c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
Available online at www.sciencedirect.com
ScienceDirect
Journal homepage: www.elsevier.com/locate/cortex
Research report
The neural bases of argument structure processing
revealed by primed lexical decision5
Donghoon Lee a, Benjamin Pruce b and Sharlene D. Newman b,*
a
b
Department of Psychology, Pusan National University, Busan, South Korea
Department of Psychological and Brain Sciences, Indiana University, Bloomington, IN, USA
article info
abstract
Article history:
Previous studies have reported anticipatory effects during sentence processing. However,
Received 30 May 2013
the source of these effects has not been clearly characterized. This study investigated the
Reviewed 22 July 2013
hypothesis that one source of anticipatory effects, particularly during verb processing, is
Revised 8 January 2014
the automatic triggering of argument structure processes. If argument structure processes
Accepted 21 April 2014
are automatically triggered it was hypothesized that the task need not require the initia-
Action editor Stefano Cappa
tion of the process, as such a primed lexical decision task was used that examined the
Published online 9 May 2014
neural priming of cross-grammatical class prime pairs (e.g., verb-noun priming). While
previous studies, as does the current study, have revealed behavioral priming for cross-
Keywords:
grammatical class and within-class (nounenoun and verbeverb) prime/target pairs, the
Priming
current results revealed significant activation differences. Enhancement effects were
fMRI
observed for cross-grammatical class priming in the language network, particularly the
Argument structure
inferior frontal gyrus (BA 47), and the posterior temporal cortex. Both regions have been
linked to argument structure processing previously. Within-class priming resulted in
neural suppression of the inferior temporal/occipital regions. Together, the data presented
suggest the automatic triggering of argument structure representations and demonstrate
that priming is a fruitful mechanism to explore aspects of sentence processing.
ª 2014 Elsevier Ltd. All rights reserved.
1.
Introduction
The language processing system has been characterized by a
set of separate processing modules which include systems
that process orthographic, phonological, syntactic, and semantic information (Levelt et al., 1999). However, there is
overwhelming evidence indicating that comprehenders use
linguistic and non-linguistic information to anticipate or
5
predict upcoming information (Altmann & Kamide, 1999;
Chambers & San Juan, 2008; Delong, Urbach, & Kutas, 2005;
Ferretti, McRae, & Hatherell, 2001; Kukona, Fang, Aicher,
Chen, & Magnuson, 2011). One of the first studies demonstrating anticipatory effects in sentence comprehension was
reported by Altmann and Kamide (1999) using the visual world
paradigm. When participants listened to a sentence that
contained the verb eat while viewing a scene that contained a
cake, ball, train and car, they immediately directed their gaze
This is an original work that has not been published elsewhere.
* Corresponding author. Department of Psychological and Brain Sciences, Indiana University, Bloomington, IN 47405, USA.
E-mail address: [email protected] (S.D. Newman).
http://dx.doi.org/10.1016/j.cortex.2014.04.013
0010-9452/ª 2014 Elsevier Ltd. All rights reserved.
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
to the object in the scene that was edible (e.g., a cake) instead
of the inedible objects (e.g., ball, train, car). However, when the
utterance contained the verb move participants delayed fixation until the direct object was fully specified. The anticipatory
effect observed at the verb was interpreted to suggest that “the
processor can predictively activate representations corresponding to a verb’s arguments” (p. 262). In other words, the
verb primes the nouns that potentially act as its arguments,
demonstrating significant interactions between semantic and
syntactic processing modules.
Behavioral priming studies also support the hypothesis
that anticipatory effects impact sentence comprehension
(Ferretti et al., 2001; McRae, Hare, Elman, & Ferretti, 2005).
Many of the constraints that drive anticipatory effects are
thought to be provided by the verb (Altmann & Kamide, 1999;
Kukona et al., 2011; McRae et al., 2005). In fact, McRae and
colleagues argue that accessing the verb also activates “highly
specific knowledge about the entities that typically participate
in the event that they encode” (page 1176); thereby driving the
priming of nouns. However, nouns can also drive anticipatory
effects in that they also activate the events that they typically
are involved in (i.e., nouns also prime verbs). In a series of
studies it has been demonstrated that verbs prime nouns
(Ferretti et al., 2001) and that nouns prime verbs (McRae et al.,
2005). Although these previous studies show that crossgrammatical class priming (e.g., nouns priming verbs) has
similar behavioral effects as within-class priming (e.g., nouns
priming nouns), it is not at all clear whether different mechanisms are responsible for the effect. If, as suggested by
Altmann and Kamide (1999) the anticipatory effects observed
for cross-grammatical class priming are related to filling thematic roles instead of a semantic association between the verb
and the object then different mechanisms may be expected for
cross-compared to within-grammatical class priming.
The primary goal of the current study was to determine
whether these anticipatory or priming effects observed in
noun/verb priming are the result of attempts to fill the argument roles of the verb. Verbs play a key role in sentence level
syntactic processing. One important lexical-syntactic feature
of verbs is its complement (entities that denote the participants involved in the event described by the verb) structure.
Another important feature of verbs is the number and types of
thematic roles (e.g., agent) the verb assigns to the complements. Essentially thematic roles describe “who did what to
whom” in a sentence. This argument structure information,
both the complement and thematic role information, are
thought to be represented in the lexical entry of the word
(Boland, 1993; Boland et al., 1990; Holmes, 1987; Shapiro et al.,
1987, 1993; Shetreet, Palti, Friedmann, & Hadar, 2007;
Tanenhaus et al., 1989; Trueswell et al., 1993). Therefore, it is
reasonable to hypothesize that this information becomes
active when the verb is activated. The question addressed
here is does the activation of this information automatically
trigger the process of filling those argument roles.
The majority of the studies examining cross-grammatical
class priming have used behavioral methods. While they have
consistently reported behavioral priming effects, it is unclear
what the source of the priming effect is. The aim of the current
study was to examine the neural architecture that supports
the behavioral priming effects observed in cross-grammatical
199
class priming with the goal of determining whether syntactic,
argument structure processes are evoked when nouns prime
verbs and vice versa. To accomplish this goal we explored
within- and cross-grammatical class priming. Crossgrammatical priming here refers to noun-verb and verbnoun priming in which the verb is an action that can be performed by the noun and therefore the noun is a potential
thematic argument of the verb. In addition to examining the
neural bases of cross-grammatical priming, the current study
also compared it to within-class priming. The withingrammatical class priming was used here as a control condition. Because there are a number of previous studies that have
examined within-class priming, using it as a control provided
a point of comparison. The within-class priming examined
nounenoun and verbeverb pairs in which the nounenoun
pairs were animals with similar characteristics (e.g., spiderscorpion) and the verbeverb pairs were manner of motion
verbs depicting a similar motion (e.g., scoot-scram).
Functional magnetic resonance imaging (fMRI) priming
studies have two potential responses, suppression or
enhancement effects. Suppression, decreased activation for
the related compared to the neutral or unrelated baseline
condition, during fMRI priming studies has been interpreted to
indicate overlap between the prime and target, either overlapping semantic features or processes (Henson, 2003;
Schacter, Wig, & Stevens, 2007). Enhancement, on the other
hand, is increased activation compared to the baseline for the
primed target and is thought to be due to different or additional processes related to the formation of new representations (Henson, 2003; Raposo, Moss, Stamatakis, & Tyler, 2006).
