings showed significantly greater activity compared
to the ambiguous stimuli with concrete meanings in an
extensive network of regions, largely the same regions
reported in the contrast above. In addition, this contrast
produced significant activation in the medial frontal gyrus
(see Figure 2). The reverse contrast showed no signifi-
cant activity.
The parallel contrast with the unambiguous stimulus
categories (abstract unambiguous stimuli vs. concrete un-
ambiguous stimuli) revealed significantly greater activation
for abstract stimuli compared to concrete stimuli in sev-
eral regions: left precuneus, left superior temporal gyrus,
left medial frontal gyrus, and left inferior occipital gyrus
(see Figure 3). The reverse contrast was not significant.
Thus, for both ambiguous and unambiguous stimuli,
abstract meanings were associated with greater activation
than were concrete meanings. This result is inconsistent
with both dual-coding and context-availability theories of
semantic representation because both of those theories
suggest that the representations for abstract meanings are
impoverished relative to those of concrete meanings. In-
stead, the present results suggest that the meanings of
abstract concepts recruit a network of cortical regions
that is bilateral in the case of ambiguous stimuli and more
left lateralized in the case of unambiguous stimuli. Many
of these regions have previously been implicated in se-
mantic processing. For instance, left middle and superior
temporal gyri have been implicated in semantic retrieval
(Martin & Chao, 2001; Perani et al., 1999; Pugh et al.,
1996). Medial frontal activation has previously been asso-
ciated with semantic processing (Harris et al., 2006),
especially the process of inhibiting inappropriate seman-
tic features during semantic search (Chou et al., 2006).
It has been argued that parietal regions may be the as-
sociation cortex, involved in integration of processes
like attention, spatial representation, and working mem-
ory (Culham & Kanwisher, 2001). Further, Chou et al.
(2006; also Koenig et al., 2005) proposed that semantic
integration processes can be localized to the inferior
parietal lobule. Activation in the posterior cingulate and
precuneus region, seen here with abstract word mean-
ings, has been tied to associative semantics (Noppeney
& Price, 2003; Cappa, Perani, Schnur, Tettamanti, & Fazio,
1998).
The supramarginal gyrus has been implicated in lexical–
semantic processing (Binder, Medler, Desai, Conant, &
Liebenthal, 2005), and our results thus suggest that ex-
tensive lexical–semantic processing was associated with
the abstract–abstract ambiguous stimuli. Ruby and Decety
(2004) argued that the right postcentral gyrus and the
inferior parietal lobule are part of a network of cortical
areas (other areas in the network include the frontopolar
cortex, portions of which were also active in these con-
trasts) that participate in semantic processing, specifically
cognition about the self/other distinction.
1416 Journal of Cognitive Neuroscience
Notably, the activation observed for abstract stimuli in
inferior occipital and inferior temporal regions suggests
that visual processing was recruited to make semantic
decisions about abstract concepts. One explanation
for this result is the notion that semantic processing
is grounded in the perceptual system (e.g., Sim & Kiefer,
2005; Barsalou, Simmons, Barbey, & Wilson, 1999). We
consider this possibility more extensively in the latter
part of the following section.
Conclusions
In the present study, behavioral results showed that
response latencies were somewhat faster for abstract
than for concrete words. This was true for unambiguous
stimuli (with only one concrete or one abstract mean-
ing) and also for ambiguous stimuli (with two concrete
meanings or two abstract meanings). We attributed this
latency difference to the fact that the abstract stimuli
were perceived to be slightly less consumable, hence,
were easier to classify as ‘‘nonconsumables’’ in our se-
mantic categorization task, than were the concrete stimu-
li. We statistically controlled for this latency difference
in the analyses of the imaging data. Notably, results of
those analyses indicated that abstract words were as-
sociated with more widespread cortical activation than
were the concrete words. Our results are consistent with
those of a number of other studies showing that there
were no cortical areas with greater activation for con-
crete relative to abstract words (Noppeney & Price,
2004; Grossman et al., 2002; Kiehl et al., 1999; Perani
et al., 1999).
