ArticlePDF Available

Abstract and Figures

Three experiments investigated factors contributing to syntactic priming during on-line comprehension. In all of the experiments, a prime sentence containing a reduced relative clause was presented prior to a target sentence that contained the same structure. Previous studies have shown that people respond more quickly when a syntactically related prime sentence immediately precedes a target. In the current study, ERP and eyetracking measures were used to assess whether priming in sentence comprehension persists when one or more unrelated filler sentences appear between the prime and the target. In experiment 1, a reduced P600 was found to target sentences both when there were no intervening unrelated fillers, and when there was one unrelated filler between the prime and the target. Thus, processing the prime sentence facilitated processing of the syntactic form of the target sentence. Experiments 2 and 3, eye-tracking experiments, showed that target sentence processing was facilitated when three filler sentences intervened between the prime and the target. These experiments show that priming effects in comprehension can be observed when unrelated material appears after a prime sentence and before the target. We interpret the results with respect to residual activation and implicit learning accounts of priming.
Content may be subject to copyright.
This article was downloaded by: [University of California Davis]
On: 20 August 2013, At: 13:25
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
Language and Cognitive Processes
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/plcp20
Evidence for priming across intervening sentences
during on-line sentence comprehension
Kristen M. Tooley a , Tamara Y. Swaab b , Megan A. Boudewyn b , Megan Zirnstein b &
Matthew J. Traxler b
a Psychology Department, Beckman Institute, University of Illinois, Urbana, IL, USA
b Psychology Department, Center for Mind and Brain, University of California, Davis, CA,
USA
Published online: 19 Apr 2013.
To cite this article: Language and Cognitive Processes (2013): Evidence for priming across intervening sentences during on-
line sentence comprehension, Language and Cognitive Processes, DOI: 10.1080/01690965.2013.770892
To link to this article: http://dx.doi.org/10.1080/01690965.2013.770892
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Evidence for priming across intervening sentences during
on-line sentence comprehension
Kristen M. Tooley
a
, Tamara Y. Swaab
b
, Megan A. Boudewyn
b
,
Megan Zirnstein
b
and Matthew J. Traxler
b
*
a
Psychology Department, Beckman Institute, University of Illinois, Urbana, IL, USA;
b
Psychology Department,
Center for Mind and Brain, University of California, Davis, CA, USA
(Received 31 May 2012; final version received 18 January 2013)
Three experiments investigated factors contributing to syntactic priming during on-line comprehension. In all of the
experiments, a prime sentence containing a reduced relative clause was presented prior to a target sentence that
contained the same structure. Previous studies have shown that people respond more quickly when a syntactically
related prime sentence immediately precedes a target. In the current study, ERP and eye-tracking measures were used
to assess whether priming in sentence comprehension persists when one or more unrelated filler sentences appear
between the prime and the target. In experiment 1, a reduced P600 was found to target sentences both when there
were no intervening unrelated fillers, and when there was one unrelated filler between the prime and the target. Thus,
processing the prime sentence facilitated processing of the syntactic form of the target sentence. In experiments 2 and
3, eye-tracking experiments showed that target sentence processing was facilitated when three filler sentences
intervened between the prime and the target. These experiments show that priming effects in comprehension can be
observed when unrelated material appears after a prime sentence and before the target. We interpret the results with
respect to residual activation and implicit learning accounts of priming.
Keywords: syntax; priming; sentence processing; syntactic priming
The nature of the mechanism behind structural prim-
ing deserves more attention.
Kay Bock and Zenzi Griffin
This article focuses on the nature of the mechanism
behind structural priming. Structural priming, also
known as syntactic priming or structural persistence,
describes the outcome when exposure to one sentence
influences the processing of a subsequent sentence that
shares aspects of grammatical form (or syntactic
structure; Bock 1986; Hartsuiker & Kolk, 1998; Pick-
ering & Branigan, 1998; Pickering & Traxler, 2004;
Tooley et al., 2009; Traxler, 2008a, 2008b; Traxler &
Pickering, 2005). As Bock and colleagues, (2007, p.
439), note, ‘Structural persistence is a tendency to echo
syntactic structures from recent experience, despite
changes in the meaning, in the wording, (or) even in
the language embodying the persistent structure’.
Syntactic priming effects have been observed in both
production and comprehension. In production, expo-
sure to a prime sentence of a given syntactic type leads
speakers to use that type with greater frequency in the
future (Bock, 1986; Hartsuiker, Bernolet, Schoonbaert,
Speybroeck, & Vanderelst, 2008; Pickering & Branigan,
1998). In experiments involving picture description, for
example, hearing or reading a passive-voice sentence
(The church was struck by lightning) makes it more
likely that a participant will produce a passive-voice
form to describe an unrelated picture (compared to
trials on which the participant had heard an active
voice sentence). Hearing an active-voice sentence
(Lightning struck the church) makes it less likely that
a passive-voice sentence will be forthcoming.
1
In
comprehension, the influence of exposure to a prime
sentence can be seen in the processing and final
interpretation of a subsequent sentence. Such influ-
ences can be seen during on-line construction of an
interpretation using methods such as eye-tracking
during reading, visual-world eye-tracking, and ERPs
(Arai, Van Gompel, & Scheepers, 2007; Ledoux,
Traxler, & Swaab, 2007; Thothathiri & Snedeker,
2008a, 2008b; Tooley, Traxler, & Swaab, 2009; Traxler
& Tooley, 2008). In behaviour, exposure to a syntacti-
cally related prime sentence leads to faster reading of a
target sentence, usually at or shortly after comprehen-
ders encounter a syntactic choice point. In ERPs,
similar exposure results in a reduced positivity at and
following syntactically disambiguating words in target
*Corresponding author. Email: mjtraxler@ucdavis.edu
Language and Cognitive Processes, 2013
http://dx.doi.org/10.1080/01690965.2013.770892
#2013 Taylor & Francis
Downloaded by [University of California Davis] at 13:25 20 August 2013
sentences. Exposure to prime sentences also influences
the final outcome of the interpretation process in
globally ambiguous sentences (Branigan, Pickering, &
McLean, 2005).
In comprehension priming experiments, speeded
processing of target sentences or modulations of
ERPs could reflect processing of various kinds, lexical,
semantic, referential, or syntactic, among others. Pre-
vious studies have therefore attempted to localize the
source of observed priming effects. These studies
indicate that syntactic processes contribute substan-
tially to observed priming effects, with semantic over-
lap playing a weaker role that is localized differently
than the syntactic effects in these experiments (see
Tooley & Traxler, 2010, for a review). Different
sentence types may be differentially affected by factors
such as lexical overlap between prime and target
sentences (Pickering, McLean, & Branigan, 2012;
Traxler, 2008a; Sturt, Keller, & Dubey, 2010), and
some facilitatory effects may be related to lexical
repetition or semantic processes, but the weight of
evidence suggests that syntactic processes, including re-
parsing, contribute to the observed effects.
Evidence for syntactic contributions to facilitatory
effects can be found in both reading time and ERP
outcomes. For example, prime sentences that have the
same overall abstract structure, such as (2) and (3), but
not completely parallel structure at a more local level
will facilitate processing of target sentences such as (1)
(Pickering, Tooley, & Traxler, 2011; Pickering &
Traxler, 2004; Traxler, 2008b; Traxler & Pickering,
2005; Traxler & Tooley, 2008):
(1) The defendant examined by the lawyer turned out
to be unreliable.
(2) The engineer who was examined by the board
passed with flying colours.
(3) The engineer who was examined passed with
flying colours.
In eye-tracking experiments, these facilitated proces-
sing effects are consistently localized to the syntacti-
cally disambiguating by-phrase (by the lawyer in 1),
although repetition priming effects at the verb (e.g.,
examined) also occur in some experiments. In ERP
experiments, facilitated syntactic processing following
prime sentences like (4) results in reduced positivities
during processing of syntactically disambiguating by-
phrase (by the board,by the lawyer) words in target
sentences, such as (1) (Ledoux et al., 2007; Tooley et
al., 2009):
(4) The engineer examined by the board passed with
flying colours.
Note that priming effects for reduced-relatives have
previously been observed only when the same verb
appears in both the prime and target sentence.
Reduced positivities in sentence processing studies
are typically referred to as the P600 effect. P600
modulations were first observed in experiments that
exclusively assessed syntactic manipulations (e.g., vio-
lations of agreement and word order, syntactic ambi-
guity, and syntactic complexity (see Kutas, van Petten,
& Kluender, 2006, for a review). More recently, P600
effects have also been observed when syntactic infor-
mation conflicts with semantic information in the
sentence (e.g., thematic violations in syntactically
unambiguous sentences; Kim & Osterhout, 2005;
Kuperberg, Caplan, Sitnikova, Eddy, & Holcomb,
2006; Nakano, Saron & Swaab, 2010). Importantly,
many ERP studies have shown P600 effects indicating
syntactic processing difficulties in syntactically ambig-
uous sentences such as those tested here.
Further, facilitated processing of syntactically dis-
ambiguating material in sentences like (1) is not
observed when the verb in the prime sentence is only
semantically related to the verb in the target. In the
Tooley et al. (2009) study, sentences like (5) did not
facilitate processing of targets like (1):
(5) The engineer inspected by the board passed with
flying colours.
In addition, in Traxler and Tooleys (2008) study,
establishing a semantic relationship between prime
and target sentences by repeating a subject noun led
to repetition priming of the subject noun, but no
priming effects occurred during processing of the
syntactically disambiguating by-phrase.
Finally, in experiments involving modifier-goal am-
biguities, such as (6), syntactically disambiguating
parts of target sentences were processed more rapidly
than normal even when the prime sentence (e.g., 7) was
semantically unrelated to the target:
(6) The girl dropped the blanket on the floor on the
bed this morning.
(7) The historian read the letter to the Romans to the
class last week.
These results suggest that priming effects observed at
and following syntactically disambiguating material in
comprehension (especially in reduced relatives, which
have received the most attention to date), are driven by
syntactic representations and processes, as opposed to
purely semantic or lexical ones. This is not to say that
semantic processes could not facilitate processing of
by-phrases in such sentences, but rather that in the
2K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
paradigm used in the current study, syntactic processes
are the most likely source of the priming effects.
Syntactic priming occurs in both production and
comprehension, but discussion continues as to the
precise mechanisms and processes that give rise to
such effects. Identifying the mechanisms and processes
that trigger syntactic priming in comprehension could
help us better understand how lexical and syntactic
representations are organized. It could also help us
understand how comprehension processes compare to
production processes. We return to these issues in the
General Discussion.
How persistent is syntactic priming in comprehension?
The three experiments reported here were designed to
test whether syntactic priming of reduced relative
clauses can be observed when one or more unrelated
sentences appear between the prime sentence and the
target. To date, syntactic priming that persists across
unrelated intervening sentences has been commonly
observed in production (Bock & Griffin, 2000; Hart-
suiker et al., 2008), whether or not specific lexical items
appear in both the prime and target utterances. Much
less is known about the existence and nature of similar
effects in on-line comprehension. However, recent
studies have shown that repeated exposure to a given
sentence type can have long-lasting facilitative effects
(Long & Prat, 2008; Tooley, 2009; Wells, Christiansen,
Race, Acheson, & MacDonald, 2009). In addition,
preliminary data from the visual world paradigm
indicate that a prime sentence may affect processing
of a target sentence when a single unrelated sentence
intervenes between the prime and target (Carminati &
van Gompel, 2009). Finally, in two ERP studies,
syntactic priming effects, as indicated by modulation
of the amplitude of the P600, were observed when a
comprehension question intervened between the prime
and the target sentence (Ledoux et al., 2007; Tooley et
al., 2009).
Finding out what happens when unrelated sentences
intervene between a prime and a target can help
constrain accounts of syntactic priming in comprehen-
sion, as it has done in production. Although we have
data from massed exposure studies (Long & Prat, 2008;
Wells et al., 2009), we do not know what happens in
comprehension under conditions that more closely
resemble typical production priming experiments. If
priming effects in comprehension do not survive at lags
greater than zero, this would suggest a qualitative
difference between production and comprehension in
terms of the mechanisms and processes that produce
priming.
Therefore, the main goal of this study was to
determine what happens during sentence comprehen-
sion when unrelated sentences appear between a prime
sentence and a target sentence. We used both event-
related potential (ERP) and behavioural (eye-tracking)
methods to assess comprehendersresponse to sen-
tences containing reduced relative clauses, such as
sentence (1):
(1) The defendant examined by the lawyer turned out
to be unreliable.
Such sentences have been shown to produce highly
robust syntactic priming effects during on-line com-
prehension in lag-zero experiments (that is, experiments
in which the target sentence immediately follows the
prime sentence; Tooley et al., 2009; Traxler & Tooley,
2008). In eye-tracking, priming effects are observable
as reduced fixation times on the syntactically disam-
biguating by-phrase region [by the lawyer in (1)], and
reduced total reading times on the verb region that
precedes the by-phrase region [examined in (1)]. In ERP
experiments, reductions in syntactic processing diffi-
culty in the target sentence manifest as reductions in
the P600 to critical disambiguating words that follow a
repeated verb in the target sentence, when the electro-
physiological response to the target sentence is com-
pared to the electrophysiological response to the prime
sentence. Presumably, these P600 effects reflect the
difficulty incorporating the syntactic form of the by-
phrase with the preceding context. Standard processing
accounts suggest that readers anticipate a noun-phrase
(as in The defendant examined the evidence) rather than
a prepositional-phrase/reduced relative clause modifier.
The clash between the likely or syntactically simpler
anticipated outcome and the actual outcome initiates
syntactic repair processes, giving rise to the P600.
Reductions in the N400 at the repeated verb have
also been observed, but contrast in polarity with effects
at and following syntactic disambiguation.
Experiment 1: ERPs at lag 1
The first experiment investigated how the presence of a
prime sentence influenced the neurophysiological re-
sponse to a target sentence that had the same syntactic
form. Consistent with prior ERP studies of syntactic
priming in comprehension, we tested participants
response to reduced-relative sentences, such as (1)
above (Ledoux et al., 2007; Tooley et al., 2009).
