Journal of Neurolinguistics 33 (2015) 96e103
Contents lists available at ScienceDirect
Journal of Neurolinguistics
journal homepage: www.elsevier.com/locate/
jneuroling
An ERP study on how subsequent sentence
context can influence previous world knowledge
constraints
Guiping Xu a, b, Weifang Zhong a, Hua Jin c, Lei Mo a, *
a
b
c
Centre for the Study of Applied Psychology, South China Normal University, Guangzhou 510631, China
Department of Biomedical Engineering, School of Medicine, Shenzhen University, Shenzhen 518060, China
Academy of Psychology and Behaviour, Tianjin Normal University, Tianjin 300074, China
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 17 March 2014
Received in revised form 13 September 2014
Accepted 16 September 2014
Available online 7 October 2014
In an ERP study, following stimuli with a world knowledge violation we added a sentence-level context which made a world
knowledge violation more acceptable in order to investigate how
participants processed the subsequent sentence context. Sentences
consisting of five Chinese words were of three experimental types:
correct sentences with correct world knowledge and coherent
sentence meaning, incoherent sentences with a world knowledge
violation and incoherent sentence meaning, and ‘re-coherent’
sentences with a world knowledge violation but coherent sentence
meaning. The results showed both incoherent and re-coherent
conditions modulated N400, elicited by the critical word which
contained the world knowledge violation, and a negative deflection elicited by the final word, in a similar way, reflecting the
immediate world knowledge integration of critical word and low
predictability of final word respectively. An additional negative
deflection was elicited only by the final word in the incoherent
condition, probably reflecting global sentence context integration.
These results provide support for immediate world knowledge
integration, and indicate that subsequent new sentence context
can immediately influence previous world knowledge constraints.
© 2014 The Authors. Published by Elsevier Ltd. This is an open
access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/3.0/).
Keywords:
World knowledge
Context
Sentence
ERPs
N400
* Corresponding author. Tel.: þ86 20 85213767.
E-mail address: [email protected] (L. Mo).
http://dx.doi.org/10.1016/j.jneuroling.2014.09.003
0911-6044/© 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/3.0/).
G. Xu et al. / Journal of Neurolinguistics 33 (2015) 96e103
97
1. Introduction
Yao-Ming is such an internationally known basketball star among Chinese people that no one would
consider this tall man to be short. So people immediately thought the sentence “That short Yao-Ming is
a basketball-star.” is nonsense (Jin et al., 2009). But to our knowledge, it is less clear how people will
understand the sentence “That short Yao-Ming is a cook”. It is unclear whether they can immediately
accept the possibility that there is a short man who is a cook and also is named Yao-Ming. The aim of
the current ERP experiment was to examine whether a subsequent global sentence context could
influence a previous world knowledge violation constraint.
In sentence comprehension, a reader not only builds up the basic meaning of a sentence according
to lexical semantics, but also activates pragmatic information such as world knowledge to judge the
truth of a sentence (Clark, 1996; Hagoort & Van Berkum, 2007; Jackendoff, 2002; Kempson, 2003;
Perry, 1997). The sentence “That short Yao-Ming is a basketball-star” has a coherent semantic interpretation but violates our world knowledge, which differs from the sentence “That sour Yao-Ming is a
basketball-star” with an incoherent semantic interpretation. With regards to the time course of integration of world knowledge and lexical semantics, recent ERP studies have consistently indicated rapid
parallel integration which supports the constraints-satisfaction model (Chwilla & Kolk, 2005; Hagoort,
Hald, Bastiaansen, & Petersson, 2004; Hald, Steenbeek-Planting, & Hagoort, 2007; Jackendoff, 2007; Jin
et al., 2009; Macizo & Herrera, 2010). This model proposed that language comprehension allows for the
rapid parallel use of multiple constraints. The various kinds of constraint information such as lexical
semantics, syntax and world knowledge are all simultaneously processed to achieve the best understanding of an utterance (Jackendoff, 2002; MacDonald, Pearlmutter, & Seidenberg, 1994; Salmon &
Pratt, 2002; Tanenhaus, Spivey-Knowlton, Eberhard, & Sedivy, 1995; Van Berkum, Hagoort, & Brown,
1999; Van Berkum, Zwitserlood, Hagoort, & Brown, 2003).
