BRAIN AND LANGUAGE 37, 232-257 (1989) Dysprosody in Broca’s Aphasia: A Case Study JACK GANDOUR AND SORANEE HOLASUIT PETTY Purdue University AND ROCHANA DARDARANANDA Mahidol University, Bangkok, Thailand A detailed acoustic analysis of timing, intensity, and fundamental frequency (F,,) at different levels of linguistic structure was conducted on the speech output of a Broca’s aphasic who was a native speaker of Thai. Timing was measured with respect to syllables, phrases, and sentences in connected speech. Intensity variation at the sentence level was measured in connected speech. FO variation associated with the five Thai tones was measured in both isolated words and connected speech. Results indicated that timing was differentially impaired depending upon complexity of articulatory gesture and size of the linguistic structure. Timing, as well as intensity, was aberrant at the sentence level. In contrast, F,, contours of the five tones were spared at all levels of linguistic structure. Findings are interpreted to support the view that dysprosody in Broca’s aphasia is more applicable to speech timing than to FO. o 1989 Academic PWSS, hc. Prosodic features of speech include three acoustical properties: timing, intensity, and fundamental frequency (F,,). The speech output of a Broca’s aphasic has traditionally been described as dysprosodic (Benson, 1979; Goodglass & Kaplan, 1983). The clinical use of the term has genThis material is based upon work supported by the National Institute of Neurological and Communicative Disorders and Stroke under Grant No. NS-24539 and the National Science Foundation under Grant No. BNS-7905854. The speech data were collected under the authority of the National Research Council of Thailand during the period from May to December 1980 while the first author was affiliated with the Division of Communication Disorders, Faculty of Medicine, Ramathibodi Hospital, Mahidol University, Bangkok, Thailand. We thank Poonpit Amatyakul (Speech and Hearing Clinic, Ramathihodi Hospital) and Athasit Vejjajiva (Neurology Unit, Ramathibodi Hospital) for their cooperation in this research. Requests for reprints should be sent to Jack Gandour, Department of Audiology and Speech Sciences, Heavilon Hall, Purdue University, West Lafayette, IN 47907. 0093-934X/89 $3.00 Copyright All rights 0 1989 by Academic Press, Inc. of reproduction in any form reserved. 232 DYSPROSODY IN BROCA’S APHASIA 233 erally connoted a deficit in all aspects of prosody with perhaps a primary deficit in the programming of F,,. However, recent findings from acoustical analysis of Broca’s speech have indicated that not all aspects of prosody are equally disturbed. While some aspects of prosody are impaired, others are clearly intact. In particular, speech timing appears to be relatively more disturbed than F, (Danly & Shapiro, 1982). Based on a review of the acoustic phonetic literature on Broca’s aphasia, Cooper and Klouda (1987) concluded that “as for the traditional clinical view that Broca’s aphasics exhibit dysprosody, it now appears that this description is more applicable to speech timing . . . than to fundamental frequency” (p. 66). In view of the possible dissociation of timing and F, in Broca’s aphasia, it is imperative that we design studies that incorporate acoustic measures of both at different levels of linguistic structure for the same subject. Without such acoustic measures, we cannot determine the extent to which timing and F,, or other aspects of prosody, may be selectively disrupted, as well as the extent to which the prosodic disruption varies as a function of the size of the linguistic structure. Intersubject variability makes it difficult to interpret the group performance of aphasic patients (Caramazza & Martin, 1983). By adopting the method of single case study, we are able to conduct an extensive analysis of Broca’s aphasia without the contaminating factor of heterogeneous variation. Danly and Shapiro (1982) hypothesized that the dissociation between FO and timing in English-speaking Broca’s aphasics resulted from their narrower-than-normal scope for linguistic planning. Because intonation production necessarily involves sentence- or phrase-sized linguistic units, the possibility remains that a deficit in F,, production occurs only when higher levels of linguistic structure are involved. To help clarify this issue, F. production needs to be investigated in word-sized units in order to eliminate that confounding variable. What is required is a tone language in which F. production is linguistically significant at the word or syllable level. If control of F. and timing are independent, then we might expect F,, patterns of tones to be preserved within syllables when they occur in polysyllabic words, phrases, and sentences in spite of abnormal timing between syllables. Such an acoustic phonetic investigation of speech prosody in Broca’s aphasia has yet to be completed in a tone language. As part of larger studies of consonant and vowel production in isolated words, timing data on word-initial stop consonants and short and long vowels in Thai have already been presented for a Broca’s aphasic (hereafter referred to as B; Gandour & Dardarananda, 1984a, 1984b). These earlier findings on consonant and vowel timing in isolated, monosyllabic words provide a point of departure for the studies on speech timing presented in this report. Thai is a language that has a maximum of three voicing categories of 234 GANDOUR, PETTY, DARDARANANDA stops in word-initial position: a three-way opposition (voiced unaspirated, voiceless unaspirated, voiceless aspirated) at the bilabial (/b, p, p”/) and alveolar (/d, t, t”/) places of articulation and a two-way opposition (voiceless unaspirated, voiceless aspirated) at the velar (/k, kh/) place of articulation. VOT values for /b, d/ lie in the long voicing lead region, for /p”, th, kh/ in the long voicing lag region, and for /p, t, k/ at zero or in the short voicing lag region (Lisker & Abramson, 1964). Gandour and Dardarananda (1984a) reported no overlap in VOT distribution between homorganic stops produced by normal speakers. B’s distribution of VOT values, however, do not fall into distinct, nonoverlapping regions. Her VOT continuum is considerably compressed. The majority of VOT values lie in the region at and just below zero. Her voiced unaspirated stops were produced with a shorter-than-normal voicing lead, voiceless aspirated stops with a shorter-than-normal voicing lag. These findings are consistent with previous reports on VOT production deficits in Broca’s aphasics or patients with apraxia of speech (e.g., Blumstein, Cooper, Goodglass, Statlender, & Gottlieb, 1980; Itoh, Sasanuma, Tatsumi, Murakami, Fukusako, & Suzuki, 1982; Ziegler & von Cramon, 1986). The production of VOT requires the coordination of two independent articulators, the tongue or lips and the larynx. The abnormal VOT patterns suggest that Broca’s aphasics have an impairment in the temporal coordination of the release of the stop closure and the onset of vocal cord vibration. However, as emphasized by Baum and Blumstein (1987), “it is not possible to determine on the basis of the VOT studies alone whether the deficit displayed by Broca’s aphasics reflects a pervasive phonetic