Académique Documents
Professionnel Documents
Culture Documents
AND
LANGUAGE
E. COOPER
AND
BARBARA SHAPIRO
Five Wernickes aphasics and five normal control subjects were tested in order
to assess several aspects of fundamental frequency (F,,) in speech production.
The clinical impression of normal prosody in Wernickes aphasia is correct
inasmuch as these patients generally exhibited F,, declination. However, F,, declination ranged over shorter domains than in normal speech. Moreover. the
increased use of F,) continuation rises by the Wernickes aphasics indicated their
inability to maintain a single F,) contour over constituents which are normally
integral. The hypermelodic quality of F,, in the speech of Wernickes aphasics
further supported the notion that speech prosody was not strictly normal. F,)
attributes tended to be normal when they corresponded to the global linguistic
variable of sentence length, while they were abnormal when they corresponded
to the processing of syntactic structure. No evidence was found that paraphasias
and neologisms directly affected the programming of F,). The results are discussed
in terms of speech processing abilities and limitations in Wernickej aphasia.
This research was supported by NIH Grants NS 06029, NS 11408, NS 13028, and NS
15059, and a Dissertation Research Award to the first author from the Department of
Psychology and Social Relations, Harvard University. The authors gratefully acknowledge
Suzanne Hamby for assistance in testing subjects: John M. Sorensen for consulting on
acoustical matters; Kenneth N. Stevens for making available the computer facilities of the
Speech Communication Group, Research Laboratory of Electronics, MIT; Kyle R. Cave
for advice on text editing; and Sheila E. Blumstein. Roger W. Brown, Jill G. de Villiers.
Susan F. Ehrlich, Charles M. Judd. Edgar B. Zurif, and two anonymous reviewers for
helpful comments on the manuscript. Please address reprint requests to: Martha Danly.
Psychology Service 116-B. Aphasia Research Center, Boston Veterans Administration
Medical Center. I50 South Huntington Avenue. Boston. MA 02130.
I
0093-934X183$3.00
Cop\r,ght
c IYU? hy Acadcn,,c
Prcr,. Ini
All rIghi\ 01 rcprmlucr~on
an <in> lwm ,r\cr\ed
DANLY,
COOPER,
AND
SHAPIRO
INTRODUCTION
WERNICKES
APHASIA
TIME
(secl
FIG. I. Fundamental frequency contour and Topline Rule predictions for the four-peak
sentence, The book on the table wa.7 a gift ,frorn my mother. (P 1 = book; P2 = table:
P3 = gift; P4 = mother.) The Topline Rule is not used to predict the first peak of the
contour, because PI consistently occurs above the declination line (Olive. 1975: Maeda.
1976: Cooper & Sorensen. 1981).
DANLY,
COOPER,
AND
SHAPIRO
patterns over relevant linguistic domains, and (2) the speakers ability
to control F0 despite paraphasias and neologisms.
F,, Resetting
In normal speech, a single declination line typically spans an entire
single-clause sentence. Within a given sentence with an established declination function, we define F, resetting as occurring at an observed F.
peak which is higher than the peak preceding it. For resetting to occur,
F, declination must be present both before and after resetting; in this
way, we exclude local perturbations in declination such as stress heightening. Resetting of the declination function normally occurs between
sentences or at clause boundaries of a multiclause sentence. As can be
seen in Fig. 2, F. resetting occurs on the word asleep, following the
relative clause. Note that the observed value of P3 is greater than the
value predicted on the basis of a single-function Topline Rule (solid line),
indicating F, resetting. For this sentence, it can be seen that F, values
are better predicted by two declination functions (broken lines) rather
than one.
Once established, F. declination can be used as a means to determine
the scope of linguistic planning in the following way. Since the F. declination
function requires some preplanning in order for the speaker to initiate
the sentence with an appropriate F. value, we may infer the organization
of linguistic units that coincide with the F, contour. For utterances in
which an intonation contour extends over the entire sentence, the preprogramming of the initial F,, value and declination indicates that, at least
160 -
70
I
1.0
I
30
I
2.0
TIME
I
4.0
I
5.0
(set)
FIG. 2. Fundamental frequency contour for the sentence, The car that Sally owned
was asleep on the large branch in the bee. (Pl = cat; P2 = owned; P3 = asleep; P4 =
branch; PS = tree.) The solid line represents the Topline Rule predictions for a single F,
declination function. Observe that P3 occurs above its predicted value, indicating F(, resetting.