Specifically, Hagoort et al. (2004) first explored this issue using the N400 effect, which is widely
considered to reflect the semantic integration of critical words with the context (Kutas & Federmeier,
2011; Kutas & Hillyard, 1980, 1984; Lau, Phillips, & Poeppel, 2008; Van Berkum, 2008). Participants
were asked to read three versions of sentences such as the following: “The Dutch trains are yellow/
white/sour and very crowded.” (critical words are in italics). Dutch people all know that Dutch trains are
yellow. So the sentences were of three types: correct (yellow), having a world knowledge violation
(white), and having a semantic violation (sour). The result showed a clear N400 effect, with equivalent
onset and peak latencies for both world knowledge and semantic violations, indicating the rapid
parallel integration of both world knowledge and lexical semantics.
Additionally, Jin et al. (2009) embedded the world knowledge violation in a sentence-initial phrase
to further explore this issue. Sentences used were of the following types: “That tall/short/sour YaoMing is a basketball-star.” (critical word is in italics, final word is in bold). The sentences were of three
types: correct (tall Yao-Ming), world knowledge violation (short Yao-Ming), and semantic violation
(sour Yao-Ming). The results also show the immediate integration of world knowledge, but more
interestingly, violations of both world knowledge and semantics elicited a similar N400 effect at the
final word of the sentence. The results indicate that the reader not only immediately integrates world
knowledge once it appears, but also re-integrates it at the end of the sentence using the global
sentence-level context to confirm the previous violations. Thus, if there is new information by the end
of the sentence which makes the previous world knowledge violation acceptable, can the reader
immediately update her understanding to acquire the re-coherent meaning?
On the discourse level, Kintsch has clearly specified the role of discourse context and world
knowledge. In Kintsch's model, it is proposed that the reader automatically relates sentences to
discourse context and establishes the overall meaning, then requires conscious effort to integrate
necessary world knowledge in her mental model representation to reach a correct understanding
(Kintsch, 1998). However, recent studies have suggested that both global discourse context and local
world knowledge or lexical semantics are integrated immediately (Hald et al., 2007; Nieuwland & Van
Berkum, 2006). Nieuwland and Van Berkum (2006) found that given a suitable discourse context (e.g.,
a story about an amorous peanut), animacy violation predicates (“the peanut was in love”) were
actually processed more easily than canonical predicates (“the peanut was salted”). Moreover, Hald
et al. (2007) presented participants with two types of discourse context: a context supporting world
98
G. Xu et al. / Journal of Neurolinguistics 33 (2015) 96e103
knowledge, or a specific local context which made a world knowledge violation more acceptable; these
were followed by sentences with correct or anomalous world knowledge. The results showed that the
N400 effect evoked by the world knowledge violation was weaker when the sentences were preceded
by a local context. These results also revealed that a previous discourse context can immediately
modulate subsequent local lexical-semantic and world knowledge integration.
To our knowledge, little is known about how readers understand the subsequent context under the
constraint of a previous world knowledge violation, especially in relation to the sentence-level context.
Here we conducted an ERP experiment to address this question. Three experimental types of sentences
(each consisting of five Chinese words) were used here (see Table 1). The critical words are in italics and
final words are in bold. If the new sentence context can immediately update the previous world
knowledge violation constraint, we would expect that in comparison with the correct condition, the
critical words in both incoherent and re-coherent conditions would elicit the same N400 effect but it
would be elicited only by the final word in the incoherent condition.
2. Material and methods
2.1. Participants
Fifteen undergraduate native speakers of Chinese (8 males, mean age ¼ 19.3 years, SD ¼ 0.8)
participated in the experiment. They were paid a small fee for their participation. All participants were
strongly right-handed and with normal or corrected-to-normal vision. None of them had any neurological impairment or had used psychoactive medication, nor had any of them participated in the pretest. All participants gave their informed consent.
