CRL Newsletter
November 1995
Vol. 10, No. 2
The newsletter of the Center for Research in Language, University of California,
San Diego, La Jolla CA 92039. 858-534-2536; email: editor@crl.ucsd.edu.
Table Of Contents
A Brain Potential Whose Latency Indexes
the Length and Frequency of Words
Jonathan W. King* and Marta Kutas*#
*Cognitive Science, UCSD; #Neurosciences, UCSD
ABSTRACT
ERPs were recorded from 24 undergraduates in an investigation
of the effects of length, frequency, and grammatical class on the brain's
response to words during sentence reading. Our results indicate that the
combined length and frequency of a word are indexed by the latency of a
negative peak maximal over left anterior regions of the scalp that we call
the Lexical Processing Negativity or LPN; the form of this relationship
mirrors that between these same lexical factors and reaction times. The
length-frequency effect on the LPN accounts for some of the known electrophysiological
differences between open class (content) and closed class (function) words.
The LPN also helps bridge a perceived gap between reaction time and electrophysiological
data as measures of cognitive processes in visual word recognition.
FEATURE ARTICLE
While the question of how language processing is subserved by the brain
has been asked concerning many levels of linguistic analysis, the bulk of
the research has dealt with the processing of single words. In particular,
the focus has been on when and where in the brain lexical access[1]occurs, and whether
these processes differ for words of different syntactic categories. With
few exceptions, studies of lexical access have relied on manual or verbal
reaction times (RTs) or eye movement gaze durations as the dependent variable.
In reading studies, these have revealed that encoding and lexical access
times are in large part a function of a word's length in letters and its
frequency of daily usage; longer words take more time to process than shorter
words, and rare words more time than common words (see, e.g., Just & Carpenter,
1980). More specifically, regression analyses suggest that each character
adds about 30 msec to the expected gaze duration on a word, while each (common)
log unit of frequency subtracts about the same amount (Thibadeau, Just & Carpenter,
1983). Indeed, extremely short, common words like "the" and "of"
are often not directly fixated at all. Most unfixated items tend to be so-called
"function words" (e.g., articles, conjunctions, prepositions,
auxiliaries) which are generally shorter, more frequent, and more predictable
from context than "content words".
Function and content words (e.g., verbs, nouns, adjectives and some adverbs)
also differ in that while new content words can be freely coined, the set
of function words is essentially "closed" to new members. Thus,
content words are often called "Open Class" words while function
words are called "Closed Class" words. It has been suggested that
closed class items, which carry information primarily about syntactic structure,
and open class words, which carry primarily semantic information, may be
processed by functionally and anatomically distinct brain areas (e.g., Swinney,
Zurif, & Cutler, 1980). In support of this hypothesis, some researchers
have reported that frequency effects are not the same for normal subjects
and Broca's aphasics, who typically show disproportionately greater difficulties
both in producing and comprehending closed than open class items. Specifically,
normal subjects show frequency effects in their RTs to open class words
only, whereas Broca's aphasics show frequency effects to both open and closed
class words (Bradley, Garrett, & Zurif 1980). However, this pattern of results
for normal subjects has not been replicated consistently, and it has been
suggested that technical problems in estimating frequency effects on very
rapid RTs may render this issue unamenable to resolution using behavioral
data alone (Gordon & Caramazza, 1985).
In contrast to the robust findings in the reaction time literature, differences
in the event-related brain potentials (ERPs) to words due to word length
and frequency have been subtle, while differences between open and closed
class items have been quite large (e. g. Van Petten and Kutas, 1991; Neville,
Mills and Lawson 1992). The largest ERP difference appears to be in the
amplitude of the N400 responses. Virtually all open class words in sentence
contexts generate an N400 whose amplitude depends on a variety of factors
including cloze probability (Kutas & Hillyard, 1984), repetition (Van Petten
et al., 1991), and frequency to a limited extent (i.e. only for the first
few open class words in sentences; Van Petten & Kutas, 1991). By contrast,
closed class words typically do not generate large, if any, N400s (but see
King & Kutas, 1995b, for a notable exception and Kluender & Kutas 1993 for
discussion). Recently, Neville, Mills, and Lawson (1992) have claimed that
the ERPs to closed class words differ from those to open class words not
only by the absence of an N400 but also by the presence a negativity at
the left anterior scalp around 280 msec (i.e. the N280). They concluded
that the N280 was a qualitative sign of a word class effect because it was
not elicited by open class words regardless of their length or frequency.
