78 Review TRENDS in Cognitive Sciences Vol.6 No.

2 February 2002

Towards a neural basis of auditory

sentence processing
Angela D. Friederici

Functional dissociations within the neural basis of auditory sentence which is characterized by serial or syntax-first
processing are difficult to specify because phonological, syntactic and models, holds that syntax is processed
semantic information are all involved when sentences are perceived. In this autonomously prior to semantic information [1,2].
review I argue that sentence processing is supported by a temporo–frontal A second view, represented by interactive or
network. Within this network, temporal regions subserve aspects of constraint-satisfaction models, claims that all
identification and frontal regions the building of syntactic and semantic types of information interact at each stage of
relations. Temporal analyses of brain activation within this network language comprehension [3,4]. Both classes of
support syntax-first models because they reveal that building of syntactic models are supported by a number of sentence-
structure precedes semantic processes and that these interact only during reading studies that use different behavioral
a later stage. paradigms (for details see 1).
None of these models addresses explicitly the role
Listening to connected speech is a task that humans of prosodic information that is available whenever
perform effortlessly each day. This is surprising spoken sentences are processed. Unfortunately,
given the short time that the processing system has the few behavioral studies that have investigated
to deal with different types of information. possible interactions between prosodic and syntactic
Segmental phonemes and suprasegmental information during auditory language comprehension
Angela D. Friederici
phonological information (prosody or pitch) as well do not provide a unitary view: although some data
Max Planck Institute of as syntactic and semantic information must be indicate an interaction between prosodic and
Cognitive Neuroscience, accessed and coordinated within milliseconds. syntactic information [5,6], others do not [7].
PO Box 500 355,
With respect to syntactic and semantic processes, These differences are attributable partly to the
04303 Leipzig, Germany.
e-mail: two alternative views have been proposed in fact that different behavioral paradigms tap into
[email protected] psycholinguistic comprehension models. One view, different processing aspects (automatic versus

Box 1. Psycholinguistic models of language comprehension

Two main classes of models have been proposed to account for the Although in both classes of models syntactic and semantic
behavioral data on language comprehension: serial, syntax-first and information are integrated during language perception to achieve
interactive, constraint-satisfaction models [a]. As these models are understanding, interaction takes place at different points during
based on data from reading, they comprise semantic and syntactic processing: interactive, constraint-based models predict early
processes but ignore prosodic processes. Serial, syntax-first models interaction, whereas serial, syntax-first models predict
assume that the parser initially constructs the simplest syntactic interaction during a later stage of processing.
structure on the basis of word-category information, independent
References
of lexical-semantic information. The latter information is processed a Mitchell, D.C. (1994) Sentence parsing. In Handbook of Psycholinguistics
during a second stage that is responsible for thematic-role (Gernsbacher, M.A. ed.), pp. 375–409, Academic Press
assignment. If the initial syntactic structure and the thematic structure b Frazier, L. (1987) Theories of sentence processing. In Modularity in
cannot be mapped onto one another, reanalysis takes place [b–d]. Knowledge Representation and Natural-language Processing (Garfield, J., ed.),
pp. 291–307, MIT Press
Recent studies, however, indicate that, for ambiguous structures,
c Frazier, L. and Clifton, C., Jr (1997) Construal: overview, motivation, and
the initial structure building is not totally independent of non- some new evidence. J. Psycholinguist. Res. 26, 277–295
structural variables such as the frequency of a particular structure or d Frazier, L. and Rayner, K. (1982) Making and correcting errors during
the semantic plausibility associated with the main verb [e,f]. This has sentence comprehension: eye movements in the analysis of structurally
led to constraint-satisfaction models in which it is assumed that, in ambiguous sentences. Cogn. Psychol. 14, 178–210
e MacDonald, M.C. et al. (1994) Lexical nature of syntactic ambiguity
the case of structural ambiguities, multiple syntactic interpretations
resolution. Psychol. Rev. 101, 676–703
are generated and weighted according to nonstructural factors. f Trueswell, J.C. and Tanenhaus, M.K. (1993) Verb-specific constraints
An influential interactive model that describes processes of in sentence processing: separating effects of lexical preference
auditory comprehension was formulated in 1980 [g]. In this from garden-paths. J. Exp. Psychol. Learn. Mem. Cogn.
19, 528–553
model, syntactic and semantic processes interact from an early
g Marslen-Wilson, W.D. and Tyler, L.K. (1980) The temporal structure of
stage during auditory language comprehension. Experiments spoken language understanding. Cognition 8, 1–71
that focus on prosodic aspects indicate that this is also true for h Warren, P. et al. (1995) Prosody, phonology, and parsing in closure
syntactic and prosodic processes [h]. ambiguities. Lang. Cogn. Processes 10, 457–486

http://tics.trends.com 1364-6613/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved. PII: S1364-6613(00)01839-8
 Review TRENDS in Cognitive Sciences Vol.6 No.2 February 2002 79