2.2. Stimulus material
The materials comprised 540 Chinese sentences. Each sentence consisted of five Chinese words.
Sentences varied in length from 10 to 13 characters. 405 of the sentences formed 135 experimental
sentence pairs. Each pair consisted of a correct sentence, an incoherent sentence that violated world
knowledge, a re-coherent sentence that violated world knowledge in the middle but added new information to make the violation more acceptable in the end. Each pair contained the same critical word
(in italics in Table 1) but different final words (in bold in Table 1). There were three versions of the
experiment so that each participant read only one type of the sentences in each pair. Each version
contained 45 exemplars of each of the three types. The other 135 sentences were unrelated filler
sentences which contained 45 correct sentences and 90 incorrect ones to make sure the items with and
without violations were equal in number. Participants read just one version of the experimental
sentences but the same 135 filler sentences, making 270 sentences in total. These sentences were
randomized. A pre-test was performed on the materials (described in below).
2.3. Pre-test
Results indicating that world knowledge violations were not based on semantic violations was
established in Jin et al. (2009). Here the 3 versions of the sentences were randomly presented to fifteen
Table 1
Example sentences and English equivalents of the five different conditions: correct, incoherent, re-coherent, correct filler, and
incorrect filler. The critical words are in italics and final words are in bold.
Sentence types
Chinese sentences
English equivalents
Correct
Incoherent
Re-coherent
Correct filler
Incorrect filler
明星。
高大的/姚明/是/篮球/明
矮小的/姚明/是/篮球/明
明星。
矮小的/姚明/是/餐厅/厨
厨师。
公司的/领导/在/召开/会议。
今年的/春节/在/农历的/七月。
That tall Yao-Ming is a basketball star.
That short Yao-Ming is a basketball star.
That short Yao-Ming is a restaurant cook.
The company leader is holding a conference.
This Spring Festival is in lunar July.
G. Xu et al. / Journal of Neurolinguistics 33 (2015) 96e103
99
subjects, who rated their acceptability on a scale from 1 (“very good”) to 5 (“very bad”). These subjects
were undergraduate native Chinese speakers as those in the formal test. The average rating for correct
sentences was 1.61 (SD ¼ 0.32), for incoherent sentences it was 4.14 (SD ¼ 0.33), for re-coherent
sentences it was 1.93 (SD ¼ 0.41), for correct filler sentences it was 1.56 (SD ¼ 0.29), and for incorrect filler sentences it was 4.43 (SD ¼ 0.21). A univariate ANOVA over the average ratings indicated
significance [F (4, 56) ¼ 372.67, P < 0.001]. Multiple comparison results indicated that the correct, recoherent and correct filler sentences were equally acceptable (P > 0.05), while the incoherent and
incorrect filler sentences were equally unacceptable (P ¼ 0.153).
2.4. Procedure
Participants were tested in a dimly lit, sound-attenuated booth. They were seated in a comfortable
reclining chair. Participants were asked to carefully read the sentences and to try to understand them as
fully as possible.
Each trial was presented word by word in white Song typeface (30-point foot size) against a dark
background in the centre of a VGA computer screen. The viewing distance was approximately 75 cm.
Each word was presented for 400 ms, followed by a blank screen for 400 ms before the next word
appeared. After the final word of the sentence, a blank screen appeared, the duration randomly varied
between 500 ms and 1100 ms, followed by an asterisk which was displayed in the middle of the screen
for 1000 ms. There was a 500e1100 ms random jitter between the disappearance of the asterisk and
the start of the next trial. Sentences were presented using E-prime 1.1. Each participant had 12 practice
trials and then completed 5 experimental blocks with short breaks in between.
2.5. EEG recording and analysis
The electroencephalogram (EEG) was continuously recorded with a sampling frequency of 1000 Hz
from the scalp with 64 Ag/AgCl sintered electrodes mounted in an elastic cap. The electrode sites were
placed according to the 10e20 international system. All electrodes were referenced to the left mastoid
and off-line re-referenced to the average of the left and right mastoids. Vertical eye movements were
monitored via supra-to to sub-orbital bipolar montage of the left eye. The EEG and electrooculogram
(EOG) recordings were amplified with two 32-channel BrainAmp MR Plus amplifiers from Brain
Products, using a high-cut-off 70 Hz (no notch filter) and a time constant of 10 s (0.016 Hz). Electrode
impedances were kept below 5 kU. The software package Vision Analyzer by Brain Vision was used to
analyze the waveforms.