Participantsresponses were assessed in two different
conditions. The first condition, the lag zero condition,
closely resembled the procedures used in the previous
ERP studies, in that the prime and target sentences
Language and Cognitive Processes 3
Downloaded by [University of California Davis] at 13:25 20 August 2013
were immediately adjacent. In the other condition, the
lag one condition, one unrelated filler sentence inter-
vened between the prime sentence and the target. By
measuring ERPs to the critical verb and the disambig-
uating words following the verb in the prime and target
sentences, we could assess how processing the prime
sentence affected the neurophysiological response to a
subsequent target sentence. If processing the prime
sentence facilitates syntactic structure-building pro-
cesses in the target, we should observe a reduced
P600 to the disambiguating words that follow the
repeated verb in the targets, compared to the primes.
Method
Participants. Sixteen undergraduates from the Uni-
versity of California, Davis, gave informed consent and
took part in the study and all participants were
compensated with course credit. All were right-handed,
native speakers of English, with no reported hearing
loss or psychiatric/neurological disorders.
Stimuli. The experimental items were 160 reduced
relative clause sentences, comprising 80 prime-target
pairs. Prime sentences were either directly followed by
target sentences (lag zero condition), or else primes and
targets were separated by one intervening filler sentence
(lag 1 condition). All prime-target sentence pairs
contained the same verb:
(Prime) The defendant examined by the lawyer
turned out to be unreliable.
(Only Lag1: Filler) The tour guide led the people
around the museum slowly.
(Target) The engineer examined by the board passed
with flying colours.
Primes and targets were counterbalanced across four
lists, such that each prime sentence in one list served as
a target sentence in another list. Versions of items were
assigned to one of four lists using a Latin-square
design. The items were presented to participants in a
fixed, randomized order (so that all participants went
through a given list of items in the same order). In
addition, prime-target pairs presented with no inter-
vening sentence in one list were presented with an
intervening sentence in another list. For example, the
sample prime listed above would serve as a prime in the
lag zero condition in one list, as a prime in the lag one
condition in another list, as a target in the lag zero
condition in the third list, and finally as a target in the
lag one condition in the last list. Each participant was
tested on one of the four lists, with equal numbers of
participants tested in each. Importantly, this counter-
balancing permitted the comparison across prime and
target items of the exact same sentences, thereby
controlling for any possible differences in length or
frequency between individual items.
In addition to the experimental items, there were 125
filler sentences, 40 of which served as intervening
sentences for lag one condition prime-target pairs.
2
Each block began with a filler sentence so that muscle-
movement artefacts that often occur early in the block
(as the participant settles in) would not impact experi-
ment items. The filler sentences were of variable
syntactic structure, and contained a mixture of sen-
tence types and subject matter. For example, some filler
sentences had a main clause structure (The professor
finished his lecture five minutes early), while others
contained relative clauses (The drummer that criticized
the guitarist quit the band). Some fillers contained
prepositional phrases (The family decorated the tree
with silver and gold ornaments). No verbs used in the
filler items were repeated in the prime-target pairs. We
assumed, on the basis of prior results (Tooley et al.,
2009; Traxler & Tooley, 2008), that prime sentences
themselves would not be primed by exposure to prior
sentences containing reduced relatives. This assump-
tion was based on the absence of priming at lag 0 when
reduced-relative prime and target sentences did not
have the same verb.
Procedure
Participants were tested in a comfortable chair in an
electrically shielded, sound-attenuating booth. The
sentences were presented one word at the time using
Presentation software, and participants were instructed
to read the sentences and to respond after each trial
with a button press, either to advance to the next
sentence or to indicate true or false in response to a
comprehension question. Twenty-five comprehension
questions were included per list. The true or false
questions did not follow prime sentences or, in the case
of the lag one condition, filler sentences that appeared
between prime-target pairs. Participants did not receive
feedback on their responses. Trials began with a white
fixation cross appearing for 1000 ms against a black
background in the centre of a computer screen
approximately 100 cm in front of the subjects. The
fixation cross was replaced by the first word of the
sentence; all sentences were presented at a rapid serial
visual presentation rate of 300 ms per word, with a 200
ms inter-stimulus interval. White horizontal bars
appeared above and below words as a fixation aid.
Five-hundred milliseconds following the offset of the
final word, which was presented with a period, either
the sentence Press for nextor a true or false
comprehension question appeared all at once on the
screen. Participants responded using a keypad, with
4K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
accuracy ranging from 80100%, averaging 90.5%
across all participants. Each of the four lists was
broken into nine blocks for presentation, each contain-
ing approximately 32 sentences.
Participants were asked to keep their eyes fixated on
the centre of the screen and to refrain from blinking or
moving when the fixation cross was visible and during
the presentation of the sentences. This was done to
minimize subject-generated artefacts in the EEG sig-
nal. Subjects were told that they could blink and move
their eyes when the Press for nextor comprehension
questions were on the screen. Stimulus codes specific to
each condition were sent at the onset of the critical
words in the sentence and these codes were used for
later off-line averaging of the EEG.
ERP recording and data reduction. EEG was re-
corded from 29 tin electrodes, mounted in an elastic
cap (ElectroCap International); additional electrodes
were placed on the outer canthi and below the left eye
in order to monitor eye movements and blinks. The
outer canthi electrodes were referenced to each other,
and the electrode under the left eye was referenced to
channel FP1 (above the left eye) in the electrode cap.
The right mastoid electrode was used as the recording
reference for all EEG electrodes and the left mastoid.
The left mastoid was recorded for later off-line
algebraic re-referencing. The EEG signal was amplified
with band pass cut-offs at 0.01 and 30 Hz, and digitized
on-line at a sampling rate of 250 Hz (Neuroscan
Synamp I). EEG was digitized continuously along
with accompanying stimulus codes used for subsequent
averaging. Impedances were kept below 5 kV.
Prior to off-line averaging, all single-trial waveforms
were screened for amplifier blocking, muscle artefacts,
horizontal eye movements and blinks over epochs of
1600 ms, starting 200 ms before the onset of the critical
target words. Averaged ERPs were computed over
artefact-free trials and were filtered off-line with a
Gaussian low-pass filter with a 25 Hz half-amplitude
cut-off. Statistical analyses were conducted on the
filtered data.
We report results for four words, the reduced-
relative verb (e.g., examined in 1), and the following
three words, by,determiner, and noun (e.g., by the lawyer
in 1).
Results and discussion
Figure 1 shows the N400 effect of repetition at the verb
for both the lag zero and the lag one conditions (e.g.,
examined in The defendant examined by ...). Figure 2
shows the comparison of the waveforms to the word by
in the lag zero and lag one conditions (e.g., by,inThe
defendant examined by ...). Figure 3 shows the priming
effects at the determiner and noun for both the lag zero
and lag one conditions (e.g., the lawyer in ...defendant
examined by the lawyer ...). The baselines used to
create the figures were identical to those used for the
statistical analysis described below.
Analyses were conducted on the mean amplitude of
the verbs and the critical words of the by-phrase (by,
determiner, noun) that followed the verbs in the prime
and target sentences. Separate analyses were conducted
in epochs corresponding to the 300500 ms (N400) and
500800 ms (P600) time windows following critical
word onset. In order to avoid distortion of the effects
due to baseline differences between conditions coming
from effects at the verb, the analysis for the critical
word by was conducted time-locked to the verb, using a
pre-verb baseline. Therefore, the N400 (300500 ms
after verb) and P600 (500800 ms after verb) time
windows for by corresponded to the 8001000 ms and
10001300 ms epochs when time-locked to the verb.
Likewise, in order to account for baseline distortion
coming from effects at the word by, analyses for the
determiner and noun in the by-phrase were conducted
time-locked to by, using a pre-by baseline. Therefore,
the N400 (300500 ms after determiner/noun) and
P600 (500800 ms after determiner/noun) time win-
dows for the determiner correspond to the 8001000
ms and 10001300 ms epochs when time-locked to by,
and for the 13001500 ms and 15001800 ms epochs
for the noun when time-locked to by
3
.
Repeated measures ANOVAs were performed with
Lag (lag zero, lag one), Type (Prime, Target), and
Electrode site (26 sites) as within-subjects factors.
Additional ANOVAs over a subset of 11 central-
posterior electrodes (C3, C4, CP1, CP2, CP5, CP6,
P3, P4, CZ, PZ, POZ) where the N400 and P600 are
typically maximal were conducted with the same
factors. The GreenhouseGeisser correction was ap-
plied to Ftests with more than one degree of freedom
in the numerator for all relevant analyses reported in
this paper. ANOVA tables for the main effects and
interactions of sentence type (prime vs. target), results
described below are summarized in Table 1.
Verb Results. A main effect of Type was found in the
N400 window to the verb (F(1,15) 5.43; p0.034),
such that waveforms to target verbs were reduced in
amplitude (less negative) compared to prime verbs. No
effect of Lag was found (FB1), nor were any interac-
tions significant (FsB1.4). No P600 effects were
obtained to the verb (FsB2).
By Results. No effects of either Lag or Type were
found in either the N400 window (FsB1.7), or in the
P600 window (FsB1.9) to the word by.
Determiner Results. A main effect of Type was found
for the determiner in the N400 time window (F16.63;
Language and Cognitive Processes 5
Downloaded by [University of California Davis] at 13:25 20 August 2013
p0.001), such that waveforms to determiners in the
prime sentences were more positive than those in target
sentences. This effect of Type also interacted with
Electrode (F4.23; p0.0042). No effects of Lag or
any interactions with Lag were significant at the
determiner (FsB1) in the N400 time window. Simi-
larly, a main effect of Type was found for the
determiner in the P600 window (F9.53; p0.007),
such that determiners in the prime conditions were
more positively deflected than those in the target
conditions. As in the earlier window, this effect of
Type also interacted with Electrode site (F4.32; p
0.0018), and no effects of Lag or interaction with Lag
were obtained (FsB1).
Noun Results. A trend towards a main effect of Type
was found for the noun in the N400 time window (F
3.93; p0.066), such that waveforms to nouns in the
prime sentences were more positive than those in target
sentences. The effect of Type significantly interacted
with Electrode (F4.3; p0.0008). No effects of Lag
or any interactions with Lag were significant at the
noun (FsB1) in the N400 time window. Similarly, a
main effect of Type was found for the determiner in the
P600 window (F7.12; p0.018), such that determi-
ners in the prime conditions were more positively
deflected than those in the target conditions. As in
the earlier window, this effect of Type also interacted
with Electrode site (F2.8; p0.02), and no effects of
Lag or interaction with Lag were obtained (FsB1).
CentralPosterior Electrode Site Results. The subset
analyses over 11 centralposterior electrode sites re-
vealed the same pattern of results to those described
above. An N400 effect of repetition priming was found
at the verb (F8.8; p0.009), but again there was no
effect of Lag (FsB1.4). Significant effects of syntactic
priming were found to the determiner in both the early
C3 CZ C4
P3 PZ P4
CP1 CP2
C3 CZ C4
P3 PZ P4
CP1 CP2
Prime
Target
800
400
800
400
Figure 1. ERP effects of verb repetition for prime target pairs in the zero lag (top) and lag 1 (bottom) conditions. Verbs in prime sentences
appear in blue, and verbs in targets sentences appear in dashed red. Negative is plotted up. The N400 time window appears shaded in light green,
whiled the P600 time window is shaded in darker green.
6K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
(F20.74; p0.0004) and late time windows (F
10.92; p0.0048), these effects also interacted with
electrode (F4.1; p0.0055; F5.15; p0.0014);
there was no effect of Lag in either time window
(FsB1). Finally, there was again a main effect of
syntactic priming at the noun in both the early (F
5.37; p0.035) and late (F8.85; p0.0094_ time
windows, which interacted with electrode (F4.09; p
0.0117; F3.18; p0.0328). There was additionally a
Lag x Electrode interaction in the N400 time window at
the noun (F3.48; p0.023).
Simple effects analysis
In order to examine the effects of priming for each lag
condition, repeated measures ANOVAs were per-
formed with Type (Prime, Target), and Electrode site
(26 sites) as within-subjects factors, separately for each
Lag condition (see Table 2).
Lag Zero Verb Results: The effect of Type did not
reach significance at the verb in either the N400 or
P600 time windows at lag zero, nor did it interact with
electrode (FsB1).
Lag Zero By Results: The effect of Type did not
reach significance at the verb in either the N400 or
P600 time windows at lag zero, nor did it interact with
electrode (FsB1).
Lag Zero Determiner Results: A main effect of Type
was found for the determiner in the N400 time
window (F7.13; p0.0175). This effect of Type
marginally interacted with Electrode (F2.54; p
0.0668). Similarly, a main effect of Type was found
for the determiner in the P600 window (F6.72; p
by : Lag Zero Condition
C3 CZ C4
P3 PZ P4
CP1 CP2
C3 CZ C4
P3 PZ P4
CP1 CP2
Prime
Tar get
800
400 1000
800
400 1000
by : Lag 1 Condition
by
Onset
by
Onset
Figure 2. Comparison of the critical word byin prime (blue) compared to target (dashed red) sentences, in the zero lag (top) and lag 1 (bottom)
conditions. ERPs are time-locked to the preceding verb; the xaxis represents a 1400 ms epoch after stimulus onset. Onset of byis at 500 ms.
Negative is plotted up. The N400 time window appears shaded in light green, whiled the P600 time window is shaded in darker green.
Language and Cognitive Processes 7
Downloaded by [University of California Davis] at 13:25 20 August 2013
0.0204), which did not interact with electrode site (FB
2.04).
Lag Zero Noun Results. The main effect of Type did
not emerge as significant at the noun in the N400 time
window, nor did it interact with electrode in this time
window (FsB2.44). The effect of Type was significant
in the P600 time window (F4.73; p0.0461), and did
not interact with electrode (FB1.4)
Lag One Verb Results: A main effect of Type was
found in the N400 window to the verb (F6.41; p
0.0231), such that waveforms to target verbs were
reduced in amplitude (less negative) compared to prime
verbs. This effect of Type did not interact with
electrode (FB1.5). No P600 effects were obtained to
the verb (FsB1).
Lag One By Results. No effects of Type were found
in either the N400 window (FsB1), or in the P600
window (FsB1) to the word by.
Lag One Determiner Results. A main effect of Type
was found for the determiner in the N400 time window
(F7.15; p0.0173), such that waveforms to determi-
ners in the prime sentences were more positive than
those in target sentences. This effect of Type did not
interact with Electrode (FB1.7). A main effect of Type
was also found for the determiner in the P600 window
(F6.51; p0.0222), such that determiners in the
prime conditions were more positively deflected than
those in the target conditions. As in the earlier window,
this effect of Type also interacted with Electrode site
(F2.5; p0.0379).
