Journal of Phonetics (1983) 11, 177-189
The fortis / Ienis question: evidence from Zapotec and Jawofi Jeri J. Jaeger Department of Linguistics, University of Pennsylvania, Philadelphia, Pennsylvania 19104, U.S.A. R eceived 1st December 1982
Abstract:
The terms "fortis/lenis" and "force of articulation" are controversial, as linguists disagree about both their definition and their validity. Languages such as Yatee Zapotec (spoken in Oaxaca, Mexico) and Jawon (spoken in Arnhem Land, Australia), in which a difference in voice onset time is not a reliable cue in the main consonantal contrast, have been called "fortis/lenis" languages. An analysis of some of the phonetic properties involved in these two languages reveal that the " fortis " consonants are long, voiceless , and have high intensity noise , while the "Ienis" are short, fluctuate in both voicing and in closure width , and have lower intensity noise . However, it is argued that the independently controlled variables are the timing of articulator gestures and glottal width, so that the terms "fortis/lenis" , implying intentional variation of "force of articulation " , are not a phoneticaly accurate characterization of this contrast.
1. Introduction (a) The fortisjlenis dis tine tion The use of the terms "fortis/lenis" to describe phonological contrasts has long been controversial. Although these terms are used frequently , and numerous attempts have been made to define and substantiate their phonetic content, linguistists continue to disagree on the interpretation or validity of such evidence, and not all linguists are convinced that the phonetic distinction indicated by these terms exists at all . The purpose of this paper is to present some data from two so-called fortis/lenis languages, and to interpret these data in terms of the ongoing fortis/lenis controversy. The terms "fortis/lenis" are used in two distinct ways. On the one hand they are used to characterize a basic phonological contrast in consonant systems which cannot be explained in terms of a voicing distinction. On the other hand , they are used as a secondary feature , intended to add additional phonetic information to a contrast which is primarily characterized as voiced/voiceless ; this is the case when phonemes are described as being " voiceless fortis " or "voiced Ienis" . In this paper I will be concerned primarily with the former usage , although the discussion will also have implications for the latter. In either case, however , when the terms "fortis/lenis" are used, the claim is being made that one member of a contrasting pair of phonemes is produced with greater "force of articulation" than the other. A wide range of phonetic phenomena have been claimed by various researchers to be involved in producing this force, including the following (see Jakobson, Fant & Halle, 1951 ; 0095-4470/ 83/020177 + 13 $03 .00/0
© 1983 Academic Press Inc. (London) Ltd .
178
J. Jaeger
Fischer-J¢rgensen, 1969; Malecot , 1970; Catford, 1977, pp. 199-208; Jakobson & Waugh, 1979, 135-9). (!)Pulmonic factors. Fortis consonants are claimed to be produced with greater pulmonic force, which is said to cause greater sub-glottal air pressure, greater oral air pressure, greater intensity of bursts or frication , and aspiration (see also Ladefoged, 1971, pp. 24, 95-97). (2) Articulation factors. Fortis consonants are said to be produced with greater force or pressure of the actual articulators , with sudden rapid release of closures, causing sharper onsets of following vowels. (3) Timing factors. Fortis consonants are analyzed as having longer durations than Ienis. ( 4) Glottal factors. Usually consonants termed fortis also tend to be voiceless , while Ienis consonants tend to either be voiced or fluctuate between voiced and voiceless productions. The main questions to be asked are, first , whether or not this list represents an accurate reflection of the phonetic factors involved in producing the distinction in question, and second, whether these phonetic properties are indeed a result of some unified and independently controlled phonetic dimension for which "force of articulation" is the correct explanation. Both of these questions need to be answered in the affirmative in order for the terms "fortis/lenis" to have phonetic substance (see Lisker, 1963 for a similar argument).