The broken lines represent the Topline Rule predictions for two declination functions.
Note also the continuation
rise on the word owned at the end of the relative clause.
WERNICKES
APHASIA
DANLY,
COOPER,
AND
SHAPIRO
& Cooper, 1980). Because the final peak of sentences occurs at approximately the same value and the F,, declination slope remains fairly constant
in single-clause sentences, a higher Pl in long sentences allows the
speaker an adequate range for the declination of F0 values. The PI effect
is examined in this study to establish whether overall sentence length,
declination, and F0 programming are properly coordinated in the speech
of Wernickes aphasics.
Sentence-Final F. Fall
In declarative sentences, the largest F, fall on a given word occurs
on the sentence-final word (Lea, 1973; Maeda, 1976). Not only does this
F. fall indicate the end of the utterance, but the minor falls which precede
it reveal the speakers intention to continue the utterance. As a demonstration that Brocas aphasics plan speech in units larger than the
single word, Danly and Shapiro (1982) showed that Brocas aphasics
produce the largest F. fall in sentence-terminal position. A similar analysis
will be undertaken here in order to examine the scope of speech planning
in Wernickes aphasia.
F. Variability and Speech Rate
To study further the degree to which Wernickes speech is prosodically
normal, we will determine whether normal subjects and Wemickes aphasics
exhibit similar valley-peak-valley
change in F,, (F,, variability). These
measurements are taken by summing the F,, change from each measured
words initial F. value (valley) to its greatest value (peak) to its final
value (valley); any additional variation due to a continuation rise is also
included. Acoustical analysis may reveal a difference between normal
and aphasic subjects that cannot be detected in subjective analysis.
As for speech rate, two recent studies have shown that speech fluency
(Kreindler, Mihailescu, & Fradis, 1980) and verbal rate (Deloche, JeanLouis, & Seron, 1979) call for more complex measures than simply the
number of words uttered per unit of time. They argue that in order to
characterize adequately the various forms of aphasia, a unidimensional
measure of speech rate is insufficient. Consideration of variables such
as overall speaking time, total number of words, number of silent pauses,
and mean duration of silent pauses is also required to provide a model
of speech rate that properly differentiates the types of aphasia. In the
present study we will relate the above observations to the relevant acoustic
data on speech timing in the reading of Wernickes aphasics. In addition,
we will comment on the relationship between speech timing and paraphasia.
As Butterworth (1979) has documented, paraphasias and neologisms are
more likely than real words to follow hesitation pauses in the spontaneous
speech of a jargon aphasic, indicating word-finding difficulty; we will
determine whether this finding extends to the present subject population
using a practiced reading paradigm.
WERNICKES
APHASIA
METHOD
Subjects
A total of IO subjects participated in this study. These included 5 right-handed male
Wernickes aphasics. ranging in age from 47 to 61. All subjects were tested at the Boston
Veterans Administration Hospital: they were the same subjects for which preliminary
results were reported in Cooper et al. (1979). Diagnoses were based on the Boston Diagnostic
Aphasia Examination (BDAE) (Goodglass & Kaplan, 1972). Each diagnosis was verified
by a CT scan which, for all 5 aphasics, showed a localized lesion in the left posterior
cerebral hemisphere due to an infarction of the brain tissue in the territory of the middle
cerebral artery. The degree of linguistic impairment in the aphasic subjects revealed a
range of severity from moderately to severely disordered. In oral reading, the least impaired
subject produced paraphasias and neologisms in 1% of the content words. while the most
impaired subject produced paraphasias and neologisms in about 90% of the content words.