These facts combined with its suggestive scalp localization over left anterior
sites has fueled speculation that the N280 is generated in or near the classical
Broca's area and has been offered as support for "...the activation
of different neural systems that are organized to process the different
kinds of linguistic information that these word classes provide" (Neville
et al., 1992).[2]
Perhaps in part because of temporal overlap with the N400, much less attention
has been given to the presence of a similar, albeit slightly later, negative
peak at left anterior sites for open class words as well (i.e. N410) (e.g.,
Neville, Kutas, Chesney and Schmidt, 1986). [3]
The possibility that the left anterior negativities generated by open class
and closed class words are identical has not yet been investigated directly,
although it is reasonable to hypothesize that they both index a process
whose dynamic timing is affected by lexical class or correlated lexical
factors such as length and frequency.[4]
If the difference in latency is strictly categorical, and related to fundamental
differences between closed and open class items, then the latencies of members
within the two classes might be fixed, unrelated to lexical factors, and
not comparable between classes. On the other hand, if both the N280 and
the N410 partially index processes affected by lexical factors, then the
duration of these processes might be reflected in the latency of this left
anterior negativity in a systematic fashion. Specifically, its latency,
both within and between lexical classes, may be a function of lexical factors
such as length and frequency which have large down-stream effects on reaction
times.[5]
Method
Subjects and Materials
Twenty-four right-handed, normal, native English monolingual UCSD students
(12 women) between 18 and 27 years of age participated in the study after
giving their informed consent in compliance with university procedures,
and received $5.00 an hour for their time. The materials included 288 sentences,
of which 72 were the critical materials for another study (King & Kutas,
1995b) and the other 216 were filler sentences of various syntactic structures.
The results reported here are based on the ERPs to the single words of the
filler sentences. These words were sorted into 10 broad classes based on
their syntactic features. Open Class types included Nouns, Verbs, and Adjectives;
Closed Class types were Infinitival "To", Definite Articles ("the"),
Indefinite Articles ("a" or "an"), Noun Phrase Prepositions
(e.g. "of"), Verb Phrase Prepositions (e.g. "for"),
Conjunctions, and forms of the verb To Be (e.g. "was", "is").
For each class, the mean length in letters for the members was computed,
as was mean "scarcity". Scarcity was calculated as the common
log of each word's frequency in the Francis and Kucera corpus lexicon (1982)
subtracted from 6 (the highest possible log frequency in a 1 million word
corpus). This transformation yields a variable, like word length, that is
positively (rather than negatively) correlated with reaction time on most
tasks.
Procedure
Subjects read sentences presented one word at a time in the center of a
CRT while their electroencephalogram (EEG) was recorded. Words were presented
for a duration of 200 msec with a 500 msec word onset asynchrony. Subjects
were instructed to read each sentence knowing that a True/False comprehension
probe would follow approximately half the sentences. ERPs were recorded
from 6 pairs of lateral electrodes on an Electro-Cap and from electrodes
placed on the left and right mastoids. All electrodes were referenced to
a noncephalic lead derived from an electrode placed at the sterno-clavicular
junction and over the seventh cervical vertebra both fed through a potentiometer
adjusted to reduce cardiac artifact. The electrodes covered both standard
10-20 sites (F7, F8, T5, T6, O1, O2), one pair approximately over left and
right primary auditory cortex, and one electrode each over Broca's area,
Wernicke's Area, and the two locations over the right hemisphere analogous
to these language areas. The electrode site that is the primary focus of
this paper is F7, which lies over the lateral aspect of the left anterior
scalp, and will be henceforth referred to as Left Frontal. Subjects' EEG
was digitized on-line with a sampling rate of 250 Hz with a time constant
of ~8 seconds; eye movement and blink artifacts were rejected off-line prior
to averaging. Further details of the experimental procedure are given in
King and Kutas (1995b).