Working memory Online processes

Primary

0
Auditory cortex
acoustic analysis
ERP components

BA 44 Phonological BA 44
Phonological (superior-posterior (superior-posterior Identification Phase 0
segmentation BA 42
memory region) portion) of phonemes N100 (100 ms)
and sequencing

STG Identification
(posterior region) of word form

Memory of BA 44 Syntactic BA 44 STG Identification of Phase 1

(superior-anterior and structure (inferior portion) (anterior region)
syntactic structure
inferior portion) building ELAN ELAN word category ELAN (150 _ 200 ms)

Time
MTG Identification of
(posterior region) lemma and morpho-
logic information

BA 45/47 Semantic
Memory of Integration of
(semantic) relations Left frontal MTL semantic and morpho- Phase 2
semantic features
BA 44/45 Thematic role LAN N400 syntactic information LAN/N400 (300 _ 500 ms)
Thematic structure (morpho-syntax) assignment

ms
General memory Syntactic Fronto-central Centro-parietal Processes of Phase 3

1000
IFG reanalysis and repair
resource integration P600 P345/P600 P600 (±600 ms)

Frontal Space Temporal

TRENDS in Cognitive Sciences

Fig. 1. Neurocognitive model of auditory sentence processing. The boxes represent the functional I argue that both psycholinguistic views,
processes, the ellipses the underlying neural correlate identified either by fMRI, PET or ERPs. autonomous processing and interactive processing,
The neuroanatomical specification (indicated by text in square brackets) is based on either fMRI or
PET data. The ERP components specified in their temporal structure (left-hand side) are assigned to
hold in principle, but describe different processing
their neural correlate by the function rather than the localization of their generator. This holds true phases during language comprehension (i.e. early
for the ERP components of phase 2 and -3 as late components are hard to localize. The different versus late). The present model is, thus, compatible
distributions of the P600 and their functional nature are discussed in Ref. [53]. The neural correlate
with both syntax-first models and interactive
of the ELAN, however, has been verified by dipole localization [54]. Abbreviations: BA, Brodmann’s
area; ELAN, early left-anterior negativity; ERP, event-related brain potential; fMRI, functional models that assume late interaction, but not with
magnetic resonance imaging; IFG, inferior frontal gyrus; MTG, middle temporal gyrus; those that claim immediate or, even, predictive
MTL, middle temporal lobe; PET, positron imaging tomography; STG, superior temporal gyrus. interaction. Although interaction between prosodic
and syntactic information during auditory
controlled) and/or different time windows during sentence comprehension is considered in the
processing (early versus late). However, we are able proposed model, the temporal structure of this
to distinguish early from late processes using interaction is not yet specified.
electrophysiological techniques that register the The functional neuroanatomy of auditory
brain’s reaction to a given item millisecond-by- language comprehension is described as a bilateral
millisecond from its onset. temporo–frontal network in which the left temporal
I propose a neurocognitive model of sentence regions support processes that identify phonetic,
comprehension, the temporal parameters of which lexical and structural elements; the left frontal
are based on electrophysiological data and cortex is involved with sequencing and the formation
neurotopographical specifications on brain-imaging of structural, semantic and thematic relations; the
data. The temporal characteristics of the model right temporal region is thought to support the
consist of three phases. Phase 1 (100–300 ms) identification of prosodic parameters; and the right
represents the time window in which the initial frontal cortex is involved in the processing of
syntactic structure is formed on the basis of sentence melody. A schematic view of the processes
information about the word category. During phase 2 that occur within the left hemisphere in this model
(300–500 ms), lexical-semantic and morphosyntactic is given in Fig. 1. This figure also sketches the role of
processes take place with the goal of thematic role working memory in the process of language
assignment. During phase 3 (500–1000 ms), the comprehension (discussed briefly in Box 2).
different types of information are integrated. The model is based on empirical evidence from
Although building of the syntactic-phrase structure neurophysiological studies using event-related brain
is autonomous and precedes semantic processes in potentials (ERPs) and magnetic fields, and from
the early-time windows, these processes interact imaging studies that include PET and fMRI. In this
only in the late-time window. From this perspective, review, the neuroanatomy and the time course and

http://tics.trends.com
80 Review TRENDS in Cognitive Sciences Vol.6 No.2 February 2002