Five normal male subjects were also tested. Matched in age, educational level. and
handedness to the experimental subjects, they were Boston-area residents who volunteered
to participate in a variety of psychology experiments. In addition, we tested five patients
with either anterior or posterior right-hemisphere damage. Our aim was to study neurological
control subjects whose brain damage was in the nondominant hemisphere. However, upon
analyzing the F,, patterns produced by the right-hemisphere patients, it was evident that
they exhibited certain gross abnormalities of prosody that would require explanations
distinct from those within the scope of this paper. Hence. the data gathered from these
subjects will receive separate treatment (Danly. Gardner. & Shapiro. 1982).
Sentence Materiuls
The sentences listed in Table I, taken from Cooper and Sorensen (lY81), were designed
to test the hypotheses outlined in the Introduction. The italicized key words were measured
to test F,) declination and F,, resetting. In sentences l-4 we measured three peaks, in
sentences 5-8 four peaks, and in sentences Y-12 five peaks. These sentences were designated
as short, medium. and long sentences, respectively. Notice that the three-peak and fivepeak sentences are paired to test the PI effect for short vs. long sentences. The order of
TABLE
and peanuts
9. The deer by the canyon could be Seen from the ;c+doM, of the car
10. The host at the parry could have packed all the glusses with the plafes.
II. The cat that Sally ottaned was asleep on the large branch in the free.
12. The
DANLY,
the sentences was randomized; the sentences were then presented to subjects with different
pseudorandom orders shown to alternate subjects. Two filler sentences appeared as the
first and last sentences of each list to eliminate the possibility of starting and finishing
effects. The sentences contained no punctuation other than sentence-final periods.
Procedure
The subjects were tested individually in a sound-insulated room. At the beginning of
each session, the experimenter told the subject that he would be given a list of sentences
to read first silently to himself and then aloud. After a practice oral reading of each sentence,
during which the experimenter monitored the recording levels and checked for emphatic
or contrastive stress, the subject read the sentence aloud for recording. On occasion,
particularly for the more disordered aphasics, the first recorded token included words that
were emphatically or contrastively stressed. If so, the experimenter asked the subject to
repeat the sentence, providing a verbal model of the sentence without such stress; the
subject was then asked to produce the sentence a second time for recording. The Wernickes
subjects appeared unable to avoid the use of stress in some sentence tokens despite their
ability to repeat single words in isolation without emphasis. In such cases we analyzed
the sentence token containing the least emphatic or contrastive stress.
Paraphasias and neologisms produced by the Wernickes subjects were not corrected
by the experimenter, as we were particularly interested in capturing F,) patterns in the
context of paraphasic and neologistic output. The subjects read each sentence aloud until
at least one clearly articulated version of the completed sentence was produced. All
utterances were recorded onto a Sony TC106A tape recorder via a Sony F-25 microphone.
Acoustical
Analysis
RESULTS
WERNICKES
APHASIA
TABLE 2
SUMMARY
OF
F,, MEASURES
FOR NORMAL
F,) measure
Normal subjects
(N = 5)
APHASICS
Wernickes aphasics
(N = 5)
I. F,, declination
12.7 Hz/peak
2. F,, resetting
Short sentences
Medium sentences
Long sentences
0%
0%
60%
5%
55%
SO%
30%
90%
50%
50%
N = 12
N=2
9.9 Hz/peak
N = 22
N = 18
0%
S%
50%
17%
45%
68%
30%
70%
80%
50%
N = II
N = 2
6.2 Hz
N = 26
N = 25
15.0 Hz
139.6 Hz
145.4 Hz
146.8 Hz
152.6 Hz
- 13.2 Hz
12.0 Hz
13.0 Hz
26.4 Hz
17.6 Hz
11.6 Hz
6.0 Hz
45.6 Hz
22.8 Hz
45.8 Hz
F, Declination
10
DANLY,
COOPER,
AND
SHAPIRO
to compare the slopes and goodness-of-fit measures with those for normal
subjects. But since the speech rate of normal subjects was almost twice
that of the aphasic subjects (the aphasics took an average of 6.54 set to
produce each sentence, while the normals took 3.35 set), we considered
a direct comparison of the declination slopes to be inappropriate. To
avoid the bias due to speech rate differences, we instead regressed F,
on the key words (peaks), thereby neutralizing the factor of time.
