HHS Public Access

Author manuscript
Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
Author Manuscript

Published in final edited form as:

Lang Cogn Neurosci. 2023 ; 38(1): 88–104. doi:10.1080/23273798.2022.2091150.

ERPs Reveal How Semantic and Syntactic Processing

Unfold across Parafoveal and Foveal Vision during Sentence
Comprehension
Chuchu Li1, Katherine J. Midgley2, Phillip J. Holcomb2
1University of California, San Diego
Author Manuscript

2San Diego State University

Abstract
We examined how readers process content and function words in sentence comprehension with
ERPs. Participants read simple declarative sentences using a rapid serial visual presentation
(RSVP) with flankers paradigm. Sentences contained either an unexpected semantically
anomalous content word, an unexpected syntactically anomalous function word or were well
formed with no anomalies. ERPs were examined when target words were in the parafoveal or
foveal vision. Unexpected content words elicited a typically distributed N400 when displayed
in the parafovea, followed by a longer-lasting, widely distributed positivity starting around 300
ms once foveated. Unexpected function words elicited a left lateralized LAN-like component
when presented in the parafovea, followed by a left lateralized, posteriorly distributed P600
Author Manuscript

when foveated. These results suggested that both semantic and syntactic processing involve two
stages—the initial, fast process that can be completed in parafovea, followed by a more in depth
attentionally mediated assessment that occurs with direct attention.

Keywords
Reading; ERPs; Parafoveal and foveal processing; Semantic and syntactic anomaly

Introduction
Readers process both semantic and syntactic information to comprehend sentences, and
content versus function words are two classes of words that differ in the degree to which
Author Manuscript

they carry semantic versus syntactic information. Content words like nouns (e.g., dog),
lexical verbs (e.g., eat), or adjectives (e.g., red) have rich semantic information. In contrast,
function words like determiners (e.g., the), pronouns (e.g., he), or prepositions (e.g., in)
carry little substantive meaning but reveal grammatical relationships between content words
(e.g., Neville, Mills, & Lawson, 1992). This difference might explain why patients with
impaired grammatical analysis abilities due to Broca’s aphasia typically report difficulties

Please send correspondence to: Dr. Chuchu Li, Department of Psychology, University of California, San Diego, 9500 Gilman Drive,
La Jolla, CA 92093-0948, USA, [email protected].
Declaration of Interest Statement
None
 Li et al. Page 2

in processing function words (e.g., Caramazza & Zurif, 1976; Berndt & Caramazza, 1980;
Author Manuscript

Bradley et al., 1980; Swinney et al., 1980; Friederici, 1983, 1985; Goodglass & Menn, 1985;
Rosenberg et al., 1985; Friederici & Kilborn, 1989; Friederici et al., 1991; Pulvermüller,
1995; Biassou et al., 1997). The investigation of neurological differences in processing
these two classes of words has been conducted since the 1980s (e.g., Bradley & Garrett,
1983; Friederici, 1985; Friederici, Opitz, & von Cramon, 2000; Nobre, Price, Turner, &
Friston, 1997; Nobre & McCarthy, 1994). The present study addressed this issue from a new
perspective — instead of directly comparing the processing of these two word classes, we
instead examined differences during sentence processing between expected and unexpected
content and function words using a modified rapid visual presentation (RSVP) paradigm.
Importantly, using the event-related potential (ERP) methods, we examined how readers
process content and function words not only when these words were presented in the fovea
but also just prior to foveation while they were presented in the parafovea. This latter feature
Author Manuscript

of the modified RSVP paradigm (henceforth RSVP with flankers) more closely resembles
natural reading.

Prior ERP studies have revealed how content and function words are processed differently
via directly comparing them in reading tasks. When reading single words or sentences
that were presented one word at a time in the standard RSVP paradigm, content words
tend to elicit more negative-going deflections that are largest over central-parietal brain
regions (Brown, Hagoort & ter Keurs, 1999; Neville et al., 1992; Nobre & McCarthy,
1994; ter Keurs et al., 1999, but see Münte, Wieringa, Weyerts, Szentkuti, Matzke, &
Johannes, 2001; Van Petten & Kutas, 1991). These negativities typically peak between
300 and 500 ms (N400) after word onset. Because the N400 is generally larger in
amplitude when content words are not supported by the semantic context of prior words
in a sentence, this component has been argued to reflect lexico-semantic operations during
Author Manuscript

sentence processing (Kutas & Hillyard, 1980, 1984). On the other hand, function words
tend to elicit more negative-going deflections that are largest over left anterior brain
regions between 250 and 500 ms. Neville et al. (1992) was the first to propose that such
left-anterior negativities (LANs) that reflect syntactic operations specific to function words
during sentence processing (a component they identified as the N280), although subsequent
studies have questioned this interpretation (ter Keurs, Brown, Hagoort, & Stegemann, 1999;
Brunelliere, Hoen, & Dominey, 2005; Osterhout, Bersick, & McKinnon, 1997). Other
studies have reported similar LAN-like components in paradigms using word-category and
agreement violations (e.g., Coulson et al., 1998; Kim & Sikos, 2011; O’Rourke & Van
Petten, 2011; see Molinaro, Barber, & Carreiras, 2011 for a review; but see Tanner, Morgan-
Short, & Luck, 2015). Such violations also tend to generate large posterior positivities
referred to as the P600 (Osterhout & Holcomb, 1992). Subsequent studies have shown
Author Manuscript

that the P600 is a positive going deflection that peaks around 600 ms post-stimulus onset
and is sensitive to grammatical anomalies and/or syntactic processing difficulty (Osterhout,
Holcomb & Swinney, 1994), such as subject–verb agreement violations (e.g., Hagoort,
Brown, & Groothusen, 1993), inflection violations (e.g., Friederici, Pfeifer, & Hahne, 1993),
or difficult syntactic integration even in grammatical sentences (e.g., Kaan, Harris, Gibson,
& Holcomb, 2000). In summary, content and function words seem to engage, at least to
some degree different neural systems, likely due to richer semantic features for the former

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 3

(which contribute to the N400) and richer syntactic features for the latter (which contribute
Author Manuscript

to the LAN and the P600).

The first aim of the present study was to examine how content versus function words
contribute differently to semantic and syntactic processing during sentence processing. To
accomplish this in some sentences (controls) all content and function words were consistent
with the meaning and syntax of the sentence while in other (experimental) sentences, either
a content or function word was replaced by another word belonging to the same word class
which rendered these sentences difficult to process from this point on. This design allows for
two important controls over potentially confounding factors. First, the same sentence frames
are used in experimental and control sentences allowing for the investigation of differences
in ERP patterns elicited by unexpected content or function words while holding sentence
context constant. Second, by only contrasting control and experimental items within a
word-class variables such as word length and frequency are controlled.
Author Manuscript

We focused on three ERP components, N400, LAN, and P600. The N400, as described
earlier, is a negative shift that usually peaks around 400 ms after word onset and sensitive
to semantic anomalies. The LAN and P600 are two ERP components that have been shown
to be sensitive to syntactic processing difficulties. While an unexpected content word should
elicit an N400, an unexpected function word should elicit both a LAN and a P600 given
these two components increased involvement in syntactic processing.