C3 CZ C4
P3 PZ P4
CP1
C3 CZ C4
P3 PZ P4
CP1 CP2
Prime
Target
Determiner
Onset
Noun
Onset
Determiner
Onset
Noun
Onset
O
O
nse
t
CP2
004008 1200 1600
-
+
2µV
004008 1200 1600
-
+
2µV
Figure 3. Comparison of the determiner in prime (blue) compared to target (dashed red) sentences, in the zero lag (top) and lag 1 (bottom)
conditions. ERPs are time-locked to the preceding word (by); the xaxis represents an 1800 ms epoch after stimulus onset. Onset of the
determiner is at 500 ms; onset of the noun is at 1000ms. Negative is plotted up. The N400 time windows appear shaded in light green, while the
P600 time windows are shaded in darker green.
8K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
Table 1. Experiment 1: Main effects of lag (lag zero vs. lag one), sentence type (prime vs. target), and interactions with electrode
site for Experiment 1. The first row represents the N400 epoch (300500 ms post-stimulus). The second row represents the P600
epoch (500800 ms post-stimulus).
All-Electrode Results
VERB
Epoch Lag Type L x T L x E T x E L x T x E
F p F p F p F p F p F p
N400 <1 0.5
5.43 0.05
<1 0.4 1.34 0.3 <1 0.4 <1 0.5
P600 1.73 0.2 <1 0.79 <1 1.0 <1 0.4 <1 0.5 1.25 0.3
BY
Epoch Lag Type L x T L x E T x E L x T x E
F p F p F p F p F p F p
N400 1.7 0.2 <1 0.8 <1 0.5 <1 0.5 <1 0.4 1.48 0.2
P600 <1 0.4 <1 0.6 <1 0.6 1.9 0.1 <1 0.7 1.14 0.3
DETERMINER
Epoch Lag Type L x
T
L x
E
T x
E
L x T x
E
F p F p F p F p F p F p
N400 <1 0.4
16.6 0.001 <1 0.9 <1 0.8 4.23 0.01
<1 0.9
P600 <1 0.6
9.53 0.01 <1 0.9 <1 0.6 4.32 0.01
<1 0.9
NOUN
Epoch Lag Type L x
T
L x
E
T x
E
L x T x
E
F p F p F p F p F p F p
N400 3.21 0.09 4.06 0.06 <1 0.7 <1 0.8 2.86 0.05
1.00 0.4
P600 <1 0.75 <1 0.4 <1 0.7 <1 0.8 1.13 0.35 <1 0.5
11-Electrode Subset Results
VERB, SUBSET
Epoch Lag Type L x
T
L x
E
T x
E
L x T x
E
F p F p F p F p F p F p
N400 <1 0.9
8.8 0.01
<1 0.4 1.42 0.3 1.21 0.3 1.01 0.4
P600 <1 0.4 <1 0.5 <1 1.0 <1 0.8 2.32 0.1 1.45 0.3
Language and Cognitive Processes 9
Downloaded by [University of California Davis] at 13:25 20 August 2013
Lag One Noun Results. No effects of Type were
found in either the N400 window (FsB2.08), or in the
P600 window (FsB3) to the noun.
Repetition of the verb in the target sentences led to a
reduction of the N400 relative to the first presentation
of the verb in the primes. The size of this effect did not
differ as a function of the lag manipulation. This
replicates findings of Ledoux et al. (2007), and Tooley
et al. (2009).
Importantly, P600 effects were obtained across both
lag conditions at the determiner and the noun of the
critical by-phrase: Determiners and nouns in target
sentences showed reduced P600 amplitudes compared
to those observed in prime sentences. This effect was
also significant in the N400 time window. However, this
difference most likely reflects an early positive shift
found for determiners and nouns in prime sentences
compared to those in target sentences, rather than a
canonical N400 effect. A larger negative shift for the
determiners/nouns in prime sentences compared to
targets would have been consistent with an N400 effect.
However the current results show a larger positive shift
for the prime sentences. Although this effect begins in
the N400 time window, we interpret our results at the
determiner as reflecting the beginning of an extended
positivity (P600). The determiner comes at a point
during the processing of the sentence that is one word
after the reader has encountered the structural diffi-
culty (i.e., just following by), and therefore it is not
surprising that the onset of the effect at the determiner
is relatively early for a P600.
In summary, the results indicate that processing of a
reduced relative target sentence is affected by the
presence of a reduced relative prime sentence. Reduced
positivity following presentation of the determiner,
whether the target sentence followed the prime im-
mediately or followed one unrelated intervening sen-
tence, provided the strongest indication of such an
influence. The priming effect in the lag zero condition
replicates similar effects in the two previous ERP
studies of syntactic priming in comprehension for this
sentence type (Ledoux et al., 2007; Tooley et al., 2009).
The persistence of such effects in the lag one condition
has not previously been observed, with the possible
Table 1 (Continued )
DETERMINER, SUBSET
Epoch Lag Type L x
T
L x
E
T x
E
L x T x
E
F p F p F p F p F p F p
N400 <1 0.5
20.7 0.001 <1 0.8 <1 0.7 4.1 0.01
<1 0.8
P600 <1 0.8
10.9 0.005 <1 0.8 <1 0.9 5.15 0.001
<1 0.9
NOUN, SUBSET
Epoch Lag Type L x T L x E T x E L x T x E
F p F p F p F p F p F p
N400 2.64 0.1 2.69 0.1 <1 0.7 <1 0.6 3.5 0.05
1.1 0.3
P600 <1 0.4 <1 0.5 <1 1.0 <1 0.8 1.6 0.2 <1 0.6
BY, SUBSET
Epoch Lag Type L x T L x E T x E L x T x E
F p F p F p F p F p F p
N400 1.46 0.3 <1 0.5 1.16 0.3 <1 0.7 1.3 0.3 1.9 0.2
P600 <1 0.5 <1 0.8 <1 0.7 <1 0.5 1.6 0.2 1.6 0.2
Note: For the main analyses, degrees of freedom for main effects are (1,15) and for interaction terms are (25,375). For the subset analyses, degrees
of freedom for main effects are (1,15) and for interaction terms are (10,150). N400 refers to the 300500 ms post-stimulus epoch. P600 refers to the
500800 ms post-stimulus epoch. Statistically significant results appear in bold.
10 K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
exception of Ledoux and colleaguesand Tooley and
colleaguesexperiments. In those ERP experiments, a
comprehension question followed every sentence in the
experiment. Thus, these experiments might be viewed
as showing a priming effect at lag one as well. However,
because the comprehension question always referred to
the immediately preceding sentence, it may have served
to preserve or refresh a trace of aspects of the prime
sentence, including its syntactic form. This cannot
explain the results of the present experiment, since an
unrelated structure intervened between prime and
target sentences in the lag 1 condition.
It is also worth noting that the P600 effect was found
in different words in the disambiguating by-phrase
that followed the repeated verb in Ledoux et al. (2007),
Tooley et al (2009), and the present study. In Ledoux
and colleagues(2007) study, modulations of the ERP
signal appeared to the noun within the by-phrase (e.g.,
lawyer in example 1). In Tooley and colleagues(2009)
study, similar effects appeared to the preposition by, the
earliest point at which syntactic disambiguation could
occur. In the current experiment, the P600 effect was
found to the determiner (the) in the by phrase, as well
as the following noun. Given that the present experi-
ment was designed to determine if syntactic priming
Table 2. Experiment 1: Simple effects comparisons, testing
the effect of sentence type (prime vs. target), and interactions
with electrode site for each lag condition in Experiment 1.
The rst row represents the N400 epoch (300500 ms post-
stimulus). The second row represents the P600 epoch (500
800 ms post-stimulus).
LAG ZERO: VERB
Epoch Type T x E
F p F p
N400 <1 0.35 <1 0.51
P600 <1 0.87 1.55 0.21
LAG ZERO: BY
Epoch Type T x E
F p F p
N400 <1 0.5 1.47 0.23
P600 <1 0.52 1.03 0.4
LAG ZERO: DETERMINER
Epoch Type T x E
F p F p
N400 7.13 0.02
2.54 0.07
P600 6.72 0.02
2.04 0.12
LAG ZERO: NOUN
Epoch Type T x E
F p F p
N400 2.44 0.14 1.96 0.13
P600 4.73 0.04
1.38 0.26
LAG ONE: VERB
Epoch Type T x E
F p F p
N400 6.41 0.02
1.47 0.24
P600 <1 0.86 <1 0.54
Table 2 (Continued )
LAG ONE: DETERMINER
Epoch Type T x E
F p F p
N400 7.15 0.02
1.7 0.16
P600 6.51 0.02 2.5 0.04
LAG ONE: NOUN
Epoch Type T x E
F p F p
N400 1.88 0.19 2.08 0.09
P600 2.95 0.11 1.5 0.2
LAG ONE: BY
Epoch Type T x E
F p F p
N400 <1 0.64 <1 0.4
P600 <1 0.9 <1 0.59
Language and Cognitive Processes 11
Downloaded by [University of California Davis] at 13:25 20 August 2013
effects survive the presentation of intervening sentences
with a different structure, variability of the precise
point in the sentence at which the modulation of the
ERPs occurs is not of great significance. Variability in
the precise timing of effects across three ERP experi-
ments is comparable to the variability that has been
observed in behavioural studies of this sentence type. In
many of the eye-tracking studies that used this para-
digm (e.g., Tooley et al., 2009, Experiment 2; Traxler &
Tooley, 2008), effects emerge in first-pass reading time
or first-pass regressions from within the by-phrase (by
the lawyer in example 1). However, such effects some-
times emerge only in measures like total time, which
can be affected by relatively lateinterpretive events.
Furthermore, the eye-tracking region of analysis in-
cludes the word by, the determiner, and the following
noun, and so does not differentiate between the
different effects on these three words. It is therefore
possible that a more fine-grained analysis of the eye-
tracking data in the previously mentioned studies
would reveal an analogous jittering in the exact timing
of the effect. Future studies may identify what drives
the exact onset of the syntactic priming effect.
For our purposes, the critical finding is that a
reduced P600 was found to the determiner and follow-
ing noun in the disambiguating by phrase of the target
sentences compared to the determiner and following
noun in the same region in the primes), and that the
reduction in the P600 amplitude was about the same in
the lag one trials as the lag zero trials. These outcomes
suggest that the beneficial effects of processing the
prime are not noticeably reduced by the presence of a
single unrelated filler sentence. It could be the case that
inserting the filler sentence delays the onset of the
priming effect. This possibility seems less likely, how-
ever, because the timing of the priming effects appeared
to be about the same in the current experiment in the
lag zero and lag one trials. While we cannot rule out all
differences between processing of the target sentences
between the lag zero and lag one conditions on the
basis of this single set of results, the results do suggest
that the broad outlines of the target-sentence response
are similar in the lag zero and lag one conditions.
Experiment 2: eye-tracking at lag 3
The ERP experiment provided evidence that syntactic
priming effects during the comprehension of reduced
relative target sentences are of similar magnitude
whether prime and target sentences are immediately
adjacent, or whether a single unrelated filler sentence
separated the prime from the target. Our second
experiment, which used eye-tracking rather than
ERPs to assess readersresponses, was designed to
further test whether priming effects in comprehension
survive across unrelated material. This experiment also
used reduced relatives as primes and targets. To test
whether priming effects could survive at longer lags,
primes and targets were separated by three unrelated
sentences. This experiment has the potential to provide
converging evidence for persistent syntactic priming in
the comprehension of sentences containing reduced
relative clauses.
Method
Participants. A total of 34 undergraduates from
UC Davis participated to complete a course require-
ment.
Stimuli. There were 72 items like (1), repeated
here:
(1) The defendant examined by the lawyer turned out
to be unreliable.
Each participant read a total of 72 sentences like (1),
36 as primes and 36 as targets.
Within each prime-target pair, the target was
separated from the prime by three unrelated sentences.
A trial thus consisted of the presentation of a prime
sentence, three unrelated filler sentences, and a target
sentence. For example, a trial would appear like this
(except each sentence was displayed on a computer
screen by itself):
(Prime) The defendant examined by the lawyer
turned out to be unreliable.
(Filler 1) The tour guide led the people around the
museum slowly.
(Filler 2) The pastry chef invented a new kind of
cookie dough.
(Filler 3) The dogs played together all day at the dog
park.
(Target) The engineer examined by the board passed
with flying colours.
By rotating items across lists, each prime sentence (on
one list) served as a target sentence (on a different list).
Thus, prime-target comparisons were made across the
same sentences. This eliminates concerns about the
length and frequency of the individual words as well as
possible effects of additional ambiguities beyond the
reduced relative-main clause ambiguity. The estimates
of the priming effects were based on measuring readers
responses to sentences like The defendant examined by
the lawyer proved to be unreliable when it appeared as a
prime and when it appeared as a target. Three
12 K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
intervening filler sentences with an unrelated structure
appeared between the prime and target sentences.
Counterbalancing of sentences across lists accom-
plished the following goals: 1) Every prime sentence
(on one list) served as a target sentence (on a different
list). 2) A participant read any particular sentence
exactly once. After experimental sentences were ran-
domly assigned to one of two lists, the experimental
and filler sentences were presented to participants in a
fixed order.
The experimental sentences were displayed along
with 122 filler sentences of various types, most of them
with simple transitive or intransitive constructions (for
example, The player tossed the bat at the fence when he
struck out). Thus, all of the fillers represented relatively
commonly occurring sentence types with fairly straight-
forward syntactic properties. Some consisted of con-
joined main clauses (The worker framed the house and
supervised the sheet rock hangers). Some had verbs with
infinitival complements (The students wanted summer
vacation to start). Some of the fillers had prepositional
phrase modifiers (The sister wore the dress with the big
blue bow on the hip.) (All of the stimulus lists are
available from the senior author upon request: mjtrax-
ler@ucdavis.edu) At least one filler sentence intervened
between each experimental trial. None of the filler
items had verbs that were used in the prime-target
pairs.
Eye-movement monitoring procedure. A Forward
Technologies dual-Purkinje image eye-tracker moni-
tored participantseye movements while they read the
sentences. The tracker has angular resolution of 10of
arc. The tracker monitored only the right eyes gaze
location. A PC displayed sentence materials on a
computer monitor seventy cm from participantseyes.
