BRAIN AND LANGUAGE SI, 318-335 (1995)

Prosodic Characteristics of Speech Pre- and
Post-Right Hemisphere Stroke

LEE XENAKIS BLONDER,*"'-4 JOHN E, PICKERING,§ Rosin L. HEATH,* 1-4
Cuarces D. Smitu,{|| AND STEVEN M, BUTLERt**

*Department of Behavioral Science, *Sanders-Brown Center on Aging, tDepariment of
Anthropology, “Department of Neurology, |Magnetic Resonance Imaging and
Spectroscopy Center, and **Department of Statistics at the University of Kentucky, and
the §Communication Disorders Program at the College of Saint Rose

Case-control studies have shown right hemisphere specialization in the produc-
tion of intonation in speech. We examined spontaneous prosody in audiotapes of
interviews with a 77-year-old right-handed woman recorded 6 months before and
6 weeks after she suffered a stroke affecting the right frontotemporo-parietal
regions and the right basal ganglia. Post-stroke. the patient had a normal Mini-
Mental Status Examination Score of 29, hemispatial neglect, and impairments in
the comprehension of facial expression and prosody. Self-rated mood was within
normal limits. We compared beginning, peak, and ending fundamental frequencies
(fo) in breath groups. the timing of these fo changes, rate of speech, pause dura-
tion, and breath-group duration. We found that post-stroke. the patient had a
more restricted fo contour, no changes in the timing of peak fo, an increased rate
of speech, less variability in pause duration, and no changes in breath-group
duration. ¢ 1998 Academie Press. Inc.

INTRODUCTION

Prosodic abnormalities have been observed in patients with damage to
varying regions of the brain, including the cerebral cortex, the basal gan-

This research was supported in part by an NIH/NINDS FIRST Award NS29082 to
L. X. Blonder and an NIH/NINDS Clinical Investigator Development Award NS01421 to
C. D. Smith. We thank the following individuals for their contributions to the project:
Thomas Arcury. Director of the Oral History of Kentucky Farm Families Research Pro-
gram; Jeffrey Suchanek of the UK Oral History Program; Sara Quandt and Gun Roos of
the UK Oral History Program and the Department of Anthropology; Sandra Kryst, Depart-
ment of Anthropology: Joseph Fox, Chief of Audiology and Speech Pathology at the Veter-
ans Administration Medical Center: John Haley, Biostatistics Consultant Service: and the
staff of Cardinal Hilt Hospital. Address correspondence and reprint requests to Lee Xenakis
Blonder, Ph.D., Center on Aging, 101 Sanders-Brown Building, University of Kentucky,
Lexington, KY 40536-0230. FAX: (606) 257-8990. E-mail: STRO03 @ UKCC.UKY.EDU.

318
0093-934X/95 $12.00
Copyright © 1995 by Academic Press, inc.
All rights of reproduction in any form reserved.
PROSODIC CHARACTERISTICS 319

glia, and the cerebellum (Monrad-Krohn, 1963; Darley, Aronson, &
Brown, 1969; Kent & Rosenbek, 1982; Tucker, Watson, & Heilman,
1977; Odell, McNeill, Rosenbek, & Hunter, 1991; Ryalls & Behrens,
1988, Cancelliere & Kertesz, 1990). These abnormalities are diverse and
include defects in syllabic stress, rate of speech, juncture, and intonation.
In spite of conflicts in the literature concerning the neural representation
of prosodic features, a large body of evidence supports a right hemisphere
specialization in the production of intonation, particularly emotional into-
nation (Tucker et al., 1977; Ross & Mesulam, 1979; Ross, 1981: Wein-
traub, Mesulam, & Kramer, 1981; Hughes, Chan, & Su, 1983; Borod,
Koff, Lorch, & Nicholas, 1985).

Tucker et al. (1977) found that patients with right parietal dysfunction
and neglect had difficulty both discriminating and producing emotional
prosody. Ross (1981) described several cases of patients with various
types of aprosodias resulting from right hemisphere lesions, including
“motor aprosody”’ characterized by monotone speech. Borod et al.
(1985) showed that right hemisphere damaged (RHD) patients suffered
impairments in emotional facial expression, prosody, and emotional
speech. Speedie, Coslett. and Heilman (1984) noted that two patients
with transcortical motor aphasia resulting from Jeft hemisphere damage
lacked emotional intonation during a prosodic repetition task. They sug-
gested that these cases of aprosodia were caused by the disconnection
of right hemisphere regions that mediate emotional prosody and the left
perisylvian area. Weintraub et al. (1981) sought to determine whether
prosodic disorders accompanying RHD were limited to emotional pro-
cessing, and found that RHD patients were impaired in nonaffective pro-
sodic comprehension, repetition, and spontaneous production.

The studies of aprosody described above rely on clinical observation
or ratings made by judges. As such, they provide indirect evidence that
the disorder is communicatively significant. A small number of investiga-
tors have undertaken acoustic quantification of speech in RHD patients.
These results delineate more precisely the mechanical, physical, and lin-
guistic nature of prosodic production following RHD.

