Journal of Phonetics (1993) 21, 479-489
Utterance-final lengthening and the duration of final-stop closures Rochele Berkovits Department of Linguistics, Tel Aviv University , 69978 Tel Aviv, Israel Received 14th June 1991 , and in revised form 19th January 1993
The distribution of utterance-final lengthening within the final stressed syllable was investigated, focusing on the relative proportion of lengthening in the vowel and final stop in Hebrew. Six native speakers of Hebrew read matched pairs of open- and closed-syllable key words occurring in non-final and sentence-final position in phonetically matched sentences. Vowels in final open and closed syllables did not differ in magnitude of lengthening . In the case of final closed syllables, the lengthening effect increased on each successive segment of the final syllable. The proportion of lengthening of the final-stop closure in utterance-final vs. non-final position was more than twice that of the preceding stressed vowel. The finding that the final plosive shows the greatest effect of utterance-final lengthening indicates that acoustic studies of final lengthening which focus on the duration of the vowel by excluding measurements of the final-stop closure may be overlooking a significant aspect of the lengthening process .
1. Introduction Among the acoustic phenomena which characterize the temporal organization of speech, the process of final lengthening has been extensively investigated, in many languages and with diverse experimental paradigms. Delattre (1966) measured syllable durations in extemporaneous speech in English, German, Spanish and French, and found that final syllables were longer than non-final syllables. Oller (1973) found that vowels in final syllables of nonsense utterances were longer than those in other syllables in the utterance and that consonants in final syllables were generally longer than those in medial syllables. In an analysis of segment duration in the readings of prepared scripts, Crystal & House (1988a) found that vowels preceding word-final consonants which appeared at a syntactic boundary were longer than those preceding non-prepausal word-final consonants. Klatt (1975) computed the average duration of each phonetic segment in a passage of connected speech and then noted the location of segments whose duration exceeded their median values. All the lengthened segments appeared in the final syllable of a word which occurred at some syntactic boundary. These studies involve syntagmatic comparisons in the sense that segments in final syllables are horizontally compared with those in non-final syllables. 0095-4470/93/040479 + 11 $08.00/0
Academic Press Limited
Final lengthening has also been investigated paradigmatically with matched sentence pairs in which the same key word appears in final vs. non-final position. Luce & Charles-Luce (1985) found that the duration of vowels in monosyllabic test words in clause-final position was greater than the corresponding segments in non-final position. Kohler (1983) compared syllable durations of key words in utterance-final and non-final positions in German, and found that the final syllable of the utterance-final word showed the greatest amount of lengthening. These studies indicate that final lengthening primarily affects the syllable, or the vowel of the syllable, that immediately precedes a structural boundary. Nevertheless the results of these studies do not reveal how the lengthening effect is distributed within this final syllable. In order to determine how speakers program final lengthening it is necessary to focus on the relative durations of the individual segments of the final syllable . Cooper & Danly (1981 , Experiment 2) investigated utterance-final lengthening as a function of vowel type and voicing of the final consonant. The key words bead, beet, bid and bit were measured from the release of the initial stop up to the offset of the vowel. This, in fact, is the same measurement convention employed by Cooper & Paccia-Cooper (1980) throughout their study of the relationship between syntactic structure and temporal aspects of speech production. Most of the key words in their experiments were monosyllabic with voiceless stops in initial and final positions. In measuring the duration of key words which appeared at structural boundaries vs. the same words which preceded no boundary, Cooper & PacciaCooper included the closure of the final stop, a release burst, and any actual pause preceding the onset of the following word, in a measure of "pause" duration. They appear to take the position that the final-stop closure essentially serves no function in the lengthening processes under consideration, saying, " ... it seems that any significant lengthening of a word segment is due primarily to elonagation of the vowel and any other sonorant or continuant phonemes within the word, while any substantial lengthening of the pause interval can be attributed primarily to pausing between words rather than to lengthening of the word-final stop consonant closure" (p. 31) . In contrast, recent studies on the articulatory implementation of prosodic effects indicate that final lengthening primarily affects the latter portion of the vocalic gesture. In an analysis of jaw movements, Edwards & Beckman (1988) and Edwards , Beckman & Fletcher (1991) found that phrase-final lengthening exerts a greater effect on the closing mandibular gesture. This suggests that the vowel nucleus may not show the primary effect of utterance-final lengthening. Though neither Cooper & Paccia-Cooper (1980) nor Cooper & Danly (1981) cite empirical evidence for the assumption that the burden of final lengthening is carried by vowels , the study that is generally referred to (e .g., in Klatt, 1975, 1976) in support of this claim is that of Oller (1973). Oller modeled his study of English speakers on a study by Lindblom (1968 , cited in Oller, 1973), who investigated final lengthening in Swedish . The results of the two studies indicated that the lengthening of vowels in final syllables was greater than final-syllable consonant lengthening. It should once again be noted, however , that these studies compared durations of segments in final syllables against comparable segments in other positions in the same utterance. The finding that consonants in final syllables are longer than, for example, consonants in medial syllables, but that this lengthening effect is not as great as the corresponding
Final lengthening and final-stop closures
vowel comparisons, does not address the issue of the relative distribution of the lengthening effect over the vowel and final consonant of the final syllable. The studies in which Cooper & Paccia-Cooper (1980) investigated the effect of phrase structure on segment duration generally involved key words with initial and final voiceless stops . On intuitive grounds alone it is not unreasonable to assume that stop sounds are the least likely candidates for segments that would show a lengthening effect. Oller (1973), in fact, found that plosives show less final lengthening than do fricatives, and Klatt (1976) characterized the lengthening of final syllables as a process in which "most of the durational increment [is] restricted to the vowel and any postvocalic sonorant or fricative consonants" (p. 1211) . On the other hand , there is evidence that the duration of the stop closure is manipulated in speech production , varying as a function of factors such as utterance length and rate of speaking. Thus , Lisker (1957) notes that closure durations are generally shorter in longer utterances . Port (1981) found that the duration of medial and final closures decreases as the number of syllables increases, and that medial stop closures are shortened with an increase in speaking rate. Moreover , in their analysis of connected speech , Crystal & House (1988a) found longer final-stop closures in words preceding syntactic boundaries than in words in non-final position. These results replicated the findings of one experiment reported by Luce & Charles-Luce (1985) which compared key words in clause-final and non-final positions , though sentence position did not affect closure duration in a second experiment in which the key words were embedded in a different phonetic environment. Whereas Oller (1973) specifically instructed his subjects to release the final voiced stops in reading the test utterances , he claims that no such instruction was necessary with respect to final voiceless stops since the latter are normally released in American English . Crystal & House (1988a) found , in fact , that voiceless stops tend to be released more often than voiced stops , particularly in word-fin al position , with a higher proportion of releases in prepausal word-final stops. Cooper & PacciaCooper (1980) , on the other hand , claim that most readings produced by their subjects did not exhibit the release burst. Since final stops are apparently always released in Hebrew (D . Gil ; A . Laufer , personal communication) , the present study was carried out in order to investigate the distribution of utterance-final lengthening within the final syllable. Evidence for utterance-final lengthening in Hebrew was reported in Berkovits (1991) on the basis of whole word measures of key lexical items appearing in phrase-final and sentence-final positions. The present study similarly involves paradigmatic comparisons between key words at these two positions , but focuses on the relative proportion of lengthening displayed by the individual segments of the final syllable. The final stressed segments in the key words of Berkovits (1991) contained inherently long vowels. Though vowel quality was not an independent variable in the present study , the vowel inventory of the key words was expanded to include all the vowels of Hebrew. Furthermore the present study included matched pairs of open- and closed-syllable key words in order to determine whether these two types of syllables are similarly affected by final lengthening. As in other languages (Maddieson, 1985), Hebrew vowels tend to be longer in open than in closed syllables (Marbe, 1971). Oller (1973) found indications that vowels in final open syllables are lengthened less than those in final closed syllables. On the other hand, Crystal & House (1988b) reported that vowels in word-final open syllables were
lengthened to a greater degree than those in closed -syllables. The present investigation will assess the interaction of final lengthening and syllable type on the basis of paradigmatic comparisons.
