Cognitke Brain Research, 1 (1993) 183-192 183

0 1993 Elsevier Science Publishers B.V. 092~6410/9

Open access under CC BY-NC-ND license.

BRESC 30013

Event-related brain potentials during natural speech processing:

effects of semantic, morphological and syntactic violations

Angela D. Friederici, Erdmut Pfeifer and Anja Hahne

Cogniritr ScienceLob Be&n, Insfifur fiir PsyBofogie, Freir Unirersiriit Be&t, Be&t IFRG)

(Accepted 20 July 1993)

Key ~orrls: Event-related potential; N4OO;Auditory word processing; Semantic priming; Syntactic priming

The present study investigated different aspects of auditory language comprehension. The sentences which were presented as connected speech
were either correct or incorrect including a semantic error fselectional restriction). a morphological error (verb inflection), or a syntactic error
(phrase structure). After each sentence, a probe word was presented auditorily, and subjects had to decide whether this word was part of the
preceding sentence or not. Event-related brain potentials (ERPs) were recorded from 7 scalp electrodes. The ERPs evoked by incorrect
sentences differed significantly from the correct ones as a function of error type. Semantic anomalies evoked a ‘classical’ N400 pattern.
Morphological errors elicited a pronounced negativity between 3Of! and 600 ms followed by a late positivity. Syntactic errors, in contrast, evoked
an early negativity peaking around I80 ms followed by a negativity around 400 ms. The early negativity was only significant over the left anterior
electrode. The present data demonstrate that linguistic errors of different categories evoke different ERP patterns. They indicate that with using
connected speech as input, different aspects of language comprehension processes cannot only be described with respect to their temporal
structure. but eventually also with respect to possible brain systems subserving these processes.

INI’RODUCTION particular syntactic analysis and interpretation, interact

on-line as the linguistic information enters the sys-
An adequate description of language comprehen- tem3’[Link] debate between these positions can thus
sion requires not only a definition of the different be reduced to the question of when during language
subsystems involved in this process, but also a specifi- processing information from the different subsystems
cation of the temporal parameters under which these come together.
subsystems are active. Most comprehension theories The specification of the temporal parameters under
agree that various functionally distinct components re- which the different subsystems interact has been sub-
sponsible for the processing of phonological, syntactic ject of a number of recent behavioral studies. The
and semantic aspects are engaged in language compre- existence of functionally distinct subsystems of the lan-
hension. They differ, however, with respect to their guage system has been supported by reaction time
assumption about whether and how these different experiments conducted with normal subjects’0*‘5”‘*4’,as
subsystems interact in time. well as by those involving subjects with circumscribed
Two distinct views have been formulated with re- brain lesions”“‘. Ferreira and Clifton’“, using mea-
spect to how these different components are activated suresof eye movements during reading, for example,
during language comprehension. One position claims provided evidence in support of distinct processing
that the different components are activated in a serial components for purely structural and interpretative
fashion with the phonological and the syntactic analysis processes. They found that syntactic processing strate-
preceding the semantic interpretation”*‘4*47. The alter- gies were applied even when those resulted in themati-
native view holds that the different subcomponents, in cally based anomaly or when they conflicted with se-

Correspondence:A. Friederici. Institut fir Psychologie, Freie Universitat Berlin, Habelschwerdter Allee 45, D-14195. Berlin. FRG. Fax: (49) (30)
838-3824.
mantie discourse biases. These data are compatible ing language comprehensiony.“‘,“.~. With respect to
with the comprehension account proposed by Fra- language processes a specific negative component
zierlzJ4 assuming two temporally distinct processing peaking around 400 ms (N400) after the onset of the
stages. On-line comprehension studies conducted with target word presentation has been identified to vary as
subjects suffering from specific left hemisphere brain a function of the context =. It was demonstrated that
lesions suggest that structural processes are selectively the amplitude of the N4OO for a sentence final word
affected by left anterior brain lesions’h*‘[Link] 16, was a monotonic function of the cloze probability. i.e.
for example, has shown that patients with lesions in the the N400 was greater in amplitude for less predii:ahle
Broca’s area, but not those with lesions in the Wer- words than for more predictable words. The variation
nicke’s area, are selectively slowed down when re- of the semantic context was shown to influence the
quired to monitor function words in auditorily pre- N400 in word context”*3s*43,w and sentence context
sented sentences. studies x*2h4x. It was suggested that the amplitude of
Thus, it seems that a ‘fine-grained look’ at the the N400 is sensitive to the build up of semantic
temporal structure of the language comprehension pro- constraints upon recognition/ processing of the suc-
cess is compatible with a view proposed by Frazier”*‘4 ceeding word2N*ZV.
assuming an initial stage during which the parser iden- The attempts to find an equally specific correlate for
tifies the structure of the input on the basis of syntactic syntactic processes have been less successful. The first
category information and a late stage during which studies that investigated the influence of syntactic
final integration of thematic and structural information structure on auditory word recognition using ERP
takes place. At this latter stage reanalysis of the initial measures were those by Brown and colleagues2-“. They
parse may turn out to be necessary in case of a mis- presented subjects identical word forms which had two
match between the initial parse and the final interpre- readings, a noun reading and a verb reading (e.g. fire).
tation is detected. This form was presented in two different disambiguat-
The critical aspect to be taken into account when ifig contexts, a noun context (e.g. sit by the fire) or a
modelling language comprehension, it appears, is the verb context (e.g. ready, aim, fire). The identical word
temporal structure of this process. The characteriza- forms which were presented repeatedly in each of
tion of the temporal structure can be done most di- these two contexts produced different ERP patterns as
rectly when investigating spoken language comprehen- a function of the preceding syntactic context. A princi-
sion on-line, as speech unrolls as a sequence in time. pal component analysis revealed an early negativity at
The on-line measure, usually employed in psycholin- around 150 ms over the left hemisphere with a larger
guistics, is the registration of reaction times either in a amplitude for the noun than for the verb context
lexical decision, a word-monitoring or in a speech- condition. This difference in amplitude was most
shadowing task. These tasks, however, require a per- prominent at left anterior sites.
son’s reaction to some given input and the measures Kutas and Hillyard” set out to investigate the pro-
are, therefore, separated in time from the actual per- cessing of morphosyntactically incorrect sentences dur-
ception process. ing reading using event-related potential measures.
One way to investigate the time course of semantic With English as the experimental language, their incor-
and syntactic processes more directly is to register rect sentences contained errors in number marking.
brain potenrials. A number of recent studies have done They reported a tendency for these errors to be associ-
SO, mostly using a word-by-word visual presentation ated with increased negativity between 200 ms and 500
mode with pauses of different length between each ms post stimulus with a slightly more frontally located
word (for a review see refs. 28 and 48). Only a few maximum than the classical N400 effect. The effect,
studies used auditory language material’*2’*32or natural however, failed to reach the required reliability.
speech7*22while registrating event-related brain poten- Miinte et al.36 investigated syntactic and mor-
tials. None of these auditory studies, however, investi- phosyntactic violations during reading using German as
gated syntactic aspects of language processing, they the experimental language. As morphosyntactic viola-
rather focussed on semantic aspects. tions, they introduced case marking errors. Syntactic
violations were realized in sentences containing phrasal
Event-related brain potentials continuations which were incorrect with respect to the
The electrical activity of the brain time-locked to word’s syntactic class (noun vs. conjunction). Both types
the presentation of a stimulus, the so-called event-re- of violations were correlated with a negativity around
lated potential (ERP), has been shown to be sensitive 400 ms.
to a variety of sensory and cognitive processes includ- Two more recent studies analyzed ERP responses
 185