A second aim was to examine semantic and syntactic violation effects in a more realistic
sentence reading task. Most of the research using ERPs to examine semantic and syntactic
processing in sentences have adopted the word-by-word RSVP paradigm where individual
words are presented at fixation. This approach is used in order to minimize the role of
Author Manuscript

saccadic artifacts during natural sentence reading but also to provide an unambiguous time-
lock point (word onset) to assess differences between conditions. However, the artificiality
of this method of sentence reading leaves many questions about the interpretation of
differences between conditions. In particular there is ample evidence that during normal
reading information from words to the right of fixation influence processing prior to their
being fixated in central (foveal) vision (e.g., Briihl & Inhoff, 1995; Gordon, Plummer,
& Choi, 2013; Inhoff, 1989a, 1989b, 1990; Inhoff & Tousman, 1990; Pollatsek, Lesch,
Morris, & Rayner, 1992; Rayner, 1998; but see Altarriba, Kambe, Pollatsek, & Rayner,
2001; Rayner, Balota, & Pollatsek, 1986; Rayner & Morris, 1992; Mirault, Yeaton, Broqua,
Dufau, Holcomb, & Grainger, 2020). In particular, readers benefited from the preview of
a semantically plausible word, suggesting early processing of parafoveal words regarding
whether they semantically fit previous context (Schotter & Jia, 2016; Veldre & Andrews,
Author Manuscript

2016, 2018). A limitation of the boundary change paradigm used in the above studies
was that there might be some confounding of eye-movements and/or the presentation of
parafoveal words (e.g., the inclusion of both valid and invalid preview). Although the
traditional word-by-word RSVP paradigm in ERP studies may address the confounding
issue, parafoveal information is not available in that paradigm.

Some recent ERP studies support concerns about the lack of parafoveal information in
the traditional RSVP paradigm. These studies strongly suggest that words presented in

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 4

parafoveal vision are not only processed at a sub-lexical level but also receive some degree
Author Manuscript

of semantic analysis as evidenced by parafoveal N400 effects (Barber, Doñamayor, Kutas,

& Münte, 2010; Barber, van der Meij, & Kutas, 2013; Li, Niefind, Wang, Sommer, &
Dimigen, 2015; Payne & Federmeier, 2017; Payne, Stites, & Federmeier, 2019; Stites, Payne
& Federmeier, 2017; Snell, Meade, Meeter, Holcomb, & Grainger, 2019; Meade, Declerck,
Holcomb, & Grainger, 2021; Mirault, et al., 2020; but see Dimigen, Kliegl, & Sommer,
2012). Barber and colleagues (2010, 2013) developed the RSVP with flankers paradigm
to examine ERPs that are elicited by targets in the parafovea. In this paradigm, sentences
are presented serially in triads. Participants are instructed to fixate on the target word at
the center, which is flanked by the upcoming word in the sentence to the right, and the
preceding word to the left. Adopting this paradigm, previous research has found N400
effects that can be elicited by unexpected content words while they are still in the parafovea
– moreover these effects have been found in multiple languages (e.g., English and Hebrew in
Author Manuscript

Barber et al., 2010; Chinese in Li et al., 2015). Note that in Hebrew the reading direction is
from the right to the left, so that the upcoming words that elicited parafoveal N400s were on
the left of the central word. Stites et al. (2017) further showed that this parafoveal N400 was
continuously graded with increasing cloze probability, replicating the graded foveal N400
pattern that is typically found in studies using the typical RSVP paradigm (e.g., Federmeier,
Wlotko, De Ochoa-Dewald, & Kutas, 2007; Kutas & Hillyard, 1984; Wlotko & Federmeier,
2012). Curiously, Stites et al. found that the presence of a parafoveal N400 mitigated or even
eliminated the N400 when the unexpected word was subsequently foveated. This seems to
suggest that readers did not only access the semantic information of the target word when
it is in the parafovea, but also largely completed this semantic analysis before the target
is foveated. The influence of words in the parafovea has also been shown to influence the
N400 in word-pair tasks. Using a trans-saccadic priming paradigm, Grainger, Midgley, and
Holcomb (2016) found foveal-to-parafoveal lexical priming effects on the N400.
Author Manuscript

Although semantic information has been processed when the target word is in the parafovea,
there is evidence that there is a second step that engages readers’ foveal attention on
the target word to complete target word processing in a sentence context. Using the
RSVP with flankers paradigm, Payne et al. (2019) found a posteriorly distributed late
positive component (LPC or P600-like component), when an anomalous content word was
fixated in the fovea. This effect was not seen when these words were in the parafovea.
Importantly, this P600-like component was only observed when participants were asked to
judge the plausibility of sentences and was not seen when sentences were read passively for
comprehension. This latter finding is consistent with Stites et al. (2017) where participants
completed a memory task after sentence comprehension, in which a P600-like effect was
not observed when the target was foveated. Payne et al. suggested that the foveal P600-
Author Manuscript

like component reflects a higher-order and task-dependent integrative process that engages
foveal attention, i.e., a result of plausibility-related integration failures (Brouwer, Crocker,
Vanhuizen, & Hoeks, 2017; DeLong, Quante, & Kutas, 2014; Van Petten & Luka, 2012).

To date ERP studies that adopted the RSVP with flankers paradigm have focused on
unexpected content words that result in semantic anomalies. However, none of this work
has examined unexpected function words that elicit syntactic anomalies, although results
from eye-tracking studies have shown some evidence of parafoveal syntactic processing in

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 5

normal reading (Brothers & Traxler, 2016; Veldre & Andrews, 2018). Accordingly, adopting
Author Manuscript

the RSVP with flankers paradigm, the present study focused on ERPs recorded to expected
and unexpected function words, first presented in the parafovea followed by presentation
to the fovea. These function word effects were then contrasted with comparable expected
and unexpected content word processing. Given that the difference between expected and
unexpected function words should involve greater differential syntactic processing we aimed
to determine whether such effects would be apparent when the target function words are
in the parafovea. Moreover, we were interested in examining whether processing function
words might also involve a two-step process as reported for content words by Payne et al.
In other words, might we see a first effect of unexpected function words when they are
presented in the parafovea followed by a second effect when these words are presented at
fixation and engage foveal attention.

Following previous studies, we predicted a robust parafoveal N400 and a foveal P600-like
Author Manuscript

component on unexpected content words. Furthermore, if readers also process syntactic

information of the previewed targets, we predicted we would observe parafoveal LAN or
P600 effects on unexpected function words when they are still in the parafovea. Lastly,
if the syntactic analysis of function words is completed during parafoveal presentation
(i.e., one-step processing), we predicted that there should be little to no foveal LAN or
P600 effects on unexpected function words. However, if as suggested by Payne et al. for
content word processing, there are also two steps involved in function word processing when
a plausibility judgment is required, then unexpected function words might also elicit an
enhanced foveal LAN or P600.

Methods
Author Manuscript

Participants
Twenty-four right-handed, native English speakers (10 males, mean age = 22.0 years,
ranging from 19 to 27) received compensation for their participation. All participants
reported normal or corrected-to-normal visual acuity with no history of neurological
impairment or language disability. All participants reported that they had learned no other
language before the age of six, and that they were not proficient in any languages other than
English.

Stimuli & Design

The critical stimuli for this study were 120 sentences ranging from 7 to 17 words. There
were three versions of each sentence: a) the normal version without semantic or syntactic
errors and containing a control content word and a control function word (see sentences
Author Manuscript

“a1” and “a2” below); b) a semantic violation version where the appropriate control content
word was replaced by an unexpected content word which produced a semantic anomaly (see
sentences “b1” and “b2” below); and c) a syntactic violation version where the appropriate
control function word was replaced by an unexpected function word which produced a
syntactic anomaly (see sentence “c1” and “c2” below). None of the target words were
sentence final words. The target content and function words were at most one word away

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 6

from each other to ensure the context around the replaced words were similar. For example,
Author Manuscript

in the sentences

a1) The employee needs to prepare the document for his boss this morning.

a2) The old man was asleep in the chair when I came back.

The word boss/chair served as the control content words, and his/the served as the control
function words. While boss/chair was replaced by water/cherry in the unexpected content
word condition, his/the was replaced by on/of in the unexpected function word condition. As
a result, the sentences in these two conditions were

b1) The employee needs to prepare the document for his water this morning.

b2) The old man was asleep in the cherry when I came back.

c1) The employee needs to prepare the document for on boss this morning.
Author Manuscript

c2) The old man was asleep in of chair when I came back.