(b) Phonological systems There are basically three types of phonological systems which have been called "fortis/lenis". The first is languages like English in which fortisness was ascribed to the voicelss stops because of their aspiration. However , Lisker & Abramson (1964) have shown that aspiration is a function of voice onset time (hereafter VOT), so that no extra "force of articulation" need be posited to account fully for this contrast. Further, for languages whose main contrasts can be explained in terms of VOT, it has been shown that the fact that voiceless consonants tend to have greater oral air pressue , airflow, and intensity of bursts can be fully explained by the position of the glottis and supra-glottal articulators, and there there is no extra pulmonic force involved (Ladefoged, 1971, p. 96; LOfqvist, 1975).In fact there is actually a decrease in subglottal air pressure during aspirated stops (Ladefoged, 1967, pp. 41-44), and studies of articulator pressure have failed to reveal any consistent differentiation between voiced and voiceless consonants in these languages (Catford, 1977, p. 201). Therefore the use of terms such as 'fortis/lenis' to describe such systems are neither as explanatory as the VOT analysis, nor are they necessary as secondary explanations for various phonetic phenomena which are associated with the voicing distinction. A second type of system usually called "fortis/lenis" is exemplified by Korean and Javanese. These languages have contrasts which cannot be explained in terms of VOT; however , these contrasts have recently been demonstrated to depend on a difference in the tension of the laryngeal musculature, and are thus better characterized as having "tense voice" and "lax voice" (see Hardcastle , 1973 on Korean , and Catford, 1977 , p . 203 on Javanese; my own informal analysis of Javanese has supported the data reported by Catford). Further, these languages have few of the other phonetic characteristics of the fortis/lenis distinction; for example, there are no consistent durational differences between tense-voice and lax-voice stops in Javanese, and the lax-voice stops have greater air flow as they are usually breathy. Again the terms "fortis/lenis " are inappropriate for explaining these contrasts. In Lisker & Abramson's (1964) paper on VOT, they note that "an examination of the phonetic literature generally fails to turn up any language which is said to possess stop categories that differ only in force of articulation" (p . 386). However, since the publication
The fortisjlenis question
179
of that paper, a few languages have come to the attention of linguists which have consonantal contrasts which clearly do not rely on VOT differences , as they have much free variation in voicing ; further , they do not have any sort of tense-voice/lax-voice distinction. Two such languages are Yatee Zapotec (spoken in the state of Oaxaca, Mexico), and Jawoii , an Australian Aboriginal language spoken in Arnhem Land. These languages (as well as other Zapotec and Australian languages with similar phonetic properties) have been termed "fortis/lenis" partly because VOT is not a reliable cue in their systems, and partly because they exhibit some of the phonetic properties noted above which are said to be caused by differences in force of articulation. Because such languages are unusual, a detailed analysis of their phonetic systems would be extremely useful to linguists interested in the fortis/lenis question . In the remainder of this paper, then, I first present a sketch of the phonetic properties of the main consonantal contrast in these two languages, and then discuss these properties in terms of whether their interrelationships suggest some unitary dimension for which the terms "fortis/lenis" are an appropriate explanation.
2. Phonetic analysis of Yatee Zapotec and Jawoii (A) Phonological systems The consonant phonemes which participate in the fortis/lenis distinction in Yatee Zapotec (hereafter "Zapotec ") are presented in Table I. This contrast occurs word-initially, medially, and finally. Table II presents the consonant phonemes of Jaw011 which are involved in the fortis/lenis contrast. As in most Aboriginal languages which have more than one series of stops, this contrast only occurs word-medially, intervocalically and between a sonorant and a vowel. Word-initially only Ienis consonants occur, and word-finally, or medially between Table I Consonant phonemes of Yatee Zapotec (Oaxaca, Mexico) which participate in the fortis/lenis distrinction
Stop Fricative/ Affricate Nasal Lateral
Labial
Dental
p b
d
Fortis Lenis Fortis Lenis Fortis Lenis Fortis Lenis
Retroflex
Palatoalveolar
t
Velar k g
(;
z N n L
~
I
Table II Consonant phonemes of Jawoii (Arnhem Land , Australia) which participate in the fortis/lenis distinction
Stops
Fortis Lenis
Labial
Alveolar
p b
t d
Retroflex
Palatal
Velar
c
k g
+
180
J. Jaeger