The fact that, in the present study, abstract word mean-
ings were associated with more widespread cortical acti-
vation suggests that abstract concepts are represented in
a network of regions tied to semantic processing, includ-
ing temporal, parietal, and frontal areas. These results are
not consistent with the notion that the representations
of abstract concepts activate less visual semantic informa-
tion than do the representations of concrete concepts,
and thus, do not support the dual-coding theory. These
results are also not consistent with the notion that ab-
stract concepts activate less contextual information, and
thus, do not support the context-availability theory.
Certainly, there are limitations to the present study.
We would not claim that, in controlling for both con-
creteness and ambiguity, we have dealt with all of the
factors that are likely to be important for neural repre-
sentation of semantics. There is evidence, for instance,
that different categories of concrete concepts are repre-
sented in different neural regions (e.g., Kiefer, 2005;
Damasio, Grabowski, Tranel, Hichwa, & Damasio, 1996;
Martin, Wiggs, Ungerleider, & Haxby, 1996). Further, in
our stimulus set, the unambiguous words had lower
mean frequency than the ambiguous words. This mis-
match in mean frequency that is present for the ambig-
uous and unambiguous stimuli in the present study
Volume 19, Number 8
prevents us from drawing strong conclusions about the
impact of ambiguity on cortical activity. The issue of how
ambiguity per se influences neural representation of se-
mantics should be addressed in future research.
Our results suggest that the representations of ab-
stract meanings are not impoverished versions of rep-
resentations for concrete meanings but, instead, are
characterized by more extensive activation in the se-
mantic network. Does any current theory of semantic
representation fit this characterization? Very few cog-
nitive theories have considered representation of ab-
stract meanings. One exception is the Perceptual Symbol
Systems’ view of Barsalou and Wiemer-Hastings (2005)
and Barsalou (1999). Barsalou proposed that semantic
knowledge is grounded in sensorimotor experience:
Knowledge is acquired through sensorimotor experi-
ence and retrieval of that knowledge involves simulation
or partial reenactment of the sensory, motor, and men-
tal states implicated at encoding. Semantic knowledge
is thus represented in terms of simulators. Activation for
a particular concept involves activating a subset of the
knowledge that is represented in a simulator and run-
ning a simulation that reenacts some of what is known
about that concept. This theory is proposed to explain
semantic representation for both concrete and abstract
concepts. Barsalou and Wiemer-Hastings argued that
there are similarities in the representations of concrete
and abstract concepts. Both are represented in terms of
situational content (the settings in which they have been
experienced), but the focal content within those situa-
tions differs. Situational content for concrete concepts is
largely about individual, spatially circumscribed objects,
whereas the situational content for abstract concepts is
distributed across many focal topics, including physical,
mental, and interpersonal events. Barsalou and Wiemer-
Hastings argued that these different types of content all
have sensorimotor properties that can be partially re-
enacted by simulators. The distributed nature of the
focal content for abstract concepts, however, is likely to
lead to more complex representations. That is, because
the meanings of abstract concepts lack a unitary focal
referent, the representations of abstract meanings are
proposed to be more complex and more likely to be
distributed across different types of content: ‘‘abstract
concepts often depend critically on multiple pieces of
information distributed across a situation, [and] com-
plex relations are needed to coordinate them’’ (p. 150).
This account seems a good fit for the results of the
present study. If, relative to concrete concepts, abstract
concepts draw upon a greater number of content areas
in order to establish a coherent meaning, then the great-
er cortical activity observed in a number of semantic
processing regions could result from the activation of
these distributed associations. Barsalou’s account is com-
patible with that offered by Grossman et al. (2002), as
both attribute the differences in representation between
concrete and abstract concepts to the presence or lack
of a stable focal referent. Furthermore, both theories
contend that the lack of stable focal features for abstract
concepts leads to a complicated distributed representa-
tion that requires integration in order for the coherent
activation of meaning.