with respect to the processing of syntactic informa- around 700 ms. Subjacency constraint violations pro-
tion encoded in verbs, i.e. subcategorization infor- duced a broad parietal late positivity similar to that
mation4’[Link] et al.42 investigated the processing observed by the studies mentioned above.
of correct and incorrect German sentences containing Thus, taken together, the data at hand suggest at
verbs which cannot be passivized (e.g. ‘Der Lehrer least three different waveforms to be related to syntac-
wurde gesehen’ (The teacher was seen) vs. ‘Der Lehrer tic processing.
wurde geweint’ (The teacher was cried)). Sentences First, Brown and colleagues’ using auditory Ian-
were presented using a visual word-by-word presenta- guage material found an early left anterior negativity
tion with interword intervals of 100 ms and 200 ms, peaking at around 150 ms as a correlate fcr the pm-
respectively. Targets preceded by the syntactically in- cessing of syntactic word category information in a
correct context elicited a negative going wave peaking principal component analysis.
around 400 ms post-stimulus which was significant over Second, a number of studies report a left anterior
the left and frontal electrode sites. The topography of negativity peaking at around 400 ms in correlation with
this wave clearly differed from that elicited by a seman- a variety of different aspects of syntactic processing
tically incorrect context for a given past participle form during reading for the processing of lexically bound
(selectional restriction violation, e.g. The president was syntactic information4*, the processing of those ele-
murdered vs. The milk was murdered) which was simi- ments that carry syntactic information, i.e. the closed
lar to those reported in the literature for semantic class elements”, or the processing of syntactically li-
anomalies. Interpreting these data with respect to their censed filler-gap dependencies*4.
temporal and topographic characteristics, the negativ- Third, some studies found a late parietal positivity
ity peaking around 400 ms was taken to reflect the at around 600 ms and later in relation to the processing
stage of full lexical access, as both types of violations of syntactic aspects during reading. Such a late positiv-
elicited such a waveform in this time range. The partic- ity was observed for sentences in which the preferred
ular topography was viewed to reflect the processing of syntactic reading did not allow plausible interpreta-
semantic versus syntactic information encoded in the tions and required reanalyses (ref. 40; see also ref. 19).
verb. These ERP data in their heterogeneity cannot easiIy
Osterhout and Holcomb4” studied violations of a be connected to current psycholinguistic models of
verb’s subcategorization frame (e.g. ‘The banker per- language comprehension. Whether the heterogeneity is
suaded to seli the stock? in comparison to correct due to differences in the various aspects of syntactic
sentences (e.g. ‘The banker decided to sell the stock’). processing tested, the different presentation modes
They found a late (about 600 msl positive deflection (auditory vs. visual) or even languages used (English,
over parietal areas of the scalp in correlation with such German, Dutch) cannot be decided on the basis of the
violations. Note, that this component may not only be a data at hand. However, when considering the temporal
reflection of a syntactic error, but also be correlated distribution of the reported effects across the different
with a so-called garden-path effect, i.e. requirement of studies, we can state a difference between lexical-
a structural reanalysis in these sentences which be- semantic effects clearly located around 400 ms and
comes evident only at the end of the sentence. A syntactic processes either being correlated with early
similarly distributed late positivity has also been ob- negativities (around 150 ms and 400 ms) or with a late
served for other types of syntactic anomalies, as for positivity (around 600 ms and later).
example, sentences containing the incorrect order of
words in a phrase or inflectionally marked incorrect The present study
subject-verb agreement in Dutch sentences”. Given The present study was designed to address the ques-
the onset polarity and topography, the late positivity tion whether different aspects of language parsing be-
could be viewed as a member of the P300 components come manifest in temporally and topographically dis-
usually elicited by unexpected eventssq9. tinct patterns of event-related potentials in the same
Finally, Neville et al.37 studying different types of subjects. In particular, we investigated the processing
syntactic violations found each to be correlated with a of three different types of information, semantic infor-
different ERP pattern. Phrase structure violation pro- mation, morphological information and syntactic infor-
duced a biphasic effect with a left frontal negativity mation. Given the assumptions about early fast and
peaking at around 400 ms and a parietal positivity automatic syntactic processes involved in the initial
peaking at around 700 ms. Specificity constraint viola- parse, we reasoned that these may be more likely to be
tions evoked a sustained left frontal negativity emerg- observed when presenting the language material on-
ing immediately after the critical word and peaking at line. We therefore decided to present the stimulus
186

