Brain (2000), 123, 2338–2349
Cortical activation patterns of affective speech
processing depend on concurrent demands on the
subvocal rehearsal system
A DC-potential study
Hans Pihan,1 Eckart Altenmüller,2 Ingo Hertrich1 and Hermann Ackermann1
1Department
of Neurology, University of Tübingen and
of Music Physiology and Performing Arts
Medicine, Academy of Music and Drama, Hannover,
Germany
2Institute
Correspondence to: Eckart Altenmüller, Institute of Music
Physiology and Performing Arts Medicine, Academy of
Music and Drama Hannover, Plathnerstrasse 35,
D-30175 Hannover, Germany
E-mail: [email protected]
Summary
In order to delineate brain regions specifically involved
in the processing of affective components of spoken
language (affective or emotive prosody), we conducted two
event-related potential experiments. Cortical activation
patterns were assessed by recordings of direct current
components of the EEG signal from the scalp. Righthanded subjects discriminated pairs of declarative
sentences with either happy, sad or neutral intonation.
Each stimulus pair was derived from two identical original
utterances that, due to digital signal manipulations,
slightly differed in fundamental frequency (F0) range or
in duration of stressed syllables. In the first experiment,
subjects were asked: (i) to denote the original emotional
category of each sentence pair and (ii) to decide which of
the two items displayed stronger emotional expressiveness.
Participants in the second experiment were asked to
repeat the utterances using inner speech during stimulus
presentation in addition to the discrimination task. In the
absence of inner speech, a predominant activation of right
frontal regions was observed, irrespective of emotional
category. In the second experiment, a bilateral activation
with left frontal preponderance emerged from
discrimination during additional performance of inner
speech. Compared with the first experiment, a new pattern
of acoustic signal processing arose. A relative decrease of
brain activity during processing of F0 stimulus variants
was observed together with increased activation during
discrimination of duration-manipulated sentence pairs.
Analysis of behavioural data revealed no significant
differences in evaluation of expressiveness between the
two experiments. We conclude that the topographical
shift of cortical activity originates from left hemisphere
(LH) mechanisms of speech processing that centre around
the subvocal rehearsal system as an articulatory control
component of the phonological loop. A strong coupling
of acoustic input and (planned) verbal output channel in
the LH is initiated by subvocal articulatory activity
like inner speech. These neural networks may provide
interpretations of verbal acoustic signals in terms of
motor programs and facilitate continuous control of
speech output by comparing the signal produced with
that intended. Most likely, information on motor aspects
of suprasegmental signal characteristics contributes to
the evaluation of affective components of spoken language.
In consequence, the right hemisphere (RH) holds a merely
relative dominance, both for processing of F0 and for
evaluation of emotional significance of sensory input.
Psychophysically, an important determinant on expression
of lateralization patterns seems to be given by the degree
of communicative demands such as solely perceptive (RH)
or perceptive and verbal-expressive (RH and LH).
Keywords: emotion; prosody; subvocal rehearsal system; event-related potential; hemispheric specialization
Abbreviations: AP ⫽ anterior, posterior; DC ⫽ direct current; EMO ⫽ emotional category; F0 ⫽ fundamental frequency;
LH ⫽ left hemisphere; MANIP ⫽ acoustic parameter manipulation; RH ⫽ right hemisphere
© Oxford University Press 2000
Prosodic competence of left hemisphere
2339
Introduction
Human behaviour in general, and speech communication
in particular, is essentially stimulated by motivational and
emotional states. Information about these psychophysical
preconditions can be perceived from the speech signal, either
explicitly from a word’s meaning ‘I am excited (happy, angry
. . .)’ or implicitly from the speaker’s tone of voice. Usually,
the latter speech component is referred to as ‘affective
prosody’ (Ackermann et al., 1993). The fact that we easily
succeed in decoding the semantic message, i.e. the meaning
of words and sentences as well as the intentional or emotional,
which is delivered by prosodic cues, points at two basic
conditions constituting the conceptual frame of this paper.
First, a perceived emotional state must have its correlate
in the acoustic of the speech signal. However, since our
impression is sometimes uncertain or wrong, the acoustic
cues may fail exhaustive specification of affective states. As
meaningful percepts result from neural activity that goes
beyond the mere processing of physical stimulus
characteristics, acoustic parameters are apt not to define,
but rather to describe prosodic contents in terms of an
approximation.
Secondly, two distinct messages, e.g. the semantic and the
emotional, are presumably represented by spatially separated
neuronal networks. According to lesion studies and functional
imaging experiments, circumscribed brain areas [predominantly in left hemisphere (LH) regions] could be determined
as prerequisites for perceptual and expressive language
functions, especially of phonological, semantic and syntactic
aspects (Posner and Raichle, 1994; Springer and Deutsch,
1995). However, no consensus has been achieved so far in
localizing the processing of intonational features subserving
affective and sociolinguistic impressions in the right
hemisphere (RH) or LH (for review, see Pell and Baum,
1997a).
In correspondence to vocal linguistic communication, the
perception of a speaker’s emotional tone of voice is based
on the decoding of various acoustic parameters. Fundamental
frequency (F0) represents the closing frequency of the vocal
cords during phonation. It determines largely its main
psychophysical correlate, i.e. the speaker’s pitch, which
presumably holds transcultural pertinence to express emotion
(Ohala, 1984) and which is also of linguistic relevance, e.g.
to realize a stress pattern or to indicate a question. The extent
of F0 variation (F0 variability, F0 range) constitutes a
variability measure of F0. Its expression in extreme either
reflects a melodic speech with vivid changes of pitch or a
monotonous way of speaking that lacks melodic elements.
Simple temporal parameters like syllable duration are
difficult to interpret, since vocal emotional expressions
inconsistently vary not only voiced but also unvoiced
elements and are also influenced by speech pauses (Bergmann
et al., 1988). The variation of vowel length seems to hold a
higher degree of explanatory value. In parallel with F0-range,
it indicates dynamic properties of a voice either typifying
happiness (high variation) or amplifying the impression of
melancholy and sadness (low variation) (Tischer, 1993).
The acoustic parameters selected for systematic variation
in this study, i.e. F0-range and duration of stressed vowels,
were chosen because they yielded consistent effects on
speaker state attribution. Bergmann and colleagues varied
different acoustic parameters in short speech utterances and
had them judged according to their affective and attitudinal
expressiveness (Bergmann et al., 1988). In agreement with
Ladd and colleagues (Ladd et al., 1985), they found that F0range and temporal parameters, especially duration of stressed
segments, independently modulate expressiveness as
continuous variables. Strong attribution effects were found
for F0-range (broad ⫽ high arousal; narrow ⫽ sadness) and
temporally modulated stressed vowels (short ⫽ joyful; long ⫽
sadness).