Kent and Rosenbek (1982) did spectrographic analysis of prosody in
patients with ataxic dysarthria, apraxia of speech, Parkinson dysarthria,
and right hemisphere dysarthria. They found that the speech of RHD
patients strongly resembled that of Parkinson patients in that both had a
normal or increased rate of speech, decreased fundamental frequency
(fo) variation, and decreased intensity in the resonance frequencies
above 500 hz, due to nasalization. Cooper et al. (1984) measured fo and
duration in sentences read by RHD patients, left hemisphere damaged
(LHD) aphasics, and normal controls. They noted that Broca’s aphasics
produced flatter intonation contours, longer utterances, and longer sen-
tence final words than Wernicke’s aphasics, RHD patients, and normal
320 BLONDER ET AL.

controls. However, their data also show that RHD patients’ intonation
contours were flatter and the duration of their utterances and sentence
final words slightly shorter than that of Wernicke’s aphasics and normal
controls. No statistical analyses were performed. Ryalls, Joanette, and
Feldman (1987) compared average fo range, contour slope, and sentence
duration in repeated sentences and found no significant differences be-
tween RHD and normal subjects. They did find a trend, however, for
RHD patients tested prior to 100 days post-onset, to have narrower fo
ranges than those tested at 100 days or more post-onset. These results
suggest that the disorder may be a transient phenomenon. Alternatively,
the non-emotional nature of the stimuli or the use of a repetition task
may have influenced the performance of the RHD patients. Ross et al.
(1988) found significant differences in fo slope, fo register, fo variation,
and delta fo (a measure of deviation of intonation from a flat contour) on
an emotional prosodic repetition task administered pre- and post-Wada
testing. Behrens (1989) demonstrated that RHD patients’ intonation con-
tour was characterized by more restricted fo dispersion in declarative
sentences and yes—no questions, and flatter fo decline in declarative sen-
tences. Shapiro and Danly (1985) asked LHD patients and RHD patients
with anterior, central, or posterior lesions, and a group of normal controls
to read sentences using linguistic (interrogative, declarative) or emotional
(happy, sad) prosody. They found that RHD patients with anterior and
central lesions demonstrated more restricted fo range in both linguistic
and emotional prosody, while patients with posterior RHD had exagger-
ated fo range during linguistic and emotional speech. The prosody of
LHD patients was similar to that of normal controls.

In contrast to these studies of intonation following right hemisphere
damage, acoustic investigations of lexical stress in RHD patients have
found preserved production. Emmorey (1987) analyzed stress contrasts
in compound nouns and noun phrases and concluded that RHD patients
performed comparably to normal controls. Behrens (1988) also found that
the production of lexical stress was preserved in RHD patients.

While the results of these studies are somewhat inconsistent, they sug-
gest that the right hemisphere may mediate the modulation of fo over the
length of an utterance (see Behrens, 1989). Although many studies find
that the right hemisphere is particularly specialized in the production of
emotional intonation, this may be a function of the research design in
which comparisons with nonemotional prosodic production are not made.

With the exception of the study by Ross et al. (1988) which employed
a within-subjects design pre- and post-Wada testing, the investigations
described above are based on analyses of read, repeated, or experimen-
tally elicited sentences produced by individuals after they have suffered
right hemisphere damage. In this case study, we had the unique opportu-
nity to examine prosody in spontaneous discourse produced by a woman
PROSODIC CHARACTERISTICS 321

6 months before and 6 weeks after she suffered a right hemisphere infarct.
Because psycholinguists have found significant correlations in inter-
viewee utterance duration, pause duration, and reaction time latency for
intervals up to 15 months (Matarazzo & Wiens, 1972). we felt justified in
our assumption that prosodic characteristics should remain stable over
the 7.5-month interval. We performed acoustic analysis of discourse us-
ing high quality audiotapes of pre and post-stroke interviews. We com-
pared intonation (changes in fo over time), rate of speech, pause duration,
and breath group duration in pre- and post-stroke speech samples. In
addition, we compared aspects of verbal communication in the pre- and
post-stroke interviews and obtained neurobehavioral and neuroimaging
data following the stroke. Based on previous research, we predicted that
our patient would manifest flatter intonation patterns and an increase in
the rate of spontaneous speech following her right hemisphere infarct.
We also predicted that indices of left hemisphere verbal competence,
such as vocabulary diversity, syllables per word, and percentage of words
that were prepositions, would remain stable.

CASE REPORT AND METHODS

P.C. is a right-handed 77-year-old woman who suffered an embolus to her right frontotem-
poro-parietal region on January 16, 1992. She had a 10th-grade level of education and was
a homemaker and a retired Soil Conservation District Supervisor. At the time of her stroke
she was widowed and living in a rural area in an apartment complex for senior citizens.
Her medical history was significant for rheumatic heart disease. hypertension, mild chronic
renal insufficiency, and chronic atrial fibrillation. Her aortic and mitral valves were replaced
in 1979, Neurologic examination at the time of her stroke showed that, with the exception
of a moderate left central VII paresis. cranial nerves II through XII were intact. She had
left-sided hyperreflexia without clonus and an upgoing toe on the left. Proprioception and
stereognosis were impaired on the left side, particularly the left upper extremity. and there
was left hemispatial neglect. A computerized tomographic (CT) scan performed on the day
of the stroke revealed mild cortical atrophy consistent with age. A repeat CT scan performed
four days later showed a large acute nonhemorrhagic infarct involving the right posterior
frontal, anterior and middle temporal, and anterior parietal lobes, and the head of the
caudate nucleus and the globus pallidus on the right. Medications at the time of our assess-
ments were nitedipine 30 mg. PO b.i.d. (Procardia XL), Lanoxin .125 mg. PO q.d., and
Coumadin 2.5 mg. PO q.d. Pre-stroke medications included all of the above with the excep-
tion that she had been taking dipyridamole (Persantin) rather than Coumadin for prophylac-
tic anti-coagulation.