disorder [in timing] . . . or whether it reflects a more selective timing impairment” (p. 10). In Thai, differences in the duration of short and long vowels are used phonemically (e.g., /hat/ “to practice” vs. /haat/ “beach”; /bbt/ “to grind” vs. /boot/ “temple”; /khun/ “you” vs. /khuun/ “to multiply”). Previous studies of the production of vowel length in normal adult speakers of Thai have shown that short and long vowel productions fall into distinct, nonoverlapping regions along the duration continuum (Abramson, 1962, 1974; Gandour, 1984). Gandour and Dardarananda (1984b) reported no overlap on the duration continuum for short (/a/) and long (/aa/) vowels produced by Thai normal speakers. Similarly, B’s short and long vowel productions were distributed into relatively distinct regions along the duration continuum with only sporadic occurrences of overlap. Her duration ratio of the means of long to short vowels (1.8) was robust. Albeit less than that of the four normal speakers (2.6) in Gandour and Dardarananda’s (1984b) study, her duration ratio was identical to the mean duration ratio that Roberson (1982, p. 95) reported for six normal speakers. DYSPROSODY IN BROCA’S APHASIA 235 These findings suggest that B’s timing of vowel duration is spared in Thai monosyllabic words by either an absolute or relative standard of measurement. Yet, B exhibited a timing deficit in the production of VOT associated with word-initial stop consonants. The stimuli in both the VOT and vowel length studies were monosyllabic. Thus, we cannot ascribe B’s inability to maintain timing control of VOT to variation in the size of the temporal domain. According to Baum and Blumstein (1987), Broca’s aphasics “seem to be particularly vulnerable to those phonetic parameters requiring the temporal integration of the movement of two independent articulators” (p. 17). Of previous studies of vowel duration in Broca’s aphasia or apraxia of speech, some (Collins, Rosenbek, & Wertz, 1983; Kent & Rosenbek, 1982, 1983; Ryalls, 1981; Shankweiler, Harris, & Taylor, 1968) have reported longer-than-normal vowel durations, while others (Duffy & Gawle, 1984; Ryalls, 1986) have found no significant lengthening of vowels. Ryalls (1987) has pointed out that the target words were polysyllabic in all the studies that found longer vowel durations for anterior patients. B’s vowel durations were not significantly different from normal. Indeed, her productions of /aa/ were, on average, somewhat shorter than those of normal speakers. The target words in the Thai study were monosyllabic. These combined findings suggest that, in addition to the complexity of articulatory implementation of temporal parameters, the degree of temporal disintegration in Broca’s aphasia is dependent upon the size of the linguistic unit over which temporal parameters are defined. In Thai, every syllable has a distinctive tone as a part of its phonemic makeup. Thai has five contrastive lexical tones on syllables ending in long vowels, diphthongs, or nasal consonants (Gandour, 1975), traditionally labeled mid (unmarked), low (‘), falling (^), high (‘), and rising (‘). The mid tone can be described as mid level with a final drop, the low tone as low falling, the falling tone as high falling, the high tone as high rising, and the rising tone as low rising (Abramson, 1962). As part of a larger study of tone production in isolated words, some F,, data on the five Thai tones produced by B have already been reported in Gandour, Petty, and Dardarananda (1988). These earlier findings on F0 in isolated, monosyllabic words provide a point of departure for the F0 analysis presented in the current study. Gandour et al. (1988) reported that listeners correctly identified the five Thai tones produced by normal speakers at a near-perfect level of accuracy. For B, the mean percentage of correct identification pooled across the five tones was 92%. The level of accuracy for all five, individual tones produced by normal speakers ranged from 95 to 100%. Of the five individual tones produced by B, mid, high, and rising were correctly identified at levels comparable to that of normals. Her productions of the low (84%) and falling (83%) tones were correctly identified 236 GANDOUR, PETTY, DARDARANANDA at levels only slightly lower than normal. Thus, B managed to achieve levels of proficiency in signaling tonal contrasts that were comparable to those of normal speakers. B’s pattern of tonal confusions differed from that of normal speakers primarily in degree rather than in kind. For the group of normal speakers, confusion between the mid and low tones accounted for 86% of the total number of listeners’ errors; for B, 35%. The overwhelming majority (71%) of B’s errors involved substitution of the mid tone for either low or falling. These findings indicate that B’s control of F0 in the production of tones in monosyllabic citation forms is relatively spared. If control of F0 and timing are independent, then we should expect F0 contours of B’s tones to be preserved within syllables when they occur in polysyllabic words, phrases, and sentences in spite of abnormal timing patterns between syllables. If so, these data would provide strong evidence in favor of the notion that dysprosody in Broca’s aphasia is not solely a manifestation of articulatory or phonatory deficits, but instead a manifestation of timing disturbances over larger-sized linguistic units. Only a few, acoustic investigations have been conducted on Thai tones in connected speech. F, contours of the five tones undergo perturbations induced by tonal coarticulation (Abramson, 1979a; Palmer, 1969) and sentence intonation (Abramson 1979b; Abramson & Svastikula, 1983). Although the shapes of the tonal contours are not identical to those that appear in monosyllabic citation forms, tonal contrasts are preserved. In the present report, we present further detailed acoustic analysis of timing, intensity, and F,, at different levels of linguistic structure for this same Thai-speaking Broca’s aphasic. Specifically, we focus on features of timing related to rhythmic patterns at the sentence level; features of intensity at the sentence level; and features of F, at the word and sentence levels. With this particular constellation of acoustic features, we are able to evaluate the extent to which various elements of prosody may be separately disrupted in Broca’s aphasia. In so doing, we bring evidence to bear on the notion that prosodic disturbance in Broca’s aphasia results primarily from a deficit in timing. CASE REPORT B, a 60-year-old right-handed Thai female, approximately 3 years poststroke, was referred to the Speech Clinic at Ramathibodi Hospital in June 1980 for speech therapy. On 3 June 1977, while eating dinner, she noticed her right arm was weak. Her right leg dragged when she walked. Examination revealed an alert, cooperative patient whose speech was markedly limited to a few words or short sentences. Her fundi showed slightly narrow and tortuous retinal vessels. Eye movements were full. There was upper motor neuron DYSPROSODY IN BROCA’S APHASIA 237 facial weakness of the right side. Her palate moved normally on phonation and the tongue was mid-line on protrusion. There was evidence of right hemiparesis with absent deep reflexes