The neurobiological basis subserving the decoding of
prosodic cues is far from being identified. Patients with
damage to the RH have been reported to be more impaired
than subjects with left-sided lesions on tasks requiring the
discrimination and/or identification of emotional tone (Tucker
et al., 1977; Weintraub et al., 1981; Heilman et al., 1984).
Other studies, however, found similar performance of patients
with unilateral damage to either hemisphere (Schlanger et al.,
1976; Van Lancker and Sidtis, 1992; Pell and Baum, 1997a).
Thus, lesion studies have not unequivocally clarified the
locus of neuronal networks primarily involved in affective
speech processing (for a comprehensive review, see Gainotti,
1991; Heilman et al., 1993). Possibly, cerebral mechanisms
underlying prosodic perception are bilaterally located, as
concluded by Pell and Baum (Pell and Baum, 1997b). Aside
from the problem of localizing specified areas subserving the
construction of emotional percepts, it is unclear to what
degree the representation as such is affected when perceptual
deficits are detectable. Blonder and colleagues suggested a
high level disruption of affective representations (Blonder
et al., 1991); also, impairments of low level mechanisms
such as the processing of one of the above-mentioned acoustic
parameters remain conceivable.
More psychophysically oriented studies yielded some
evidence that the RH has an advantage over the LH in
extracting pitch information from complex auditory stimuli.
Patients with RH lesions were found to make no use of F0variability, but rather relied on duration cues to assess
affective prosody (Van Lancker and Sidtis, 1992). Robin
and colleagues observed that right temporoparietal lesions
disrupted the discrimination of tones, but not the perception
of time patterns, while lesions in the homologous regions of
the LH had the opposite effects (Robin et al., 1990). Possibly,
the expected contribution of the RH to the processing of
affective speech prosody reflects the discrimination of pitch
contours rather than the evaluation of emotional significance
of verbal utterances.
Studies in healthy subjects represent a further approach to
2340
H. Pihan et al.
investigate emotion processing. For example, a PET study
reported by George and colleagues examined the topography
of cerebral blood flow change while subjects identified
either the affective tone, the emotional propositional content
(sentence meaning) or the second word in acoustically
presented sentence utterances (George et al., 1996). The
judgement on the emotional propositional content correlated
with a bilateral prefrontal activation, while the discrimination
of affective intonation evoked a lateralized right frontal
response. Considering the aforementioned lesion studies, one
would have expected, in addition, significant activation of
temporoparietal fields. Their absence may partly be due to
limitations inherent to the PET approach. Its low temporal
resolution amounting to 1 min or longer leads to integration
of multiple cognitive operations which presumably correspond to widespread cortical activation. With respect to
intersubject averaging, effects of distributed neural networks
tend to be blurred (Chertkow and Murtha, 1997).
The recording of evoked potentials provides an alternative
in this regard as it allows assessment of cortical activation
with a high temporal resolution. The present investigation
recorded direct current (DC) components of the EEG signal
at the scalp. This method has been proven valuable for the
identification of the language-dominant hemisphere, e.g. the
search for synonyms yielded a localized maximum of cortical
activation over left dorsolateral prefrontal areas (Altenmüller
et al., 1993). Investigations on melody perception have
shown a predominant involvement of right frontal and right
central fields, especially in musically untrained subjects
(Altenmüller, 1986).
In our preceding study on affective speech processing, the
same areas were found to be strongly activated. Left/right
comparison of hemisphere activation revealed a highly
significant lateralization towards the RH. This effect appeared
largely independent of emotional stimulus category (Pihan
et al., 1997). Post hoc evaluation of subjects’ strategies to
resolve the perception task revealed that one male test subject
repeated the stimuli in parallel with the ongoing presentation
using inner speech. His cortical activation pattern was
balanced between the LH and RH, interrupting the clear RH
effects observed in the others. Although it is known that
lateralization effects can be masked if linguistic operations
are concurrently performed with non-linguistic tasks (e.g.
after raise of numbers of response categories, demonstrated
by Tompkins and Flowers, 1985), the neural mechanism
underlying this effect remains to be clarified. Possibly,
the conflict of lesion studies on the issue of hemispheric
specialization in affective speech processing is a problem
related to variable LH and RH involvement due to acoustic,
biological or cognitive determinants that are not sufficiently
controllable or even yet considered. Also, understanding
cerebral processing of linguistic and affective components of
speech perception might depend critically on the degree to
which we succeed in evaluating and separating linguistic
task demands and strategies from non-linguistic ones.
By measuring cortical activation during discrimination of
Table 1 Test sentences of the present study
Sie wollte seh’n, was das Leben im Süden bieten kann.
(She wanted to see what life in the South could give her.)
Sie gab nach, und verflogen war der ganze Unmut.
(She gave in and all annoyance vanished.)
Er kam spät am Abend und ging früh am Morgen.
(He came late at night and left early in the morning.)
Er sprach mit langen Sätzen, und niemand hörte zu.
(He spoke long winded sentences and nobody listened.)
Vowels of accented syllables are in bold-italic letters (English
translation in parenthesis).
emotional expressiveness that is digitally altered by temporal
and spectral manipulation of natural speech, the present study
aims to elaborate further LH and RH mechanisms of nonverbal acoustic communication.
Material and methods
Experimental procedure
A non-invasive EEG technique was applied to investigate
cortical activation during discrimination of affective
intonations of spoken language. DC (slow) components of
the EEG signal can be recorded from the surface of the scalp
provided that mental activity extends over a time interval
of several seconds. They presumably represent sustained
excitatory input to apical dendrites of cortical pyramidal cells
(Rockstroh et al., 1989). Local distributions of these surfacenegative low frequency DC potentials constitute cortical
activation patterns. Since DC potentials are lower in voltage
than the ongoing background EEG, the signal to noise ratio
has to be enhanced by averaging task-related EEG activity
over several trials. Activation patterns obtained with this
method are highly task-specific and intra-individually
reproducible (for neurophysiological details of the method,
see Altenmüller and Gerloff, 1999).