P.C. was participating in a study of emotional communication following unilateral stroke
when we discovered that she had taken part in the University of Kentucky Family Farm
Oral History Program six months before her stroke. In this project, residents of diverse
counties are interviewed individually regarding their farming practices and dietary behav-
iors. Each of these interviews is recorded on a Sony TCM 5000 stereo cassette recorder
using a Lavallier microphone in quiet conditions. The interview with P.C. was conducted
in her home on July 12, 1991. The duration of this interview was 131 min. This interview
had been transcribed, and both the original audiotapes and the transcription were made
available to us. This release of data was permitted according to the Oral History project
consent form approved by the University of Kentucky Institutional Review Board and
322 BLONDER ET AL.

signed by the participant. We studied the contents of the transcript, paying particular atten-
tion to the types of questions asked, and performed a similar interview with P.C. post-
stroke. The patient was simply told that we wanted to gain information on her background
and past life, and was not informed as to our interest in the earlier interview or speech
prosody. She was interviewed on February 29, 1992 (44 days post-stroke and 7.5 months
after the initial interview) in a quiet room in the rehabilitation hospital where she was an
inpatient. by RLH who recorded the interview on the same equipment that had been used
to make the pre-stroke audiotape. The patient was a natural historian and did not remark
on our interest in her past life or the similarity between our questions and those of the
interviewer in the Oral History Project. The duration of the post-stroke interview was 126
min,

To analyze P.C.’s intonation before and after the right hemisphere stroke, the two tran-
scribed oral history interviews were separated into breath groups. A breath group is a unit
of speech prosody separated by inhalation (Lieberman and Blumstein, 1988). Acoustic
analysis of selected breath groups was done using Kay Elemetrics Visi-Pitch (Model 6095/
6097) interfaced with an IBM-compatible microcomputer. Each breath group was individu-
ally played into the Visi-Pitch using a Marantz Model PMD 430 stereo cassette recorder.

In preliminary comparisons, we found that the resolution and reliability of the Visi-Pitch
were highest when using a 2-sec window and 0-600 hz range. Thus, we limited the sample
of breath groups analyzed to those not exceeding two seconds in length. The actual means
and standard deviations for these restricted breath groups are 1.74 (.23) sec pre-stroke and
1.78 (.20) sec post-stroke. This comparison was not statistically significant [7(59) = — 1.03,
p = 31), ensuring that potential pre- and post-stroke differences in fo range would not be
secondary to differences in utterance length and concomitant reallocation of pulmonary air
pressure (Ryalls 1982). Additionally, the pre- and post-stroke means in the duration of
breath-groups in the restricted samples did not differ significantly from the pre and post-
stroke means in two unrestricted samples of 40 breath groups each [pre-stroke mean; (SD):
2.003 {.99). post-stroke mean 1.998 (.96}], suggesting that potential pre- and post-stroke
differences in fo measures would not be a function of differences in breath group duration
between the restricted and unrestricted samples.

We sampled 60 breath groups of 2 sec or less in duration from each interview. The breath
groups sampled were distributed across the interviews. They were played into the Visi-Pitch
and analyzed acoustically. The Visi-Pitch provided a visual image corresponding to the
intonation pattern produced for each breath group. Five acoustic measures were obtained
for each breath group based on the visual image: (1) the fo at the beginning of the breath
group (foB); (2) the highest fo obtained in the breath group, or peak frequency (foP);
(3) the fo at the end of the breath group (foE); (4) the time between the beginning of the
breath group and the peak fundamental frequency: and (5) the time between the peak
fundamental frequency and the ending fo of the breath group. The fundamental frequency
is the lowest frequency component of a complex sound wave and represents the number
of cycles per second of vocal fold vibration (Crystal, 1992). It has been measured repeatedly
in research on intonation (see Ross et al., 1988, for further description). We chose to
measure the beginning, peak and ending frequencies because these points represent the
rise—fall intonation pattern characteristic of unmarked breath groups (Lieberman and
Blumstein, 1988). The peak frequency consistently corresponded to the listener's perception
of the stressed syllable within the breath group. Further, the fo measures were obtained
from voiced segments, principally vowels. The three fo measures and the two duration
measures provided the information needed to analyze the speaker's intonation patterns.

In addition to these measures, we compared rate of speech (syllables per minute) in 50
one minute samples of speech per interview, duration of pauses (in seconds) in a distributed
sample of fifty pauses per interview, and duration of breath groups in a sample of 40 per
interview (see above). Syllables per minute were counted using a digital stop watch. Pause

PROSODIC CHARACTERISTICS 323

duration and duration of breath group were calculated using wide-band spectrograms gener-
ated by the Computerized Speech Laboratory (Kay Elemetrics).

In addition to the acoustic analyses, we performed select neurobehaviora! tests post-
stroke and examined aspects of verbal competence in the interviews. These latter measures
included syllables per word. percentage of words that were prepositions, and vocabulary
diversity. using a modification of the type-token ratio (Templin, 1957). The type-token ratio
is the number of unique words divided by the total number of words used in a speech
sample. We used the first 800 words of each interview as the denominator and the number
of unique words in that sample as the numerator.

