.3 .3 1‘3 11*31 COARTICULATION AS AN ADJUNCT T0 SPEECH PERCEPTION. Thesis for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY WALTER H. MANNING 1 9 7 2 This is to certify that the thesis entitled COARTICULATION AS AN ADJUNCT TO SPEECH PERCEPTION presented by Walter H. Manning has been accepted towards fulfillment of the requirements for Ph.D. dcgreein Audiology and Speech Sciences Daniel S. Beasley, Ph.D. Mnjorprofeuor Date July 11, 1972 0-7639 RARY BINDERS Inn-r mm- ABSTRACT COARTIbULATION AS AN ADJUNCT TO SPEECH PERCEPTION BY Walter H. Manning This study tests the hypothesis that a particular unit of speech production would be more apt to be per- ceived correctly if it were to be more highly coarticulated than if that same unit were to possess a lesser degree of coarticulation. Recent research dealing with human encoding of speech has demonstrated that continuous speech cannot be appropriately thought of as discrete, independent, and commutable phonemic-sized segments. Such research in speech production has suggested that ongoing speech be defined in terms of a dynamic and synergistic process which implies the occurrence of coarticulated supra- phonemic units. Models of speech production which postu- late such supra-phonemic units were employed in the creation of an operational definition of "degree" of co- articulation. Walter H. Manning Fifty subjects listened to 360 two-word combina- tions which provided 180 conditions of greater /sp/, /st/ and /sk/ coarticulation, and 180 conditions of lesser /sp/, /st/ and /sk/ coarticulation. Each subject re- sponded to the stimuli by writing down the phoneme /s/ as they perceived it in the presence of auditory masking via controlled experimental procedures. Subjects were ran— domly divided into five groups of ten individuals each and administered the stimuli in one of five signal-to- noise (S/N) ratios ranging from -6 to +14 dB. Percent- ages of correct responses to the stimuli were calculated and the mean percentage correct scores across all vari- ables (coarticulation, signal-to-noise, consonant cluster and stressing) were computed. The results of this research are in agreement with the findings of previous studies which emphasize the ap- parent relationship between articulatory encoding and auditory perceptual processing. That is, the physio- logical and acoustical overlapping of phonemic features which has been demonstrated in studies of speech produc- tion has also been shown to play a role in the perception of speech. While the results did not support the con- tention that degree of coarticulation alone serves as a critical variable in perception, they did reveal that, where stressing was not present, the more-coarticulated consonant clusters were perceived significantly better Walter H. Manning than the less-coarticulated consonant clusters. Thus, where differences in stressing between the stimulus word pairs could not be employed in the perceptual decision, the amount of coarticulatory information became a critical factor. The results of this study also offer support for models of articulatory production which postulate supra- phonemic units of speech encoding. A production/ perception unit composed of one or more consonant seg- ments followed by a vowel (or vocalic null) was sup- ported as one possible unit of such a system. It would appear that coarticulatory information is carried by both consonants and vowels with the greatest amount of in- formation being available in the consonant-consonant relationship. Lastly, the findings of this study may be interpreted as complementing a motor theory of speech production/perception. Based on the results of this study, models of articulatory production and models of auditory perceptual processing, suggestions are offered for a more complete model of articulatory processing. The findings are also viewed as to clinical application in the areas of arti- culation and stuttering. \/,,—/-7 /‘ .. x J .- 4 7 Approved Date ' Ayn / 2V 46/ / COARTICULATION AS AN ADJUNCT TO SPEECH PERCEPTION BY Walter H. Manning A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Audiology and Speech Sciences 1972 Accepted by the faculty of the Department of Audiology and Speech Sciences, College of Communication Arts, Michigan State University, in partial fulfillment of the requirements for the degree of Doctor of Philosophy. ‘> 7 241 1.0 Director Danie S. Beasley, Ph.D. //,..>...‘..?. QM“ ....._._. William F. ' .D. Thesis Committee: David E. Wessel, PE.D. ii ACKNOWLEDGMENTS The author wishes to express his appreciation to the members of the thesis committee (Dr. William Rintel- mann, Dr. Herbert Oyer, Dr. Robert Ebel and Dr. David weasel) who offered their time and knowledge toward the completion of this research. Most especially, acknowledg- ment is made to Dr. Daniel S. Beasley, who as a thesis director, advisor and friend, has provided insight and guidance which will always be remembered. And to my wife Anne, whose encouragement and support, makes the years of work which this dissertation represents, truly worthwhile. iii TABLE OF CONTENTS LIST OF TABLES . . . . . . . . LIST OF FIGURES . . . . . . . . INTRODUCTION . . . . . . . . . Coarticulation and Speech Production Models of Coarticulatory Production Nature of the Coarticulatory Unit . Coarticulation and Speech Perception Interaction Between Coarticulatory Production and Perception . . Rationale for a Coarticulatory Unit Additional Coarticulatory Stimuli . Effect of Stress on Coarticulation Summary and Statement of Problem . EXPERIMENTAL PROCEDURES . . . . . Subjects . . . . . Design and Stimuli . . Presentation Procedures Analysis . . . . . “SULTS O I I O O O O O O 0 Main Effect of Coarticulation . . Main Effect of Signal-to-Noise Condition Main Effect of Consonant Cluster . Main Effect of Stressing . . . . swat-y O O O O I O O O 0 DISCUSSION 0 O O O O O O O O Coarticulation and Speech Processing The Role of Stressing . . . . . iv Page vi vii Suggestions for a Model of Coarticulatory Processing . . . . . . . . . Problems Inherent in the Present Study . . Implications for Therapy and Future ReseaICh I O O O O O O O O O 0 LIST OF REFERENCES . . . . . . . . . APPENDICES Appendix A. List of Randomized Word Pairs Used in This Study . . . . . . . . B. Instructions Which were Presented to Each Subject Prior to Adminis- tration of Practice and Master LiSts 0 O O O O O O O 0 O O C. ReSponse Sheet and Scoring Form Used by the Subject and Experimenter . . D. Relative Duration of Each Wbrd of the Stimulus Pairs (in msec) . . . . E. Relative Intensity of Each Wbrd of the Stimulus Pairs . . . . . . . . F. Tabled Values of AverageDuration and Intensity of Each Wbrd of the Stimulus Pairs . . . . . . . . G. Mean Percentage Correct Scores for Each Variable Across Each Level of Signal- to-Noise Condition . . . . . . Page 56 59 60 65 71 73 74 84 87 90 91 LIST OF TABLES Table Page 1. Summary of an Analysis of Variance Performed on the Percentage Correct Scores at Two Levels of Coarticulation (C and NC) Factor (A), five levels of signal—to-noise ratio (-6, 0, +4, +10, +14 dB) factor (B), three levels of consonant cluster (sp, st, sk) factor (C) and two levels of stressing (stressed and unstressed) factor (D) . . . 36 2. Summary Table of the Mean Percentage Correct Main Effects of Two Factors: Coarticulation (C and NC) and Stressing (Stressed and Unstressed) . . . . . . . . . . . 37 3. Summary Table of the Mean Percentage Correct Main Effects of the Three Factors: Signal- to-Noise Ratio (-6, 0, +4, +10, +14 dB) Consonant Cluster (sp, st, sk) and Stressing (Stressed and Unstressed) . . . 41 F.l. Summary Table of Average Relative Intensity for All Stimuli . . . . . . . . . . 9O F.2. Summary Table of Average Duration for All Stimuli O O O O O O O C O I O O 90 6.1. Mean Percentage Correct Scores for Each Variable Across Each Level of Signal- to-Noise Condition . . . . . . . . . 91 vi LIST OF FIGURES Figure Page 1. Kozhevnikov and Chistovich's Model of Coarticulatory Speech Production . . . . 4 2. Ohman's Model of Coarticulatory Speech Production . . . . . . . . . . . 6 3. Henke's Model of Coarticulatory Speech Production . . . . . . . . . . . 8 4. Graphic DiSplay of the Mean Percentage Correct Scores at Each of Six Signal- to-Noise Ratio Listening Levels as Found by Manning, Beasley and Beachy (1972) . . . . . . . . . . . . . 16 5. Mean Percentage Correct Scores for Both Levels of Coarticulation and Both Levels of Stressing at Each Signal-to- Noise Condition . . . . . . . . . . 39 6. Mean Percentage Correct Scores for the Three Levels of Consonant Cluster at Each Signal-to-Noise Condition . . . . . 40 7. Mean Percentage Correct Scores for the Three Levels of Consonant Cluster and the Two Levels of Stressing at Each Signal-to-Noise Condition . . . . . . 42 8. A Synergistic Model of Coarticulatory Processing . . . . . . . . . . . 57 vii INTRODUCTION Coarticulation and Speech Production Recent research has demonstrated that encoded speech signals, rather than being simply the production of a linear string of discrete phonemes, are actually the re- sult of the complex process of coarticulation (Daniloff, 1971). That is, the acoustic and physiological charac- teristics of a specific phone can be seen to overlap and influence the characteristic features of preceding and succeeding phones, termed "forward” and ”backward” co- articulation respectively. While forward coarticulation has been attributed to neuromuscular programming of the articulatory mechanism, backward coarticulation has been attributed to mechanoinertial factors associated with articulatory movement (Ohman, 1966; Daniloff and Moll, 1968: Daniloff, 1971). Both forms of coarticulation, forward and backward, have been found to occur in acoustic, electromyographic, and articulatory movement data. Stevens and House (1963), Lindblom (1963), Stevens, House and Paul (1966), and Ohman (1966) have demonstrated coarticulation in the acoustic wave for CVC and VCV syllables in which vowels and consonants mutually influence each other in forward and backward directions, over as many as two phonemes. Fromkin (1966), Ohman (1967), Lubker (1967), and MacNeilage and DeClerk (1967) have presented electromyographic tracings which suggest extensive coarticulation of speech gestures dependent upon phonemic environment in both for- ward and backward directions. Also, cinefluorographic and high speed motion picture studies of articulatory production (Truby, 1959; Fujimura, 1961a: 1961b: Kozhevnikov and Chistovich, 1966: Lindblom, 1968; Daniloff and Moll, 1968) have indicated extensive vowel-consonant coarticulation of lip and jaw configuration in forward directions. Such findings have led to the development of various models which postulate supra-phonemic units of speech encoding. Models of Coarticulatory Production Utilizing various electropalatographic and photographic techniques, Kozhevnikov and Chistovich (1965) studied the articulation of Russian. Among their con- clusions they stated: . . . in syllables of the consonant-vowel type all the movements of a vowel which are noncontradictory (i.e. involve the same articulator(s) to perform dIfferent articulations) to the articulation of the consonant begin with the beginning of the syllable. (p. 543) Kozhevnikov and Chistovich proposed a model of articulatory production wherein motor commands for the entire syllable are specified simultaneously with the start of the first consonant when the commands are non-competing.1 Where .competing commands are present, execution is dependent upon the order of the consonants and their specified move— ments. Based on their electromyographic observations and data concerning relative time invariances of different speech segments, Kozhevnikov and Chistovich advanced the concept of the "articulatory syllable" consisting of any number of consonants followed by a vowel, as the basic unit of the articulatory program. With priority being given to the coordination of articulatory movements within the syllable, there should be maximum coarticulation within such segments and minimum coarticulation between (see Figure 1). ‘ Ohman's (1967) numerical model of coarticulation advances the notion that a vowel-consonant-vowel (VCV) syllable is the most appropriate coarticulatory unit. The construction of this VCV unit involves a diphthongal vowel gesture proceeding from the initial to the final vowel. Superimposed upon this vowel configuration is an inde- pendently produced consonant gesture. Coarticulation 1Non-competing commands are those that do not involve the same articulatory structure or the same portion of the structure. .cOAuosooumnooomm Shouussowuuoou mo Hove: m.n0a>oumasu can >och>onuoxuu.d Gunman 4m)”.— 4m>3 , 4§§8< >SA§SE< Sign .35 «23.: . Aegis? 1 tie V . QR . 1|! Sues? i l 89$ _ .4 M. >88 ‘ >68 >68 Tl >08 +|>T Tl >8 9. a w .w A n 3 . e geese between and among the vowels and consonants of this unit (in terms of tongue shape and jaw position, for example) is a function of the commonality of the articulators in- volved. This particular model, however, has not received strong support in the literature. That is, while Ohman's system may be used to describe the coarticulation of lip rounding as found in Daniloff and Moll's-(l968) data, the model does not provide an explanation of such results. Daniloff and Moll did find that their data was somewhat more consistent with the coarticulatory models of Kozhevnikov and Chistovich (1965) and Henke (1967). Amerman, Daniloff and M011 (1970) found that their data on coarticulation of jaw lowering and lip retraction only partially support the Kozhevnikov-Chistovich and Henke models of articulation. The authors make no mention, however, of their findings relative to Ohman's model. Various attempts have been made (Henke, 1967; Coker, 1967) to create computer-based articulation models for use in the development of synthetic Speech. These models are usually based on the assumption that the articulatory structures seek articulatory targets corre- sponding to phonemes. Henke's model describes "articula- tory goals" wherein each goal is an ideal position and/or shape of some portion of the articulators. The basic input-output unit of the system is supra-phonemic but sub- syllabic in nature. More important, the model is a dynamic .cofluosooum nooomm huoumasofluumoo uo Hoooz m.cmen0I|.N ousmwm £3: .35.. .3589. 