Mental activity is maintained over seconds when changes
of a speaker’s emotional state are perceived from his or her
voice. This takes place during verbal communication, e.g.
while the listener perceives spoken language. In this study,
stimuli were presented as sequences of two successive
sentences, each of them with identical wording. The paired
stimuli represented variants of the same emotional category
differing in expressiveness. Perceptual effects were created
by systematic variation of single acoustic parameters.
Subjects were asked, first, to recognize the emotional
category, i.e. neutral or sad or happy, and secondly, to indicate
whether the first or the second stimulus of the pair sounded
happier or sadder or more excited (in instances of neutral
sentences). In a second task, subjects were asked to repeat
simultaneously the wording of the presented utterances using
inner speech. Subjects were requested to avoid any phonation
or articulation during this task as electric fields from muscular
activity and movement effects represent major sources of
artefacts in DC recordings. Four stimulus pairs were chosen
to help subjects to become acquainted with the task demands
Prosodic competence of left hemisphere
2341
prior to the start of recordings. All subjects gave informed
consent to participate in the study which was approved by
the ethics committee of the University of Tübingen.
Stimulus generation
Four different declarative sentences were digitally recorded
by a professional actress, each with neutral, happy and
sad intonation. The actress was instructed to stress four
predetermined syllables and to avoid any inconsistencies with
respect to sentence focus and contrastive stress patterns
(Table 1). Propositional content of the resulting 12 utterances
was compatible with each of the three prosodic modulations
considered. All sentences were digitally adapted to equal
levels of perceived loudness prior to the following
manipulations.
By means of commercially available software (LPC
Parameter Manipulation/Synthesis Program of the
Computerized Speech Lab CSL 4300, Kay Elemetrics, New
York, USA), three stimuli differing in F0-range were
resynthesized from each of the 12 spoken sentences. In line
with psychoacoustic standards, the pitch contours of the
original utterances were extended or reduced in relation to
the sentence-final F0 level, which was kept unchanged (Ladd
et al., 1985). Figure 1A illustrates the procedure and presents
the average of each maximum F0 range and its standard
deviation for all resynthesized stimuli.
A copy of the original utterances was used to create a
second set of test sentences in which emotional expressiveness
was altered by varying vowel duration of stressed syllables.
From each original sentence, three variants with either short,
middle or long duration of stressed syllables were obtained
by either cutting out or doubling single pitch periods of the
acoustic signal. On average, the shortened and lengthened
variants had durations of 80, 110 or 125% of the original
vowels, respectively. The rhythm of each sentence was largely
kept constant as manipulation of vowel length was performed
in proportion to its absolute duration. Figure 1B exemplifies
the time-manipulation and its effect on total stimulus duration.
Altogether, a corpus of 72 sentences (12 utterances
produced by an actress ⫻ 3 F0 variants plus 3 ⫻ 12 durational
variants) was created. The naturalness of all the sound was
verified by two certified speech pathologists as well as
the authors. At perceptual evaluation, the pitch- and timemanipulated variants of happy and sad utterances differed in
the degree of perceived intensity of the respective emotion.
In contrast, sentences with neutral intonation sounded more
or less excited. Forty-eight pairs of test sentences were
assembled with the following characteristics. The two paired
utterances were always derived from the same original
recording and, therefore, had identical wording and belonged
to the same emotional category. At the acoustic level they
either differed in F0 range or in duration of stressed syllables.
Half of the stimulus pairs were characterized by a maximal
difference in the respective acoustic parameter, i.e. smallest
versus largest F0 range or shortest versus longest vowel
Fig. 1 (A) Acoustic signal (upper panel) and three synthetic pitch
contours (lower panel) of the test sentence ‘Sie gab nach, und
verflogen war der ganze Unmut’, produced with neutral
intonation. Voiced signal portions are indicated by bars below the
acoustic signal, numbers indicate peak and end-point frequencies
in Hz. In the lower panel, F0 range of the medial pitch contour is
slightly increased compared with the original utterance. All test
sentences in Table 1, each spoken with happy, sad and neutral
intonation, were subjected to F0 range manipulation in the same
manner. Averages of F0 range (highest minus lowest value) over
four test sentences of each emotional and synthetic category are
listed below (standard deviation in parenthesis). (B) Durational
variants of the vowel /o/ of the stressed syllable in ‘verflogen’.
In the upper acoustic signal, vowel duration has been reduced to
143 ms by cutting out single pitch periods; the middle and lower
signals were lengthened to 195 ms and 232 ms, respectively, by
doubling single pitch periods. Other accented vowels of each test
sentence in Table 1 were shortened/lengthened accordingly giving
rise to 4 ⫻ 3 durational variants (short, medial or long vowel
duration) within each emotional category. Averages of total
stimulus duration are listed below (standard deviation in
parenthesis).
duration, the other half by the contrast, minimal versus
medium F0 range or minimal versus medium vowel
duration.
2342
H. Pihan et al.
Fig. 2 Time course of a single discrimination trial displaying
presentation periods and time-correlated DC-potentials at
electrode position FC4 (grand average across all trials of a single
test person). Sentence durations are indicated by horizontal bars;
the light parts at the right ends indicate the range of durational
variability between stimuli.
Participants and procedure
Altogether, 24 right-handed healthy students and staff (10
males, 14 females; age 19–34 years) from the University of
Tübingen participated in the present study. In the first
experiment, 16 subjects (8 males, 8 females) discriminated
the stimuli without additional demands. In order to evaluate
the role of inner speech on affective prosody perception, we
conducted a second experiment in which eight test persons
(2 males, 6 females) performed inner speech in addition to
the discrimination task. Handedness was assessed by means
of the Edinburgh Inventory (Oldfield, 1971). None of the
participants was familiar with the aims of the research work.
Forty-eight stimulus pairs were presented in a randomized
order and repeated with the sequence reversed. A second run
was performed in the same way (one individual time course
is displayed in Fig. 2). At the end of each trial the subjects’
forced choice was to specify the emotional category of each
pair (happy, sad or neutral) and indicate the sentence (first
or second) which showed stronger expression of the perceived
emotion (happier, sadder) or speaker’s arousal (more excited).
An answer was considered ‘correct’ if the sentence with
broader F0 range or shorter vowel duration was recognized
as ‘more expressive’. In sadly intonated pairs the utterance
with longer syllable duration was expected to be labelled
as ‘sadder’.
Recording procedure
DC potentials were recorded from the scalp by 26 nonpolarizable AgCl electrodes (impedance below 10 kΩ).