The participantsgaze location was sampled every
millisecond and the PC software recorded the trackers
output to establish the sequence of eye fixations and
their start and finish times. Before the experiment, the
participant was seated at the eye tracker and positioned
with a chin cup and head rests to minimize head
movements, and then the tracker was aligned and
calibrated. Participants were instructed to press a
button as soon as they had finished reading each
sentence. After 22 of the filler sentences, the participant
responded to a comprehension question. Participants
did not receive feedback on their responses. The
average percent correct was 93.7% (standard
deviation 5.98%), with a minimum of 77% and a
maximum of 100%. Between each trial, a pattern of
squares appeared on the computer screen along with a
cursor that indicated the participantscurrent gaze
location. If the tracker was out of alignment, it was
recalibrated before proceeding with the next trial.
Analyses. Four standard eye-movement measures
were computed for each participant: 1) First-pass time
is the sum of all fixation durations beginning with the
first fixation in a region until the readers gaze leaves
the region, left or right; 2) First-pass regression reflects
eye-movements that cross a regions left-hand bound-
ary immediately following a first-pass fixation; 3)
Regression-path time includes all of the fixations
from the first fixation in a scoring region until the
reader fixates somewhere to the right of the scoring
region (Konieczny, Hemforth, Scheepers, & Strube,
1997; Traxler, Bybee, & Pickering, 1997). This measure
includes time spent re-fixating previous regions; 4)
Total time is the sum of all fixation durations in a
region, regardless of order.
Three scoring regions were analysed. The verb
region consisted of the verb inside the relative clause
(e.g., examined in 1). The PP region consisted of the
prepositional phrase (e.g., by the lawyer). The PP
region was analysed as a whole because it is a
linguistically defined unit that has been used in
previous priming studies using the eye-tracker (e.g.,
Tooley et al., 2009; Traxler & Tooley, 2008), and
previous studies involving the reduced relative sentence
type that did not involve priming manipulations (e.g.,
Clifton, et al., 2003). In addition, as in prior eye-
tracking studies on this sentence type, we analysed a
post-PP region that consisted of the two words
immediately following the end of the PP region.
Prior to determining fixation durations, an auto-
matic procedure incorporated fixations of less than 80
ms into the largest fixation within one character. In the
next stage, the procedure eliminated all individual
fixations greater than 1000 ms and less than 80 ms.
Subsequently, first pass times and total times of less
than 120 ms were excluded from the analyses. Further,
first-pass or total times exceeding 2000 ms for a given
scoring region were excluded.
The data were analysed using hierarchical linear
modeling (HLM; Raudenbush & Bryk, 2002; Traxler,
Williams, Blozis, & Morris, 2005). In the particular
application used here, HLM functions very similarly to
repeated measures ANOVA, the chief difference being
the use of random slopes and intercepts for each
individual unit of analysis (participants and items; for
a fuller discussion of similarities and differences
between ANOVA and HLM, see Blozis & Traxler,
2007). Reading times were modelled as a function of
sentence characteristics (whether the sentence appeared
in the list as a prime sentence or a target). Participants
had up to 72 responses (36 prime-target pairs).
Separate models were run for each dependent measure
for each scoring region. At the first level of the model,
outcomes (dependent measures) were considered a
function of condition (prime vs. target). At the second
Language and Cognitive Processes 13
Downloaded by [University of California Davis] at 13:25 20 August 2013
level of the model, parameter slopes and error terms
were allowed to vary freely between individuals.
Finally, a second set of models was run in which RTs
were considered as nested within items rather than
participants.
4
Results and discussion
First pass, first-pass regressions, regression path time,
and total time results for the verb and PP scoring
regions are summarized in Table 3.
No statistically significant effects were observed in
any scoring region in the first-pass, first-pass regres-
sions, or regression-path time data. However, signifi-
cant effects of condition (prime vs. target) were
obtained, as follows:
Verb Region. Total time in the verb region was 19 ms
shorter in the lag 3 targets than in the corresponding
primes. This effect was significant by subjects [t
1
(33)
2.48, pB0.05], but marginal by items [t
2
(71) 
1.75, 0.05 BpB0.10]. Priming effects in total time at
the verb are not accompanied by effects in other
measures (first-pass time, first-pass regressions, or
regression-path time) that presumably reflect earlier
processing events. Thus, in this experiment, as in most
other eye-tracking experiments of this kind of sentence,
priming effects did not begin to appear until after
readers fixated the disambiguating PP region.
PP Region. Total time in the PP region was 32 ms
shorter in the lag 3 targets than in the corresponding
primes. This effect was significant by both participants
and items [t
1
(33) 2.90, pB0.01; t
2
(71) 2.01, pB
0.05].
Post-PP Region: Total time in the spillover region
was 35 ms shorter in the lag 3 targets than in the
primes. This effect was significant by both participants
and items [t
1
(33) 3.53, pB0.005; t
2
(71) 2.58, p
0.01].
Supplemental Analyses. As noted above, we assume
based on prior results that little or no priming occurs
from one trial to the next. If processing a reduced-
relative prime and target on one trial influences
processing of the prime sentence on the following trial,
these eye-tracking results may actually under-estimate
the absolute magnitude of the priming effects. A
correlation analysis revealed a non-significant negative
correlation between serial position within the experi-
ment and total reading time for the disambiguating
region within the prime sentences (r0.12, NS).
This non-significant effect could relate to some small
between-trial priming effect or overall facilitation of
reduced relative clause processing, but it could just as
well reflect adjustment to the eye-tracking procedure.
In addition, an analysis of reading times for the
disambiguating region of the prime and target sen-
tences revealed no relationship between serial position
within the experiment and the size of the priming effect
(r0.011, NS). Hence, neither the baseline reading
time nor the size of the priming effect was related to
position within the experimental list of items.
These results indicate that exposure to a reduced
relative prime sentence can facilitate processing of a
reduced relative target sentence even when three
unrelated filler sentences appear between the prime
and the target. The eye-movement patterns suggest that
the priming effects occur only after participants
encounter the syntactically disambiguating by-phrase.
Experiment 3: lag 3 with main clause fillers
Experiments 1 and 2 indicated that priming effects in
sentence comprehension can occur when the prime and
target sentences are separated by up to three unrelated
sentences. Experiment 3 was designed to test whether a
filler sentence that had the same verb as the prime and
target but a different syntactic structure would disrupt
the priming effects. We therefore manipulated proper-
ties of the filler sentences. As in Experiment 2, three
filler sentences intervened between reduced relative
prime and target sentences. As in Experiment 2, the
prime and target sentences shared a verb (because these
are the conditions under which priming has been
shown to occur for the reduced relative sentence
type). However, unlike Experiment 2, in Experiment
3, the second of the three filler sentences had the same
verb as the prime and target, but had a main clause
structure, as in (8):
Table 3. Experiment 2. Mean rst-pass, rst-pass regres-
sions, regression-path time, and total reading time by scoring
region and condition.
Verb Region PP Region Spillover
First Pass Time
Prime 329 (8.11) 512 (23.1) 418 (15.6)
Target 319 (5.08) 504 (8.6) 412 (12.4)
First Pass Regressions
Prime 9.95% (1.29) 13.0 (1.61) 12.9 (1.15)
Target 9.95% (1.29) 11.03 (1.37) 12.1 (1.16)
Regression Path Time
Prime 378 (11.9) 580 (24.1) 526 (19.0)
Target 370 (9.24) 565 (10.4) 527 (10.7)
Total Time
Prime 410 (11.7) 628 (25.6) 525 (20.4)
Target 391
$
(7.7) 596* (11.2) 490** (9.81)
Note: Standard errors appear in parentheses. *indicates the
parameter differs from zero at pB.05 in both by-participants and by-
items analyses; **indicates by-participants and by-items pB.01. (*)
indicates .05 BpB.10.
14 K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
(8) The tourist examined the map before entering the
museum.
In previous studies (at lag zero) sentences such as (8)
have had no effect on the processing of reduced relative
target sentences (Pickering & Traxler, 2004; Traxler &
Pickering, 2005; Traxler & Tooley, 2008).
Predicting what effect this filler manipulation will
have depends on assumptions about the nature of
syntactic representations and assumptions about how
those syntactic representations are deployed during on-
line comprehension. We will return to this topic in
greater detail in the General Discussion, but briefly: If
syntactic structure representations are accessed via
lexical entries (MacDonald, Pearlmutter, & Seidenberg,
1994; Boland & Blodgett, 2006; Vosse & Kempen,
2000; Pollard & Sag, 1994), and if the most recent
experience with a lexical item exerts the greatest
influence on syntactic processing, then the amount of
priming that occurs in Experiment 3 should be
substantially less than that observed in Experiment 2.
In Experiment 3, at the point in time where readers
encounter the target sentence, their most recent experi-
ence with the critical verb (e.g., examined) will involve a
syntactic structure that is different than the structure
required for the target sentence. If priming survives
under these conditions, then some influence other than
the most recent experience must be driving that effect.
In prior experiments where a main clause prime
sentence has immediately preceded a reduced relative
target sentence, no priming effects have been observed.
Thus, priming under these conditions would indicate
that the less recent prime sentence is exerting a greater
influence on processing than the more recent filler
sentence.
Determining why a more distant prime sentence
should have a greater influence than a more recent filler
sentence again requires some assumptions about how
priming works. One possibility, raised in previous
studies, is that rarer syntactic structures respond
differently to priming manipulations compared to
more frequently occurring structures (Segaert, Menen-
ti, Weber, & Hagoort, 2011; Segaert, Menenti, Weber,
Petersson, & Hagoort, 2012). Some production prim-
ing studies find differential susceptibility to priming
across alternating forms, and this may be related to
their relative frequencies (Bock, 1986; Luka & Barsa-
lou, 2005; Pickering, Branigan, & McLean, 2002). In
this experiment, sentences with reduced relative clauses
served as primes and targets, while main clause
sentences (like 8) served as fillers. These forms differ
greatly in their overall frequency, as main clause
constructions appear far more often than reduced
relatives. Undertaking the mental processes necessary
to parse and interpret a reduced relative may, therefore,
have a greater persistent effect on performance than
processing a main clause sentence. If so, the main
clause filler sentences will not disrupt the effects
resulting from processing of the reduced relative prime
sentence.
A trial in Experiment 3 consisted of a reduced
relative prime sentence, a reduced relative target
sentence, and three filler sentences. Two of the three
filler sentences were identical to those used in Experi-
ment 2. However, the second of the three fillers was
changed so that it had the same verb as the prime and
target sentences, but a main clause structure, rather
than a reduced relative clause structure. A trial there-
fore consisted of the following five elements (with
example sentences):
(1) Prime Sentence: The defendant examined by the
lawyer was convicted yesterday.
(2) (Filler 1): The tour guide led the people around
the museum slowly.
(3) Filler 2 (Main Clause Filler): The cook examined
the tomatoes before making the soup.
(4) (Filler 3): The dogs played together all day at the
dog park.
(5) Target: The engineer examined by the board
passed with flying colours.
Method
Participants. A total of 48 UC Davis undergradu-
ates participated under the same terms as Experiment
2.
Stimuli. The prime and target stimuli were iden-
tical to Experiment 2. The second of the three filler
sentences from that experiment were replaced by a
sentence that had the same verb as the prime and
target, but a main clause structure (as in 8, above).
Apparatus and procedure. The eye-tracking and
data analysis procedures were identical to Experiment
2.
Results and discussion
Table 4 presents mean values for the four dependent
measures by region (verb, PP, spillover) and condition
(prime vs. target). 5.4% of the data were excluded from
the analyses using the same criteria as Experiment 2.
Participants averaged 95% correct on the comprehen-
sion questions (range 69100%).
Verb Region. No priming effect was observed in the
first-pass, first-pass regressions, or regression-path time
data from the verb region. However, total time was
shorter in the targets than in the primes [403 vs. 426 ms;
t
1
(47) 2.82, pB.01; t
2
(71) 2.70, pB0.01]. These
Language and Cognitive Processes 15
Downloaded by [University of California Davis] at 13:25 20 August 2013
results closely resemble the results for the verb region
from the preceding Experiment 2.
PP Region. Significant priming effects were ob-
served in the PP region in first-pass time (590 vs. 633
ms) by participants [t
1
(47) 2.13, pB0.05] but not by
items [t
2
(71) 0.76, NS]. Fewer first-pass regressions
occurred in the PP region in the target sentences than
the primes [10.8% vs. 12.8%; t
1
(47) 1.98, pB0.05;
t
2
(71) 1.93, pB0.06]. Like first-pass time, the regres-
sion path time data produced a significant priming
effect in the by-participants analysis but not in the by-
items analysis [578 vs. 618 ms; t
1
(47) 3.50, pB0.002;
t
2
(71) 1.50, p0.14]. Analysis of the total time data
indicated significant priming in the PP region [t
1
(47) 3.10, pB0.01; t
2
(71) 1.98, pB0.05].
Post-PP Region. The only significant effect in the
post-PP region appeared in the first-pass regressions
data. There were fewer first-pass regressions from the
post-PP region in the target sentences than in the
primes [t
1
(47) 1.97, p0.05; t
2
(71) 2.11, pB0.05].
Supplemental Analyses. As in Experiment 2, we
conducted two correlational analyses to assess whether
baseline reading time or the size of the priming effects
changed across the course of the experiment. A
correlation analysis revealed a non-significant negative
correlation between serial position within the experi-
ment and total reading time for the PP-region within
the prime sentences (r0.07, NS). In addition, an
analysis of reading times for the PP-region of the prime
and target sentences revealed no relationship between
serial position within the experiment and the size of the
priming effect (r0.12, NS). Hence, as in Experi-
ment 2, neither the baseline reading time nor the size of
the priming effect was related to position within the
experimental list of items.
Cross-Experiment Comparison. Each trial in Experi-
ments 2 and 3 differed only in the content of the second
of three filler sentences. In Experiment 2, the second
filler was random. In Experiment 3, the second filler
had the same verb as the target sentence and the prime,
but a different syntactic structure. We performed a
further set of analyses to assess the degree of similarity
in the results of Experiments 2 and 3. To do so, we
analysed total time data from the disambiguating by-
phrase (the PP-region), the main locus of the priming
effects. These data were subjected to multilevel models
with Experiment (2 vs. 3), condition (prime vs. target),
and their interaction as random factors. These analyses
produced no main effect of Experiment (tsB1, NS),
and no interaction of Experiment with condition (ts B
1, NS). There was a main effect of condition [t
1
(81)
3.25, pB0.01, standard error 9.82; t
2
(71) 2.92, p
0.03, standard error 14.4]. Thus, the analyses did not
reveal any substantial differences in the results across
the two experiments.