Though we included sentences with various types of semantic and syntactic anomalies
we did not control the number of sentences of each subtype, which was not the focus of
the present study. For example, for sentences with semantic anomalies, 12 out of the 120
were due to animacy violations (but none were role-reversed sentences), while others were
implausible events. For sentences with syntactic anomalies, in 29 out of the 120 sentences an
article was replaced by a preposition or vice versa (like c2); in 27 a preposition was replaced
by a pronoun or vice versa (like c1); in 21 a preposition was replaced by another preposition
(e.g., of-for); in nine a conjunction was replaced by a pronoun or vice versa (e.g., and-us);
in seven a conjunction was replaced by a preposition or vice versa (e.g., and-of); in six
an article was replaced by a pronoun or vice versa (e.g., the-he), in five a quantifier was
Author Manuscript

replaced by a preposition or vice versa (e.g., some-in); in five a preposition was replaced by
an auxiliary verb (e.g., on-will); and the remaining 11 between other types of function word
substitution.

We pre-tested each sentence in a group of six native English speakers that did not participate
in the ERP study to determine if readers would be aware of the anomalies produced by the
critical words at the point these words are presented in an RSVP format. On a 7-point scale
they rated the acceptability of each sentence through the point of the critical words (1 not
acceptable, and 7 completely acceptable). The ends of the sentences after the critical word
were not presented. For example,

• The employee needs to prepare the document for on

• The old man was asleep in the cherry

Author Manuscript

Raters made their judgment based on whether it’s acceptable to have the word on after
the context The employee needs to prepare the document for. See Table 1 for detailed
characteristics of critical words in each condition. We also recruited a different group of
English monolinguals (n = 26) to complete a cloze task in order to assess the predictability
of the critical sentences (e.g., they saw the incomplete sentence frame The employee needs
to prepare the document for___ and completed it with whatever words or phrases came into
their minds first). Predictability is indicated by the number of responses in which target

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 7

words in the normal version (i.e., the control condition) were entered as the first words
Author Manuscript

of their responses (e.g., his boss and his job both counted in function words, as in the
control condition his was the target word in the sentence The employee needs to prepare the
document for his boss). No participant used the target words from the unexpected conditions
(i.e., the semantic violation and syntactic violation versions) in any sentences. Although
function words were slightly more predictable than content words (M = 6.45 vs. 5.30;
meaning that for each critical sentence, out of 26 responses in each word class, 6.45 were
target function words vs. 5.30 were target content words), this difference was not significant
(t =1.42, p = .16).

In the ERP experiment the 120 critical sentences were evenly distributed in three
experimental lists in a Latin-square design. In this way, each list contained 40 critical
sentences in each condition, and no sentence repeated within a list. Stimuli were randomized
once within each list and then presented in the same order for each participant. Participants
Author Manuscript

were randomly assigned to a list. To elicit the higher-order and task-dependent P600-like
effect that engages foveal attention, we instructed participants to judge whether a sentence
makes sense after they finished reading each sentence. Their task was to press one button on
a gamepad labeled “yes” for good sentences and another labeled “no” for bad sentences. In
order to balance yes/no decisions 40 filler sentences without semantic/syntactic errors were
included in each list. Each participant read 160 sentences in total.

Procedure
All stimuli were presented on a 24-inch LCD gaming monitor set to a refresh rate of 100 Hz
and located 145 cm directly in front of the participant. The testing began with 10 practice
sentences to acclimate participants to the paradigm and experimental conditions.
Author Manuscript

A visual hemi-field RSVP with flankers paradigm was adopted in which each sentence was
presented three words at a time, with the target word at central fixation (foveal target),
the upcoming word on the right (parafoveal target), and the preceding word on the left.
For every triad, participants were instructed to fixate on the centered word throughout the
experiment (i.e., the foveal target word). To facilitate attention to the central foveal target
this item was displayed as white letters on a black background. The other two flanking
words were displaced as slightly dimmer grey letters. In addition to the color difference,
we also added two yellow vertical bars, one above and one below the foveal target to help
participants maintain fixation on the center of the screen. All words were presented in a
fixed width font (New Courier) with each character occupying a 20 × 40 pixel matrix.
Each triad was presented for 400ms, following which the parafoveal target became the
foveal target in the following 400 ms epoch (see Figure 1 – note this method of display
Author Manuscript

makes the sentence seem to “slide” one word at a time from right to left). Each sentence
presentation started with a fixation mark (“+”) at the center of the screen (i.e., where the
foveal target would be presented), and ended with a purple question mark (“?”) at the
center that indicating it was time to make the acceptability judgment. Participants were
asked to maintain fixation on the central stimulus and to minimize blinking during sentence
presentation, but were encouraged to make movements/blinks during the presentation of the
fixation and question mark, which did not disappear until participants pressed a continue

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 8

button at which point the fixation mark of the next trial was displayed. There were three
Author Manuscript

longer rest breaks every 40 sentences.

EEG Recording
Participants were seated in a comfortable chair in a sound attenuated darkened room. An
electro-cap fitted with tin electrodes was used to record continuous EEG from scalp 29 sites
(See Figure 2). Four additional electrodes were used: one over the left mastoid bone that
served as the reference site for all scalp electrodes; one over the right mastoid that was used
to monitor for differential mastoid activity; one below the left eye which together with the
FP1 electrode was used to monitor for vertical eye movement (i.e., blinks), and one at the
outer canthus of the right eye to monitor for horizontal eye movements. Impedances were
kept below 2.5 kΩ for all electrodes. EEG signals were amplified using Neuroscan Synamp
RT amplifiers with a bandpass of DC to 100 Hz. The signals were digitally sampled at 500
Author Manuscript

Hz throughout the experiment.

EEG Analysis
Separate ERPs from four types of critical words (control content vs. unexpected content,
control function vs. unexpected function) at two sentence positions (foveal target, parafoveal
target) were calculated time-locked to the onset of the target triad. Averaging began 100 ms
prior to stimulus onset and continued 1,000 ms thereafter. The resulting data were baselined
to the mean voltage in a period from −100 to 0 ms pre-target onset for parafoveal and foveal
targets, respectively. Trials with muscle artifact or eye movements (less than 5% of trials in
total) were excluded. We were especially careful to monitor for and reject trials with any
evidence of horizontal eye movements.

Repeated measures analyses of variance (ANOVAs) were used to analyze the ERP data.
Author Manuscript

We analyzed the critical content and function words at each sentence position (foveal vs.
parafoveal) separately. For both content and function words, for each position, the analysis
included within participant factors of Word Type (control vs. unexpected), and two electrode
position factors, Laterality (left vs. midline vs. right) and Ant-post (anterior-posterior, FP vs.
F vs. C vs. P vs. O – see Figure 2 for a diagram of the sites used). For all statistical analyses
Geisser-Greenhouse correction was used for all repeated measures factors with greater than
1 degree of freedom in the numerator (Geisser & Greenhouse,1959).

Results
Behavioral Results
Participants judged an average of 88.9% of the sentences correctly (SD = 3.0%). The
Author Manuscript

accuracy rates in conditions with a no response (unexpected function words: M = 91.5%,

SD = 9.1%; unexpected content words: M= 93.2%, SD = 5.1%) were higher than sentences
with a yes response (control: M= 85.4%, SD = 7.3%; fillers: 85.3%, SD = 6.1%; ps < .05).
This was probably due a high threshold for yes responses. The > 85% accuracy rate in all
conditions confirmed attentive reading.

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 9

Visual Inspection of ERPs

Author Manuscript

The parafoveal results are plotted in Figures 3 and 4 and are the grand mean ERP waveforms
for content and function words time-locked to the onset of the parafoveal targets. In these
figures, the epoch starts with two broadly distributed components peaking at around 100
ms (N1) and 250 ms (P2). Neither of these components appears to be influenced by the
Word Type variable. Starting at around 300 ms and continuing through 700 ms there was a
larger negative-going component that appears to be larger for unexpected words compared to
expected words.