RESULTS
Neurobehavioral Assessment

Neurobehavioral testing, begun approximately 3 weeks post-stroke,
revealed that P.C. performed eight out of eight tasks with her right hand,
had no history of childhood left-handedness, and had no left-handed par-
ents, siblings, or children. She obtained a score of 29 out of 30 on the
Mini-Mental Status Examination (Folstein, Folstein. & McHugh, 1975),
having lost one point as a result of constructional apraxia. She scored in
the normal range on the Zung Depression Scale (Raw score: 27, SDS
Index: 34: Zung, 1965), but was impaired in the perception of facial and
prosodic emotion, and in cross-modal matching of facial and prosodic
emotion on the Florida Affect Battery (Bowers et al.. 1991; see Table
1). We also administered the Positive and Negative Affective Schedule
(PANAS; Watson et al., 1988), which consists of 10 words that refer to
positive emotions and 10 words that describe negative emotions. The
subject responds by indicating the extent to which s/he feels each emo-

TABLE 1
Post-Stroke Neurobehavioral Profile
Test Score Performance

Mini Mental Status Exam 29 Normal
Zung Depression Scale 34 (SDS Index) Normal
PANAS 33 (Pos) If (Neg) Normal
Florida Affect Battery:

Facial 1D Discrimination 50% Impaired*

Facial Emot Discrimination 80% Normal

Name Emotional Face 60% Impaired*

Point to Emotional Face 606% Impaired*

Match Emotional Face 4S% Impaired*

Neutral Prosody Discrimination 56% Impaired*

Emot Prosody Discrimination 65% Impaired*

Name Emotional Prosody 55% Impaired*

Match Emot Prosody to Face 40% Impaired*

Match Emot Face to Prosody 28% Impaired*

* Impaired score greater than 2 SD below norma] contro} mean (Blonder et al.. 1991)
324 BLONDER ET AL.

tion on a 5-point scale. We asked P.C. to provide a rating based on the
last few days. P.C. achieved a positive emotion score of 33 and a negative
emotion score of 11. Both of these scores were within approximately one
standard deviation of the normative means reported by Watson et al.
(1988). P.C. did not manifest neglect on the Line Bisection or Draw a
Clock subtests of the Florida Neglect Battery (Heilman et al., 1983), but
did have left-sided neglect on the Crossing Out subtest.

Neuroanatomical Mapping

To determine more precisely the location of the patient’s infarct, an
experienced neurologist trained in neuroimaging (CDS) compared tem-
plates published by Damasio and Damasio (1989) and Matsui and Hirano
(1978) to CT X-ray images of the brain acquired for clinical purposes
at a community hospital using a standard CT scanner (General Electric
CT-9800). Fourteen 5-mm axial slices oriented 17° to the inferior orbital-
external acoustic meatal baseline were acquired (interslice gap: slices 1-7
(posterior fossa), 3.75 mm:; slices 7-14, (cerebrum), 6.25 mm]. Determina-
tion of lesion location was done without knowledge of the patient's clini-
cal or neurobehavioral findings.

On images acquired the day of the stroke, there was minimal general-
ized cerebral atrophy consistent with age. In addition, a very small area
of hypodensity (1.5 x 2 mm) in the medial head of the left caudate
nucleus was seen and interpreted as an old lacunar infarct (see Fig. 1).
This lacunar infarct was clinically silent, as the patient had no docu-
mented prior history of stroke. Four days post-stroke, a well-defined
hypodensity in the right hemisphere was accompanied by a 2-mm shift
of the intraventricular septum to the left. No region of hyperintensity
suggestive of hemorrhage was observed. Brain tissue hypodensity in the
distribution of the right middle cerebral artery included virtually all of
the pre-frontal, pre-central, and central branch territories in the frontal
lobe and the temporo-polar, anterior temporal, middle temporal, and pos-
terior temporal branch territories in the temporal lobe. The lenticulostri-
ate vascular territory likewise showed hypodensity. This area includes
the anterior limb of the internal capsule, the head of the caudate nucleus,
the putamen, and the anterior portion of the globus pallidus. The poste-
rior limb of the internal capsule was spared. A thin ribbon in the territory
of the anterior insular branches showed relatively preserved intensity on
these images. Although there was a loss of sulcal detail in the right poste-
rior parietal and angular vascular territories (including the supramarginal
gyrus), image intensity was preserved in these regions. Right hemispheric
vascular territories of Heubner’s artery, the anterior choroidal artery,
and all branches of the anterior and posterior arteries were spared. In
the left hemisphere, the only abnormality was the old lacunar infarct
PROSODIC CHARACTERISTICS 325

Ficure |

previously described. In summary, CT imaging showed an acute right
fronto-temporal, anterior parietal middle cerebral artery infarct which
included the right lenticulostriate vascular territory (anterior limb of the
internal capsule, the head of the caudate nucleus, the putamen, and the
anterior portion of the globus pallidus), but spared the angular gyrus and
the superior parietal region (Figs. 2A and 2B).
326 BLONDER ET AL.

a

Ficurr 2

Acoustic Analysis

Because the observations under study derived from a single individual,
we could not assume that the data points were independent and that the
t and F would be appropriate statistical tests. We therefore performed
Pearson Product-Moment correlations on adjacent pairs to assess auto-
correlation (see Kazdin, 1982, for further discussion). The pre-stroke au-
PROSODIC CHARACTERISTICS 327

a

Fic. 2—Continued

tocorrelations for the variables under study were as follows: beginning
fo: —.11, p = .41: peak fo: .08, p = .53; end fo .20, p = .12; duration
in seconds from beginning to peak fo: —.0070, p = .96; duration in
seconds from peak to end fo: .08, p = .54; syllables per minute: .15,
p = .31; pause duration: .11, p = .44; and breath group duration: .05,
p = .77. None of these variables displayed significant autocorrelation
(a = .05) so they were analyzed using analysis of variance (ANOVA) or
the ¢ test.

