Publications of Dr. Martin Rothenberg:
The Breath-Stream Dynamics of Simple-Released-Plosive Production

6. Change of Glottal Adjustment in Plosive Production

6.1 Some Definitions

A primary goal in this chapter is to discuss some ways in which the dynamic constraints in the laryngeal mechanism may be significant in simple-released-plosive production. Before going on to this task it will be necessary to define some terminology concerning the adjustments, and changes of adjustment, of the glottis. Among the various phonetically distinct adjustments of the glottis which can be held or continued1, there are five adjustments of most interest in the study of plosives. They can be described as follows:

(1) Open. An adjustment in which the glottis is held at or near its most open position. This adjustment is approximately that used in respiration, though from a phonetic point of view, any adjustment can be called open that would not result in an appreciable generation of acoustic energy at normal volume velocities of the breath stream.

The air flow resistance of the glottis during normal respiration has been measured to be approximately 2 W (National Academy of Sciences, 1958). Though there seems to be no report in the literature of an experimental or theoretical determination of the range of air flow resistance associated with the open glottal state in speech, one might suggest a range of approximately 1 to 5 W as an initial guess.

(2) Whispered. A glottal adjustment in which some portion of the glottis is sufficiently constricted that turbulcnce is produced at normal volume velocities of the breath stream, but in which no oscillation of the vocal folds occurs. It is difficult to find any estimates of the range of air flow resistance associated with this glottal state. However, we might guess that during a whisper the air flow resistance of the glottis would be somewhere between 3 and 10 W.

(3) Breathy or breathy-voiced. A glottal adjustment in which a normal transglottal pressure produces both oscillation of the true vocal folds and a degree of turbulence (breathy voicing). This adjustment is often accomplished by holding the true vocal folds sufficiently closed so that they will oscillate when driven by the breath stream but not so close that they will come to a complete closure at any time in the vibratory cycle. The range of air flow resistance in this state has also not been charted. An initial guess might be from 8 to 40 W; however, the upper limit is somewhat arbitrary, since it is difficult to find a clear boundary between 'breathy-voiced' and 'voiced' (see below) that has a phonetic significance. Another problem in the definition of a breathy adjustment arises from the possibility of there being more than one glottal mechanism for the production of breathy voicing.

(4) Voiced. A glottal adjustment that at normal transglottal pressures produces a voiced glottal action, i.e., a vibration of the true vocal folds in which they form a closure or near closure during each period of the oscillation. During a normal voiced glottal action (voicing) the average glottal resistance is high and most of the acoustic energy generated by the air flow pattern will be quasi-periodic. The average glottal resistance during steady-state, non-breathy voicing appears to be in the vicinity of 40 to 100 W (CHIBA and KAJIYAMA, 1941; VAN DEN BERG et al., 1957; LADEFOGED and McKINNEY, 1963). With no transglottal pressure, a voiced glottal adjustment will result in a closed or almost closed glottis.

(5) Tightly-voiced. A glottal adjustment that results in a closure of the true vocal folds at low transglottal pressures, and a vibration of the true vocal folds at high transglottal pressures. The average air flow in the vibratory state may be expected to be small.

Certain other phonetic distinctions have also been made between various vibratory patterns of the glottis; however, such distinctions, when they do occur regularly in a dialect, are not usually associated with plosive phone types. In this work we will therefore consider the term 'voiced' to include any adjustments of the larynx which at normal transglottal pressures produce a vibratory glottal action with a low average air flow and no appreciable turbulence. Similarly, all glottal adjustments will be lumped under 'tightly-voiced' that produce a vibratory glottal action with a low average air flow at high transglottal pressures, but that result in a closed glottis at low or moderate transglottal pressures.

A distinct glottal adjustment of 'closed' has not been defined, though it would be necessary for certain sound types not considered here. Note, however, that a glottis which is closed will be considered to be in a 'voiced' or 'tightly-voiced' adjustment if an appropriate transglottal pressure will initiate oscillations of the true vocal folds. The terminology refers to the adjustment, not the action.