There have been a number of fMRI studies of semantic priming and the results have been mixed. Many of these neuroimaging studies have reported activation suppression for the
primed target (Copland, et al., 2003; Gold, et al., 2006; Kotz,
Cappa, von Cramon, & Friederici, 2002; Mummery, Shallice,
& Price, 1999; Rissman, Eliassen, & Blumstein, 2003; Ruff,
Blumstein, Myers, & Hutchison, 2008; Wheatley, Weisberg,
Beauchamp, & Martin, 2005). The cortical region most often
found to show suppression effects is the inferior occipital/
temporal region. However, there are also studies that report
enhancement effects (Kotz et al., 2002; Raposo et al., 2006;
Rossell, Price, & Nobre, 2003). The enhancement effects have
been in the right hemisphere, left middle temporal cortex and
inferior parietal cortex. It has been suggested that the differences across studies may be due to differences in methods,
tasks and materials (Raposo et al., 2006). However, by using
the same task in the same participants and overlapping
stimuli, it should make comparing within- and cross-grammatical class priming more reliable.
Cross-grammatical class priming was predicted to reveal
significant enhancement effects. These effects were predicted
to result from integration processes that are automatically
triggered when a thematically-related noun and verb are
presented. Activation enhancement was predicted in the regions implicated in complement processing, such as middle
and superior temporal cortex (Assadollahi, Meinzer, Flaisch,
Obleser, & Rockstroh, 2009; Bornkessel, Zysset, Friederici,
von Cramon, & Schlesewsky, 2005) and thematic role assignment, which involves the left inferior frontal gyrus (Hirotani,
Makuuchi, Ruschemeyer, & Friederici, 2011, Newman, Ikuta,
200
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
& Burns, 2010; Newman, Just, Keller, Roth, & Carpenter, 2003)
along with the temporal regions mentioned.
Choosing the appropriate baseline condition to compare
within- and cross-grammatical class priming was a challenge.
Although the standard procedure in the semantic priming
literature is to use an unrelated priming condition as the
baseline (McNamara, 2004), a neutral baseline condition
(pseudoword prime) was used here. The neutral priming
condition has been used previously to differentiate facilitation
(faster response times than neutral) and inhibition (slower
response times than neutral; Gold et al., 2006; Neely et al.,
1989). Typically this is done by comparing related and unrelated priming conditions to the neutral condition. In this case
the neutral condition was not used to compare related and
unrelated conditions but semantic and thematic priming. As
mentioned above it is hypothesized that semantic priming
would result in fMRI activation suppression while thematic
priming in enhancement. While inhibition and facilitation are
not directly equivalent to suppression and enhancement, they
are analogous.
2.
Methods
2.1.
Participants
Eighteen right-handed native English speakers (age: 20e35; 10
males) participated in the study for pay. All participants reported no history of neurological or psychiatric disorders and
provided written informed consent, as approved by the
Institutional
Review
Board
of
Indiana
University,
Bloomington.
2.2.
Materials
Several steps were used to prepare the stimuli for the primed
lexical decision task. First, a list of animal names were
generated, and a list of manner of motion verbs were taken
from Levin (1993). From the lists, forty animaleanimal name
pairs of the same biological class, and forty manner of motion
verb pairs describing similar motions were created for the
within-class condition. Additionally, forty animal-verb and
forty verb-animal pairs were created for the crossgrammatical class condition. The animal and verb names
were chosen to ensure that they were matched across lexical
features (see below). Target words in each category were
repeatedly paired with four prime types comprising the other
experimental conditions: a within-class related, cross-class
related, unrelated, and a baseline condition. For the unrelated condition, the animal name targets were paired with tool
names and the manner of motion verb targets were paired
with cognition verbs. For the baseline condition, target words
were paired again with orthographically legal and pronounceable pseudowords with perceptual characteristics
(e.g., length) similar to other prime stimuli (see Appendix for
full lists of stimuli). Finally, the four prime stimuli types were
repeatedly paired with 40 nonword targets. The unrelated
condition was treated as a filler condition.
Lexical characteristics as well as mean response time (RT)
of the lexical decision for the current word stimuli were
examined using the English Lexicon Project (ELP: http://
elexicon.wustl.edu/, Balota, et al., 2007). Statistically, there
was no significant difference between noun and verb conditions in terms of word frequency [t(39) < 1; p > .75], length
[t(39) < 1; p > .43], and number of phonemes [t(39) < 1; p > .87],
as well as the lexical decision time extracted from the database [t(30) ¼ 1.5; p > .08]. Mean lexical characteristics (except
word frequency) of the word stimuli were entered as constraints in the web application of ELP for generating pseudoword prime stimuli, which were pronounceable, and
orthographically legal pseudowords. Thus, the lexical characteristics of pseudoword stimuli were very similar to those of
word stimuli in the current experiment. Mean scores and
standard deviations (SD) of the characteristics of stimuli in
each condition are summarized in Table 1.
In the second step, the semantic relatedness of the
generated word pairs was examined. First, we retrieved semantic similarity scores for the stimuli from Latent Semantic
Analysis (LSA). The semantic similarity scores were computed
by measuring the co-occurrence rates (see Table 2; Landauer &
Dumais, 1997). For this, we used a web application of LSA
(http://lsa.colorado.edu/) and retrieved the scores based on
general reading materials for up to 1st year college students,
with 300 semantic factors. The LSA semantic similarity scores
of word pairs was highest for the AnimaleAnimal name pairs
(mean ¼ .38, SD ¼ .17) and about the same for the other 3
conditions (see Table 2). The second semantic relatedness
measurements were subjective ratings of 14 native English
speakers, who did not participate in the current fMRI study.
The degree of semantic relatedness of the current stimulus
pairs were rated using a scale from one to five (one ¼ not
related to five ¼ strongly related). The ratings were highest for
Table 1 e Lexical characteristics of stimuli.
Lexical info.
Animal nouns conditions
Prime stimuli
Length
Frequencya
No. of phoneme
No. of morpheme
Imageabilityb
a
Action verbs conditions
Target stimuli
Prime stimuli
Target stimuli
Animal
Action
pse
Animal
pse1
Animal
Action
pse
Action
pse2
5.9
7.1
4.9
1.1
4.4
5.7
7.0
4.5
1.0
3.6
5.9
6.1
7.1
5.0
1.1
4.6
5.9
6.1
6.9
5.1
1.1
4.5
5.8
6.8
4.7
1.2
3.4
5.8
5.7
7.1
4.5
1.1
3.5
5.7
Word frequency is the log-transformed HAL frequency which is based on the Hyperspace Analog to Language (HAL) corpus (Lund and Burgess,
1996). The ELP recommends to use the HAL frequency rather than the old Kucera and Francis (1967).
b
The imageability was scored by 5-to-1 scales.
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
Table 2 e Lexical characteristics of stimuli.
Similarity score
ani-ANI act-ANI ani-ACT act-ACT
LSA similaritya
.38 (.17)c
b
Relatedness judgment 3.25 (.70)
.18 (.18)
2.3 (.78)
.18 (.2)
1.90 (.95)
.17 (.19)
3.01 (.84)
a
The scale of LSA similarity is arbitrary. See (Landauer & Dumais,
1997).
b
Similarity judgment was conducted with 5 scales (maximum 5 to
minimum 1).
c
The number in parentheses presents the standard deviation.
the AnimaleAnimal and MotioneMotion conditions with no
significant difference between them (see Table 2).
Finally, the stimuli were distributed into eight lists to
control possible confounds from the repetition of targets.