525852 azzzzcsa 4.95 «we... {SEE k Tax}. 2358 Fag SEEKS-Koo on Resume gag toss ox secexstegefi as. at E >o> +II. >o>+|| >o>+lnl > > \aflufi gm. 0.3th Suits hotssosd fianuekkaud Egia one which employs an ongoing description of the status of the articulatory structures. Thus while each unit repre- sents a discrete and finite set of articulatory features, the Speed of production characteristic of ongoing speech results in significant modifications of the articulatory end-product. With this introduction of the time factor into the process, the articulatory features, while they do not blend into one another, form a composite goal con- sisting of present and future input segments. To account for this process Henke prOposed a high level look-ahead or scanning of the as yet unproduced supra-phonemic units. Such a scanning process allows for a coarticulatory anti- cipation of future units when this anticipation does not conflict with the articulatory features of more immediate phonemes. This scanning process, along with the subse- quent creation of composite goals, Henke has termed pre or forward coarticulation. However, the allophonic effects of previously articulated phonemes also influence production and result in additional modification of the articulatory pattern. While these phonemes, once articu— lated, are forgotten, the affect of their production on the position, velocity and force of the articulatory structpres influence the articulators as they are directed to current target goals. When the articulators cannot move rapidly enough to satisfy the original target commands, post or backward coarticulation occurs. .cofiuOBooum nooomm mucuMHDOHuumoo mo Hoooz m.oxcomun.m ousoflm 4mg dam 455$ >u8§2§< Saigofi .364 guru... MW gets Ea. Sages Tll «356 all... Resins... . ski «gagkxigst hug g... on -g gfigé , . Raga Egg zoFfizuFaau 92353 292.532.5300 g m2: Thus in Henke's model, backward coarticulation results from inertial-physiological effects, while forward co- articulation is the result of neurological programming. Nature of the Coarticulatory Unit There is much disagreement in the literature concerning the nature and size of the basic coarticula- tory unit. Indeed, it has been proposed that the size of such units may vary with the demands of the communi- cation medium in terms of time, stress, distortion, and information (Daniloff, 1971). Fromkin (1966), Henke (1967), Coker (1967) and Wickelgran (1969) employed sub-syllabic units in their discussions of coarticulatory behavior. Other researchers, while they agree on some type of syllable as the most basic unit of coarticulatory production, are not in accord as to the make-up of such syllables, i,e, whether they are of a CVC-form (Stevens, House and Paul, 1966), a CV-form (Fumimura, 1961a,b), a C1C2C3C4V-form (Kozhevnikov and Chistovich, 1965) or a VCV-form (Ohman, 1966). Daniloff (1971) suggested that such a syllable might be defined as that size unit over which coarticulation can occur, regardless of its phonemic form. Although the above models may vary relative to definition of a coarticulatory unit, there is agreement that more coarticulatory behavior is apt to occur within 10 a specified programmed unit than between such units. For example, Kozhevnikov and Chistovich (1965) maintain that coarticulatory overlapping of gestures within a syllable is as complete as possible. They further stated that the degree of overlap within the consonant c1uster(s) of the unit (C1C2C3C4V) may be even more pronounced than that between consonant and vowel. In addition, research re- sults suggest that the degree to which the features of neighboring phonemes overlap may be at least partially determined by various phonemic and supra-phonemic in- fluences, such as rate of utterance (Henke, 1967), intona- tion and stress patterns (Lindblom, 1965), morphemic boundaries (Lehiste, 1962), and phonetic context (Stevens and House, 1963; Ohman, 1966). Such limitations may vary, however, depending upon the articulatory behavior under study. Coarticulation and Speech Perception If coarticulation occurs during Speech production, the question arises as to its role during perceptual processing of speech stimuli. That a listener perceives overlapping acoustic and/or articulatory features of phones has been shown by Ali, Gallagher, Goldstein and Daniloff (1971) for the characteristic of nasality. Ali 22 31. employed CVC and CVVC syllables in which the final consonants were either nasal or non-nasal. With the final 11 consonant and its vowel-consonant transition spliced away, listeners were asked to predict whether the missing con- sonant was nasal or non-nasal. The authors found that the presence of nasal consonants could be predicted better than chance. Non-nasal consonants were also perceived better than chance. The authors hypothesized that the listeners in their study utilized coarticulatory information to lighten the phoneme-processing load. They also reasoned that since they were able to demonstrate perceptually significant coarticulation of velar movements across CVVN sequences, there were very likely other coarticulatory movements which would be observable. Further, since their subjects were able to employ coarticulatory information in the experimental task, perception, at least in this in- stance, appeared to use or follow production. Lehiste and Shockey (1971) compared the perceptual accuracy of released final plosives (p, t, k) following vowels (i, a, a, u) to the perception of final plosives derived from VCV utterances. In the VCV utterances the C (p, t, k) and the second vowel (l, a, a, u) were deleted by manual mechanical procedures (shears). In such VC syllables, the transition from the vowel to the consonant is influenced by the anticipation of the terminal vowel. For all consonants combined, listeners achieved correct scores of 80% for released and 39% for unreleased final 12 plosives. Initial plosives as well as the mechanically manipulated utterances were equally easily identifiable. In another study Shockey and Lehiste (1971) investigated whether, in a VCV sequence, the initial VC combination carried significant perceptual cues as to the quality of the following vowel. Random lists of VC utterances were played to listeners, who were asked to indicate which vowel had been deleted. The results indi- cated that coarticulation effects did not provide suffi- cient cues to the perception of the vowels under study. Shockey and Lehiste stated that such results " . . . suggest that some coarticulation effects may not have to be taken into consideration in speech synthesis." Interaction Between Coarticulatory Production and’Perception The above findings tend to suggest a possible relationship between articulatory encoding and perceptual decoding. It has been theorized, for example, that pro- duction (encoding) and perception (decoding) are inex- tricably related in a symmetrical way. That is, produc- tion may facilitate correct perception of speech. The most extreme form of such an argument may be found in the motor theory of speech perception (Liberman, Shankweiler, and Studdert-Kennedy, 1967). According to this theory, perception is mediated by reference to neural command signals which are used during production. 13 Assuming that coarticulation does play a signi- ficant role in the production of ongoing Speech, and that the argument for a symmetrical relationship between the encoding and decoding processes of speech stimuli is a sound one, it might be suggested that coarticulation plays a significant role in the perceptual processing of such stimuli. A review of the literature reveals that several studies have hypothesized the importance of coarticulation as an aid in the perception of speech stimuli; others have gone further and investigated the presence of perceptual cues in acoustically altered coarticulatory units. No study, however, has been concerned with differential perceptability as it relates to amount or degree of co- articulation. More specifically, it could be hypothesized that a particular unit of speech production would be more apt to be perceived correctly if it were to be more highly coarticulated than if that same unit were to possess a lesser degree of coarticulation. In order to study this question, Manning, Beasley and Beachy (1972) conducted a study which utilized a pro— duction unit which exhibits both a high and a low degree of coarticulation, depending upon its embedded contextual environment. Such a unit is the commonly occurring con- sonant cluster /st/. Sixty subjects listened to 50 two- word combinations which provided 25 conditions of greater /st/ coarticulation (Condition C) and 25 conditions of 14 lesser /st/ coarticulation (Condition NC). Subjects were randomly divided into six groups of ten individuals and administered the stimuli binaurally in one of six signal- to—noise ratios ranging from -10 to +15 dB, in 5 dB steps. Wideband white noise (also presented binaurally) was used to create the various signal-to-noise conditions. Each subject responded to the stimuli by writing down the two-word combination as they perceived it. Percentages of correct responses to the stimuli were calculated and the mean percentage correct scores for the 10 subjects in each of the six signal-to-noise conditions were found. Graphic representation of the results demon- strated that across the six levels of signal-to-noise ratios, the percentage correct identification for the more highly coarticulated (C) condition remained rela- tively consistent (within ten percentage points). How- ever, the curve for the stimuli with the lesser degree of coarticulation, the NC condition, revealed progress- ively poorer perception as the signal-to-noise ratio decreased. More specifically, the percentage correct scores between the two conditions were essentially the same at the optimum signal-to-noise ratio. However, differences between the two conditions tended to occur as the signal-to-noise relationship became less favor- able. The differences in perceptual accuracy of the two lists remained fairly constant through signal-to-noise 15 ratios of +10, +5 and 0 dB. Beyond 0 dB the differences became progressively greater (see Figure 4). The findings of the above study were in basic agreement with earlier studies of a similar nature in emphasizing the apparent relationship between articulatory encoding and auditory perceptual processing. More speci- fically, such results suggested that, as listening condi- tions became more difficult, listeners were more apt to correctly perceive stimuli which exhibit a relatively high degree of coarticulation. The authors, however, employed only one consonant cluster (/st/) in their study. Also, no attempt was made to systematically vary the potentially significant variable of stressing in the word pairs. Rationale for a Coarticulatory Unit The findings of Manning, Beasley and Beachy (1972) would appear to further support the hypothesis that co- articulatory behavior occurring during encoding is an asset to accurate decoding of linguistic messages. The authors suggested that additional study should provide important information relative to the further clarifica- tion of such interaction. Thus it was decided to expand upon the general design of this original study in order to further explore such relationships. Again, as in the original study, a production unit which exhibited both a high and a low degree of coarticulation was needed. As 16 PERCENTAGE CORRECT :00 '1” 40 .L 80 '4’ PM 7f).h .z””‘ ’2" (90 w 4// 50 “r [r I ‘40’-- I! C‘lO——————O I NC =o----o 30 Iip I I I a“) u: / [I ’0 In I// 0’ ~10 {5' o *5 +10 #5 S/N RATIO (dB) Figure 4.--Graphic display of the mean percentage correct scores at each of six signal-to-noise listening levels as found by Manning, Beasley and Beachy (1972). 17 cited above, the original study employed the commonly occurring phoneme consonant cluster /st/. The following is a presentation of the rationale for the use of such stimuli. The consonant cluster /st/, when occurring intra- morphemically, as in "cap stab," is traditionally referred to as a blend. However, if the two phonemes /s/ and /t/ appear intermorphemically and/or between lexical items, as in ”caps tab," they may not be considered a blend, at least traditionally (Shriner and Daniloff, 1971). Rather, they may be considered to be distinct, though perhaps coarticu- lated, phonemes. Thus, in a blend condition a temporal gap occurs prior to the /st/, whereas in the non-blend situation the gap occurs between the /s/ and the /t/. It appears likely that the articulatory features of the /s/ and /t/ phonemes in the blend /st/ would overlap (or co- articulate) to a greater degree than in a situation whereby the adjacent phonemes /s/ and /t/ were separated by a temporal gap between lexical-bound morphemic units. Ad- mittedly, the time difference between the articulation in the /st/ combination and the context of /kap steb/ and that context provided by /kups tub/ may be rather small. This difference, however, according to Henke's model of speech production, would result in a correspondingly greater or lesser degree of coarticulation. According to Henke (1967), if the temporal interval between neuromuscu- lar programmed input segments to the articulators was of 18 sufficient length, specific articulatory movements asso- ciated with these segments could be performed, target positions achieved, and the articulatory realization of the segments could approach a steady state. However, due to the rapidity of the input to the articulatory mechanism, this does not normally occur during speech production. Further, the more events which occur in a given time per- iod, the more likely it is that the articulatory parameters for the programmed segments will overlap: the more temporal overlap in articulatory movements, the greater the amount or degree of coarticulatory activity. Further support for the use of stimuli embedded in such a context may be found within the Kozhevnikov and Chistovich (1965) model of speech production. They ad- vanced evidence to