Generalizing across sentence length, we found that the Wernickes
aphasics produced a significant declination slope, of 9.9 Hz/peak: t(53)
= 8.87, p < .OOl, r = .77. For the normal subjects, the declination
slope was 12.7 Hz/peak (t(53) = 14.70, p < .OOl, Y = .90), which does
not differ significantly from the Wernickes declination slope (t( 115) =
1.38, p = .17, Y = .13). These results indicate that Wernickes aphasics
adequately program F. declination over sentence strings and that the
total decrease in F. is comparable to normal declination. This particular
finding coincides with the clinical impression that Wernickes aphasics
do not suffer from a disorder of speech prosody.
While the Wernickes aphasics, taken together, exhibited F. declination,
it is noteworthy that this effect is primarily attributable to four of the
five aphasic subjects (SM, WH, LMc, and ES), whose declination slopes
ranged from 10.2 to 17.0 Hz/peak. The fifth subject, RD, who was the
most severely aphasic of the group, produced virtually no declination
whatsoever, at 0.9 Hz/peak: t(9) = .78, p > .40, Y = .25. From this
observation we may conclude that F, declination is spared in Wernickes
aphasia, except perhaps in the severest of cases.
Having established F, declination in the aphasic subjects, we now turn
to the role of paraphasia in the programming of Fo. One might consider
that a disorder in the lexical selection process, exhibited as a paraphasia
or neologism, might disrupt the maintenance of F. declination. In keeping
with such a claim is the fact that RD, the most severely aphasic of the
group, produced virtually no declination and was almost entirely paraphasic
and neologistic (88% of the words). However, since practically all of
RDs content words were either paraphasic or neologistic, his speech
does not allow us to directly test the effect of paraphasia on declination;
likewise there is a similar lack of variability in subjects ES, WH, and
SM, whose paraphasias contributed to no more than 10% of their output.
Therefore, we focused on the speech of LMc, who produced verbal
paraphasias and neologisms in approximately half (54%) of the content
words. We considered two possible ways in which these errors might
affect declination: (1) the paraphasias and neologisms are consistently
accompanied by F, peaks that fall either above or below the declination
line (but not both), or (2) F. peaks of errors generally miss the declination
function more than the peaks of correct words. For this analysis, we
combined verbal paraphasias with neologisms into a single category of
WERNICKES
APHASIA
II
lexical selection errors, which we will call paraphasia in this analysis for
the sake of brevity.
To test these hypotheses we performed two sets of multiple regressions.
For the first, we regressed the F0 of each key word produced by LMc
on peak number, sentence length, and paraphasia. The results showed
that, when the other two independent variables are controlled for, paraphasias had no consistent additive or subtractive effect on F, (b = -7.9
Hz, t(44) = - 1.63, p > .lO). The second analysis was implemented in
two steps. We first regressed the F, of each key word on peak number
and sentence length. The resultant residuals were squared to eliminate
the negative values and then regressed on paraphasia. If the second
hypothesis were true, we would expect to find the large residuals paired
with the paraphasic words. Yet this second analysis revealed that the
paraphasias were no more inaccurately programmed for F,, than correct
words (t(46) = 1.23, p > .20, r = . IS). This last result suggests that,
at least in terms of F,, declination programming, errors in lexical selection
do not adversely affect intonation.
The Topline Rule as a predictor of F, is useful at this point in the
analysis. Having demonstrated F, declination and its lack of interaction
with paraphasia, we now turn to a more subtle comparison of F,, patterns
in normal and Wernickes subjects. As mentioned in the Introduction,
the Topline Rule mathematically describes the relation between F0 and
time and is used to predict F, peak values between the initial and final
peaks of a given sentence. One use of the Topline Rule has been to test
the invariance of F, across a wide range of speaking variables. such as
speakers sex, speaking rate, and grammatical category of the measured
words (Cooper & Sorensen, 1981). The rules utility is also seen in the
analysis of F, resetting, which occurs when observed peak values are
greater than those predicted on the basis of the Topline Rule. However,
the Topline Rule presupposes declination as a property of the F,, peaks:
therefore it is not an ideal test of the declination slope per se.