Similar to Experiment 2, this experiment produced
evidence for facilitated target processing occurring
after participants had read three filler sentences, one
of which shared a verb with the target but had a
different syntactic structure. Previous studies have
shown that the main clause filler, by itself, is ineffective
at facilitating processing of reduced relative target
sentences (Ledoux et al., 2007; Traxler & Tooley,
2008). Thus, priming effects at the target must have
been driven by exposure to the prime.
General discussion
In Experiment 1, ERPs were measured to critical words
in temporarily ambiguous target sentences that were
preceded by structurally identical prime sentences. The
primes and targets were either immediately adjacent
(lag zero condition) or separated by one unrelated filler
sentence (lag one condition). Both in the lag zero and
the lag 1 conditions, a modulation of the N400 was
found to the repeated verb in the target relative to the
prime sentences. This effect was of similar magnitude in
both lag conditions and indicates facilitated processing
as a function of lexical repetition. However, in the
disambiguating by-phrase following the verb, facili-
tated processing in the target sentence was instead
observed on the P600; a reduced P600 was found to the
determiner in the target sentences, even when there was
an intervening filler sentence with a different structure.
This P600 effect also did not differ in size as a function
of the lag manipulation. Finally, a reduction in the
amplitude of the N400 was found to the noun in the by
Table 4. Experiment 3. Mean rst-pass, rst-pass regres-
sions, regression-path time, and total reading time by scoring
region and condition.
Verb Region PP Region Spillover
First Pass Time
Prime 338 (8.49) 511 (17.9) 419 (13.1)
Target 331 (4.91) 493 (8.44) 418 (7.61)
First Pass Regressions
Prime 9.3% (1.00) 12.8 (1.42) 13.7 (1.19)
Target 8.0% (0.97) 10.8* (1.00) 11.5* (1.12)
Regression Path Time
Prime 403 (14.3) 618 (24.8) 486 (17.6)
Target 384 (9.86) 578 (11.5) 479 (11.8)
Total Time
Prime 426 (12.2) 633 (23.9) 514 (20.2)
Target 403** (8.29) 590* (13.9) 494 (10.2)
Note: Standard errors appear in parentheses. *indicates the
parameter differs from zero at pB.05 in both by-participants and by-
items analyses; **indicates by-participants and by-items pB.01. (*)
indicates .05 BpB.10.
16 K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
phrase following the verb in the target sentence in both
lag zero and lag 1 conditions. The change in the nature
of the electrophysiological signature of facilitated
processing of the repeated verb (N400), the adjacent
words that follow in the disambiguating region (P600),
and the noun that follows the disambiguating region
(N400) suggests that the nature of the facilitated
processing in the target sentence changes as more
information becomes available to the reader. The
N400 modulation to the noun in the target sentence
suggests that semantic integration processes after
syntactic disambiguation are facilitated as well.
One potential problem with the ERP experiment is
the fact that words were presented one at the time (to
prevent eye-movement artefacts in the ERP signal),
and at a fixed rate, which could interfere with natural
reading processes. However, the results of the two eye-
tracking experiments of the present study validate the
ERP findings, since faster reading times (in the eye-
tracking experiments) and modulations of the N400
and P600 occurred at similar positions in the sentences.
This suggests that sub-optimal reading conditions did
not alter the on-line reading process in a significant
way. Thus, a reduction in the amplitude of the P600 to
the determiner in the disambiguating by-phrase in the
target sentences relative to the primes supports the
conclusion that participants had less difficulty proces-
sing the reduced relative structure when it was preceded
by the same syntactic structure. To our knowledge, this
is the first demonstration of syntactic priming when
target sentences follow primes after an unrelated filler
sentence with a different structure.
The ERP experiment was followed by two related
eye-tracking experiments. In both eye-tracking experi-
ments, target sentences were separated from the primes
by three filler sentences. In Experiment 2, none of the
filler sentences repeated the verb, and the results
showed that total time in the verb, PP, and spillover
scoring regions were shorter in the target sentences
than in their corresponding prime sentences. These
findings indicate that processing the prime sentence
facilitated processing of the target, despite the presence
of unrelated filler material.
In Experiment 3, the prime and target sentences
were also separated by three filler sentences. However,
in contrast to experiment 2, the second of the three
fillers contained the same verb as the prime and target,
but had a main clause structure, rather than a reduced
relative. Despite this intervening main clause filler, the
target sentence had shorter total time in the verb and
PP scoring regions, as well as fewer first-pass regres-
sions from the spillover region. Thus, the presence of a
main clause filler with the same verb as the target did
not eliminate the processing benefit resulting from
exposure to the reduced relative prime sentence.
Before interpreting these results further, we must
first address two recurring issues in syntactic priming
studies, namely the potential influence of participant
strategies and potential semantic influences on target
processing.
It is possible that because of repeated exposure to
sentences containing reduced relatives, or exposure to
prime-target pairs that have the same verb, participants
developed a strategy that allowed them to more rapidly
interpret the target sentences. A number of empirical
findings make this hypothesis less likely, however: 1)
Syntactic priming effects in eye-tracking occur in the
absence of reliable strategic cues (Pickering & Traxler,
2004; Traxler, 2008a; Traxler & Pickering, 2005). 2)
Syntactic priming effects in eye-tracking experiments
are as large on the first trial, and on the first pair of
trials, as they are in the experiments as a whole (Tooley
et al., 2009). Similarly, there was no relationship
between serial order in the experiment and the
magnitude of priming effects in the current Experi-
ments 2 and 3. These two types of results show that
strategic cues are not necessary for priming effects to
occur. 3) Syntactic priming fails to occur in the
presence of valid strategic cues (Tooley et al., 2009).
For example, repeating the subject noun rather than the
critical verb does not lead to facilitated processing of
the disambiguating PP region. 4) Explicitly warning
participants to anticipate the immediate arrival of a
reduced relative clause does not lead to facilitated
processing of the disambiguating PP region (Traxler &
Tooley, 2008). These latter findings indicate that
strategic cues are not sufficient for priming to occur.
If strategic cues are neither necessary nor sufficient for
syntactic priming to occur during on-line sentence
comprehension, then the explanation for the observed
priming effects must lie elsewhere.
In principle, facilitated processing of target sen-
tences in this study could reflect speeded semantic
processing (akin to the kinds of processes that cause
the word hat to prime the word glove in word priming
studies; Balota et al., 2007). This semantic priming
hypothesis has been addressed empirically in previous
articles (Ledoux et al., 2007; Traxler, 2008a; Traxler &
Tooley, 2008). Most recently, Tooley and colleagues
(2009) showed that priming effects for reduced relatives
occurred when verbs were repeated across primes and
targets, but not when the verb in the prime sentence
was a synonym of the verb in the target sentence, in
both ERP and eye-tracking experiments. This experi-
ment showed that, for reduced relatives at least, a
semantic relationship between the prime and target is
not sufficient for facilitated processing to occur. In
Traxler and Tooleys (2008) article, semantic overlap
between subject nouns produced repetition priming of
the subject noun but no observable effects on proces-
Language and Cognitive Processes 17
Downloaded by [University of California Davis] at 13:25 20 August 2013
sing of the relative clause. In the two published ERP
experiments involving reduced relatives, decreased
P600 amplitude (rather than N400-like effects) point
towards syntactic, rather than semantic, effects. The
P600 is very sensitive to syntactic ambiguity, and even
though the P600 effect has also been found in proces-
sing situations where there is a thematic or selection-
restriction violation (...the eggs were eating ...), this
is not the case in the kinds of sentences that were tested
here. Further, in experiments with repeated verbs, but
without repeated structure (as in main clause primes),
there is no evidence for facilitated processing, especially
in the relative clause region (the PP-region) that is the
locus of the effects obtained in the current and previous
experiments. In addition, facilitated processing of the
PP region of reduced relatives occurs when reduced-
relative target sentences follow full relatives(The
defendant who was examined by the lawyer ...) and
short relatives(The defendant who was examined was
convicted ...). These results indicate that the abstract
relative clause structure, rather than semantic or
function-word overlap, is capable of facilitating proces-
sing of reduced relative target sentences, suggesting
again that the facilitatory behavioural effects and the
ERP results observed with these types of structures are
driven by syntactic rather than semantic processes.
So, while we can point to a substantial body of
experimental evidence indicating that facilitated pro-
cessing reflects syntactic rather than semantic pro-
cesses, we know of no evidence that semantic overlap
produces similar outcomes. In our view, this state of
affairs lays the burden of proof on advocates of the
semantic hypothesis.
Implicit learning and comprehension priming
To be maximally parsimonious, an account of syntactic
priming in comprehension should also accommodate a
wide range of findings from production (such as the
rapid extinction of the lexical boost; Hartsuiker et al.,
2008). It is always possible that the details of processing
during comprehension differ from production, due to
differences in inputs and outputs, the kinds of goals
comprehension and production systems pursue, and
the time courses over which they operate. We will
therefore attempt to articulate an account for compre-
hension priming, and consider it an added bonus if the
account successfully accommodates the full range of
production priming phenomena as well.
For comprehension priming, the two most salient
outcomes that need explanation are the lag priming
effects (Experiments 13 here) and the lexical boost
(Arai et al., 2007; Carminati & van Gompel, 2009;
Pickering & Traxler, 2004; Pickering et al., 2011; Tooley
et al., 2009; Traxler & Tooley, 2008). The mere
existence of priming across intervening material would
appear to support some form of implicit learning
during on-line sentence interpretation (Chang, Dell,
Bock, & Griffin, 2000, Chang, Dell, & Bock, 2006;
Luka & Barsalou, 2005). That is, an account under
which exposure to a reduced relative strengthens the
structure building processes for that syntactic structure
would predict that processing of subsequent sentences
with the same form will be speeded up, whether the
target sentence appears immediately or after a delay.
This, by itself, could account for relatively long-lived
priming effects.
However, these lag effects might simply reflect the
slow decay of activation within combinatorial nodes
associated with lemma representations (Pickering et al.,
2002). In the limit, slow decaying activation would be
indistinguishable from implicit learning. However, we
believe that residual activation amongst lemma-combi-
natorial node representations does not provide a
satisfactory explanation for the outcome of the current
Experiment 3. In Experiment 3, the nearest related
stimulus to the target had a main clause, rather than a
reduced relative structure. Residual activation from this
sentence should have reduced the accessibility of the
reduced relative syntactic form. Thus, the more distant
related stimulus, the reduced relative prime, appeared
to exert greater influence on target processing in the
current Experiment 3 than the more recently encoun-
tered main clause filler. This outcome would seem to
largely rule out residual activation, even slowly decay-
ing residual activation, as the source of the observed
priming effects.
Nevertheless, not all of the present findings neatly
align with an implicit learning mechanism. For exam-
ple, if participants immediately strengthen the repre-
sentation of the reduced relative syntactic form via
exposure and implicit learning, one would expect faster
processing of that form during the experiment with
increased exposure. Hence, reading times for reduced-
relative sentences towards the end of the experiment
should have been faster than reading times at the
beginning of the experiment. The correlational analyses
did not detect significant reductions in overall proces-
sing time for reduced-relative sentences across the
course of the experiment, so this prediction was not
borne out.
5
Further, other aspects of the current findings may
provide support for the residual activation account.
First, the absolute magnitude of the priming effects in
the two eye-tracking experiments are about half as big
as the priming effects that have been obtained on nearly
identical stimuli in five previous eye-tracking experi-
ments (Pickering et al., 2011; Traxler & Tooley, 2008;
Tooley et al., 2009). In these previous eye-tracking
studies, the average total time effect in the PP region
18 K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
was 86 ms, with a range of 46130 ms. Thus, it is
possible that the syntactic priming effects that we
obtained in the eye-tracking experiments at lag three
are smaller than the syntactic priming effects that
occurred when similar sentences were tested at lags of
zero. If this were the case, the difference could be
attributed to the decay of residual activation with lag
(comparable to Hartsuiker et al.s, 2008, production
results). Also, the precise timing of the onset of effects
may have differed between the current (lag 3) and
previous (lag 0) studies. Previous eye-tracking studies
have produced relatively early priming effects (as
evidenced by reductions in first-pass times during
processing of the PP region), while the current experi-
ments showed effects only in total time. In addition, the
P600 effects of syntactic priming have been found to
different words in the critical disambiguating by-phrase
following the repeated verb, also indicating that the
time course of priming effects may vary. Perhaps
residual activation is responsible for earlier appearance
of priming effects in previous lag zero experiments.
Concluding that patterns of priming really differ
between lags of zero, one, and three may be premature,
however. We only have three estimates of priming
effects across lags of one (from ERP) and three (from
eye-tracking), and a maximum of seven estimates of
zero-lag priming effects (2 ERP and 5 eye-tracking
experiments, depending on how one views the two
preceding ERP experiments; Ledoux et al., 2007;
Tooley et al,. 2009). It may be too early, therefore, to
strongly conclude that priming effects are smaller at
lags of one or three than they are at lag zero. However,
if they were, this would seem to indicate that residual
activation plays a role over and above the influence of
implicit learning in producing syntactic priming effects,
at least at lags of zero. Minimally, we would need some
principled explanation for why priming effects are
bigger at lag zero than at greater lags.
Certainly, the residual activation and implicit learn-
ing mechanisms need not be mutually exclusive. The
existence of one need not rule out the possible existence
of the other. In fact, some accounts of syntactic
priming explicitly incorporate separate mechanisms to
deal with different experimental phenomena (Chang et
al., 2006; Hartsuiker et al., 2008). Tooley and collea-
gues incorporated this assumption into their dual-
mechanism account of syntactic priming in comprehen-
sion (Tooley, 2009; Tooley & Traxler, 2010; see also
Traxler & Tooley, 2008). Like Hartsuiker and collea-
gues(2008) account, the dual mechanism account
incorporates aspects of both the combinatorial node
and implicit learning accounts. According to this
account, exposure to a prime sentence temporarily
boosts the activation of the relevant combinatorial
nodes. The combinatorial node that encodes the actual
syntactic relationships that underlie the sentences final
interpretation ends up with the highest degree of
activation. When the input is parsed and the stimu-
lus-appropriate combinatorial node has been selected,
the connection between the lemma and the selected
combinatorial node is strengthened.