show that the CV-type syllable is the minimal unit of speech timing, coarticulation and produc- tion. They stressed that the C stands for a single conso- nant or a consonant cluster of any size. As stated above, Kozhevnikov and Chistovich maintained the existence of a high degree of coarticulatory overlapping of gestures within a syllable, with the degree of overlap within the consonant cluster being even more complete than that be- tween consonant and vowel. Since the cluster in a Kozhevnikov-Chistovich type syllable may cross ordinary syllable and word boundaries, their definition of a "cluster" differs somewhat from that previously referred 19 to as a blend which occurs intrasyllabically only. How- ever, the important point is that coarticulation and mutual influence are minimized at C1C2. . .CnV syllable boundaries. It is at these boundaries where smooth flow of articulatory gestures are interrupted by a lesser degree of articulation, or by stress patterns, or by syntactically dictated sentence pauses, that consonant production may be expected to be most difficult (Shriner, Prutting and Daniloff, 1971). In their discussion of the Kozhevnikov-Chistovich model, Shriner, Prutting and Daniloff suggested that the temporal gap between lexical items would also provide a form of syllabic boundary. That is, the temporal gap requires the articulators to operate as though a vowel were programmed to occur between the lexical items. Although one would expect the articu- lators to assume a vowel configuration during the temporal gap, it would appear inaccurate to state that an actual vowel is occurring as part of the programmed utterance. This "non-vowel-temporal-gap" might best be referred to as a "vocalic null" which would at the same time symbolize the overlapping features of coarticulatory production and the lack of a specific programmed vowel. The Greek letter phi /¢/ has been used to designate the vocalic null ’in this study. Investigations of the perception of coarticulatory segments (Lehiste and Shockey, 1971; Shockey and Lehiste, 20 1971) have revealed apparent differences in the co- articulatory cues provided by VC segments of VCV syllables. That is, while the VC segment of the syllable tended to provide cues as to the release or non-release of the C-plosives in the original VCV syllable, this same VC segment did not provide coarticulatory cues as to the quality of the final V in the VCV syllable. Additionally, plosives derived from VCV utterances by the deletion of the initial vowel were also easily identifiable. In re- gard to the stimuli employed in the Manning, Beasley and Beachy study, one would predict that the vowel and the vowel-consonant transition at the end of the first word of the stimulus-pair would contribute coarticulatory information as to the release of the final consonant, and possible additional information as to whether or not another consonant was immediately following. In every instance, the vowels which appear to provide coarticula- tory information about adjacent or nearby consonants are one, or at the most, two positions away from the experi- mental consonant cluster and could be expected to provide a coarticulatory effect involving the /st/ blend. Finally, if the assumption of the existence of a symmetrical relationship between production and percep- tion is valid, then it may be predicted that the intra- morphemic /st/ blend would be more accurately perceived 21 in a difficult listening situation than the /s ¢ t/ combi- nation occurring intermorphemically and between lexical items. Additional‘Coarticulatory Stimuli Assuming that such a rationale would also apply to other consonant clusters, an attempt was made to find additional stimuli which possessed the advantages of the /st/ combination. Two additional consonant clusters, /sp/ and /sk/, were selected. Like the /st/ cluster, the /sp/ and /sk/ combinations have a high frequency of occur- rence in English (Moser, 1969; Roberts, 1965). As with the /st/ cluster, both the /sp/ and /sk/ combinations, when embedded in one of two different contextual environ- ments, exhibit a correspondingly high and low degree of coarticulation. That is, with the /sp/ and /sk/ clusters appearing intramorphemically and intersyllabically as in /kap ¢ span/ and kap ¢ skIt/, respectively, there is, according to the above rationale, a greater degree of co- articulation within the clusters than when these same two clusters appear intermorphemically and intersyllabically as in the context /kups ¢ pan/ and /k¢ps ¢ krt/. Finally, these three consonant clusters (st, sp, sk) were chosen because of their classification in terms of distinctive feature theory (Miller and Nicely, 1955; Chomsky and Halle, 1968). That is, by using phonemes which were at the same time voiceless and non-nasal these two variables were able 22 to be controlled. Further, while the initial /s/ of the consonant cluster is characterized by affrication and rela- tively long duration, the other members of the clusters (the plosives /t/, /p/ and /k/) are differentiated only by place of articulation. Thus the three major points of constriction of the vocal tract--front, middle and back-- are represented by each of the three phonemes /p/, /t/ and /k/ respectively. Effect of Stress on Coarticulation Finding that stress/juncture production tasks were not disrupted by blocking of auditory and tactile feedback channels, Gammon, Smith, Daniloff and Chin (1971) hypothesized that the projection of stress/juncture rules upon the articulatory mechanism occurs at a high level in the decoding process (Kozhevnikov and Chistovich, 1965) and is probably feedback free. Thus, just as the produc- tion/perception of coarticulatory units would appear to be a higher level of perceptual processing, so does the pro- duction/perception of stress and juncture. It has been suggested that stress provides organ- izational cues during the perception of ongoing speech. Jakobson, Fant and Halle (1951) postulated that stressing serves to divide the chain of sound into grammatical units. Bond (1972) investigated this possibility by examining the relationship between stress patterns 23 and the frequency of substitution and ordering errors in the perception of obstruent (speech sounds in which the breath is wholly or partly obstructed) clusters. Bond employed eighteen CVCCVC nonsense words presented under three conditions of stressing and found that subject perceptual performances were best either when both syllables of the word were stressed or when the first syllable was stressed. Bond cOncluded that the explana- tion for her findings could be best attributed to the organizational function of stress and that "stress serves to signal the division of the utterance into grammatical or perceptual units" (p. 4). Looking at the effect of stress patterns on vowel production, Fant (1962) found that a decrease in stress, which is in most cases associated with a decrease in the duration of the vowel, shifts the formant pattern of the vowel towards that of a schwa. Commenting upon the effect of symmetrical consonantal environments on stressed American English vowels, Stevens and House (1963) found that the influence of such a context upon vowels is regularly manifested as a displacement of vowel-formant frequencies away from their.target frequencies. In British as well as American varieties of English, there is a tendency for most vowels in weakly stressed syllables to approach schwa in quality. This phenomenon has been commonly referred to as "gradation" 24 and implies that the reduction in vowel quality varies along a continuum: the amount of "vowel reduction" being related to degree of stress placed on the vowel (Lindblom, 1963). Stetson (1951) observed a regular series of "re- duced values" of vowel production which, at the extreme end of the continuum, resulted in production of the schwa. With an increase in the rate of production, all vowels in unstressed syllables displayed a centralizing effect and arrived at a common schwa. In a Spectrographic study of eight Swedish vowels, Lindblom (1963) looked at the effect of consonantal con- text, consonant and vowel duration, and syllable stress on the phenomenon of vowel reduction. 0n the basis of his findings he contended: . . . timing is the primary variable in determining the reduction of sounds and that the articulatory imprecision or laxness that may hypothetically be associated with reduced stress can be neglected in this contention (p. 1780). Lindblom explained that it is of no consequence whether a given vowel is produced by rhythm or amount of stress; vowel duration appears to be the critical determinant of reduction. Lindblom further noted that such reduction in vowel quality may be thought of as a coarticulatory effect. In an attempt to study how word level stress is communicated, Medress, Skinner and Anderson (1971) ob- tained acoustic measurements of vowel durations, peak, integral, and average energy values in addition to 25 fundamental frequency contour shapes and peak values within each vowel interval of 40 multisyllabic words. These parameters were then examined to determine (1) how they carry primary stress information, (2) which are the most useful in locating vowels with primary stress and (3) how the various correlates can be used together to find pri- mary stressed vowels. Among their results the authors found that energy integral (which reflects both vowel duration as well as energy level) is the most effective of the various cues. Also, fundamental frequency information appears to be as helpful as energy integral data only when both fundamental frequency contour shape and fundamental frequency peak value are used simultaneously. While the above discussion dealt largely with the effects of duration/energy stress factors on vowel pro- duction, the possibility exists that such changes in stress might also influence consonantal production. That is, consonants, like vowels, appearing in a context which allows more time or less time for production due to stress factors, will in turn result in more coarticulation or less coarticulation. For example, Pickett and Decker (1960) demonstrated that duration of the closure during the production of /p/ in the word "tapic" affected the listener's perceptual decision as to whether they per- ceived "tapic" or "top pick.” In view of this factor it 26 appears necessary to study the variable of stress as related to coarticulation and perceptual processing of speech stimuli. Summary and Statement of PrOblem A review of the literature_reveals that production of continuous Speech cannot beapprOpriately thought of as discrete, independent, commutable phonemic-sized segments in a linear array. Rather, more recent research has defined on-going speech production as a dynamic and synergistic process implying mutually articulated supra- phonemic units (Ohman, 1966; Kozhevnikov and Chistovich, 1965; Henke, 1967). Further, it has become obvious, both theoretically as well as behaviorally, that production of continuous speech involves mutually-articulated or co- articulated motor movements (Stevens and House, 1963: Stevens, House and Paul, 1966; Ohman, 1966; Fromkin, 1966; Ohman, 1967; Lubker, 1967; MacNeilage and DeClerk, 1967; Truby, 1959; Fujimura, 1961a; Fujimura, 1961b, Kozhevnikov and Chistovich, 1965; Lindblom, 1968; Daniloff and M011, 1968: Amerman, Daniloff and M011, 1970; Ali, Gallagher, Goldstein and Daniloff, 1970). While virtually all studies in the area of co- articulation have looked at the process from the stand- point of production, relatively few researchers have studied the effects of coarticulatory activity in terms 27 of speech perception. Those studies which have been undertaken in the area of perception of coarticulatory information have revealed results of a random and in- conclusive nature. Thus there would appear, from a review of the literature, to be a clear need for a well designed study of the affects of coarticulation on the perception of meaningful Speech stimuli. The present study is an attempt to seek basic information concerning the relation- ships between degree of coarticulation and perceptual processing of meaningful speech stimuli. The purpose is to compare the perceptual performance of listeners pre- sented Speech stimuli containing consonant clusters characterized by high and low degrees of coarticulation. Differential perceptibility of stressed and unstressed stimuli will also be considered. More specifically, the following questions were investigated: 1. Will the degree of coarticulation influence the accuracy of the perceptual processing of the Speech stimuli? 2. Will differential perceptability according to degree of coarticulation be reflected in the articulatory features associated with the consonants /p/, /t/ and /k/ as they occur intra- and intermorphemically with the consonant /S/? 