Stabilization of the electrode potential and reduction of the
skin impedance was ascertained by applying the method for
high quality DC recordings described by Bauer (Bauer et al.,
1989). Subjects were instructed and trained not to perform
any muscular movement during the recording period and
especially to avoid articulatory gestures in the second
experiment (inner speech condition). Signals were amplified
and digitized using Neuroscan soft- and hardware (NeuroScan
Inc., Herndon, Va., USA). Electrode positioning was
conducted according to the Jasper 10/20 System (Jasper,
1958) using unipolar leads with linked mastoid electrodes as
a reference. Simultaneously, electro-oculogram was assessed
through an additional diagonal bipolar recording between
nasion and the anterior zygomatic arch. For further data
processing, only trials without artefacts were accepted.
Generally, between 10 and 20 trials per condition and subject
were averaged and analysed relating the DC amplitudes
during the presentation periods to a baseline taken from a
1.5 s prestimulus period. As shown in Fig. 2, activation of
the second analysis interval was considerably higher
compared with the first. We specifically tried to elaborate
the interaction between the inner speech condition and
comparative processing of sentence pairs. Therefore, mean
amplitudes within analysis period 2 provided the basis for
data normalization and statistical evaluation. Analysis period
1 was not considered as we could not exclude influences
from a varying onset of inner speech performance and of
phonological encoding as wording was not known before
presentation start. Evoked potentials within the first second
of stimulus presentation do partly reflect unspecific activation
and orientation responses and, therefore, were not analysed
either.
Prior to statistical analysis the data were normalized. All
values of a given subject were divided by his/her most
negative mean value of a single analysis period and scaled
such that the minimum value was ‘–1’. A repeated measures
ANOVA (analysis of variance) was performed, in which the
factor group (Experiment 1, no inner speech; Experiment 2,
with inner speech) was entered as a between-subjects factor
since different subjects participated in the two experiments.
Emotional category (EMO: happy, sad, neutral) and acoustic
parameter manipulation (MANIP: F0, time) were taken as
within-subject factors. Recorded DC potentials at 26 electrode
positions constituted the dependent variables and were used
to create the maps shown in Fig. 4. In order to evaluate
topographic differences in cortical activation, mean values
of the following electrodes were grouped as follows:
F7, F7F3, F3, FC3: anterior left (AL)
F8, F8F4, F4, FC4: anterior right (AR)
T3, T5, PT3, P3: posterior left (PL)
T4, T6, PT4, P4: posterior right (PR)
The grouping categories were considered in the ANOVA by
introducing two additional within-subject factors: side (left,
right) and AP (anterior, posterior).
Results
Evaluation of response behaviour showed that subjects of
the first experiment correctly identified the intended affective
prosody of the original utterances (neutral, sad, happy) (Fig.
3, left panel, dark grey boxes). Few sentences with happy or
Prosodic competence of left hemisphere
2343
Fig. 3 Percentage of ‘correct’ answers across subjects obtained during Experiment 1 (spontaneous discrimination, n ⫽ 16, dark grey
boxes) and Experiment 2 (discrimination plus inner speech, n ⫽ 8, light grey boxes). The left panel refers to identification of ‘emotional
category’, the right panel to discrimination of ‘expressiveness’. F0-varied stimulus conditions are indicated by the suffix ‘-P’, durationmanipulated conditions are marked by ‘-T’.
Table 2A Significant effects and interactions in the
repeated measures ANOVAs performed on ranked
behavioural data
Effect
d.f.
F
P
Emotional category
Group
1.22
3.63
0.0700
Dimensional evaluation
Group
EMO
EMO ⫻ MANIP
1.22
2.44
2.44
0.30
15.37
11.76
0.5931
0.0001
0.0001
sad tone were categorized as instances of neutral intonation.
Discrimination of expressiveness within a given emotional
category yielded less consistent results, with the highest
percentage of expected answers in the group of pitch-varied
happy and time-varied neutral sentences (Fig. 3, right panel,
dark grey boxes). Few subjects rated expressiveness at
random or contrary to the predictions made. In the second
experiment, additional linguistic demands imposed on
subjects by the inner speech task did not disrupt their ability
to discriminate the sentence pairs (Fig. 3, light grey boxes).
Rating results of emotional categorization and dimensional
evaluation were ranked and subjected to an ANOVA taking
emotions and acoustic parameter manipulations as repeated
measures. No significant differences between the two
experiments were observed with respect to discrimination
of expressiveness. However, a tendency emerged towards
reduced performance in denoting the correct emotion (Table
2A, Group, upper panel). In addition, evaluation of
expressiveness yielded a significant main effect for emotion
and significant interaction of emotion and acoustic parameter
manipulations (EMO ⫻ MANIP). This indicated, first, a
relatively low performance in discriminating expressiveness
of sentence pairs with sad intonation and, secondly, more
consistent attributions made to pitch-varied happy and
duration-manipulated neutral stimuli as opposed to their timeand pitch-varied counterparts (Fig. 3, right panel, and Fig.
6, left panel). Results of the statistics on behavioural data
are summarized in Table 2A.
Comparing the two grand average plots in Fig. 4, a clear
effect of inner speech can be assessed in terms of dissolving
the predominant right frontocentral activation. Without inner
speech, discrimination of affective intonations evoked brain
potentials with clearly higher amplitudes over the RH
compared with the LH, whereas additional performance of
inner speech resulted in a bilateral activation with left frontal
preponderance.
Several major findings emerged from the statistical analysis
of DC potentials by repeated measures ANOVA (all
significant effects and interactions are listed in Table 2B).
As indicated by a significant main effect of AP, activation
was more pronounced in anterior compared with posterior
regions. The effects of inner speech can be characterized
as follows. First, significant interactions with localization
(side ⫻ group and AP ⫻ side ⫻ group; see Fig. 4C, upper
panel, and Fig. 5) demonstrate a shift of activation pattern,
from predominant RH and frontal activation without inner
speech to pronounced left frontal activation in Experiment
2. Since data were normalized by subject and since different
subjects participated in the two experiments, no final
conclusion can be drawn as to what extent inner speech
resulted in RH deactivation and/or additional LH
involvement. However, visual inspection of the brain maps
derived from non-normalized data (Fig. 4) gives the
impression that the left shift of stimulus processing during
inner speech might, in part, be due to a reduction of RH
activation.