6
When a target
sentence is encountered immediately following a prime
sentence, syntactic priming effects reflect the temporary
increased activation of the combinatorial nodes shared
between the prime and target as well as changes in
representational strength caused by implicit learning. If
a target sentence follows a prime at some distance,
because unrelated sentences intervene, priming effects
can still be observed based on changes in the strength
of the connection between a lemma and its correspond-
ing combinatorial nodes as a function of implicit
learning. In production, this means that the structure
is more likely to be selected, or that production
latencies will be reduced. In comprehension, this means
that the relevant structure is more rapidly activated or
constructed.
How might the dual-mechanism account be imple-
mented? A distinction between short-lived residual
activation and longer-lived implicit learning mechan-
isms may be rooted in different memory systems, as
proposed by some language learning accounts (e.g.,
Ullman, 2001).
7
Another possibility is that implicit
learning and residual activation apply to different
components of the memory systems that encode lexical
and syntactic structure knowledge.
8
Assume for the
moment that word-related knowledge, including syn-
tactic structure possibilities, is encoded in a distributed
fashion. (Figure 4 shows lemma and syntactic structure
knowledge as localist representations, but this is not
necessary and, in fact, the dual mechanism account
could work just as well with distributed lexical and
syntactic representations.) Assume that the verb exam-
ined has associations with the structural possibilities
labelled in the Syntactic Structurescomponent of
Figure 4. Comprehending a reduced relative clause
involving the verb examined (The defendant examined
by the lawyer ...) would temporarily enhance the
connection between the lemma representation and the
reduced relative pattern within the syntactic structure
representations. This enhanced connectivity could be
short-lived. Implicit learning processes might simulta-
neously facilitate the procedures used to arrive at the
reduced relative analysis within the syntactic represen-
tation system. This enhanced procedural facilitation
could be longer lived. If so, the system could produce a
short-lived lexical boost and a longer-lasting general
syntactic priming effect. The degree to which lexically
independent priming could occur would be influenced
by the relative strength of the different syntactic
structure representations. In the case of reduced
Language and Cognitive Processes 19
Downloaded by [University of California Davis] at 13:25 20 August 2013
relatives, they may be so dominated by other structural
possibilities (transitive and passive, for example) that
the combination of lexical boost and implicit learning
may be required before a facilitated response (priming
effect) can be observed when a single prime sentence
precedes a target. Other structures, where the possibi-
lities are more balanced, may be more amenable to
lexically independent priming and may show different
patterns of priming across intervening material. This is
potentially rich ground for future empirical research
and modeling.
One advantage that various dual mechanism ac-
counts share is that they can deal with a broader range
of phenomena than either residual activation or
implicit learning can by themselves. For instance,
Chang and colleagues (2006) note that their implicit
learning model does not produce the lexical boost when
it attempts to simulate certain production priming
results, The dual-path model ...does not exhibit
increased structural priming when there is lexical or
morphological overlap ...The lexical enhancement of
priming from content words is the only significant
deviation between the model and priming data that we
know of (p. 256).However, as noted above, implicit
learning offers a plausible explanation for priming
across intervening material where residual activation
may not. Thus, even advocates of implicit learning
accounts of production priming recognize a need for a
lexically based mechanism to produce less durable
priming effects. ...Lexicalstructural interaction re-
flects short-term bindings that are not as durably
represented in the production system.
Open questions and future directions
These experiments provided evidence that syntactic
priming occurs when unrelated fillers appear between
the prime and the target. Because all of the conditions
here involved repeating a verb across the prime and the
target, these experiments do not address the question
whether such effects can occur when verbs are not
repeated across the prime and the target sentence. To
date, there is limited evidence for verb-independent
priming in comprehension (Traxler, 2008a; Thothathiri
& Snedeker, 2008a, 2008b). The evidence we have may
indicate that lexically independent priming is more
robust for some sentence types than others, that the
number of prime sentences matters, or that the type of
task is critical. An important next step in this line of
research is to see whether the kinds of effects obtained
here extend to other sentence types (such as dative
alternations or other forms of ambiguity) and whether
the effects can be observed when the prime and target
sentences do not share verbs (or other kinds of lexical
items). Further, it will be important to determine
whether the position of the related filler sentence
influences the magnitude of priming across lags.
A final major open question is the degree to which
syntactic priming in comprehension relies on the same
mechanism as syntactic priming in production. Lexi-
cally independent priming has been reported in com-
prehension (Thothathiri & Snedeker, 2008a, 2008b;
Traxler, 2008a), but appears to be more robust in
production priming studies. Production studies have
produced very long-lived priming effects, lasting as
long as several days (Bock & Griffin, 2000; Kaschak &
Borreggine, 2008). Evidence for persistent priming is
more limited in comprehension, but massed-priming
studies involving extensive exposure to specific sen-
tence types in comprehension have shown evidence for
long-term learning effects (Long & Prat, 2008; Wells et
al., 2009). The lexical boost, enhancements in priming
effects when specific words are repeated across primes
and targets, also appears to occur in both production
and comprehension. Thus, there appear to be substan-
Figure 4. Lemma and syntactic structure representations.
20 K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
tial similarities in the way priming effects manifest
across production and comprehension. Because syn-
tactic priming studies in comprehension are still a
relatively new field of inquiry, it remains to be seen
whether irreconcilable differences will be established
between the two modalities. If such differences are
established, they may reflect the nature of the tasks. In
production, priming is often measured as changes in
the likelihood of a binary choice. In comprehension,
the effects are typically graded and may differ both in
terms of overall processing time and the onset of
effects. Recently, studies of production priming have
begun to focus more on time-course (latency) issues,
but the overall focus is still on the outcome of some
structural choice. The most conservative position at the
moment appears to be that much more research is
needed, especially on comprehension, before convin-
cing conclusions about the overlap of mechanisms
between production and comprehension are justified.
Finally, these results, along with compatible results
from massed priming experiments (e.g., Wells et al.,
2009; Long & Prat, 2008; Luka & Barsalou, 2005) may
have implications more generally for models of parsing.
Constraint-based lexicalist accounts of parsing empha-
size the role of experience in shaping the way compre-
henders respond to input (e.g., MacDonald et al.,
1994). Accounts of this type are fully consistent with
changes in performance resulting from exposure to
particular sentence types. While the mere existence of
priming effects in comprehension does not, by itself,
pose a major challenge for two-stage, heuristic-driven
parsing, the existence of relatively long-lived priming in
comprehension suggests that aspects of syntactic
processing may flexibly adapt to changes in the input.
Conclusions
The ERP findings of this study offer, to our knowledge,
the first evidence that syntactic priming modulates the
P600 when an unrelated sentence intervenes between a
prime and target sentence. The eye-tracking data
provide the first behavioural evidence for syntactic
priming across lags of three during on-line comprehen-
sion, even when one of the intervening fillers repeats
the verb of the prime sentence, but in a different
sentence structure. This set of findings could help bring
the literature on comprehension priming a step closer
to the literature on production priming. Long-lived
priming effects have been attested in the production
literature for over a decade. However, there is almost no
evidence for comparable effects in comprehension. This
study begins to provide that evidence. Finally, while
further research is required, we believe that an account
of the current findings, as well as the broader syntactic
priming literature, will require two mechanisms: a
residual activation mechanism operating on combina-
torial nodes (or some similar representational sub-
strate) and an implicit learning mechanism that can
produce long-lasting changes in the pattern of connec-
tions in the underlying system of syntactic representa-
tions (see also Hartsuiker et al., 2008).
Acknowledgements
This project was supported by awards from the
National Science Foundation (#1024003) and from
the National Institutes of Health (#1R01HD048914).
The order of authorship is arbitrary.
Notes
1. Recent experiments provide evidence that exposure to a
given syntactic form may have separable inuence on the
likelihood of selection and the timing of planning
processes during speech production (Segaert et al.,
2011, 2012).
2. The full list of items for this experiment, as well as
Experiments 2 and 3, is available upon request from the
senior author: mjtraxler@ucdavis.edu
3. In order to address whether baseline differences could
have contributed to the pattern of results reported here,
we conducted additional analyses for the determiner,
using a pre-verb baseline instead of a pre-by baseline. The
use of a pre-verb baseline for all critical words required
extracting EEG epochs extending beyond 2300ms in
length, which resulted in a large amount of rejected trials
due to the presence of artifacts (epochs over 1000ms are
atypical, due to signal drift and participant movement).
For this reason, it was impossible to analyze effects at the
noun using a pre-verb baseline. However, the results for
the determiner using a pre-verb baseline were consistent
with those reported below, which used a pre-by baseline.
This suggests that the current pattern of results cannot be
attributed to the choice of baseline.
4. The by-participants multi-level models were congured
as follows:
Level 1: RT for person i, for item j BoiB1i
(sentence type)j eij
Level 2: Boi goo uo
B1i g10u1
For the by-items models, transpose person and item.
5. One might argue that the implicit learning is for the
connection between lexical entries and syntactic struc-
tures. If so, implicitly strengthening the connection
between the verb examinedand the reduced relative
clause structure would not necessarily facilitate proces-
sing a sentence that contained a different verb.
6. Segaert et al. (2011) have proposed a more detailed
account under which specic combinatorial nodes are
tied to specic lexical entries, but exposure to a given
syntactic form affects all related combinatorial nodes.
Language and Cognitive Processes 21
Downloaded by [University of California Davis] at 13:25 20 August 2013
7. We thank an anonymous reviewer for introducing this as
a possibility.
8. We thank an anonymous reviewer for raising this issue.
References
Arai, M., Van Gompel, R. P. G., & Scheepers, C. (2007). Priming
ditransitive structures in comprehension. Cognitive Psychology,
54(3), 218250. doi:10.1016/j.cogpsych.2006.07.001
Balota, D. A., Yap, M. J., Hutchison, K. A., Cortese, M. J., Kessler, B.,
Loftis, B., ... Treiman, R. (2007). The english lexicon project.
Behavior Research Methods,39(3), 445459.
Blozis, S. A., & Traxler, M. J. (2007). Analyzing individual differences
in sentence processing performance using multilevel models.
Behavior Research Methods,39(1), 3138. doi:10.3758/BF03192841
Bock, J. K. (1986). Syntactic persistence in language production.
Cognitive Psychology,18(3), 355387. doi:10.1016/0010-
0285(86)90004-6
Bock, K., Dell, G., Chang, F., & Onishi, K. (2007). Persistent structural
priming from language comprehension to language production.
Cognition,104(3), 437458. doi:10.1016/j.cognition.2006.07.003
Bock, K., & Grifn, Z. M. (2000). The persistence of structural
priming: Transient activation or implicit learning? Journal of
Experimental Psychology: General,129(2), 177192. doi:10.1037/
0096-3445.129.2.177
Boland, J. E., & Blodgett, A. (2006). Argument status and PP-
attachment. Journal of Psycholinguistic Research,35(5), 385403.
doi:10.1007/s10936-006-9021-z
Branigan, H. P., Pickering, M. J., & McLean, J. F. (2005). Priming
prepositional phrase attachment during comprehension. Journal of
Experimental Psychology: Learning, Memory, and Cognition,31,
468481. doi:10.1037/0278-7393.31.3.468
Carminati, M., & van Gompel, R. P. G. (2009, March). The Time
Course of Structural Priming in Comprehension. Paper presented to
the 22nd Annual CUNY Conference on Human Sentence Proces-
sing. Davis, CA.
Chang, F., Dell, G. S., & Bock, K. (2006). Becoming syntactic.
Psychological Review,113(2), 234272. doi:10.1037/0033-
295X.113.2.234
Chang, F., Dell, G. S., Bock, K., & Grifn, Z. M. (2000). Structural
priming as implicit learning: A comparison of models of sentence
production. Journal of Psycholinguistic Research,29(2), 217229.
doi:10.1023/A:1005101313330
Clifton, C., Traxler, M. J., Williams, R., Mohammed, M., Morris, R.
K., & Rayner, K. (2003). The use of thematic role information in
parsing: Syntactic processing autonomy revisited. Journal of
Memory and Language,49(3), 317334. doi:10.1016/S0749-
596X(03)00070-6
Hartsuiker, R. J., Bernolet, S., Schoonbaert, S., Speybroeck, S., &
Vanderelst, D. (2008). Syntactic priming persists while the lexical
boost decays: Evidence from written and spoken dialogue. Journal
of Memory and Language,58(2), 214238. doi:10.1016/
j.jml.2007.07.003
Hartsuiker, R. J., & Kolk, H. H. J. (1998). Syntactic persistence in
Dutch. Language and Speech,41, 143184.
Kaschak, M. P., & Borreggine, K. L. (2008). Is long-term structural
priming affected by patterns of experience with individual verbs?
Journal of Memory and Language,58(3), 862878. doi:10.1016/
j.jml.2006.12.002
Kim, A., & Osterhout, L. (2005). The independence of combinatory
semantic processing: Evidence from event-related potentials. Jour-
nal of Memory and Language,52(2), 205225. doi:10.1016/
j.jml.2004.10.002
Konieczny, L., Hemforth, B., Scheepers, C., & Strube, G. (1997). The
role of lexical heads in parsing: Evidence from German. Language
and Cognitive Processes,12(2), 307348. doi:10.1080/
016909697386871
Kuperberg, G. R., Caplan, D., Sitnikova, T., Eddy, M., & Holcomb, P.
J. (2006). Neural correlates of processing syntactic, semantic, and
thematic relationships in sentences. Language and Cognitive Pro-
cesses,21(5), 489530. doi:10.1080/01690960500094279
Kutas, M., van Petten, C. K., & Kluender, R. (2006). Psycholinguistics
electried II. In M. J. Traxler & M. A. Gernsbacher (Eds.), The
handbook of psycholinguistics (2nd ed) (pp. 659724). Elsevier
Academic Press.