28 3. How does the variable of stressing interact with degree of coarticulation in the perception of the speech stimuli? EXPERIMENTAL PROCEDURES This study utilized 50 subjects randomly assigned to one of five signal-to-noise conditions. Each condition provided stimuli which varied according to degree of co- articulation, phonemic content and stressing. Subjects The subjects used in this study were 50 adult (45 female and 5 male) college students between the ages of 18 and 29. All subjects exhibited normal hearing (+15 dB, ANSI 1969) and bilaterally between the frequencies of 250 to 8000 Hz and no subject had any history of auditory impair- ment. Also, in order to insure that the desired signal-to- noise ratios were presented to each ear, no subject had a three frequency (500, 1000 and 2000 Hz) average between ears which differed by more than 5 dB. Design and Stimuli The stimuli for the study consisted of 360 pairs of lexically distinct monosyllabic CVC units which were ‘ constructed to conform to either the form /CVC o CSCVC/, as in /knp d stab/, /kap ¢ Span/, and /kap ¢ skrt/, or the 29 30 form /CVCC2s ¢ CVC/, as in /kups cp tnb/, Amps ¢ pen/, and lkups ¢ kIt/. Thus there were 180 pairs of stimuli with a vocalic null /¢/ between the blend /st/, /sp/ and /Sk/ and the terminal phone of the first word of the word pair (condition C), and 180 pairs with the vocalic null /¢/ between the terminal morpheme /s/ and the word-initial /t/, /p/ or /k/ (condition NC). The terminal phone of the initial unit of each lexical pair was either /t/, /p/ or /k/. These phones were chosen because they are plosive stop consonants which differ by only a single articulatory feature--that of anterior, medial or posterior (Miller and Nicely, 1955; Chomsky and Halle, 1968). Thus any perceptual differences which would occur between the C and NC conditions could be investigated relative to this feature. In addition, each of these con- sonants have a high frequency of occurrence in Spoken English (Moser, 1969: Roberts, 1965). It should be pointed out that each stimulus pair (for example cap stab and caps tab) was used twice, once in a nonstressed manner and once with primary stress on the second word of the pair. That is, those word pairs providing the more highly coarticulated contexts (C condi- tion) for the /st/, /sp/ and /sk/ clusters (n = 180) were produced one-half of the time with equal stress on both words and the other half of the time with primary stress on the second word. Those word pairs providing a lesser 31 degree of coarticulation (NC condition) for the consonant clusters (n = 180) were also produced one-half of the time with equal stress on both words and one-half of the time with primary stress on the second word. 7 The speaker was a phonetically—trained adult male speaking a General American dialect. During the recording of the unstressed stimuli, an attempt was made to produce the monosyllabic phrases with a monotone type stress and intonation pattern and with a normal rate of utterance and within 0 to -3 VU. As the stressed stimuli was recorded an attempt was made to stress the second word of the pair by means of a slight increase in duration and intensity. The stimulus word pairs were recorded in a random fashion on the master tape. During the recording procedure, approximately 5 seconds of silent interval response time was inserted between the stimulus pairs. The master tape was then played to a group of ten graduate students in Audiology and Speech Sciences under optimal listening conditions in order to determine whether the syllabic differences had been produced and recorded as intended. Each stimulus was agreed upon by 80% of the group before it was included on the master list. Presentation Procedures Most experiments which have investigated the per- ception of Speech in noise have employed white noise. The band of noise from 100 to about 10,000 Hz, which is 32 characteristic of white noise, covers all the component frequencies in speech and is quite adequate for most purposes (Miller, 1951). Also, previous investigations of the perception of isolated words in (white) noise suggests that signal-to-noise ratios in the range of approximately -12 to +15 dB would be most apprOpriate (Miller, Heise and Lichten, 1951). Five groups of 10 subjects each were randomly assigned to individual listening sessions under one of five levels of S/N ratio (+14 dB, +10 dB, +4 dB, 0 dB, -6 dB). The signal-to-noise ratios were determined by presenting the stimuli at 50 dB SPL embedded in 36 to 56 dB SPL wideband white noise, depending upon the ratio desired. The presentation was conducted inside a two-room' testing suite with the experimenter in the control room and the listener in a prefabricated double-walled sound treated booth (IAC 1200 series). The ambient noise in the test chamber was less than 20 dB SPL as measured on the A-scale of a Bruel and Kjaer type 2203 sound level meter. The signal-to-noise ratio during the experimental presentation was controlled via a Grason-Stadler Model 162 dual channel speech audiometer and was presented to the listeners binaurally via TDH39-lOz earphones with Mx-41/AR cushions. The stimuli were presented at 7.5 ips by means of a Ampex Model AG 600-2 tape recorder connected to the Grason-Stadler audiometer. 33 Each subject attended a Single 1.5 hour experi- mental session during which time they received, a pure tone audiometric threshold test, standardized instruc- tions (see Appendix B), and both practice and master response sheets (see Appendix C). Also during this time each subject was auditorially presented with six practice stimuli followed by the complete 360 item experimental list. On each rSSponse sheet were listed, in the order of their presentation, the initial and terminal phones of the 360 items of the randomized version of the stimuli. For example, in the case of /keps ¢ heb/ and /hnp ¢ stub/, the appropriate space on the response sheet appeared as cap tab. The listener was instructed to complete the two words in written form for each auditorially presented stimulus pair. The listener was instructed to guess if necessary. In order to control for a possible learning ef- fect as well as possible fatigue factors, the entire master list of 360 stimuli was divided into six lists of 60 stimulus items each. The first 25 subjects then re- ceived the stimuli in lists 1 through 6 (stimuli 1-360) with rest periods of approximately two minutes after each 60 stimuli. The next 25 subjects, however, received the stimuli in a different order with stimuli 181-360 being presented first (lists 4 through 6), followed by 34 stimuli 1-180 (lists 1 through 3). These subjects also received a two-minute rest after each 60-item list. Analysis The data was hand-scored by the experimenter. The number of items correctly responded to was converted into percentage correct scores for each subject. Con— sidering all possible combinations of the variables (two levels of coarticulation, two levels of stressing, and three levels of consonant cluster), twelve such percentage correct scores were obtained for each subject. The data was placed into a Winer multi-factor (four factor) analysis of variance with repeated measures design (Case II), and suitable F-tests were performed (computerized). There were ten subjects per each signal- to-noise condition for a total of fifty subjects. Each subject received all 360 stimuli which represented all conditions of coarticulation, stressing and consonant cluster. RESULTS The results of this study do not support the thesis that degree of coarticulation alone provides a significant influence in the accurate perception of selected consonant clusters. The over-all results, however, demonstrate that the main effects of singal-to-noise and stressing play a Significant role in the subject's ability to perform the experimental task. Also, the interactions of signal-to- noise x consonant cluster, coarticulation x stressing, and signal-to-noise x consonant cluster x stressing all func- tion as important factors in the perception of the stimulus pairs. Table 1 depicts the results of a multifactor (four factor) analysis of variance with repeated measures (Case II, Winer) which was performed on the data. The mean score for each factor under consideration with all other factors is presented in Appendix G. Main Effect of Coarticulation Table 1 reveals that the main effect of coarticula- tion is not significant (p>.05). Thus the over-all means 35 36 Table l.--Summary of an analysis of variance performed on the percentage correct scores at two levels of coarticulation (C and NC) factor (A), five levels of signal-to-noise ratio (-6, 0, +4, +10, +14 dB) factor (B), three levels of consonant cluster (sp, st, sk) factor (C), and two levels of stressing (stressed and unstressed) factor (D). Source SS df MS F A 8.86 l 8.86 -- AB 1988.70 4 497.17 1.13 B 28001.36 4 7000.34 56.20** C 127.98 2 63.99 2.29 BC 668.42 8 83.55 2.99* D 4410.25 1 4410.25 77.81** BD 354.67 4 88.67 1.56 AC 79.97 2 39.99 -- ABC 262.80 8 32.85 -- AD 1669.67 1 1669.67 17.36** ABD 92.62 4 23.15 -- CD 52.39 2 26.19 -- BCD 699.17 8 87.40 3.24* ACD 91.10 2 45.55 1.74 ABCD 348.17 8 43.52 1.67 *p<.01 **p<.0005 37 of 90.5 per cent and 90.2 per cent for the two groups of stimuli C and NC, respectively, when averaged over two conditions of stress, three consonant cluster combinations, and five Signal-to-noise conditions, do not differ signi- ficantly (see Table 2). This suggests that degree of Table 2.-—Summary table of the mean percentage correct main effects of two factors: coarticulation (C and NC) and stressing (stressed and unstressed). Degree of Coarticulation More Less . Coarticulation Coarticulation Joint Stressed 91.5 94.6 93.1 Unstressed 89.4 85.8 87.6 Joint 90.5 90.2 coarticulation alone does not play a significant part in the perception of such stimuli. There is, however, a significant (p<.0005) coarticulation x stressing inter- action which reveals that stressing had a significant differential effect upon the results of the two levels of coarticulation in this task. As illustrated in Table 2, the more highly coarticulated (C) stimuli remained more stable across stressed-unstressed conditions than did the less coarticulated (NC) stimuli. Specifically, the mean percentage correct score is somewhat higher for the NC stimuli under the stressed condition while the score for the C stimuli is greater in the unstressed condition. 38 This would suggest that degree of coarticulation plays an important role in the perception of such consonant cluster segments where both word pairs are equally stressed. This, trend toward more accurate perception of the unstressed stimuli due to degree of coarticulation (see Figure 5) is independent of both signal-to-noise condition and conso— nant cluster. This is supported by the non-significant interactions of coarticulation x signal-to-noise condition, coarticulation x consonant cluster, and coarticulation x signal-to-noise condition x consonant cluster x stressing. Main Effect of Signal-to-Noise Condition The highly significant F-ratio (p<.0005) for signal-to-noise condition (see Table 1 and Figure 6) suggests that as the signal—to-noise ratio becomes less favorable, the task of correctly identifying the stimuli becomes considerably more difficult. Table 1 also re- veals two significant interactions involving the factor of signal-to-noise condition. The significant (p<.05) interaction of signal-to-noise x consonant cluster re- veals that signal—to-noise condition had a significant differential effect upon the results of perceptual accuracy of the three consonant clusters utilized in this task. AS illustrated in Table 3 and Figure 7, the consonant cluster /sk/ was perceived slightly more accurately at S/N ratios of 0 dB and +4 dB and slightly 39 ‘00 a. H QO .. .;._m{}____:::: ,J' O Ip // 8 / / E'WOT- ff 8 6,, .. {'3 C stressed . o____o E 50 T’ NC stressed a D————CI g C unstressed - O----o E 40 “P NC unstressed - 0......5 30 do 20 J- .0 Il- *7 4f fil p '1 ‘6 O *4 +10 +14 s/N RATIO Figure 5.--Mean percentage correct scores for both levels of coarticulation and both levels of stressing at each signal-to-noise condition. 40 100 1» «o ,_ 80 d- . 8 e‘ 2 7" T 8 60 + 8 5 E 50 -- lsp/ '*——-* 2 /st/ se--—--e E 40 «(7- /3k/ 3...”...u.‘ 30 4. 20 i" )0 + .4, ,3 .5. .io J4 S/N RATIO Figure 6.--Mean percentage correct scores for the three levels of consonant cluster at each signal- to-noise condition. 41 Table 3.-—Summary table of the mean percentage correct main effects of the three factors: signal-to-noise ratio (-6, 0, +4, +10, +14 dB) consonant cluster (sp, st, sk) and stressing (stressed and un- stressed). S/N Ratio -6 0 +4 +10 +14 /SP/ Stressed 80.3 91.2 97.2 97.0 99.2 Unstressed 76.3 86.5 91.3 91.9 95.8 Combined 78.3 88.9 94.2 94.5 97.5 90.7 /S'I‘/ Stressed 81.7 90.3 96.7 96.2 97.7 Unstressed 76.3 87.3 89.4 86.3 94.7 Combined 79.0 88.8 93.1 91.3 96.2 89.7 /SK/ Stressed 81.7 93.8 97.8 96.0 98.5 Unstressed 69.2 86.7 94.2 91.5 96.5 Combined 75.5 90.3 96.0 93.8 97.6 90.6 Overall Total 77.6 89.3 94.4 93.2 97.1 PERCENTAGE CORRECT 42 IOO‘" so .. go .. 7o .. cf 60 «r so .. 40 '1' /sp/ stressed :- f 3 /st/ stressed -(}> ;<) 3° "' /sk/ stressed ' I? 4: /sp/ unstressed sO-----O 10 1- /st/ unstressed 'O-----O /sk/ unstressed 'D"-'--~D .0 el- . ’9 0 0'; NO +34 S/N RATIO Figure 7.--Mean percentage correct scores for the three levels of consonant cluster and the two levels of stressing at each signal-to-noise condition. 