There were 640 total trials: 160 animal name target trials (4
prime stimuli; within-class, cross-class, pseudoword and
filler 40 animal name targets), 160 manner of motion verb
target trials (4 prime stimuli 40 motion verb targets), 320
pseudoword target trials paired with the four types of prime
stimuli. Since the current experiment was a within-subject
design, repetition of stimuli occurred for each target and
prime item. To make the repetition effects the same across
conditions, the order of conditions was counterbalanced.
Eight lists of stimuli using a Latin square permutation were
created. Once the repetition order of the stimuli was manipulated across lists, the order of presentation of trials in each
list was pseudo-randomly scheduled with a variable interstimulus interval (ISI) which is necessary for the rapid
event-related design used (see below).
2.3.
Experimental design and procedure
The experiment was composed of eight runs which were 6 min
24 sec, making each imaging session approximately an hour.
Presentation of stimuli in each list was pseudo-randomly
scheduled as required for a rapid event-related fMRI design
with variable ISIs (mean ¼ 4 sec, ranged from 2.5 sec to 7 sec). A
trial was initiated by a green plus sign presented in the middle
of the screen for 250 msec. Once the green plus sign disappeared, a prime stimulus was presented in the middle of
screen for 200 msec and masked by a series of hatch signs
(#######) for 50 msec; then a target stimulus was presented for
750 msec. Thus, the SOA between the prime and the target
stimulus was 250 msec. Subjects were asked to judge, as soon
as possible, whether the target stimulus was a real word or not
by pressing a button with the left index finger for ‘yes’, and
with the right index finger for ‘no’. Behavioral responses were
collected until 1500 msec after the target presentation. The
response button mapping was constant across subjects.
2.4.
fMRI data acquisition and analysis
Functional MRI scanning was conducted on a Siemens 3T Tim
TRIO scanner in the Imaging Research Facility at Indiana
University. Structural and functional images were acquired
with a 32-channel radio frequency head coil in the scanner.
Functional images were acquired with a gradient echo planar
T2* imaging (EPI) sequence for the blood oxygenation level
dependent (BOLD) contrast repetition time (TR) ¼ 2000 msec,
201
echo time (TE) ¼ 25 msec, flip angle ¼ 70 , matrix
size ¼ 128 128, field of view (FOV) ¼ 220 220 mm2, 35
oblique axial slices with 3 mm thickness, no gap between
slices, original voxel size ¼ 1.725 1.725 3 mm3). A structural
image was acquired with a high resolution MPRAGE anatomical sequence (TR ¼ 1900 msec; TE ¼ 4.15 msec; TI ¼ 1100 msec;
1 mm3 isotropic voxel; 256-mm field of view.
fMRI data were analyzed with SPM8 (Wellcome Trust Centre
for Neuroimaging; http://www.fil.ion.ucl.ac.uk/spm). fMRI
data were preprocessed in several steps including slice timing
correction, motion correction by realignment, co-registration
between functional and anatomical scans, spatial normalization and smoothing. All functional data were resampled to
2 mm3 isomorphic voxels normalized to the Montreal Neurological Institute (MNI) template. For the spatial smoothing a
6 mm FWHM Gaussian kernel was applied. On the preprocessed fMRI data of individual subjects, a canonical statistical analysis based on the general linear model (GLM)
and Gaussian random field theory was performed (Friston,
Frith, Turner, & Frackowiak, 1995). The event-related hemodynamic response for the target stimuli in each condition was
modeled with a canonical hemodynamic response function
(HRF) built on the onsets of the prime stimuli with a duration
that includes the stimulus presentation and reaction time. For
each individual data analysis, 16 regressors were built for 2
response types (word vs pseudoword) 2 target types (animal
vs action) 4 primes (within-class, cross-class, pseudoword,
and filler). Additionally, a regressor for incorrect responses and
6 regressors from the realignment step were included in the
model to remove unexpected effects from careless responses
and noise from head movement. Contrast images for withinclass, cross-class and pseudoword primed conditions for animal name targets and motion verb targets were obtained for
each individual subject and entered in the group-level analysis
for the two by three (target prime) factorial analysis. Since the
current experiment was repeated within-subject design, the
subject variable was used as a basic factor in the flexible
factorial analysis to estimate the main effect of target, the main
effect of prime and the interaction. In post-hoc tests, semantic
and thematic priming effects for each target category (i.e.,
animal and manner of motion verbs) were examined by contrasting each condition to the pseudoword prime condition.
For the contrasts examined we applied a Monte Carlo
simulation of the brain volume to establish an appropriate
voxel contiguity threshold. The threshold obtained from the
simulation has the advantage of higher sensitivity to smaller
effect sizes like those seen in neural priming studies (Slotnick
& Schacter, 2004). The result of the Monte Carlo simulation
indicated that a cluster size of 23 contiguous resampled voxels
using an uncorrected threshold of p < .005 would be appropriate to control type I error, at a p < .05 corrected for the
multiple comparisons in the whole brain volume analysis.
3.
Results
3.1.
Behavioral results
A two by three repeated measures ANOVA was performed and
revealed a main effect of target [F(1,17) ¼ 30.25, p ¼ .000,
202
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
h2p ¼ .640], a main effect of prime [F(2,34) ¼ 49.52, p ¼ .000,
h2p ¼ .744], and an interaction [F(2,34) ¼ 11.52, p ¼ .000, h2p ¼ .404],
see Fig. 1. Post-hoc t-tests (using a one-tailed test) comparing
the within-class priming and the cross-class priming conditions to neutral, pseudoword priming revealed that for both
the animal and action target conditions a significant effect of
within-class priming was observed [animal: t(17) ¼ 3.01,
p < .005; action: t(17) ¼ 2.18, p < .05]. The animal target condition [VeN; t(17) ¼ 1.8, p < .05] revealed a significant crossclass priming effect; the action target condition (NeV; t < 1)
failed to show an effect.
When examining the pseudoword target conditions no
significant effects were observed [target: F(1,17) ¼ 1.62, p ¼ .219,
h2p ¼ .087; prime: F(2,34) ¼ 1.59, p ¼ .225, h2p ¼ .086;
target prime: F(2,34) ¼ 1.69, p ¼ .206, h2p ¼ .091].
3.2.
fMRI results
All of the fMRI results reported are for the animal and action
targets. Therefore, pseudoword condition below refers to the
pseudoword priming condition.
3.2.1.
Suppression effects
Within-class suppression effects were determined by examining the pseudoword minus both the animaleanimal and
actioneaction conditions. This contrast revealed priming effects in inferior occipital/temporal cortex including the fusiform and lingual gyri, as well as in bilateral prefrontal cortex,
medial prefrontal cortex, posterior cingulate cortex and the
cerebellum (see Table 3 and Fig. 2). Cross-class suppression
effects were found only in the right medial prefrontal cortex
as well as medial parietal cortex.
3.2.2.
Enhancement effects
Within-class enhancement effects (greater activation for the
semantic condition compared to the pseudoword prime condition) were found in left BA 47 and the middle STG. Crossclass enhancement effects were also observed in traditional
language processing regions including the middle and superior temporal cortex in the left hemisphere, the inferior frontal
gyrus (BA 47) as well as the postcentral gyrus (see Table 4 and
Fig. 3).
3.2.3.
4.