Ledoux, K., Traxler, M. J., & Swaab, T. Y. (2007). Syntactic priming in
comprehension: Evidence from event-related potentials. Psycholo-
gical Science,18(2), 135143. doi:10.1111/j.1467-9280.2007.01863.x
Long, D. L., & Prat, C. S. (2008). Individual differences in syntactic
ambiguity resolution: Readers vary in their use of plausibility
information. Memory & Cognition,36(2), 375391. doi:10.3758/
MC.36.2.375
Luka, B. J., & Barsalou, L. W. (2005). Structural facilitation: Mere
exposure effects for grammatical acceptability as evidence for
syntactic priming in comprehension. Journal of Memory and
Language,52(3), 436459. doi:10.1016/j.jml.2005.01.013
Nakano, H., Saron, C., & Swaab, T. Y. (2010). Speech and span:
Working memory capacity impacts the use of animacy but not of
world knowledge during spoken sentence comprehension. Journal
of Cognitive Neuroscience,12, 28862898.
Pickering, M. J., & Branigan, H. P. (1998). The representation of verbs:
Evidence from syntactic priming in language production. Journal of
Memory and Language,39(4), 633651. doi:10.1006/jmla.1998.2592
Pickering, M. J., Branigan, H. P., & McLean, J. F. (2002). Constituent
structure is formulated in one stage. Journal of Memory and
Language,46(3), 586605. doi:10.1006/jmla.2001.2824
Pickering, M. J., McLean, J. F., & Branigan, H. P. (2012). Persistent
structural priming and frequency effects during comprehension.
Journal of Experimental Psychology: Learning, Memory, and
Cognition.
Pickering, M. J., Tooley, K. M., & Traxler, M. J. (2011). Thematic
processing during sentence comprehension. Manuscript submitted for
publication.
Pickering, M. J., & Traxler, M. J. (2004, March). Grammatical repetition
and garden path effects. Paper presented to the CUNY Sentence
Processing conference. College Park, MD.
Pollard, C., & Sag, I. A. (1994). Head-driven phrase structure grammar.
Stanford, CA: Center for the Study of Language.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models:
Applications and data analysis methods (2nd ed). Newbury Park,
CA: Sage.
Segaert, K., Menenti, L., Weber, K., & Hagoort, P. (2011). A paradox
of syntactic priming: Why response tendencies show priming for
passives, and response latencies show priming for actives. PLoS
ONE,6(10), e24209. doi:10.1371/journal.pone.0024209
Segaert, K., Menenti, L., Weber, K., Petersson, K. M., & Hagoort, P.
(2012). Shared syntax in language production and language
comprehension an fMRI study. Cerebral Cortex,22(7), 1662
1670. doi:10.1093/cercor/bhr249
Sturt, P., Keller., F., & Dubey., A. (2010). Syntactic priming in
comprehension: Parallelism effects with and without coordination.
Journal of Memory and Language,62(4), 333351. doi:10.1016/
j.jml.2010.01.001
Thothathiri, M., & Snedeker, J. (2008a). Give and take: Syntactic
priming during spoken language comprehension. Cognition,108(1),
5168. doi:10.1016/j.cognition.2007.12.012
Thothathiri, M., & Snedeker, J. (2008b). Syntactic priming during
language comprehension in three- and four-year-old children.
22 K.M. Tooley et al.
Downloaded by [University of California Davis] at 13:25 20 August 2013
Journal of Memory and Language,58(2), 188213. doi:10.1016/
j.jml.2007.06.012
Tomasello, M. (2000). The item-based nature of childrens early
syntactic development. Trends in Cognitive Sciences,4(4), 156
163. doi:10.1016/S1364-6613(00)01462-5
Tooley, K. M. (2009). Syntactic priming in sentence comprehension:
Mechanisms and time course (Unpublished doctoral dissertation).
University of California, Davis.
Tooley, K. M., & Traxler, M. J. (2010). Syntactic priming effects in
comprehension: A critical review. Language and Linguistics Com-
pass,4(10), 925937. doi:10.1111/j.1749-818X.2010.00249.x
Tooley, K. M., Traxler, M. J., & Swaab, T. Y. (2009). Electrophysio-
logical and behavioural evidence of syntactic priming in sentence
comprehension. Journal of Experimental Psychology: Learning,
Memory, Cognition,35,1945. doi:10.1037/a0013984
Traxler, M. J. (2008a). Lexically independent priming in online sentence
comprehension. Psychonomic Bulletin and Review,15(1), 149155.
doi:10.3758/PBR.15.1.149
Traxler, M. J. (2008b). Structural priming among prepositional phrases:
Evidence from eye-movements. Memory & Cognition,36(3), 659
674. doi:10.3758/MC.36.3.659
Traxler, M. J., Bybee, M. D., & Pickering, M. J. (1997). Inuence of
connectives on language comprehension: Eye-tracking evidence for
incremental interpretation. Quarterly Journal of Experimental
Psychology A,50, 481497.
Traxler, M. J., & Pickering, M. J. (2005, March). Syntactic priming in
comprehension. Paper presented to the CUNY Sentence Processing
Conference. Tucson, AZ.
Traxler, M. J., & Tooley, K. M. (2008). Priming in sentence comprehen-
sion: Strategic or syntactic? Language and Cognitive Processes,
23(5), 609645. doi:10.1080/01690960701639898
Traxler, M. J., Williams, R. S., Blozis, S. A., & Morris, R. K. (2005).
Working memory, animacy, and verb class in the processing of
relative clauses. Journal of Memory & Language,53(2), 204224.
doi:10.1016/j.jml.2005.02.010
Ullman, M. T. (2001). The neural basis of lexicon and grammar in rst
and second language: The declarative/procedural model. Bilingual-
ism, Language, and Cognition,4, 105122. doi:10.1017/
S1366728901000220
Vosse, T., & Kempen, G. (2000). Syntactic structure assembly in human
parsing: A computational model based on competitive inhibition
and lexicalist grammar. Cognition,75(2), 105143. doi:10.1016/
S0010-0277(00)00063-9
Wells, J. B., Christiansen, M. H., Race, D. S., Acheson, D. J., &
MacDonald, M. C. (2009). Experience and sentence processing:
Statistical learning and relative clause comprehension. Cognitive
Psychology,58(2), 250271. doi:10.1016/j.cogpsych.2008.08.002
Language and Cognitive Processes 23
Downloaded by [University of California Davis] at 13:25 20 August 2013
... Here we only focus on the ambiguity effect. We also note that a separate line of work has begun to investigate trial-to-trial syntactic priming effects on RC/MV garden paths (Tooley et al., 2015;Tooley & Traxler, 2018;Traxler et al., 2014). We describe these studies in the general discussion. ...
... This work has approached questions similar to the one addressed here, though from a different perspective. As we discuss, the lack of adaptation effects on first-pass measures has parallels in these studies on structural priming (Tooley et al., 2015;Tooley & Traxler, 2018;Traxler, Tooley, & Pickering, 2014). We then close by discussing different possible explanations for the lack of adaptation effects on first-pass measures in these two lines of work. ...
... Three eye-tracking reading studies have investigated trial-to-trial structural priming in MV/RC ambiguity (Tooley et al., 2015;Tooley & Traxler, 2018;Traxler et al., 2014). One study compared reading measures on the disambiguation region of the RC, depending on whether the preceding prime trial had been and RC or MV (Traxler et al., 2014). ...
Preprint
Implicit expectations play a crucial role during language processing. These expectations are often assumed to be static once acquired, or to change only at relatively slow time scales.Some theoretical proposals, however, hold that comprehenders continuously adapt their expectations based on recent input, and that this adaptation underlies phenomena such as syntactic priming. Existing evidence from research on sentence processing has confounded adaptation to the experimental task with evidence for expectation adaptation. Here we eliminate this confound by, for the first time, investigating expectation adaptation during natural reading. In two eye-tracking experiments, subjects read sentences that contain a higher than expected proportion of a previously unexpected structure (reduced relative clauses). We inspect whether and how syntactic adaptation manifests in the eye-movement measures: if subjects adapt their syntactic expectations, their reading times for unexpected syntactic structures should decrease over the course of the experiment. This prediction is confirmed in both experiments, providing evidence for expectation adaptation during naturalreading. Specifically, we find evidence for syntactic adaptation primarily for regression-related measures but not first pass reading measures. We discuss the different possible accounts of this pattern of syntactic adaptation in the eye-movement record.
... One concern, however, is that the majority of these findings have been tested in priming for production and have been less extensively discussed in terms of comprehension priming, where the processing of repeated structures results in facilitatory processing (for a review on comprehension priming, see Tooley & Traxler, 2010). For example, Tooley, Swaab, Boudewyn, Zirnstein, and Traxler (2014) found facilitation effects for repeated structures in comprehension using both event-related potentials and eye-tracking measures when the prime and target shared the same verb. These lexically dependent priming effects, or "lexical-boost" effects, have not been found to last long in production priming studies (Hartsuiker, Bernolet, Schoonbaert, Speybroeck, & Vanderelst, 2008) and are theorized to rely on the residual activation of combinatorial nodes in a connectionist network. ...
... However, the extreme discrepancy in frequencies between RR and MV resolutions in real language comprehension (Hare et al., 2007) should still contrast strongly with the manipulated frequencies in this experiment. It could also be that removing the rote repetition of the RR GP structure by adding filler items may take away their impact and lead to a lack of facilitation for these sentences; however, structural priming in comprehension has been shown to take place with multiple intervening filler sentences (Tooley et al., 2014). Furthermore, with sheer repetition of the structure, it becomes difficult to conclude whether effects are due to an adaptation of binary resolution expectations versus an adaptation to the specific structure itselfhard sentences become easier over time, all reading times typically get faster throughout an experiment, and even ungrammatical sentences become more acceptable (i.e., syntactic satiation effects). ...
Article
Full-text available
Convergent probabilistic cues do not trigger syntactic adaptation 2 ABSTRACT Previous work has ostensibly shown that readers rapidly adapt to less predictable ambiguity resolutions after repeated exposure to unbalanced statistical input (e.g., a high number of reduced relative-clause garden-path sentences), and that these readers grow to disfavor the a priori more frequent main verb resolution after exposure (Fine, Jaeger, Farmer, & Qian, 2013). However, recent work has failed to replicate effects indicating a penalty for the preferred, more frequent continuation, despite finding a speedup in syntactic repair times after initial exposure to the dispreferred, infrequent structure (Harrington-Stack, James, & Watson, 2018). The current study reports three self-paced reading experiments that test whether co-occurring cues (explicit comprehension questions, preceding semantic cues, and font color) help facilitate adaptation to reduced relative/main verb garden-path sentences. Results suggest that readers do not overcome pre-existing expectation biases by rapidly adapting to statistically novel linguistic contexts even with convergent probabilistic cues. An emphasis is placed on the difference between syntactic satiation effects and expectation adaptation, the latter of which we argue can only be determined through a penalty for an a priori preferred resolution after repeated exposure to its a priori less preferred counterpart.
... The P600 is a positive component arising between 500 and 900 ms over posterior sites that typically indexes syntactic processing (e.g., Osterhout and Holcomb, 1992;Hahne and Friederici, 1999). Previous studies of comprehension structural priming using ERPs have typically found a reduction in the P600 effect to indicate facilitated processing of primed syntactic structures (Ledoux et al., 2007;Tooley et al., 2009Tooley et al., , 2013Boudewyn et al., 2014; for a review, see Tooley and Traxler, 2010). Earlier positivities (Ledoux et al., 2007) and N400 effects to repeated verbs (Ledoux et al., 2007;Tooley et al., 2009) have also been found. ...
... The other cross-modal priming direction, priming from production to comprehension, has been addressed in only two published studies (to our knowledge), one of which used an offline sentence-picture matching task (Branigan et al., 2005) and one from the perspective of neural substrates rather than online processing (Segaert et al., 2012). Previous studies of comprehension-to-comprehension priming using ERPs have typically found reductions in the P600 (Ledoux et al., 2007;Tooley et al., 2009Tooley et al., , 2013Boudewyn et al., 2014; for a review, see Tooley and Traxler, 2010), though all studies examined comprehension-to-comprehension priming using visual presentation. The current study, the first to study cross-modality structural priming into comprehension with ERPs and the first ERP comprehension priming study to use auditory presentation, found a priming effect in terms of a reduction of the N400, and no P600 effect. ...
Article
Full-text available
Natural language involves both speaking and listening. Recent models claim that production and comprehension share aspects of processing and are linked within individuals (Pickering and Garrod, 2004, 2013; MacDonald, 2013; Dell and Chang, 2014). Evidence for this claim has come from studies of cross-modality structural priming, mainly examining processing in the direction of comprehension to production. The current study replicated these comprehension to production findings and developed a novel cross-modal structural priming paradigm from production to comprehension using a temporally sensitive online measure of comprehension, Event-Related Potentials. For Comprehension-to-Production priming, participants first listened to active or passive sentences and then described target pictures using either structure. In Production-to-Comprehension priming, participants first described a picture using either structure and then listened to target passive sentences while EEG was recorded. Comprehension-to-Production priming showed the expected passive sentence priming for syntactic choice, but not response time (RT) or average syllable duration. In Production-to-Comprehension priming, primed, versus unprimed, passive sentences elicited a reduced N400. These effects support the notion that production and comprehension share aspects of processing and are linked within the individual. Moreover, this paradigm can be used for the exploration priming at different linguistic levels as well as the influence of extra-linguistic factors on natural language use.
... Block order was counterbalanced across participants exposed to each respective list using a Latin square design, ensuring that each block occurred equally often in each of the eight possible list positions. This was done so as to avoid confounding of the conditions of interest with presentation order, since reading times are known to be affected by previous exposure to similar structures (e.g., Fine et al., 2010Fine et al., , 2013Tooley et al., 2014;Tooley and Traxler, 2018;Yan and Jaeger, 2020). ...
Article
Full-text available
A central component of sentence understanding is verb-argument interpretation, determining how the referents in the sentence are related to the events or states expressed by the verb. Previous work has found that comprehenders change their argument interpretations incrementally as the sentence unfolds, based on morphosyntactic (e.g., case, agreement), lexico-semantic (e.g., animacy, verb-argument fit), and discourse cues (e.g., givenness). However, it is still unknown whether these cues have a privileged role in language processing, or whether their effects on argument interpretation originate in implicit expectations based on the joint distribution of these cues with argument assignments experienced in previous language input. We compare the former, linguistic account against the latter, expectation-based account, using data from production and comprehension of transitive clauses in Swedish. Based on a large corpus of Swedish, we develop a rational (Bayesian) model of incremental argument interpretation. This model predicts the processing difficulty experienced at different points in the sentence as a function of the Bayesian surprise associated with changes in expectations over possible argument interpretations. We then test the model against reading times from a self-paced reading experiment on Swedish. We find Bayesian surprise to be a significant predictor of reading times, complementing effects of word surprisal. Bayesian surprise also captures the qualitative effects of morpho-syntactic and lexico-semantic cues. Additional model comparisons find that it-with a single degree of freedom-captures much, if not all, of the effects associated with these cues. This suggests that the effects of form-and meaning-based cues to argument interpretation are mediated through expectation-based processing.