It seems clear that in most if not all languages, a primary feature of plosive production is a change of glottal adjustment, i.e., a glottal gesture or a glottal movement, that is timed relatively consistently with respect to the period of articulatory closure.

A glottal movement (or gesture) will be considered an opening or closing movement if the change in glottal adjustment is to a more open or to a more closed adjustment of the glottis, respectively; the adjustments are ordered: open (the most open adjustment), whispered, breathy, voiced, and tightly-voiced.

Most of the plosive production mechanisms fall into one of the following three categories:
(1) those in which no glottal gesture is employed (as in /aba/ or /sps/ in English);
(2) those in which a single, unidirectional glottal gesture, either opening or closing, is employed (as in /aps/ or /spa/ in English);
(3) those in which a cyclic glottal gesture, either opening-closing or closing-opening, is employed (as the opening-closing movement in /apa/ in English. A closing-opening movement does not occur regularly during English plosive phonemes.).

The remainder of this chapter will be concerned with the dynamic constraints involved in the production of a unidirectional or a cyclic glottal gesture.

6.2 The Dynamic Constraints in Glottal Movements

Though the dynamic constraints in opening and closing movements of the glottis are obviously important in estimating the produceability of certain glottal gestures, no direct measurements could be found of the actual time constants involved. The work of MARTENSSON and SKOGLUND mentioned in Chapter 3, concerning the basic contraction properties of certain laryngeal muscles, is a significant contribution in this area, and using their data, one might go far toward estimating the dynamic constraints in a glottal gesture. Though it is beyond the scope of this study to analyze in detail the laryngeal mechanisms employed in the closing and opening gestures of the glottis, it might be noted that from the results of MARTENSSON and SKOGLUND (Section 3.6) it is not unreasonable to expect that a unidirectional glottal movement could be accomplished in considerably less than 100 msec, with the closing movements the fastest. In the remainder of this section, some rough estimates of the time constants of glottal movements will be obtained from spectrographic data.

To estimate the minimum time necessary for a cyclic movement of the glottis in plosive production, it is desirable to have some technique by which the adjustment of the glottis can be monitored during the period of articulatory closure. (With no articulatory closure, the adjustment of the glottis is indicated in the parameters of the radiated pressure wave.) For the important case of the voiced-open-voiced cyclic glottal movement (as in an intervocalic unvoiced aspirated plosive) it is only necessary to monitor the adjustment of the glottis as it leaves and enters the voiced adjustment. Toward this end the following procedure was employed.

A piece of polyethylene tubing about three cm in length and having a 3/16" inside diameter (approximately) was inserted into the corner of the mouth. This tube served as an outlet for the air in the supraglottal cavity during a bilabial closure. Since the resistance to air flow of the tube was low enough to prevent appreciable pressure buildup, the transglottal pressure was almost equal to the subglottal pressure during the closure; thus a voiced glottal adjustment would normally result in voicing.

The end of the tubing was placed close to a microphone that was simultaneously recording the radiated acoustic wave on the tape loop of a spectrograph (described previously). As a result, a noise-like signal of high amplitude was recorded whenever the lips were closed and the glottis opened, and a noise-like signal with a strong periodic component was recorded when the lips were closed and the glottis was in a voiced adjustment. A spectrogram made from this composite signal (speech wave and tube output) provided an easily interpretable qualitative indication of the state of the glottis during the period of bilabial closure. With a breathy-voiced glottal adjustment, the composite signal provided a reasonably good indication of the relative degree of breathiness. The only ambiguous areas on the spectrogram are intervals of little or no recorded energy, which could be due either to no bilabial closure and an open or whispered glottal adjustment or to a complete glottal closure.