two sonorants or a vowel and a sonorant, only fortis consonants occur. There are no fricatives or affricates in Jawoii , and the sonorants do not participate in the contrast. 1 . (B) The data The Zapotec data to be analyzed here consist of 514 words spoken in isolation, 62 words spoken in frame sentences , and 22 natural Zapotec sentences, all recorded on high-quality tape in the field by myself and Robert D. Van Valin, Jr. (The words to be analyzed all have eve, evev, or eveve syllable structures.) The data were collected from four adult speakers, two men and two women. The Jawoii data consist of 85 utterances taken from a field elicitation tape with a single adult female speaker, collected by Francesca Merlan . Spectrograms and Mingograf tracings of the waveform were made of all the utterances , from which duration, voicing, and closure type data were taken . Amplitude tracings were made of the Zapotec utterances using an intensity meter with a 500Hz high-pass filter and a sampling rate of 10 ms. (C) Voicing and closure type In order to test the hypothesis that voicing is not a reliable cue for this distinction , all obstruent tokens were analyzed as to whether they were fully voiced (i.e . voice bars occurring throughout the entire closure), partly voiced (several voice bars occurring at either the beginning or end of the closure , with lack of voice bars throughout the remainder), or voiceless (no voice bars throughout closure or burst). Since it was further noted that there was much fluctuation in the closure types of the lenis obstruents, these consonants were analyzed in terms of whether they were produced as stops, fricatives, affricates, or glides. As the contrast in the laterals and nasals in Zapotec was found to be almost entirely one of duration, with no voicing or manner fluctuation, they were not included in this analysis. In Zapotec, the fortis obstruents are always voiceless , and show no variation in closure type ; they are unaspirated initially and medially , but are often heavily aspirated wordfinally. However, the lenis obstruents show much free variation in voicing, as shown in Table III ; they are partly or fully voiced in 60% of the word-initial tokens, 93% of the medial tokens , and 69% of the word-final tokens . It is interesting to note that there is more voicing in general in tokens spoken in context than in those spiken in isolation, with 86% of tokens in context containing some voicing, as compared to 68% of the isolation tokens ; this is obviously due to the fact that word-initial consonants spoken in context usually become intervocalic, in which environment voicing is favored. Likewise, there is variation in the completeness of the closure of the lenis stops and affricate, in that they actually contain a full stop closure in only 43 % of the tokens , with 57% being produced as fricatives . The completeness of closure varies according to the position in the word, with stop closures occurring in 58% of word-initial tokens, 30% of medial tokens, and 6% of final tokens .
1 The phonological analysis of Jawon assumed here is actually open to question. Most Australian languages have only one series of stops which flu ctuate in voicing in a manner similar to those of1awon, i.e. generally voiced medially and voiceless initially and finally (if they occur finally; see Dixo n 1980, p. 137) . For those languages with a medial contrast, a number of different phonemicizations have been proposed. The one assumed here is the one which is supported by the phonetic facts presented below: the initial and medial consonants which are shorter and fluctuate in voicing and manner are group ed into the Ienis series, while the finals and the medials which are lon ger and usually voiceless are grouped into the fortis.
181
The fortis/ Ienis question
Table III Voicing and closure type variations of Ienis obstruents in Yatee Zapotec, words spoken in context vs isolation. Numbers represent the percentage of tokens with each voicing/closure type Stop closure Partly Voiced voiced Vless Isolation /bdg/ Context
Isolation
fz'?f Context Isolation
Iii Context
Initial Medial Final Initial Medial Final Initial Medial Final Initial Medial Final Initial Medial Final Initial Medial Final
Fricative closure Partly voiced Vless Voiced
0 0 0 44 16 0
0 6 0 22 15 33
57 4 0 15 6 0
35 88 0 15 60 33 37 73 0 25 43
7 0 67 0 3 0 34 27 80 45 28.5
16 7 0 29 0
8 33 0 71 37
0 7 0 0 II
4 53 0 0 26
48 0 0 0 26
1 2 33 4 0 33 29 0 20 30 28.5 24 0 100 0 0
Tota!N
N 104 51 12 27 62 3 38 22 15 20 7 0 25 15 2 7 19 0 = 429
Table IV Voicing and closure type variations of fortis and Ienis stops in Jawoii. Numbers represent percentage of tokens with each voicing/closure type Stop closure Partly Voiced voiced Fortis
Lenis
Medial (+contrast) Medial (-contrast) Final Initial Medial (+contrast)
% of segments with stop closure % fully or partly voiced: Initial Medial Final
Vless
Fricative/glide closure Partly Voiced voiced Vless
N
0
17
83
53
0 0 26
56 0 14
44 100 40
18 9 35
49
9
Fortis
Lenis
100
59
27 0
60 99
17
3
37
4
0
0 79 Tota!N= 194
182