Results
Figure 1 contrasts the grand average ERP waveforms for Open and Closed Class
words that are the focus of this study.[6]
There are clearly several differences between the ERPs to open and closed
class words. For example, the ERPs to open class words are characterized
by greater negativity between 250 and 500 msec which is larger over right
hemisphere sites (the N400), and greater negativity for closed class relative
to open class words over left anterior sites. Specifically, the left anterior
negativity seen for closed class words includes an early peak with an approximate
latency of 280 msec (N280), and a later, broader negativity (N400-700) that
overlaps the ERP to the following word (Neville et al. 1992). We should
briefly note the N400-700 does seem to be diagnostic of closed class words,
but only in sentence contexts; Nobre and McCarthy (1994) detected no such
wave in their studies using unstructured word lists. King and Kutas (1995b)
and Van Petten and Kutas (1991) have offered explanations of the N400-700
that emphasize the role of closed class words in heading up syntactic constituents
in sentential contexts. For open class words, the ERP also includes a negative
peak at approximately 315 msec at the left frontal electrode site F7, which
is clearly different from the N400 and appears to be identical to the N410
(Neville et al., 1986) and the N330 (Nobre and McCarthy 1994).
Figure 1. Grand average ERPs (n=24) from 14 electrode sites
for the Open Class (solid line) and Closed Class words considered in this
study. The left hemisphere is plotted on the left, and negative voltages
are plotted up.
Our hypothesis that the N280 and the N315 might reflect a common processing
stage predicts that they should vary in latency with lexical factors; Figure
2 shows that the latencies of the N280 and our N315 do seem to vary in the
ERPs to the different word types, being earliest (~270 msec) for the shortest
and most frequent closed class words such as definite articles, later (~300
msec) for longer and less frequent closed class items such as prepositions
commonly used in adverbial phrases, and later still (~ 315 msec) for longer
and less frequent open class items such as nouns. Note that some of the
other differences between word classes apparent in Figure 2 are statistically
reliable, most notably the greater positivity (P2) preceding the N315 elicited
by nouns and adjectives; these and other lexical class differences will
not be discussed further in this paper.
Figure 2. Grand average ERPs (n=24) at the Left Frontal (F7)
electrode site for representative word types that are subclasses of the
broad Open vs. Closed Class data. Dashed line is at 280 msec; asterisks
mark peak latencies for the word types.
The real test of our hypothesis, however, is to determine whether the mean
latencies of the left frontal negativity to words from a greater variety
of Open and Closed Class items can be predicted from their respective lengths
and frequencies. To assess this possibility, the peak latency of this negativity
for words in each of the ten classes listed in the Methods section was regressed
onto the sum of its respective mean length and scarcity.[7]
For the mean peak latency data, this resulted in a highly reliable regression
(p .001) accounting for 86% of the variance. The left panel of Figure 3
shows both the grand mean latency observations and the best fitting line,
whose equation is:
P = 256 + 4.8*(L+S)
where P is peak latency in milliseconds, L is length in characters and S
is scarcity.[8]
By way of comparison, Thibadeau, Just and Carpenter (1983) found that length
and frequency together accounted for 76% of the variance in gaze duration
for (individual) fixated words, collapsed over subjects. It should also
be noted that our regression equation is sensible in that it cannot predict
a peak latency for this negative wave that would precede the P2 component,
due to the location of the intercept. At the same time, this equation does
predict latency increases with increasing length or scarcity.[9]
Figure 3. Panel A shows the regression of the Lexical Processing
Negativity mean peak latency (in msec.) onto the Length+Scarcity predictor
(solid line). Points indicate observations from the 10 lexical types used
in the regression, with ALL CAPS used for category labels, oblique lower
case used for prototypical category exemplars, and roman lower case used
for definitive category exemplars. Panel B shows the superimposed regression
lines for all 24 subjects to demonstrate the variability of fits to individual
subject data.