After re-referencing to the average of both mastoids, the EEG signal was filtered with a low pass of
30 Hz. Ocular artifacts were automatically corrected. Epochs that ranged from 150 ms to þ800 ms,
related to the onset of critical word and final word, were segmented to a 150 ms pre-onset baseline.
Segments with potentials exceeding ±80 mV were rejected. Finally, ERPs time-locked to the onset of
critical word and final word were averaged respectively, for each condition (correct, incoherent, recoherent) for each participant at each electrode site.
Repeated measures analyses of variance (ANOVA) with the factors Condition (correct, incoherent,
re-coherent) and Position (left hemisphere: F3, FC3, C3, CP3; midline: Fz, FCz, Cz, CPz; right hemisphere: F4, FC4, C4, CP4) were carried out. When appropriate, the estimated Greenhouse-Geisser coefficient ε was corrected for violations of the sphericity assumption (Geisser & Greenhouse, 1958). All
reported P-values were based on corrected degrees of freedom, while the stated degrees of freedom
were uncorrected.
3. Results
3.1. ERPs elicited by critical word
The visual inspection of the waveforms time-locked to the onset of critical word (see Fig. 1) showed
an N1 followed by a P2 and an N400. Here we focused on the particular ERP component N400. At
around 300 ms, the incoherent and re-coherent conditions deviated from the correct condition.
100
G. Xu et al. / Journal of Neurolinguistics 33 (2015) 96e103
Fig. 1. Grand-average ERPs time-locked to the onset of critical word at Cz (left): correct condition (full line), incoherent condition
(dashed line), and re-coherent condition (dotted line). Spline-interpolated isovoltage maps display the topographic distributions of
the time window of 300e500 ms (right): correct condition (top), incoherent condition (middle), re-coherent condition (bottom).
Moreover, the topographic distribution showed that the N400 was maximal over central and parietal
sites, without a clear hemispheric dominance. A latency window of 300e500 ms after critical word
onset was used to compute the amplitude of N400.
Repeated measures ANOVA failed to show a significant Condition Position interaction [F (4,
56) ¼ 1.304, P ¼ 0.288]. However, there was a significant main effect of Condition [F (2, 28) ¼ 6.333,
P ¼ 0.005]. Additionally, both incoherent and re-coherent conditions resulted in a larger N400
amplitude than the correct condition (P ¼ 0.019, P ¼ 0.032 respectively). However there was no significant difference between incoherent condition and re-coherent condition (P ¼ 1). In addition, the
main effect of position reached significance [F (2, 28) ¼ 6.475, P ¼ 0.008]. The amplitudes at midline
sites were larger than that at left and right hemisphere sites (P ¼ 0.078, P ¼ 0.002 respectively); no
significant difference was found between the left and right hemisphere sites (P ¼ 0.982).
3.2. ERPs elicited by final word
The visual inspection of the waveforms time-locked to the onset of final word (see Fig. 2) showed an
N1 followed by a P2 and a negative deflection with a bi-modal morphology, peaking at about 320 ms
and 420 ms respectively. At around 300 ms, the incoherent and re-coherent conditions deviated from
the correct condition. This effect was followed by a different modulation between 380 ms and 480 ms,
with peak latency slightly later than usually seen for N400. Here they have been labelled as N320 and
N420 respectively.
Firstly, the amplitudes in the time window of 300e380 ms were tested. As can be seen in Fig. 2, both
the incoherent and re-coherent conditions elicited a clear negative deflection that peaked around
320 ms, labelled as N320 here. For this time window, repeated measures ANOVA on the amplitude
showed an approaching significant Condition Position interaction [F (4, 56) ¼ 2.514, P ¼ 0.075]. There
was a significant main effect of condition [F (2, 28) ¼ 5.925, P ¼ 0.007]. Both incoherent and recoherent conditions resulted in a larger N320 amplitude than the correct condition (P ¼ 0.043,
P ¼ 0.002 respectively), while there was no significant difference between incoherent condition and recoherent condition (P ¼ 1). Moreover, the main effect of position failed to reach significance [F (2,
28) ¼ 2.005, P ¼ 0.17].