2. Method 2.1. Materials Five pairs of key words in Hebrew were chosen. The words in each pair contained the same phonetic segments, differing only in the fact that one word ended in an open syllable while the second was closed by a final voiceless stop /t/. In three of the five pairs, the final stop functioned as the feminine marker added to a noun, a verb or an adjective, whereas in the remaining two pairs, the addition of the stop resulted in a morphologically unrelated lexical item. The words, which are listed in Table I, contained two or three syllables with the final syllable receiving stress, as is TABLE
I. Key word pairs and test sentences
" wondrous" (masc.) " wondrous" (fern.)
sone so net
" hates" (masc.) " hates" (fern.)
Jam a Jam at
" heard" " dropped"
" his friend" (masc.) " friends" (fern.)
" a squash" " a decoration"
(a) yosi ra?a ru?ax pi/?i (Yossi saw a wondrous [m.] spirit [m.].) (b) yosi ra?a ru?ax pif?i baxalom (Yossi saw a wondrous [m.] spirit [m.] in the dream.) (c) yosi ra?a ru?axpif?it (Yossi saw a wondrous [f.] spirit [f.] .) (d) yosi ra?a ru?ax pif?it baxalom (Yossi saw a wondrous [f.] spirit [f.] in the dream.) (a) xumus yona sone (Yona [m.] hates [m.] hummus.) (b) xumus yona sone le?exol (Yona [m.] hates [m .] to eat hummus.) (c) xumus yona sonet (Yona [f.] hates [f.] hummus .) (d) xumus yona sonet le?exol (Yona [f.] hates [f.] to eat hummus.) (a) et hataklit haze avi lo fama (A vi did not hear this record.) (b) et hataklit haze avi lo fama bixlal (A vi did not hear this record at all.) (c) et hataklit haze avi lo famat (Avi did not drop this record.) (d) et hataklit haze a vi lo Jamat bixlal (A vi did not drop this record in fact.) (a) ra?iti oto medaber im xavero (I saw him talking with his friend [m .].) (b) ra ?iti oto medaber im xavero leyad hakafiteriya (I saw him talking with his friend [m .] near the cafeteria.) (c) ra?iti oto medaber im xaverot (I saw him talking with friends [f.].) (d) ra?iti oto medaber im xaverot leyad hakafiteriya (I saw him talking with friends [f.] near the cafeteria.) (a) ron it hitbakJa lehavi kifu (Ronit was asked to bring a squash.) (b) ronit hitbakJa lehavi kifu lamesiba (Ronit was asked to bring a squash to the party.) (c) ro nit hitbakJa lehavi kifut (Ronit was asked to bring a decoration.) (d) ron it hitbakJa lehavi kifut lamesiba (Ron it was asked to bring a decoration to the party.)
Final lengthening and final-stop closures
generally the case in Hebrew (Bolozky, 1978, 1982). Each of the five vowels of Hebrew (Bolozky , 1978) was in the final stressed syllable of one of the word pairs. The words of each pair were embedded in identical sentences which were constructed in two versions. In one version the key word appeared in sentence-final position, whereas in the other, it appeared in non-final position. Table I gives the four sentences which were constructed for each of the five word pairs. The 20 test sentences, together with 26 fillers, were distributed over eight lists of sentences . Each list contained either five or six sentences, such that no list included more than one version of each key word pair.
2. 2. Subjects and procedure
Six female native speakers of Hebrew read the 46 sentences. They were instructed to first read each sentence silently so as to understand its meaning. They then read the sentence aloud, once for practice and a second time for recording. They were asked to speak naturally , using a moderate speaking rate. Sentences which they felt had not been appropriately read could be repeated. The material was recorded in a recording studio at the Language Instruction Center of Tel Aviv University. The recordings were made at 7.5 in/son a Sony TC-270 tape recorder using a Sony F-25 microphone.