Box 2. The role of Broca’s area in sentence comprehension

Broca’s area has been anatomically defined to include BA 44 and BA 45

[a]. A classical view that Broca’s area is the locus of syntax [b–d] is
supported by brain-imaging studies showing that either BA 44 or BA
44/45 is active when comparing less complex subject-first sentences,
such as ‘The juice that the child spilled stained the rug.’, with more
complex object-first sentences, such as ‘The child spilled the juice that
stained the rug.’ [e–g]. However, because the object noun ‘juice’ in the
second sentence is not in its canonical position (subject–verb–object),
these sentences differ not only in their complexity, but also in their
working memory requirement. Aware of this, the authors claim that
syntactic processes and the required memory recourses are responsible Fig. 2. Brodmann areas (BA) in the left hemisphere. The inferior frontal
for the increase of activation in Broca’s area. gyrus (IFG) is shown in green, the superior temporal gyrus (STG) in red
and the middle temporal gyrus (MTG) in blue. (Adapted from Ref. [55].)
A recent study in German varied the factor complexity (object first
versus subject first) independently of the factor memory, that is the
distance between the object-noun phrase and its original position in the possible interplay of syntactic and semantic
structure (long versus short) [h]. Results using fMRI demonstrated that processes are specified. The neuroanatomy of
increased activation of Broca’s area (BA 44) was triggered by the factor prosodic processes and their interaction with
syntactic memory but not by complexity. This finding is compatible with syntactic processes are also discussed.
the view that Broca’s area is not the locus of syntax per se [i], but that it
supports aspects of syntactic memory. Local phrase-structure building Syntactic and semantic processes
seems to recruit the inferior tip of BA 44 and the frontal operculum, in Neuroanatomy
particular [j]. The functional neuroanatomy of speech
It should be noted that the description presented above concerns the perception prior to syntactic and semantic processes
role of Broca’s area in language comprehension. There is no doubt that has been described in detail recently by Hickok and
this area also supports language production [k,l]. Moreover, this area is Poeppel [8]. As the present review focuses on
involved in the processing of musical sequences [m], the perception of the sentence-level processes, this processing stage is not
rhythm of motion [n] and the imagery of motion [o]. A common feature of considered here.
these tasks is that they involve an aspect of sequencing, which suggests Studies on the functional neuroanatomy of
that Broca’s area supports the processing of sequences in both language semantic processes at the sentence level are rare.
and non-language domains. Rather, most imaging studies of semantic processes
References
are conducted at the word level. Such studies indicate
a Amunts, K. et al. (1999) Broca’s region re-visited: cytoarchitecture and intersubject that the left middle temporal gyrus (MTG), the
variability. J. Comp. Neurol. 412, 319–341 angular gyrus and the left inferior frontal gyrus (IFG)
b Damasio, H. and Damasio, A.R. (1989) Lesion Analysis in Neuropsychology, Oxford support semantic processes [9–12]. It is proposed that
University Press
the frontal cortex is responsible for strategic and
c Berndt, R.S. and Caramazza, A. (1980) A redefinition of the syndrome of Broca’s
aphasia: implications for neuropsychological model of language. Appl. Psycholinguist.
executive aspects of semantic processing [13–15].
1, 225–278 Studies investigating semantic processes at the
d Zurif, E.B. (1995) Brain regions of relevance to syntactic processing. In An Invitation to sentence level report a variety of activation loci,
Cognitive Science, (Vol. I, 2nd edn) (Gleitman, L. and Liberman, M., eds), pp. 381–398, including the left IFG (Brodmann area, BA 45/47)
MIT Press
[16], the right superior temporal gyrus (STG) and the
e Stromswold, K. et al. (1996) Localization of syntactic comprehension by positron
emission tomography. Brain Lang. 10, 132–144 left MTG [17], as well as the left posterior temporal
f Caplan, D. et al. (1998) Effects of syntactic structure and propositional number on region [18] (see Fig. 2). Studies that identified
patterns of regional cerebral blood flow. J. Cogn. Neurosci. 10, 541–552 activation in temporal regions used a task in which
g Just, M.A. et al. (1996) Brain activation modulated by sentence comprehension. subjects had to ‘judge whether the sentence made
Science 274, 114–116
sense’. Frontal activation of BA 45/47 was observed
h Fiebach, C.J. et al. (2001) Syntactic working memory and the establishment of filler-gap
dependencies: insights from ERPs and fMRI. J. Psycholinguist. Res. 30, 321–338 when subjects were asked to judge whether two
i Grodzinsky, Y. (2000) The neurology of syntax: language use without Broca’s area. sentences presented successively ‘meant the same’.
Behav. Brain Sci. 23, 1–21 Although both tasks require an explicit judgment,
j Friederici, A.D. et al. (2000) Auditory language comprehension: an event-related fMRI the latter can only be performed after comparing the
study on the processing of syntactic and lexical information. Brain Lang. 74, 289–300
k Damasio, H. and Damasio, A.R. (1992) Brain and language. Sci. Am. 267, 88–95
two sentences held in memory and, therefore,
l Indefrey, P. et al. (2001) A neural correlate of syntactic encoding during speech requires memory resources. Overall, the combined
production. Proc. Natl. Acad. Sci. U. S. A. 98, 5933–5936 findings indicate that semantic processes are mainly
m Maess, B. et al. (2001) Musical syntax is processed in the area of Broca: an MEG study. subserved by the left temporal region and that the
Nat. Neurosci. 4, 540–545
frontal cortex is recruited when strategic and/or
n Schubotz, R.I. and von Cramon, D.Y. (2001) Interval and ordinal properties of sequences
are associated with distinct premotor areas. Cereb. Cortex 11, 210–222
memory aspects come into play.
o Binkofsi, F. et al. (2000) Broca’s region subserves imagery of motion: a combined Studies on the functional neuroanatomy of
cytoarchitectonic and fMRI study. Hum. Brain Mapp. 11, 273–285 syntactic processes demonstrate involvement of the
inferior frontal cortex and the anterior portion of