The same simple model also fits data from individual subjects, as demonstrated
by the clear similarities in the best-fitting lines shown in the right panel
of Figure 3. Statistically, 16 of the 24 individual regressions were significant
at the .05 level, 4 others were marginal (p .15), and the others were insignificant
(p > 0.2). Over all 24 subjects, the median proportion of variance accounted
for was 44%; this increases to 52% when only subjects with reliable individual
regressions are considered.
Discussion
Previously we have argued that the ERPs to all words, closed and open class
alike, include an N400, so its presence cannot be used as a definitive marker
of lexical class. Similarly, we now suggest that the ERPs to all words,
closed and open class alike, also include a negativity over left anterior
sites whose latency varies with the eliciting word's lexical characteristics,
i.e. its length and frequency. Accordingly, we propose that this ERP effect
would be more aptly called the Lexical Processing Negativity (LPN), and
that its latency be taken as an electrophysiological measure of the effect
of lexical factors on word processing. Like the N400, the LPN (aka N280)
cannot serve as a qualitative classifier of closed versus open classedness.
One interesting feature of the LPN is that the size of its variation in
peak latency (less than 5 msec per log unit of frequency or word length)
appears to be much smaller than would be expected for either gaze duration
or button-pressing reaction times with the same stimuli. Rather than being
a matter for concern, however, this finding can be taken to reveal how and
where frequency of a word's daily usage influences word recognition. Reaction
time data necessarily reflect the cumulative effects of processing at all
stages, whereas ERP latency data can reflect neural activity taking place
at particular earlier stages of analysis. Thus, the fact that frequency
effects are larger at the final output than at this intermediate stage indicates
that frequency impacts processing at multiple points during word recognition
and that these effects are cumulative. McRae, Jared, and Seidenberg (1990)
provide evidence from word naming in favor of this position, and suggest
that such effects could best be explained using models of distributed lexical
access processes throughout a broader parallel processing network (such
as Seidenberg & McClelland's [1989] word naming model). Other evidence in
favor of the accumulation of frequency effects throughout stimulus analysis
comes from the ERP study of Polich and Donchin (1988), who noted that differences
of ~2 log units of word frequency led to a (mean) 20 msec shift in the peak
latency of the P3 in a lexical decision task, but had an appreciably larger
(110 msec) effect on the associated lexical decision times.
In any case, establishing the sensitivity of the LPN to word length and
frequency, while essential in understanding its functional role during reading,
does not, however, reveal what that role is. To date the only hypothesis
concerning the functional role of the LPN was based on the assumption that
it was a fixed latency N280 elicited only by closed class words and localized
over Broca's area (Neville et al. 1992). On this view the N280 is presumed
to index "...the activation of processes important in the look up and/or
identification of words in a system that only includes representations of
closed class words" and perhaps "... processes concerned with
parsing sentence structure." (page 251). Our results showing that the
N280 is neither specific to closed class words nor fixed in latency suggest
a re-consideration of this proposal, although the LPN may, nevertheless,
reflect activity linked to a stage of syntactic processing that is sensitive
to frequency effects of some kind.[10]
Another possibility is that the LPN reflects processing more directly tied
to reading, such as the control of gaze or the planning and generation of
saccadic eye movements. For the remainder of this report, we outline this
alternative hypothesis whose eventual validity should in no way detract
from our primary finding of a continuum in LPN latency between open and
closed class items.
According to a gaze control hypothesis, the LPN should be sensitive to the
same factors that control gaze duration in reading of which length and frequency
are two of the more important (Just and Carpenter, 1980). At first glance,
the 270-plus millisecond latency of the LPN may seem too long to reflect
eye movement-related processes, but it is important to remember that peak
latency is a conservative measure of the timing of mental operations and
merely serves as an upper limit of the time by which some processing must
have occurred. It is also possible that the LPN could be reflecting the
activity of an inhibitory process whose timing would naturally place it
close to that of eye movements.