material not in a word-by-word manner but as running s!!!m!a

km. 200 ms)
speech.
The prediction for the semantic violation condition
was that it should elicit a classical N400 wave. The
predictions for the other two types of violations were
less straight forward. given the state of the art dis-
cussed above. Moreover, as we were interested in
studying the processing of the different types of viola-
tion under the condition of natural speech input, the
basis for clear predictions was even more sparse. The correct sentence

only study using natural speech as input while record-

ing ERPs had only investigated semantic aspects of Inconect, semantic

sentence processing”.
incorrect, morphological : Das Park&t
MATERIALS AND METHODS (RNpu9u~~yYpoashl

incorrect syntactic : Der Freund Im iiesucht.

Suhjecrs ~eMsndwu!nlheVidw6J
Six&u young adults (eight male. eight female. mean age = 3.9
yearsj were paid DM IO.-- per hour to S~IVX as subjects. All hut one Fig. I. Examples 01 stimulus items for each condition. It displays one
were right handed as evaluated hy the Edinburgh Inventory (Old- example of a sentence as an oscillogram. in which the first critic?1
field, 1971). They were all native speakers of German with no known point. i.e. the beginning of the auxiliary ‘wurde’ and the second
hearing deficit. critical point. i.e. the beginning of the main verb are marked.

A set of IhO rxprrimental sentence?, was constructed. These fell tion: 40 sentences contained a complete prepositioral phrase. i.e.
into 4 different categories: (1141 sentences were correct *. (2) 40 preposition plus noun phrase (20 with full nouns, 20 with pronouns).
were semantically incorrect containing a selectional restriction viola- and 40 sentences with the structure art-noun-auxiliary-past partici-
tion, (3) 40 were morphologically incorrect containing a verb which ple.
was inflected inmrrectly. and (4) 40 were syntactically incorrect Each sentence was spoken by a female speaker. The speaker was
containing a violation of the phrase structure. Examples of each trained to produce correct and incorrect sentences with normal
sentences type are displayed in Fig. 1. intonation or a best approach. These were produced in analogy to
The selectional restriction error in sentence type (3 concerned a correct sentences containing a complete prepositional phrase with a
mismatch between the preceding noun and the sentence final verb. pronoun (e.g. incorrect: Der Freund wurde im hesucht/The friend
The morphological error (3) was caused by a mismatch between the was in the visited: correct: Der Freund wurde von ihm hesucht/The
preceding auxiliary marking passive voice and the sentence final verb friend was by him visited) **. Sentences were first recorded on
form (1st person singular instead of a past participle form). In this analogue tape and were then digitized (20 kHz, 12 bit resolution).
condition phrase structure assignment is possible as a main verb is As the experiment used a probe verification paradigm with trials
required in this position. however. in a different inflectional form. in which each sentence was followed by a pause of 800 ms after
The violation of the phrase structure in sentence type (4) was due to which a probe word was presented auditorily. the following trials
a mismatch between the preceding head of the phrase (here the were constructed. Half of the prohe words were so-called true
preposition indicating the heginning of a prepositional phrase requir- probes. i.e. words that had occurred in the sentence immediately
ing a noun or pronoun as the next word) and the following word in preceding the probe. Words from each position could serve as a
sentence final position (here an inflected verb form. a category which prohe. This word position-probe relation was equally distributed over
is syntactically illegal as successor after a preposition). All 40 exem- the entire set of the true probes for the correct sentences and the
plars of each sentence category were of the same type (i.e. article- semantically and the morphologically incorrect sentences with a
noun-auxiliary-(prepositionbmain verb) rangmg in length between 6 quarter each. For the syntactically incorrect sentences three quarters
and 9 syllables (article = I syllable. noun = l-3 syllables, auxiliary = Z were filled with nouns, auxiliaries and main verbs. The fourth quar-
syllables. main verb = 2-3 syllables) or IO syllables in case of those ter consisted of articles and prepositions to equal parts. Half of the
sentences containing a preposition. This corresponds to a mean prohe words were so-called false probes. as they occurred not in the
sentence !cngth of 1.7 s (correct sentences), 1.X s (semantic condi- previously heard sentence. False probes were morphological variants
tion). 1.7 s (morphological errors) and I.9 s (syntactic condition). In of words that had occurred in the sentence, e.g. an article with a
all the sentences, the sentence final main verb form was the critical different case or gender marking, a noun with a different number
word. These I60 critical words were all two- or three-syllable words. marking. an auxiliary with a different tense or number marking, and
The 80 two-syllable and the X0 three-syllable words were equally a main verb with a different tense or number marking. We chose to
distributed over the different conditions. We chose to control for the use morphological variants as probes in order to keep the auditory
critical words’ and the sentences’ length on the basis of a number of attention at a high level. Pretests had shown the present procedure
syllables, as a syllable can be considered to be a critical unit in to be successful. Again these four categories were equally distributed
speech ptrceptionJ”.“. over the false probe set for the correct sentences and the semanti-
In addition to these experimental sentences, 80 correct filler cally and morphologically incorrect sentences. For the syntactically
sentences were added to the stimulus list with the following composi- incorrect sentences. three quarters of the probes fell in the three