J. Jaeger (a) Duration of fortis vs Ienis consonants in Yatee Zapotec (ms) Table V tokens spoken in isolation and in context combined. Number in parentheses is number measured
/p/ /t/ /k/ /s/
M /c/ /N/ /L/
Initial
Medial
Final
146 ( 4) 165 (55) 149 (3) 202(13) 165 (16) 207 (3) 108 (17) 121 (21)
176(13) 150(14) 144 (31) 183 (8) 174 (3) 161 (18) 125 (27) 138 (30)
216 (2) 209 (13) 198 (8) 215 (24) 236 (8) 24 7 ( 4) 172 (17) 221 (13)
Initial
/b/ /d/ /g/
/z/ 1'?1 /j/ /n/ /1/
63 ( 41) 60 (4) 59 (23) 123(32) 125 (26) 132 (7) 58 (44) 70 (36)
Medial 58 53 74 98 94 98 65 61
Final
(39) (31) (37) (19) (1 O) (17) (31) (64)
64 (9) 68 (6) 144(2) 136 (10) 197 (13) 220 (2) 79 (15) 119(12) Total N = 921
Duration of Ienis as percentage of duration of fortis Initial Medial Final Stops 40% 39% 44% Fricatives 68% 54% 74% Affricate 64% 61% 89% Nasal 54% 52% 46% Lateral 58% 44% 54% Further, there are more stop closures in the tokens spoken in context (53%) as compared to those spoken in isolation (37%). The data on voicing and closure type for Jawoii. are shown in Table IV. It is quite clear that voicing cannot be considered a necessary cue for the distinction in this language, as there is fluctuation in voicing for both the fortis and lenis consonants. However, there is less fluctuation in environments where the two series contrast than in environments where only one series occurs. Although initially only 60% of the lenis consonants are voiced, medially 99% are voiced. In medial positions where there is a contrast, 83% of the fortis consonants are voiceless; but where there is no contrast, only 44% are completely voiceless. However, worcj-finally, where there is no contrast, 100% of the fortis consonants are voiceless. The sa111e sort of closure fluctuation in the lenis stops occurs in Jawoii. as was seen in Zapotec, except that the Jawoii. stops can be either fricated or produced as glides; this happens more often medially (59% stop closures) than initially (80% stop closures). (D) Duration
The durations of the consonants in question were measured from the spectrograms, and these measurements were checked against the waveform tracings. Only tokens flanked either by vowels or by silence (if possible) were measured; consonants in clusters were not included. Medial consonants were measured from the offset of the preceding vowel to the onset of the following vowel, so that the burst and any aspiration following a stop was included as part of the duration of the consonant. Initial consonants were measured from the onset of noise and/or voicing , and final consonants to the end of the frication and/or voicing, or to the stop release, including aspiration. (Obviously there were a number of initial stop tokens spoken in isolation which could not be measured for the Zapotec data.) Table V presents duration data from Zapotec. (Although the durations for tokens spoken in context were generally shorter than those spoken in isolation , the two sets of measurements have been combined for this comparison.) There are consistent and usually large differences between the duration of the fortis and lenis consonants, with the lenis being on
The fortis/len is question
183
Table VI Duration of fortis vs Ienis consonants in Jawon (ms). Number in parenthesis is number measured
#_ Fortis
/p/ /t/
It/ /c/ Lenis
/k/ /b/ /d/
51 (8)
Nl It/ /g/
61 (2) 58 ( 4) 52 (9)
{~on} _ Son
v v
Son V
138(10) 109 (6) 139 (9) 18 7 (8) 153(10) 58 (8) 36 (8) 40 (8) 60 (3) 60 (10)
165 (4)
140 (2) 122 (1) 71 (3)
144 48 39 23 36 36
102 (8)
(11) (7) (1) (3) (5) (14)
_#
N
16 7 226(4) 16 8 185 (2) 31 23 9 13 12 33 TotalN= 168
Duration of Ienis as percentage of duration of fo rtis: V _ V= 35%; So n _ V = 23 %
the average about 50% of the duration of the fortis word-medially , 55% word-initially , and 60% word-finally . The differences are greatest for the stops , and least for the affricates, but in nearly all cases they are well above the just-noticable differences discussed by Lehiste (1970 , pp. 10-13). In the only case in which the duration difference is probably not enough to be noticable , that of word -final affricates , the distinction relies primarily on other cues, as the fortis phoneme is realized as an aspirated affricate, whereas the Ienis is usually a voiceless fricative . Duration data for Jawoii is given in Table VI. Again there is a large and regular difference in duration between fortis and Ienis consonants, with the Ienis being from 23 % to 35% of the duration of the fortis in the environments in which they contrast. In the environments in which they do not contrast, the Ienis consonants are slightly longer , with a mean of 53 ms initially vs 44 ms medially. The fortis consonants are somewhat shorter in non-contrasting environments , with an average duration of 102 ms in non-contrasting environments vs 147 ms in contrasting environments; however they are the longest when word-final (mean 212 ms) which is a position of non-contrast. It is clear that the duration differences are intrinsic to the articulation of these consonants, as they occur both in contrasting and noncontrasting environments.