Secondly, inner speech showed significant interaction with
the acoustic parameters manipulated to create differences in
expressiveness (MANIP ⫻ group; Table 2B). During inner
speech, activation was more pronounced in response to timevaried compared with pitch-manipulated sentences, whereas
the reverse pattern was observed during spontaneous
discrimination in Experiment 1 (Fig. 4C, lower panel). No
significant interaction could be detected between the inner
2344
H. Pihan et al.
Fig. 4 (A and B) Grand average data of activation amplitudes. Mean amplitude values corresponding to grey levels and their units are
indicated by the reference bar of map (A) and (B). Interelectrode values were computed by a linear smoothing algorithm taking into
account the nearest four electrode values (Buchsbaum et al., 1982). (C) Plot of normalized amplitudes averaged across all subjects
including values of the second presentation/analysis periods at frontal and temporoparietal electrode positions (listed above). The upper
panel displays mean activation of LH and RH in Experiment 1 (white rhombi) and Experiment 2 (inner speech, black rhombi). The
lower panel shows mean activation during discrimination of F0 and duration-varied stimuli in Experiment 1 (inner speech: no) and
Experiment 2 (inner speech: yes).
Fig. 5 Plot of normalized amplitudes averaged across all subjects including values
of the second presentation/analysis periods at frontal (white background) and
temporoparietal electrode positions (grey background). Mean values are plotted
separately for the LH and RH, and for pitch- and duration-varied stimuli in
Experiment 1 (white rhombi) and Experiment 2 (inner speech, black rhombi).
speech condition and the emotional category (P ⬎ 0.1).
However, the factor EMO showed a significant interaction
with the kind of acoustic parameter manipulations
(MANIP ⫻ EMO; Table 2B). Pitch manipulation in neutrally
intonated sentences and duration variation in happy stimuli
evoked higher activation compared with their pitch- and timevaried emotional counterparts, whereas overall activation was
similar in sadly intonated sentence pairs (Fig. 6, right panel).
The two intrasubject factors EMO and MANIP did not yield
significant main effects, indicating that overall activation was
independent both of emotional content and of acoustic
parameters used for manipulation of expressiveness.
Discussion
With respect to the behavioural data, subjects’ responses
demonstrated that expressiveness of emotional intonations
can be predictably varied by changing F0 range or duration
of stressed syllables. Inner speech performed concurrently
with stimulus discrimination elicited a tendency to attribute
Prosodic competence of left hemisphere
more often a neutral intonation to sentence pairs with happy
or sad tone of voice, whereas no effect on discrimination of
expressiveness became apparent. The main results of the
brain activation study showed that recognizing emotional
intonations and discriminating their expressiveness leads
to a predominant activation of the RH with right frontal
preponderance only in the absence of linguistic task demands.
Inner speech performed in addition and concurrently with
the identification/discrimination task gave rise to a balanced
RH/LH activation pattern with left frontal preponderance.
This lateralization shift was linked to an inverse response of
cortical activity to F0- or time-varied stimuli. Without inner
speech, pitch manipulations resulted in higher activation
compared with durational manipulations, whereas the reverse
effect, that is higher cortical activity in case of durational
variations, was observable in subjects instructed to use
inner speech.
Significant interactions of acoustic parameter manipulations with emotions emerged both in the behavioural
analysis and, with reverse features, in cortical activation
strength (Fig. 6). They presumably result from varying
cognitive efforts required to derive an impression. In other
words, pitch might represent a stronger cue to expressiveness
in happy intonations, whereas durational differences seem to
Table 2B Significant effects and interactions in the
repeated measures ANOVAs performed on normalized
cortical activation values
Effect
d.f.
F
P
AP
GRP ⫻ SIDE
GRP ⫻ SIDE ⫻ AP
GRP ⫻ MANIP
EMO ⫻ MANIP
1.22
–
–
–
2.44
44.37
9.74
5.01
5.92
6.23
0.0001
0.0050
0.0356
0.0236
0.0041
The between-subjects factor group (with, without inner speech)
and the repeated factors AP (anterior, posterior), side (left, right),
MANIP (pitch, time) and EMO (sad, neutral, happy) were tested.
All effects not included here were non-significant (P ⬎ 0.1;
cortical activation analyses only).
2345
be more usable in neutral utterances. Similar effects of varying
cognitive demands on corresponding cerebral activation have
been described in a linguistically oriented functional MRI
study on comprehension of sentences with different structural
complexity (Just et al., 1996).
Acoustic structure and perception of affective
prosody
As in other psychophysically oriented studies, the pertinence
of the observed effects depends on clarifying the interdependence of physical stimulus characteristics and the
construction of a mental representation. Different theories of
emotion perception approach this still unresolved issue. The
category conception, which considers emotional perception
and expression to manifest in separate, unrelated stimulus
classes, faces a major problem which inheres in studies on
acoustic affective communication. Few emotional categories
present unique acoustic features and few acoustic features
correspond uniquely to given emotional categories (Pakosz,
1983). Observations that misclassifications of affective
expressions are based on certain similarities between emotions
gave rise to dimensional models which conceptualize
emotional perception as a result of multidimensional
evaluation. Typical dimensions introduced are activation (e.g.
excited versus calm), valence (e.g. pleasant versus unpleasant)
and control (e.g. intentional versus unintentional). Binary
terms were used to express contrasting poles and to permit
further investigations in perceptual–acoustic interrelations
(for a detailed discussion of this topic, see Frijda, 1969).
Recognition of speakers’ affective states does change
during the course of their utterances. Tischer presented
affectively intonated, naturally spoken phrases with an
increasing number of words and asked subjects to rate
them dimensionally (Tischer, 1993). He demonstrated that
differentiation of the valence dimension occurs only in the
course of the phrase and continues to develop throughout the
whole utterance. In contrast, differentiation of the activation
dimension was already completed in the shortest phrase,
Fig. 6 Interaction of emotional intonation and acoustic parameter manipulations. Left panel: means of
correct answers (ranked values), averaged over all subjects of Experiments 1 and 2. Right panel: plot of
normalized amplitudes averaged over all subjects (Experiments 1 and 2) including values of the second
presentation/analysis periods at all electrode positions considered for topographical analyses.
2346
H. Pihan et al.
which comprised the first two words of a short sentence.
This effect resulted in a typical change of affective state
attribution: while the shortest phrase with happy intonation
was often mistaken for an angry one, increasing stimulus
length reduced false negative and produced positive (happy)
impressions. It can be concluded from his study that
evaluation and discrimination of emotional intonations
presumably require the integration of different dimensional
aspects in a continuously updated impression.