-- _
* 1 nitty-seven ot these sentences were of an average cloze probability of 0.43 (ranging from 0.10 t!lO.901. As three of the original sentences had
a cloze probability of zero. these were excluded from later analysis.
** Attemps were made to splice the auditory material of correct sentences and use these parts for the incorrect sentences. This procedure,
however, resulted in most unnatural sounding material, we !herefore decided to use the procedure described in the method section.
 187

categories of noun, auxiliary and main verb. The last quarter con- ms between the offset of the sentence and the onset of the probe
sisted of articles and prepositions to equal parts. preventing the motor response from contaminating the ERP to the
These probe words were spoken as individual words by the same final word.
female speaker who bad spoken the sentence material on analogue Subjects were instructed to give their response as fast and as
tape. Like the sentences. probe words were digitized and were ~nred accuratelv ‘rr pnssitie. After a silent pause o! 2(100 ms the next trial
in separate disk files. Using this technique. trials with standard began. Throughout the task subject wile piewtriicu’ ;ci:h m oa!!inr
interstimulus intervals of 800 ms between sentence offset and probe of a small black rectangle in front of him/her. They were told that
word onset were constructed. This was true for all critical trials as the fixation of this rectangle would ease the required restriction of
well as for the filler trials. All sentences and probes were reassem- eye movements during the task. Ail subjects were confronted with
bled and output in a quasi random order on an analogue tape the IWO tapes containing malerial set A and material set B in a fifed
together with IO practice trials preceding the experimental trials. ordvr.
Two differently randomized analogue audio tapes were constructed
from the same material (set A and set B). Set A and set B differed in
ERP recording
their combination of sentencesand probe types: those sentences that
The EEG was recorded from 7 scalp electrodes including 3 places
were combined with a true probe in set A were combined with a
of the standard international IO-20 system locations at the midline
false probe in set B, and vice versa.
(Fz, Cz, Px) and 4 non-standard locations: Broca’s region (BI was
defined in the fronto-temporal region as crossing point between
Procedure T3-Fz and F?-Cz) and the right hemisphere homologue (Br, i.e.
The probe verification paradigm required the subjectsto listen to crossing point betwera T4-Fz and F8-Cz) and the Wemicke’s
the auditorily presented sentence and a target following after a region (WI was defircd in the posterior-temporal region as crossing
pause of 800 ms. They were asked to indicate whether or not the point between T3-P3 and C3-PS) and the right hemisphere homo-
probe word had appeared in the preceding sentence. This response logue IW, i.e. crossing point between T4-P4 and C4-P6). All scalp
was delayed by the introduction of an interstimulus intenal of 800 electrodes were reR%;encedto linked mastoids. Eye movements and

PV
Grand sveragc: SEllANTICS (n*16) FZ
- correct -2.5 : .... .. .
_ _

,“‘_..,....~..,““,““ff

0 0.5 1 1.5

0 0.s I 1.5
2.5-

Fig. 2. Grand averages of ERPs in the semantic violation condition. Superimposed are the ERPs obtained for the sentence final main verb (the
baseline consists of the first 250 ms of the auxiliary ‘w&e not displayed here) in correct sentences (solid line) and in semantically incorrect
(broken line) sentences. The panels with event-related activity are arranged according to the placement of the electrodes: left panels are
‘Broca’-left (Bl), ‘Wernicke’-left (WI), middle panels are midline frontal (Fz), midline central (Cz), midline parietal (Pz); right panels are
“Broca’right (Br) and ‘Wernicke’-right (Wr). The vertical line indicates the beginning of the main verb 3. Negativity is up in this and the
following figures.
188

blink artifacts were monitored by the electrooculogram (EOG) syntactically incorrect condition, however. the word directly preced-
recorded from electrodes placed ahove and beneath the left eye. All ing the main verb was always a preposition. In order to allow a direct
electrodes used were Ag/AgCI electrodes. All impedances were comparison over the different conditions. prestimulus baseline was
maintained below 5 kf1. The EEC and EOG channels were ampli- taken for the same word type. here the auxiliary ‘wurde’ (see Fig. 1).
fied hy ESMED amplifiers with 1.6 s time constant and a low pass The second critical point marked the onset of the sentence final
filter at 70 Hz. -3 Db/octave roll-off. The EEG and EOG were word. ERPs to the sentence final words were quantified in three
recorded continuously for each block and were A/D converted with different latency windows (LOO-250 ms. 2X&700 ms and 700-1200
13bit resolution at a rate of 36 Hz. Data collection was controlled ms). These latency windows were defined after visual inspection of
by a 3864BM compatible computer. The EEG was digitized on-line the grand averages of the four different conditions, and the visual
and stored for later analysis. identification of the main differences in amplitudes.
Off-line separated ERPs were averaged for each subject at each Repeated measures analyses of variance (ANOVAs) were used to
electrode site from trials free of EOG artifact for the sentence final analyze these data. Factors in the analysis included electrode site
words in each condilion. The rejected trials due to artefacts were (Fz, Cz, Pz, 91, Br. WI, Wr). time epoch flOO-250 ms. 250-7M) ms,
distributed equally over the different conditions with a mean of 700-1200 ms) and condition (correct, semantic violation, morpholog-
%.75% (S.D. = 12.27) for the correct sentences. a mean of 25.00% ical violation. syntactic violation). In cases where specific predictions
(S.D. = 12.94) for the semantically incorrect sentences. a mean of were made for final words, the overall ANOVA was followed by
2531% (SD. = 12.1 I) for the morphologically incorrect sentences planned painvise comparisons.
and a mean of 26.41% (S.D. = I?.661 for the syntactically incorrect The general linear model procedure of SAS Institute Inc. was
sentences. used for the statistical analyses. All tests were adjusted for violations
For each sentence. two critical points were used to allow time- of sphericity which is inherent in repeated measure analysis accord-
locked EEG averages for the sentence final words: the first heing at ing to the formulas of Huynh and Feldt’>.
the onset of the auxiliary ‘wurde’ and the second heing at the onset
of the main verh. The YO ms following the first critical point were
used to calculate the prestimulus baseline. We choose these ?Stt ms RESULTS
of the auxilia~ ‘wurde’ as the word which was closest in time to the
critical word and present in the sentences of each condition. The
auxiliary directly preceded the main verh in the correct. the semanti- As behavioral data, error rates were computed sepa-
cally incorrect and the morphologically incorrect condition. In the rately for each condition and probe type (true vs. false