(E) Amplitude The peak amplitude of the bursts of stops and affricates and the peak and average amplitudes of the frication for fricatives were measured from the Mingograf tracings of the output of the intensity meter ; amplitudes were measured in dB above the base line of the tracing. These measurements were made only for the Zapotec data, and only for words spoken in isolation ; the quality of the Jawoii recording was not high enough to allow accuracy of measurement. The results of these measurements are shown in Table VII . The data show that the fortis consonants have consistently higher peak and average amplitudes than those of the Ienis consonants. (F) Summary of data At this point it will be useful to sum up the main properties of the fortis /lenis distinction
184
J. Jaeger
Table VII Peak amplitude of burst of stops and affricates, and peak and average amplitude of frication of fricatives in Yatee Zapotec; measured in dB above base line (500Hz filter, I 0 ms sampling rate) . Total N = 424
fpf /t/ /k/
fC!
/sf M
Peak amplitude Mean SD
N
2.05 3.17 4.66 11 .60 10.13 10.47
13 23 32 34 28 14
0.87 1.37 1.99 2.80 2.85 1.63
Average amplitude Mean SD N
/b/ /d/
fgf
5.70 6.42
1.85 0.92
28 14
m fzf I?!
Peak amplitude Mean SD N
Average amplitude Mean SD N
1.35 1.24 1.91 6.37 5.00 7 .29
3.37 4.77
1.00 0.70 1.39 1.84 2.00 2.24
37 9 13 44 38 30
1.43 37 1.72 30
which have been found in the two languages under discussion. In each, the prototypical fortis obstruent is long and voiceless, with no variation in closure type, and higher amplitude noise. The prototypicallenis consonant is short, usually voiced but often voiceless, has much variation in closure type, and lower amplitude noise.
3. Discussion (A) Fortisjlenis vs VOT languages Before looking specifically at the questions raised earlier about the appropriateness of the fortis/lenis terminology, it would be instructive to make a brief comparison of the data from these and a few other fortis/lenis languages to data from VOT languages along the phonetic dimensions discussed above. The purpose of this comparison is to affirm that the phonetic cues signalling the distinction in these languages are in fact different from those found in VOT languages. First, in languages in which the main consonantal contrast is termed one of voicing, the actual phonetic factors involved in the contrast may not always be the actual presence or absence of glottal pulses; in some cases it may be, eg. simultaneous vs. delayed onset of voicing in relation to release of closure (i.e. unaspirated vs. aspirated) . Likewise, such a contrast may depend on different cues in different positions in the word (e.g. in English the distinction in the stops is aspirated/unaspirated initially, but often voiced/voiceless medially) . However, there is nearly always a distinction which can be explained in terms of a VOT difference in every position. It is clear from the data presented above that VOT differences are not consistent in these fortis/lenis languages, in that, e.g., both fortis and Ienis stops can be voiceless unaspirated, and in Jawoii fortis stops can sometimes be partly voiced. This fluctuation in voicing indicates that the primary cues for the distinction must be sought elsewhere. Second, the sort of free variation in closll{e type found in Jawoii and Zapotec is rare in VOT languages. However, it is quite common in those Australian Aboriginal languages which have two series of stops, where the fortis or voiceless series is longer in duration than the Ienis or voiced; the fluctuation is always in the shorter phonemes. This alternation occurs in Wal)kumara (McDonald & Wurm, 1979), Thargari (Klokeid , 1969), Burduna (P. Austin, personal communication), Margany and Gunya (Breen, 1981), and Nakkara (B. Eather, personal communication), among others. Further, it is interesting to note that Lehiste (1966, p. 38) reports that in Estonian, for intervolcalic consonants of the shortest duration,
The fortisflenis question
185
"allophones of /t/ and /k/ were often voiced or fricated". There seems to be some connection between fluctuation in closure completeness (and probably voicing) and closure duration. Third, in languages termed fortis/lenis, there is always reported to be a consistent difference in the duration of the fortis vs Ienis consonants. For example, according to the data on Breton presented in Lehiste (1970, p. 28, based on Falc'hun, 1951), initial Ienis stops are on the average 66% of the duration of the fortis (approximately 60ms vs 90ms), and medial Ienis stops have 50% of the duration of the fortis (50ms vs lOOms). However, a review of the literature reveals no such reliable duration difference in languages with VOT contrasts. For example, while it has been shown that English speakers can use longer duration as a cue to voiceless consonants (Lisker, 1957; Cole & Cooper, 1975), the data from Umeda (1977) show that this duration difference is not actually consistently produced in natural speech. As another example, in KiMvita Swahili the voiceless unaspirated stops are consistently longer than the voiced imploded, and the voiceless fricatives longer than the voiced, but