The parameters considered for stimulus manipulation in
this study, i.e. F0 range and duration of stressed syllables,
presumably influence both the activation and the valence
dimension, depending on whether dynamic or absolute aspects
of stimulus perception are considered. Variations of dynamic
characteristics (pitch range, modulation of vowel duration)
were found to point at differences within the valence
dimension. They gain importance with increasing utterance
length (Tischer, 1993). At the same time these manipulations
also varied aspects of perception like mean pitch height and
speech rate which predominantly link with the activation
dimension (Frick, 1985).
Subjects in this study correctly identified the intended
affective prosody of the original utterances. In contrast,
evaluation of expressiveness yielded less consistent results
with a broader range of performance (Fig. 3). As stimuli
with happy and sad intonation showed marked differences
in constant and dynamic aspects (i.e. activity and valence
dimension) of emotion perception, correct identification of
emotional category was expected. Accordingly, difficulties
in evaluating expressiveness might at first be thought of as
resulting from only small variations of the activity dimension,
introduced by the acoustic manipulations. However, regarding
the large variation of performance in Fig. 3 (right panel)
another explanation becomes evident that refers to the inherent
ambiguity of terms like ‘happier’, ‘sadder’ or ‘more excited’
with respect to dimensional specifications. For example, an
utterance can be perceived as sadder if it expresses passivity
and less speaker activation (depression) or if it expresses
despair when high speaker arousal and activity is signalled.
Even happier might equally denote a more extrovert, highly
excited state (e.g. enthusiasm) as well as a rather relaxed,
self-content attitude like placidity. Consequently, it cannot
be excluded that some subjects changed their attitude during
the experiment alternately labelling stimuli with higher or
lower activation as more expressive. In order to avoid
direction of the subjects towards a pure evaluation of the
activity dimension and, possibly, towards a decision strategy
that relies on physical stimulus characteristics, response terms
were intentionally kept general. As controlling of this effect
was not looked upon as being essential with respect to the
activation study, it was not further investigated.
Lateralization of affective speech processing
As lesion studies could not unequivocally outline specific
cortical areas in which emotional meaning derived from
affective intonation is represented (for a comprehensive
review, see Gainotti, 1991; Heilman et al., 1993; Pell and
Baum, 1997b), the current theories of emotional processing
still await biological confirmation. While the valence
hypothesis suggests RH dominance for negative and LH
dominance for positive emotions, the RH hypothesis proposes
a RH superiority, regardless of valence. A third hypothesis
states that bilateral mechanisms underlie the perception of
affective intonation. They might be based on a relative RH
dominance in processing pitch-related parameters and a
predominant LH processing of temporal cues (Robin et al.,
1990; Van Lancker and Sidtis, 1992).
In our study, a predominant RH activation with right frontal
preponderance was observed when subjects discriminated
affective speech without any additional demands. Within
limitations, this result suggests a dominating RH function
for emotion processing according to the RH hypothesis.
However, specific effects of inner speech on affective
intonation processing question a simple LH/RH dichotomy
projected on to language and emotion perception.
Under additional inner speech demands, a balanced RH/
LH activation pattern arose with left frontal preponderance
and reverse responses to pitch- and time-related information
(Fig. 4C). Although our intersubject design does not allow
clear distinctions in terms of RH deactivation and/or additional LH involvement, further evaluation of lateralization
effects can be performed by considering the different task
demands.
First of all, effects of inner speech might be attributed to
an inclusion of LH resources by a supervening linguistic
performance. The cognitive function of repeating words ‘in
our head’ has been conceptualized in terms of a phonological
loop (Baddeley, 1995). It involves two components, a memory
store retaining phonological information and an articulatory
control process. Paulesu and colleagues were able to
demonstrate the underlying anatomy and localized the
phonological store to the left supramarginal gyrus (Paulesu
et al., 1993). The subvocal rehearsal system could be attached
to Broca’s area, whereas the left superior temporal gyrus
seems to contribute to memory-independent phonological
processing. These findings corroborate our results which
revealed a preponderance of activation recorded from the left
frontal electrode group (Figs 4B and 5). The number of
electrodes used and the restrictions of spatial resolution
inherent in the DC approach prevented reliable discrimination
of temporal and parietal activation. However, the selectivity
of temporoparietal brain responses towards frequency- and
time-related information in this area (Fig. 5) can hardly be
attributed to a general phonological storage, supporting our
view that no significant phonological memory component
was involved. Besides, there was no need to make use of a
phonological store while listening to the second sentence and
performing inner speech. This part of the task was not
directed to any further linguistic stimulus analysis. It can be
characterized as a process that quickly became automated in
the absence of considerable interference (subjective and
Prosodic competence of left hemisphere
behavioural) with prosodic task demands. According to the
RH hypothesis of emotion processing one would expect a
rather independent activation of corresponding left- and rightsided neural networks and would have to interpret the
activation pattern in Experiment 2 as resulting from an
increase of LH activity due to pure inner speech. Regarding
the comparatively little effort to perform this additional task,
this suggestion seems not to be justified. Comparing the
brain maps derived from non-normalized data (Fig. 4) and
the behavioural results (Fig. 3, left panel, and Table 2A),
indications of a reduction of RH activation under inner
speech might be seen to correspond with the behavioural
tendency to falsely denote an increased number of stimulus
pairs as neutral. However, this effect represents only a
tendency, leaving recognition of emotions basically intact.
We presume that the lateralization effects observed are neither
attributable to a repeated low level linguistic performance
like ‘inner speech’ nor to some kind of linguistically provoked
distraction from discrimination of emotional intonations, for
which we hold no behavioural evidence. As a consequence,
results from the second experiment do not support the RH
hypothesis.
LH processing of affective prosody?
Cortical activation patterns during affective speech processing
have been demonstrated to be dependent on concurrent
recruitment of the subvocal rehearsal system. As no significant
emotion main effect emerged, the results do not support
the valence hypothesis, which suggests RH dominance for
negative and LH dominance for positive emotional contents.
In addition, interaction between emotional category and inner
speech remained clearly non-significant. Our data rather show
that concurrent demands on the subvocal rehearsal system
result in a bilateral activation of the two hemispheres,
irrespective of emotional category. The striking finding is
that this effect resulted from a reversal of preference towards
the use of duration- or frequency-related acoustic features
compared with spontaneous discrimination (Fig. 4C, lower
panel). Although not significant in the omnibus statistics, a
differential use of left and right temporoparietal regions
towards the processing of either parameter is indicated in
Fig. 5, which demonstrates left temporoparietal effects of
temporal cue activation and right temporoparietal effects
when pitch-related information is processed.