Uv
Grand aversge: NORPt+OSYNTAX (n-16) FZ
- CorrecI -2.5
.. norpn. Inc. ..( ...

-2.

-2, ,5

2.

5
0 0.5 I 1.5
2.5-
‘. ._ ,:.
~“‘,““,““,“‘~.~ “‘,,. n-rls
0 0.5 1 I.5
Fig. 3. Grand averages of ERPs in the morphological violation. Superimposed are ERPs obtained from the sentence fhnd main verb (the baseline
*cutsisIs of the first 30 ms)of the auxiliary ‘wurde’ (not displayed here) in correct sentences (solid line) and in morphologically incorrect (broken
line) sentences. Layout as Fig. 2.
 I89

probe). Descriptive analyses showed that overall error and virtually non-existent at the posterior electrodes
rates amounted to 1.1% and distributed equally over (Fig. 4). In the time domain between 250-700 ms we
conditions. Error rates were too low to be analyzed by observed a negative going wave for the semantically
means of inferential statistics. incorrect condition which had the clear distribution of
The average voltage in the 250 ms following the first the classical N400 (Fig. 2). Morphological errors also
critical point (i.e. the onset of the auxiliary ‘wurde’) elicited a negativity around 400 ms, however, with a
was examined for significant differences as a function smaller amplitude than the ‘semantic N4OO’,peaking
of experimental corrdition. Since no systematic effects somewhat earlier and merging into a positivity at
were found in an ANOVA analysis, this interval was around 600 ms (Fig. 3). The syntactically incorrect
used as a baseline for all further ERP amplitude mea- condition also elicited a negativity in this time domain
sures. which, however, was most prominent at anterior and
Visual inspection of the grand average ERPs elicited frontal electrodes and basically absent at posterior
by the final words per violation condition against the electrodes (Fig. 4). Given these observations we chose
correct condition plotted in Figs. 2-4 indicates differ- three different time windows (i.e. time epochs 100-250
ent waveforms for each of the incorrect condition, with ms/250-700 ms/700-1200 ms) for the following calcu-
respect to their temporal and topographic distribution. lations.
A more detailed inspection of the time course of the An overall ANOVA with the factors Condition (cor-
ERPs revealed a very early negativity peaking at I80 rect/ semantic/ morphological/ syntactic violation) X
ms only for the syntactic condition. Although this nega- Time epoch (100-250 ms/250-700 ms/700-1200 ms)
tivity can be seen at El, Br and Fz, it was largest at BI x Electrode (Fz, Cz, Pz, Bl, Br, WI, Wr) revealed a

Grand average: SYNTAX (n-16) ‘FZ

- correct
....... Syn . In c .

- PV I
@iI
I
.._._......_
_._...
--...-....__..

-2.5

Fig. 4. Grand averages of ERPs in the syntactic violation. Superimposed are ERPs obtained from the sentence final main verb (the baseline
consists of the first 250 ms) of the auxiliary ‘wurde’ (not displayed here) in correct sentences (solid line) and in syntactically incorrect (broken
line) sentences. Layout as Fig. 2.
190