Unfortunately, it is sometimes not easy to determine the time of occurrence of the articulatory closure or release on the resulting spectrograms. If there is voicing at the time of closure or release, a good clue is the change of spectral energy distribution (similar to the change that occurs at the closure or release of a nasal). If there is an open glottis, the closing or opening of the articulators is indicated by the onset or offset, respectively, of the aperiodic component generated by the air flow through the tube which was directed at the microphone. However, these clues can be marginal, especially if the glottis is adjusted for breathy voicing. Another problem occurs with plosive articulations in which the articulatory release is normally partially forced by the build-up of supraglottal pressure. In this case, reducing the supraglottal pressure will delay the opening of the articulators2. In this work explicit mention will be made of the time of closing or release of the articulators only when it is felt that the error due to all of the above factors is less than about 30 msec.

Experiments in which air pressures in the vocal mechanism are artificially forced to deviate from their normal values are interpretable only if the physiological variables of interest can be considered to follow their natural time course in such unnatural situations. This assumption seems plausible for the bypass method of monitoring the glottal adjustment during an articulatory closure if the subject does not react to the perturbation by speaking in a grossly unnatural manner (this is a problem, of course, in most experimentation on speech in which physiological measurements are being taken), and if the unnatural situation is not retained over such long periods of time that unnatural compensatory speech motor patterns are learned.

In order to estimate the duration of the cyclic opening movement of the glottis in intervocalic production of the English phoneme /p/, five productions of the utterance 'I put the apple back in the pig's mouth' were made by the author, using the procedure described above. In the several productions of this phrase various rates of speech were used, from moderate to fast. The same sentence was also recorded once by another male, adult native English speaker. In each utterance a stress pattern was used that had no one predominant stress, although slight variations in the stress pattern existed between utterances.

Spectrograms made from two of the repetitions in this series, as spoken by the author at a moderate and a fast rate of speech, are shown in Figure 6.2.1. Also shown in Figure 6.2.1 are spectrograms of two other repetitions of the test sentence, as produced under normal speaking conditions (with no bypass of the closure), at moderate and fast rates of speech. For the four bilabial plosives in each utterance, an estimated interval of bilabial closure is indicated by a double bar below the spectrograms. For spectrogram D, the double bars for the /p/ and /b/ of 'apple back' terminate in arrowheads to indicate that there was no clear period in which the lips opened for the intermediate /l/. (It is likely that the lips did not open, due to the absence of intraoral pressure. This opinion is based on the interpretation of the spectrogram and on direct observation of the lips during the production of similar repetitions of the utterance.) For the two spectrograms from the test series, an interval during which a cyclic opening movement of the glottis seemed to take place is indicated by a single bar.

The measurements from this series of records can be summarized by the following observations :
(1) The time for a cyclic opening movement of the glottis, voiced-open-voiced, was always between approximately 80 and 150 msec.
(2) Where the interval was about 140 to 150 msec, the offset and onset of the voiced adjustment was relatively sudden; if a period of breathy voice was detectable, it appeared to last for less than 20 msec.
(3) When the interval for the cyclic movement appeared to last for less than about 120 msec, there usually seemed to be a transition period of breathy voice in the opening movement, the closing movement, or both.
(4) At moderate rates of speech, the time taken by a cyclic opening movement of the glottis varied from about 110 msec to 150 msec. (As might be expected, in other utterances recorded with the same technique, cyclic glottal movements of a duration considerably longer than 150 msec were found for some productions of /p/ occurring near places of high stress.)
(5) In fast speech, the cyclic glottal opening movements became briefer. Some of the fastest movements measured are illustrated in the examples from fast speech in Figure 6.2.1. The case of the /p/ of 'apple' in Figure 6.2.1D is especially interesting. An opening gesture was apparently made, since the periodic component of the signal almost fades out near the middle of the closure. However, it appears that there was not time enough for the glottis to reach a fully opened state and thus the opening gesture was in effect voiced-breathy-voiced, instead of voiced-open-voiced.

These observations might be summarized by the following : It seems that the shortest interval in which the glottis can consistently make a cyclic transition of voiced-open-voiced (or voiced-whispered-voiced) is of the order of 80 to 100 msec. In addition, in the most rapid movements some control is apparently lost over the duration and timing of the transitions from and to the voiced state. As a result, in those production mechanisms that require a relatively precise control over the opening or closing movement, (usually with respect to the articulatory gesture) the minimum time for the cyclic transition can be considered somewhat higher, say 110 to 130 msec. It does not seem unreasonable to refer to a cyclic opening movement of the glottis of voiced-open-voiced or voiced-whispered-voiced, as ballistic, in the sense discussed in Section 3.7, if the duration of the movement is less than about 150 msec at a normal rate of speech, or 125 msec at a fast rate of speech.