Evidence from the literature indicates that bilateral
mechanisms may underlie decoding of affective intonation.
RH- and LH-damaged patients were found to show no
difference in performance when asked to identify four
different emotional intonations of short sentences (Van
Lancker and Sidtis, 1992). The authors demonstrated in a
meta-analysis on subjects’ errors that RH-damaged patients
preferentially relied on duration cues to make affective
judgements, whereas LH-damaged subjects appeared to make
use of F0 information. However, both groups performed
poorly compared with normal controls, indicating that
2347
preferential use of either parameter was not sufficient for a
good performance. Further evidence for different LH and
RH specialization in analysing acoustic signals was obtained
by Robin and colleagues (Robin et al., 1990). They showed
left temporoparietal lesions to impair the ability of gap
detection and pattern perception when sequences of tones
were presented, while frequency perception was completely
preserved. Lesions of homologous regions of the RH had
opposite effects with normal processing of temporal
information. The authors suggested that RH-damaged patients
might be impaired in making prosodic judgements by a
deficit in processing frequency-related information. Temporal
processing capabilities of the LH were looked upon as
referring to language and (linguistic-) prosody perception.
This seems evident, as important linguistic information of
the speech signal such as voice onset time or fast formant
transitions depends highly on temporal signal properties.
In this study, inner speech resulted in a balanced RH/LH
activation pattern with left frontal preponderance. Compared
with Experiment 1, this shift in lateralization effects was
paralleled by higher activation during discrimination of
duration-varied stimuli and less activity elicited by pitchvaried sentences. As duration of stressed vowels was changed
relative to their absolute length, the manipulations failed to
yield any difference in rhythm or stress pattern. It can be
assumed that duration variations did not introduce any
linguistically relevant difference into the speech signal.
Presumably, the only pertinence of perceived temporal
information resided within the prosodic task demand, which
was the discrimination of emotional expressiveness.
We assume that the shift of lateralization during inner
speech reflects a neural involvement in the LH that goes
beyond the activity of a subvocal rehearsal system.
Conceivably, inner speech results in a bilateral hemisphere
involvement together with new characteristics of acoustic
signal processing. Weighting of acoustic parameters relevant
for evaluation of emotional content changes, resulting in an
increase of LH activation by temporal cues and reduced RH
processing of frequency-related information. This change in
preferential use of either parameter proved to be compatible
with a good performance, at least in non-brain-damaged
subjects.
Coupling of acoustic input and verbal output
channel: suggestion for a new approach
towards lateralization of affective speech
processing
Cortical activation during discrimination of affective speech
has been shown to be dependent on concurrent demands
on the subvocal rehearsal system. The resulting bilateral
processing as such is not surprising as the two hemispheres
are closely connected through the corpus callosum and
other commissures. Rather striking is the observation that a
predominant RH activation during spontaneous discrim-
2348
H. Pihan et al.
ination changed into a bilateral pattern together with an
inverse RH and LH preference towards temporal- and
spectral-related information.
We assume that bilateral involvement during inner speech
reflects inherent LH functional coupling of acoustic input
and verbal output channels. This mechanism has been shown
to provide motor control information in the early phase of
speech analysis, enabling subjects to shadow (i.e. to repeat
as fast as possible) speech segments with latencies at the
lower limit for execution of motor gestures (Porter and
Lubker, 1980). These observations indicate parallel processes
of auditory analysis and generation of an articulatory gesture.
The purpose of this mechanism has been assumed to be to
facilitate continuous control of speech output by comparing
the signal produced with that intended. In more general
terms, shadowing studies suggest that perceived speech is
rapidly represented by neural networks, which provide an
interpretation of the acoustic structure in terms of a motor
program. Other input channels like the visual do not seem
to have the same coupling strength to verbal output. Shaffer,
for example, showed that auditory-vocal shadowing of
continuous speech can be successfully combined with visually
controlled copy-typing (Shaffer, 1975). In contrast, copytyping of continuous speech was not compatible with a
vocal reproduction of a written text, even for highly skilled
audio-typists.
To our knowledge, there are no studies investigating to
what degree the construction of an articulatory representation
of perceived speech also refers to suprasegmental intonational
aspects. These processes must be expected close to neural
structures in the frontal lobes, which are functionally
connected to articulophonatory programs. They could provide
feedback and control of prosody production, for example, if
someone tries to speak intentionally in a happy or angry
voice. Unfortunately, the neural structures underlying prosody
production have not, been identified yet as recently concluded
by Baum and Pell (Baum and Pell, 1997a). Conceivably, the
high activation over left frontal regions observed in this study
indicates the localization of the proposed mechanism.
Early representation of suprasegmental information may
also be relevant for evaluation of linguistic information like
syntactic phrase structure. In this context, an early left anterior
negativity was described by Friederici and colleagues in an
event-related potential study (Friederici et al., 1999). LHdamaged patients have been shown to improve in repetition
and discrimination of affective prosody under reduced verbalarticulatory load (Ross et al., 1997). We suggest that the
underlying mechanism is a decoupling of LH acoustic input
and verbal output channel resulting in a facilitated RH
stimulus processing and performance improvement.
B8: Altenmüller, B10: Ackermann/Daum) and the
Graduiertenkolleg Neurobiologie, University of Tübingen.
References
Ackermann H, Hertrich I, Ziegler W. Prosodische Störungen bei
neurologischen Erkrankungen – eine Literaturübersicht. Fortschr
Neurol Psychiatr 1993; 61: 241–53.
Altenmüller E. Hirnelektrische Korrelate der cerebralen
Musikverarbeitung beim Menschen. Eur Arch Psychiatry Neurol
Sci 1986; 235: 342–54.
Altenmüller EO, Gerloff C. Psychophysiology and the EEG. In:
Niedermeyer E, Lopes da Silva F, editors. Electroencephalography.
4th ed. Baltimore: Williams and Wilkins; 1999. p. 637–55.
Altenmüller E, Kriechbaum W, Helber U, Moini S, Dichgans J,
Petersen D. Cortical DC-potentials in identification of the languagedominant hemisphere: linguistical and clinical aspects. Acta
Neurochir Suppl (Wien) 1993; 56: 20–33.
Baddeley A. Working memory. In: Gazzaniga MS, editor. The
cognitive neurosciences. Cambridge (MA): MIT Press; 1995.
p. 755–64.
Bauer H, Korunka C, Leodolter M. Technical requirements for highquality scalp DC recordings. Electroencephalogr Clin Neurophysiol
1989; 72: 545–7.