main effect of Condition (F,, = 3.71, P< 0.02). a effect of Electrode (Fevw = 24.47, P < 0.001) and Con-
main effect of Time epoch (F2.3, = 10.25, P < 0.001), dition by Electrode interaction (F,,,* = 5.63, P < 0.001).
and a main effect of Electrode (F,, = 17.42, P <
0.~001). All interactions were significant (Time epoch DISCUSSION
x Condition: F,., = 6.58, P < 0.0001; Time epoch X
Electrode: F,,.,, = 14.43, P : 0001; Condition X The three bmctionally different types of linguistic
Electrode: F,x_Z,O= 2.71, P < 0.01; Time epoch X violations presented in connected speech each elicited
..
Condnron x Electrode: [Link] = 4.34. P < 0.0001). ERPs which were clearly different in their temporal
The local ANOVA for the time epoch 100-250 ms and topographic parameters. Semantic selectional re-
with the factor Condition (correct vs. syntactically in- striction errors became manifest in a negative going
correct) by Electrode revealed a significant main effect effect in the time domain at around 400 ms with a
of Electrode (F,.,, = 5.20, P < 0.003) and a significant broad distribution over both hemispheres similar to the
interaction ( Fhew= 3.53, P < 0.03). Planned r-tests classical N400 effect. The syntactic phrase structure
showed that the difference between the correct and the errors realized as a violation of syntactic word cate-
syntactically incorrect condition was only significant at gory, in contrast, evoked an early negativity peaking
Bl (fi, = -2.21, P < 1).05x around 180 ms with a maximum over frontal and
The local ANOVA for the time epoch 250-700 ms anterior lateral electrode sites, most salient over the
with the factor Condition (correct, semantically incor- electrode ‘Broca left’. This early negativity was fol-
rect, morphologically incorrect and syntactically incor- lowed by a second negative going wave peaking at
rect) and Electrode (BI, Br, Fz, Cz, Pz, WI, Wr) re- around 400 ms which, however, was less widely dis-
vealed a significant main effect of Condition (F3.45 = tributed than the semantic N400 effect and had its
8.68, P < 0.0004), a main effect of Electrode ( Fhaul,= maximum at frontal and anterior sites. The detection
8.54, P < 0.001) and a significant interaction (F,,,,,, = of a morphological violation evoked a negativity at
3.75, P < 0.001). This interaction was analyzed further around 400 ms with a maximum over anterior sites,
by conducting planned t-tests. The difference between followed by a weak late positivity with parietal and
the’correct and the incorrect condition was significant posterior lateral distribution.
for the semantically incorrect sentences at all elec- The topography and the timing of the semantic
trodes tall P’s < 0.01, except for WI with a P < 0.09. ERP-effect as observed in the present study using
For the morphologically incorrect sentences this differ- auditory language input is in line with previous studies
ence was reliab!e at the fronto-temporal electrodes (BI, using visual and auditory presentation modes22,d8.
Br) and at Pz with P < 0.05. For the syntactically Compared to the N400 effect as a correlate for
incorrect sentences this negativity was only significant semantic processes, syntactic effects are much less es-
at the fronto-temporal electrodes (Bl. Br). tablished. In the tntroduction we discussed effects re-
In order to approach the specific question whether ported for each of the time domains under analysis in
the observed negativities for the syntactic violation the present study. In the time domain between 100 and
condition in the first and in the second tkne epoch 250 ms we found an early negativity left anterior for
(loo-250 ms and 250-700 ms) are to be attributed to the phrase structure violation induced by an incorrect
different generators, an additional analysis was con- word category. A similar early negativity was observed
ducted on normalized data. The normalization was by Neville et al.“’ in correlation with phrase structure
carried out according to McCarthy and Wood”“. This violations. However, a negative going wave as early as
analysis over Time epoch (2 levels) and Electrodes (7 50-100 ms at anterior sites and most prominent in the
levels) revealed a significant interaction ( Fhqqo= 2.79, left hemisphere was also found in correlation with
P < 0.05). semantic anomalies when presented in a connected
The local ANOVA for the time epoch 700-1200 ms speech mode2’. The authors interpret their finding to
with the factor Condition (correct vs. morphologically reflect an interaction between the semantic and
incorrect) and Electrode revealed no significant main prosodic or coarticulatory cues. Connolly, Stewart and
effect or interaction. A look at the visually displayed Phillis’ investigating the processing of high and low
waveforms suggested that this may be due to the fact probability words in spoken sentences, found both an
that the negativity in the preceding time window only early negativity (‘N200’) and the N400 for the low
spanned between 250 and 600 ms and that the observ- probability in contrast to high probability sentence-fi-
able positivity started at 600 ms. Therefore an addi- nal words, interpreting the first negativity to reflect the
tional ANOVA was calculated using the time epoch acoustic analysis and the second the semantic analysis.
600- 1200 ms. This ANOVA revealed a significant main Given these data from the auditory language process-
 191