of the limbs and an extensor plantar response on that side. Sensory testing showed no abnormality. Her pulse rate was regular, 72/min, B.P. 160/100 mm Hg. There was a grade 2/6 ejection systolic murmur over the aortic area, but there were no arterial bruits audible in the neck. Other systems were otherwise normal. Investigations including urinalysis, EKG, Hb, Hct, WBC and differential count, BUN, blood Cr, uric acid, cholesterol, SGOT and PT were within normal limits, but her fasting blood glucose was 145 mg%. A CT scan was performed a week after clinical onset and showed evidence of infarction at the left lateral frontal region (see Fig. 1). When B was discharged from the hospital after 2 weeks, her right arm and leg improved slightly in strength but her speech was virtually unchanged. Three days after leaving the hospital, B had convulsions which began in the right limbs. When examined on readmission to the hospital, she had a complete right hemiplegia and motor aphasia. Phenytoin and dexamethasone were given. She improved slowly, and when discharged from the hospital 3 weeks later, she was able to walk with support. When seen at the out-patient clinic approximately 3 months after the stroke, B could say only a few isolated words. Another CT scan at 39 months postonset (17 September 1980) revealed an irregular area of decreased density at the left frontoparietal lobe with ipsilateral enlargement of left lateral ventricle and cortical sulci, compatible with the old infarct. A Thai adaptation of the Boston Diagnostic Aphasia Examination (Gandour, 1982; Gandour, Dardarananda, Buckingham, & Viriyavejakul, 1986) was administered on 30 July 1980, approximately 3 years after the stroke. The results from the Thai aphasia exam were fully compatible with a diagnosis of Broca’s aphasia (Gandour, Dardarananda, & Vejjajiva, 1985). Spontaneous speech was nonfluent, dysprosodic, and agrammatic; repetition was poor; however, comprehension of spoken language was relatively good in ordinary conversation. B (= Bl in Gandour & Dardarananda, 1984a,b) was born in 1920 in Bangkok. She had completed the 10th grade in local schools. She was a monodialectal speaker of Central Thai, the official language of Thailand. Timing Method Subjects. Data were obtained from B for this study. For comparison purposes, data were taken from a previously published report on five normal adult speakers (Luangthongkum, 1977). Speech muterids. Four readings of a passage were obtained from B. Data based on one FIG. 1. CT scan showing area of decreased density in left lateral frontal change in appearance was evident after enhancement with conray 420. 238 lobe. No DYSPROSODY IN BROCA’S APHASIA 239 reading of the passage by each of five normal speakers were taken from Luangthongkum’s (1977) study. The passage itself was a Thai adaptation of the story of a parrot as told by Blanche DuBois in Tennessee Williams’ play A Sfreetcnr Named Desire. The majority of words in the reading passage were monosyllabic. The Thai adaptation was carefully designed by Luangthongkum for the purpose of making syllable duration measurements. A phonemic transcription according to Haas (1964) and an English translation of the reading passage are presented in the Appendix. Measurement procedure. Recordings of four readings of the passage by B were played into a Pitch Instruments F,-intensity meter. The audio, FU, and intensity signals were played into three different channels of a Honeywell 1508 oscillograph operating at a chart speech of 100 mm/set. Identification of syllable boundaries was determined from inspection of the audio, F,,, and intensity-meter signals (cf. Gandour, Weinberg, Petty, & Dardarananda, 1986). For the majority of adjacent syllables in the passage either the offset of the first syllable or the onset of the second syllable was voiceless, or they were separated by a syntactic pause. In those cases, syllable onset was operationally defined to occur when there was a sharp increase in intensity from the baseline, syllable offset when there was a sharp decrease in intensity to the baseline. In the few remaining cases, adjacent syllables had different tones. Syllable onset and offset were determined on the basis of shifts in direction of F0 associated with the different tones. A pause was operationally defined to occur when the speech waveform approached baseline for at least 200 msec. The portion of the waveform tracing between two pauses was designated as a phrase. In this context, phrases were phonetically structured entities that did not necessarily coincide with syntactic phrases. Syllable and pause durations were measured by the first author to the nearest centisecond using a transparent millimeter rule. All syllable and pause durations were reanalyzed on a separate occasion for an estimate of intrajudge reliability (r = .99). The total number of syllables subjected to duration measurements differed between readings because of the speaker’s omissions of words in each of her four readings of the paragraph or the occurrence of unmeasurable portions of the oscillographic records. Identification of stressed syllables in the readings of the passage by B was performed by the second author, a native speaker of Thai with a postgraduate degree in linguistics. She listened to the tape recordings of the reading passage and identified which syllables she perceived to be stressed. Each recording of the paragraph was presented in two separate sessions spaced approximately 1 week apart. Her stress judgments were identical for the two presentations of each recording of the paragraph. Stressed syllables in the readings of the passage by the five speakers were taken from Luangthongkum (1977). Duration measurements of syllables and pauses in the readings of the passage by the five normal speakers were taken directly from Luangthongkum (1977). Her measurements to the nearest centisecond were determined in an identical manner from inspection of oscillographic records. Results and Discussion Mean durations of syllables and pauses based on four readings of the passage by B and one reading each of the passage by Luangthongkum’s (1977) five normal speakers are summarized in Table 1. B took a longer time to read the passage aloud than normal speakers. Her total reading time was about four times greater than that for normal speakers. As measured in syllables per second, B and the normal speakers spoke at a rate of 0.9 and 3.3 syllables/set, respectively. In terms of words per 240 GANDOUR, PETTY, DARDARANANDA TABLE 1 DURATION OF PAUSES AND SYLLABLES Speakers B Normal Measure M SD M SD Total duration of passage (set) Total number of pauses Total duration of pauses (set) Mean pause duration (csec) Percent pause duration Total number of syllables Total duration of syllable (set) Mean syllable duration (csec) Syllables per second Pause-to-syllable duration ratio 208.4 130.8 117.7 90.1 56.3 175.3 87.7 50.0 0.9 1.4 16.6 8.9 15.0 10.0 2.7 4.0 3.3 1.3 0.1 0.1 54.5 23.2 14.1 60.0 25.7 177.0 40.4 23.0 3.3 0.3 2.5 1.3 1.6 7.0 1.8 1.3 1.2 1.0 0.1 0.1 minute, B spoke at a rate of 46 wpm, while normal speakers spoke at a rate of 165 wpm. All pause time measures for B were larger than for normal speakers (see Table 1). B exhibited a larger number of pauses, a greater