G. Xu et al. / Journal of Neurolinguistics 33 (2015) 96e103
101
Fig. 2. Grand-average ERPs time-locked to the onset of final word at Cz, epochs ranged from 100 to 500 ms for graphic display
(left): correct condition (full line), incoherent condition (dashed line), and re-coherent condition (dotted line). Spline-interpolated
isovoltage maps display the topographic distributions of the time window of 300e380 ms and 380e480 ms, respectively (right):
correct condition (top), incoherent condition (middle), re-coherent condition (bottom).
Secondly, the amplitudes in the time window of 380e480 ms were also tested. As can be seen in
Fig. 2, the incoherent condition elicited another clear negative deflection that peaked around 420 ms,
labelled as N420 here. For this time window, repeated measures ANOVA on the amplitude showed an
approaching significant Condition Position interaction [F (4, 56) ¼ 3.057, P ¼ 0.067]. There was a
significant main effect of condition [F (2, 28) ¼ 6.456, P ¼ 0.005]. Only the incoherent condition
resulted in a larger amplitude than the correct condition (P ¼ 0.007). However, there was no significant
difference between re-coherent condition and correct condition (P ¼ 0.14). Additionally, the main
effect of position failed to reach significance [F (2, 28) ¼ 1.75, P ¼ 0.184].
To sum up, the results indicate that incoherent and re-coherent conditions modulate N400 elicited
by the critical word, and N320 elicited by the final word, in a similar way in terms of latency, amplitude
and topographical distribution, while there was a different pattern for N420, elicited by the final word
in incoherent condition and re-coherent condition.
4. Discussion
We conducted an ERP experiment to investigate how readers integrate sentence context under the
constraints of previous world knowledge violation. Unlike previous studies (Hald et al., 2007;
Nieuwland & Van Berkum, 2006), in this experiment the sentence-level context which made the
world knowledge violation more acceptable was placed after the violation occurred. Here, we found
similar N400 and N320 effects elicited by the critical word and the final word respectively, both in
incoherent and re-coherent conditions, and an additional N420 effect elicited only by the final word in
the incoherent condition.
In line with previous studies, the critical word elicited a clear N400 effect, with equivalent onset and
peak latencies for both incoherent and re-coherent conditions, indicating the rapid parallel integration
of world knowledge in the middle of the sentence as in Jin et al. (2009). Taken together, these findings
indicate that world knowledge is immediately integrated whenever it is encountered by a reader.
Additionally, the final words elicited two negative deflections during the 300e500 ms window.
Both the incoherent and re-coherent conditions displayed an N320 effect, while only the incoherent
condition displayed an N420 effect. The results suggest that subsequent global sentence context
immediately influences a previous world knowledge violation constraint so that the global coherent
102
G. Xu et al. / Journal of Neurolinguistics 33 (2015) 96e103
sentence meaning is obtained. Based on the time window in which the two deflections occurred and
the semantic processing in which they were involved, we analyze them as sub-components of the
€nen & Marantz, 2003) have also proposed that the N400 can be
N400. Pylkk€
anen and Marantz (Pylkka
decomposed. We suggest that the N320 and N420 might reflect two different aspects of final word
integration. The N320 effect, elicited by the final word in both incoherent and re-coherent conditions,
may reflect the integration of the final word itself with its low predictability, because the reader cannot
predict from earlier information whether the final word will remain anomalous or become re-coherent
according to previous information. The amplitude of N400 can also be modulated by the degree of
predictability (Kutas & Federmeier, 2011; Kutas & Hillyard, 1984; Lau et al., 2008), while the N420
effect, elicited only by the final word in the incoherent condition may reflect the integration of the final
word with the global sentence context (George, Mannes, & Hoffinan, 1994; George, Mannes, &
Hoffman, 1997; Kutas, Van Petten, & Besson, 1988; Lau et al., 2008; Salmon & Pratt, 2002; Van
Berkum, Brown, Zwitserlood, Kooijman, & Hagoort, 2005).