Fig. 1 e Behavioral results show a decrease in reaction time
for both cross-class and within-class priming compared to
the pseudoword prime condition. The cross-class priming
for the Action target condition, however, was not
statistically significant. The stars indicate a significant
difference from the pseudoword baseline (prime neutral)
condition.
Simple contrasts
Contrasts were also performed to explore the differential
activation linked to Animal-Action and Action-Animal priming. The Animal-Action condition elicited increased activation
the inferior frontal gyrus (BA 47) compared to Action-Animal
priming while Action-Animal priming elicited greater activation in bilateral middle frontal cortex and the thalamus (See
Table 5 and Fig. 4).
To further explore the data the simple contrasts examining
the enhancement effects for the Action-Animal and AnimalAction priming conditions with different “baseline” conditions were computed (see Fig. 5). It should be noted that
although there is an unrelated prime condition it was not the
focus of this study because it is a semantically unrelated
condition and was not controlled for thematic relatedness. For
example, repel-sniff is an actioneaction stimulus. While repel
and sniff are semantically unrelated they may be thematically
related in that the two words together could be interpreted as
attempting to suppress a sniff. A comparison with this unrelated
baseline is presented here for completeness. As shown, the
results of this analysis echo those described above e the
enhanced activation of BA 45/47 is primarily observed for the
Animal-Action condition. Also, increased activation of STG is
strongest when compared to a pseudoword prime condition
and absent when compared to an unrelated condition which
supports the suggestion that this unrelated condition may be
inadvertently thematically related.
To explore the differences between within-class and crossclass priming a contrast comparing them was performed. The
cross-class minus within-class contrast revealed significant
activation in a number of regions including the posterior STG,
and the occipital cortex (see Table 6).
Discussion
The current study examined the neural bases of crossgrammatical class priming. The results presented here show
that cross-grammatical class priming resulted primarily in
enhancement effects (increased activation relative to baseline) in classical language processing regions. Additionally,
differences were observed when examining the two crossgrammatical class priming conditions (NeV and VeN) such
that a behavioral priming effect was observed only for the
VeN condition while the NeV condition elicited a larger left
inferior frontal gyrus response. Finally, when comparing
cross-grammatical class to within-grammatical class priming
the cross-grammatical class priming condition elicited greater
activation in the left posterior superior temporal gyrus, a region associated with argument structure processing. The data
presented here suggest the automatic triggering of argument
structure representations and demonstrate that priming may
be a fruitful paradigm to explore aspects of sentence
processing.
203
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
Table 3 e fMRI suppression effects.
Region
Pseudoword-Within-class
Right
Middle frontal
Left
Middle frontal
Left
Cuneus
Left
Lingual gyrus
Right
Inferior occipital
Left
Cuneus
Left
Cerebellum
Left
Cerebellum
Right
Medial frontal
Left
Medial frontal
Left
Precuneus
Right
Precuneus
Right
Precuneus
Left
Cerebellum
Pseudoword-Cross-class
Right
Medial frontal
Left
Anterior cingulate
Right
Frontal
Right
Middle frontal
Left
Lingual gyrus
Right
Anterior cingulate
Left
Anterior cingulate
Right
Caudate
Left
Parahippocampal gyrus
Right
Precuneus
Right
Cuneus
Right
Medial frontal
BA
Extent
Z-score
x
y
z
10
10
17
17
18
19
77
61
624
8
9
7
7
7
23
70
25
23
4.29
4.05
3.97
3.96
3.66
3.94
3.53
2.84
3.29
3.14
3.07
2.97
2.79
2.96
34
28
2
14
34
26
36
26
8
2
4
24
20
22
62
60
94
100
90
88
82
74
20
42
54
54
58
68
0
12
2
8
6
34
32
40
46
18
56
48
42
24
4.05
3.07
3.84
3.83
3.68
3.61
2.77
3.26
3.24
3.22
3.18
3.16
2
8
24
20
12
4
2
22
24
12
18
18
28
34
0
34
96
42
36
26
36
58
92
62
10
4
56
38
6
12
14
28
12
26
0
4
46
50
33
11
32
6
8
18
32
32
36
31
17
10
84
48
92
107
69
23
54
23
36
23
Fig. 2 e The figure depicts the regions that revealed a within-class (A) and cross-class (B) suppression effect. As shown
suppression effects for within-class priming were observed in the inferior occipital cortex bilaterally as well as bilateral
prefrontal cortex while effects for cross-class priming were observed in medial prefrontal regions.
204
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
Table 4 e fMRI enhancement effects.
Region
Within-class e Pseudoword
Left
Inferior frontal
Left
Inferior frontal
Left
Inferior frontal
Left
Parahippocampal gyrus
Right
Precentral gyrus
Right
Putamen
Left
Amygdala gyrus
Left
Middle temporal
Left
Middle temporal
Right
Superior temporal
Right
Superior temporal
Left
Middle frontal
Right
Putamen
Right
Precentral gyrus
Right
Precentral gyrus
Right
Cerebellum
Right
Cerebellum
Right
Postcentral gyrus
Right
Superior frontal
Left
Inferior parietal
Right
Insula
Cross-class e Pseudoword
Left
Middle temporal
Right
Precentral gyrus
Left
Superior temporal
Left
Middle temporal
Left
Inferior frontal
Left
Inferior frontal
Left
Middle frontal
Left
Postcentral gyrus
Right
Postcentral gyrus
Right
Postcentral gyrus
Right
Insula
Left
Supramarginal gyrus
Right
Putamen
Right
Superior temporal
Right
Cerebellum
Left
Putamen
Left
Cuneus
BA
Extent
Z-score
x
y
z
47
10
46
36
6
719
4.65
4.12
3.78
4.18
4.03
3.98
3.96
3.91
3.82
3.74
3.72
3.66
3.59
3.58
3.57
3.54
3.3
3.5
3.37
3.3
3.14
50
44
48
40
24
26
34
56
60
66
44
36
30
54
64
38
30
52
10
46
40
34
46
38
28
18
18
6
40
34
18
0
28
8
6
4
64
66
14
4
52
2
6
4
8
20
54
14
18
4
0
2
14
18
4
12
30
38
38
54
66
48
2
4.67
4.63
4.32
4.31
4.28
3.9
3.89
4.15
4
3.1
3.66
3.36
3.28
3.24
3.23
3.17
2.81
56
64
46
52
50
46
36
40
52
56
36
64
28
50
30
28
24
40
4
54
64
34
40
34
20
12
20
16
50
8
14
74
2
82
2
20
14
12
6
18
16
54
52
44
12
28
4
10
26
2
32
22
21
22
38
46
43
6
50
60
48
36
230
50
80
37
148
364
57
3
6
40
13
43
28
41
33
22
6
39
39
47
47
11
3
3
2
13
40
493
91
510
22
19
Relating fMRI priming to behavioral priming results is far
from straightforward. One reason is that while behaviorally
priming is defined by improved processing or faster
responding to a stimulus. This intuitively means that there
should be less processing taking place when there is priming.
However, this is not necessarily what is observed. fMRI
priming studies have found both suppression or reduced
activation akin to faster responding, as well and enhancement
or increased activation. Therefore, the behavioral response to
priming can be driven by multiple factors and this is what is
shown here. An open question is how does enhancement or
the engagement of additional processes result in faster
response times? In order for the engagement of additional
processes to facilitate the behavioral response the system
must be parallel and interactive and the additional process
must be fast. In other words, lexical decision and argument
structure processing must take place in parallel and the
argument structure processing must take place rapidly and
automatically and then feedback onto the lexical decision
process.