The foveal results are shown in Figures 5 and 6 and are the grand mean ERPs for content
and function words time-locked to the onset of words in foveal target position. Just as in
the parafoveal plots (Figures 3 & 4) the ERPs in these figures start with broadly distributed
N1 and P2 components. Following these, for content words, the first visible component that
Author Manuscript

appears to differ between Word Types was a negative-going deflection between 0 and 150
ms after stimulus onset that was followed by a larger positive-going deflection starting from
300 ms post stimulus onset. For function words, the first and only visible difference for
unexpected words compared to expected words was a positive-going deflection starting from
650 ms after stimulus onset. Figure 7A shows voltage maps at the 350 – 550 ms and 650
– 950 ms time windows for parafoveal targets, while Figure 7B shows voltage maps at the
same time windows for foveal targets.

Epoch Analysis of ERPs

Parafoveal Targets 350 – 550 ms—For content words, an omnibus ANOVA on the
mean amplitude values in this epoch revealed significant main effects of Word Type (F (1,
23) = 9.14, p = .006), indicating that unexpected content words elicited more negative-going
Author Manuscript

ERPs than control words. There was also a significant interaction between Word Type and
Laterality (F (2,46) = 4.22, p = .021) with the Word Type effect size at the midline (F (1, 23)
= 11.62, p = .002) and the right column (F (1, 23) = 7.96, p = .010) being stronger than that
at the left column (F (1, 23) = 7.09, p = .014). There was also a marginally significant Word
Type × Ant-post interaction (F (4, 92) = 3.76, p = .051), further analyses of which showed
that the Word Type effect was significant at F sites (F (1, 23) = 5.0, p = .035), C sites (F (1,
23) = 18.95, p < .001), P sites (F (1, 23) = 12.2, p < .001) and O sites (F (1, 23) = 9.31, p =
.006), but not at FP sites (F (1, 23) <1, p = .83).

For function words, the omnibus ANOVA showed a significant Word Type × Laterality
interaction (F (2, 46) = 6.11, p = .011). Further analyses showed that the Word Type effect
was only significant at the left hemisphere column (F (1, 23) = 5.25, p = .031) and the
midline (F (1, 23) = 5.11, p = .033), but not at the right hemisphere column (F (1, 23) = 1.35,
Author Manuscript

p = .26).

Parafoveal Targets 650 – 950 ms—In this epoch, for both content and function words,
Word Type did not show a significant main effect or interaction with the two electrode
position factors (ps > .37).

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 10

Foveal Targets 350 – 550 ms—Analyses of this epoch showed that unexpected content
Author Manuscript

words produced more positive-going ERPs than control words (Word Type main effect: F
(1, 23) = 7.97, p = .010). In addition, Word Type showed a significant interaction with both
Laterality (F (2, 46) = 3.72, p = .047) and Ant-post (F (4, 92) = 8.11, p = .001). Follow-up
analyses showed that the positivity was significant in all three columns (left: F (1, 23) =
5.30, p = .030; midline: F (1, 23) = 7.85, p = .010; and right: F (1, 23) = 10.08, p = .004),
although numerically it was largest over the right hemisphere (M = 0.92, 1.21, and 1.31 on
the left, midline, and right, respectively). In addition, the positivity was significant at C sites
(F (1, 23) = 7.84, p = .010), P sites (F (1, 23) = 13.06, p = .001), and O sites (F (1, 23) =
16.36; p < .001), but was not significant at FP and F sites (ps > .10).

For function words, the omnibus ANOVA showed a significant interaction between Word
Type and Laterality (F (2, 46) = 7.36, p = .007). Follow-up analyses suggested that
unexpected function words had a trend to elicit more positive-going deflections at the left
Author Manuscript

column and more negative-going deflections at the right column than control words, but
none of the three lines (i.e., left, midline, and right) showed a significant Word Type effect
(ps > .10).

Foveal Targets 650 – 950 ms—For content words, the omnibus ANOVA revealed
significant main effects of Word Type (F (1, 23) = 9.14, p < .001), indicating that unexpected
content words elicited more positive-going ERPs than control words. There was also a
significant interaction between Word Type and Ant-post (F (2, 46) = 4.22, p = .021).
Follow-up analyses showed that the difference between unexpected vs. control words was
significant at C sites (F (1, 23) = 18.84, p < .001), P cites (F (1, 23) = 27.77, p < .001), and
O sites (F (1, 23) = 29.71, p < .001), marginally significant at F sites (F (1, 23) = 4.08, p =
.055), but was not significant at FP sites (F < 1).
Author Manuscript

For function words, although the main effect of Word Type was not significant (F (1, 23) =
2.12, p = .16), there was a significant interaction between Word Type and Laterality (F (2,
46) = 7.88, p = .007). The Word Type effect was only significant at the left column (F (1, 23)
= 5.12, p = .033), marginally significant at the midline (F (1, 23) = 3.57, p = .071), and was
not significant at the right column (F < 1). There was also a significant interaction between
Word Type and Ant-post (F (4, 92) = 14.32, p = < .001). Follow-up analyses showed that the
difference between unexpected and control function words was significant at P sites (F (1,
23) = 19.32, p < .001) and O sites (F (1, 23) = 11.26, p = .003), but was not significant FP, F,
and C sites (ps > .12).

Word Class Difference Wave Analyses—We further calculated difference between

Author Manuscript

waves of unexpected and control words in each critical time window and then analyzed
this difference as a function of Word Class (content vs. function)1, Laterality (left vs.
midline vs. right) and Ant-post (FP vs. F vs. C vs. P vs. O) as three independent variables.
These analyses showed that the distribution of the above word type effects was significantly
different between content and function words as follows. The analyses of parafoveal target

1Note that using difference waves computed in this manner removes any word category specific differences prior to analysis of
differences in the anomaly effects (i.e., word type effects) as a function of word class.

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 11

words between 350 and 550 ms showed a significant interaction between Word Class and
Author Manuscript

Laterality (F (2, 46) = 5.68, p = .012). Combining the findings in separate analyses for
content and function words, the negativity was right lateralized for unexpected content
words but left lateralized for unexpected function words. See Figure 7A for the spatial
distribution that illustrates these interactions.

The analyses of foveal targets between 350 and 550 ms showed a significant interaction
between Word Class and Laterality (F (2, 46) = 12.28, p < .001), a marginally significant
interaction Word Class and Ant-post (F (4, 92) = 3.36, p = .062), and a significant three-way
interaction (F (8, 184) = 2.61, p = .032). As shown above, while unexpected content words
elicited a right lateralized positivity at posterior sites, unexpected function words did not
elicit any significant effects. See Figure 7B for the spatial distribution that illustrates these
interactions.
Author Manuscript

The analyses of foveal targets between 650 and 950 ms showed a significant interaction
between Word Class and Laterality (F (2, 46) = 7.31, p = .006) and a significant three-way
interaction (F (8, 184) = 3.56, p = .007). While unexpected words in both classes showed
a positivity in this time window, the effect elicited by unexpected function words was more
left lateralized and posteriorly distributed. See Figure 7B for the spatial distribution that
illustrates these interactions.