Baum SR, Pell MD. Production of affective and linguistic prosody
by brain-damaged subjects. Aphasiology 1997; 11: 177–98.
Bergmann G, Goldbeck T, Scherer KR. Emotionale Eindruckswirkung von prosodischen Sprechmerkmalen. Z Exp Angew Psychol
1988; 35: 167–200.
Blonder LX, Bowers D, Heilman KM. The role of the right
hemisphere in emotional communication. Brain 1991; 114: 1115–27.
Buchsbaum MS, Rigal F, Coppola R, Cappelletti J, King C,
Johnson J. A new system for gray-level surface distribution maps
of electrical activity. Electroencephalogr Clin Neurophysiol 1982;
53: 237–42.
Chertkow H, Murtha S. PET activation and language. [Review].
Clin Neurosci 1997; 4: 78–86.
Frick RW. Communicating emotion: the role of prosodic features.
Psychol Bull 1985; 97: 412–29.
Friederici AD, von Cramon DY, Kotz SA. Language related brain
potentials in patients with cortical and subcortical left hemisphere
lesions. Brain 1999; 122: 1033–47.
Frijda NH. Recognition of emotion. In: Berkowitz L, editor.
Advances in experimental social psychology. New York: Academic
Press; 1969. p. 167–223.
Acknowledgements
Gainotti G. Disorders of emotions and affect in patients with
unilateral brain damage. In: Boller F, Grafman J, editors. Handbook
of neuropsychology, Vol. 3. Amsterdam: Elsevier; 1991. p. 345–61.
We wish to thank the anonymous referee of a previous
version of the paper for valuable comments. This study was
supported by the German Research Foundation (SFB 307;
George MS, Parekh PI, Rosinsky N, Ketter TA, Kimbrell TA,
Heilman KM, et al. Understanding emotional prosody activates
right hemisphere regions. Arch Neurol 1996; 53: 665–70.
Prosodic competence of left hemisphere
2349
Heilman KM, Bowers D, Speedie L, Coslett HB. Comprehension
of affective and nonaffective prosody. Neurology 1984; 34: 917–21.
ME, editors. Images of mind. New York: Scientific American
Library; 1994. p. 105–30.
Heilman KM, Bowers D, Valenstein E. Emotional disorders
associated with neurological diseases. In: Heilman KM, Valenstein E,
editors. Clinical neuropsychology. 3rd ed. New York: Oxford
University Press; 1993. p. 461–97.
Robin DA, Tranel D, Damasio H. Auditory perception of temporal
and spectral events in patients with focal left and right cerebral
lesions. Brain Lang 1990; 39: 539–55.
Jasper HH. Report of the Committee on Methods of Clinical
Examination in Electroencephalography. Electroencephalogr Clin
Neurophysiol 1958; 10: 370–5.
Just MA, Carpenter PA, Keller TA, Eddy WF, Thulborn KR. Brain
activation modulated by sentence comprehension. Science 1996;
274: 114–6.
Ladd DR, Silverman KEA, Tolkmitt F, Bergmann G, Scherer KR.
Evidence for the independent function of intonation contour type,
voice quality, and F0 range in signaling speaker affect. J Acoust
Soc Am 1985; 78: 435–44.
Ohala JJ. An ethological perspective on common cross-language
utilization of F0 of voice. Phonetica 1984; 41: 1–16.
Oldfield RC. The assessment and analysis of handedness: the
Edinburgh inventory. Neuropsychologia 1971; 9: 97–113.
Pakosz M. Attitudinal judgements in intonation: some evidence for
a theory. J Psycholinguist Res 1983; 12: 311–26.
Paulesu E, Frith CD, Frackowiak RS. The neural correlates of the
verbal component of working memory. Nature 1993; 362: 342–5.
Pell MD, Baum SR. The ability to perceive and comprehend
intonation in linguistic and affective contexts by brain-damaged
adults. Brain Lang 1997a; 57: 80–99.
Pell MD, Baum SR. Unilateral brain damage, prosodic
comprehension deficits, and the acoustic cues to prosody. Brain
Lang 1997b; 57: 195–214.
Pihan H, Altenmüller E, Ackermann H. The cortical processing of
perceived emotion: a DC-potential study on affective speech prosody.
Neuroreport 1997; 8: 623–7.
Porter RJ Jr, Lubker JF. Rapid reproduction of vowel-vowel
sequences: evidence for a fast and direct acoustic-motoric linkage
in speech. J Speech Hear Res 1980; 23: 593–602.
Posner MI, Raichle ME. Interpreting words. In: Posner MI, Raichle
Rockstroh B, Elbert T, Canavan A, Lutzenberger W, Birbaumer N.
Slow cortical potentials and behaviour. 2nd ed. Baltimore: Urban
& Schwarzenberg; 1989.
Ross ED, Thompson RD, Yenkosky J. Lateralization of affective
prosody in brain and the callosal integration of hemispheric language
functions. Brain Lang 1997; 56: 27–54.
Schlanger BB, Schlanger P, Gerstman LJ. The perception of
emotionally toned sentences by right hemisphere-damaged and
aphasic subjects. Brain Lang 1976; 3: 396–403.
Shaffer LH. Multiple attention in continuous verbal tasks. In: Rabbit
PMA, Dornic S, editors. Attention and performance V. London:
Academic Press; 1975. p. 157–67.
Springer SP, Deutsch G. Linkes Rechtes Gehirn. Heidelberg:
Spektrum; 1995.
Tischer B. Äusserungsinterne Änderungen des emotionalen
Eindrucks mündlicher Sprache: Dimensionen und akustische
Korrelate der Eindruckswirkung. Z Exp Angew Psychol 1993; 40:
644–75.
Tompkins CA, Flowers CR. Perception of emotional intonation by
brain-damaged adults: the influence of task processing levels. J
Speech Hear Res 1985; 28: 527–38.
Tucker DM, Watson RT, Heilman KM. Discrimination and evocation
of affectively intoned speech in patients with right parietal disease.
Neurology 1977; 27: 947–50.
Van Lancker D, Sidtis JJ. The identification of affective-prosodic
stimuli by left- and right-hemisphere-damaged subjects: all errors
are not created equal. J Speech Hear Res 1992; 35: 963–70.
Weintraub S, Mesulam MM, Kramer L. Disturbances in prosody.
Arch Neurol 1981; 38: 742–4.
Received November 11, 1999. Revised June 6, 2000.
Second revision July 3, 2000. Accepted July 14, 2000