658
198
147
36
28
37
26
30
33
26
4.1.
Argument structure processing
The primary goal of this study was to explore whether argument structure processing is automatically triggered when a
verb and a potential complement are presented, in this case
within a primed lexical decision task. The fMRI enhancement
effects found for cross-grammatical class priming suggests that
these argument structure processes are indeed triggered. The
regions that revealed enhancement effects e the inferior frontal gyrus (IFG) and middle and superior temporal cortex in the
left hemisphere e have been found previously to be involved in
sentence processing (Newman et al., 2003, 2010). For example,
the region of the left IFG activated here has been implicated in
thematic role assignment previously (Hirotani et al., 2011;
Newman et al., 2003, 2010), BA 47/45. In a study examining the
impact of semantic relatedness on syntactic processing
Newman et al. (2010) suggested that the semantic relationship
between the noun and verb may allow for the priming of the
verb representation which includes argument structure information, causing facilitation of thematic role assignment.
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
205
Fig. 3 e The figure depicts the regions that revealed a cross-class enhancement effect. As shown left hemisphere language
processing regions revealed enhancement effects. Both within-class (A) and cross-class (B) priming revealed significant
enhancement effects in BA 45/47 and middle STG. Only cross-class priming revealed an effect in posterior STG.
As mentioned earlier, the posterior superior temporal region activated here has been found to be involved in sentence
level processing (Bornkessel, et al., 2005; Caplan, Stanczak, &
Waters, 2008; Friederici, Opitz, & Von Cramon, 2000;
Kuperberg et al., 2003; Sakai, Tatsuno, Suzuki, Kimura, &
Ichida, 2005). Bornkessel et al. (2005) have argued that during
sentence processing the posterior superior temporal cortex is
involved in syntax-semantic integration in that it integrates
information into the argument hierarchy. Additionally,
the region has been linked specifically to the access of
Table 5 e Cross-class minus within-class priming.
Region
Left
Left
Right
Left
Left
Left
Right
Left
Right
Right
Right
Right
Right
Right
Left
Left
Cuneus
Superior occipital
Middle occipital
Precentral
Cingulate
Cingulate
Precuneus
Superior temporal
Precuneus
Precuneus
Precuneus
Superior occipital
Middle temporal
Middle temporal
Postcentral
Middle frontal
BA Extent Z-score
19
19
18
4
31
24
7
39
31
19
19
19
39
39
5
6
253
84
58
60
33
37
128
88
46
28
51
4.64
3.54
3.94
3.72
3.67
2.92
3.67
3.56
3.32
3.27
3.18
3.3
3.25
3.23
3.21
2.99
x
y
z
26
30
44
40
10
4
26
46
28
22
22
42
46
58
32
38
86
90
86
18
22
14
54
54
72
82
70
80
76
58
40
2
32
24
2
54
42
40
52
12
28
40
34
30
22
12
58
60
subcategorization options for verbs (Shetreet et al., 2007). For
example, Shetreet et al. (2007) found that activation of STG was
correlated with the number of complement options a verb can
take during sentence comprehension. Also, Shetreet et al.
(2007) reported both the STG and BA47 were involved in subcategorization processes. However, these two regions may be
involved in different processes. Shetreet and colleagues suggested that BA 47 was involved in selecting the appropriate
noun (semantic processing) to incorporate into the subcategorization frame while the frame was accessed via the STG. In
this way the retrieval of the frame may occur when a verb is
encountered but the presence of another word, preferably a
noun, triggers the integration process, supporting the hypothesis that subcategorization information is linked to the
verb representation.
The data presented here provides support for a differential
role of BA 47 and the STG in argument structure processing.
BA 47 failed to show a significant difference in the crossgrammatical class minus within-class contrast with the region revealing enhancement effects for both conditions, supporting its role in semantic processing, generally. However,
the posterior STG did reveal a significant increase in activation
for cross-grammatical class compared to within-class priming, supporting the hypothesis that this region is involved in
argument structure processing, more specifically. These differences in the response of BA 47 and posterior temporal
cortex further support that these regions make unique contributions to argument structure processing.
206
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
Fig. 4 e The figure depicts the activation resulting from the simple contrasts e verb-noun minus noun-verb (left) and nounverb minus verb-noun (right). As shown, the noun-verb condition elicited greater activation in BA 47 while the verb-noun
condition elicited greater activation more anterior regions of the middle frontal gyrus.
The data presented here suggests that argument structure
processes are rapid enough to impact lexical decision. This is
an interesting finding that supports earlier theories of syntactic processing that suggest that there are two passes, a fast,
heuristic analysis followed by a slower more algorithmic
analysis (Hahne & Friederici, 1999), with argument structure
processing being related to the first pass. There is some evidence from electrophysiology to support the hypothesis that
argument structure processing occurs rapidly and automatically. In a recent MEG study syntactic information related to
the word category elicited visual cortex activity 120 msec after
stimulus onset (Dikker, Rabagliati, Farmer, & Pylkkanen, 2010).
The responses observed here may also be linked to the eLAN,
an electroencephalography (EEG) response that occurs
100e300 msec after stimulus onset and is related to initial
phrase structure building based on syntactic word category
information (Friederici, 2004). eLAN effects are observed in
electrodes located over the prefrontal cortex, suggesting that
the effect may be more related to the integration process
taking place in BA 47, or as Shetreet and colleagues stated, the
incorporation of the appropriate item into the subcategorization frame. If this is the case then it provides some support for
the use of priming paradigms to more extensively explore the
eLAN as well as argument structure processing.
Fig. 5 e The figure depicts the separate contrasts for the action-animal (VN) condition (left column) and the animal-action
(NV) condition (right column). The first row is the VN(NV) minus the pseudoword prime contrast. The second row is the
VN(NV) minus the related NN(VV) contrast. The bottom row is the VN(NV) minus the unrelated NN(VV) contrast.
207
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
Table 6 e Simple contrasts.
Region
Animal-Action minus Action-Animal
Left
Inferior frontal
Left
Inferior frontal
Left
Inferior frontal
Right
Inferior parietal
Right
Postcentral
Left
Postcentral
Right
Posterior cingulate
Right
Parahippocampal gyrus
Left
Middle occipital
Right
Precuneus
Left
Amygdala
Right
Middle temporal
Right
Inferior frontal
Action-Animal minus Animal-Action
Left
Middle frontal
Left
Superior frontal
Right
Medial frontal
Right
Cingulate gyrus
Right
Middle frontal
Right
Medial frontal
Left
Paracentral
Left
Posterior cingulate
Left
Medial frontal
Left
Medial frontal
Left
Superior frontal
Right
Superior parietal
Right
Inferior parietal
4.2.