Figure 3 shows the Topline Rule predictions for the four-peak and
five-peak sentences, along with the actual peak values produced by both
groups of subjects. Observe that F, resetting appears to occur at the
subject-predicate boundary (between P2 and P3) for both the Wernickes
and normal subjects in the five-peak sentences, yet only for the aphasic
It also should be noted that despite the relationship noted between extreme paraphasia
and the loss of declination in patient RD. the individual declination slopes of the remaining
four aphasic subjects were unrelated to the number of paraphasias that each subject
produced.
The declination slope for an N-peak sentence as defined by the Topline Rule is: 213
[(PN - Pl)/(TN
- TI)]. From this formula it is clear that if PI equals PN, the slope of
the declination line is zero. Thus the Topline Rule cannot be used to establish whether
or not the peaks in a given utterance actually decline: rather, it is used to predict peaks
intermediate to the initial and final peaks.
12
DANLY,
160 150 ;;
SENTENCES
:\
140 -
130 -
G
E
120 -
110 -
ks
&I.
100 -
160 -
150 -
ii
Q
140
2
3
Y
130 -
120 II0
100 -
Wernickes
Aphasics
Normal
Subjects
5-PEAK
SENTENCES
\
Y\
h;;\zts
PI
P2
P3
PEAK
P4
Subjects
P5
NUMBER
FIG. 3. F0 declination functions for the four-peak and five-peak sentences produced by
normal speakers and Wernickes aphasics. The Topline Rule indicates similar declination
slopes for the two groups of subjects. In addition, note that F,) resetting occurs at P3 for
both the Wernickes aphasics and the normal speakers in the long sentences, but only for
the aphasic subjects in the medium-length sentences (see text).
WERNICKES
APHASIA
Normal
Speakers
Wernickes
Aphasics
/
short
medium
SENTENCE
long
LENGTH
of short,
14
DANLY,
COOPER,
AND
SHAPIRO
7E 100
_
pz 00-
Normal
Speakers
L
E
a
60
--
----q
Wernickes
b 40
Aphasics
;/
z
it?
2 20
2
E
,
OPrepositional Phrase
BOUNDARY
,
Relative Clause
TYPE
WERNICKES
APHASIA
=
6FBO_
;
60-
F
6
u
40-
Wernickes
Aphasics
/
/
/ -Normal
/
Speakers
I
long
medium
short
SENTENCE
LENGTH
subjects indicates that they are generally more likely than normal speakers
to intonationally break up constituents.
Evidence for the aphasics lack of syntactic differentiation is shown
in Fig. 7 by the frequency of continuation rises in contrasting boundaries.
Examining the occurrence of continuation rises at prepositional and relative
clause boundaries in the five-peak sentences, it is evident that normal
speakers produced more continuation rises at the larger of the two syntactic
boundaries (the relative clause), whereas the Wernickes aphasics did
not. This interaction was tested in a two-way repeated-measures ANOVA
(subject type x boundary type), which revealed that the normal and
2
100 -
2
vr
a
k!
BO-
::
E
NOVlKll
Speakers
o- Prepositional
I
Phrase
BOUNDARY
Relative
Clause
TYPE
16
DANLY,
COOPER,
AND
SHAPIRO
aphasic subjects tended to treat the contrast between the two boundaries
in different manners: F(1, 8) = 3.92, p = .08, n2 = .35. While this
interaction is not significant at (Y = .05, the effect size is relatively large.