43 less accurately at the S/N ratio of -6 dB than were either of the consonant clusters /sp/ and /St/. Also, the conso- nant cluster /st/ was perceived somewhat less accurately at S/N ratios of +4, +10 and +14 dB than were either of the consonant clusters /sp/ and /sk/. In addition, the signi- ficant (p<.05) interaction of signal-to-noise x consonant cluster x stressing reveals that these three variables have a significant differential effect upon one another. AS illustrated in Table 3 and Figure 7, the accuracy of perception of the stressed consonant clusters across the five levels of signal-to-noise conditions remained rela- tively consistent with a slightly better score at S/N ratios of +4 and 0 dB for the consonant cluster /Sk/. Perception of the unstressed consonant clusters was some- what less consistent. That is, the unstressed /st/ was perceived less accurately than the other combinations at S/N ratios of +4, +10 and +14 dB and the consonant cluster /sk/ was perceived less accurately at the S/N ratio of -6 dB. Main Effect of Consonant Cluster Table 3 reveals that the main effect of consonant cluster is not significant (p>.05). Thus the over-all means of 90.7 per cent, 89.7 per cent and 90.6 per cent for the three consonant clusters /sp/, /st/ and /sk/, respectively, when averaged over two levels of 44 coarticulation, five levels of signal-to-noise condition and two levels of stressing do not Significantly differ. There are, as stated above, significant interactions of signal-to-noise x consonant cluster and signal-to-noise x consonant cluster x stressing. Table 3 and Figures 6 and 7 illustrate these significant interactions. In addition, Figure 6 reveals that the consonant cluster /sk/ is perceived more accurately than the other two consonant clusters in three of the five signal-to-noise conditions. On the other hand, the /st/ consonant cluster is perceived less accurately than the other two consonant clusters in four of the five signal—to-noise conditions. Main Effect of Stressing Table 3 reveals that the main effect of stressing is significant (p<.0005). Table 2 and Figure 7 reveal that the unstressed stimuli are gradually perceived less accurately through the signal-to-noise ratio of -6 dB. The stressed stimuli on the other hand are gradually per- ceived less accurately through the signal-to-noise ratio of 0 dB, followed by a rather pronounced drop in accuracy at the S/N ratio of -6 dB. As stated above, there are two significant interactions involving the factor of stressing. The significant (p<.0005) interaction of coarticulation x stressing suggests that the presence or lack of stressing makes a significant difference in the role that co- articulation takes in the perceptual task. Specifically, 45 for the stressed word pairs, it was the less-coarticulated stimuli which were perceived most accurately in 4 of the 5 signal-to-noise conditions. An inverse effect was noted for the unstressed stimuli. That is, the percentage correct scores for the more-coarticulated stimuli were greater in 4 of the 5 signal-to-noise conditions. As expected, there was a significant (t<.05) learn- ing effect between the first and second half of the randomized master list as it was presented to the subjects. The presentation procedure outlined above, however, can- celled out the learning effect so that overall percentage correct scores for first and second half performance on the experimental task was non-significant (t>.05). Summary In summary, the results revealed that the single factor of coarticulation is not a significant influence in the perception of both the stressed and unstressed stimuli employed in this study. Degree of coarticulation, however, interacted with the variable of stressing such that those consonant clusters which were embedded in the context of the stressed stimulus pairs were perceived significantly better than those consonant clusters which were placed within the context provided by the unstressed stimulus pairs. Further, the mean percentage correct scores were significantly higher for the NC (less- coarticulated) consonant clusters under the stressed 46 condition while the C (more-coarticulated) consonant clusters were perceived more accurately under the un- stressed condition. Overall performance decreased as the signal-to-noise relationship became less favorable. DISCUSSION Coarticulation and Speech Processing The findings of this study are in basic agreement with previous research in that there is an apparent re- lationship between the articulatory encoding and perceptual processing of language signals. That is, the physiological and acoustical overlapping of phonemic features which has been demonstrated in studies of speech production, has also been shown to play a role in the perception of Speech. The results reported are consistent and predictable from Henke's (1967) model of articulatory encoding. The model suggests that the shorter the time of input (to the articulators) between programmed articulatory attributes, the more overlap occurs in the behavioral approximations of these attributes. Assuming at least a relatively symmetrical relationship between encoding and decoding processes, from Henke's encoding model it would be predicted that the greater the degree of temporally biased articula- tory feature overlap, the more accurately a unit of pro- duction would be perceived in a difficult listening situa- tion. The results of this study offer support for such a 47 48 contention. That is, while the main effect of degree of coarticulation was found to be non-significant, the signi- ficant coarticulation x stressing interaction revealed that when both words were equally stressed, coarticulation became an important variable. Thus, where differences in stressing could not be employed in the decision of /s/- placement, coarticulatory cues appeared to assume greater importance in the perceptual task. In such instances of non-stressed pairs, accuracy in perceptual decoding was apparently aided by the redundant acoustical and/or arti- culatory information carried by adjacent phonemes via for- ward and backward coarticulation. Such information would ease the listener's decoding task by providing simultaneously-occurring perceptual cues across phonemic- sized segments. Thus, there is a decrease inthe necessity for a listener to make a series of extremely rapid per- ceptual decisions based on a string of phones and their associated features. Whether the listener refers to stored acoustical or articulatory features is, in effect, irrele- vant to this contention. The point is that the listener does this processing of several phonemic features simul-4 taneously, in supra-phonemic coarticulatory units. This interpretation is consistent with the postu- lations for encoding by Shriner and Daniloff (1971). They theorized that in a blend, traditionally viewed as occurring intra-morphemically, adjacent consonants facilitate their 49 reSpective productions. If the consonants, however, are separated by one or more vowels, then the consonants would be less apt to behave facilitatively. In the present study, the vocalic null /¢/ is considered to assume the role of a vowel during the coarticulation of successive consonants across word boundaries. As the temporal dis- tance between the consonants decreases, however, there is a greater degree of coarticulatory overlap between these successive consonants. Where both words of the stimulus- pair received equal stressing, the accuracy of perceptual processing appeared to be facilitated by the mutual in- formation carried by successive consonants. Further, this facilitation was maximal when the consonants were adjacent. The results reported here also offer additional support for the findings of Manning, Beasley and Beachy (1972). That is, for the non-stressed stimuli in the present study, the more-coarticulated clusters were per- ceived more accurately than the less-coarticulated clusters across the five signal-to-noise conditions. Further, the two additional consonant clusters, /sp/ and /sk/, also resulted in a similar response by the listeners. Although the rather dramatic difference in percentage correct scores between the more- and less-coarticulated stimuli demonstrated in the original study was not replicated in the present study, degree-of-coarticulation remained an important factor in the perceptual task. There are 50 several possible reasons for the less pronounced difference between C and NC lists in this study. First, and perhaps most important, the task in the present study involved placing only the letter /s/ either at the end of the first word or at the beginning of the second word in the pair. In the initial study, each subject was asked to place three letters (either a /p/ or /k/, an /s/ and a /t/) into the written context of CV VC. That is, the present study presented the subjects with a considerably easier task. This may account for the higher percentage correct scores for all stimuli across the five signal-to-noise conditions as well as account for the less pronounced differences between more— and less-coarticulated stimuli. Secondly, while in the Manning, gt_gl. (1972) study all word pairs received equal stressing on both the first and second word, the present study also employed stimuli with relatively greater stressing on the second word of the stimulus pair. This also made the experimental task somewhat different and may have contributed to the lessen- ing of degree of coarticulation as a single influential factor. Considering the model of articulatory production prOposed by Kozhevnikov and Chistovich (1965), it might be hypothesized that the /sk/ cluster could, because of the less-competitive portions of the articulatory struc- tures involved in its rapid production, possess a greater 51 degree of coarticulation than the other two consonant clusters and thus be perceived more accurately across the five listening conditions. That is, the posterior por- tion of the tongue could be responding to commands direct- ing it toward its target position for the /k/ while simul- taneously the anterior portion of the tongue could already be in the process of approximating the target position for the production of the /s/. While there was not a signi- ficant difference across this variable, there was a trend for the /sk/ to be perceived better than the other conso- nant clusters in three of the five listening conditions (see Figure 7). Further, the /St/ cluster might be thought of as possessing a lesser degree of coarticulation due to the more competitive nature of the articulatory structure involved in the rapid articulation of the two phonemes: the /s/ phoneme must be nearly completed before movement toward the target position for the /t/ can be initiated. Accordingly, this cluster was perceived less accurately in four of the five listening conditions. Thus the re- sults of this study also offer some support for Kozhevni- kov and Chistovich's model of coarticulatory encoding. The implications posed by the results of this study add to the previous findings on perception of co- articulatory information. For example, Ali 32 31. (1971) found that coarticulatory effects can be demonstrated across as many as two vowels such that a decision could 52 be made as to whether the following consonant was either nasal or non-nasal. While the results of the present study also support the occurrence of such coarticulatory infor— mation aiding in perception, the model employed points to coarticulation across consonants cumulating in either a vowel or a vocalic null at the end of the coarticulatory unit. Perhaps, as Ali gt 31. suggest, there is indeed co-, articulation across vowels providing information as to the nature of the following consonant(s). However, as Kozhevnikov and Chistovich suggest, there may be a greater amount of coarticulation across two or more consonants within a cluster. It would appear that there is a con- tinuum of coarticulatory information which is carried by consonants and vowels in terms of their relationships with other consonants and vowels. That is, consonants apparently provide considerable coarticulatory information for other consonants, especially adjacent consonants. Consonants do not, however, carry significant perceptual information as to the nature of following vowels (Shockey and Lehiste, 1971). On the other hand, while vowels appear to carry information as to the nature of the following consonant (Lehiste and Shockey, 1971) such vowels are not as likely to provide information as to following vowels. Thus co- articulatory information is apparently carried by both consonants as well as vowels with the greatest amount of 53 such information being available in the consonant-consonant relationship. The Role of Stressing The variable of stressing plays an obviously signi- ficant role in the listener's ability to correctly perceive the stimuli used in this study. That is, while a greater degree of coarticulation appears to aid the subjects in the perception of the stimuli, differential stressing permits them to make an even more accurate decision. Where differ- ential stressing was not present (each word of the pair equally stressed), degree of coarticulation apparently became an important factor in the perception of the stimuli. Where the two words received an unequal amount cf stress, however, (slightly greater duration and intensity accorded to the second word of the pair) such stressing became an over-riding cue as to the make-up of the stimuli. The listeners may, as Jakobson, Fant and Halle (1951) and Bond (1972) have suggested, use such stressing to signal the division of the utterance into grammatical units. Bond found that her two-syllable stimuli were perceived more accurately if both the first and second syllables received equal stress or if the first syllable only received primary stress. She found that the most errors occurred when the second syllable was stressed. Contrary to the findings of Bond, the stimuli in the present study which received stress on the second word were perceived more accurately 54 than the stimuli receiving equal stress on both words. The most notable difference between the stimuli used in the present study and that employed by Bond is the semantic quality of the stimuli. While Bond used bisyllabic nonsense items, the present study employed two mono- syllabic meaningful words. Perhaps the role of stressing is somewhat differentdepending on whether the speech is of a meaningful or non-meaningful nature.. Perhaps also, the influence of stressing varies according to the time interval separating the syllabic units in question. At any rate, stressing appears to provide a more over-riding cue than does the information carried within the co- articulatory unit. The coarticulatory unit in turn pro- vides a greater amount of perceptual information than does the individual phoneme, especially during the processing of ongoing Speech. Such findings on the role of stressing in the present study also offer support for the contention that stressing occurs at a relatively high level in the encoding process (Kozhevnikov and Chistovich, 1965; Gammon gt_al., 1967). Further, as Gammon st 31. suggested, both stressing and vowel production appear to be largely feedback free. That is, while Kozhevnikov and Chistovich (1965) have postulated that projection of stress is largely feedback free, other researchers (Gammon 2E’31., 1971; McCrosky, 1958; Ringle and Steer, 1963) have found that vowel 55 production is unaffected by both auditory and tactile feed- back. If then, as several studies (Fant, 1962; Stevens and House, 1963: Lindblom, 1963; Stetson, 1951; Medress, Skinner and Anderson, 1971) seem to indicate, perception of stressing is closely correlated with vowel duration/ intensity, then perhaps programming for both stressing and vowel production are also closely related. As the present study suggests, stressing (and perhaps vowel) cues provide more basic perceptual information than does co- articulatory information. Again, where stressing informa- tion is less apparent, coarticulatory cues (consonant- consonant and vowel—consonant relationships) become more important. The findings of the present study may also be interpreted as complementing a motor theory of speech production/perception (Liberman, Shankweiler, and Studdert- Kennedy, 1967). That is, models of coarticulatory pro- ' duction are compatible with the necessary invariance of basic neural commands. The overlap of acoustic and physiological features which characterizes the term co- articulation may be thought of as the effect of forward coarticulation on the invariant neural input signals being fed into the programming process. The analysis-by-synthesfla process as described in the motor theory (the one-to—one correSpondence of the neural commands) could still provide a valid approach for the description of speech processing. 