BA
Extent
Z-score
x
y
z
47
10
47
40
3
3
30
19
19
7
251
3.96
3.49
3.65
3.61
2.99
3.57
3.48
3.06
3.36
3.3
3.3
3.28
3.03
48
48
32
40
44
24
24
28
44
14
32
54
14
32
42
26
30
18
34
64
56
82
60
2
8
24
6
4
18
44
52
54
12
2
14
52
12
16
18
3.92
3.02
3.56
3.5
3.4
3.36
3.3
3.27
3.26
3.07
3
2.87
2.85
26
18
4
4
30
6
4
4
6
6
22
38
42
58
58
56
34
12
24
38
48
22
32
38
68
60
10
18
4
42
60
66
64
20
64
34
42
46
44
21
47
10
10
10
31
6
6
4
30
6
9
8
7
40
The impact of noun/verb order
The behavioral results revealed a larger priming effect for the
verb-noun compared to the noun-verb condition. This mirrors
the activation results in that the noun-verb condition resulted
in increased activation in BA 47 compared to the verb-noun
condition; both the behavioral and the neural results show
greater facilitation for the verb-noun condition. One interpretation of this result is that, as predicted, the verb prime
activates the argument structure which leads to the anticipation of thematic role fillers. As a result of the activation of
the argument structure, the verb as the prime for the noun
resulted in greater priming. Interestingly, one region that
revealed differences between these two conditions was in the
left IFG, a region that overlaps with one that has been previously linked to controlled semantic processing (ThompsonSchill, D’Esposito, Aguirre, & Farah, 1997) as well as thematic
role assignment. The results presented here suggest that
these two processes may be related.
Unlike the within-class priming condition the crossgrammatical class priming pairs were not semantically
related; however, they may be considered to be associatively
related. When examining the relatedness judgments the
cross-grammatical class pairs were rated less related than the
within-class pairs but their scores were in the middle of the
range suggesting that participants detected some level of
relatedness. A recent study by Sachs et al. (2011) found differential activation patterns for categorical (pot-kettle) and
associative (pot-stove) priming. The categorical priming effect
was observed as a suppression effect in the right middle
frontal gyrus and was linked to the retrieval of perceptual
38
76
27
58
44
32
24
28
29
115
86
57
72
95
62
27
24
36
37
features of objects. However, when prime-target pairs were
associated with each other the left IFG (BA 47) revealed
enhancement effects. While not significantly different, the
verb-noun condition was rated more related than the nounverb condition. Together these results suggest that BA 47 is
involved in semantic retrieval processes, generally, and that
its activation is modulated by the difficulty of that retrieval.
4.3.
A neural account
There is still the question of how the neural system accomplishes this fast, heuristic process. One possible clue may be
found in a theory of neural processing, the predictive coding
framework (Mumford, 1992; Price & Delvin, 2011). The basic
idea of predictive coding is that neural regions receive both
bottom-up input from lower-level processes and top-down
predictions from higher-level processes. The predictions
attempt to match the bottom-up input with internal knowledge. Here when a verb is encountered its argument structure
is activated and predictions are generated regarding what
word or word type will follow. This prediction may be expected to come from higher-level semantic processing regions, like BA 47, while the posterior superior temporal gyrus
(STG) integrates information received from bottom-up processes and the predictions from BA 47. Some support for this is
seen in the enhancement effects for BA 47 for both withinclass and cross-class priming, suggesting that it is always
making predictions while the posterior superior temporal
gyrus (STG) shows enhancement when a verb is present. Also,
based on the predictive coding theory, if the prediction does
not match the bottom-up input an error is elicited and there is
208
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
an increase in neural activation (Price & Delvin, 2011) likely
due to an attempt to resolve the conflict. This may explain the
effect of noun/verb order. When the verb is presented first the
prediction regarding the next word is likely to be more accurate because the subcategorization frame provides for more
information to make a prediction. As a result, there should
less activation in both BA 47 and posterior STG for the verb/
noun compared to the noun/verb order because there is less
predictive error; and this is what is seen. While this model is
somewhat speculative, it fits the data presented and may be a
good starting point for future studies of cross-grammatical
class neural priming.
4.4.
previous studies have suggested, is involved in semanticsyntactic integration processes.
A confound in the current study is that relatedness/similarity is confounded with within-class/cross-class priming in
that the within-class pairs are more related than the crossgrammatical class pairs. This has been tangentially discussed above and has likely impacted the activation differences observed when comparing cross- and withingrammatical class priming. In an ideal study relatedness
would be equated across the semantic/thematic dimension,
meaning that the noun-verb pairs would be semantically
related just as the nounenoun and verbeverb pairs are.
However, that may not be possible.
Limitations
While the current study found enhancement effects for crossgrammatical class priming, other studies have shown suppression effects for a similar manipulation (Obleser & Kotz,
2010). For example, in an auditory sentence study using a
cloze probability manipulation the high probability condition
elicited less activation than the low probability condition in a
similar region of the posterior STG. There are a number of
differences between the Obleser and Kotz (2010) study and the
current one that may account for these differences. The tasks
and stimuli are different with the previous study not including
a task but instead passively presenting auditory sentences.
Also, the contrasts were different. Here we see greater activation for the cross-grammatical class priming condition
compared to the pseudoword prime condition in which no
integration processes are triggered. In the Obleser and Kotz
study, greater activation was observed for the less predicted
final word of a sentence where integration is thought to occur
in both conditions. What is important here is that the manipulations of both studies appear to show priming in a
similar region of posterior STG, suggesting that this region, as
5.
Conclusions
Together, the results presented here show that crossgrammatical class priming activates a similar network as
that reported in the sentence comprehension literature. This
indicates that priming is a promising paradigm to explore
argument structure processing. Priming may be preferred
over sentence comprehension in some populations (e.g.,
Alzheimer’s patients and young children) and contexts.
Additionally, the results presented here suggest that crossgrammatical class priming, in particular noun-verb priming, may elicit different neural mechanisms than withinclass priming even though behaviorally the effects are
similar.
Acknowledgment
This work was partially supported by National Institute of
Health Grant (R03 HD051579-01).
Appendices
Appendix A-1 e List of prime stimuli for animal noun targets.
Prime stimuli
Animal
Spider
Heron
Squid
Zebra
Hare
Python
Mastiff
Catfish
Wasp
Carp
Moose
Turtle
Canary
Terrier
Husky
Crocodile
Herring
Salamander
Crab
Target stimuli
Filler
Motion
Pseudoword
Chisel
Basin
Scarf
Blade
Sofa
Blouse
Blender
Slipper
Sash
Cage
Cannon
Dagger
Basket
Earring
Skirt
Nightgown
Goggles
Corkscrew
Cask
Crawl
Glide
Wriggle
gallop
Flee
Slither
Scuffle
Splash
Buzz
Dive
Roam
Creep
Squirm
Rush
Tramp
Saunter
Wiggle
Camouflage
Prowl
Givets
Colob
Coreg
Colls
Fass
Fonvex
Congelb
Tracong
Ohal
Boup
Pames
Tefine
Hotion
Drolics
Soeth
Idvesting
Tholera
Honsiderate
Lenk
Animal nouns
Scorpion
Pelican
Octopus
Pony
Rabbit
Cobra
Bulldog
Mackerel
Bee
Trout
Elk
Tortoise
Finch
Poodle
Collie
Alligator
Eel
Chameleon
Lobster
Pseudoword
Ronelier
Buieter
Borbles
Durp
Reperp
Ouler
Kraining
Fliskered
Fic
Yone
Gri
Alaiting
Cleot
Soiset
Rounch
Frobation
Ker
Frassness
Incerep
209
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
Appendix A-1 e (continued )
Prime stimuli
Animal
Butterfly
Lice
Beaver
Haddock
Raven
Gazelle
Pigeon
Swan
Dolphin
Panther
Monkey
Badger
Locust
Lark
Armadillo
Parrot
Frog
Skunk
Seal
Goat
Pheasant
Target stimuli
Filler
Strainer
Vise
Cradle
Curtain
Baton
Cabinet
Carton
Tray
Necktie
Platter
Crayon
Kettle
Mallet
Bowl
Detergent
Abacus
Sofa
Tiara
Oven
Rug
Bracelet
Motion
Flutter
Stray
Bound
Splatter
Twirl
Scurry
Flock
Float
Plunge
Ambush
Scratch
Crouch
Bounce
Soar
Scramble
Swagger
Squat
Waddle
Slip
Leap
Meander
Pseudoword
Dorrupted
Noal
Tadist
Deaters
Keams
Lonceit
Tookie
Pige
Prabbed
Dinters
Perein
Bocer
Mevies
Chru
Trothered
Boower
Boug
Floser
Kons
Hoos
Parbines
Animal nouns
Pseudoword
Moth
Flea
Weasel
Anchovy
Crow
Antelope
Dove
Mallard
Whale
Coyote
Baboon
Squirrel
Grasshopper
Sparrow
Porcupine
Peacock
Toad
Otter
Walrus
Cow
Quail
Kass
Noen
Goicet
Reneral
Basu
Feterans
Fole
Voupsh
Coips
Bualte
Borce
Lerriers
Prissgoffer
Fonarch
Bepletion
Prading
Corg
Corth
Biserf
Inp
Poide
Appendix A-2 e List of prime stimuli for motion verb targets.