the differences are not as great as those found in Jawon or Zapotec (see Table VIII). In French (Table VIII) the voiceless stops are on the average 23 ms longer than the voiced, which would probably not be a noticable difference in duration. In Danish (Table VIII) the closures of the unaspirated stops are an average of 15 ms longer than those of the aspirated; Fischer-J¢rgensen (1969) claims that the unaspirated are fortis and the aspirated Ienis, although the duration data do not support this claim, as the 15 ms difference in duration would clearly not be a noticable difference. Further, it is not clear that a difference in the duration of the closure, as opposed to a difference in the total length of the consonant, serves as a phonetic cue. In sum, then, while voiceless obstruents tend to be longer than their voiced counterparts in VOT languages, the differences are neither as great nor as reliable as in fortis/lenis languages. Fourth, all other things being equal, voiceless consonants have higher intensity noise than voiced, due to the greater oral air pressure build-up and greater air flow allowed by the open glottis. Thus VOT languages as well as fortis/lenis languages generally have consistent Table VIII
Duration data from three VOT languages
KiMvita Swahili 1 Duration vless asp. stop vless unasp. stop vd (imploded) stop vless fricative vd fricative French 2 Duration of closure /b, d, gf jp, t, k/
200 134 83 148 113
SD
N
10 17 13 13 23
8 19 14 16 16
Closure period for voiced stops is 80% duration of voiceless unaspirated.
139 162
Danish 2 Closure
Aspiration
Total
122 137
73 18
!9S ISS
vless aspirated vless unaspirated 1
2
Voiced stops are 62% duration of voiceless unaspirated. Voiced fricatives are 76% duration of voiceless fricatives.
Data gathered by Jaeger; all tokens V _ V. From Fisher-J~rgensen (1969).
Closure period for aspirated is 89% of unaspirated
186
f . Jaeger
differences along this dimension. However , it is not clear that the greater intensity of fortis consonants is related to some factor other than voicelessness, as fortis consonants are nearly always voiceless. In fact in Danish, where the voiceless unaspirated stops are claimed to be fortis and the aspirated lenis, there is no difference in the intensity of the bursts (FischerJ¢rgensen 1969, p . 72). This point will be returned to below.
· (B) Is ''force of articulation" an independently controlled variable? As noted earlier, in order for the terms "fortis/lenis" or "force of articulation" to be considered phonetically accurate terms, it must be shown that they correspond to some unitary and independently controlled phonetic parameter. Looking first at the duration data discussed above , it is quite clear that there are large differences in duration between the two consonant series in Zapotec and Jawoii. However , it is not at all obvious that the duration differences imply differences in "force of articulation". The timing of the gestures of the muscles involved in speech production is an independently and voluntarily controlled variable; the only way in which it takes any extra force to sustain a gesture for a longer period of time is in the rather circular sense that the overall energy expended by a muscle will be more if it fires longer . However, it is much more phonetically explanatory to term these variations a matter of duration which is due to intentionally varied timing of articulator movement, than to ascribe them to some vaguely defined notion of "force of articulation" . Second, the differences in intensity of bursts and fricative noise in the Zapotec obstruents are clear and consistent , and indicate differences in oral air pressure and air flow (although no data were collected on these factors). However, again it is not clear that "force of articulation" is the correct explanation. In the first place , other things being equal, a segment with a complete stop closure held for a longer time will by definition have higher oral air pressure than one with a fricative closure held for a shorter time. Further, the fluctuation in voicing of the lenis obstruents suggests that the vocal cords are probably being consistently adducted, although sometimes not enough to allow voicing to occur. Similarly , the vocal cords are most likely being fully abducted when the fortis consonants are produced (although in the case of Jawoii there may be a timing lag in this abduction in some instances). This would of course contribute to the higher oral air pressure of the fortis consonants. Although it is possible that, as Malckot (19 70) argues (see below) , a synesthetic feeling of greater oral air pressure is the main cue correlated with the use of the term "fortis", the phonetic factors which are actually being manipulated to produce this feeling are duration , glottal width, and completeness of closure. Again I would argue that the distinction can be explained more accurately in these terms , rather than referring to the notion of "force" , which as far as oral air pressure is concerned is not an independently controlled variable. Third, there is the question of greater muscular tension. Although there has been a great deal of research done on the correlation between either oral or respiratory muscular tension and fortis vs lenis articulations, the results have been contradictory at best (Fischer-J ¢rgensen 1969, 1972; Malecot, 1970; Catford, 1977, p . 