Also of significance are the limits in the accuracy with which the speech mechanism can synchronize the gesture at the glottis with the articulatory closure or the release. It is of interest in this regard that in most of the productions of the phoneme /p/ at moderate rates of speech, the glottal opening gesture seemed to be almost coincident with the articulatory closing gesture. This property of English plosives has been noted before in experiments employing a measurement of the light transmitted through the glottis from a source placed in the mouth, when the light is picked up by a sensor on the throat (see, for example, LISKER and ABRAMSON, 1964).

Thus in the closure of a postvocalic /p/ in English, and apparently in similar plosives of many other languages, it seems important that the articulatory closing gesture come during the breathy-voiced transition period of the glottal opening movement. When the glottal opening movement is fast enough that the transitional period of breathy voicing lasts less than about 20 msec, it seems likely that a difference of only 10 or 15 msec in the timing of the articulatory closing gesture relative to the glottal opening gesture could have a significantly different acoustical (and auditory) result. For example, a small delay of the cyclic glottal opening movement with respect to the articulatory closure was found in the production of the English sequence /bp/. To determine the timing of the glottal opening movement in /bp/, a series of nine repetitions of the English sentence 'I prefer the cab polished' and one repetition of the sentence 'The bear cub put its paw down' were produced by the author at a normal rate of speech with a tube bypassing the bilabial closure, as described above. The repetitions of the first sentence were made with various stress patterns. The second sentence was produced with no predominant stress.

The spectrograms resulting from these utterances showed that the glottal opening gesture followed the articulatory closing gesture by between 10 and 90 msec. This range is illustrated by the three spectrograms of Figure 6.2.2, which are taken from three utterances of the series in which the sequence /bp/ was spoken with a rather neutral stress. (The exact timing in Figure 6.2.2A is difficult to determine, since the occurrence of the articulatory closing gesture and the glottal opening gesture were almost simultaneous.)

It may be mentioned in passing that in Figure 6.2.2A the /p/ of 'prefer' was apparently produced with a glottal opening gesture that was somewhat delayed with respect to the articulatory closure; it followed the closure by about 20 to 40 msec. This seemed to occur in a number of the repetitions of the sentence. This timing pattern may represent a dialectal idiosyncracy of the speaker.

Apparently the speech mechanism has the capability of accurately synchronizing the opening of the glottis and the closing of the articula tors (as in a normal production of the English /p/), at least at moderate rates of speech. In fact, the necessity of an accurately timed articulatory gesture might be one reason for the faster preclosure articulatory movements noted in some languages for postvocalic unvoiced, aspirated phone types (for example LISKER, 1957, and LEHISTE and PETERSON, 1961). The motion of the articulators would need be fast in order to match the rapidly opening glottis and it would seem that the timing of the critical portion of an articulatory gesture (during which Ra becomes large with respect to Rg) would be easier to control in a rapid motion. It should also be noted that a 'firmer' closing gesture might be necessary at certain places of articulation in order to interrupt the increasing air flow and to withstand the rapidly rising supraglottal pressure.

6.3 Some Additional Definitions

It seems worthwhile at this point to explain some rather unconventional terminology that will be used to illustrate certain aspects of the role of the glottis in simple-released-plosive production. One of the more significant changes of glottal adjustment in a prevocalic simple-released-plosive is a closing gesture of the glottis from an open or whispered adjustment to either a voiced or tightly-voiced adjustment, with the new adjustment continued into the succeeding phonetic unit. Such a gesture may occur as a unidirectional movement if an open or whispered glottal adjustment is present in the preceding phonetic unit (or after a breath pause), or as a cyclic movement if the preceding phonetic unit is voiced (i.e., intervocalically). In the following discussion it will be convenient to refer to the time at which a closing gesture occurs with respect to the articulatory release (i.e., the first air flow through the articulatory constriction). We will use the following terminology to refer to the time from the articulatory release until the glottal adjustment first enters the range of the target adjustment (voiced to tightly-voiced).