Continuation rises, like F0 resetting, rarely occur at clause-internal
locations in normal speech. When they consistently occur at such locations,
they may reflect a defect in or the absence of a syntactic analysis of the
sentence. Table 2 compares the number of continuation rises normal and
aphasic subjects produce at the subject-predicate boundary (appropriate)
vs. other locations (inappropriate). The Wernickes aphasics were significantly more likely than normal subjects to produce continuation rises
at minor, within-clause boundaries (t(8) = 2.92, p < .02, Y = .72). This
finding allows two interpretations. One, the aphasics produced more
continuation rises at smaller though reasonable boundaries; that is,
their threshold for production of continuation rises is lower than normal,
although some syntactic hierarchy of boundaries exists, however rudimentary. Two, continuation rises occurred randomly within sentences,
without regard to syntactic boundary strength. For the five-peak sentences,
there was a total of 20 measured words, 16 of which occurred in positions
at which it is reasonable to produce continuation rises-that is, nonsentencefinal position. Of these 16 key words, 4 (25%) occur between the subject
and predicate, the strongest syntactic boundary. In these five-peak sentences, the Wernickes aphasics produced a total of 25 continuation rises,
14 (56%) of which occurred at the subject-predicate boundary, more
often than predicted by random placement. A matched-pairs t test comparing the frequency of continuation rises at major vs. minor boundaries
showed that they were significantly more likely to occur at the major
boundary (t(4) = 2.81, p < .05, Y = .64). In a parallel analysis of F0
resetting, we found that the Wernickes aphasics also reset intonation
contours significantly more often at the largest syntactic boundary of the
five-peak sentences: t(4) = 7.50, p < .Ol, r = .81. Thus, the findings
for F0 resetting and continuation rises, taken together, show that Wernickes
aphasics produce F0 patterns that reflect the distinction between major
and minor syntactic boundaries, though clear deficits exist in the processing
of complex structure.
The PI Effect
The results of the test of the PI effect revealed that the Wernickes
aphasics were able to combine the factors of sentence length and declination
in order to produce higher values of Pl for longer sentences. Both the
control and aphasic subjects produced PI values that were 5.8 Hz higher
for the long vs. the short sentences, which is a significant main effect
for sentence length in a two-way repeated-measures ANOVA (sentence
WERNICKES
APHASIA
17
F,, Full
18
DANLY,
COOPER,
AND
SHAPIRO
WERNICKES
APHASIA
19
observation matches our findings for this subject. Compared to real words,
neologisms were significantly more likely to be preceded by pauses of
250 msec or greater: x2(1) = 4.80; p < .05; 4 = .38. We therefore
replicate Butterworths finding for neologisms with a Wernickes aphasic
in the practiced reading paradigm. Moreover, we establish a dissociation
between the programming of two prosodic features: speech timing and
F0 declination (upon which neologisms apparently had no effect).S This
difference between the two prosodic features indicates the independence
of their relationships to lexical selection in Wernickes aphasia.
DISCUSSION
20
DANLY,
COOPER,
AND
SHAPIRO
WERNICKES
APHASIA
21
22
DANLY,
COOPER,
AND
SHAPIRO
Blumstein,
Blumstein,
WERNICKES
23
APHASIA
Buckingham. H. W.. & Kertesz. A. 1974. A linguistic analysis of fluent aphasia. Brain
and Lunguqe,
1, 43-62.
Butterworth. B. 1979. Hesitation and the production of verbal paraphasias and neologisms
in jargon aphasia. Bruin rind Lunguuge.
8, 133-161.
Cohen. A., & tHart. J. 1967. On the anatomy of intonation. Linguu. 19, 177-192.
Cooper, W. E.. Danly, M.. & Hamby, S. 1979. Fundamental frequency (F,,) attributes in
the speech of Wernickes aphasics. In J. J. Wolf & D. H. Klatt (Eds.). Speec,h
c~or,lmunicrrtion
America.
New
pupcrs
presenred
ut the 97th
Mcrting
of tiw
Acorr.slicu/
Society
of
P.tvc,l~o/o,~ic,tr/
B7rllctin.
72, 273-276.
Freidman. H. 1968. Magnitude of experimental effect and a table for its rapid estimation.
P.sycho/o,qicu/
Bullrtin,
70, 245-25 I.
Gibson. E. J.. & Levin, H. 1975. The psychology
ofreuding.