Discussion
In the present study we used the RSVP with flankers paradigm combined with ERP
recording to compare parafoveal-foveal content versus function word processing during
sentence comprehension. The first notable finding was that we replicated previous studies,
whereby an unexpected semantically anomalous content word elicited a widely distributed
Author Manuscript

parafoveal N400 (e.g., Barber, et al., 2010, 2013; Li, et al., 2015; Payne & Federmeier,
2017; Payne, et al. 2019; Stites, et al., 2017). In addition, a broadly distributed positivity
but no N400 was shown when the unexpected content word was foveated. A novel finding
of the current study was that unexpected function words in the parafovea also elicited a
negativity around 400 ms. At first blush this finding would seem to suggest that unexpected
function words, like content words produce an N400 effect. However, careful examination
of the negativity produced by unexpected function words suggests this effect was more
left lateralized compared to the comparable effect seen for content words. Therefore, we
cautiously suggest this effect might better be classified as a parafoveal LAN. Interestingly,
function words did not elicit a parafoveal P600 effect but did show a clear P600 when they
were foveated. In summary, for both content and function words, different ERP components
were shown when the targets were presented in the parafovea versus in the fovea. Consistent
Author Manuscript

with the results of Payne et al (2019), these results indicate there are two stages that
differently engages attention in sentence comprehension, regardless of the part of speech of
target words.

Content Words
As mentioned above, the parafoveal N400 elicited by unexpected content words is consistent
with previous studies with similar manipulations. Because these words only moved to

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 12

the fovea after 400ms, this result strongly suggests that the effect cannot be due to
Author Manuscript

foveal processing. One potential caveat is that participants might have on some trials
made rightward saccades during the presentation of the critical word in the parafovea,
thus allowing foveal processing of the critical items. To guard against this possibility,
we carefully monitored horizontal eye movements and rejected all trials containing such
movements.

When target words were foveated, unexpected content words no longer elicited an N400
effect, consistent with previous findings that a parafoveal N400 mitigated or even eliminated
the following foveal N400 (Stites et al., 2017). Instead, a broadly distributed positivity was
seen starting around 300 ms after targets were foveated. It seems likely that this positivity
reflects the same process that the late positivity reported by Payne et al. (2019), namely, a
failure of sentence level integration for foveated unexpected content words. An alternative
possibility is that the foveal content word effect is a member of the semantic P600. Semantic
Author Manuscript

P600s had been observed in sentences with semantic anomalies with a plausible non-surface
interpretation, typically in role-reversed sentences (e.g., The mouse is chasing the cat)
or sentences with animacy violation (e.g., The hearty meal was devouring) in which the
error could be attributed to syntactic properties like word order or the characteristics of
the agent/patient of a verb (e.g., Chow & Phillips, 2013; Herten, Kolk, & Chwilla, 2005;
Kim & Osterhout, 2005; Kuperberg, 2007). Although none of our sentences involved role
reversals and only 13 out of 120 sentences with semantic anomalies involved animacy
violation, other factors may also elicit semantic P600s, such as the severity of the conflict
between expected words and the presented words. For example, van de Meerendonk,
Kolk, Vissers, and Chwilla (2010) found that deeply implausible sentences (e.g., The
eye consisting of among other things a pupil, iris, sticker…) elicited a larger P600 than
mildly implausible sentences (e.g., The eye consisting of among other things a pupil, iris,
Author Manuscript

eyebrow….). We suggested that the positivities shown in the present study were less likely
to be semantic P600 though, considering that the semantic P600 has been observed in both
active and passive comprehension tasks in previous research, while our effects, according
to Payne et al. (2019), were likely only present in tasks that required participants to make
an overt judgment (see Leckey & Federmeier, 2020 for review). Nevertheless, either a
result of plausibility-relevant integration failures or a semantic P600, our suggestion is
that the component elicited by unexpected content words engages foveal attention. Both
interpretations suggest difficulty in sentence-level re-analysis.

One interesting difference between the current foveal P600-like effects and the positivities
reported by Payne et al (2019) is that the current effects seem to have a somewhat earlier
time course. One possible explanation of this difference might be that Payne et al (2019)
Author Manuscript

used the same baseline for parafoveal and foveal targets, while in the present study we reset
the baseline for foveal targets. However, our foveal P600-like effects were already robust in
the foveal N400 time window. Although additional analyses showed that the effects were
no longer significant if we used the baseline for parafoveal targets (ps > .13), unexpected
content words still elicited numerically more positive waves in this time window. In contrast,
previous studies using the baseline for parafoveal targets showed negativities in the foveal
N400 time window (Payne et al., 2019; Stites et al., 2017). The opposite direction suggested
that different baseline at least was not the only reason. Importantly, this difference does

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 13

not affect our most critical comparison between content versus function words, given the
Author Manuscript

baselines of the two word classes were the same. Another factor that may contribute to this
different P600 time course might be the experimental procedure. In the present study, each
trial was presented for 400 ms with no blank screen between words. In Payne et al.’s (2019)
study each word was presented for 100 ms but was followed by a 350 ms blank screen. It
is possible that the 50 ms shorter SOA or the 300 ms longer presentation duration of each
parafoveal target in our study contributed to the earlier target P600 onset. Although more
research is needed to confirm this speculation, neither of the above possibilities affected our
critical claims. That is, our results showed evidence that initial semantic processing occurs
early in parafovea, while the later stage semantic processing at the higher order sentence
level engages foveal attention, overall consistent with previous findings.

Function Words
Author Manuscript

The novel manipulation in the present study was on function words. Unexpected function
words elicited a negativity between 350 and 550 ms when presented in the parafovea.
However, the distribution of this effect was greater over midline and left-hemisphere sites,
whereas the comparable content word N400 effect was larger over midline and right
hemisphere sites. Moreover, while the parafoveal N400 effect for content words had a focus
at central to posterior electrodes, the effect for function words showed a flatter distribution
along the anterior-posterior axis, an absence of Word Type × Ant-post interaction. Therefore,
the negativity was relatively more anterior on function than content words. Given the
difference in spatial distribution, the parafoveal negativity elicited by unexpected function
words seems more likely to related to previous reports of LAN effects for syntactic
violations. Although the LAN is typically distributed more anteriorly while our effect shows
a broad distribution across anterior and posterior sites, this could be because the LAN
Author Manuscript

and N400 are not categorically distinct ERP components; instead, the distribution reflects
a continuum about agreement errors on semantic and syntactic processing (Molinaro,
Barber, Caffarra, & Carreiras, 2015; also see Barber & Carreiras, 2005; Tanner & Van
Hell, 2014). Namely, the more semantic information processed, the more N400-like are
the effects; the more syntactic information processed, the more LAN-like are the effects.
Therefore, this different N400 versus relatively more LAN-like distribution elicited by
content versus function words did indicate more semantic process versus relatively more
syntactic processing in parafovea, respectively. In addition, the absence of foveal LAN
further suggested that initial syntactic processing of function words was completed before
targets are foveated, similar to the initial semantic processing of content words.

Interestingly we did not observe any P600-like activity when function words were in
peripheral vision. The presence of parafoveal LAN and absence of parafoveal P600
Author Manuscript

supported the claim that the LAN and the P600 are two functionally dissociable components
(Mancini, Molinaro, Rizzi, & Garreiras, 2011). While the LAN reflects earlier and more
automatic syntactic processing, the P600 seems more likely to reflect a higher-order/
sentence level structural repair or reanalysis process that occurs at a later stage (Friederici,
2002). This difference was also consistent with previous research that adopted the co-
registration approach of eye movements and fixation-related-potentials (FRPs), which
showed that regressions were strongly associated with the P600 effect in natural reading

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 14

(i.e., when the whole sentence was available to readers in each trial; Metzner, Von Der
Author Manuscript

Malsburg, Vasishth, & Rösler, 2017; also see Degno & Liversedge, 2020 for a review of this
co-registration approach). According to Metzner et al. (2017), regressions reflect readers’
attempt to find alternative interpretations in response to words that do not match built-up
expectations, thus eliciting P600 effects which reflect sentence level reanalysis. In addition,
P600 effects were triggered by both semantic and syntactic anomalies in Metzner et al.
(2017), although in that study both types of anomalies were elicited by unexpected content
words. This result was consistent with the present study, that both unexpected content and
function words in fovea elicited P600 effects, supporting the claim that P600 might not
be specific to syntactic processing, but reflect a general process of reanalysis. The present
study further suggested that the reanalysis process might involve direct/foveal attention to
the words that elicit anomalies.