http://tics.trends.com
Review TRENDS in Cognitive Sciences Vol.6 No.2 February 2002 81

(a) (b) (c)

–5 µV N400 –5 µV ELAN –5 µV
Cz F7 Pz

0 0 0

5 5 5
0 0.5 1 1.5s 0 0.5 1 1.5s 0 0.5 1 1.5s
P600

Das Hemd wurde bebügelt. Das Hemd wurde bebügelt.

F7 The shirt was 'ironed'.
The shirt was 'ironed'.
Cz
Das Gewitter wurde gebügelt. Pz Die Bluse wurde am gebügelt
The thunderstorm was ironed. The blouse was on ironed.

TRENDS in Cognitive Sciences

Fig. 3. The three language related components in the ERP: (a) N400, and the relevant frontal regions are left BA 44 and
(b) very early left-anterior negativity (ELAN), and (c) P600. Shown are the adjacent frontal operculum. Although a larger
average ERPs for the semantic- and syntactic-violation condition at
selected electrode sites. Solid lines represent the correct condition,
portion of BA 44 seems to support aspects of syntactic
and dotted lines the incorrect condition. working memory, the inferior tip of BA 44 and the
frontal operculum are required specifically for local
the temporal cortex. A consistent finding in studies phrase-structure building.
that compare brain activation during simple and
complex sentences (for details see Box 2) is that Time course
complex sentences are accompanied by increased Although many studies have investigated the time
activation of the left inferior frontal cortex (BA 44/45) course of syntactic and semantic processes, only a
[19–23]. These studies, however, use sentence few have investigated their direct interplay. The
materials in which the factor ‘syntactic complexity’ is electrophysiological outcome of semantic processes is
confounded with the factor ‘working memory’. A more a negative wave, the so-called N400, that peaks
recent study in which these two factors were varied about 400 ms after the word onset [31] and occurs in
independently demonstrates that activation of BA 44 response to words that cannot be integrated
is due to aspects of working memory rather than semantically into the preceding context [32].
syntactic complexity [24]. Syntactic processes are correlated with two ERP
Anterior and posterior temporal activation has components, a left-anterior negativity (LAN), which
been reported during sentence processing [16,18,22]. occurs during an early time window (between
In particular, the anterior STG (planum polare) is 100–500 ms) and a late centro-parietal positivity,
active in a number of studies. This is accompanied termed P600, which occurs between 600–1000 ms.
by either substantial activation of the inferior Within the early time window, a very Early LAN
frontal gyrus [16,18] or minimal or no activation in (ELAN) correlates with rapidly detectable word-
Broca’s area, although activation in the left frontal category errors [33–36] whereas the LAN correlates
operculum is sometimes observed [25–27]. with morphosyntactic errors [37–40]. The P600
Interestingly, the latter studies used auditory correlates with outright syntactic violations
stimuli, which suggests that there is a partial (following the ELAN), with ‘garden-path’ sentences
difference in sentence comprehension between that require syntactic revision, and with processing
auditory and reading tests. It is probable that the of syntactically complex sentences [41–44]. The three
involvement of the IFG during sentence reading different ERP components and example sentences
occurs because of the process of phonological are displayed in Fig. 3.
recoding during reading, a process that is attributed The electrophysiological data clearly support the
to the IFG on the basis of studies at the phoneme- three-phase neurocognitive model presented at the
and word level [9,28–30]. start of this review. However, additional evidence is
Thus, the combined neuroimaging data indicates needed before the claims about modular syntactic
that both semantic and syntactic processes involve processes during the early-time window (phase 1)
parts of the temporal and the inferior frontal cortex. and interactivity between semantic and syntactic
The left MTG and BA 45/47 are the relevant areas in processes during the late-time window (phase 3)
the semantic domain, although activation of BA 45/47 can be justified. This evidence is provided by
appears to depend on the amount of strategic and/or experiments in which the critical word in the
memory processes required. In the syntactic domain, sentence violates both the syntactic and semantic
the relevant temporal region is the anterior left STG constraints set by the prior context, thus leading to