We performed a 2 x 3 repeated-measures ANOVA with no grouping
factor and two within-group factors [stroke status (pre, post) and funda-
mental frequency (foB, foP, and foE). We found no main effect of stroke

328 BLONDER ET AL.

status, meaning that the overall fo level before and after stroke was com-
parable [Mean (SD): 189.3 (36.5) hz pre and 191.6 (28.3) hz post). There
was a main effect of fo [F(2, 118) = 213.7, p < .0001]. Post-hoc compari-
sons using Fishers test revealed that overall, both the beginning fre-
quency and the end frequency were significantly lower than the peak
(p < .001). This was expected given that declarative sentences are char-
acterized by a rise and then a fall in fo. We also found a frequency by
status interaction [F(2, 118) = 9.24, p = .0002] such that the beginning
fo pre-stroke was significantly lower than the beginning fo post-stroke (p
< .01), and the peak fo pre-stroke was significantly higher (p < .05) than
the peak frequency post-stroke. There were no statistically significant
differences between ending fo values. We also calculated fo range for
beginning, peak, and ending values pre- and post-stroke by subtracting
the lowest recorded fo value from the peak fo value for each breath
group. We found statistically significant differences [T(59) = 3.32, p =
002] such that the fo range pre-stroke was greater [99.1 hertz (SD: 46.4]
than the fo range for these measurements post-stroke [73.9 hertz (SD:
30.2)]. Taken together, these results signify that the change in fo over
time was more restricted post-stroke, resulting in a flatter intonation con-
tour (see Fig. 3).

We performed a2 x 2 repeated-measures ANOVA using stroke status
(pre, post) and relative duration (time from beginning fo to peak fo; time
from peak fo to end fo) as the within-group factors. There was no main
effect of stroke status, meaning that the occurrence of a right hemisphere

Frequency in Hertz
&

245

Fa

155 bus . TS rT FSP TFT TPP PT SP

i Ol 02 03 0.4 os 0.6 0.7 08 a9 1 Vw 12 W3
Duration in Seconds from Beginning to Peak Duration in Seconds from Peak 10 End

@ Pre-Stroke _M™ Post-Stroke

Main effect of Frequency: 2,118) = 213.7, p< .0001; Frequency x Stroke Status Interaction: F(2,118) = 9.24, p < 0.0002;
Main Effect of Duration: F(1,59) = 7.40, p< .01.

Fic. 3. Fundamental frequency in breath groups. Pre- and post-right hemisphere stroke.
PROSODIC CHARACTERISTICS 329

infarct did not affect the average time change (beginning to peak and
peak to end fo). There was a main effect of relative duration [F(1, 59) =
7.40, p < .01] such that the overall time from beginning to peak fo was
significantly less than the time from peak to end fo. Again, this pattern
is expected for declarative sentences. Finally, there was no relative dura-
tion by status interaction, suggesting that the timing of the peak frequency
in relation to the beginning and end frequencies did not differ as a function
of stroke (Fig. 3).

In order to examine the effects of this patient's right hemisphere stroke
on rate of speech, we compared syllables per minute in 50 I-min samples
of pre- and post-stroke discourse. These values were then subjected to a
matched pairs f test. We found that P.C.’s rate of speech was significantly
faster after her stroke than before [7(49) = 4.78, p < .0001]. In particular,
she spoke an average of 219.2 syllables per minute post-stroke (SD: 19.2)
compared with 198.1 (22.2) syllables per minute pre-stroke (see Fig. 4).
Variances were homogeneous. We also compared the duration in millisec-
onds of 50 pauses before and 50 pauses after stroke and did not find
significant differences [7(49) = 1.68, p = .10; Pre-stroke mean: .48 (.25);
Post-stroke: .55 (.17)]. These variances were not homogeneous [Levene
(1, 98) = 7.08, p = .009], indicating that variance in pause duration
pre-stroke was greater than variance in pause duration post-stroke (see
Fig. 5). Finally, we compared the duration in seconds of a non-restricted
sample of 40 breath groups produced before the stroke and 40 produced

Syllables per Minute
270

10 2» 40 30 10 w 30 40 50
Time in Minutes Pre-Stroke Time in Minutes Post-Stroke

T (49) = 4.78, p<.0001
Fic. 4. Syllables per minute. Pre- and post-right hemisphere stroke.
330 BLONDER ET AL.

Duration in Seconds

0.35

Pre-Stroke Post-Stroke
T (49) = 1.68, p<.10; Levene's Test for Equality of Variance (1,98)=7.08, p=.009

Fic. 5. Duration of pauses. Pre- and post-right hemisphere stroke.

after the stroke and found no significant difference, either in the mean or
the variance [7(39) = —.02. p = .99; Levene (1, 78): .23, p = .64].