Moreover, considering the same world knowledge violation condition in Jin et al. (2009) and the
current study, the final word in the former study elicited an N400 effect with only one negative
deflection, while in this study an N400 effect with two negative deflections was elicited. We assume
that the different experimental situations in the two studies may account for this difference. In Jin et al.
(2009), the global context was consistent with the previous violation, whereas in the current study, the
global context could be either consistent or inconsistent with the previous violation. Participants in the
current study were therefore unable to make a confident prediction of the nature of the final word,
leading to an additional N320 effect. From the difference, we can see the rapid attempts by the brain to
immediately process the new information to understand a sentence.
To sum up, we placed world knowledge before sentence context and we found the new sentence
context got round the previous world knowledge violation constraint. Depending on the global sentence context integration, the reader can finally judge whether an earlier world knowledge violation is
a real violation or just new semantic information to be used for constructing a coherent sentence
meaning. “Here-and-now view” of “constructionist theory” in discourse comprehension held that the
reader was constantly interpreting current information, integrating the current sentence with previous
ones, and updating the established mental representation based on the new information (Bower &
Morrow, 1990). Based on the result we found here, “Here-and-now view” could also be applied to
sentence comprehension.
Acknowledgements
This study was supported by grants from the National Natural Science Foundation of China
(#31170997), and National Basic Research Program of China (973 Program) (#2012CB720701). We
thank Zhixin Ye, Shufang Wang, Luojin Zhong, Nairui Guo, Hongwei Gao, Zhuoer Ye, Meng Fan for their
help in data collection, and all the participants in this experiment. Special thanks go to Johnny and
Elizabeth Pang for their helpful comments on our first draft and Professor Bencie Woll for her very
valuable comments on the final manuscript.
References
Bower, G. H., & Morrow, D. G. (1990). Mental models in narrative comprehension. Science, 247(4938), 44e48.
Chwilla, D. J., & Kolk, H. H. J. (2005). Accessing world knowledge: evidence from N400 and reaction time priming. Cognitive
Brain Research, 25(3), 589e606. http://dx.doi.org/10.1016/j.cogbrainres.2005.08.011.
Clark, H. H. (1996). Using language. Cambridge University Press.
Geisser, S., & Greenhouse, S. W. (1958). An extension of Box's results on the use of the F distribution in multivariate analysis. The
Annals of Mathematical Statistics, 885e891.
George, M. S., Mannes, S., & Hoffinan, J. E. (1994). Global semantic expectancy and language comprehension. Journal of Cognitive
Neuroscience, 6(1), 70e83.
George, M. S., Mannes, S., & Hoffman, J. E. (1997). Individual differences in inference generation: an ERP analysis. Journal of
Cognitive Neuroscience, 9(6), 776e787.
Hagoort, P., Hald, L., Bastiaansen, M., & Petersson, K. M. (2004). Integration of word meaning and world knowledge in language
comprehension. Science, 304(5669), 438e441.
Hagoort, P., & Van Berkum, J. J. A. (2007). Beyond the sentence given. Philosophical Transactions of the Royal Society B: Biological
Sciences, 362(1481), 801e811. http://dx.doi.org/10.1098/rstb.2007.2089.
G. Xu et al. / Journal of Neurolinguistics 33 (2015) 96e103
103
Hald, L. A., Steenbeek-Planting, E. G., & Hagoort, P. (2007). The interaction of discourse context and world knowledge in online
sentence comprehension. Evidence from the N400. Brain Research, 1146, 210e218. http://dx.doi.org/10.1016/j.brainres.2007.
02.054.
Jackendoff, R. (2002). Foundations of language: Brain, meaning, grammar, evolution. Oxford University Press.
Jackendoff, R. (2007). A parallel architecture perspective on language processing. Brain Research, 1146, 2e22. http://dx.doi.org/
10.1016/j.brainres.2006.08.111.