ing domain, the possibility must be considered that the speech as input combined with ERP measures clearly
early negativity observed in the present study using indicates that there are ways to identify different tem-
connected natural speech may not solely be due to the poral stages during language comprehension and to
syntactic violation introduced, but may bf: influenced connect these to distinct neural systems in the human
by hidden prosodic cues. Interestingly, in the present brain.
study no early difference between the correct and
incorrect sentence was observed for the semantic and Acknowledgements. This work was supported by the Alfried Krupp
the morphological condition suggesting that possible von Bohlen und HalbachScience Award assigned to the fii author
and by a grant fmm the German Research Foundation (DFG-FR
prosodic influences must have been tied to incorrect 519/12-l) assigned to the first aulhor and to Axe1 Mecklinger. The
phrase structure, and not to incorrectness per se. This authors wish to thank Burkhard Mae6 for his advice and support
makes it most likely that the observed effect is indeed concerning the hardware and software used for the present cxpcri-
merits, and Axe1 Mecklinger for his advice in anaIyzing the data and
due to the syntactic violation introduced. his comments on the manuscript.
For the morphosyntactic violation induced by an
incorrect verb inflection we found a late positivity at
REFERENCES
the central posterior site and a negativity around 400
ms preceding it. The observed negativity may be similar
I Bentin, S.. Kutas, M. and Hillyard. S.A., Electrophysiological
to the negativity reported by Kutas and Hillyard% for evidence for task effects on semantic priming in auditow word
morphosyntactic number ma&ing. The late positivity processing, Psychophysiu!~, 30 (1993) 161-169.
2 Brown, S.S.. Lchmann, D. and Marsh, J.T., Linguisticmeauing-
may be connected to the ‘late positive shift’ reported
related differences in evoked potential topography: English,
by Hagoort et aLI in correlation with verb inflection SwissGerman and imagined, Brain Lang., 11 (1980) 340-353.
errors marking subject-verb-agreement errors in Dutch. 3 Brown,W.S., Marsh,R.E. and Smith, J-C.. Contextual meaning
effects on speech-evoked potentials, Behar. Biof., 9 (1973) 75%
These authors, however, found a late positivity also in 761.
correlation with a particular phrase structure violation 4 Bmwn. W.S., Marsh, R.E. and Smith, J.C., Evoked potential
requiring a reanalysis of a preferred syntactic reading wave-form differences produced by the perception of different
meanings of ambiguous phrase, Electtwncephabgr. Clin. Neuro-
of a given sentence. Given its distribution, this late pkyol., 41 I19761 113-123.
positivity may be a member of the P300 component, 5 Brown, W.S., Marsh, R.E. and Smith, J.C., Principal component
often found following unexpected stimuli which require analysis of ERP differences related to the meaning of an ambigu-
ous word, Eleclrdencephalogr. Clin. Neurophysiio..46 (1979) 709-
an updating of preceding information8*20. 761.
The data of the present study evaluating three dif- Clifton, C., Speer, S. and Abney, S., Parsingarguments:phrase
structure an argument structure as determinants of initial parsing
ferent aspects of language processing in the same sub- decisions, .I. Mem. Lang.? 30 (1991) 251-271.
jects suggest that these aspects are processed at differ- Connolly, J.F., Stewart, S.H. and Phillis, NA., The effects of
ent points in time during comprehension. Processing of processing requirements on neumphysiological responses to spa-
ken sentences, Brain Lang.. 39 (1990) 302-318.
word category information, as in rhe phrase structure Donchin, E., Surprise! . . . surprise!, PsychophysioLogy. I8 (1981)
violation condition, seems to precede lexical processes 493-513.
in auditory language comprehecsion; as indicated by 9 Donchin, E. and Coles, M.C.H.. On the conceptual foEnd?tions
of cognitive psychophysiology, Behac. Brain Sci., II (1988) 408-
the early negativity evoked by a word category error 418.
preceding the negativity at 400 ms correlated with 10 Ferreira, F. and Clifton, C., The independence of syntactic pro-
cessing, J. Mem. Lang., 25 (1986) 348-368.
lexical processes. The late positivity observed in some
11 Forster, ICI., Levels of processing and the structure of the
studies’9*40may reflect the detection of the need for language processor. In W.E. Cooper and E. Walker (Eds.), Sen-
structural reanalyses in a given sentence. Once these tence Processing: PsychoIinguistic Studies Presented to MerriU Gar-
rett, Erlbaum Associates, Hillsdale, NJ, 1979, pp. 27-85.
effects observed in correlation with syntactic process- 12 Frazier, L., On comprehending sentences: syntactic parsing
ing, i.e. the early ixgativities and the late positivity, are strategies, University of Connecticut, Doctoral Dissertation, 1978.
validated by further experiments, they may be taken to I3 Frazier, L., Sentence pmcessing: a tutorial review. In M. Colt-
heart (Ed.), Attention and Perfbrmance XII, Erlbaum Associates,
support a two-stage module of syntactic parsing, with a Hillsdale, NJ, 1987, pp. 559-586.
first stage during which a structure-driven parser as- I4 Frazier, L.. Exploring the architecture of the language-processing
system. In G.T.M. Altmann (Ed.), Cognitive ModeIs of Speech
signs an initial structure to the input based on major Processing, MlT Press, Cambridge, MA, 1990. pp. 409-433.
category information, and a second stage during which 15 Frazier, L. and Rayner, K., Making and correcting errors during
structural and semantic aspects are made available sentence comprehension: eye movements in the analysis of struc-
turally ambiguous sentences, Cogn. PsychoI., 14 (1982) 178-210.
before final interpretation takes place’0V13-‘5. I6 Friederici, A.D.. Aphasics’ perception of words in sentential
Further research is clearly needed before we will be context: some real-time processing evidence, [Link],
able to adequately describe the underlying temporal 21 (1983) 351-358.
17 Friederici, A.D., Levels of processing and vocabulary types: evi-
and neurotopological structure of the language com- dence from on-line comprehension in normals and agrammatics,
prehension process. The present study using connected Cognition, 19 (1985) 133-166.
192