The Perceptual Symbol Systems account also offers
an explanation as to why the current results differ from
those found by Binder, Westbury, et al. (2005). Our imag-
ing results were different than those reported by Binder,
Westbury, et al., despite the fact that most of the items
within the current concrete unambiguous and abstract
unambiguous categories were taken from the Binder,
Westbury, et al. study. A crucial difference between the
present study and that reported by Binder, Westbury,
et al. is the task used. Barsalou and Wiemer-Hastings
(2005) argued that abstract concepts may generate less se-
mantic activation in tasks that can be performed with only
shallow semantic processing, such as LDT. Abstract con-
cepts are usually instantiated in rich, but highly specific,
guiding contexts (e.g., the abstract concept OPTIMISM
can be instantiated in the context of describing the traits
of a friend, or in assessing the financial future of a cor-
poration). When an abstract concept is presented in iso-
lation, no specific context of instantiation is highlighted,
initiating competitive activity across numerous relevant
contexts. Barsalou and Wiemer-Hastings argued that
the lack of an individual context immediately asserting
dominance for abstract concepts may lead to decisions
based upon the processing of word associations (cued
by the presentation of a word stimulus), not the activa-
tion of semantic representations that constitute abstract
conceptual meaning. Thus, in an LDT, the more focal
semantic representations of concrete concepts are more
likely to be active, providing the standard concreteness
advantage in behavioral data, whereas the shallow pro-
cessing of abstract meanings would lead to limited acti-
vation observed in the imaging data. In tasks that require
more extensive semantic processing, such as semantic
categorization, the representations for abstract concepts
are more likely to be fully activated and activation of
a complex and distributed system of representation
is observed. A number of studies have demonstrated
that the behavioral effects observed in LDT can be very
different than the behavioral effects observed in se-
mantic categorization (e.g., Hino et al., 2006; Piercey
& Joordens, 2000). Given that behavioral results vary
across these tasks, it is perhaps not surprising that im-
aging results also vary. Because the present semantic
categorization task depends more heavily on semantic
processing than does LDT, we would argue that the pres-
ent results may provide a more direct examination of
semantic representation.
The results of the present study provide new insight
about how the meanings of concrete and abstract con-
cepts are represented in the brain. The mixed results
produced in previous studies (see Table 1) seem, in part,
to be attributable to the variety of tasks used in those
Pexman et al. 1417
studies. Our findings suggest that, although ambiguity
might influence the extent to which observed activation
is bilateral, the mixed results produced in previous
studies are likely not attributable to stimulus ambiguity.
That is, regardless of stimulus ambiguity, we observed
that the meanings of abstract concepts invoked more
activation than did the meanings of concrete concepts in
a distributed network of association areas. Further, there
were no cortical areas that were activated more strongly
for concrete concepts than for abstract concepts. This
pattern of activation is best accounted for by a theory
that captures representations for abstract concepts as
complex multimodal associations.
Acknowledgments
This research was supported by a research grant to P. M. P.
from the Natural Science and Engineering Research Council
(NSERC) of Canada.
Reprint requests should be sent to Penny M. Pexman, Depart-
ment of Psychology, University of Calgary, 2500 University Drive
NW, Calgary, AB, Canada T2N 1N4, or via e-mail: pexman@
ucalgary.ca.
Notes
1. The practice of presenting critical items on ‘‘no’’ trials in
semantic tasks has been implemented in several behavioral stud-
ies (e.g., Siakaluk, Pexman, Sears, & Owen, 2007; Hino et al.,
2006; Pexman, Hino, & Lupker, 2004; Klinger & Greenwald,
1995) and analyses suggest that responses to these trials are
sensitive to the same lexical and semantic variables as are the
responses to the ‘‘yes’’ trials.
2. In a standard LDT, there is typically a concreteness ad-
vantage (faster response latencies for concrete words than for
abstract words; e.g., Schwanenflugel & Akin, 1994; James,
1975). In order to determine whether our stimuli produced
the typical effects of concreteness in visual LDT, we collected
mean LDT reaction time and error data for our items from the
English Lexicon Project (Balota et al., 2002; http://elexicon.
wustl.edu). Results included a concreteness advantage, sug-
gesting that our stimuli produce standard effects in LDT.
REFERENCES
Balota, D. A., Cortese, M. J., Hutchison, K. A., Neely, J. H.,
Nelson, D., Simpson, G. B., et al. (2002). The English
Lexicon Project: A web-based repository of descriptive
and behavioral measures for 40,481 English words and
nonwords. Retrieved from http://elexicon.wustl.edu/
on 12 October 2005. Washington University.