amount of total pause time, a greater mean pause duration, and a greater percentage of pause time. B’s pause-to-syllable duration ratio (total pause time divided by total phrase time) was almost five times as large as that for the normal speakers. Total duration of syllables and mean syllable duration for her were both about twice as large as those for normal speakers. Phrase timing measures for B and the normal speakers are presented in Table 2. There is an inverse relationship between the total number of TABLE 2 PHRASE TIMING Speakers Normal B Measure M SD M SD Total number of phrases Total phrase duration (set) Mean phrase duration (csec) Maximum syllables per phrase Mean syllables per phrase 136.8 25.2 18.3 4.3 1.3 3.9 2.8 2.1 0.9 0.1 24.2 32.3 134.0 19.0 7.3 1.3 1.1 8.0 0.0 0.4 241 DYSPROSODY IN BROCA’S APHASIA phrases and the total duration of phrases. B produced a greater number of phrases in a shorter total duration of phrase time, while the normal speakers produced fewer phrases in a longer total duration of phrase time. In terms of mean phrase duration, maximum number of syllables per phrase, and mean number of syllables per phrase, B produced clearly shorter phrases. The time between the onset of one stressed syllable and the onset of the next stressed syllable was designated as an interstress interval. A one-syllable interstress interval indicates that two adjacent syllables are stressed; a two-syllable interstress interval indicates that an unstressed syllable occurs between two stressed syllables; and so on. To provide a measure of the extent to which the temporal speech patterns were syllable timed or stress timed, mean durations of interstress intervals were computed, as well as their frequencies of occurrence. Important characteristics of interstress intervals are summarized in Table 3. All of B’s interstress intervals were one syllable. While the majority of the normal speakers’ interstress intervals were also one syllable, 35% of them were two syllable. B’s one-syllable interstress intervals were, on average, three times longer than those of normal speakers. A comparison of syllable durations as a function of stress is presented in Table 4. The ratio of stressed-to-unstressed syllables was 1.7 for the TABLE 3 OCCURRENCE AND DURATION OF INTERSTRESS INTERVALS Speakers B Interstress interval M One syllable Occurrence” Duration’ Two syllable Occurrence Duration - Three syllable Occurrence Duration - Four syllable Occurrence Duration Normal SD M SD 100 0 119.0 6.9 56.3 39.2 1.4 1.9 - 35.3 52.0 1.2 3.1 - 3.9 107.4 0.5 - - u Values are expressed in percentages. ’ Values are expressed in centiseconds. - 2.9 94.0 7.3 0.8 10.3 242 GANDOUR, PETTY, DARDARANANDA TABLE 4 EFFECT OF STRESS ON SYLLABLE DURATION Speakers B Normal Measure M SD M SD Total number of stressed syllables Total number of unstressed syllables Mean stressed syllable duration (csec) Mean unstressed syllable duration (csec) 175.3 4.0 1.3 - 112.0 65.0 28.0 14.0 2.3 2.3 0.2 0.1 50.0 - normal speakers; the stressed-unstressed duration ratio was 2.0. Because all of her syllables were stressed, no ratios were computed for B. A comparison of syllable durations as a function of whether or not they preceded a pause is presented in Table 5. B produced about three times as many prepausal as nonprepausal syllables; normal speakers, on the other hand, produced about seven times as many nonprepausal as prepausal. The prepausal-nonprepausal duration ratio was 1.Ofor B, 1.7 for the normal speakers. These combined findings clearly indicate that B exhibited an impairment in timing in connected speech. In comparison to normal adults, B’s temporal patterns in syllables and phrases were abnormal. Her slower overall speaking rate, shorter phrases, uniform syllable durations regardless of stress or sentence position, and exclusive use of one-syllable interstress intervals all contributed to the dysrhythmic quality of B’s speech. These findings were not restricted to oral reading. The dysrhythmic quality of B’s oral reading directly reflects her spontaneous speaking abilities. B tended to produce speech on a syllable-by-syllable basis. In so doing, an economy in timing patterns was achieved insomuch as temporal dependencies among syllables were minimized. Such timing patterns appear to reflect simplification of motor programming of larger sized linguistic units. Interestingly, B’s syllable-by-syllable production is not unlike that reported in ataxic dysarthria (Kent, Netsell, & Abbs, 1979). A few polysyllabic words, phrases, and sentences were selected from the passage to further illustrate the nature of B’s breakdown in timing in larger sized temporal domains (Figs. 2-4). First, B’s polysyllabic words, phrases, and sentences are markedly longer than normal on an absolute scale (see Figs. 2-4). Second, conspicuous, lengthy gaps of silence are evident between syllables in the majority of B’s productions (see Figs. 2-4). Third, the relative temporal relationships among B’s syllables deteriorate in larger sized linguistic units. In Thai, initial syllables in bisyllabic words are unstressed and final syllables are stressed. 243 DYSPROSODY IN BROCA’S APHASIA TABLE 5 EFFECT OF POSITION ON SYLLABLE DURATION Speakers B Normal Measure M SD M SD Total number of prepausal syllables Total number of nonprepausal syllables Mean prepausal syllable duration (csec) Mean nonprepausal syllable duration (csec) 130.8 44.5 50.1 8.9 6.2 I.2 23.2 154.2 1.3 2.5 I.9 50.0 1.9 35.7 20.8 0.6 The average duration ratio of final to initial syllable in the word /kaafaxe/ was 3.2 for normal speakers, 1.1 for B (see Fig. 2); in the prepositional phrase /naj baan/ 3.1 for normal speakers, 0.8 for B (see Fig. 3). In the Thai sentence /man sari pen baa laaj w&j/, an alternating stress pattern manifested by the differential lengthening of syllables is evident in normal speech only (see Fig. 4). The average duration ratio of last to first syllable in /man s&r/, /pen baa/, and /laaj wa’oj/, respectively, was 1.4, 2.1, and 1.9 for normal speakers and 0.9, 0.9, and 0.7 for B. In each case, the last syllable was longer than the first for the normal speakers, whereas the opposite was true for B. Following Kent and Rosenbek (1982, pp. 266-267), an estimate of degree of the temporal isolation of syllables in this sentence was obtained by dividing total syllable duration by the total sentence duration. The smaller the ratio, the greater the degree of separation of syllables. The average ratio was 1.0 for normal speakers, but kaa 1 FIG. 2. Durations of syllables and silent intervals for the word /kaafaz/ “coffee” produced by a Thai Broca’s aphasic (B) and five normal speakers (N). The duration of /kaa/ is shown by the white portion of the bar, /fax/ by the stippled portion, and the silent interval by the black portion. For B, each of the numbered bars (l-4) represents an individual production; for N, the bar represents a mean based on one production from each of the five normal speakers. 