Cambridge, MA: MIT Press.
Goodglass. H., & Geschwind. N. 1976. Language disorders (aphasia). In E. C. Carterette
Vol. VIII: Lang17ugc~
und .sprd7.
& M. F. Friedman (Eds.), Handbook oJperc,ep/ion.
New York: Academic Press. Pp. 389-428.
Goodglass, H.. & Kaplan, E. 1972. The ussessrnent of uphmitr trnd rclrrted disordc,rs.
Philadelphia: Lea & Febiger.
Green, E. 1969. Phonological and grammatical aspects of jargon in an aphasic patient: A
case study. Lunguuge und Speech, 12, 103-l 18.
Hecaen. H.. & Albert, M. L. 1978. Hun7ur7
nrl,rop.~vc,ho/o,~~.
New York: Wiley.
Howes, D. 1964. Application of the word frequency concept to aphasia. In A. V. S. De
of lunguuge. London: Churchill.
Rueck & M. OConnor (Eds.), Disorders
Kreindler, A., Calvarezo, C.. & Mihailescu, L. 1971. Linguistic analysis of one case of
jargon aphasia. Rr1,rte romuine de nerrrologie,
8, 209-228.
Kreindler, A.. Mihailescu, L.. & Fradis. A. 1980. Speech fluency in aphasics. Brtrin und
Lunguuge.
9, 199-205.
Lea, W. A. 1973. Segmental and suprasegmental influences on fundamental frequency
contours. In L. M. Hyman (Ed.), Consonunr fypes und tone. Los Angeles: U. S. C.
Press.
Lecours. A., & Rouillon, F. 1976.Neurolinguistic analysis ofjargonaphasia andjargonagraphia.
In H. Whitaker & H. A. Whitaker (Eds.), Srrrdics in nerrrolingctistics.
New York:
Academic Press. Vol. 2.
Lehiste, 1. 1970. Supruscgtnentuls.
Cambridge. MA: MIT Press.
Lesser, R. 197X. Lingui.stic, irll.e.stigutions
o.fuphmsiu. New York: Elsevier.
24
DANLY,
Lieberman, P. 1967. Intonation, perception, and language. Cambridge, MA: MIT Press.
Maeda, S. 1976. A characterization ofEnglish intonation. Unpublished Ph.D. dissertation,
Massachusetts Institute of Technology.
McAllister, R. 1971. Predicting physical aspects of English stress. Quarter/y Progress
Report, Speech Transmission Laboratory, Royal Institute of Technology. Stockholm,
Sweden, 1, 20-29.
Olive, J. P. 1975. Fundamental frequency rules for the synthesis of simple declarative
sentences. Journal of the Acoustical Society of America, 51, 273-276.
OShaughnessy, D. 1976. Modelling fundamental frequency and its relationship to syntax,
semantics, and phonetics. Unpublished Ph.D. dissertation, Massachusetts Institute of
Technology.
OShaughnessy, D. 1979. Linguistic features in fundamental frequency patterns. Journal
of Phonetics, 7, 119-145.
Parisi, D., & Pizzamiglio, L. 1970. Syntactic comprehension in aphasia. Cortex, 6, 204215.
Pierrehumbert, J. 1979. Intonation synthesis based on metrical grids. In J. J. Wolf & D.
H. Klatt (Eds.), Speech communication papers presented uf the 97th Meeting of the
Acoustical Society of America. New York: Acoustical Society of America.
Rinnert, C., & Whitaker, H. A. 1973. Semantic confusions by aphasic patients. Cortex,
9, 56-81.
Sorensen, J., & Cooper, W. E. 1980. Syntactic coding of fundamental frequency in speech
production. In R. A. Cole (Ed.), Perception andproduction affluent speech. Hillsdale,
NJ: Erlbaum.
Zurif, E. B., & Caramazza, A. 1976. Psycholinguistic structures in aphasia: Studies in
syntax and semantics. In H. Whitaker & H. Whitaker (Eds.), Studies in nemolinguistics.
New York: Academic press. Vol. 1, pp. 261-292.