56 Whether analysis, especially during rapid ongoing speech, is performed on a phoneme-to-phoneme neural command basis or whether, as this study suggests, a "chunking" into co- articulatory units occurs, is open to question. In any case, a rather symmetrical relationship between output and input of speech stimuli does indeed appear to exist. Suggestions foria Model of Cdarticulatory,Processing Based on the findings of this study one might con- clude that models of coarticulatory production Should appropriately include a system of decoding as well as en- coding. Drawing from models of articulatory production (Kozhevnikov and Chistovich, 1965: Henke, 1967), models of auditory perceptual processing (Broadbent, 1957; Aaronson, 1967), and the results of this study, a more complete description of coarticulatory processing might be suggested (see Figure 8). In such a model, articulatory production would be initiated by a series of motor commands associated with specific phonetic-sized segments. During speech output the ”scanning" process associated with forward coarticulation and the mechanical-inertial effects of backward coarticulation would result in overlapping physiological target positions and acoustical smearing of the phonemic units. The unit of production over which such coarticulatory overlap would occur could be any one of the following: Cd, CV, CCV, CCCV, or CCCCV. Reception of 57 l J zany l J . FORWARD CDNUTNRA?ON 4 4 CO [edflv'flms CV 4“ m I'" ccv ' 666V ’ CCCOV STRESSINQ EMQHfiIIIMBL 4MZWSHCAL BEDDING LEVEL [W | zenv W 1 .5114. um . ‘1‘ m%esggg ___1 _ CO cg; "'——'" °"""V J cccv , , £531., STRESSINQ MHMEPFNM. PROCBRNNQ Figure 8.—-A synergistic model of co- articulatory processing. 58 such units would take place initially in the lower levels of the central nervous system. The series of phonemic- sized segments would be taken into a short term memory storage system until the processed segments comprised one of the above coarticulatory units. This last concept would coincide somewhat with the investigations which reveal that stop-consonant perception is at least a two part judgment, 3.3., perception of the burst portion of a stop consonant is held in short term memory until the vowel transition can be identified (Winitz, Schaib and Reeds, 1972). Perhaps it is at this point where the perceptual cues associated with stressing are processed. The coarticulatory units are then passed on to a second stage system in serial fashion for further processing (syntactic and semantic evaluation, for example). Per- ceptual strategies would vary according to the temporal demands made upon the system. That is, at slower rates one would most likely employ a more immediate processing of the Speech stimuli and possibly, were the rate of in- coming stimuli slow enough, process the speech in near- phonemic-Sized units. With an increase in rate, however, there would be a greater degree of coarticulatory overlap and the listener would be more likely to adopt a delayed strategy (postponement of a perceptual decision until two or more phoneme-segments have been processed) and deal with the stimuli in supra-phonemic coarticulatory units. 59 Problems_Inherent in the Present Study Caution regarding the above interpretations must be exercised, since in creating the stimuli and the five different listening conditions used in this study, several potential problems arose. First, the listener was required to process two lexical items representing two morphemes in the C condition, whereas he had to process two lexical items with three morphemes in the NC condition. The latter, it may be argued, required more processing on the part of the listener, thus making the task more difficult. On the other hand, it may be argued that the NC condition provided more information because of the presence of three morphemes, which would suggest that the perceptual task involving the NC stimuli would be linguistically easier to perform. A second difficulty arises with the consonant durations of the particular phonemic positions utilized in this study. Lehiste (1964) and Hoard (1966) suggested that a word-initial consonant following a syllabic boundary has a longer duration than a corresponding final consonant. Thus, in this study the /S/ in the C condition may have been longer in duration, permitting it to be more easily perceived than the /s/ in the NC condition. However, Barnwell III (1970) was unable to substantiate the hypo- thesis of Lehiste and Hoard. Further, such a hypothesis does not negate the findings of this study, Since it may 60 be argued that such durational characteristics are due, at least in part, to coarticulatory programming. Lastly, it might be argued that normal listeners diSplay a perceptual preference for consonant clusters which occur in the initial portions of words over similar clusters which occur in the terminal portions of words. Such a preference would of course, bias the subjects in favor of more accurate perception of the consonant clusters as they appeared in the C or more-coarticulated contexts. At the present time, however, this author is not aware of any research which would support such an argument. This question is an empirical one, and will be investigated in the future. Implications for Therapy and Future Researdh The present study is a rather theoretical one. Nevertheless, there are several implications for clinical application and future research which become apparent. One might hypothesize, for example, that perceptual pro- cessing of coarticulatory information would Show a domi- nance effect when presented dichotically. That is, draw- ing from research which relates perception of stressing and vowels and research in dichotic perception of conso- nants and vowels (Shankweiler and Studdert-Kennedy, 1967) it could be postulated that perception of stressing would take place in the right hemisphere while perception of 61 coarticulatory information and consonants occurs in the left hemisphere. One might also consider a similar inves- tigation using meaningful and non-meaningful stimuli. The results of such research might well add to the present models of articulatory processing and provide a basis for extension of these models into the areas of higher level articulatory activity. Normative data on such tasks might be useful in the evaluation of individuals who have suf- fered central nervous system damage with resulting arti- culatory and/or language disorders. Further, in terms of normative data on such a task, one might wish to investi~ gate whether there is a point at which the normal human develops the ability to make use of such coarticulatory units in the processing of ongoing speech. Perhaps the infant is "programmed" from birth to handle such informa- tion or perhaps this ability is acquired as the child develops skill in the encoding of speech. Were the younger child to possess less skill in the processing of coarticulatory information, he would undoubtedly benefit less from articulation therapy which emphasizes produc- tion within the context of coarticulatory units than would the older child who has learned to make use of such information. Research in the area of stuttering has recently begun to consider the process of coarticulation. It has been asserted, for example, (Agnello and Wingate, 1972) 62 that ". . . the transitional features within the syllabic structure of stuttered speech point to a disruption of coarticulated events" (p. 66). Considering such statements in the light of relatively recent findings with delayed auditory feedback and stuttering behavior, (Soderberg, 1968: Soderberg, 1969) one might hypothesize atypical per- ceptual processing of coarticulatory information by stut- terers. Perhaps stuttering individuals would vary from normal speakers across various levels of auditory delay and degree of coarticulation. That is, stutterers may be apt to demonstrate no difference from more to less co— articulation while normal speaking subjects would demon- strate a difference across degree of coarticulation where stressing information is not present. Several other approaches might be taken into account when using stimuli similar to that employed in this study. For example, a gating procedure1 could be employed which would enable one to segment the /s/ phoneme from its phonetic context. Then, rather than using white noise and asking subjects to perform the perceptual task under varying signal-to-noise conditions, the subjects could Simply place the /s/ into the two-word stimulus pair where they thought it would be most accurate. Per— haps the /s/ placement would be more accurate in the 1Gating procedures involve segmentation of the acoustic wave via electronic switching equipment. 63 instances of greater coarticulation. Another interesting approach might take the form of comparing the perception of human and non-human (computer) Speech in a similar perceptual task. In this case, it might be hypothesized that the computer-based speech would be perceived less accurately due to the lack of coarticulatory information from the output. The use of white noise presented several problems in this study. Not only was there the possibility of the white noise interacting with the aspiration of nearby consonants but also the white noise did not provide a range of difficulty in terms of the perceptual task. That is, the task was not extremely difficult (as evidenced by the rather high percentage correct scores) through the signal-to-noise ratio of 0 dB. At the signal-to-noise ratio of -6 dB however, the task suddenly became somewhat more difficult. Perhaps if the "masking” effect took place at a higher level in the perceptual processing system, ideally at the same level where the processing of coarticulatory information takes place, a greater range of difficulty would be demonstrated and the effect of degree of coarticulation might be more pronounced; For example, one might replicate the present experiment using competing message procedures (Speaks, Jerger and Trammell, 1970) to achieve such masking at a higher level in the perceptual system. 64 Lastly, one might consider the temporal gap between consonants which reflects upon the degree of co— articulatory overlap of features. Obviously, there are potential limits to both the amount of permissible arti- culatory overlap, as well as the size of the temporal gap which inhibits accurate perception. One might look at the size of this gap over which coarticulation could be demonstrated. Perhaps coarticulatory effects between consonants could be demonstrated as long as the two conso- nants were within some 150 - 200 msec of one another: the durations at which perceptual breakdowns begin to occur as shown by studies in delayed auditory feedback (Fair- banks, 1955). Possibly this range of "critical duration" for perceptual processing can also be applied to the processing of coarticulatory information. At any rate, coarticulation behavior which occurs during encoding certainly appears to provide valuable information in the process of accurate decoding of linguistic messages. While such information appears to be only one of several sources employed in the perceptual process, based on the results of this research, coarti— culatory processing would appear to warrant further investigation as an important variable. LI ST OF REFERENCES LIST OF REFERENCES Aaronson, D. Temporal factors in perception and short- term memory. Psychological Bulletin, 1967, Jan-June, 130-143. Agnello, J. G., and Wingate, M. E. Some aSpects of stuttered Speech. Paper presented at the 83rd Meeting of the Acoustical Society of America, 1972, April 18-21, Buffalo, N.Y. Ali, L., Gallagher, T., Goldstein, J., and Daniloff, R. Perception of coarticulated nasality. J. acoust. soc. amer., 1971, 49, 539-540. Amerman, J. Cineflurographic investigation of coarticula- tory behavior of the apex and body lingual articulator, 1970, unpublished doctoral disserta- tion, University of Illinois, Champaign. Amerman, J., Daniloff, R., and Moll, K. Lip and jaw coarticulation for the phoneme /12/. Journal of Speech and Hearing Research, 1970, 13, 147-161. Barnwell, III, T. P. Some syllabic junctural effects in English. MIT Quarterly Progress Rept., 1970, 99, 147-159. Bond, 2. Stress patterns and types of perceptual errors. Paper presented to the 83rd Meeting of the Acoustical Society of America, 1972, April 18-21, Buffalo, N.Y. Broadbent, D. E. Perception and Communication. New York: Pergamon Press, 1958. Chomsky, N., and Halle, M. The Sound Pattern of English. New York: Harper and Row, 1968. 65 66 Coker, C. H. Synthesis by rule from articulatory parameters. Reprints of the Conference on Speech Communication and Processing, 1967, Cambridge, Mass. Daniloff, R. G. Normal articulation processes. In Normal Aspects of Speech, Hearing, and Language, Edited by Minifie,F ., Hixon, T., and Williams, F., Englewood Cliffs: Prentice-Hall, In Press: 1972. Daniloff, R. G., and Moll, K. L. Coarticulation of lip rounding. Journal of Speech and Hearing Research, 1968, 2, 707-721. Fairbanks, G. Selective vocal effects of delayed auditory feedback. Journal of Speech and Hearing Dis- orders, 1955, 22, 333-346} Fant, G. Den akustiska fonetikens gunder. Kungl. Tek. HOgSkol., 1962, Taltransmissionslab. Reappt. No. 7. Fromkin, V. Neuro-muscular Specifications of linguistic units. Language and Speech, 1966, a, 170-199. Fujimura, O. Bilabial stop and nasel consonants: a motion picture study and its acoustical implica~ tions. Journal of Speech and Hearing Research, 1961a, 4,“233-247. Fujimura, 0. Effects of vowel context on the articulation of stOp consonants. J. acoust. soc. amer., 1961b, 33, 842. Gammon, S. A., Smith, P. J., Daniloff, R. G., and Kim, C. W. Articulation and stress/juncture production under oral anesthetization and masking. Journal of Speech and Hearing Research, 1971,14, 271-282. Harris, K. 8., Huntington, D. A., and Scholes, G. N. Coarticulation of some disyllabic utterances measured by EMG techniques. J. acoust. soc. amer., 1966, 32, 1219. Henke, W. Preliminaries to speech synthesis based upon an articulatory model. 