244 GANDOUR, PETTY, DARDARANANDA naj b&n 1 FIG. 3. Durations of syllables and silent intervals for the phrase /naj ba’an/ “in the house” produced by a Thai Broca’s aphasic (B) and five normal speakers (N). The duration of /naj/ is shown by the white portion of the bar, /baan/ by the stippled portion, and the silent interval by the black portion. Also see caption for Fig. 2. only 0.6 for B. Thus, her pattern of syllable dissociation clearly disrupted timing on a relative as well as absolute scale of measurement in multisyllabic word-, phrase-, and sentence-sized linguistic units. In contrast to the study of vowel duration in isolated, monosyllabic words, the present findings on timing control in larger sized linguistic units indicate that B exhibits a severe timing deficit. B’s timing control is spared in the production of vowel duration within isolated, monosyllabic words, but impaired in the production of syllables that occur in series within words, phrases, and sentences. These results suggest that the degree of timing disintegration in Broca’s aphasia depends, in part, on the length of the speech planning unit. Additional support for this hypothesis can be found in previous studies of duration in English-speaking Broca’s aphasics and apraxia of speech patients. Danly and Shapiro FIG. 4. Durations of syllables and silent intervals for the sentence /man s&n pen ba^a laaj w&j/ “How come the day is so short” produced by a Thai Broca’s aphasic (B) and five normal speakers (N). The duration of /man/ is shown by the white portion of the bar; /San/, /bah/, and /wahj/ by stippled portions; /pen/ and /laaj/ by hatched portions; and the silent intervals by black portions. Also see caption for Fig. 2. DYSPROSODY IN BROCA’S APHASIA 245 (1982) reported that sentence-final lengthening was absent in Broca’s aphasics. Kent and Rosenbek (1983) found that the magnitude of the lengthening effect on vowel duration in apraxia of speech patients increased as the length of the utterance increased. On the basis of the VOT, vowel duration, and rhythm studies, it appears that B’s timing problem is a manifestation of a broader problem in the dynamics of speech production. Her speech timing difficulties lie not only in the integration of articulatory movements for a single phonetic segment (VOT study), but also in the integration of articulatory movements from one phonetic syllable to another (rhythm study). However, B’s timing is relatively normal when no temporal integration between two independent articulators is required for a single phonetic segment within an isolated, single syllable (vowel duration study). Intensity Relative peak intensity was measured using the Kay Elemetrics VisiPitch Model 6095 interfaced to an Apple II microcomputer. Peak intensities were measured for each syllable in a test sentence selected from the same passage used in the rhythm study. In the sentence /khraj daan phaan maa phoo dii/, intensity values were expressed relative to the syllable with the maximum peak intensity among the six syllables. The remaining syllables were assigned negative intensity levels in dB. As shown in Fig. 5, less variation in relative peak intensity across syllables was observed for B than for the normal speaker. Following Kent and FIG. 5. Relative peak intensities for the words in the sentence /khraj daan p%an maa p”oo dii/ “Do you know who was passing by?” produced by a Thai Broca’s aphasic (B) and a normal female adult speaker (N). For B, each of the four dashed lines represents an individual production; for N, the solid line represents a mean based on five productions. 246 GANDOUR, PETTY, DARDARANANDA Rosenbek (1983, p. 236), the intensity values for the syllables in the test sentence were averaged for each speaker to provide a quantitative measure of intensity variation from syllable to syllable. The smaller the negative number, the lesser the degree of intensity variation. The mean results were -2.3 and -5.3 for B and the normal speaker, respectively. Furthermore, B failed to give the greatest intensity to the same syllable as the normal speaker in three of her four productions. Thus, B’s intensity envelope differed from that of the normal speaker both qualitatively and quantitatively. The observed flattening of B’s intensity envelope presumably reflects her inability to produce appropriate variations in stress patterns in larger-sized temporal domains. Such abnormalities in B’s intensity envelope are consistent with auditory impressions of her conversational speech as well. This tendency toward reduced intensity variation from syllable to syllable in B’s speech has also been reported in earlier acoustic investigations of apraxia of speech (Kent & Rosenbek, 1983; Lebrun, Buyssens, & Henneaux, 1973). Fundamental Frequency Method For comparison purposes, F,, data on tones produced in isolation were taken from a previously published report on five normal adult speakers (Gandour et al., 1988). Further details regarding speech materials, recording procedure, and measurement procedure can be found therein. A measure was obtained of intraspeaker variability in F0 contours of each of the five Thai tones produced by B and one normal, female speaker. To compare the point-to-point variability of F. contours for repetitions of each tone from each speaker, an ensembleaveraging approach was taken (cf., Atkinson, 1976, pp. 441-442). All repetitions of each tone made up an ensemble of F,, contours for that tone by that speaker. Since ail repetitions of each tone were not of equal duration, the F. contours were normalized for duration on a percentage scale. The ensemble mean and standard deviation were determined at each 10% interval. These ensemble-averaged statistics made it possible to evaluate changes in ensemble characteristics of intraspeaker contours for each tone as a function of time. Polysyllabic words, phrases, and sentences were selected from Luangthongkum’s (1977) reading passage. Data were obtained from B and one normal, male adult speaker. F. contours were analyzed by means of F,, displays from the Kay Elemetrics Visi-Pitch Model 6095 and narrow-band spectrograms from the Kay Elemetrics Digital Sona-Graph Model 7800. Results and Discussion Average F,-, contours normalized for time as a percentage of total duration of each of the five tones produced by one of the normal, female speakers and B are displayed in Fig. 6. For the normal speaker, FO contours associated with the mid and low tones fall steadily throughout. FO contours associated with the falling, high, and rising tones change in slope approximately halfway through their duration. The falling tone is relatively level and then falls sharply; the high and rising tones fall slightly DYSPROSODY IN BROCA’S APHASIA 20 40 00 fDIE ca 247 100 (x) FIG. 6. Average F0 contours of the Thai tones in Hz normalized for time on a percent scale for monosyllabic words produced in isolation by a Thai Broca’s aphasic (B; top panel; adapted from Gandour, Petty, & Dardarananda, 1988) and a normal female adult speaker (N; bottom panel). and then rise. Average F,, contours for the other four normal speakers were similar. B’s average FO contours associated with all five tones are similar in shape and position within the tone space to those of normal speakers. 