Label   Time from the articulatory release

'slightly aspirated'

  over 40 msec
0 to 40 msec
less than zero

Whenever these terms are used in this special sense, they will be enclosed in quotation marks. When used without the marks, the terms aspirated and unaspirated will refer to the more traditional categories; for example, as defined by PETERSON and SHOUP (1966). The category of 'unaspirated' will be considered to include that case in which a voiced glottal adjustment is present at the closing gesture and is held through the closure until after the release.

6.4 An Illustration of the Significance of the Cyclic Opening Movement of the Glottis in Plosive Production

It is argued above that the time at which a postrelease closing transition of the glottis occurs can be greatly influenced by the transition in glottal adjustment, if any, which occurred at or near the time of articulatory closure. The general principle proposed here is that if there was an opening transition at or near the time of articulatory closure, the shortest time in which the postrelease closing transition can be effected depends strongly on the time of the first transition and the basic dynamic physical and physiological limitations of the laryngeal structures. Though this observation is in some sense evident, its explanatory power in linguistic description does not seem to have been fully utilized. In the present section we will attempt to show the relevance of this principle to certain aspects of the phonology of Korean.

The phonetic structure of Korean contains three contrasting unvoiced plosive phone types for each place of articulation. One of the three, in which the glottis attains a voiced or tightly-voiced state before the release (i.e., is 'unaspirated') is discussed in Sections 6.5 and 6.6. The remaining two categories are illustrated in Figure 6.4. The spectrograms in this figure are also made from the recordings described in Section 3.4.

There is an obvious difference of glottal action in the two unvoiced plosives in Figure 6.4. Figure 6.4A shows a very small period of aspiration and indicates that the glottis was in the process of closing at the time of release. A breathy voicing starts within 15 msec from the release and develops into a full harmonic-rich voicing in about 30 to 40 msec. In the terminology introduced in Section 6.3, this phone type can be referred to as 'slightly aspirated'. The spectrograms of Figure 6.4B show a period of aspiration in which there is no sign of voicing for about 50 msec after the time of the release. The rate at which the higher harmonics become stronger after the onset of voicing indicates that the glottis was closing at about the same rate as in the contrasting phone type above, except for the difference in timing. This last phone type is 'aspirated', in the terminology introduced in Section 6.3.

However, according to MARTIN (1951) the above-mentioned three-way contrast among Korean unvoiced plosives does not exist intervocalically. The contrast can only be found in those environments in which the glottis is open before the articulatory closure, and, therefore, in which all three unvoiced plosives would be produced with a unidirectional closing gesture of the glottis. Thus it appears that the previously hypothesized capability of the speech mechanism to control the timing of a unidirectional glottal movement is used in Korean to distinguish between two phonemes. However, what then are the allophones of these phonemes which occur intervocalically?

According to MARTIN the 'slightly aspirated' phone type does not occur intervocalically, but, instead, is replaced by either a simple voiced, unaspirated plosive [b], or by the 'unaspirated' phone type in which the glottis is closed at the time of release. Thus in both cases there is no aspiration, so that all aspirated plosives that occur intervocalically can be identified unambiguously as signifying the same phonemes as the 'aspirated' allophones which occur after an unvoiced phonetic unit. In other words, it seems that a distinction in the length of aspiration is not sufficiently controllable in the intervocalic position to support a phonemic distinction.

Of interest are the observations of HAN and WEITZMAN (1965) that the intervocalic allophones of the 'aspirated' plosives of Korean are less aspirated intervocalically. In their measurements, the interval of aspiration was less than half as long intervocalically as in the initial position, and averaged about 45 msec in length (including most of the breathy period). Durations of this magnitude are close to what has been found for the intervocalic unvoiced, aspirated plosives of a number of languages, and probably indicate that the intervocalic aspirated plosives of Korean are generated with the same 'ballistic' cyclic glottal movement discussed above for English.