1967 Conf. Speech Comm. Processing, 1967, Office of Aerospace Research, A1r Force Cambridge Research Laboratories, USAF, Bedford, Mass. 67 Hoard, J. E. Juncture and syllabic structure in English. Houde, R. A. A study of tongue body motion during selected Speech sounds. 1967, Unpublished doctoral dissertation, University of Michigan, Ann Arbor, Michigan. Kozhevnikov, V. A., and Chistovich, L. A. S eech: articulation and perception. EngliSE Transl. from Russian, WaShington, D.C.: U.S. Dept. Commerce, 1965. Lehiste, I. Acoustical characteristics of selected English consonants. Communic. Sci. Lab. Rept. No. 9, 1962, University of Michigan, Ann Arbor, MiEKigan. Lehiste, I. Juncture. Proc. of the Fifth International Congress of Phonetic Sciences, 1964. Lehiste, I., and Shockey, L. Coarticulation effects in the identification of final plosives. Paper presented to the 82nd Meeting of the Acoustical Society of America, 1971, October 19-22, Denver, Colorado. Liberman, A. M., C00per, F. S., Shankweiler, D. P., and Studdert-Kennedy, M. G. Perception of the speech Lindblom, B. Spectrographic study of vowel reduction. J. acoust. soc. amer., 1963, 35, 1773-1781. Lindblom, B. Articulatory activity in vowels. Speech Transmission Lab. Quart. Prog. Status Rep., 1964, 4, 7:10. Lubker, J. F. An electromyographic-cinefluorographic investigation of velar function during normal speech production. Cleft Palate J., 1968, 5, 1-18. MacNeilage, P. F. Electromyographic and acoustical study of the production of certain final clusters. J. acoust. soc. amer., 1963, 35, 461-463. MacNeilage, P. F. The motor control of serial ordering of speech. Psych. Rev., 1970, 11, 182-196. 68 MacNeilage, P. F., and DeClerk, J. L. On the motor control of coarticulation in CVC monosyllables. J. acoust. soc. amer., 1969, 33, 1237. Manning, W. H., Beasley, D. S., and Beachy, T. S. Co- articulation as an adjunct to perception. Paper presented to the 83rd Meeting of the Acoustical Society of America, 1972, April 18—21, Buffalo, N.Y. McCrosky, R. The relative contributions of auditory and tactile cues to certain aSpectS of Speech. Southern Speech J., 1958, 33, 84-90. Medress, M., Skinner, T. E., and Anderson, D. E. Acoustic correlates of word stress. Paper presented to the 82nd Meeting of the Acoustical Society of America, October 19—22, 1971, Denver, Colorado. Miller, G. A. Lan ua e and Communication. New York: McGraw-Hil? Book Company Inc., 1951. Miller, G. A., Heise, G. A., and Lichten, W. The intelli- gibility of Speech as a function of the context of test materials. J. Exp. Psychol., 1951, 41, 329-335. — Miller, G. A., and Nicely, P. E. An analysis of perceptual confusions among some English consonants. 3. acoust. soc. amer., 1955, 31, 338-352. Moll, K. L. Cinefluorographic techniques in speech re— search. Journal of Speech and Hearing Research, 1960, 3, 227-241. Moll, K. L. Ve10pharyngeal closure on vowels. Journal of Speech and Hearing Research, 1962, 3, 30-37. Moll, K. L., and Daniloff, R. G. An investigation of the timing of velar movements during Speech. 3. acoust. soc. amer., 1971, 33, 678-684. Moser, H. M. One-S llable words. Columbus, Ohio: Charles E. Merr111 Publishing Company, 1969. Ohman, S. E. G. Coarticulation in VCV utterances: Spectrographic measurements. J. acoust. soc. amer., 1966, 33, 151-168. Ohman, S. E. G. Perception of segments of VCCV utter- ances. J. acoust. soc. amer., 1966, 33, 979-988. ,M (‘1‘, 69 Pickett, J. M., and Decker, L. R. Time factors in per- ception of a double consonant. Language and Speech, 1960, 3, 11-17. Ringel, R. L., and Steer, M. D. Some effects of tactile and auditory alterations on speech output. Journal of Speech and Hearing Research, 1963, 3, 369-378. Roberts, A. H. §_Statistical Linguistic Analysis of American English. The Hague: Mouton and Company, 1965. Shankweiler, D. P., and Studdert-Kennedy, M. Identifica- tion of consonants and vowels presented to the left and right ears. J. Exper. Psychol., 1967, 13, 51-57. Shockey, I., and Lehiste, I. Perception of coarticulation in English VCV sequences. Paper presented to the 82nd Meeting of the Acoustical Society of America, 1971, October 19-22, Denver, Colorado. Shriner, T., and Daniloff, R. Response to "Comments on the relationship between articulatory deficits and syntax in speech defective children," by J. McNutt and R. Keenan, Journal of Speech and Hearing Research, 1971, 31, 442-443. Shriner, T. A., Prutting, C. A., and Daniloff, R. G. Synertistic aspects of language performance. 1971, Unpublished doctoral dissertation, Univer- sity of Illinois, Champaign. Soderberg, G. Delayed auditory feedback and stuttering. Journal of Speech and Hearing Disorders, 1968, Soderberg, G. Delayed auditory feedback and the Speech of stutterers. Journal of Speech and Hearing Disorders, 1969, 31, 20-29. Stetson, R. H. Motor Phonetics. Amsterdam: North Holland PubliShing Company, 1951. Stevens, K. N., and House, A. G. Perturbations of vowel articulations by consonantal context: an acoustical study. Journal of Speech and Hearing Research, 1963, SJ III:T28i 70 Stevens, K. N., House, A. S., and Paul, A. P. Acoustic description of syllable nuclei: an interpretation in terms of a dynamic model of articulation. J. acoust. soc. amer., 1966, 33, 123-132. Truby, H. M. Acoustical-cineradiographic analysis con- siderations with especial reference to certain consonantal complexes. Acta Radio. Suppl., 1959, 182. Wicklegren, W. A. Context sensitive coding, associative memory, and serial order in (speech) behavior. Psych. Review, 1969, 13, 1—15. Winer, B. StatisticalPrinci 1eS in Experimental Design. New York: McGraw-HiIE Inc., 1962. Winitz, H., Scheib, M. E., and Reeds, J. A. Identification of steps and vowels for the burst portion of /p, t, k/ isolated from conversational Speech. 3. acoust. soc. amer., 1972, 33, 1309-1317. APPENDICES APPENDIX A LIST OF RANDOMIZED WORD PAIRS USED IN THIS STUDY H oomqmmhuun—d H [.5 I 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. SO. hat spike cook stop cap 220.91 caps £3£_ hat scold hats tar cook steam hats top hats ppke hat scan cooks tack cooks tool hats pool cooks till hats cuff cook hats tab cap scab hats team hats tick cooks 32p. cook still cook Skid cooks tack cook scar hat Spoke cap spill hat spgwn cook spill cook stool hat star hats _t_u3 hats 235 cap stick cook still cap stab hat Span hats pawn caps till cooks pike hats 313p hat still cap stub caps tar cap stool cap scar caps camp cook stall cap star caps tab stack 51. 52. 53. S4. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. cooks tub hats l§_i_t_ hats pine cooks tool cooks Egp_ cooks team cook stick cook scold caps tack hat steam hat spin hat steam cooks cuff cap stack cook stack hats till hat Speak hat stab hats £313 cook star hat skill cook stick cooks £3£_ hat scar cooks team cooks tall cook scamp cook scold cook stOp hat scagp hat stop hats 533_ cap stall cap steam caps ggp_ hat skit cap steam cook scuff cooks poke hats team cooks tall hats tack hat stick hat scan hats E]; cap stack hats pin hats tall hats 9.3.2 hats tub 71 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135. 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149. 150. cap score cooks pawn hats pike cook Spoke cooks peak hat Spike hat stool cook stool cooks peak cook Span cooks kid cook score cap skill cap scamp caps pin cap scar cap 212:. cap scuff cap stool hat skill hat scold caps ppke hat skid hat span hat stub cook spin hats peak caps team hats core cooks camp hats pawn caps tall cap Spine cap 22% cap sees cooks kill hat Speck cap skill cap sppke caps E22. hats cold cap score hats peck cooks tick hat scab caps p533 hat stub cap £19.12. cooks can cook scar 151. 152. 153. 154. 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. 170. 171. 172. 173. 174. 175. 176. 177. 178. 179. 180. 181. 182. 183. 184. 185. 186. 187. 188. 189. 190. 191. 192. 193. 194. 195. 196. 197. 198. 199. 200. cook scab hats cold cooks 2122. caps peak cooks kit hats p33 cap skid cap stub hat star cap skid hats car hat stack cooks p32. caps kid cooks £335. cook Skill hats pan cap stop hat Spine hats pip caps pip§_ hat stall hat stool caps tick cooks core hat Spawn cap Spool hat scuff caps cold caps p333_ cook 3223. cooks kill cooks l_<_i_t'; caps cuff cook 222: caps 2233. caps tack cooks ki__d cooks gp3g_ cap Spike hat stick caps tap cap scold hat 2211:: hat 211.1 cook scab caps pawn hats tool cap spawn cooks peg§_ 201. 202. 203. 204. 205. 206. 207. 208. 209. 210. 211. 212. 213. 214. 215. 216. 217. 218. 219. 220. 221. 222. 223. 224. 225. 226. 227. 228. 229. 230. 231. 232. 233. 234. 235. 236. 237. 238. 239. 240. 241. 242. 243. 244. 245. 246. 247. 248. 249. 250. cooks ppol hat stab caps can caps pine hats peck cap Spawn cook scamp cooks p32_ hat scamp hats core caps 123 cooks £22. cap Spoke caps pool cook Skid cook spool hat spill caps p201 hats pike cap scan cook Spill hats cuff caps pawn cap still cook sEck ccok Spike cook stab cap scold cooks pin cook stall caps pan hats kit caps kit cap Skit cooks pawn cook steam cap Spine hats gap caps till cooks ggp_ caps car caps cuff cook Speak cap stab caps team cooks pine cooks core caps core hats tall caps top 251. 252. 253. 254. 255. 256. 257. 258. 259. 260. 261. 262. 263. 264. 265. 266. 267. 268. 269. 270. 271. 272. 273. 274. 275. 276. 277. 278. 279. 280. 281. 282. 283. 284. 285. 286. 287. 288. 289. 290. 291. 292. 293. 294. 295. 296. 297. 298. 299. 300. 72 caps core hats pine cook m1 cooks pill cook scuff cap stop hat Skid cooks pill cook score hat stop cooks till hat Scar caps pike caps tub caps peak cap spill caps kill caps kill cap scamp hats peak cap scuff caps 23£_ hat speck cooks top hats cab cooks pan caps cab cooks pike cap Spin cook scan hats tick cap star cook spawn hat scuff cook span caps tool hat stack cap Speck caps tool cap Speak cook ppke hats poke cap still hats tool cook Spawn cook stub caps k_i_c_l_ cook stab caps 222. cap scab 301. 302. 303. 304. 305. 306. 307. 308. 309. 310. 311. 312. 313. 314. 315. 316. 317. 318. 319. 320. 321. 322. 323. 324. 325. 326. 327. 328. 329. 330. 331. 332. 333. 334. 335. 336. 337. 338. 339. 340. 341. 342. 343. 344. 345. 346. 347. 348. 349. 350. hat Spool cook Spine cooks pool cooks tar caps 291 caps pill cooks cold caps poke cap Spin cook scan caps camp hats can caps tall cap 2.6.22 caps ggp_ caps pike hats 5:3; hat ma hats tack cooks 233;. caps peck cook sack—e cook skit hats camp hat Spin cooks peck cooks ggp_ cap Span hat spill cooks car cook skill cook Spin cap stall cooks cab hats kid hat score hats SEEP. cooks camp cook Speak hat stall cap Span cook skit cap spike caps p1ne caps peck hats p333_ hat Speak hat score hat 2222. hats pill 351. 352. 353. 354. 355. 356. 357. 358. 359. 360. hats kill cook EEEE. cook spike cooks gg£_ hats kill hats till hat skit hat Spoke cook speck cooks tab APPENDIX B INSTRUCTIONS WHICH WERE PRESENTED TO EACH SUBJECT PRIOR TO ADMINISTRATION OF PRACTICE AND MASTER LISTS INSTRUCTIONS TO SUBJECTS Through the earphones you will be hearing a list of word pairs in a background of white noise. White noise sounds like sh........sh. Both the word pairs and the noise will be coming through both earphones at the same time. On the word list you will be hearing through the earphones there is an s-sound occurring either at the end of the first word or at the beginning of the second word. At no time is there more than one (1) s-sound. Notice on your response sheet there are no such s-letters. You are asked to do the following: 1. Using the response sheet you are to take the pencil and write a Single letter-s so that you will then have in written form what you just heard through the earphones; and 2. After placing each letter-s on the response sheet indicate your confidence in your response by making a check (V) in the appropriate Space to the right of the word pair. You will have approximately five (5) seconds between word pairs during which to mark your responses. Respond to each item, guessing if necessary. We'll first try a Six-item practice list. Are there any questions? 73 APPENDIX C RESPONSE SHEET AND SCORING FORM USED BY THE SUBJECT AND EXPERIMENTER PRACTICE RESPONSE SHEET book Spot bat tack book Apot tap care bat tack tap care Guess 74 Uncertain Certain 75 owouuoo aflouuou xmm ICD mmoou oxflm xooo Haws emu czom non com. no: no» moo Hawu xooo xoflu moo woo no: no» uos you no: Hoou xooo Haws xooo csoml no: Hawm moo oxom. non uoo xooo xoou xooo ofix xooo HHHu xooo moo xooo .ov .mm .mm .hm .mm .mm .wm .mm .Nm .Hm .om .mN .mm .hN .mm .mm .vm .mm .NN .HN omd cwouuou ouoo awouuou -GD mmooo xowu non Eoou so: ooo .moo poo poo xoou xooo uwoo no: Haw» xooo HoomT, no: Hooo xooo xoou xooo coo ooc oxom no: mo» use Some xooo no» no: oaoo no: sou moo Hoom, moo moo xooo exam. use .om .mH .ma .hH .ma .mH .va .MH .NH .HH .OH 00...... HNMQ‘IDWFCDO‘ GEMZ 76 2.3 9.6 .3 .II 53 use .2. as.» use .mm III chm, use .Ho v.3 no: .3 III Eoou poo .om moo no: 23 ll xooo moo .mm mEoo no: .3 III oaoo xooo .3 mo» xooo .mb III xowu xooo .hm oaoo xooo .2. III Eoou xooo .mm memo xooo .E I no» xooo .mm :3 xooo .2. ll Hoop xooo .3 53 xooo .2. I. 93.3 use .3 woo no: .2. I vex no: .3 no» xooo .2. .ll sou xooo IS .33 xoom .2. II no» as... .3 32 use .2. II no» 98 .3 you xooo .2. III Haou xooo .me 93 £2 .3 II. as... m3 .2 no» no: .3 III uoo moo .3 SEN use .3. .II Hoop moo .3 SB. use .8 II .8» m3 .3 xoou. xooo .mm iul no» moo .2. .83 m3 .3 II 23 us: .2 $8 xooo .8 ll m3 use .3 cwopuoo cwouuoo moose aflouuoo cwouuoo mmosw IGD ICD 77 cwouuoo cHouuoo Eoou moo xoom non cam xooo no» no: soul. non wax no: oxom. moo oHoo uon Ha“: no: Hoou .mmo «mso mmo pox moo uoo mmo cam moo mEoo moo Hawx moo ouoo xooo wax xooo com}, xooo xoomi, xooo Hoou xooo Hoop no: .mNH .hNH .mNH .mNH .vNH .MNH .NNH .HNH .ONH .mHH .mHH .bHH .mHH .mHH .vaa .MHH .NHH .HHH .OHH .moa .moa .hoa aflouuou cfiouuou oxflm. no: xomm. xooo oxom xooo mxom “on czom. xooo onoo .moo as» non coo no: Haou no: cam, no: xoou moo Hoomlw uon coo no: xowu uon xoou no: Haou xooo Eoou no: oxom, xooo mmso xooo Eoou moo uwx no: noo .moo .moa .moa .