248 GANDOUR, PETTY, DARDARANANDA It is noted, however, that the shape of B’s mid and high tones differ somewhat from those of the normal, female speaker depicted in Fig. 6. Instead of displaying a steadily falling slope, B’s mid tone is relatively level for over half of its duration and then falls sharply. Instead of displaying a continuous rise in the terminal portion, B’s high tone exhibits a falling slope. These F,, contours associated with B’s mid and high tones are not abnormal. Such contours have been reported in earlier studies of Thai tones produced by normal speakers in citation forms of monosyllabic words (Abramson 1962, 1976). The mean F,, and standard deviation of each tone for B and one normal, female speaker are given in Table 6. B’s mean F0 (212 Hz) across the five tones was significantly lower than that (221 Hz) of the normal female speaker, t(lOO) = 2.38, p < .019. Ryalls (1982) reported that the mean F0 on a sentence-repetition task was of the same order of magnitude for French-speaking Broca’s aphasics and normal speakers. However, Ryalls (1984) found that the overall F0 of vowels on a read speech task produced by English-speaking anterior aphasics were significantly higher than those produced by normal speakers. It was suggested that the higher average F,, of the anterior aphasics could be due to their higher psychological stress in read speech. In this study, B produced read speech, yet she did not exhibit a higher-than-normal mean FO. More research is required to resolve these conflicting findings with respect to the mean F,, of Broca’s aphasics and normal speakers. The average F0 ranges of the five tones for B and a group of five, normal adult speakers were of the same order of magnitude (see Table 7). B’s F0 ranges for the mid and high tones were slightly higher than normal, and for the low, falling, and rising tones were slightly lower than normal. None of B’s F,, ranges that were lower than normal fell outside minus one standard deviation of the normal range for that tone. Ryalls TABLE 6 DESCRIPTIVESTATISTICSOF THAI TONES Speaker B Normal” Tone M SD M SD Mid Low Falling High Rising 206 189 222 234 211 12 13 24 12 20 219 199 249 235 207 11 14 23 7 15 Note. Values are expressed in F,, (Hz). a Normal = one normal, female adult speaker. 249 DYSPROSODY IN BROCA’S APHASIA TABLE 7 F,, RANGES OF TONES Tone Speaker Mid Low Falling High Rising B Normal” 4.1 3.9 4.1 5.1 6.2 6.8 2.4 2.3 4.4 4.5 Note. Values are expressed in F. (semitones). ’ Normal = Five, normal adult speakers from Gandour et al. (in press). (1982) and Cooper, Soares, Nicol, Michelow, and Goloskie (1984), on the other hand, in their production studies of sentence intonation in French- and English-speaking Broca’s aphasics, respectively, both reported a restricted F, range for Broca’s aphasics when compared to normals, Ryalls’ measure of F, range was peak-to-valley fluctuation within sentences; Cooper et al.‘s measure was differences among F. peaks within sentences. Interestingly, Danly and Shapiro (1982) reported that English-speaking Broca’s aphasics actually produced more fluctuation in F, than normal speakers. They measured F,, variation by summing valley-peak-valley F,, differences within words embedded within sentences. Their findings suggest that “the perception of flat intonation in Broca’s speech may be based on restricted word-to-word F0 variation, while within-word variability may actually be increased” (Cooper & Klouda, 1987, p. 62). The absence of any compression of the F0 range in B’s production of tones on isolated, monosyllabic words supports the view that restricted F,, variation in Broca’s speech is a function of the size of the temporal domain rather than a function of F, itself. The point-to-point variability in F0 contours of each of the five tones produced by B and one normal, female speaker is displayed in Fig. 7. The ensemble mean can be considered as the target F,, contour for a particular tone; the ensemble standard deviation as a measure of the random perturbations that result in deviations from the target contour. As shown in Fig. 7, B’s intraspeaker contours appear to be more variable than those of the normal, female speaker for all tones except rising. To provide a quantitative comparison of intraspeaker variability, the ensemble standard deviations were averaged over the 11 measurement points of each tone for all five tones. Overall average variability in F, was significantly greater for B than for the normal, female speaker, t(108) = 2.18, p < .03. This finding is consistent with the increased variability reported in B’s production of VOT (Gandour & Dardarananda, 1984a) and vowel duration (Gandour & Dardarananda, 1984b). And more generally, this finding agrees with earlier acoustic investigations that have revealed greater variability in speech produced by anterior aphasics, Mm FIG. 7. Variability of <,, contours of the Thai tones in semitones normalized for time on a percent scale for monosyllabic words produced in isolation by a Thai Brow’s aphasic (B; top panel) and a normal female adult speaker (N; bottom panel). Vertical bars represent plus or minus one standard deviation about the ensemble average F. contour. 250 DYSPROSODY IN BROCA’S APHASIA 2.51 measured in terms of VOT (Blumstein et al., 1980), vowel duration (Ryalls, 1986), formant frequencies (Ryalls, 1981, 1986), and F0 (Ryalls, 1984). F,, contours for a three-syllable noun compound /jaajsdawk&=/, a seven-syllable clause /man ca d5j khit wa’a muiurt ~&XI/, and a sevensyllable sentence /ru’u khamjdap ma^akmaajltiakaan/ are shown in Figs. 8-10, respectively. In comparison to the normal speaker, B’s speaking rate and rhythm were markedly impaired. Her utterances were considerably longer than normal. Lengthy pauses predominated at syllable boundaries. Timing adjustments required by stress patterns in a series of syllables were absent. This pattern of temporally isolated syllables supports the view that B’s speech is produced on a syllable-by-syllable basis. These deviant timing patterns notwithstanding, B apparently had no difficulty in producing the F0 contours of all five Thai tones in connected speech (see Figs. 8-10). She was able to do so regardless of sentence position or adjacent tone. In Fig. 8, the sequence of tones was mid-rising-low; in Fig. 9, mid-mid-falling-high-falling-falling-high; and in Fig. 10, high-mid-low-falling-mid-rising-mid. When compared to the normal speaker, B’s F,, contours appear to resemble more closely the shape of tonal contours that occur prepausally or in monosyllabic citation forms (see Figs. 9 and 10). The temporal domain for Thai tones is the syllable. B is apparently able to plan and execute F0 contours successfully within this unit of speech production. Because of F, perturbations induced by tonal coarticulation and sentence intonation, the normal speaker’s tonal contours actually exhibit less F(, modulation than those of B. The effects of tonal coarticulation and sentence intonation are minimized in B’s speech due to her inability to control F, over a temporal domain larger than a single syllable. It is because Thai tones extend over a syllable-sized unit of speech programming that tonal contours are preserved in B’s connected speech. The findings of the present study are consistent with previous investigations of intonation and timing in English-speaking Broca’s aphasics (see Cooper & Klouda, 1987, for review). Of particular interest is the study by Danly and Shapiro (1982). Based on three acoustic measures of speech prosody (sentence-final