The foveal P600 elicited by function words appeared later, was more left lateralized, and
Author Manuscript

more posteriorly distributed than the foveal P600-like effects elicited by content words.
These differences seemed consistent with early findings in Kutas and Hillyard (1983), which
directly compared content versus function words using the word-by-word RSVP paradigm
— in sentence comprehension, ERPs elicited by content words were overall more positive
than those elicited by function words from 200–700 ms over most scalp sites. The word
class effects on foveal targets probably reflect general different mechanisms of content
versus function words processing at the sentence level, although the exact mechanism(s) is
not clear. It is possible that the foveal P600 elicited by unexpected content words was more
about plausibility-relevant integration failures, while the foveal P600 elicited by unexpected
function words was more relevant to syntactic re-analysis at the sentence level. While future
research is needed to examine the exact mechanism(s), one factor that is unlikely to be
the cause of the different distribution and time course in the present study are differences
Author Manuscript

in expectancy, as our rating results showed that unexpected content and function words
were less acceptable than their controls to the similar extent before the context after the
critical words are presented. One conclusion that does seem warranted is that the processes
involved in both P600-like effects likely include foveal attention. In other tasks such as
passive reading or memory tasks, unexpected content words have not been shown to elicit
the foveal positive component, and while no RSVP passive reading studies have looked at
function word anomalies it seems reasonable that the function word P600 effect seen here
also resulted in part from the attentional demands of the task. Future research is needed to
verify this speculation.

Limitations and Future Research

Like most studies of complex cognitive/linguistic questions, the present study had several
Author Manuscript

limitations. First, while the RSVP with flankers paradigm is a clear improvement over the
word-by-word RSVP task in terms of approximating the stimulus conditions encountered
in natural reading, it is still possible that participants in this task use a less than natural
reading strategy that impacted the pattern of ERP anomaly effects reported. One possibility
is that despite instructions to fixate the highlighted word at fixation, participants nevertheless
consistently biased their attention to the right of the central word allowing the word to the
right to receive more in-depth processing than occurs in natural reading. This might make

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 15

the reported parafoveal effects reported here and in previous RSVP with flankers studies
Author Manuscript

more akin to typical foveal effects in RSVP studies. This might happen because of a longish
SOA (400 ms in our study) which results in a slightly slower than typical reading rate
thus encouraging rightward shifts of attention. However, because we carefully monitored
participants’ horizontal eye movements to guard against the possibility of saccades to and
from the rightward parafoveal words, one thing that we can be sure of is that any such
attentional bias occurred in the absence of accompanying eye movements. If participants
were consistently covertly attending to the word to the right of fixation, they were doing
so without moving their eyes. Importantly, even if participants in the RSVP with flankers
task consistently covertly attend to the parafoveal word, the ERP results obtained clearly
demonstrate that it is possible for words in the parafovea to influence both semantic and
syntactic processing, a process followed by a second step engages foveal attention.

Second, the results of the present study might be dependent on requirements to make
Author Manuscript

plausibility judgments to each sentence. As suggested in Payne et al. (2019), the foveal
P600 effect elicited by semantic anomalies was only present in their experiment requiring
plausibility judgments but was not present in passive comprehension. Therefore, some
effects reported here for function words might also be selectively present in the plausibility
judgment task.

Third, although we did not directly compare content versus function words, these two word
classes typically appear alternately in a sentence. As a result, parafoveal content words
might have, on average, been closer to foveal vision than parafoveal function words which
could account for some of the differences between word classes we are reporting.

Fourth, given the 400 ms presentation duration of each triad, what we are reporting as a
Author Manuscript

foveal P600 effects might alternatively be a somewhat delayed parafoveal P600 (see Risse &
Kliegl, 2012 for the similar argument in eye-tracking studies).

Lastly, in the present study we showed P600-like effects in both semantically and
syntactically anomalous sentences. While we argued that both effects reflect sentence-level
integration and re-analysis difficulties, there were differences in terms of time course
and scalp distribution suggesting that semantics and syntax may affect the family of late
positivities differently.

To address these limitations, future research with more diverse paradigms (e.g., FRPs,
passive comprehension), other languages (e.g., Chinese in which word length difference
is less salient across content versus function words, so that visual angle of flanker
words can be better controlled), and more factors controlled (e.g., orthographic similarity
Author Manuscript

between target words and unexpected words, sub-word class within content and function
words such as the comparison between nouns vs. verbs, preposition vs. conjunctions) are
needed to provide a more comprehensive picture. Future research should also include more
detailed manipulations of semantic and syntactic variables such as role-reversed sentences,
morphosyntactic violations and agreement violations to elicit semantic P600 versus syntactic
P600 via different components to shed light on the meaning of late positivities in sentence
comprehension. Despite the above limitations and the clear need for additional studies, we

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 16

would argue that our results and those of previous RSVP with flankers studies (e.g., Payne
Author Manuscript

& Federmeier, 2017; Payne, et al., 2019; Stites, et al., 2017) provide important additional
information beyond that gleaned from the traditional RSVP paradigm about the neural
mechanisms involved in semantic and syntactic processing during sentence comprehension.

Conclusion
In summary, the present study suggested that in sentence comprehension content versus
function words elicit more semantic versus syntactic processing, respectively. More
importantly, similar to content words, the processing of function words also appears to
include two stages that rely on attention to different extents. In sentence comprehension in
natural settings, readers are able to first perform an initial semantic and syntactic assessment
of the upcoming word when it is presented in the parafovea. At this stage, semantic and
syntactic processing is fast, perhaps automatic, and does not have to rely on direct attention.
Author Manuscript

In contrast, at the second stage readers seem to perform a more in depth attentional mediated
assessment. Both semantic and syntactic processing at this stage seems to be about sentence-
level integration or re-analysis, which engages foveal attention.

Acknowledgement
This research was supported by grants from the Eunice Kennedy Shriver National Institute of Child Health
and Human Development (HD099325, HD25889). Any opinions, findings, and conclusions or recommendations
expressed in this material are those of the authors and do not necessarily reflect the views of NIH.

Reference
Altarriba J, Kambe G, Pollatsek A, & Rayner K (2001). Semantic codes are not used in integrating
information across eye fixations in reading: Evidence from fluent Spanish-English bilinguals.
Author Manuscript

Perception & Psychophysics, 63(5), 875–890. [PubMed: 11521853]

Barber HA, Doñamayor N, Kutas M, & Münte T (2010). Parafoveal N400 effect during sentence
reading. Neuroscience letters, 479(2), 152–156. [PubMed: 20580772]
Barber HA, van der Meij M, & Kutas M (2013). An electrophysiological analysis of contextual and
temporal constraints on parafoveal word processing. Psychophysiology, 50(1), 48–59. [PubMed:
23153323]
Bornkessel-Schlesewsky I, & Schlesewsky M (2008). An alternative perspective on “semantic P600”
effects in language comprehension. Brain research reviews, 59(1), 55–73. [PubMed: 18617270]
Bradley DC, & Garrett MF (1983). Hemisphere differences in the recognition of closed and open class
words. Neuropsychologia, 21(2), 155–159. [PubMed: 6866257]
Briihl D, & Inhoff AW (1995). Integrating information across fixations during reading: The use of
orthographic bodies and of exterior letters. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 21(1), 55–67.
Brothers T, & Traxler MJ (2016). Anticipating syntax during reading: Evidence from the boundary
Author Manuscript

change paradigm. Journal of Experimental Psychology: Learning, Memory, and Cognition, 42,
1894–1906. [PubMed: 27123753]
Brown CM, Hagoort P, & Keurs MT (1999). Electrophysiological signatures of visual lexical
processing: Open-and closed-class words. Journal of cognitive neuroscience, 11(3), 261–281.
[PubMed: 10402255]
Brunelliere A, Hoen M, & Dominey PF (2005). ERP correlates of lexical analysis: N280 reflects
processing complexity rather than category or frequency effects. Neuroreport, 16(13), 1435–1438.
[PubMed: 16110266]