6.5 The Timing of the Cyclic Gesture with Respect to the Period of Articulatory Closure

In the previous section, it is argued that in the production of certain plosives in the intervocalic position, an important part of the action of the speech mechanism is a cyclic ballistic opening movement of the glottis, closed-open-closed. In this section a differentiation will be made between some of the possible patterns of timing between a cyclic glottal opening movement and the period of articulatory closure. Toward this end, the diagrammatic presentation of Figure 6.5.1, will be used. In each diagram in Figure 6.5.1, a single bar of fixed length is used to represent the time required for a ballistic opening movement of the glottis (as voiced-open-voiced, or voiced-whispered-voiced). The length of this bar can be thought of as representing the approximate time required for a ballistic glottal opening movement at normal rates of speech, or about 125 msec. The period of articulatory closure is represented by a double bar. Except in one case, this period is assumed to be about 100 msec long. Though the duration of 100 msec is a somewhat arbitrary choice, it is used because closure durations of as short as 100 msec seem to be attainable at normal rates of speech for most places of articulation. For one pattern of timing, number 3, it will be shown that the duration of the articulatory closure must be somewhat longer than the cyclic opening movement of the glottis, and therefore a closure duration of 140 msec is represented in Diagram 3. The patterns of timing shown in Figure 6.5.1 are discussed, in turn, below.

(1) In the preceding section, the significance of a ballistic cyclic opening movement of the glottis has been illustrated for the production of an intervocalic unvoiced, aspirated plosive, such as might be produced for the sequence /apa/ in English. The timing of the glottal movement in this case is illustrated by Diagram 1 in Figure 6.5.1. The opening gesture of the glottis and the articulatory closure occur approximately simultaneously. The articulatory release occurs before the glottis has returned to a voiced adjustment and therefore there is a short interval of aspiration (high air flow).

(2) The timing pattern shown in Diagram 2 of Figure 6.5.1 has also been illustrated previously. The spectrograms of Figure 6.2.2 show some examples of an intervocalic /bp/ sequence for a dialect of English. In these productions, the cyclic glottal opening movement begins some time after the closing of the articulators, and ends after the articulatory release. This relationship is indicated in Diagram 2 of Figure 6.5.1. It should be noted that a maximally short period of articulatory closure would result in a rather long period of aspiration. In order to conserve the breath stream, it might be expected that this type of plosive would often be produced with a longer than average period of articulatory closure.

It also might be noted that a minimum time required for the production of this type of plosive is the interval from the beginning of the articulatory closure to the termination of the aspiration. This period is long compared to the time required for the intervocalic version of a simple unvoiced, aspirated plosive (type 1 above) or for a voiced plosive (in which no glottal movement is necessary). This longer minimum duration could explain the fact that this type of production is rarely if ever found with a separate phonemic identity; it occurs most commonly as the geminated version of a sequence of two plosive phonemes.

(3) The timing pattern illustrated in Diagram 3 of Figure 6.5.1 also occurs most commonly as the geminated version of a sequence of two plosive phonemes, as for example in the sequence /apba/ in English. In this type of production, the articulatory closing movement and the glottal opening movement occur approximately simultaneously, as in the simple unvoiced, aspirated plosive (type I above). The articulatory closure, however, is held until after the glottis returns to a voiced adjustment, as in some types of voiced plosive. An example from English of this type of production is shown in Figure 6.5.2. Spectrograms are shown for two productions, by the author, of the sentence 'Please buzz the bellhop boy'. A rather even stress pattern was used. In production A, a short length of tubing was used to bypass the bilabial closures, as described above. The periods indicated for the articulatory closures and the cyclic glottal opening movements of production A were interpreted from the spectrogram. Of interest here is the /pb/ sequence in 'bellhop boy' .