http://tics.trends.com
82 Review TRENDS in Cognitive Sciences Vol.6 No.2 February 2002

(a) Syntactic violation (b) Combined violation

F7 F3
FZ
FT7 FC3
ELAN CZ ELAN
CP5
–5 FZ P3 PZ
–5 F7 F3 FZ
F7 F3

µV µV
0.5 1.0 1.5 0.5 1.0 1.5
(s) (s)
no N400
FT7 FC3 CZ Correct FT7 FC3 CZ
Incorrect

CP5 P3 PZ CP5 P3 PZ

P600 P600

Das Hemd wurde bebügelt. Das Hemd wurde bebügelt.

The shirt was 'ironed'. The shirt was 'ironed'.
Die Bluse wurde am gebügelt Das Gewitter wurde im gebügelt.
The blouse was on ironed. The thunderstorm was in ironed.
TRENDS in Cognitive Sciences

Fig. 4. Average ERPs for the syntactic violation (a) and the combined and syntactic processing. Processes of
violation (b). Solid lines represent the correct condition, and dotted identification (word category and meaning) that are
lines the incorrect condition. (Adapted from Ref. [47].)
assumed to be encoded in the mental lexicon
might be located primarily in temporal structures,
difficulties in both processing domains. When a and the construction of syntactic relations
word-category violation, usually reflected by the (structure building) and semantic relations
ELAN (phase 1), and a semantic violation, usually (categorization and selectional restriction) appear
reflected by the N400 (phase 2), are combined in one to involve the frontal cortex. Sentence
target word, only an ELAN is observed [45,46] comprehension consists of three functionally
(see Fig. 4). The absence of an N400 in double- distinct phases: an initial parsing phase (phase 1),
violation conditions possibly occur because a target which precedes processes of thematic assignment
word that is not licensed by the syntax is not lexically based on semantic and morphosyntactic information
integrated. This finding indicates that syntactic- (phase 2), and a late phase of revision during which
structure building precedes semantic processes. interaction between semantic and syntactic
When combining a morphosyntactic violation information might take place (phase 3).
(e.g. syntactic gender) usually reflected by the LAN So far I have discussed the processing of
(phase 2) and a semantic anomaly usually reflected semantic and syntactic information contained in
by the N400 (phase 2), both ERP components are sentences presented visually and auditorily.
present and independent of one another. In this However, prosodic information encoded in the
condition, the amplitude of the P600 varies as a auditory presentation mode is an additional,
function of both the semantic and syntactic factors, relevant parameter.
thus suggesting an interaction between these factors
in the late-time window (phase 3) [39]. Prosodic processes
There are some reports of experiments using The functional neuroanatomy of prosodic processes
ERPs that do not show ELAN effects in response to has been specified in recent studies using PET
syntactic violations. We have demonstrated [47,48] and fMRI. At the segmental level, pitch
that most of these studies used sentence material that discrimination in speech syllables correlates
did not contain outright syntactic violations, rather with an increased activation in the right prefrontal
they contained correct, although unusual, structures. cortex [49]. Violations of pitch for lexical elements
The finding that correct but unusual structures evoke in a tonal language, such as Thai, results in
only a P600, and not an ELAN, is expected on the modulation of activation in the left frontal
basis of the present model and suggests that the brain operculum adjacent to Broca’s area [50]. Processes
reacts in accordance with the grammar. at the suprasegmental level, in which pitch
In summary, different subparts of the left modulations act as syntactic markers appear,
temporal- and frontal cortices subserve semantic instead, to involve the right hemisphere. A recent

http://tics.trends.com
 Review TRENDS in Cognitive Sciences Vol.6 No.2 February 2002 83