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 17

Chiarello C, & Nuding S (1987). Visual field effects for processing content and function words.
Neuropsychologia, 25(3), 539–548. [PubMed: 3683811]
Author Manuscript

Chow WY, & Phillips C (2013). No semantic illusions in the “Semantic P600” phenomenon: ERP
evidence from Mandarin Chinese. Brain research, 1506, 76–93. [PubMed: 23422676]
Coulson S, King JW, & Kutas M (1998). Expect the unexpected: Event-related brain response to
morphosyntactic violations. Language and cognitive processes, 13(1), 21–58.
DeLong KA, Quante L, & Kutas M (2014). Predictability, plausibility, and two late ERP positivities
during written sentence comprehension. Neuropsychologia, 61, 150–162. [PubMed: 24953958]
Donchin E (1981). Surprise!… surprise?. Psychophysiology, 18(5), 493–513. [PubMed: 7280146]
Donchin E, & Coles MG (1988). Is the P300 component a manifestation of context updating?
Behavioral and brain sciences, 11(3), 357–374.
Federmeier KD, Wlotko EW, De Ochoa-Dewald E, & Kutas M (2007). Multiple effects of sentential
constraint on word processing. Brain research, 1146, 75–84. [PubMed: 16901469]
Friederici AD (1985). Levels of processing and vocabulary types: Evidence from on-line
comprehension in normals and agrammatics. Cognition, 19(2), 133–166. [PubMed: 4017514]
Friederici AD, Meyer M, & Von Cramon DY (2000). Auditory language comprehension: an event-
Author Manuscript

related fMRI study on the processing of syntactic and lexical information. Brain and language,
74(2), 289–300. [PubMed: 10950920]
Friederici AD, Opitz B, & Von Cramon DY (2000). Segregating semantic and syntactic aspects of
processing in the human brain: an fMRI investigation of different word types. Cerebral cortex,
10(7), 698–705. [PubMed: 10906316]
Friederici AD, Pfeifer E, & Hahne A (1993). Event-related brain potentials during natural speech
processing: Effects of semantic, morphological and syntactic violations. Cognitive brain research,
1(3), 183–192. [PubMed: 8257874]
Gordon B, & Caramazza A (1982). Lexical decision for open-and closed-class words: Failure
to replicate differential frequency sensitivity. Brain and Language, 15(1), 143–160. [PubMed:
6184120]
Gordon PC, Plummer P, & Choi W (2013). See before you jump: Full recognition of parafoveal
words precedes skips during reading. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 39(2), 633–641. [PubMed: 22686842]
Author Manuscript

Grainger J, Midgley KJ, & Holcomb PJ (2016). Trans-saccadic repetition priming: ERPs reveal on-line
integration of information across words. Neuropsychologia, 80, 201–211. [PubMed: 26656872]
Hagoort P, Brown C, & Groothusen J (1993). The syntactic positive shift (SPS) as an ERP measure of
syntactic processing. Language and cognitive processes, 8(4), 439–483.
Van Herten M, Kolk HH, & Chwilla DJ (2005). An ERP study of P600 effects elicited by semantic
anomalies. Cognitive brain research, 22(2), 241–255. [PubMed: 15653297]
Hinojosa JA, Martin-Loeches M, Casado P, Munoz F, Carretie L, Fernandez-Frias C, & Pozo MA
(2001). Semantic processing of open-and closed-class words: an event-related potentials study.
Cognitive Brain Research, 11(3), 397–407. [PubMed: 11339989]
Inhoff AW (1989). Parafoveal processing of words and saccade computation during eye fixations in
reading. Journal of Experimental Psychology: Human Perception and Performance, 15(3), 544–
555. [PubMed: 2527961]
Inhoff AW (1989). Lexical access during eye fixations in reading: Are word access codes used to
integrate lexical information across interword fixations? Journal of Memory and Language, 28(4),
444–461.
Author Manuscript

Inhoff AW (1990). Integrating information across eye fixations in reading: The role of letter and word
units. Acta Psychologica, 73(3), 281–297. [PubMed: 2353591]
Inhoff AW, & Tousman S (1990). Lexical integration across saccades in reading. Psychological
Research, 52(4), 330–337. [PubMed: 2287696]
Kim A, & Osterhout L (2005). The independence of combinatory semantic processing: Evidence from
event-related potentials. Journal of memory and language, 52(2), 205–225.
Kutas M, & Hillyard SA (1980). Event-related brain potentials to semantically inappropriate and
surprisingly large words. Biological psychology, 11(2), 99–116. [PubMed: 7272388]

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 18

Kutas M, & Hillyard SA (1983). Event-related brain potentials to grammatical errors and semantic
anomalies. Memory & cognition, 11(5), 539–550. [PubMed: 6656613]
Author Manuscript

Kutas M, & Hillyard SA (1984). Brain potentials during reading reflect word expectancy and semantic
association. Nature, 307(5947), 161–163. [PubMed: 6690995]
Leckey M, & Federmeier KD (2020). The P3b and P600 (s): Positive contributions to language
comprehension. Psychophysiology, 57(7), e13351. [PubMed: 30802979]
Li N, Niefind F, Wang S, Sommer W, & Dimigen O (2015). Parafoveal processing in reading Chinese
sentences: Evidence from event-related brain potentials. Psychophysiology, 52(10), 1361–1374.
[PubMed: 26289548]
Meade G, Declerck M, Holcomb PJ, & Grainger J (2021). Parallel semantic processing in the flankers
task: Evidence from the N400. Brain and Language, 219, 104965. [PubMed: 33975227]
Mirault J, Yeaton J, Broqua F, Dufau S, Holcomb PJ, & Grainger J (2020). Parafoveal-on-foveal
repetition effects in sentence reading: A co-registered eye-tracking and electroencephalogram
study. Psychophysiology, 57(8), e13553. [PubMed: 32091627]
Münte TF, Wieringa BM, Weyerts H, Szentkuti A, Matzke M, & Johannes S (2001). Differences in
brain potentials to open and closed class words: Class and frequency effects. Neuropsychologia,
Author Manuscript

39(1), 91–102. [PubMed: 11115658]

Neville HJ, Mills DL, & Lawson DS (1992). Fractionating language: Different neural subsystems with
different sensitive periods. Cerebral cortex, 2(3), 244–258. [PubMed: 1511223]
Nobre AC, Price CJ, Turner R, & Friston K (1997). Selective processing of nouns and function words
in the human brain. Neuroimage, 5 (4):53.
Nobre AC, & McCarthy G (1994). Language-related ERPs: Scalp distributions and modulation by
word type and semantic priming. Journal of cognitive neuroscience, 6(3), 233–255. [PubMed:
23964974]
Osterhout L, Bersick M, & McKinnon R (1997). Brain potentials elicited by words: word length
and frequency predict the latency of an early negativity. Biological Psychology, 46(2), 143–168.
[PubMed: 9288411]
Osterhout L, & Holcomb PJ (1992). Event-related brain potentials elicited by syntactic anomaly.
Journal of memory and language, 31(6), 785–806.
Osterhout L, Holcomb PJ, & Swinney DA (1994). Brain potentials elicited by garden-path sentences:
Author Manuscript