Verbal Competency

in our analysis of vocabulary diversity as indexed by the type-token
ratio, we found that P.C. produced 270 unique words in a sample of 800
words taken from the beginning of the pre-stroke interview and 272
unique words in a sample of 800 words taken from the beginning of the
post-stroke interview. These produced a type-token ratio of .34 both pre-
and post-stroke, which is comparable to the mean TTR in 800 word sam-
ples of interview discourse produced by five healthy spouses of Alzhei-
mer’s patients in a prior study [spouse mean: .34 (SD .03); Blonder, Kort,
& Schmitt, 1994]. We also computed the mean number of syllables per
word in each 800 word sample pre- and post-stroke and found that P.C.’s
pre-stroke mean was 1.280 and the post-stroke mean [.269. Both of these
values are within one standard deviation of our normal control mean in
the Alzheimer’s study [1.304 (.063); Blonder, et al., unpublished data).
Lastly, we calculated the percentage of words that were prepositions in
these pre- and post-stroke interview samples and found that 10.75% of
the pre-stroke 800 word sample were prepositions and 13% of the post-
stroke 800 word sample were prepositions. Again, both of these values
are within one standard deviation of our normal control mean [11.38
(1.94); Blonder et al., unpublished data].
PROSODIC CHARACTERISTICS 331

DISCUSSION

We analyzed the prosodic characteristics of speech in a right-handed
77-year-old woman using audiotaped interviews conducted before and
after she suffered a massive right hemisphere stroke. During the subacute
period post-stroke this patient demonstrated intact mental status and
mood, but impaired spatial hemiattention and receptive facial and pro-
sodic processing. Samples of spontaneous speech recorded 6 months be-
fore and 6 weeks after her stroke revealed expected verbal competence
in vocabulary diversity, syllables per word. and percent of prepositions
used. However, as predicted. significant differences in prosodic produc-
tion were found on acoustic analysis. In particular, this patient produced
significantly higher beginning fo and significantly lower peak fo after the
stroke. Frequencies at the end of the breath groups did not differ. These
results suggest that the post-stroke abnormality occurred during the rise
portion of the rise—fall contour. Furthermore, the overall range in begin-
ning, peak, and ending fo was significantly more restricted post-stroke.
Ryalls (1982, 1986) has suggested that fo range may be a percentage of
the average fo. Thus higher fundamental frequencies would be associated
with wider fo range. However. since there were no differences in our
patient’s mean fo pre- and post-stroke, the post-stroke restriction in range
is not secondary to a lowering in the average frequency of the patient's
voice. Ryalls (1982) also suggested that restrictions in fo range found in
Broca’s aphasics during a sentence repetition task may have been due to
a loss of pulmonary air pressure resulting from elongating utterances.
Since the average pre- and post-stroke breath group duration did not
differ in our patient, the decrease in fo range is also not the result of the
patient elongating her utterances. It is noteworthy that although P.C.’s
intonation is more restricted post-stroke, she did continue to produce a
rise-fall contour. Furthermore, the relative timing of the patient's peaks
in fo did not differ as a function of the stroke. These results suggest that
some aspect of the ‘‘rule’’ used to generate declarative sentence intona-
tion was preserved.

In addition, the patient's rate of speech as measured in syllables per
minute was significantly faster following the stroke. Given that the pa-
tient’s rate increased post-stroke in the absence of an increase in the
duration of her breath groups, we propose that the neural mechanisms
that control rate of speech and duration of breath groups may differ.
Furthermore, although the mean duration of P.C.’s pauses did not differ,
the post-stroke variance in pause duration was more restricted. If repli-
cated in a Jarger sample of RHD patients, these results may signal an
abnormality in the rhythm of discourse which in turn contributes to the
monotonous impression often noted in RHD patients’ speech.
332 BLONDER ET AL.

Template analysis of this patient’s CT scan performed 4 days post-
stroke revealed involvement of the right fronto-temporal and anterior
parietal regions as well as the anterior limb of the internal capsule, the
head of the caudate nucleus, the putamen, and the anterior portion of the
globus pallidus. The angular gyrus and the superior parietal region were
spared. Our patient's receptive and expressive prosodic deficits are pre-
dictable based on the size and location of this lesion (see Ross, 1981). It
is significant that this patient, and many of the aprosodic patients in
previous studies, including Parkinsonians, had sustained damage to the
basal ganglia. Recently, Cancelliere and Kertesz (1990) found that disor-
ders of emotional prosodic expression were most often associated with
basal ganglia damage, and that the frequency of such disorders did not
differ as a function of hemispheric side of lesion. Lacking comparative
data in patients with lesions limited to the right or left basal ganglia and
the right or left cerebral cortex. the differential contribution of unilateral
cortical versus subcortical structures to prosody remains uncertain. Ina
previous study, Blonder, Gur, and Gur (1989) found expressive prosodic
deficits in patients with both right and left hemiparkinsonism, suggesting
that the neural control of intonation may be bilateral at the level of the
basal ganglia. This is consistent with an evolutionary model in that non-
human primate calls, which resemble intonation in acoustic structure, are
mediated by bilateral, subcortical brain regions (Jurgens 1979).