The type 3 production, like the type 2 production, takes longer to produce than a simple, unvoiced, aspirated plosive (type 1) or a voiced plosive. The important point is that the period of articulatory closure must be larger than the time required for a ballistic cyclic opening movement of the glottis. (For example, in each production of Figure 6.5.2, compare the length of the period of closure for the /pb/ sequence with the length of the closure for the /b/ of 'bell'.) For this reason we might expect that in most language dialects this type of production would not represent a single phoneme but would more commonly indicate a sequence of two plosives, as noted above. There are, however, cases in which a type 3 plosive is used regularly to indicate a single phoneme, and one such case will be illustrated at this point.

The three contrasting unvoiced plosive categories of Korean are mentioned in Section 6.4 above, and two of the categories, the 'slightly aspirated' and 'aspirated' plosives, respectively, are discussed at that point. We are concerned here with the third or 'unaspirated' category3. In plosives of this category, a strong, harmonic-rich voicing is usually found to begin almost immediately after the articulatory release (LISKER and ABRAMSON, 1964; HAN and WEITZMAN, 1965). It may thus be assumed that the glottis attains a voiced or tightly-voiced adjustment some time before the release. (Naturally a prevocalic or intervocalic position is being assumed in this description.)

It is important to the understanding of an 'unaspirated', unvoiced plosive to know when during the period of articulatory closure there normally is a voiced or tightly-voiced glottal adjustment, i.e., to know to what extent the plosive is unvoiced because of an open glottal adjustment. If the plosive occurs in an intervocalic position, it is important to know whether or not a cyclic opening movement of the glottis is employed. If there is an opening movement of the glottis, the timing of the glottal and articulatory movements should be determined.

To investigate these questions for the unvoiced 'unaspirated' plosives of Korean, measurements were taken from an adult male speaker of standard Korean. Each of three Korean sentences was produced twice, once with and once without the bilabial closure bypassed using the procedure described above. The sentences contained various bilabial plosives, mostly of the unvoiced 'unaspirated' category. The subject was told to produce each sentence at a normal volume and rate of speech, as monitored by the experimenter.

The resulting records indicate that in the intervocalic position, the Korean bilabial unvoiced 'unaspirated' plosives are produced with a cyclic opening movement of the glottis. The timing of this movement with respect to the articulatory movements is similar to the timing noted for the intervocalic /pb/ sequence in English; the glottal opening movement tends to occur synchronously with the articulatory closing movement, and the glottal closing movement occurs some time before the articulatory release. Referring to Figure 6.5.1, the mechanism used seems to fall clearly into category 3.

Some typical spectrographic records from the above sequence are shown in Figure 6.5.3. Spectrograms are shown for the sentence /aBagaosInda/ ('Daddy is coming') as produced with and without the bilabial closure bypassed. Referring to the record made with the closure bypassed, it can be seen that the unvoiced, 'unaspirated' /B/ in /aBaga/ ('daddy' as subject) is clearly produced with a ballistic cyclic opening movement of the glottis. The opening movement of the glottis appears to begin just after the articulatory closure is attained, and to terminate in a voiced or tightly-voiced adjustment, somewhat before the release. (It should be kept in mind that the reduction of supraglottal pressure in the non-normal production may have delayed the release somewhat.)

There is a rather long period of articulatory closure for this plosive in both the normal and non-normal productions of Figure 6.5.3. It seems clear that this long period of articulatory closure results from the hypothesized requirement that the period of articulatory closure must encompass the entire cyclic opening movement of the glottis.

(4) and (5) Diagrams 4 and 5 in Figure 6.5.1 illustrate two possible patterns of timing that would result in an interval of 'preaspiration' (high air flow before the articulatory closure is attained). Aside from the period of preaspiration, the resulting productions would be acoustically similar, respectively, to the type 1 production (as the /p/ of English) and the type 3 production (as the /pb/ of English). In fact, one might expect plosives of type 4 or type 5 to occur occasionally as allophonic variants of plosives of type 1 and type 3, respectively. An example of a type 4 or a type 5 plosive occurring as a canonical allophone can be found in Icelandic. (See also WOLTER, 1965.)