Delexicalized vs normal speech

LH RH

Delexicalized speech Z < 3.1 8

Normal speech Z > -3.1 -11

TRENDS in Cognitive Sciences

Fig. 5. Aspects of prosodic processing apparent from functional and the possible attempt to revise the initial
magnetic resonance imaging data. Functional brain activation in structure (P600) built on the available syntactic and
different subjects was averaged and superimposed onto a
white-matter segmented, normalized anatomical volume.
prosodic information. These ERP findings indicate
Comparing normal speech with delexicalized speech (filtered normal that both types of information interact but that they
sentence that leaves the FO contour intact but filters out all lexical are mute with respect to the temporal structure
information) reveals that left perisylvian areas are strongly involved
because the measure was taken words after the
in processing grammatical information whereas right perisylvian
areas subserve the processing of slow prosodic modulations in misguiding prosodic information. Overall, although
spoken sentences. limited, the data available indicate that a
temporo–frontal network that is predominantly
fMRI experiment that systematically varied the within the right hemisphere supports prosodic
presence of pitch information (normal intonation processes and that prosodic information can
versus synthesized, flattened intonation) and of influence syntactic processes.
syntactic information (normal speech versus
synthesized, delexicalized speech) at the sentential Conclusion
level identified modulations in activity of the right In summary, I have argued that a bilateral
peri-sylvian cortex. In particular, the right superior temporo–frontal network subserves auditory-
temporal region and the fronto-opercular cortex sentence comprehension. Although syntactic and
were identified as regions that support the semantic information are processed predominately
processing of suprasegmental information [51] by the left hemisphere, processing of prosodic
(see Fig. 5). information occurs predominantly in the right
Although the available neuroanatomical data are hemisphere. Temporal regions support
suggestive, the temporal structure of the processing identification processes, with syntactic processes
of prosodic information with respect to other involving the left anterior STG, semantic processes
information types is still an empirical issue. There recruiting the left MTG and prosodic processes
is electrophysiological evidence that prosodic involving the right posterior STG. Frontal regions,
information interacts with syntactic information at by contrast, support the formation of relationships,
some point [52], although the time course of this with syntactic relationships involving BA 44 and
interaction is not yet specified. This evidence stems the frontal opercular cortex, and semantic
from an ERP experiment conducted in German, in relationships recruiting BA 45/47. These different
which syntactic and prosodic phrasing either did or areas within the network must be activated and
did not match [52]. In the prosodic-mismatch coordinated to achieve auditory sentence
condition, a prosodic-phrase boundary (which comprehension. The timing of the syntactic
indicates a transitive-verb structure) present two processes of structure building precedes, and are
Acknowledgements words before a transitive verb caused problems for initially independent of, semantic processes,
This study was supported listeners in integrating the verb into the prior although both interact during a later processing
by the Leibniz Science
context. These problems were evidenced in a phase. Prosodic processes influence syntactic
Prize awarded to A.D.F.
by the Deutsche biphasic N400–P600 pattern, reflecting the processes, however, the exact timing of this is a
Forschungsgemeinschaft. difficulty of lexical-semantic integration (N400) subject for future research.

http://tics.trends.com
84 Review TRENDS in Cognitive Sciences Vol.6 No.2 February 2002