Balota, D. A., Ferraro, R. F., & Connor, L. T. (1991). On the early
influence of meaning in word recognition: A review of the
literature. In P. J. Schwanenflugel (Ed.), The psychology of
word meanings (pp. 187–221). Hillsdale, NJ: Erlbaum.
Balota, D. A., Paul, S. T., & Spieler, D. H. (1999).
Attentional control of lexical processing pathways during
word recognition and reading. In S. C. Garrod & M. J.
Pickering (Eds.), Language processing (pp. 15–57).
Hove, UK: Psychology Press.
Barsalou, L. W. (1999). Perceptual symbol systems.
Behavioral & Brain Sciences, 22, 577–660.
1418 Journal of Cognitive Neuroscience
Barsalou, L. W., Simmons, W., Barbey, A. K., & Wilson, C. D.
(1999). Grounding conceptual knowledge in modality-
specific systems. Trends in Cognitive Sciences, 7, 84–91.
Barsalou, L. W., & Wiemer-Hastings, K. (2005). Situating
abstract concepts. In D. Pecher & R. A. Zwaan (Eds.),
Grounding cognition: The role of perception and action
in memory, language, and thinking (pp. 129–163).
Cambridge, UK: Cambridge University Press.
Beckmann, C., Jenkinson, M., & Smith, S. M. (2003).
General multi-level linear modeling for group analysis
in fMRI. Neuroimage, 20, 1052–1063.
Binder, J. R., Medler, D. A., Desai, R., Conant, L. L., &
Liebenthal, E. (2005). Some neurophysiological constraints
on models of word naming. Neuroimage, 27, 677–693.
Binder, J. R., Westbury, C. F., McKiernan, K. A., Possing,
E. T., & Medler, D. A. (2005). Distinct brain systems for
processing concrete and abstract concepts. Journal of
Cognitive Neuroscience, 17, 1–13.
Birn, R. M., Cox, R. W., & Bandettini, P. A. (2002). Detection
versus estimation in event-related fMRI: Choosing the
optimal stimulus timing. Neuroimage, 15, 252–264.
Breedin, S. D., Saffran, E. M., & Coslett, H. B. (1994).
Reversal of the concreteness effect in a patient with
semantic dementia. Cognitive Neuropsychology, 11,
617–660.
Cappa, S. F., Perani, D., Schnur, T., Tettamanti, M., &
Fazio, F. (1998). The effects of semantic category and
knowledge type on lexical–semantic access: A PET Study.
Neuroimage, 8, 350–359.
Chou, T., Booth, J. R., Bitan, T., Burman, D. D., Bigio, J. D.,
Cone, N. E., et al. (2006). Developmental and skill
effects on the neural correlates of semantic processing
to visually presented words. Human Brain Mapping, 27,
915–924.
Coltheart, M. (1980). Deep dyslexia: A review of the
syndrome. In M. Coltheart, K. E. Patterson, & J. C. Marshall
(Eds.), Deep dyslexia (pp. 22–47). London: Routledge
& Kegan Paul.
Coltheart, M. (1981). The MRC psycholinguistic database.
Quarterly Journal of Experimental Psychology, Series A,
Human Experimental Psychology, 33, 497–505.
Crutch, S. J., & Warrington, E. K. (2005). Abstract and concrete
concepts have structurally different representational
frameworks. Brain, 128, 615–627.
Culham, J. C., & Kanwisher, N. (2001). Neuroimaging of
cognitive functions in human parietal cortex. Current
Opinion in Neurobiology, 11, 157–163.
Damasio, H., Grabowski, T. J., Tranel, D., Hichwa, R. D., &
Damasio, A. R. (1996). A neural basis for lexical retrieval.
Nature, 380, 499–505.
Farah, M. J. (2000). The neural bases of mental imagery.
In M. S. Gazzaniga (Ed.), The new cognitive neurosciences
(2nd ed., pp. 965–974). Cambridge: MIT Press.
Fiebach, C. J., & Friederici, A. D. (2003). Processing concrete
words: fMRI evidence against a specific right-hemisphere
involvement. Neuropsychologia, 42, 62–70.