The role of such inhibitory processes may be highlighted in ERP reading
studies due to the requirement that subjects maintain fixation in the center
of the screen. This requirement prevents readers not only from making long
saccades from word to word, but also from moving their eyes to the "preferred"
fixation point within a presented word (i.e., slightly to the left of center;
McConkie & Rayner, 1975). Thus, the LPN might reflect some aspect of withholding
or controlling eye movements, consistent with a contribution from the frontal
eye fields (FEFs) in coordination with nearby anterior language areas. Data
from both cortical stimulation (e.g., Luders et al., 1992) and positron
emission tomography (PET) localization studies (Fox, et al., 1985) indicate
that the brain regions that play a vital role in the generation of volitional
eye movements (including the FEFs) in humans are located anterior to the
motor strip and dorsal to the classical Broca's area. While direct electrical
stimulation of sites in this region can elicit (contralateral) eye movements,
stimulation of other nearby areas can also lead to the cessation of ongoing
volitional eye movements, or even the cessation of both eye movements
and speech (Luders et al., 1992). Intriguingly, ERP studies requiring button
presses have shown that during trials on which usually appropriate responses
had to be inhibited (i.e. "no-go" trials), one observes a negative
peak similar the LPN at electrode sites over nearby premotor cortex (Sasaki
et al., 1993).[11]
The left laterality of the LPN may reflect either the specific link with
language processing or the prevalence of rightward saccades in reading English.
One way to adjudicate between these alternative positions would be to investigate
the ERPs of subjects reading a language like Hebrew, where saccades are
primarily leftward and see if the LPN is altered in its laterality.[12] Another way
to explore the laterality of the LPN would be to look at the ERPs to words
and sentences in American Sign Language (ASL), given that ASL speakers must
suppress saccadic eye movements towards salient, peripheral stimuli moving
in either direction.
Accounts that tie the LPN directly to reading processes also provide an
alternative explanation of the reported effects on the N280 in individuals
for whom reading is a less well-practiced skill. For example, Neville, Coffey,
Holcomb and Tallal's (1993) data showed longer latency or less well-articulated
LPNs in individuals who were less skilled in reading, whether the comparison
was between children and adults, or between children with specific language
impairments and those without. Similarly, second language learners such
as native ASL speakers who have poorer reading skills in English also were
found to show abnormal N280 effects when reading English text (Neville et
al. 1992). At the other end of the developmental spectrum, preliminary data
in our lab also indicate that the morphology and latency of the LPN change
in normal aging, consistent with proposals that older adults suffer declines
in inhibitory motor processing (King & Kutas, 1995a).
Whether or not this gaze control hypothesis is eventually confirmed, we
believe that the study of the LPN may help us better understand reading
and supplement eye movement measures, which now provide most of the compelling
data on the nature of online sentence processing. Also, because the LPN
can be observed in single word studies (e.g. Nobre & McCarthy, 1994) and
appears to be sensitive to length and frequency effects occurring before
reaction time, its measure may prove useful in studies of priming and the
organization of semantic and lexical memory, in a manner complementary to
the more domain-general P300. Last but not least, the fact that LPN can
be elicited in paradigms where no overt behavior is required, makes it a
more readily accessible measure of a word's processing in members of populations
where reaction time data may be difficult to collect or to interpret.
ACKNOWLEDGEMENTS
Work reported in this paper was supported by NICHD grant HD22614 and by
NIA grant AG08313 to Marta Kutas. Jonathan W. King was also supported in
part by a McDonnell-Pew postdoctoral fellowship from the San Diego Center
for Cognitive Neuroscience, funds from the Center for Human Information
Processing (CHIP; NIMH grant MH-14268), and funds from the Center for Research
on Language (CRL; NIH grant T32 DC00041-01) at UCSD. We thank Elizabeth
Bates, Seana Coulson, Adele Goldberg, and Jill Weckerly for helpful comments
on previous versions of this manuscript.
NOTES
[1] We use
the term lexical access to refer to the execution of the process(es) that
brings a word's semantic meaning, syntactic category, and possibly other
information into working memory from long term memory representations during
a language comprehension task.
[2] Statements
such as this one presuppose a strong association between Broca's area and
receptive grammatical processing that is based on the classical English
language aphasia literature. More recent and cross-linguistic research sheds
doubt on the strength of this association (see, e. g., Bates, Wulfeck, and
MacWhinney, 1991).