Jin, H., Zhong, W. F., Xu, G. P., Cai, M. X., Yang, Y. F., & Mo, L. (2009). The time course of world knowledge integration in sentence
comprehension. Acta Psychologica Sinica, 41(7), 565e571.
Kempson, R. (2003). Pragmatics: language and communication. In M. Aronoff, & J. Rees-Miller (Eds.), The handbook of linguistics
(pp. 394e427). Oxford: Blackwell Publishers Inc.
Kintsch, W. (1998). Comprehension: A paradigm for cognition. Cambridge University Press.
Kutas, M., & Federmeier, K. D. (2011). Thirty years and counting: finding meaning in the N400 component of the event-related
brain potential (ERP). Annual Review of Psychology, 62(1), 621e647. http://dx.doi.org/10.1146/annurev.psych.093008.131123.
Kutas, M., & Hillyard, S. A. (1980). Reading senseless sentences: brain potentials reflect semantic incongruity. Science, 207(4427),
203e205.
Kutas, M., & Hillyard, S. A. (1984). Brain potentials during reading reflect word expectancy and semantic association. Nature, 307,
161e163.
Kutas, M., Van Petten, C., & Besson, M. (1988). Event-related potential asymmetries during the reading of sentences. Electroencephalography and Clinical Neurophysiology, 69(3), 218e233.
Lau, E. F., Phillips, C., & Poeppel, D. (2008). A cortical network for semantics: (de)constructing the N400. Nature Reviews
Neuroscience, 9(12), 920e933. http://dx.doi.org/10.1038/nrn2532.
MacDonald, M. C., Pearlmutter, N. J., & Seidenberg, M. S. (1994). Lexical nature of syntactic ambiguity resolution. Psychological
Review, 101(4), 676e703.
Macizo, P., & Herrera, A. (2010). What is heavier, a kilo of feathers or a kilo of lead? What you know determines what you
understand. NeuroReport, 21(14), 938e942. http://dx.doi.org/10.1097/WNR.0b013e32833e3310.
Nieuwland, M. S., & Van Berkum, J. J. A. (2006). When peanuts fall in love: N400 evidence for the power of discourse. Journal of
Cognitive Neuroscience, 18(7), 1098e1111.
Perry, J. (1997). Indexicals and demonstratives. In R. Hale, & C. Wright (Eds.), Companion to the philosophy of language (pp.
586e612). Oxford: Blackwells Publishers Inc.
€nen, L., & Marantz, A. (2003). Tracking the time course of word recognition with MEG. Trends in Cognitive Sciences, 7(5),
Pylkka
187e189.
Salmon, N., & Pratt, H. (2002). A comparison of sentence-and discourse-level semantic processing: an ERP study. Brain and
Language, 83(3), 367e383.
Tanenhaus, M. K., Spivey-Knowlton, M. J., Eberhard, K. M., & Sedivy, J. C. (1995). Integration of visual and linguistic information
in spoken language comprehension. Science, 268(5217), 1632e1634.
Van Berkum, J. J. A. (2008). Understanding sentences in context what brain waves can tell us. Current Directions in Psychological
Science, 17(6), 376e380.
Van Berkum, J. J. A., Brown, C. M., Zwitserlood, P., Kooijman, V., & Hagoort, P. (2005). Anticipating upcoming words in discourse:
evidence from ERPs and Reading times. Journal of Experimental Psychology: Learning, Memory and Cognition, 31(3), 443e467.
http://dx.doi.org/10.1037/0278-7393.31.3.443.
Van Berkum, J. J. A., Hagoort, P., & Brown, C. M. (1999). Semantic integration in sentences and discourse: evidence from the
N400. Journal of Cognitive Neuroscience, 11(6), 657e671.
Van Berkum, J. J. A., Zwitserlood, P., Hagoort, P., & Brown, C. M. (2003). When and how do listeners relate a sentence to the wider
discourse? Evidence from the N400 effect. Cognitive Brain Research, 17(3), 701e718. http://dx.doi.org/10.1016/s09266410(03)00196-4.