F(, fall, declination of F0 peaks throughout a sentence, and sentence-final segmental lengthening), they similarly found that speech timing was relatively more disturbed than F,,. Broca’s aphasics exhibited F,, fall in utterance-final position. F,, declination was present in short sentences, but absent in longer sentences. For longer sentences, Broca’s aphasics reset their declination function many times within the sentence. Segmental lengthening in utterance-final position was absent. In fact, sentence-final words were actually shorter than their sentence-initial and sentence-medial counterparts. Their findings indicated that Broca’s aphasics were able to carry out F,, programming over 252 GANDOUR, PETTY, DARDARANANDA F,---L NI /jaaj 200 ms sa”aw ka&e/ FIG. 8. F. contours of the compound word /jaajsa2wk&e/ “spinster” produced by a Thai Broca’s aphasic (B; top panel) and a normal male adult speaker (N; bottom panel). A dotted line is drawn through the tenth harmonic in the narrowband spectrograms (8 kHz frequency range; 45 Hz analysis filter). domains that are much shorter than normal. Because of the shorter-thannormal domains of prosodic programming, Broca’s aphasics were unable to carry out timing programming for segments in utterance-final position. The programming of F0 necessarily intersects with temporal planning in sentence production. Thus, the disruption of intonation contours need not indicate a malfunction in F,, mechanisms, but instead a secondary effect due to a malfunction in timing. B’s preservation of tones in isolated words and connected speech contrasts sharply with her impaired production of consonants. First, her production of the voicing contrast in word-initial stop consonants was aberrant (Gandour & Dardarananda, 1984a). Second, a phonological analysis of B’s conversational speech revealed that consonant errors were 253 DYSPROSODY IN BROCA’S APHASIA 2 kH.z / man co /man ca d;j m&At la&w / 3 kHz .._. kh6 IV& muk.n MatBw/ FIG. 9. F0 contours of the clause /man ca daj khit was muim la&w/ “This would make him think that it was dark .” produced by a Thai Broca’s aphasic (B; top panel) and a normal male adult speaker (N; bottom panel). A dotted line is drawn through the fifth harmonic in the narrowband spectrograms (4 kHz frequency range; 75 Hz analysis filter). frequent, whereas tone errors rarely occurred (Gandour et al., 1985). All occurrences of velar stops were replaced with a voiceless unaspirated alveolar stop. Third, B produced the high and rising tones without regard to the phonological constraint of Central Thai which prohibits their occurrence on syllables beginning with voiceless unaspirated stops. These findings together point to a dissociation between segmental and suprasegmental aspects of B’s speech. Despite deterioration of articulatory control of consonants, B maintained phonatory control of the five Thai tones. GENERAL DISCUSSION The unqualified use of the term dysprosody implies that the deficit is generalized to all prosodic aspects of speech, including F,, intensity, and timing. Characterizing B’s speech as simply dysprosodic would be inaccurate and misleading. Such a characterization implies that various elements of B’s speech prosody suffer mutual deterioration as the result 254 GANDOUR, PETTY, DARDARANANDA F,- - +-WV - BI. \ /ru’u I n’ kham r’Y j&p ----- F, - N- /ru’u -- aahns kham j&p m&k 2ooms fnaaj t I& Ill& - rna:k keen/ - maaj IGo keen/ FIG. 10. F,, contours of the sentence /ru’u k”amjaap maakmaaj lthakaan/ “He knew a lot of dirty words” produced by a Thai Broca’s aphasic (B; top panel) and a normal male adult speaker (N; bottom panel). A dotted line is drawn through the fifth harmonic in the narrowband spectrograms (B: 2 kHz frequency range, 37.5 Hz analysis filter; N: 8 kHz frequency range, 45 Hz analysis filter). of damage to Broca’s area. To the contrary, some elements of B’s speech prosody remain intact, while others are clearly impaired. Despite the fact that sentence timing and intensity were abnormal in her connected speech, B retained an ability to produce FOpatterns associated with the five Thai tones. In particular, the findings from this case study reinforce the view that dysprosody in Broca’s aphasia is more applicable to speech timing than to FO (Cooper & Klouda, 1987, Danly & Shapiro, 1982). Moreover, it is because differences in FO patterns are linguistically significant at the word level in a tone language that the data of B provide unequivocal evidence in support of Danly and Shapiro’s hypothesis that dysprosody in the speech of Broca’s aphasics stems primarily from their smaller units of speech programming. With this different conceptualization of dysprosody in Broca’s aphasia, it may be inappropriate to speak of a tone or intonation deficit per se. It is not the control of FO for tone or intonation specifically that is in question, but rather the control of FO over variable-sized temporal domains. The critical variable is the size of the domain over which the F, DYSPROSODY IN BROCA’S APHASIA 25.5 patterns extend. By definition, F, patterns associated with intonation extend over larger sized temporal domains than those associated with tone. If dysprosody in Broca’s aphasia is primarily due to a narrowerthan-normal scope for speech programming, it is predicted that tones will be more resistant to disruption than intonation. To produce intonation successfully, one must be able to control timing over phrase- or sentence-sized units; to produce tone successfully, one must be able to control timing over syllable-sized units of speech programming. Timing disturbances at the sentence level necessarily disrupts FO contours associated with intonation patterns. As exemplified in this case study of a Thai-speaking Broca’s aphasic, F. contours associated with tone are not disrupted as a consequence of deviant timing at the sentence level. APPENDIX Thai Phonemic Transcription and English Translation of the Story of a Myna Bird pee nyg duu k3on ryaq thfi khaaj 1%~p&a nyg %g l&w 3aw rfian n6gVag dii kwAa nA mii n6gYag tua n$q man dAa keg pen faj laaj r6u khamj&ab m&agmaaj lyakaon MEW k3 wlthii diaw thfi c&p haj man ghb dBaj n&n k3 khyy t3n 3aw phaa khlum kroq sia man c&p d&aj khft wAa myyd Mew dbj weelaa noon sla thii piitaa nfi n& rQgch&aw wan n$q jaajstiwktc cBwkh3og n6g n&n k3 paad paw ph%a thfi khlum krou %g jaag khoaj ruu maj khraj daan phaan maa phoodii b&adluag c&3 jaajc?lwkh%g kcc rfib wig klab maa khlum krog c&w&g j&au rew waj ICcw k3 wft~ paj thii pratuu chaan b&adlfrarj kh&w maa naj baan n6g man k3 nlg chiaw ICP ICcw jaajsHawk&c k3 th%am ph3obtidlfran wia haj s&j namtaan kti ch3on naj kaafcc 3&j c%wnbg man k3takoon phl&q %g maa w&a mEc h&j taaj th53 nfP %j klaau wan nfi man sari pen b&a laaj w&j Hm . . . let me think . . . a story that I have told before. Ah! I remember now . . . the story of a myna bird. There was a myna bird who was very keen in using bad language. He knew a lot of dirty words. The only way that one could make him quiet was to cover his cage with a piece of cloth. This would make him think that it was dark and that it was the time to go to bed. Well, one early morning, the spinster who was the bird’s owner came to uncover the cage as usual. Do you know what was passing by? A Roman Catholic priest! The owner of the bird rushed back to cover the cage as quickly as she could. Then, she ran to the door and invited the priest to come in. The bird was quiet. Finally, the spinster asked the priest, “How much sugar do you want in your coffee?” The bird, then, shouted out loudly, “Damn it! How come the day is so short! 