Prime stimuli
Motion
Scoot
Gnaw
Scuttle
Stroll
Lurch
Adhere
Flop
Growl
Scamper
Wobble
Condense
Escape
Spin
Dangle
Zigzag
Mangle
Chew
Linger
Elongate
Mount
Pursue
Gulp
Ramble
Jog
Struggle
Bustle
Poke
Evade
Squeeze
Snort
Target stimuli
Filler
Animal
Pseudoword
Motion verbs
Pseudoword
Amuse
Deem
Verify
Forge
Crave
Predict
Rile
Infer
Enrage
Loathe
Persuade
Conceive
Muse
Affirm
Ponder
Ruffle
Swear
Grieve
Construe
Mourn
Inspire
Fret
Sicken
Vex
Refresh
Adore
Deny
Speculate
Reassure
Repel
Hyena
Dragonfly
Ostrich
Jackal
Lamb
Worm
Hound
Tiger
Deer
Hamster
Caterpillar
Chipmunk
Gull
Gorilla
Cricket
Mantis
Shark
Lizard
Snapper
Koala
Cougar
Viper
Swine
Calf
Orangutan
Alpaca
Ant
Raccoon
Rattlesnake
Elephant
Bairg
Glop
Rowir
Prader
Pifth
Geborn
Kump
Ofary
Tling
Detter
Nistress
Commeter
Veep
Bisth
Troker
Bryet
Domud
Maraed
Comferts
Afoub
Rether
Halg
Cosple
Iro
Consilt
Fonfer
Gont
Follinate
Ristern
Taret
Scram
Devour
Hasten
Wander
Stagger
Cling
Tumble
Roar
Hurry
Jiggle
Shrink
Evacuate
Hover
Swing
Waggle
Carve
Munch
Dawdle
Stretch
Climb
Chase
Crunch
Promenade
Trot
Wrestle
Trudge
Sting
Sneak
Twist
Sniff
Toopy
Larr
Itvert
Insidu
Phiding
Roner
Maltef
Taru
Tissent
Laared
Casteb
Clieer
Twops
Polden
Reliek
Codmle
Bemi
Ponsul
Gelpers
Sivot
Huick
Recher
Tontinent
Tret
Attergen
Brofit
Angeb
Gapal
Vuint
Kweller
(continued on next page)
210
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
Appendix A-2 e (continued )
Prime stimuli
Motion
Stride
Strike
Stomp
Vibrate
Hustle
Stumble
Disappear
Clinch
Target stimuli
Filler
Pacify
Concede
Abide
Relieve
Devise
Analyze
Encourage
Concur
Animal
Puma
Chimp
Rhino
Penguin
Cheetah
Kangaroo
Hippo
Boa
references
Altmann, G. T. M., & Kamide, Y. (1999). Incremental interpretation
at verbs: restricting the domain of subsequent reference.
Cognition, 73, 247e264.
Assadollahi, R., Meinzer, M., Flaisch, T., Obleser, J., &
Rockstroh, B. (2009). The representation of the verb’s
argument structure as disclosed by fMRI. BMC Neuroscience,
10(1), 3.
Boland, J. (1993). The role of verb argument structure in sentence
processing: distinguishing between syntactic and semantic
effects. Journal of Psycholinguistic Research, 22, 133e152.
Boland, J. E., Tanenhaus, M. K., & Garnsey, S. M. (1990). Evidence
for the immediate use of verb control information in sentence
processing. Journal of Memory and Language, 29, 413e432.
Balota, D. A., Yap, M. J., Cortese, M. J., Hutchison, K. A., Kessler, B.,
Loftis, B., et al. (2007). The English lexicon project. Behavior
Research Methods, 39(3), 445e459.
Bornkessel, I., Zysset, S., Friederici, A. D., von Cramon, D. Y., &
Schlesewsky, M. (2005). Who did what to whom? the neural
basis of argument hierarchies during language
comprehension. NeuroImage, 26(1), 221e233.
Caplan, D., Stanczak, L., & Waters, G. (2008). Syntactic and
thematic constraint effects on blood oxygenation level
dependent signal correlates of comprehension of relative
clauses. Journal of Cognitive Neuroscience, 20(4), 643e656.
Chambers, C., & San Juan, V. (2008). Perception and
presupposition in real-time language comprehension: insights
from anticipatory processing. Cognition, 108, 26e50.
Copland, D. A., de Zubicaray, G. I., McMahon, K., Wilson, S. J.,
Eastburn, M., & Chenery, H. J. (2003). Brain activity during
automatic semantic priming revealed by event-related
functional magnetic resonance imaging. NeuroImage, 20(1),
302e310.
Delong, K., Urbach, T., & Kutas, M. (2005). Probabilistic word
preactivation during language comprehension inferred from
electrical brain activity. Nature Neuroscience, 8, 1117e1121.
Dikker, D., Rabagliati, H., Farmer, T. A., & Pylkkanen, L. (2010).
Early occipital sensitivity to syntactic category is based on
form typicality. Psychological Science, 21, 629e634.
Ferretti, T., McRae, K., & Hatherell, A. (2001). Integrating verbs,
situation schemas, and thematic role concepts. Journal of
Memory and Language, 44, 516e547.
Friederici, A. D. (2004). Event-related brain potential studies in
language. Current Neurology and Neuroscience Reports, 4,
466e470.
Friederici, A. D., Opitz, B., & Von Cramon, D. Y. (2000). Segregating
semantic and syntactic aspects of processing in the human
brain: an fMRI investigation of different word types. Cerebral
Cortex, 10(7), 698e705.
Pseudoword
Reters
Reeline
Thant
Rateral
Vopped
Tandits
Fonserves
Parver
Motion verbs
Tread
Thrash
Dash
Tremble
Sprint
Vault
Lurk
Embrace
Pseudoword
Utage
Dralas
Parg
Fupates
Coziep
Rujor
Kest
Phoking
Friston, K. J., Frith, C. D., Turner, R., & Frackowiak, R. S. (1995).
Characterizing evoked hemodynamics with fMRI. NeuroImage,
2(2), 157e165.