References 23 Inui, T. et al. (1998) A functional MRI analysis of 43 Hagoort, P. et al. (1993) The syntactic positive
1 Frazier, L. (1987) Theories of sentence processing. comprehension processes of Japanese sentences. shift as an ERP measure of syntactic processing.
In Modularity in Knowledge Representation and NeuroReport 9, 3325–3328 Lang. Cogn. Processes 8, 439–483
Natural-language Processing (Garfield, J., ed.), 24 Fiebach, C.J. et al. (2001) Syntactic working 44 Kaan, E. et al. (2000) The P600 as an index of
pp. 291–307, MIT Press memory and the establishment of filler-gap syntactic integration difficulty. Lang. Cogn.
2 Frazier, L. and Fodor, J.D. (1978) The sausage dependencies: insights from ERPs and fMRI. Processes 15, 159–201
machine: a new two-stage model of the parser. J. Psycholinguist. Res. 30, 321–338 45 Friederici, A.D. et al. (1999) Lexical integration:
Cogniton 6, 291–325 25 Müller, R-A. et al. (1997) Receptive and expressive sequential effects of syntactic and semantic
3 Marslen-Wilson, W.D. and Tyler, L.K. (1980) language activations for sentences: a PET study. information. Mem. Cogn. 27, 438–453
The temporal structure of spoken language NeuroReport 8, 3767–3770 46 Hahne, A. and Friederici, A.D. (1998)
understanding. Cognition 8, 1–71 26 Friederici, A.D. et al. (2000) Auditory language ERP evidence for autonomous first-pass parsing
4 McClelland, J.L. et al. (1989) An interaction comprehension: an event-related fMRI study on processes in auditory language comprehension.
model of context effects in letter perception: the processing of syntactic and lexical J. Cogn. Neurosci. (Suppl.) S125
Part I. An account of basic findings. Lang. Cogn. information. Brain Lang. 74, 289–300 47 Hahne, A. and Friederici, A.D. Differential task
Processes 4, 287–336 27 Meyer, M. et al. (2000) Neurocognition of auditory effects on semantic and syntactic processes as
5 Warren, P. et al. (1995) Prosody, phonology, and sentence comprehension: event-related fMRI revealed by ERPs. Cogn. Brain Res. (in press)
parsing in closure ambiguities. Lang. Cogn. reveals sensitivity to syntactic violations and task 48 Friederici, A.D. The neural basis of sentence
Processes 10, 457–486 demands. Cogn. Brain Res. 9, 19–33 processing: a neurocognitive model. In Towards
6 Kjelgaard, M.M. and Speer, S.R. (1999) Prosodic 28 Zatorre, R.J. et al. (1996) PET studies of phonetic a New Functional Anatomy of Language.
facilitation and interference in the resolution of processing of speech: review, replication, and Cognition (Special Issue) (Hickok, G. and
temporary syntactic closure ambiguity. J. Mem. reanalysis. In Cortical Imaging-Microscope of the Poeppel, D., eds) (in press)
Lang. 40, 153–194 Mind (Raichle, M. and Goldman-Rakic, P.S., eds), 49 Zatorre, R.J. et al. (1999) Auditory attention to
7 Stirling, L. and Wales, R. (1996) Does prosody (Special Issue) Cereb. Cortex 6, 21–30 space and frequency activates similar cerebral
support or direct sentence processing? 29 Poldrack, R.A. et al. (1999) Functional specialization systems. NeuroImage 10, 544–554
Lang. Cogn. Processes 11, 193–212 for semantic and phonological processing in the left 50 Gandour, J. et al. (2000) A crosslinguistic PET study
8 Hickok, G. and Poeppel, D. (2000) Towards a inferior prefrontal cortex. NeuroImage 10, 15–35 of tone perception. J. Cogn. Neurosci. 12, 207–222
functional neuroanatomy of speech perception. 30 Pugh, K.R. et al. (1997) Predicting reading 51 Meyer, M. et al. Towards the cerebral substrates of
Trends Cogn. Sci. 4, 131–138 performance from neuroimaging profiles: sentence-level syntactic and prosodic processing.
9 Demonét, J-F. et al. (1992) The anatomy of the cerebral basis of phonological effects in Speech Commun. (in press)
phonological and semantic processing in normal printed word recognition. J. Exp. Psychol. Hum. 52 Steinhauer, K. et al. (1999) Brain potentials
subjects. Brain 115, 1753–1768 Percept. Perform. 23, 299–318 indicate immediate use of prosodic cues in natural
10 Price, C.J. et al. (1997) Segregating semantic from 31 Kutas, M. and Federmeier, K.D. (2000) speech processing. Nat. Neurosci. 2, 191–196
phonological processes during reading. J. Cogn. Electrophysiology reveals semantic memory use 53 Friederici, A.D. et al. Distinct neurophysiological
Neurosci. 9, 727–733 in language comprehension. Trends Cogn. Sci. patterns reflecting aspects of syntactic complexity