256 GANDOUR, PETTY, DARDARANANDA REFERENCES Abramson, A. S. 1962. The vowels and tones of standard Thai: Acoustical measurements and experiments. Bloomington, IN: Indiana University Research Center in Anthropology, Folklore, and Linguistics, Publication 20. (Also International Journal of American Linguistics, 28(2), Part III, 1962.) Abramson, A. S. 1974. Experimental phonetics in phonology: Vowel duration in Thai. Pusan, 4, 71-90. Abramson, A. S. 1976. Thai tones as a reference system. In T. W. Gething, J. G. Harris, & P. Kullavanijaya (Eds.), Tai linguistics in honor of Fang-Kuei Li. Bangkok: Chulalongkorn Univ. Press. Pp. l-12. Abramson, A. S. 1979a. The coarticulation of tones: An acoustic study of Thai. In V. Panupong, P. Kullavanijaya, K. Tingsabadh, & T. Luangthongkum (Eds.), Studies of Tai and Mon-Khmer phonetics in honour of Eugenie J. A. Henderson. Bangkok: Chulalongkorn Univ. Press. Pp. l-9. Abramson, A. S. 1979b. Lexical tone and sentence prosody in Thai. In E. Fischer-Jorgensen, J. Rischel, & N. Thorsen (Eds.), Proceedings of the Ninth International Congress ofPhonetic Sciences. Copenhagen: University of Copenhagen. Vol. II, pp. 380-387. Abramson, A. S., & Svastikula, K. 1983. Intersection of tone and intonation in Thai [summary]. In S. Hattori & K. Inoue (Eds.), Proceedings ofthe XZllth International Congress of Linguists. Tokyo: Comite International Permanent des Linguistes. Atkinson, J. E. 1976. Inter- and intraspeaker variability in fundamental voice frequency. Journal of the Acoustical Society of America, 60, 440-445. Baum, S., & Blumstein, S. E. 1987. Consonant production deficits in aphasia. In J. Ryalls (Ed.), Phonetic approaches to speech production in aphasia and related disorders. Boston: College-Hill. Pp. 3-21. Benson, D. F. 1979. Aphasia, alexia, and agraphia. New York: Churchill Livingstone. Blumstein. S. E., Cooper, W. E., Goodglass, H., Statlender, S., & Gottlieb, J. 1980. Production deficits in aphasia: A voice-onset time analysis. Brain and Language, 9, 153-170. Caramazza, A., & Martin, R. C. 1983. Theoretical and methodological issues in the study of aphasia. In J. B. Hellisge (Ed.), Cerebral hemisphere asymmetry: Method, theory, and application. New York: Praeger. Pp. 18-45. Collins, M., Rosenbek, J., & Wertz, R. 1983. Spectrographic analysis of vowel and word duration in apraxia of speech. Journal of Speech and Hearing Research, 26, 224-230. Cooper, W. E., & Klouda, G. 1987. Intonation in aphasic and right-hemisphere-damaged patients. In J. Ryalls (Ed.), Phonetic approaches to speech production in aphasia and related disorders. Boston: College-Hill. Pp. 59-77. Cooper, W. E., Soares, C., Nicol. J., Michelow, D., & Goloskies, S. 1984. Clausal intonation after unilateral brain damage. Language and Speech, 27, 17-24. Danly, M., & Shapiro, B. 1982. Speech prosody in Broca’s aphasia. Brain and Language, 16, 171-190. Duffy, J., & Gawle, C. 1984. Apraxic speaker’s vowel duration in consonant-vowelconsonant syllables. In J. Rosenbek, M. McNeil, & A. Aronson (Eds.), Apraxiu of speech: Physiology, acoustics, linguistics, management. Boston: College-Hill. Pp. 167-196. Gandour, J. 1975. On the representation of tone in Siamese. In J. G. Harris & J. R. Chamberlain (Eds.), Studies in Thai linguistics in honor of William J. Gedney. Bangkok: Office of State Universities. Pp. 170-195. Gandour, J. 1982. A diagnostic aphasia examination for Thai. Linguistics of the TibetoBurmun Area, 6(2), 65-76. Gandour, J. 1984. Vowel duration in Thai. Crossroads: Southeast Asian Studies, 2(l), 59-62. An Interdisciplinary Journal of 257 Gandour, J. 1987. Tone production in aphasia. In J. Ryalls (Ed.), Phonefic upproaches to speech production in aphasia and related disorders. Boston, MA: College-Hill. Pp. 45-57. Gandour, J., & Dardarananda, R. 1984a. Voice onset time in aphasia: Thai. II. production. Brain and Language, 23, 177-204. Gandour, J., & Dardarananda, R. 1984b. Prosodic disturbance in aphasia: Vowel length in Thai. Brain and Language, 23, 206-224. Gandour, J., Dardarananda, R., Buckingham, H., & Viriyavejakul, A. 1986. A Thai adaptation of the Boston Diagnostic Aphasia Examination. Crossroads: An Znterdisciplinary Journal of Southeast Asian Studies, 2(3), l-39. Gandour, J., Dardarananda, R., & Vejiajiva, A. 1985. Case study of a Thai Broca aphasic with an adaptation of the Boston Diagnostic Aphasia Examination. Journul of the Medicul Association of Thailand, 68, 552-563. Gandour, J., Petty, S., & Dardarananda, R. 1988. Perception and production of tone in aphasia. Brain and Language, 35, 201-240. Gandour, J., Weinberg, B., Petty, S., & Dardarananda, R. 1986.Rhythm in Thai esophageal speech. Journal of Speech and Hearing Research, 29, 563-568. Goodglass, H., & Kaplan, E. 1983. The as.sessment of aphasia and related disorders. Philadelphia: Lea & Febiger. 2nd ed. Haas, M. R. 1964. Thai-English student’s dictionary. Stanford, CA: Stanford Univ. Press. Itoh, M., Sasanuma, S., Tatsumi, I. F., Murakami, S., Fukusako, Y.. & Suzuki, T. 1982. Voice onset time characteristics in apraxia of speech. Bruin and Language, 17, 193210. Kent, R., Netsell, R., & Abbs, J. 1979. Acoustic characteristics of dysarthria associated with cerebellar disease. Journal of Speech and Hearing Research, 22, 627-648. Kent, R., & Rosenbek, J. 1982. Prosodic disturbance and neurologic lesion. Brain and Language, 15, 259-291. Kent, R., & Rosenbek, J. 1983. Acoustic patterns of apraxia of speech. JournuI of Speech und Hearing Research, 26, 231-249. Lebrum, Y., Buyssens, E., & Henneaux, J. 1973. Phonetic aspects of anarthria. Cortex, 9, 126-135. Lisker, L., & Abramson, A. S. 1964. A crosslanguage study of voicing in initial stops. Word, 20, 384-422. Luangthongkum, T. 1977. Rhythm in standard Thai. Unpublished doctoral dissertation, University of Edinburgh. Palmer, A. 1969. Thai tone variants and the language teachers. Languuge Learning, 19, 287-299. Roberson, J. F. 1982. Bangkok Thai tones. Unpublished doctoral dissertation, University of Southern California, Los Angeles. Ryalls, J. 1981. Motor aphasia: Acoustic correlates of phonetic disintegration in vowels. Neuropsychologia, 19, 365-374. Ryalls, J. 1982. Intonation in Broca’s aphasia. Neuropsychologia, 20, 355-360. Ryalls, J. 1984. Some acoustic aspects of fundamental frequency of CVC utterances in aphasia. Phonetica, 41, 103-l 11. Ryalls, J. 1986. A study of vowel production in aphasia. Brain and Language, 29, 48-67. Ryalls, J. 1987. Vowel production in aphasia: Towards an account of the consonant-vowel dissociation. In I. Ryalls (Ed.), Phonetic approaches to speech production in aphasia and related disorders. Boston: College-Hill. Pp. 23-43. Shankweiler, D., Harris, K., & Taylor, M. 1968. Electromyographic studies of articulation in aphasia. Archives of Physical Medicine and Rehabilitation, 49, 1-8. Ziegler, W., & von Cramon, D. 1986. Timing deficits in apraxia of speech. European Archives of Psychiatry and Neurological Sciences, 236, 44-49.