Gold, B. T., Balota, D. A., Jones, S. J., Powell, D. K., Smith, C. D., &
Andersen, A. H. (2006). Dissociation of automatic and strategic
lexical-semantics: functional magnetic resonance imaging
evidence for differing roles of multiple frontotemporal
regions. The Journal of Neuroscience, 26(24), 6523e6532.
Hahne, A., & Friederici, A. D. (1999). Electrophysiological evidence
for two steps in syntactic analysis: early automatic and late
controlled processes. Journal of Cognitive Neuroscience, 11,
194e205.
Henson, R. N. A. (2003). Neuroimaging studies of priming. Progress
in Neurobiology, 70(1), 53e81.
Hirotani, M., Makuuchi, M., Ruschemeyer, S.-A., & Friederici, A. D.
(2011). Who was the agent? The neural correlates of reanalysis
processes during sentence comprehension. Human Brain
Mapping, 32, 1775e1787.
Holmes, V. M. (1987). Syntactic parsing: in search of the
garden path. In M. Coltheart (Ed.), Attention and performance XII
(pp. 587e599). Hillsdale (NJ): Erlbaum.
Kotz, S. A., Cappa, S. F., von Cramon, D. Y., & Friederici, A. D.
(2002). Modulation of the lexical-semantic network by
auditory semantic priming: an event-related functional MRI
study. NeuroImage, 17(4), 1761e1772.
Kucera, H., & Francis, W. N. (1967). Computational analysis of
present-day American English. Providence: Brown.
Kukona, A., Fang, S.-Y., Aicher, K. A., Chen, H., & Magnuson, J. S.
(2011). The time course of anticipatory constraint integration.
Cognition, 119, 23e42.
Kuperberg, G. R., Holcomb, P. J., Sitnikova, T., Greve, D.,
Dale, A. M., & Caplan, D. (2003). Distinct patterns of neural
modulation during the processing of conceptual and syntactic
anomalies. Journal of Cognitive Neuroscience, 15(2), 272e293.
Landauer, T. K., & Dumais, S. T. (1997). A solution to Plato’s
problem: the latent semantic analysis theory of acquisition,
induction, and representation of knowledge. Psychological
Review, 104(2), 211e240.
Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of
lexical access in speech production. Behavioral and Brain
Sciences, 22, 1e38.
Levin, B. (1993). English verb classes and alternations: A preliminary
investigation. Chicago: University of Chicago Press.
Lund, K., & Burgess, C. (1996). Producing high-dimensional
semantic spaces from lexical co-occurrence. Behavior Research
Methods, Instruments, & Computers, 28(2), 203e208.
McNamara, T. P. (2004). Semantic priming: Perspectives from memory
and word recognition. Psychology Press.
McRae, K., Hare, M., Elman, J. L., & Ferretti, T. R. (2005). A basis for
generating expectancies for verbs from nouns. Memory and
Cognition, 33, 1174e1184.
c o r t e x 5 7 ( 2 0 1 4 ) 1 9 8 e2 1 1
Mumford, D. (1992). The role of cortico-cortical loops. Biological
Cybernetics, 66, 241e251.
Mummery, C. J., Shallice, T., & Price, C. J. (1999). Dual-process
model in semantic priming: a functional imaging Perspective.
NeuroImage, 9(5), 516e525.
Neely, J. H., Keefe, D. E., & Ross, K. L. (1989). Semantic priming in
the lexical decision task: roles of prospective prime-generated
expectancies and retrospective semantic matching. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 15(6),
1003.
Newman, S. D., Ikuta, T., & Burns, T. (2010). the effect of semantic
relatedness on syntactic analysis: an fMRI study. Brain and
Language, 113, 51e58.
Newman, S. D., Just, M. A., Keller, T. A., Roth, J. K., &
Carpenter, P. A. (2003). The differential effects of syntactic and
semantic processing on the two subregions of Broca’s area.
Cognitive Brain Research, 16, 297e307.
Obleser, J., & Kotz, S. A. (2010). Expectancy constraints in
degraded speech modulate the language comprehension
network. Cerebral Cortex, 20, 633e640.
Price, C. J., & Delvin, J. T. (2011). The interactive account of ventral
occipitotemporal contributions to reading. Trends in Cognitive
Sciences, 15, 246e253.
Raposo, A., Moss, H. E., Stamatakis, E. A., & Tyler, L. K. (2006).
Repetition suppression and semantic enhancement: an
investigation of the neural correlates of priming.
Neuropsychologia, 44, 2284e2295.
Rissman, J., Eliassen, J. C., & Blumstein, S. E. (2003). An eventrelated FMRI investigation of implicit semantic priming.
Journal of Cognitive Neuroscience, 15(8), 1160e1175.
Rossell, S. L., Price, C. J., & Nobre, A. C. (2003). The anatomy and
time course of semantic priming investigated by fMRI and
ERPs. Neuropsychologia, 41, 550e564.
Ruff, I., Blumstein, S. E., Myers, E. B., & Hutchison, E. (2008).
Recruitment of anterior and posterior structures in
lexicalesemantic processing: an fMRI study comparing
implicit and explicit tasks. Brain and Language, 105, 41e49.
Sachs, O., Weis, S., Zellagui, N., Sass, K., Huber, W.,
Zvyagintsev, M., Mathiak, K., & Kircher, T. (2011). How
211
different types of conceptual relations modulate brain
activation during semantic priming. Journal of Cognitive
Neuroscience, 23(5), 1263e1273.
Sakai, K. L., Tatsuno, Y., Suzuki, K., Kimura, H., & Ichida, Y. (2005).
Sign and speech: amodal commonality in left hemisphere
dominance for comprehension of sentences. Brain, 128(Pt 6),
1407e1417.
Schacter, D. L., Wig, G. S., & Stevens, W. D. (2007). Reductions in
cortical activity during priming. Current Opinion in Neurobiology,
17(2), 171e176.
Shapiro, L. P., Zurif, E., & Grimshaw, J. (1987). Sentence processing
and the mental representation of verbs. Cognition, 27, 219e246.
Shapiro, L. P., Gordon, B., Hack, N., & Killackey, J. (1993).
Verb-argument structure processing in complex sentences in
Broca’s and Wernicke’s aphasia. Brain and Language, 45,
423e447.
Shetreet, E., Palti, D., Friedmann, N., & Hadar, U. (2007). Cortical
representation of verb processing in sentence comprehension:
number of complements, subcategorization, and thematic
frames. Cerebral Cortex, 17(8), 1958e1969.
Slotnick, S. D., & Schacter, D. L. (2004). A sensory signature
that distinguishes true from false memories. Nature
Neuroscience, 7, 664.
Tanenhaus, M. K., Boland, J., Garnsey, S. M., & Carlson, G. N.
(1989). Lexical structure in parsing long-distance
dependencies. Journal of Psycholinguistic Research, 18, 37e50.
Thompson-Schill, S. L., D’Esposito, M., Aguirre, J. K., & Farah, M. J.
(1997). Role of the left inferior prefrontal cortex in retrieval of
semantic knowledge: a reevaluation. Proceedings of the National
Academy of Sciences USA, 9, 14792e14797.
Trueswell, J. C., Tanenhaus, M. K., & Kello, C. (1993). Verb-specific
constraints in sentence processing: separating effects of
lexical preference from garden-paths. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 19, 528e553.
Wheatley, T., Weisberg, J., Beauchamp, M. S., & Martin, A. (2005).
Automatic priming of semantically related words reduces
activity in the fusiform gyrus. Journal of Cognitive Neuroscience,
17(12), 1871e1885.