201). In the case of Zapotec and Jawoii, the one possible indication of a variation in muscular tension is the fact that the lenis stop phonemes vary between stop articulations and fricative or glide realizations , which are traditionally considered to be more "lax" articulations. However, Hardcastle ( 1976, p. 130-131) has shown that stop closures are achieved by a single ballistic movement of the articulator(s), while fricative closures require a precise closure width and therefore more fine-tuned muscular control ; it may be , then , that the fricatives demand more work from the muscles than do the stops. If the fricative closures were more "lax" in some sense, one
The fortisflenis question
187
would predict that they would be more frequent in contextual speech than in isolated words, but in fact exactly the opposite was found in the Zapotec data . Actually there is some question as to whether Hardcastle's conclusions are applicable to the fricatives in question, as his analysis was based on English fricatives, particularly sibilants, which tend to be quite long and have long relatively steady-state tongue positions. An alternative explanation for the variation in closure width would be that the short duration of the Ienis consonants does not allow time for a complete stop closure , so that the fricative pronunciations are the result of falling short of the target tongue height. In this case the explanation again depends on timing factors (not 'laxness') ; and one could hypothesize that the prevalence of fricative articulations in the tokens spoken in isolation is due to speakers trying to maximize the contrast in careful speech by exploiting the extremes of possible pronunciation. Obviously these questions require further research. However, whatever the explanation, these variations in closure type cannot be taken to support the argument that the fortis consonants are articulated with greater muscular tension. In sum, the data presented here do not support the notion of "force of articulation" being a unitary or independently controlled phonetic dimension. 4. Conclusions In his 1970 overview of research on the fortis/lenis question , Malecot (1970, p . 1588) argues that the use of the terms "fortis/lenis" by naive speakers (and presumably by linguists as ~m
.
is a case of synesthesia, in that it has little or nothing to do with articulatory energy but is rather a mistaken proprioceptive impression based on intrabuccal air pressure resulting from the airvalving action of the glottis, the occlusion or constriction of the buccal passage, and the velopharyngeal sphincter, and perhaps also involving closure duration. Similarly, Fischer-J¢rgensen (1972) found that naive speakers' judgments of "greater general effort " in the production of Danish stops correlated more closely with their judgments of "more forceful airstream" than "stronger organic pressure" (which , incidently, caused them to consider the aspirated stops to involve greater general effort than the unaspirated). I have argued above that in the fortis/lenis contrasts examined here, the actual phonetic factors being manipulated to produce the contrast are primarily duration, glottal width, and possibly closure width, and that such factors as oral air pressure and air flow are the results of these manipulations; Male cot's claims in fact support my argument. However, I would further argue that superimposing the notion of "force of articulation" on the contrast by the use of the terms "fortis/lenis" does not add to the explanation of the phonetic factors involved, but in fact obscures them by the vagueness , and probably incorrectness, of the notion of "force". I therefore conclude that the synesthetic feeling of extra force is not justification enough for using these terms when in fact other phonetic explanations are available. However, it is clear that the explanation for the contrast discussed here is not yet complete, and that more research is needed on the interrelationships of the phonetic factors involved . A clue to the direction which this research should take comes from the fact that in both the Australian languages and the Zapotec languages with this contrast, the fortis phonemes are analysed as coming historically from earlier consonant clusters, and the Ienis from earlier single segments (Swadesh, 1947; Dixon, 1980, pp. 215-216). Because of this, it is likely that the duration differences is the most significant factor, and that the
188
J. Jaeger