2.3. Acoustic analysis Broad-band spectrograms were prepared for each of the test sentences on a Kay Digital Sona-Graph, Model 7800, at the Center for Research in Speech and Hearing Sciences of the City University of New York. Durational measures were made using a grid which allowed measurement accuracy to within 10 ms. Corresponding measurement points were taken in the two versions of each key word. In addition to whole word measures, durations were obtained for the final stressed syllable in each of the 10 key words. Syllables closed by final voiceless stops were measured up to the final release burst. In the case of the five key words terminating in a stop, additional measures were taken for the final syllable up to, but not including, the closure of the final stop, and for the closure itself. In two key word pairs, segments preceding the final syllable were included in the duration of the final stressed syllable . The final "syllable" in xavero I xaverot included the duration of the immediately preceding vowel , since the syllable-initial uvular /r/ could not be easily segmented. The initial glottal stop of the final syllable in pif?i/ pif?it is often not produced. In this case the final "syllable" interval was measured from the vowel of the initial syllable , since the intervocalic /1/ in the resulting pili/pi/it could not be easily isolated. With respect to the remaining three key word pairs , sone/sonet, fama/famat, and kifu/kifut, which proved to be more easily segmentable, measures were obtained for each segment of the final stressed syllable : the initial consonant, the vowel nucleus, and , in the case of the closed syllables, the closure of the final stop . Vowels in open syllables in utterance-final position were measured up to the end of visible energy in the first formant. Vowels in closed syllables were measured up to the onset of closure.
R. Berkovits II. Means (and standard deviations) of durations (ms) for the five key word pairs (n = 29)
Syntactic position Final syllable Open Key word Final " syllable" Closed Key word Final " syllable" Final " syllable" up to closure Final closure
454 (29) 260 (32)
370 (19) 198 (12)
554 (37) 370 (37) 238 (25) 132 (16)
413 (34) 246 (27) 189 (21) 57 (11)
3. Results Table II shows the means and standard deviations of the durational measures for the five key word pairs averaged across the six speakers . Analyses were performed across speakers on the means of the five utterances. In one instance, the closed-syllable key word sonet in non-final sentence position was incorrectly read by one of the speakers . This token was excluded from the analysis, together with the speaker's reading of the same key word in sentence-final position and the corresponding two open-syllable key word tokens. Key words with final open syllables were 84 ms (23 % ) longer in sentence-final vs. non-final position, t(S) = 11.40, p < 0.001. The final " syllables" of these words showed 31% lengthening (62 ms) in sentence-final position , t(5) = 5.84, p < 0.01. Proportion of lengthening was calculated by taking the difference between durations in sentence-final and non-final positions and dividing by the duration in non-final position . Key words with final closed syllables were 141 ms (34 % ) longer in sentence-final position , t(5) = 9. 77, p < 0. 001; the final " syllables" of these words , which were measured up to the stop release , showed 50% lengthening (124ms), t(5)=8.71 , p
Final lengthening and final-stop closures
(28 ms) in sentence-final position , t(5) = 5. 75, p < 0.01 , and the vowel showed 45 % lengthening (37 ms), t(5) = 5.87, p < 0.01. The corresponding key words whose final syllables were closed by voiceless stops were 147 ms (37%) longer in sentence-final position, t(5) = 10.20, p < 0.001. The initial syllable in these words showed 10% (16 ms) lengthening , t(5) = 5.56, p < 0.01, while the final syllable was lengthened 58% (131 ms), t(5) = 9.19, p < 0.001. In this final syllable , the initial consonant was 21 ms (23 %) longer in sentence-final position , t(5)=3.92, p<0.05, the vowel was 38ms (54%) longer , t(5)=5.48, p<0.01, and the final closure was 72ms (116%) longer , t(5)=8.91, p < 0. 001. The magnitude of lengthening of the final closure was greater than that of the vowel, t(5) = 3. 93, p < 0.05 . In non-final sentence position , the difference between the duration of the initial consonant in open vs. closed final syllables was not statistically significant. In sentence-final position, the initial consonant was 16 ms (14%) longer in open syllables, t(5) = 5.08, p < 0.01. Vowels were 16% (11 ms) longer in open vs. closed final syllables in non-final sentence position , t(5) = 2. 78, p < 0.05, and 10% (11 ms) longer in open syllables in sentence-final position , t(5) = 3. 90, p < 0.05. No difference was revealed between the magnitude of utterance-final lengthening of the initial consonant of the final syllable in open and closed syllables. Similarly, the difference between the magnitude of lengthening of the stressed vowel in open and closed syllables was not significant. 4. Discussion
These results indicate that the proportion of lengthening of the final-stop closure in utterance-final vs. non-final position (116 % ), is more than twice that of the
Mean duration (ms)
Figure 1. Mean durations (ms) for the segments of three open syllable (sone , fama , kifu) and closed syllable (sonet, famat , kifut) key word pairs in two syntactic positions (n = 17).