18 Friederici. A.D. and Kilbom. K.. Temporal constraints on lan- 35 Mecklinger. A.. Kramer, A.F. and Strayer. D., Event-related
guage processing: syntactic priming in Broca’s aphasia. 1. Cogs. potentials and EEG components in a semantic memory search
Neurosci., I ( 19893262-272. task. Psychophysiology, 29 (1992) 104-l 19.
19 Hagoort. P.. Brown. C. and Groothusen. J.. The syntactic positive 36 Miinte. Tb.F.. Heinze, H.-J. and Prevedel, H.. Ereigniskorrelierte
shift as an ERP-measure of syntactic processing. Lung. Cogrr. Himpotentiale rellektieren semantische und syntaktische Fehler
/4ores.r.. in press. hei Sprachverarbeitung. 2. EEG EMG wrwundte Gebiere. 21
2tt Hillyard. S.A. and Picton, T.W.. Electrophysiology of cognition. (1990) 75-81.
In V.B. Mountcastle. F. Plum and S.R. Geiger tEds.1. Handbook 37 Neville. H.J., Nicole. J.. Barss. A.. Forster. K. and Garret, M.,
of Phvsiology, Vol. V. Higher /%nctiotrs of I/W Bruin. Part 2. Syntactically based sentence processing classes: evidence from
American Physiological Society, Bethesda, 1987. pp. 519-584. event-related brain potentials, L Cogn. Neurosci., 3 (1991) l55-
21 Holcomb. P.J. and Neville, H.. Auditory and visual semantic 170.
priming in lexical decision: a comparison using event-related 38 Neville. H.J.. Mills. D.L. and Lawson, D.L., Fractionating lan-
hrain potentials. Long. CO~II.t’rorcw.. 5 (1990) 281-312. guage: different neural subsystems with different sensitive peri-
22 Holcomh. P.J. and Neville. H.J.. Natural speech processing: an ods. Cerebral Cor~fl, 2 (1992) 244-258.
analysis using event-related brain potentials, &c/robioIogy. I9 39 Oldfield. R.C., The assessment and analysis of handedness: The
t 1991) 286-300. Edinburgh Inventory, Neuropsychologia, 9 f 1971) 97- 113.
23 Huynh. H. and Feldt, LA.. Conditions under which mean square 40 Osterhout, L. and Holcomb. P.J.. Event-related brain potentials
tatios in repeated measurement designs have exact F-distrihu- elicited by syntactic anomaly, 1. Mem. Lang.. 31 (1992) 785-804.
lions, _I.&I. Sruf. Assoc.. 65 (1970) 1582-1589. 41 Rayner. K. Carlson. M. and Frazier, L., The interaction of syntax
24 Kluender. R. and Kutas. M.. Bridging the gap: evidence from and semantics during sentence processing: eye movement in the
ERPs tin the processing of unbounded dependencies. J. Cogu. analysis of semantically biased sentences. 1. Verbal Learn. Verbal
Neurosri.. 5 (1993) 196-214. Eehu:. 22 (1983) 358-374.
25 Kutas, M. and Hillyard, S.A.. Reading between the lines: event- 42 RBsler. F.. Friederici. A.D.. Piitr. P. and Hahne. A., Event-re-
related brain potentials during natural sentence processing, &ah lated brain potentials while encountering semantic and syntactic
Lung.. I I t 1980) 354-373. constraint violations, 1. Cogn. Neurosci., 5 (1993) 345-362.
26 Kutas. M. and Hillyard, S.A.. Event-related brain potentials to 43 Rugg. M.D.. The effects of semantic priming and word repetition
grammatical errors and semantic anomalies. Mem. Cogn.. II on event-related potentials. Psychpitysiobgy, 22 (1985) 642-647.
t 1983) 539-550. 44 Rugg. M.D.. Dissociation of semantic priming, word and non-word
27 Kutas. M., Lindamood. T. and Hillyard, S.A.. Word expectancy repetition effects by event-related potentials, Quart. 1. hp. &-
and event-related potentials during sentence processing. In S. C/IO!.. 39 (1987) 123-148.
Komhlen and J. Requin (Eds.). Prepurarog Smres and PWC~W 45 Segui. J.. Dupoux. E. and Mehler. J., The role of the syllable in
kg. Erlh.+um Press. New Jersey. 1984. pp. 217-237. speech segmentation, phoneme identification and lexical access.
28 Kutas, M. and Van Petten, C., Event-related potential studies of In G.T.M. Altmsnn (Ed.), Cognitilr Models of Speech hcessing,
language. In P.K. Ackles. J.R. Jennings and M.G.H. Coles (Eds.). hfIT Press, Cambridge, MA, 1990, pp. 263-280.
Adrartces it, Psychopi~ysiology , Vol. 3. JAI Press, Greenwich. 46 Schriefers. H.. Friederici, A.D. and Kuhn, K., The processing of
1988, pp. 139-187. !ocai ambiguous relative clauses in German, submitted.
29 Kutas. M., Van Petten. C. and Besson, M., Event-related poten- 47 Swinney. A.D., [Link] access during sentence comprehension:
tial asymmetries during reading sentences, EIccmw~~~eplzalop. treyconsideration of eontext effects, 1. V&u/ Ileum. b!erhd Be-
ciin. Necrropi~ysioL.29 t 1988) 2 18-233. h:, I8 (19793 645-6.59.
30 Marslen-Wilson. W.D.. Functional parallelism in spoken word 48 Van Petten. C. and Kutas, M.. Influences of semantic and syntac-
recognition, Co&ion. 25 (1987) 71-102. tic context on open and closed class words. Mem. Cogn., 19
31 Marslen-Wilson. W.D. and Tyler. L.K., The temporal structure of (1991) 95-112.
spoken language understanding, Cognition, 8 (1980) I-71. 49 Zwitserlood, P.. The locus of the effects of sentential-semantic
32 McCallum. WC.. Farmer, SF. and Pocock, P.K., The effect of context in spoken-word processing, Cognifion. 32 (1989) 25-64.
physical and semanlic incongruities on auditory event-related 50 Zwitserlood. P., Schriefers, H., Lahiri, A. and van Donselaar. W.,
potentials. Elcc~raencepi~alogr. Ciiu. NectrophysioL. 59 ( 19x4) 447- The role of syllables it] the perception of spoken Dutch, J. [Link].
488. Psych. Leaming Memory Cognition, 19 (1993) 260-271.
33 McCarthy. G. and Wood. C.C.. Scalp distributions of event-re- 51 Zurif. E.. Swinne):. D. and Garrett. M.. Lexical processing and
latcd potentials: an ambiguity associated with analysis of variance syntactic comprehension in aphasia. In A. Caramazza (Ed.), Cogn.
models. Elertroencepiudogr. C/in. Nerrrophysioi.. 62 ( 1985) 203- Newopsychol. Newo&:dstics, Hillsdale, NJ, Lawrence Erlhaum,
208. 1990. pp. 123-146. - ’
34 McClelland, J.L.. St. John, M. and Tarahan, R., Sentence EOM-

prehension: a parallel distributed processing approach, Latrg.

Cogn. Procw , 1 (1989) 2X7-336.