Several studies suggest that aprosody following RHD is emotional in
nature (Tucker et al., 1977; Ross and Mesulam, 1979; Borod et al., 1985;
Ross et al., 1988). However, these investigators did not assess the ability
to produce non-emotional prosody. In a study using observer ratings to
examine the effects of RHD on linguistic prosody, Weintraub and col-
leagues (1981) found that RHD patients were impaired in the repetition
of statement-question intonation and the production of emphatic stress.
Likewise, Danly and Shapiro (1985) used acoustic analysis to examine
patients ability to produce interrogative. declarative, happy or sad pros-
ody in sentences that were content neutral. They found that anterior
RHD patients spoke with less fo variability while posterior RHD patients
produced increased fo variability across both emotional and linguistic
utterances.

Our prosodic data derive from interviews concerning the patient’s
early-life experiences. We surmise that the recounting of this information
evoked a range of feelings in the speaker, and that these feelings should
have been conveyed in part through prosody. However, the material is
not structured so as to permit us to separate emotional and linguistic
prosody. Furthermore, spontaneous discourse is rarely devoid of speaker
attitude, and speaker attitude is more or less emotionally charged (Cou-
per-Kuhlen, 1986). Many linguists challenge the view that a dichotomy
PROSODIC CHARACTERISTICS 333

exists between emotional and linguistic prosody (Bolinger, 1972; Ladd,
1980). For example, Bolinger (1972, p. 233) dispels the notion that gram-
matical and emotional intonation are distinct by pointing out that ques-
tions are an expression of a speaker’s uncertainty or curiosity, “‘which
is suspiciously close to emotion.’’ Ladd (1980, p. 109) hypothesizes that
“intonation meaning involves both all-or-none contrasts between linguis-
tic categories and dimensions of gradiance within categories.”’ In gradiant
phenomena, increases or decreases in fo range, loudness, and tempo tend
to correspond to increases or decreases in the speaker's feelings about
the subject matter under discussion (see Bolinger, 1986).

In a spectrographic study of lexical stress in compound words and
noun phrases, Emmorey (1987) found that RHD patients’ utterances were
intact. Likewise, Behrens (1988) studied contrastive stress production
using acoustic quantifiers and found that RHD patients were able to pro-
duce stress contrasts. These studies suggest that all-or-none contrastive
intonation is spared in RHD. Our study and those by Kent and Rosenbek
(1982) and Behrens (1989) suggest that gradiant aspects of intonation are
disrupted in RHD. Since gradiant properties of intonation are often used
to express emotion, the restriction variability in fo and pause duration
that we report, may reflect a stroke-induced restriction in the patient's
capacity to experience emotion. However, this is unlikely in P.C.’s case
since she did not exhibit post-stroke depression according to her score
on the Zung Depression Scale, nor did she display a reduction in her
capacity to experience positive and negative emotions as indexed by her
PANAS scores.

The characteristics of prosody that we have identified as impaired in
this patient following her right hemisphere infarct, are precisely those
used by speakers to create drama in speech. Thus, the prosodic disorder
that we have observed may signify a loss of knowledge regarding the
appropriate use of these cues to convey emotional or attitudinal meaning.
Alternatively, this impairment could simply reflect a mechanical ineffi-
ciency on the part of the speaker to regulate fo, rate, and pause. which
in turn results in a failure to arouse emotions in the listener. This would
explain the repeated attempts by researchers and clinicians to character-
ize monotone speech in RHD patients as an aspect of affective flattening.

The findings derived from this case study point to several avenues of
future research. These include case-control studies of speech rate and
pause duration in spontaneous discourse by RHD versus LHD patients,
listener arousal in response to RHD patients’ spontaneous prosody; the
relationship among RHD patients’ emotional state, prosody, and laryn-
geal function; and comparisons of these functions in patients with discrete
lesions in varying regions of right and left cortical and subcortical struc-
tures.
334 BLONDER ET AL.

REFERENCES

Behrens, S. J. 1988. The role of the right hemisphere in the production of linguistic stress.
Brain and Language, 33, 104-127.

Behrens, S. }. 1989. Characterizing sentence intonation in a right hemisphere-damaged
population. Brain and Language, 37, 181-200.

Blonder, L. X., Gur, R. E.. & Gur, R. C. 1989. The effects of right and left hemiparkinson-
ism on prosody. Brain and Language, 36, 193-207.

Blonder, L. X., Bowers, D., & Heilman, K. M. 1991. The role of the right hemisphere in
emotional communication. Brain, 114, 1115~1127,

Blonder, L. X.. Kort, E. D., & Schmitt. F. A. 1994, Conversational discourse in patients
with Alzheimer’s disease. Journal of Linguistic Anthropology, 4 (1), 50-71.

Bolinger, D. 1972. Jrztonation. Harmondsworth, England: Penguin.

Bolinger, D. 1986. Intonation and its parts. Stanford: Stanford Univ. Press.

Borod, J. C., Koff, E.. Lorch, M. P., & Nicholas, M. 1985. Channels of emotional expres-
sion in patients with unilateral brain damage. Archives of Neurology 42, 345-348.
Bowers, D., Blonder, L. X., & Heilman, K. M. 1991. The Florida affect battery. Gaines-

ville, FL: University of Florida Press.

Cancelliere, A. E. B.. & Kertesz, A. 1990. Lesion localization in acquired deficits of emo-
tional expression and comprehension. Brain and Cognition, 13, 133-147.