«oa .MOH .moa .Hoa .ooa .mm .ma .bm .mm .mm .vm .mm .mm .Hm .om .mm .mm .hm .mm .mm 78 cwouuou cwouuoo IGD mmoso Haou uon xofiu moo cam no: ocflm no: mmu moo com. no: Hawx xooo xofiu xooo cox .moo com! xooo xoou no: you uo: cox moo no» oon as» \moo cox moo comi. no: wax xooo xoom, moo ocwm, xooo oaoo no: noo xooo .NhH .HhH .ona .me .mma .hma .mmH .mma .va .mwa .NmH .Hma .omH .mmH .mmH .hma .mma .mma .va .mma .Nma .Hma aflouuou Gwouuou ICD mooso uoo xooo coo xooo xofiu moo no» no: cam} .moo noo nos xowu xooo xoom no; ouoo moo oaoo no: no» moo oxom moo Haflx moo xoom non Hawx xooo xoom. .moo xoom\. moo ocflm. moo Haou moo :3om. no: mEoo xooo ouoo non .omH .moa .mva .h¢a .mva .mwa .vea .moa .Nva .HvH .ova .ama .mmH .hma .mma .mmH .vma .mma .NMH .Hma .OMH .mNH 79 cwouuou cwouuou Hoom. xooo wax xooo Hoom moo oxoml, moo as» xooo uwx moo ouoo no: mEoo uoc cam xooo mEoo xooo c3om. moo xoom uoc ocflm .moo coo .moo nou uoc Hoom, xooo xoom. xooo c3om «mmo Hoou uoc c3om moo coo xooo Haflu uoc .mHN .mHN .vHN .MHN .NHN .HHN .oam .mom .mom .bom .mom .mom .vom .mom .mom .How .oom .mma .mma .hmH .oma .mmH CHMUHQU CamuHmU ocflm\. uoc oaoo moo mo» moo xowu uoc oxflm moo oaoo xooo wax xooo xoou moo oHoo moo ocflm xooo mmco moo uwx xooo Hawx xooo as» xooo HHHm. .moo oaoo moo mmco uoc Hoom moo c3mm. no: ouoo xooo xowu moo Hoou uoc .me .mma .Nma .HmH .omH .mma .me .hma .mma .mma .va .mmH .NmH .HmH .oma .mha .mba .hha .mba .mha .vna .mha 80 cflopuoo cflouuou [CD mmoso mmu ooc ouoo xooo HHHm xooo wax uoc mdu mmw mmco xooo Hmoml, xooo Homm. xooo ocwm, uoc ouoo moo mou moo ammo ooc ouoo moo ouoo xooo ocwm xooo Eoou moo nou moo xoom xooo mmso moo uoo .moo noo xooo ammo moo .omm .mmm .mmm .hmm .mmN .mmm .vmm .mmm .Nmm .Hmm .omm .mqm .mwm .nvm .wvm .mvm .vvm .mwm .N¢N .me .ovm .mmm cflouuoo cwouuou ICD mmocw coo uoc ocflm moo Boo» xooo czom xooo pox .moo wax moo wax uoc com moo HHou xooo cam xooo Uaoo moo no» xooo oxwml xooo xoom xooo ammo moo czoml, moo mmco uoc Hmmm xooo coo moo oxom own Hoom. moo Hmmm ooc .mmm .nmm .mmm .mmu .vm~ .mmm .mmm .Hmm .omm .mmm .mmm .hNN .mmm .mmm .oum .mmm .mmm .HNN .omm .mam .mam .hHN 81 cuounou cuouuou you xooo Hoom xooo ocfim, xooo Hoom, uoc noo .lmoo ncu moo nou xooo omx .moo nsu xooo c3om. xooo Hoou uoc ammo .moo oxom\ uoc oxom xooo xoom .imoo aoou moo xoom .moo xoou uoc Hoou .moo com xooo mmso uoc c3om. xooo .vom .mom .Nom .Hom .oom .mmm .mmm .hmm .mmm .mmm .cmm .mmm .mmm .Hmm .omm .mmm .mmm .mmm .mmm .mmN .vmm .mmm cuouuoo cwouuou uou moo xowu uoc coo xooo com moo oxom, xooo noo moo com. xooo poo uoc mou xooo xoom uoc uoo Jmoo mwco .moo xoom» uoc maoo moo Hamx moo Haux moo Haum. moo xoom moo ncu moo oxum .mmo uoo uoc Hauu xooo .Nmm .Hmm .omu .mhm .mbm .hhm .mhm .mhm .ehm .mww .NhN .Hbm .onw .mmm .mmw .bom .mmm .mwm .emm .nmm .Nmm .Hmm 82 cuouuoo cuouuoo IGD mmoco ouoo uoc xoom uoc ccom poo xoom moo occm. moo oxflm mMo uflx xooo com moo Haou uoc xoom xooo mEoo xooo WEoo uoc ouoo uoc wax uoc coo xooo Haou .moo cum. xooo Haux xooo uoo xooo 2w “:2 cmm. moo coo xooo .mqm .hvm .mom .mvm .wvm .mvm .Nvm .va .ovm .mmm .mmm .bmm .mmm .mmm .vmm .mmm .Nmm .Hmm .omm .mmm .mmm .hmm cflouuou cwouuou ICD mmocu xoom\. xooo cum uoc mEoo uoc uflx xooo oxom. xooo xoom moo wmco xooo xoou uoc Hoom. uoc moo uoc oxum, moo coo moo coo moo Haou moo coo uoc mEoo mmo coo xooo cum .moo oxom. Jmoo oHoo xooo HHHm, .moo com moo .mmm .mmm .vnm .mmm .NNm .me .omm .mam .mHm .me .mam .mam .vam .mam .NHm .Ham .oam .mom .mom .hom .mom .mom 83 cfiouuou cuouuoo IGD mmeU cou xooo xoom. xooo oxom. uoc uwx uoc Hawu uoc Hawx uoc uoo xooo oxuml xooo uou xooo Haux uoc 3cm poo coo uoc .omm .mmm .mmm .nmm .mmm .mmm .vmm .mmm .Nmm .Hmm .omm .mem APPENDIX D RELATIVE DURATION OF EACH WORD OF THE STIMULUS PAIRS (in msec) I23 \omqmmawnw HH HO on 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. SO. Wbrd-l Wbrd-Z 550 750 450 650 550 900 700 600 450 1050 700 600 300 900 650 500 600 550 450 1000 650 600 650 550 650 550 750 550 650 600 350 900 850 650 550 1000 750 650 700 450 800 600 400 900 500 1000 700 600 400 950 500 850 600 950 1000 950 500 900 500 1000 400 950 700 650 800 550 500 900 500 950 550 950 550 1050 800 700 750 600 700 550 700 600 500 900 500 850 850 700 500 1000 500 1100 800 650 450 1100 450 1000 750 650 no. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. word-1 Wbrd-Z 700 600 700 700 750 750 450 400 800 500 550 500 700 500 400 750 500 550 750 400 500 450 700 500 700 750 400 450 450 500 550 750 500 500 750 500 500 400 700 700 600 800 550 500 700 500 750 750 700 800 84 650 700 800 650 800 800 950 950 700 1000 950 1000 500 850 850 650 850 950 650 1000 900 800 650 1000 650 650 800 1000 900 900 800 600 950 950 700 800 1100 950 500 700 650 550 750 1000 650 850 600 750 800 600 no. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. ' 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135. 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149. 150. Word-1 Word-2 500 700 700 450 750 450 500 400 750 450 800 450 450 500 850 500 450 450 500 550 450 900 450 450 500 400 800 850 700 700 700 850 450 500 500 700 400 450 450 850 750 450 850 750 450 750 400 450 750 450 950 700 600 800 650 950 900 850 550 1000 550 900 950 950 650 950 950 950 950 950 1000 700 1000 1000 900 1050 550 700 550 700 800 600 1200 900 900 650 950 900 950 700 850 900 550 400 900 650 950 850 600 900 no. 151. 152. 153. 154. 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. 170 171. 172. 173. 174. 175. 176. 177. 178. 179. 180. 181. 182. 183. 184. 185. 186. 187. 188. 189. 190. 191. 192. 193. 194. 195. 196. 197. 198. 199. 200. Word-1 Word-2 400 750 700 800 700 800 450 400 400 450 750 450 650 800 700 400 700 500 500 800 850 550 450 800 700 500 450 500 900 850 350 850 750 800 450 700 800 700 700 500 500 750 500 450 500 500 700 700 450 700 950 750 850 500 500 700 850 800 850 950 550 950 750 550 550 900 650 850 950 700 450 950 900 450 650 850 950 950 600 700 950 550 500 600 1150 700 550 600 700 800 850 500 950 1100 950 1100 750 700 950 600 no. 201. 202. 203. 204. 205. 206. 207. 208. 209. 210. 211. 212. 213. 214. 215. 216. 217. 218. 219. 220. 221. 222. 223. 224. 225. 226. 227. 228. 229. 230. 231. 232. 233. 234. 235. 236. 237. 238. 239. 240. 241. 242. 243. 244. 245. 246. 247. 248. 249. 250. 85 Wbrd-l Wbrd-Z 700 700 400 1100 700 700 800 800 700 400 450 1100 350 1100 650 700 450 800 700 750 800 600 600 700 450 850 750 650 400 750 350 950 400 1000 750 650 750 550 450 900 400 1050 700 500 750 650 500 900 300 950 350 1000 400 900 450 1000 750 600 400 900 800 700 750 550 800 550 600 800 700 800 300 1050 450 1150 750 750 750 600 600 750 650 600 700 600 350 750 400 900 750 650 650 650 550 650 700 600 650 650 700 450 no. 251. 252. 253. 254. 255. 256. 257. 258. 259. 260. 261. 262. 263. 264. 265. 266. 267. 268. 269. 270. 271. 272. 273. 274. 275. 276. 277. 278. 279. 280. 281. 282. 283. 284. 285. 286. 287. 288. 289. 290. 291. 292. 293. 294. 295. 296. 297. 298. 299. 300. 600 600 300 650 300 400 350 600 350 500 650 400 750 700 650 500 700 700 450 750 500 700 500 700 750 700 700 700 450 400 700 500 500 500 400 700 450 500 800 450 700 700 450 700 450 400 700 400 650 500 Word-1 Word—2 650 650 900 500 750 800 850 550 850 800 650 800 450 550 500 850 500 600 800 450 850 600 700 500 600 600 600 550 1000 1050 550 850 1100 750 1100 650 800 800 600 900 550 500 850 550 900 850 650 950 550 1050 86 29, WOrd-l Wbrd-Z 29: Wbrd-l word-2 301. 450 850 351. 700 650 302. 400 950 352. 350 1100 303. 700 600 353. 350 850 304. 700 550 354. 650 800 305. 700 700 355. 700 550 306. 750 600 356. 750 600 307. 700 650 357. 500 950 308. 700 500 358. 400 800 309. 450 950 359. 400 800 310. 450 900 360. 700 600 311. 700 650 312. 750 650 313. 700 600 314. 450 1000 315. 700 750 316. 750 550 317. 650 700 318. 450 1000 319. 750 650 320. 700 650 321. 400 750 322. 350 900 323. 350 750 324. 700 500 325. 500 1050 326. 700 400 327. 700 800 328. 500 900 329. 450 1000 330. 750 650 331. 400 1000 332. 400 850 333. 450 900 334. 700 650 335. 700 550 336. 450 900 337. 600 600 338. 700 550 339. 400 900 340. 450 850 341. 450 1100 342. 300 850 343. 450 950 344. 650 600 345. 700 500 346. 700 700 347. 400 950 348. 450 850 349. 450 1050 350. 650 500 APPENDIX E RELATIVE INTENSITY OF EACH WORD OF THE STIMULUS PAIRS no. Word-1 Word-2 E“ Word-1 Word-2 33. Word-1 Word-2 1. 31 33 51. 31 34 101. 33 34 2. 25 31 52. 33 31 102. 29 31 3. 31 33 53. 32 31 103. 32 32 4. 32 32 54. 31 37 104. 31 34 5. 31 33 55. 32 33 105. 33 30 6. 31 32 56. 31 34 106. 32 29 7. 22 31 57. 33 32 107. 33 33 8. 28 32 58. 26 35 108. 32 32 9. 28 32 59. 34 33 109. 30 30 10. 30 28 60. 29 32 110. 30 32 11. 26 32 61. 33 33 111. 32 33 12. 28 33 62. 32 31 112. 31 34 13. 29 33 63. 27 34 113. 32 33 14. 26 33 64. 34 30 114. 33 31 15. 30 32 65. 30 32 115. 34 31 16. 24 29 66. 32 35 116. 32 32 17. 33 33 67. 33 32 117. 33 33 18. 32 33 68. 33 35 118. 32 30 19. 33 31 69. 32 34 119. 33 33 20. 32 30 70. 26 34 120. 33 35 21. 31 33 71. 33 35 121. 31 35 22. 30 34 72. 31 30 122. 34 35 23. 31 33 73. 31 33 123. 33 31 24. 32 33 74. 33 34 124. 31 29 25. 32 32 75. 28 30 125. 30 33 26. 31 32 76. 30 35 126. 32 30 27. 31 32 77. 31 32 127. 33 27 28. 31 31 78. 30 33 128. 33 35 29. 30 33 79. 32 28 129. 29 34 30. 32 34 80. 33 30 130. 29 33 31. 32 34 81. 33 33 131. 30 32 32. 34 35 82. 34 35 132. 31 33 33. 32 32 83. 32 31 133. 33 32 34. 33 33 84. 32 32 134. 32 31 35. 33 33 85. 33 34 135. 31 31 36. 33 33 86. 33 34 136. 28 34 37. 35 31 87. 33 31 137. 33 32 38. 32 31 88. 26 28 138. 32 33 39. 33 34 89. 28 31 139. 32 29 40. 32 32 90. 32 34 140. 32 33 41. 34 33 91. 27 32 141. 31 36 42. 31 34 92. 36 32 142. 31 33 43. 31 33 93. 33 32 143. 33 26 44. 33 33 94. 33 33 144. 29 25 45. 33 36 95. 33 36 145. 30 32 46. 33 33 96. 33 32 146. 30 31 47. 32 30 97. 33 30 147. 31 31 48. 33 34 98. 34 31 148. 30 29 49. 32 34 99. 32 35 149. 29 31 50. 31 33 100. 34 35 150. 27 33 87 88 22. Wbrd-l Word-2 22: Word—1 word-2 22, Word-1 Word-2 151. 26 31 201. 29 36 251. 29 33 152. 31 32 202. 31 31 252. 29 31 153. 30 34 203. 32 30 253. 22 33 154. 31 22 204. 31 34 254. 28 31 155. 28 27 205. 31 24 255. 22 32 156. 33 35 206. 32 34 256. 30 34 157. 33 33 207. 29 32 257. 27 36 158. 32 34 208. 23 33 258. 25 31 159. 29 33 209. 33 30 259. 24 35 160. 32 31 210. 33 37 260. 31 33 161. 30 31 211. 30 32 261. 27 31 162. 31 28 212. 25 31 262. 29 34 163. 27 31 213. 31 30 263. 30 32 164. 31 33 214. 32 33 264. 32 34 165. 29 26 215. 26 34 265. 31 32 166. 27 32 216. 24 32 266. 32 36 167. 31 29 217. 29 32 267. 29 32 168. 31 32 218. 31 37 268. 33 35 169. 33 32 219. 31 32 269. 32 31 170. 33 33 220. 30 30 270. 31 27 171. 34 20 221. 27 29 271. 32 27 172. 34 35 222. 30 31 272. 31 35 173. 33 37 223 33 29 273. 33 32 174. 31 29 224. 31 34 274. 29 34 175. 27 34 225. 24 31 275. 30 30 176. 31 30 226. 24 31 276. 28 32 177. 32 32 227. 27 32 277. 32 32 178. 32 29 228. 30 36 278. 31 31 179. 33 32 229. 26 28 279. 31 29 180. 31 33 230. 26 31 280. 27 31 181. 28 36 231. 32 29 281. 29 29 182. 29 31 232. 32 31 282. 30 32 183. 30 28 233. 32 31 283. 32 30 184. 32 30 234. 33 33 284. 31 33 185. 30 29 235. 30 30 285. 29 30 186. 33 35 236. 27 31 286. 30 36 187. 29 30 237. 31 32 287. 31 32 188. 31 29 238. 32 32 288. 32 31 189. 27 35 239. 32 33 289. 31 32 190. 32 30 240. 28 32 290. 31 31 191. 33 30 241. 29 32 291. 30 33 192. 32 30 242. 30 32 292. 30 29 193. 31 32 243. 24 30 293. 31 32 194. 31 31 244. 30 35 294. 29 32 195. 31 35 245. 32 31 295. 25 33 196. 25 31 246. 24 30 296. 28 35 197. 32 33 247. 28 33 297. 30 33 198. 30 35 248. 31 33 298. 27 27 199. 29 30 249. 30 35 299. 30 33 200. 30 26 250. 30 36 300. 32 32 89 22, Word-1 Word-2 £9: Wbrd-l Wbrd-Z 301. 30 32 351. 30 34 302. 28 30 352. 23 31 303. 28 31 353. 24 29 304. 28 32 354. 28 35 305. 31 35 355. 30 32 306. 31 31 356. 31 32 307. 26 31 357. 31 31 308. 30 31 358. 31 32 309. 32 31 359. 26 26 310. 29 32 360. 27 31 311. 32 31 312. 32 30 313. 31 33 314. 32 32 315. 31 33 316. 30 28 317. 30 34 318. 31 32 319. 33 34 320. 23 27 321. 31 23 322. 22 32 323. 27 29 324. 31 29 325. 32 31 326. 27 25 327. 28 36 328. 31 31 329. 32 33 330. 31 32 331. 26 29 332. 24 30 333. 30 31 334. 28 30 335. 30 31 336. 30 32 337. 31 33 338. 27 29 339. 29 29 340. 31 33 341. 32 31 342. 24 31 343. 33 31 344. 31 32 345. 32 31 346. 32 34 347. 29 32 348. 33 34 349. 31 30 350. 29 30 APPENDIX F TABLED VALUES OF AVERAGE DURATION AND INTENSITY OF EACH WORD OF THE STIMULUS PAIRS Table F.l.--Summary table of average relative intensity for all stimuli. Average Intensity Word No. 1 Word No. 2 Difference Coarticulated 30.20 31.75 1.55 Non-Coarticulated 30.98 31.66 0.68 Stressed 30.47 31.95 1.48 Unstressed 30.67 31.74 1.07 Combined 30.58 31.76 Table F.2.--Summary table of average duration for all stimuli. Average Duration (in msec) Word No. 1 Word No. 2 Difference Coarticulated 452.1 924.7 472.6 Non—Coarticulated 724.0 619.4 104.6 Stressed 584.0 733.5 149.5 Unstressed 595.1 733.7 138.6 Combined 588.8 752.8 90 APPENDIX G MEAN PERCENTAGE CORRECT SCORES FOR EACH VARIABLE ACROSS EACH LEVEL OF SIGNAL-TO-NOISE CONDITION 91 00.00 00.N0 0m.m0 00.N0 00.05 N5.00 00.N0 00.00 NN.00 50.05 HMHOB flfimuw H0.00 00.00 50.00 00.N0 0H.m5 00.50 00.N0 00.00 00.00 50.55 HMHOB 00.50 0m.00 00.H0 00.m0 00.00 mm.00 00.00 0M.H0 mm.m0 mm.00 H0.05 00mmeUmcD 00.n0 50.00 50.50 mm.00 H0.00 H0.N0 0m.50 00.00 50.00 50.00 Nm.H0 Ummmwuum \Mm\ 00.00 00.00 0m.H0 Hm.m0 50.05 0m.00 H0.N0 00.00 0H.00 0m.H0 HMDOB 00.00 00.N0 Nm.N0 0m.00 H0.N0 mm.H5 50.00 00.00 mm.00 00.N0 0m.H0 UmmmmhumcD 00.00 0m.00 50.00 mm.00 00.00 00.00 00.00 50.n0 H0.m0 00.00 0m.H0 00mmwhum \em\ 00.00 No.00 00.N0 0m.H0 5H.05 5H.00 00.00 00.00 0m.00 00.00 HmuOB 5m.00 mm.00 00.00 50.50 00.00 00.m5 00.50 00.00 00.00 H0.00 00.05 00mmwhumc0 50.N0 0m.00 H0.00 mm.00 00.00 0m.05 00.00 00.00 00.00 50.00 0m.H0 memmhum \mm\ 0H+ 0H+ 0+ 0 0| 0H+ 0H+ 0+ 0 0| nocucsoo o o so no co oHuom Z\m HHoHo>o o u H flu. 00: z oouoasouunooo .couuuocoo omwocuouuaocmum mo Ho>oa cooo mmouoo oHcoHHo> cooo How mouoom uoouuoo omoucoouom coozll.a.0 oHcoB IIHIH I 1293 03196 3931 U IIHIWIII