preceding stressed vowel (54 % ). In the five closed-syllable key words of the present study , 40% of the average lengthening of the final " syllable" was due to the lengthening of the portion of the final "syllable" up to the final closure , whereas 60% was due to the lengthening of the closure itself. With respect to the three key words which lent themselves more easily to segmentation , the lengthening of the initial consonant of the final syllable accounted for 16% of the lengthening of the final syllable, the lengthening of the stressed vowel accounted for 28%, while 56% of the overall lengthening of the final syllable was due to the lengthening of the final closure. Thus in the case of final stressed syllables closed by voiceless stops, the final plosive shows the greatest effect of utterance-final lengthening. The relatively small standard deviations associated with the closure durations in this study contrast with the considerable variability reported by Cooper & Paccia-Cooper (1980) for their "pause" data , indicating that , unlike the idiosyncratic nature of pausing, the process of final lengthening exerts a consistent effect on the duration of final-stop closures. The present results also reveal the systematic nature of the distribution of lengthening over the individual segments of the final syllable. The proportion of lengthening of the stressed vowel (54%) was more than twice that of the initial consonant of the syllable (23 % ) , and lengthening of the final-stop closure (116%) was more than twice that of the vowel. Thus the lengthening effect appears to be progressively distributed from left to right across the segments of the final syllable. This left to right increase in lengthening within the final syllable is analogous to Kohler's (1983) finding that utterance-final lengthening has the strongest effect on the final syllable and gradually decreases towards the beginning of the word. Thus the pattern of final lengthening which characterizes the relative duration of syllables within the word is repeated within the final syllable itself. The initial syllable of key words with final open and closed syllables was significantly longer in utterance-final position . The final lengthening effect thus not only increased the duration of each segment of the final syllable, but also extended back to the initial unstressed syllable. These results were also found in a recent study (Berkovits, 1993) with Hebrew bisyllabic key words terminating in fricatives. Moreover, in the latter study , final fricatives showed significantly more utterance-final lengthening than the preceding stressed vowel. This indicates that the pattern of progressive lengthening within the final syllable, uncovered in the present study, applies to a wider set of segment types in Hebrew. Furthermore, the effect appears to be generalized to other languages as well. In a recent study with nonsense monosyllables, comparing ataxic dysarthric speakers of French and normal controls (Beii-Berti, Gelfer, Boyle & Chevrie-Muller, 1991), final-stop closures in the normal group showed more utterance-final lengthening than the vowel. Finally, in a pilot experiment (Berkovits , unpublished data) using the English sentence pairs of Cooper & Danly (1981, Experiment 2) with consistently released stops, closures of final voiced and voiceless stops were lengthened to a greater extent in sentence-final position than the preceding stressed vowels. Thus the results of the present study are neither restricted to final-stop closures, nor unique to Hebrew. The evidence for progressive utterance-final lengthening suggests an increasing deceleration of the speech gestures during the production of successive segments of the final syllable. As such , the overall results of the present study are consistent with the "edge effect" at phrase boundaries described in the articulatory studies of Edwards & Beckman (1988) and Edwards, et al. (1991). Nevertheless, the
Final lengthening and final-stop closures