Cooper, W., Soares. C.. Nicol. J.. Michelow, D., & Goloskie, S. 1984. Clausal intonation
after unilateral brain damage. Language and Speech, 27, (7-24,

Couper-Kuhlen, E, 1986. An introduction to English prosody. London: Edward Arnold.

Crystal, D. 1992. A dictionary of linguistics and phonetics, Oxford, UK: Blackwell.

Damasio, H., & Damasio, A. R. 1989. Lesion analysis in neuropsychology. New York:
Oxford.

Darley, F., Aronson. A., & Brown, J. 1969. Clusters of deviant speech dimensions in the
dysarthrias. Journal of Speech and Hearing Research, 12, 462-496,

Emmorey, K. D. 1987. The neurological substrates for prosodic aspects of speech. Brain
and Language, 30, 305-320.

Folstein, M. F.. Folstein, S. E., & McHugh, P. R. 1975, Mini-mental state. Journal of
Psychiatric Research, 12, 189-198,

Heilman. K., Bowers, D.. Valenstein, E., & Watson, R. 1983. Florida neglect battery.
Unpublished test, University of Florida.

Hughes, C. P., Chan, J. L., & Su, M. S. 1983. Aprosedia in Chinese patients with right
cerebral hemisphere lesions. Archives of Neirology, 40, 723-736.

Jurgens, U. 1979. Neural control of vocalization in nonhuman primates. In H.D. Steklis &
M. J. Raleigh (Eds.), Neurobiology of social communication in primates. New York:
Academic Press. Pp. 11-44.

Kazdin, A. E. 1982. Single-case research designs. New York: Oxford Univ. Press.

Kent, R. D., & Rosenbek, J. C. 1982. Prosodic disturbance and neurologic lesion. Brain
and Language, 18, 259-291.

Ladd, D. R. 1980. The structure of intonation meaning. Bloomington, IN: Indiana Univ.
Press.

Lieberman, P., & Blumstein, S. E, 1988. Speech physiology, speech perception, and acous-
tic phonetics, Cambridge: Cambridge Univ. Press.

Matarazzo, J. D., & Wiens, A. N. 1972. The interview. Research on its anatomy and
structure. Chicago/New York: Aldine.

Matsui, T., & Hirano, A. 1978. An atlas of human brain for computerized tomography.
Tokyo: Igaku-Shoin.

Monrad-Krohn, G. 1963. The third element of speech: prosody and its disorders. In L.
Halpern (Ed.), Problems ef Dynamic Neurology. Jerusalem: Hebrew Univ. Press. Pp.
101-117.

PROSODIC CHARACTERISTICS 335

Odell, K.. McNeill, M. R., Rosenbek. J. C.. & Hunter. L. 1991. Perceptual characteristics
of vowel and prosody production in apraxic, aphasic, and dysarthric speakers. Journal
of Speech and Hearing Research, 46, 67-80.

Ross. E. D. 1981. The aprosodias: Functional-anatomic organization of the affective compo-
nents of language in the right hemisphere. Annals of Neurology, 38, 561-589.

Ross. E. D.. & Mesulam, M. M. 1979. Dominant language functions of the right hemi-
sphere? Prosody and emotional gesturing. Archives of Neurology, 36, 144-148.

Ross. E. D., Edmondson, J. A., Seibert. G. B., & Homan, R. W. 1988. Acoustic analysis
of affective prosody during right-sided Wada test: A within-subjects verification of the
right hemisphere’s role in language. Brain and Language, 33, 128-145.

Ryalls. J. 1982. Intonation in Broca’s aphasia. Neuropsyvchologia 20, 355-360.

Ryalls. J. 1986. What constitutes a primary disturbance of speech prosody? A reply to
Shapiro and Danly. Brain and Language, 29, 183-187.

Ryalls, J.. Joanette. Y., & Feldman, L. 1987. An acoustic comparison of normal and right
hemisphere damaged speech prosody. Cortex, 23, 685-694.

Ryalls. J... & Behrens. S. J. 1988. An overview of changes in fundamental frequency associ-
ated with cortical insult. Aphasiology, 2, 107-115.

Shapiro, B. E.. & Danly, M. 1985S. The role of the right hemisphere in the control of speech
prosody in propositional and affective contexts. Brain and Language, 25, 19-36.
Speedie. L. J., Coslett. H. B.. & Heilman. K. M. 1984. Repetition of affective prosody in

mixed transcortical aphasia. Archives of Neurology, 41, 268-270.

Templin, M. 1957. Certain language skills in children. Their development and interrelation-
ship. Minneapolis: Univ. of Minnesota Press.

Tucker, D., Watson, R., & Heilman, K. 1977. Discrimination and evocation of affectivelty
intoned speech in patients with right parietal disease. Neurology, 27, 947-950.

Watson, D., Clark, L. A., & Tellegan, A. 1988. Development and validation of brief mea-
sures of positive and negalive affect: The PANAS scales. Journal of Personality and
Social Psychology, 54, 1063-1070.

Weintraub. S.. Mesulam, M., & Kramer. L. 1981. Disturbances in prosody: a right hemi-
sphere contribution to language. Archives of Neurology, 38, 742-744.

Zung., W. W. K. 1965. A self-rating depression scale. Archives of General Psychiatry, 12,
63-70.