other factors are related to duration in some direct way. I have shown elsewhere that if a language has an asymmetry in its system of long obstruents, it will have only voiceless (as opposed to only voiced) long obstruents (Jaeger, 1978); this asymmetry has clear aerodynamic causes, since the longer an occlusion is held, the more likely the air pressure will equalize above and below the glottis, making voicing impossible. Therefore the fact that the long obstruents developed into the voiceless series is phonetically natural. Further, I have suggested above that the short durations of the Ienis obstruents may be contributing to their fricativization, since there is less time to reach the full stop closure target. This is an area that clearly needs further investigation; however, I submit that the interrelationships among duration, voicing, and closure type will be the most fruitful avenues for further research into the contrast discussed here . The fact remains that there are languages such as Zapotec and Jawoii which, since they do not rely primarily on VOT, need some sort of cover term to indicate their main contrast so that it can be discussed phonologically. It seems likely that the traditional terms "fortis/lenis" will continue to be used for this purpose, as developing more phonetically accurate terms must await further clarification of the actual factors involved in the contrast. Therefore, I would conclude that when these terms are used, they should always be accompanied by a careful explanation of the phonetic properties of the segments so described ; as Catford (1977, p . 203) similarly concludes: "the terms tense/lax, strong/weak, fortis/lenis, and so on, should never be loosely and carelessly used without precise phonetic specification". I would like to thank the following people for assistance in the collection and analysis of the data, and for useful discussions of the questions involved: Leigh Lisker, Francesca Merlan, Phil Rose, Franz Seitz, and Robert Van Valin . I would also like to thank our Zapotec speakers, Don Sixto Garcia Francisco and his family, and our Jawon speaker, Mrs. Alice Mitchell. The phonetic analysis was carried out in part at the Phonology Laboratory, U.C. Berkeley, and in part at the Phonetics Laboratory, Australian National University; the paper was written while I was a Visiting Fellow in the Linguistics Department at the Australian National University. References Breen, J. G. (1981) . Margany and Gunya. In: Handbook of Australian Languages (R. M. W. Dixon and B. J. Blake eds). Vol. 2, Canberra: A.N.U. Press. Catford, J. C. (1977). Fundamental Problems in Phonetics. Bloomington: Indiana University Press. Cole , R . A. & Cooper, W. E. (197 5). Perception of voicing in English affricates and fricatives. Journal of the Acoustical Society of America 58, 1280-1287. Di xon , R . M. W. (1980) . Th e Languages of Australia. Cambridge: Cambridge University Press. Falc'hun, F. ( 1951). Le systeme consonantique du breton avec une etude comparative de phonetique experimentale. Rennes : Imprimeries reunies. Fischer-J ¢rgensen , E. ( 1969). Voicing , tenseness and aspiration in stop consonants, with special reference to French and Danish. ARJPUC, 3, 63-114 . Fischer-J ¢rgensen , E. ( 1972). Kinesthetic judgement of effort in the production of stop consonants. ARIPUC, 6, 59-73. Hardcastle, W. J. (1973). Some observations on the tense-lax distinction in initial stops in Korean. Journal of Phonetics I, 263-272 . Hardcastle, W. J. ( 1976) . The Phy siology of Speech Production. New York: Academic Press. Jaeger, J. J . (197 8). Speech aerodynamics and phonological universals. In : Proceedings of the Fourth Annual Meeting of the Berkeley Linguistics Society. pp. 311-329. Jakobson, R., Fant , C. G. M., & Halle , M. (1951) . Preliminaries to Speech Analysis: The Distinctive Features and Their Correlates. Cam bridge: M.l.T. Press. (1963 edition.) Jakobson , R. & Waugh, L. R . (1979) . The Sound Shape of Language. Bloomington : Indiana University Press. Klokeid , T . J. (1969). Thargari phonology and morphology. Pacific Linguistics, Series B, No. 12.
The fortis/len is question
189
Ladefoged , P. (1967) . Three Areas of Experimental Phonetics. London: Oxford University Press. Ladefoged, P. (1971). Preliminaries to Linguistic Phonetics. Chicago: University of Chicago Press. Lehiste, I. (1966). Consonant Quantity and Phonological Units in &tonian. Bloomington: Indiana University Research Center in Anthropology, Folklore and Linguistics. Lehiste, I. (1970). Suprasegmentals. Cambridge: M.I.T. Press. Lisker, L. (1957). Closure duration and the intervocalic voiced-voiceless distinction in English. Language, 33, 42-49. Lisker, L. (1963). On Hultzen's "Voiceless Ienis stops in prevocalic clusters". Word, 19,376-387. Lisker, L. & Abramson , A. S. (1964). A cross-language study of voicing in initial stops: acoustical measurements. Word, 20,384-422. Li:ifqvist, A. (1975) . A study of subglottal pressure during the production of Swedish stops. Journal of Phonetics, 3, 175-189. Malecot, A. ( 1970). The Ienis-fortis opposition: its physiological parameters. Journal of the Acoustical Society of America, 47, 1588-1592. McDonald, M. & Wurm, S. (1979) . Basic materials in Wal)kumara (Galali): Grammar, sentences and vocabulary. Pacific Linguistics, Series B, No . 65 . Swadesh , M. (1947). The phonemic structure of Proto-Zapotec . International Journal of American Linguistics, 13, 220-230. Umeda, N. (1977) . Consonant duration in American English. Journal of the Acoustical Society of America, 61, 846-858.