Forman, S. D., Cohen, J. D., Fitzgerald, M., Eddy, W. F.,
Mintun, M. A., & Noll, D. C. (1995). Improved assessment
of significant activation in functional magnetic resonance
imagine (fMRI): Use of a cluster-size threshold. Magnetic
Resonance Medicine, 33, 636–647.
Gilhooly, K. J., & Logie, R. H. (1980). Meaning-dependent
ratings of imagery, age of acquisition, familiarity, and
concreteness for 387 ambiguous words. Behaviour
Research Methods & Instrumentation, 12, 428–450.
Grossman, M., Koenig, P., DeVita, C., Glosser, G., Alsop, D.,
Detre, J., et al. (2002). The neural basis for category-specific
knowledge: An fMRI study. Neuroimage, 15, 936–948.
Volume 19, Number 8
 Harris, G. J., Chabris, C. F., Clark, J., Urban, T., Aharon, I.,
Steele, S., et al. (2006). Brain activation during semantic
processing in autism spectrum disorders via functional
magnetic resonance imaging. Brain and Cognition, 61,
54–68.
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 & Performance, 22, 1331–1356.
Hino, Y., Pexman, P. M., & Lupker, S. J. (2006). Ambiguity and
relatedness effects in semantic tasks: Are they due to semantic
coding? Journal of Memory and Language, 55, 247–273.
James, C. T. (1975). The role of semantic information in
lexical decisions. Journal of Experimental Psychology:
Human Perception & Performance, 1, 130–136.
Jenkinson, M., Bannister, P., Brady, M., & Smith, S. (2002).
Improved optimization for the robust and accurate linear
registration and motion correction of brain images.
Neuroimage, 17, 825–841.
Jenkinson, M., & Smith, S. M. (2001). A global optimization
method for robust affine registration of brain images.
Medical Image Analysis, 5, 143–156.
Katz, R. B., & Goodglass, H. (1990). Deep dysphasia: Analysis
of a rare form of repetition disorder. Brain and Language,
39, 153–185.
Kiefer, M. (2005). Repetition-priming modulates category-
related effects on event-related potentials: Further
evidence for multiple cortical semantic systems. Journal
of Cognitive Neuroscience, 17, 199–211.
Kiehl, K. A., Liddle, P. F., Smith, A. M., Mendrek, A., Forster,
B. B., & Hare, R. D. (1999). Neural pathways involved in
the processing of concrete and abstract words. Human
Brain Mapping, 7, 225–233.
Klinger, M. R., & Greenwald, A. G. (1995). Unconscious
priming of association judgments. Journal of Experimental
Psychology: Learning, Memory, & Cognition, 21, 569–581.
Koenig, P., Smith, E. E., Glosser, G., DeVita, C., Moore, P.,
McMillan, C., et al. (2005). The neural basis for novel
semantic categorization. Neuroimage, 24, 369–383.
Kroll, J. F., & Merves, J. S. (1986). Lexical access for concrete
and abstract words. Journal of Experimental Psychology:
Learning, Memory, & Cognition, 12, 92–107.
Kučera, H., & Francis, W. (1967). Computational analysis
of present-day American English. Providence, RI: Brown
University Press.
Marschark, M., & Paivio, A. (1977). Integrative processing
of concrete and abstract sentences. Journal of Verbal
Learning and Verbal Behavior, 16, 217–231.
Martin, A., & Chao, L. L. (2001). Semantic memory and the
brain: Structure and processes. Current Opinion in
Neurobiology, 11, 194–201.
Martin, A., Wiggs, C. L., Ungerleider, L. G., & Haxby, J. V.
(1996). Neural correlates of category-specific knowledge.
Nature, 379, 649–652.
Medin, D. L., Goldstone, R. L., & Gentner, D. (1993). Respects
for similarity. Psychological Review, 100, 99–130.
Noppeney, U., & Price, C. J. (2003). Functional imaging of
the semantic system: Retrieval of sensory-experienced and
verbally learned knowledge. Brain and Language, 84,
120–133.
Noppeney, U., & Price, C. J. (2004). Retrieval of abstract
semantics. Neuroimage, 22, 164–170.