11 Wise, R. et al. (1991) Distribution of cortical 4, 463–470 and syntactic repair. J. Psycholinguist. Res.
neural networks involved in word comprehension 32 Kutas, M. and Hillyard, S.A. (1984) Brain (in press)
and word retrieval. Brain 114, 1803–1817 potentials during reading reflect word expectancy 54 Friederici, A.D. et al. (2000) Localization of early
12 Vandenberghe, R. et al. (1996) Functional and semantic association. Nature 307, 161–163 syntactic processes in frontal and temporal
anatomy of a common semantic system for words 33 Neville, H.J. et al. (1991) Syntactically based cortical areas: a magnetoencephalographic study.
and pictures. Nature 383, 254–256 sentence processing classes: evidence from Hum. Brain Mapp. 11, 1–11
13 Fiez, J.A. (1997) Phonology, semantics, and the event-related brain potentials. J. Cogn. Neurosci. 55 Duvernoy, H.M. (1991) The Human Brain,
role of the left inferior prefrontal cortex. 3, 151–165 Springer-Verlag
Hum. Brain Mapp. 5, 79–83 34 Friederici, A.D. et al. (1993) Event-related brain
14 Thompson-Schill, S.L. et al. (1997) Role of left potentials during natural speech processing:
inferior prefrontal cortex in retrieval of semantic effects of semantic, morphological and syntactic
knowledge: a re-evaluation. Proc. Natl. Acad. Sci. violations. Cogn. Brain Res. 1, 183–192
U. S. A. 94, 14792–14797 35 Hahne, A. and Friederici, A.D. (1999)
15 Poldrack, R.A. et al. (1999) Functional specialization Electrophysiological evidence for two steps in
Manage your references
for semantic and phonological processing in the left syntactic analysis: early automatic and late using BMN Reviews
inferior prefrontal cortex. Neuro-Image 10, 15–35 controlled processes. J. Cogn. Neurosci. 11, 194–205
16 Dapretto, M. and Bookheimer, S.Y. (1999) Form 36 Hahne, A. and Jescheniak, J.D. (2001) What’s left Did you know that you can now
and content: dissociating syntax and semantics in if the Jabberwock gets the semantics? An ERP download selected search results
sentence comprehension. Neuron 24, 427–432 investigation into semantic and syntactic from BioMedNet Reviews directly into
17 Kuperberg, G.R. et al. (2000) Common and distinct processes during auditory sentence
your chosen reference managing
neural substrates for pragmatic, semantic, and comprehension. Cogn. Brain Res. 11, 199–212
syntactic processing of spoken sentences: an fMRI 37 Osterhout, L. and Mobley, L.A. (1995) software? After performing a search,
study. J. Cogn. Neurosci. 12, 321–341 Event-related brain potentials elicited by failure simply click to select the articles you
18 Ni, W. et al. (2000) An event-related neuroimaging to agree. J. Mem. Lang. 34, 739–773 want, and choose the format required.
study distinguishing form and content in sentence 38 Gunter, T.C. et al. (1997) When syntax meets The bibliographic details, abstract
processing. J. Cogn. Neurosci. 12, 120–133 semantics. Psychophysiology 34, 660–676 and the link to the full-text article will
19 Stromswold, K. et al. (1996) Localization of 39 Gunter, T.C. et al. (2000) Syntactic gender and
then download into your desktop
syntactic comprehension by positron emission semantic expectancy: ERPs reveal early autonomy
tomography. Brain Lang. 10, 132–144 and late interaction. J. Cogn. Neurosci. 12, 556–568
reference manager database.
20 Caplan, D. et al. (1998) Effects of syntactic 40 Coulson, S. et al. (1998) Expect the unexpected: BioMedNet Reviews is available on an
structure and propositional number on patterns of event-related brain responses of morphosyntactic
institute-wide subscription. If you do
regional cerebral blood flow. J. Cogn. Neurosci. violations. Lang. Cogn. Processes 13, 21–58
10, 541–552 41 Osterhout, L. and Holcomb, P.J. (1992) not have access to the full-text articles
21 Caplan, D. et al. (2000) Activation of Broca’s area Event-related potentials and syntactic anomaly. in BioMedNet Reviews, ask your
by syntactic processing under conditions of J. Mem. Lang. 31, 785–804 librarian to contact us at
concurrent articulation. Hum. Brain Mapp. 42 Osterhout, L. et al. (1994) Brain potentials [email protected]
9, 65–71 elicited by garden-path sentences: evidence of the
22 Just, M.A. et al. (1996) Brain activation modulated application of verb information during parsing.
by sentence comprehension. Science 274, 114–116 J. Exp. Psychol. Learn. Mem. Cogn. 20, 786–803

http://tics.trends.com