Evidence of the application of verb information during parsing. Journal of experimental

psychology: Learning, memory, and cognition, 20(4), 786–803. [PubMed: 8064247]
Payne BR, & Federmeier KD (2017). Pace yourself: Intraindividual variability in context use revealed
by self-paced event-related brain potentials. Journal of Cognitive Neuroscience, 29(5), 837–854.
[PubMed: 28129064]
Payne BR, Stites MC, & Federmeier KD (2019). Event-related brain potentials reveal how multiple
aspects of semantic processing unfold across parafoveal and foveal vision during sentence reading.
Psychophysiology, 56(10), e13432. [PubMed: 31274200]
Pollatsek A, Lesch M, Morris RK, & Rayner K (1992). Phonological codes are used in integrating
information across saccades in word identification and reading. Journal of Experimental
Psychology: Human perception and performance, 18(1), 148–162. [PubMed: 1532185]
Rayner K (1998). Eye movements in reading and information processing: 20 years of research.
Psychological bulletin, 124(3), 372–422. [PubMed: 9849112]
Rayner K, Balota DA, & Pollatsek A (1986). Against parafoveal semantic preprocessing during eye
fixations in reading. Canadian Journal of Psychology/Revue canadienne de psychologie, 40(4),
Author Manuscript

473–483.
Rayner K, & Morris RK (1992). Eye movement control in reading: Evidence against semantic
preprocessing. Journal of Experimental Psychology: Human Perception and Performance, 18(1),
163–172. [PubMed: 1532186]
Risse S, & Kliegl R (2012). Evidence for delayed parafoveal-on-foveal effects from word n+2 in
reading. Journal of Experimental Psychology: Human Perception and Performance, 38(4), 1026–
1042. [PubMed: 22428669]

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 19

Schotter ER, & Jia A (2016). Semantic and plausibility preview benefit effects in English: Evidence
from eye movements. Journal of Experimental Psychology: Learning, Memory, and Cognition, 42,
Author Manuscript

1839–1866. [PubMed: 27123754]

Segui J, Mehler J, Frauenfelder U, & Morton J (1982). The word frequency effect and lexical access.
Neuropsychologia, 20(6), 615–627. [PubMed: 7162585]
Snell J, Meade G, Meeter M, Holcomb P, & Grainger J (2019). An electrophysiological investigation
of orthographic spatial integration in reading. Neuropsychologia, 129, 276–283. [PubMed:
31002854]
Stites MC, Payne BR, & Federmeier KD (2017). Getting ahead of yourself: Parafoveal word
expectancy modulates the N400 during sentence reading. Cognitive, Affective, & Behavioral
Neuroscience, 17(3), 475–490.
Ter Keurs M, Brown CM, Hagoort P, & Stegeman DF (1999). Electrophysiological manifestations of
open-and closed-class words in patients with Broca’s aphasia with agrammatic comprehension: An
event-related brain potential study. Brain, 122(5), 839–854. [PubMed: 10355670]
Van De Meerendonk N, Kolk HH, Vissers CTW, & Chwilla DJ (2010). Monitoring in language
perception: Mild and strong conflicts elicit different ERP patterns. Journal of cognitive
Author Manuscript

neuroscience, 22(1), 67–82. [PubMed: 19199401]

Van Petten C, & Kutas M (1991). Influences of semantic and syntactic context on open-and closed-
class words. Memory & Cognition, 19(1), 95–112. [PubMed: 2017035]
Van Petten C, & Luka BJ (2012). Prediction during language comprehension: Benefits, costs, and ERP
components. International Journal of Psychophysiology, 83(2), 176–190. [PubMed: 22019481]
Veldre A, & Andrews S (2016). Is semantic preview benefit due to relatedness or plausibility?
Journal of Experimental Psychology: Human Perception and Performance, 42, 939–952. [PubMed:
26752734]
Veldre A, & Andrews S (2018). Beyond cloze probability: Parafoveal processing of semantic and
syntactic information during reading. Journal of Memory and Language, 100, 1–17.
Author Manuscript
Author Manuscript

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 20
Author Manuscript
Author Manuscript

Figure 1.
The visual hemi-field RSVP-flanker paradigm. The foveal target of each triad is presented
centered in white.
Author Manuscript
Author Manuscript

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 21
Author Manuscript
Author Manuscript
Author Manuscript

Figure 2.
Electrode montage and the 15 analysis sites used for ANOVAs. The Ant-post factor is
defined by the five sites (FP, F, C, P, O) in each of the three Laterality columns, which are
indicated with connecting lines. The four additional electrodes were A1 (the reference site
over the left mastoid bone), A2 (the other reference cite to monitor for differential mastoid
activity); LE (the eye electrode to monitor for blinks), and HE (the eye electrode to monitor
for horizontal eye movements).
Author Manuscript

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 22
Author Manuscript
Author Manuscript

Figure 3.
ERPs time locked to parafoveal target onset from the 15 sites that were included in analyses,
comparing unexpected vs. control content targets. The black line is from the control content
targets, while the red line is from the unexpected content targets. We also plotted the ERPs
from the channels that monitor horizontal eye movements (HE). While 0 ms refers to the
parafoveal target onset, the targets moved to the foveal position at 400 ms.
Author Manuscript
Author Manuscript

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 23
Author Manuscript
Author Manuscript

Figure 4.
ERPs time locked to parafoveal target onset from the 15 sites that were included in analyses,
comparing unexpected vs. control function targets. The black line is from the control
function targets, while the red line is from the unexpected function targets. We also plotted
the ERPs from the channels that monitor horizontal eye movements (HE). While 0 ms refers
to the parafoveal target onset, the targets moved to the foveal position at 400 ms.
Author Manuscript
Author Manuscript

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 24
Author Manuscript
Author Manuscript

Figure 5.
ERPs time locked to foveal target onset from the 15 sites that were included in analyses,
comparing unexpected vs. control content targets. The black line is from the control content
targets, while the red line is from the unexpected content targets. We also plotted the ERPs
from the channels that monitor horizontal eye movements (HE). While 0 ms refers to the
foveal target onset, the targets moved to the left position (i.e., being the preceding word) at
400 ms.
Author Manuscript
Author Manuscript

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 25
Author Manuscript
Author Manuscript

Figure 6.
ERPs time locked to foveal target onset from the 15 sites that were included in analyses,
comparing unexpected vs. control function targets. The black line is from the control
function targets, while the red line is from the unexpected function targets. We also plotted
the ERPs from the channels that monitor horizontal eye movements (HE). While 0 ms refers
to the foveal target onset, the targets moved to the left position (i.e., being the preceding
word) at 400 ms.
Author Manuscript
Author Manuscript

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 26
Author Manuscript
Author Manuscript
Author Manuscript

Figure 7.
Voltage maps calculated from the mean difference (in microvolts) between unexpected
words and control words at 29 scalp sites at the 350 – 550 ms (left column) and 650 –
950 ms (right column) time windows. Panel A shows the maps for parafoveal targets, while
Panel B shows the maps for foveal targets.
Author Manuscript

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.
 Li et al. Page 27

Table 1.

Mean word frequency, length, and acceptability judgment scores (standard deviation in the parentheses) across
Author Manuscript

conditions.

a b
Frequency Length Acceptability
Unexpected 84.34 (201.34) 5.46 (1.49) 1.33 (.24)
Content Words Control 108.37 (216.22) 5.67 (1.83) 6.74 (.17)
Difference p = 0.37 p = 0.29 p <.001
Unexpected 7347.07 (8504.04) 3.07 (1.16) 1.69 (.42)
Function Words Control 11752.85 (9397.60) 2.84 (1.06) 6.56 (.55)
Difference p < .001 p = 0.13 p <.001

a
The frequency refers to the number per million based on SUBTLEX_US (Brysbaert & New, 2009).
b
The difference score between unexpected and control words for the content vs. function words were also compared (p= .16), indicating that the
Author Manuscript

unexpected content vs. function words were equally unexpected compared to their corresponding control words.
Author Manuscript
Author Manuscript

Lang Cogn Neurosci. Author manuscript; available in PMC 2024 January 01.