There are apparently dialects of Icelandic containing preaspirated voiceless plosives which, when occurring intervocalically, contain little or no aspiration after the release (for example, HAUGEN, 1958, p. 72). However, the exact nature of the plosive used is difficult to determine from the literature. For example, it would be of some interest to determine whether these plosives are consistently of type 4 or of type 5, as by an acoustic analysis of the release, or by some measurement of the glottal adjustment during the period of closure (as by using a bypass of the articulatory closure or transillumination of the glottis).

6.6 The Unidirectional Glottal Movement

A unidirectional glottal movement in the production of a plosive may be either opening (as in /aps/ or /abs/ in English) or closing (as in /#pa/, /spa/ or /sba/ in English). In this work, only a closing movement (open or whispered to voiced or tightly-voiced) will be considered. In effect, we will then be considering only plosives that are preceded by a breath pause or unvoiced phonetic unit and followed by a voiced phonetic unit. In this section, the term 'prevocalic' will only be used to refer to such an environment, i.e., in this section a prevocalic plosive cannot directly follow a voiced phonetic unit.

A principle dynamic constraint in a unidirectional closing movement of the glottis in plosive production appears to be the constraint on the ability of the speech mechanism to locate the glottal movement accurately in time with respect to the articulatory release. Thus in any one language dialect, there is only a limited number of contrasts that can be maintained in the duration of an interval of aspiration in a prevocalic unvoiced plosive. This maximum number appears to be two, not including 'unaspirated' as a separate class (LISKER and ABRAMSON, 1964). Two contrasting degrees of aspiration can be found in a number of language dialects, such as in the example of Korean already cited.

The above comments refer to contrasts between glottal movements in which a voiced adjustment is attained after the release. There is little information available on the ability of the vocal mechanism to time a glottal movement accurately before the release. One reason for the lack of information concerning this subject is that it is uncommon to find contrasts between plosives which differ primarily in the time before the release in which the glottis attained a voiced adjustment.

For example, in the /b/ of 'buzz' in the English utterance for Figure 6.5.2A, as produced by the author, the voiced adjustment was attained almost simultaneously with the release (and therefore the plosive would have been entirely unvoiced)4. However, this production could not contrast in the author's dialect with one in which the voiced adjustment was attained early in the period of closure (and which therefore might have been voiced). Similar records for other prevocalic /b/ variants in other sentences produced by the author showed a considerable variation in the length of the interval before the release in which a voiced glottal adjustment was present. The strong tendency, though, was for the voiced adjustment to be attained late in the period of articulatory closure.

There are language dialects, however, in which the timing of a prerelease closing glottal movement might be an important factor in the production mechanism of a plosive. One such example appears to be the unvoiced, 'unaspirated' plosives of Korean. Figure 6.5.3, presented above, contains spectrographic records of intervocalic variants of these plosives. Other spectrographic records from the same series indicate that the prevocalic versions of these unvoiced, 'unaspirated' plosives are produced with a unidirectional closing movement of the glottis (open to voiced or tightly-voiced) shortly before the release, as in the case of the prevocalic unvoiced /b/ of English. However, one difference to be expected between the Korean plosives and the prevocalic unvoiced /b/ of English is that the Korean sounds, being always or almost always completely unvoiced, would be more consistently produced with a glottal closing movement that occurred close to the time of release (i.e., late in the period of articulatory closure). The spectrographic records taken with the bilabial closure bypassed seemed to support this result. However, in view of the fact that there were only a small number of samples from one speaker of Korean, it seems that this hypothesis concerning the timing of the glottal gesture should be verified more extensively, perhaps using experimental techniques more suitable than the techniques used in the present study. If, indeed, the production of the unvoiced, 'unaspirated' plosives of Korean required a glottal closing movement as late as possible in the period of articulatory closure, then a study of the actual timing of these movements might contribute toward an understanding of how close the speech mechanism can synchronize the articulatory and glottal movements.

Chapter 5 Air Flow Resistance at the Articulatory Constriction
Last Chapter
  Chapter 7 A Voiced or Breathy-Voiced Glottal
Next Chapter


Table of Contents

E-mail Dr. Rothenberg