most direct interpretation of their model would not predict the observed lack of interaction between syllable structure and the magnitude of final lengthening of the vowel. Since the vowel gestures in open syllables are utterance-final , vowels in open syllables should show a greater degree of final lengthening than those in closed syllables. The results indicated, however, that vowels in open and closed syllables are similarly affected by utterance-final lengthening. The finding that final lengthening in closed syllables is greater than in open syllables appears to be due entirely to the final-stop closure, since no differences in magnitude of lengthening between open and closed syllables were revealed with respect to either the initial consonant or the vowel of the final syllable. In both syllable types, vowels were approximately 37 ms longer in sentence-final vs. non-final position. The finding that vowels in utterance-final position are lengthened to the same extent regardless of whether the syllable is open or closed is parallel to Kohler's (1983) observation that words in utterance-final position in German receive a constant duration increment regardless of the number of syllables that they contain . These results suggest that the process of utterance-final lengthening may, in some instances, be independent of other conditioning factors which affect the duration of segments. Vowels in open syllables in the present study were 11 ms longer than in closed syllables, in both sentence-final and non-final positions. A possible confound in non-final position is that the following segment in the sentences was voiced for the open syllables but voiceless for the closed. However, this probably does not fully explain the difference in vowel duration, since there was as large an effect of syllable structure in sentence-final position, where the vowel in the open syllable was not followed by a voiced segment. The finding that vowels in final open and closed syllables do not differ in magnitude of lengthening contrasts both with Oller's (1973) results indicating greater vowel lengthening in closed syllables and Crystal & House's (1988b) report of greater vowel lengthening in open syllables. Vowel lengthening in Oller's study was established syntagmatically by comparing the duration of vowels in final syllables with those in other positions in the same nonsense utterance. Crystal & House determined lengthening by averaging the durations of all the vowels in their corpus of connected discourse. Moreover they study a general effect of prepausal lengthening, in which the duration of vowels preceding any syntactic boundary is compared with that of vowels which precede no boundary. The present results are based on paradigmatic comparisons between vowels in key words appearing in utterance-final and non-final positions in matched sentence pairs. The diverse results may be attributed to the different experimental paradigms employed, indicating that differences in the comparisons on the basis of which lengthening is defined may produce contradictory findings. The present results indicating progressively greater lengthening within the utterance-final syllable have obvious implications for speech synthesis. The effects of utterance-final lengthening are not included in the algorithm presented by Cooper & Paccia-Cooper (1980), which predicts the duration of lexical items in spoken utterances on the basis of syntactic and extrasyntactic factors. The finding that final-stop closures carry the burden of utterance-final lengthening is furthermore relevant to the methods of duration measurement employed in acoustic studies of speech timing. As was confirmed by the present results, studies of final lengthening which focus on the duration of the vowel may, in fact, be overlooking a significant
aspect of the lengthening effect. In addition to the systematic exclusion of the duration of final-stop closures in investigations of final lengthening (Cooper & Paccia-Cooper, 1980; Cooper & Danly, 1981) noted earlier, similar measurement conventions are found in studies of other durational processes. Weismer & Ingrisano (1979) used the utterance Bob hit the big dog to study the durational effects of emphatic stress and speaking rate. Based on considerations of ease of measurement , the content words, all of which terminated in final stops, were measured up to the offset of the vowel, with the final stop ignored throughout. The finding that the final-stop closure shows the greatest effect of utterance-final lengthening does not necessarily imply a similar pattern of lengthening in the case of emphatic stress. Further studies are needed to determine whether the distribution of lengthening uncovered in the present investigation applies to other lengthening processes. Current research (Berkovits, in preparation) compares the lengthening effect in utterance-final position to that in contrastive stress and in gapped constructions, and examines the extent to which the pattern of lengthening found in final stressed syllables is revealed in the case of final syllables that are unstressed. I am grateful to the Center for Research in Speech and Hearing Sciences of the City University of New York for the use of their laboratory facilities . In addition, I would like to thank David Gil and Ronit Shoshany for their assistance in constructing the test materials .