[3] The original
experiment defining the N410 (Neville et al., 1986) used text presented
vertically in either the left or right visual field; with normal text, the
latency of the N410 generally decreases to, for instance, 330 msec in the
work of Nobre and McCarthy (1994).
[4] To date,
differences in the ERPs elicited by words which are due to frequency, length,
and lexical class have been observed primarily as modulations in amplitude
(or presence) of certain components rather than in their latency. This relative
lack of latency effects is worth noting as one might expect that ERP data
would be especially useful in providing information about the pre-response
chronometry of cognitive events.
[5] Individual
ERP peaks are undoubtedly more complex than we have suggested since they
often reflect the activity of multiple neural generators operating simultaneously.
Thus, shifts in the latency of a particular peak might still be taken as
evidence that some of generators active in response to a stimulus are altered
in their timing.
[6] Note that
the ERPs in this figure are grand means, and that the peak latency of a
component in a grand mean is not necessarily a simple function of the mean
peak latencies for individual subjects.
[7] The mean
length and frequency variables for our 10 classes were so highly correlated
(r(10) = .93) that we could not expect to obtain stable estimates of the
parameters in such a regression. Thus, a single predictor was calculated
from the sum of these two variables.
[8] Similarly
significant regressions were performed for Length alone (P = 258 + 7.5*L,
p .001) and Scarcity alone (P = 255 + 11.8*S, p .001. Our choice of the
combined variable was motivated primarily from the results of reading studies
showing the effect of both variables.
[9] This increase
is also found among just closed class items, where the regression equation
is P = 261 + 3.6*(L+S). For scarcity alone, the regression is virtually
identical in form to that for the whole data set (P = 255 + 12.0*S). In
both cases, however, the concentration on closed class items has reduced
our degrees of freedom, and probably seriously affected our p-values (which
are p = .23 and p = .15, respectively). Length alone barely varied between
closed class groups in this materials set.
[10] Such
effects are not without precedent. Most recently, MacDonald (1994) has demonstrated
that differences in frequency underlying verb argument structure preferences
influence the resolution of syntactic ambiguity. Thus, in principle, the
LPN could be related to the extraction or use of syntactic category information
(e.g. its part of speech) subsequent to one or several such processes sensitive
to the word's lexical properties.
[11] Note
that so far there is no conclusive evidence that the LPN/N280 is actually
generated in or near any specific cortical area, and the fact that it is
larger over certain anterior sites cannot be taken as firm localization
information. Our argument is that the component peak is no less consistent
with a generator in the FEF or premotor cortex than in anterior language
areas, as suggested by Neville et al. (1992).
[12] While
readers of Hebrew show the standard RVF advantage for laterally presented
words (e.g., Faust, Kravetz & Babkoff, 1993), they also show a left-of-fixation
enlargement of the window of visual attention (Pollatsek, Bolozky, Well,
& Rayner, 1981), which is compatible with the scanning hypothesis.
REFERENCES
Bates, E., Wulfeck, B., & MacWhinney, B. (1991). Cross-linguistic research
in aphasia: an overview. Brain and Language, 41, 123-148. Bradley, D. C.,
Garrett, M. F., & Zurif, E. B. (1980). Syntactic deficits in Broca's aphasia.
In D. Caplan (Ed.), Biological studies of mental processes (pp. 269-286).
Cambridge: MIT Press.
Faust, M., Kravetz, S., & Babkoff, H. (1993). Hemispheric specialization
or reading habits: Evidence from lexical decision research with Hebrew words
and sentences. Brain & Language, 44(3), 254-263.
Fox, P. T., Fox, J. M., Raichle, M. E., & Burde, R. M. (1985). The role
of cerebral cortex in the generation of voluntary saccades: a positron emission
tomographic study. Journal of Neurophysiology, 54(2), 348-69.
Francis, W. N., & Kucera, H. (1982). Frequency analysis of English usage:
lexicon and grammar. Boston: Houghton Mifflin.