... Alternatively, the practice trials might have helped children to cope with the linguistic demands imposed by the test question. Research with both children and adults suggests that hearing or producing a particular sentence structure facilitates subsequent processing of that sentence structure (e.g., Bock, 1986;Branigan & McLean, 2016;Rissman, Legendre, & Landau, 2013;Thothathiri & Snedeker, 2008;Tooley, Swaab, Boudewyn, Zirnstein, & Traxler, 2014). For instance, hearing the sentence ''Give the ball to the lion" facilitates 3-year-olds' subsequent interpretation of a sentence with the same structure (e.g., ''Give the birdhouse to the sheep") but not of a sentence with a different structure (e.g., ''Give the sheep the birdhouse") (Thothathiri & Snedeker, 2008). ...
Article
Researchers have argued that traditional elicited-response false-belief tasks involve considerable processing demands and hence underestimate children’s false-belief understanding. Consistent with this claim, Setoh et al. (2016) recently found that when processing demands were sufficiently reduced, children could succeed in an elicited-response task as early as 2.5 years of age. Here we examined whether 2.5-year-olds could also succeed in a low-demand elicited-response task involving false beliefs about identity, which have been argued to provide a critical test of whether children truly represent beliefs, while also clarifying how the practice trials in Setoh et al.’s task facilitated children’s elicited-response performance. 2.5-year-olds were tested in a version of Setoh et al.’s elicited-response task in which they heard a location or identity false-belief story. We varied whether the practice trials had the same type of wh-question as the test trial. Children who heard the same type of wh-question on all trials succeeded regardless of which story they heard (location or identity) and performance did not differ across belief type. This replicates Setoh et al.’s positive results and demonstrates that when processing demands are sufficiently reduced, children can succeed in elicited-response tasks involving false beliefs about object location or identity. This suggests that children are capable of attributing genuine false beliefs prior to 4 years of age. However, children performed at chance if the practice trials involved a different type of wh-question than the test trials, suggesting that at this age practice with the wh-question used in the test trial is essential to children’s success.
... It is worth noting that there have been some studies that have observed lexically boosted priming effects that persist across a few intervening sentences (Pickering, McLean, & Branigan, 2013;Tooley, Swaab, Boudewyn, Zirnstein, & Traxler, 2014). Furthermore, there are learning models that can accommodate different longevities of the lexical boost and abstract priming effects (e.g., Jaeger & Snider, 2013;Malhotra, Pickering, Branigan, & Bednar, 2008). ...
Article
While many recent studies focused on abstract syntactic priming effects have implicated an error-based learning mechanism, there is little consensus on the most likely mechanism underlying the lexical boost. The current study aimed at refining understanding of the mechanism that leads to this priming effect. In two eye-tracking during reading experiments, the nature of the lexical boost was investigated by comparing predictions from competing accounts in terms of decay and the requirement of structural overlap between primes and targets. Experiment 1 revealed facilitation of target structure processing for shorter relative to longer primes, when there were fewer intervening words between prime and target verbs. In Experiment 2, significant lexically boosted priming effects were observed, but only when the target structure also appeared in the prime, and not when the prime had a different structure but a high degree of lexical overlap with the target. Overall, these results are most consistent with a short-lived mechanistic account rather than an error-based learning account of the lexical boost. Furthermore, these results align with dual-mechanism accounts of syntactic priming whereby different mechanisms are claimed to produce abstract syntactic priming effects and the lexical boost.
... For instance, Hartsuiker and Kolk (1998) manipulated the number of filler sentences between primes and targets and found that structural priming effects in adults were long-lasting, and work by Bock and Griffin (2000) indicated that adults were primed even when there was intervening material (up to ten filler sentences) between primes and targets. Long-term priming effects have also been demonstrated in comprehension; Tooley, Swaab, Boudewyn, Zirnstein and Traxler's (2014)'s study revealed that adults' processing of target sentences was facilitated despite three fillers appearing between prime and target sentences. Similar effects have been demonstrated in studies with children (Huttenlocher, Vasilyeva, & Shimpi, 2004;Kidd, 2012;Savage, Lieven, Theakston, & Tomasello, 2003, 2006). ...
Article
Full-text available
A long‐standing question in child language research concerns how children achieve mature syntactic knowledge in the face of a complex linguistic environment. A widely accepted view is that this process involves extracting distributional regularities from the environment in a manner that is incidental and happens, for the most part, without the learner's awareness. In this way, the debate speaks to two associated but separate literatures in language acquisition: statistical learning and implicit learning. Both fields have explored this issue in some depth but, at present, neither the results from the infant studies used by the statistical learning literature nor the artificial grammar learning tasks studies from the implicit learning literature can be used to fully explain how children's syntax becomes adult‐like. In this work, we consider an alternative explanation—that children use error‐based learning to become mature syntax users. We discuss this proposal in the light of the behavioral findings from structural priming studies and the computational findings from Chang, Dell, and Bock's (2006) dual‐path model, which incorporates properties from both statistical and implicit learning, and offers an explanation for syntax learning and structural priming using a common error‐based learning mechanism. We then turn our attention to future directions for the field, here suggesting how structural priming might inform the statistical learning and implicit learning literature on the nature of the learning mechanism.
Chapter
Language input is often noisy and ambiguous. Yet, humans are able to successfully communicate their thoughts to one another by leveraging contextual information to make inferences about the intended meaning of the input. Though there is broad agreement on the role and importance of context, the term is used so flexibly that it can be difficult to know what it refers to and how to make progress on studying it. Here, we propose that a key dimension which spans domains of context is the timescale over which the contextual information is sampled for the purpose of inferring meaning. We review the literature on context effects in language comprehension, organizing our discussion around a few relevant temporal windows, milliseconds, seconds, and minutes, and then propose that the individual's lifetime of language experience and historical time both constitute relevant contexts for language processing as well. We then discuss how construing of context in this way reveals gaps in the existing research landscape and propose new avenues of inquiry to address them.
Article
Native (L1) and nonnative (L2) speakers sometimes misinterpret temporarily ambiguous sentences like “When Mary dressed the baby laughed happily.” Recent studies suggest that the initially assigned misinterpretation (“Mary dressed the baby”) may persist even after disambiguation, and that L2 speakers may have particular difficulty discarding initial misinterpretations. The present study investigated whether L2 speakers are more persistent with misinterpretation compared with L1 speakers during sentence processing, using the structural priming and eye tracking while reading tasks. In the experiment, participants read prime followed by target sentences. Reading times revealed that unambiguous but not ambiguous prime sentences facilitated processing of the globally correct interpretation of ambiguous target sentences. However, this priming effect was only observed when the prime and target sentence shared the same verb. Comprehension accuracy rates were not significantly influenced by priming effects but did provide evidence of lingering misinterpretation. We did not find significant L1/L2 differences in either priming effects or persistence of misinterpretation. Together, these results suggest that initially assigned misinterpretations linger in both L1 and L2 readers during sentence processing and that L1 and L2 comprehension priming is strongly lexically mediated.
Article
The nature of the facilitation occurring when sentences share a verb and syntactic structure (i.e., lexically-mediated syntactic priming) has not been adequately addressed in comprehension. In four eye-tracking experiments, we investigated the degree to which lexical, syntactic, thematic, and verb form repetition contribute to facilitated target sentence processing. Lexically-mediated syntactic priming was observed when primes and targets shared a verb and abstract syntactic structure, regardless of the ambiguity of the prime. In addition, repeated thematic role assignment resulted in syntactic priming (to a lesser degree), and verb form repetition facilitated lexical rather than structural processing. We conclude that priming in comprehension involves lexically associated abstract syntactic representations, and facilitation of verb and thematic role processes. The results also indicate that syntactic computation errors during prime processing are not necessary for lexically-mediated priming to occur during target processing. This result is inconsistent with an error-driven learning account of lexically-mediated syntactic priming effects.
Article
Full-text available
An eye-tracking experiment investigated whether incremental interpretation applies to interclausal relationships. According to Millis and Just's (1994) delayed-integration hypothesis, interclausal relationships are not computed until the end of the second clause, because the processor needs to have two full propositions before integration can occur. We investigated the processing of causal and diagnostic sentences (Sweetser, 1990; Tversky & Kahneman, 1982) that contained the connective because . Previous research (Traxler, Sanford, Aked, & Moxey, 1997) has demonstrated that readers have greater difficulty processing diagnostic sentences than causal sentences. Our results indicated that difficulty processing diagnostic sentences occurred well before the end of the second clause. Thus comprehenders appear to compute interclausal relationships incrementally.
Article
Full-text available
Event-related potentials were measured as subjects read sentences presented word by word. A small N400 and a robust P600 effect were elicited by verbs that assigned the thematic role of Agent to their preceding noun-phrase argument when this argument was inanimate in nature. The amplitude of the P600, but not the N400, was modulated by the transitivity of the critical verbs and by plausibility ratings of passivised versions of these sentences (reflecting the fit between the critical verb and the inanimate noun-phrase as the verb's Theme). The P600 was similar in scalp distribution although smaller in amplitude, than that elicited by verbs with morphosyntactic violations. Pragmatically unlikely verbs that did not violate thematic constraints elicited a larger N400 but no P600 effect. These findings support the theory that the cost of syntactic processing on a verb is influenced by the precise thematic relationships between that verb and its preceding arguments.
Article
Full-text available
Two eye-tracking experiments and two self-paced reading experiments investigated processing of sentences containing reduced relative clauses. Processing of a reduced relative is facilitated when it is preceded by a sentence that has the same syntactic structure, as long as the preceding sentence contains the same critical verb as the target (LeDoux, Traxler, & Swaab, 2007; Pickering & Traxler, 2004; Traxler & Pickering, 2005). This study tests whether facilitative effects result from readers using the prime sentence as a strategic cue to predict an upcoming reduced relative target. This hypothesis was tested in three ways. In the first experiment, properties of the filler items were manipulated to reduce or eliminate the strategic value of repeated verbs as a structural cue. In the second experiment, noun overlap, rather than verb overlap, created a valid strategic cue to the upcoming sentence structure. In the third experiment, readers were warned that a reduced relative sentence would be forthcoming. Significant priming occurred in the absence of a valid cue in the first experiment; and no structural priming occurred in the presence of a valid cue in the second and third experiments. The fourth experiment demonstrated that the self-paced reading method is sensitive to priming effects. These results are consistent with previous studies showing the same dissociation between strategic cues and priming effects and therefore further indicate that priming effects are the result of a set of processes intrinsic to the parser.
Article
Full-text available
Theories of sentence processing can be classified by the role that lexical heads play in first-pass parsing. In some theories, syntactic structure cannot even be built before it is licensed by a head (Abney, 1989; Prichett, 1992), whereas in other models, at the opposite extreme, only major category information in lexical items is used during the rule-based initial structure building process (Ferreira & Henderson, 1990; M itchell, 1987). In this paper, we will introduce a fully increm ental model that accounts for attachment preferences by the linear order of lexical heads in the surface structure and their thematic properties. The principle of "parameterised head attachment" and a serial variant, the SOUL mechanism, will be discussed on the basis of three on-line experim ents on NP- and PP-attachment in German verb-second and verb-final sentences.
Article
Activation processes appear to have an important impact on the mechanisms of language use, including those responsible for syntactic structure in speech. Some implications of this claim for theories of language performance were examined with a syntactic priming procedure. On each priming trial, subjects produced a priming sentence in one of several syntactic forms. They then viewed an unrelated event in a picture and described it in one sentence. The probability of a particular syntactic form being used in the description increased when that form had occurred in the prime, under presentation conditions that minimized subjects' attention to their speech, to the syntactic features of the priming sentences, and to connections between the priming sentences and the subsequent pictures. This syntactic repetition effect suggests that sentence formulation processes are somewhat inertial and subject to such probabilistic factors as the frequency or recency of use of particular structural forms. Two further experiments showed that this effect was not appreciably modified by variations in certain conceptual characteristics of sentences, and all three experiments found evidence that the effects of priming were specific to features of sentence form, independent of sentence content. The empirical isolability of structural features from conceptual characteristics of successive utterances is consistent with the assumption that some syntactic processes are organized into a functionally independent subsystem.
Article
Theories of sentence processing can be classified by the role that lexical heads play in first-pass parsing. In some theories, syntactic structure cannot even be built before it is licensed by a head (Abney, 1989; Prichett, 1992), whereas in other models, at the opposite extreme, only major category information in lexical items is used during the rule-based initial structure building process (Ferreira & Henderson, 1990; Mitchell, 1987). In this paper, we will introduce a fully incremental model that accounts for attachment preferences by the linear order of lexical heads in the surface structure and their thematic properties. The principle of "parameterised head attachment" and a serial variant, the SOUL mechanism, will be discussed on the basis of three on-line experiments on NP- and PP-attachment in German verb-second and verb-final sentences.
Chapter
In 1994, there were only two dominant noninvasive techniques to offer insight about the functional organization of language from its brain bases: the behavior of brain-damaged patients (neuropsychology), and event-related brain potential (ERPs). Positron emission tomographic and magnetoencephalographic (MEG) measures begin to contribute in understanding neuropsychology. Over the ensuing decade plus, these have been joined by functional magnetic resonance imaging, transcranial magnetic stimulation, event related spectral changes in the electroencephalogram (EEG), and noninvasive optical imaging. These methods are closely related in their neural and physical bases: ERPs, event-related frequency changes in the EEG and MEG. The amplitude of the EEG is considerably smaller than invasively recorded field potentials because the skull is a strong electrical insulator. Like field potentials, the amplitude and polarity of the EEG depends on the number and amplitude of the contributing synaptic potentials, on whether current is flowing into or out of cells (i.e., movement of positive or negative ions, excitatory or inhibitory synaptic potentials), and on the geometric relationship between the synapses and electrode.