References Bell-Berti, F. , Gclfer , C., Boyle , M. & Chevrie-Muller , C. (1991) Speech timing in ataxic dysarthria , Proceedings of the X lith international congress of phonetic sciences, Vol. 5, pp. 262-265. Aix-enProvence: University of Provence. Berkovits, R. ( 1991) The effect of speaking rate on evidence for utterance-final lengthening, Phonetica , 48,57- 66. Berkovits , R. ( 1993) Progressive utterance-final lengthening in syllables with final fricatives , Language and Speech , 36, 89-98. Berkovits , R. (in preparation) Durational e ffects in final lengthening, gapping and contrastive stress. Bolozky , S. (1978) Some aspects of Modern Hebrew phonology. In R. A . Berman , Modern Hebrew structure pp. 11-67. Tel Aviv: University Publishing Projects. Bolozky, S. (1982) Remarks on rhythmic stress in Modern Hebrew , Journal of Linguistics, 18, 275-289. Cooper, W. E. & Danly , M. (1981) Segmental and temporal aspects of utterance-final lengthening, Phonetica, 38, 106-115. Cooper, W. E. & Paccia-Cooper, J. (1980) Syntax and speech. Cambridge: Harvard University Press. Crystal, T. H. & House, A. S. (1988a) Segmental durations in connected-speech signals: current results, Journal of the Acoustical Society of America, 83, 1553-1573. Crystal , T. H. & House , A . S. (1988b) Segme ntal durations in connected-speech signals: syllabic stress, Journal of the Acoustical Society of America , 83, 1574-1585. Delattre, P. (1966) A comparison of syllable length conditioning among languages , International Review of Applied Linguistics , 4, 183-198. Edwards , J. & Beckman , M. E. ( 1988) Articulatory timing and the prosodic interpretation of syllable duration , Phonetica, 45 , 156-174. Edwards, J ., Beckman , M. E. & Fletcher, J. ( 1991) The articulatory kinematics of final lengthening, Journal of the Acoustical Society of America , 89, 369-382. Klatt , D. H. (1975) Vowel lengthening.is syntactically determined in a connected discourse, Journal of Phonetics , 3, 129-140. Klatt , D. H. (1976) Linguistic uses of segmental duration in English: acoustic and perceptual evidence, Journal of the Acoustical Society of America, 59, 1208-1221. Kohler , K. J. (1983) Prosodic boundary signals in German , Phonetica, 40,89-134. Lisker, L. (1957) Closure duration and the intervocalic voiced-voiceless distinction in English, Language, 33,42-49. Luce , P. A. & Charles-Luce, J. (1985) Contextual effects on vowel duration , closure duration, and the consonant/vowel ratio in speech production , Journal of the Acoustical Society of America, 78, 1949-1957.
Final lengthening and final-stop closures
Maddieson, I. (1985) Pho netic cues to syllabification . In Ph onetic linguistics: essays in honor of Peter Ladefoged (V. Fromkin , editor) pp . 203-221. Orlando : Academic Press. Marbe, A. J. (1971) Th e phonetics of Modern (Israeli) Hebrew. Unpublished Doctoral dissertation, University of Edinburgh. Oller, D . K. ( 1973) The effect of position in utterance on speech segme nt duration in English, Journal of the Acoustical Society of America, 54, 1235- 1247. Port, R. F. (1981) Linguistic timing factors in combination, Journal of the Acoustical Society of America, 69, 262-274. Weisme r, G. & Ingrisano , D. ( 1979) Phrase-level timing patte rns in English: effects of e mphatic st ress location and speaking rate , Journal of Speech and Hearing Research, 22, 5 16-533.