Paivio, A. (1991). Dual coding theory: Retrospect and current
status. Canadian Journal of Psychology, 45, 255–287.
Perani, D., Cappa, S. F., Schnur, T., Tettamanti, M., Collina, S.,
All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them Pexman
Rosa, M. M., et al. (1999). The neural correlates of verb
and noun processing: A PET Study. Brain, 122, 2337–2344.
Pexman, P. M., Hino, Y., & Lupker, S. J. (2004). Semantic
ambiguity and the process of generating meaning from
print. Journal of Experimental Psychology: Learning,
Memory, & Cognition, 30, 1252–1270.
Piercey, C. D., & Joordens, S. (2000). Turning an advantage
into a disadvantage: Ambiguity effects in lexical decision
versus reading tasks. Memory and Cognition, 28, 657–666.
Pugh, K. R., Shaywitz, B. A., Constable, R. T., Shaywitz, S. A.,
Skudlarski, P., Fulbright, R. K., et al. (1996). Cerebral
organization of component processes in reading. Brain,
119, 1221–1238.
Ruby, P., & Decety, J. (2004). How would you feel versus
how do you think she would feel? A neuroimaging study
of perspective-taking with social emotions. Journal of
Cognitive Neuroscience, 16, 988–999.
Sabsevitz, D. S., Medler, D. A., Seidenberg, M., & Binder,
J. R. (2005). Modulation of the semantic system by word
imageability. Neuroimage, 27, 188–200.
Schwanenflugel, P. J. (1991). Why are abstract concepts hard to
understand? In P. J. Schwanenflugel (Ed.), The psychology
of word meanings (pp. 223–250). Hillsdale, NJ: Erlbaum.
Schwanenflugel, P. J., & Akin, C. E. (1994). Developmental
trends in lexical decisions for abstract and concrete words.
Reading Research Quarterly, 29, 251–264.
Schwanenflugel, P. J., Harnishfeger, K. K., & Stowe, R. W.
(1988). Context availability and lexical decisions for
abstract and concrete words. Journal of Memory and
Language, 27, 499–520.
Siakaluk, P. D., Pexman, P. M., Sears, C. R., & Owen, W. J.
(2007). Multiple meanings are not necessarily a disadvantage
in semantic processing: Evidence from homophone effects
in semantic categorization. Language and Cognitive
Processes, 22, 453–467.
Sim, E. J., & Kiefer, M. (2005). Category-related brain activity
to natural categories is associated with the retrieval of
visual features: Evidence from repetition effects during
visual and functional judgments. Cognitive Brain Research,
24, 260–273.
Sirigu, A., Duhamel, J.-R., & Poncet, M. (1991). The role of
sensorimotor experience in object recognition. Brain,
114, 2555–2573.
Smith, S. (2002). Fast robust automated brain extraction.
Human Brain Mapping, 17, 143–155.
Talairach, J., & Tournoux, P. (1988). Co-planar stereotaxic
atlas of the human brain. New York: Thieme.
Thompson-Schill, S. L., D’Esposito, M., Aguirre, G. K., & Farah,
M. J. (1997). Role of left inferior prefrontal cortex in retrieval
of semantic knowledge: A reevaluation. Proceedings of
the National Academy of Sciences, U.S.A., 94, 14792–14797.
Warrington, E. K. (1981). Concrete word dyslexia. British
Journal of Psychology, 72, 175–196.
Warrington, E. K., & Shallice, T. (1984). Category-specific
semantic impairments. Brain, 107, 829–854.
Wilson, M. (1988). The MRC Psycholinguistic Database.
Retrieved from http://www.psy.uwa.edu.au/mrcdatabase/
uwa_mrc.htm on 12 October 2005.
Woolrich, M. W., Behrens, T. E. J., Beckmann, C. F., Jenkinson,
M., & Smith, S. M. (2004). Multi-level linear modeling for
fMRI group analysis using Bayesian inference. Neuroimage,
21, 1732–1747.
Woolrich, M. W., Ripley, B. D., Brady, J. M., & Smith, S. M.
(2001). Temporal autocorrelation in univariate linear
modeling of fMRI data. Neuroimage, 14, 1370–1386.
immediately.
et al. 1419