Gordon, B., & Caramazza, A. (1985). Lexical access and frequency sensitivity:
Frequency saturation and open/closed class equivalence. Cognition, 21(2),
95-115.
Just, M. A., & Carpenter, P. A. (1980). A theory of reading: From eye fixations
to comprehension. Psychological Review, 87, 329-354.
King, J. W., & Kutas, M. (1995). Do the waves begin to waver? ERP studies
of language processing in the elderly. In T. Bashore & Ph. Allen (Eds.),
Age Differences in word and language processing (pp. 314-344). Amsterdam:
Elsevier.
King, J. W., & Kutas, M. (1995). Who did what and when? Using word- and
clause-related ERPs to monitor working memory usage in reading. Journal
of Cognitive Neuroscience, 7, 378-397.
Kutas, M., & Hillyard, S. A. (1984). Brain potentials during reading reflect
word expectancy and semantic association. Nature, 307, 161-163.
Luders, H. O., Lesser, R. P., Dinner, D. S., Morris, H. H., Wyllie, E.,
Godoy, J., & Hahn, J. H. (1992). A negative motor response elicited by electrical
stimulation of the human frontal cortex. Advances in Neurology, 57, 149-157.
MacDonald, M .C. (1994). Probabilistic constraints and syntactic ambiguity
resolution. Language and Cognitive Processes, 9, 157-201.
McConkie, G. W., & Rayner, K. (1975). The span of the effective stimulus
during a fixation in reading. Perception & Psychophysics, 17(6), 578-586.
McRae, K., Jared, D., & Seidenberg, M. S. (1990). On the roles of frequency
and lexical access in word naming. Journal of Memory & Language, 29(1),
43-65.
Neville, H. J., Coffey, S. A., Holcomb, P. J., & Tallal, P. (1993). The
neurobiology of sensory and language processing in language-impaired children.
Journal of Cognitive Neuroscience, 5, 235-253.
Neville, H. J., Kutas, M., Chesney, G., & Schmidt, A. L. (1986). Event-related
brain potentials during initial encoding and recognition memory of congruous
and incongruous words. Journal of Memory and Language, 25, 75-92.
Neville, H. J., Mills, D. L., & Lawson, D. S. (1992). Fractionating language:
different neural subsystems with different sensitive periods. Cerebral Cortex,
2(3), 244-58.
Nobre, A. C., & McCarthy, G. (1994). Language-related ERPS - scalp distributions
and modulation by word type and semantic priming. Journal of Cognitive Neuroscience,
6(3), 233-255.
Polich, J., & Donchin, E. (1988). P300 and the word frequency effect. Electroencephalography
and Clinical Neurophysiology, 70, 33-45.
Pollatsek, A., Bolozky, S., Well, A. D., & Rayner, K. (1981). Asymmetries
in the perceptual span for Israeli readers. Brain & Language, 14(1), 174-180.
Sasaki, K., Gemba, H., Nambu, A., & Matsuzaki, R. (1993). No-go activity
in the frontal association cortex of human subjects. Neuroscience Research,
18(3), 249-252.
Seidenberg, M. S., & McClelland, J. L. (1989). A distributed, developmental
model of word recognition and naming. Psychological Review, 96, 523-568.
Swinney, D. A., Zurif, E. B., & Cutler, A. (1980). Effects of sentential
stress and word class upon comprehension in Broca's aphasics. Brain & Language,
10, 132-144.
Thibadeau, R., Just, M. A., & Carpenter, P. A. (1983). A model of the time
course and content of reading. Cognitive Science, 6, 157-203.
Van Petten, C., & Kutas, M. (1991). Influences of semantic and syntactic
context on open and closed class words. Memory and Cognition, 19, 95-112.
Van Petten, C., Kutas, M., Kluender, R., Mitchiner, M., & McIsaac, H. (1991).
Fractionating the word repetition effect with event-related potentials.
Journal of Cognitive Neuroscience, 3, 131-150.
[CRL
Newsletter Home Page] [CRL Home Page]
Center for Research in Language
CRL Newsletter November 1995 Vol. 10, No. 2