1. Introduction

Vowels involving friction and syllabic fricatives are relatively rare among the languages of the world, but despite this are geographically fairly widespread. They have been reported in a number of Sino-Tibetan languages, including Standard Chinese, Liangshang Yi (Ladefoged & Maddieson, 1996), Lanzhou and Wenzhou (Dell, 1994); in the Salish languages of the Pacific coast of Canada and the United States (Hoard, 1978), including Bella Coola; and in the Afroasiatic languages, some Berber lects exhibit a range of syllabic consonants, including syllabic fricatives (Coleman, 1996). Among the languages of Africa south of the Sahara, Lendu, a Central Sudanic language spoken in NE Zaire has been the most closely examined (Dimmendaal, 1986; Kutsch-Lojenga, 1989; Demolin, 1993). In Bantoid, at least some of the Grassfields Bantu languages are known to have friction somehow associated with vowels (Elias, Leroy & Voorhoeve, 1984; Fransen, 1995) while Kelly (1974) reports fricative (labiodentalized) vowels in Fang, a Narrow Bantu language spoken in Gabon and Southern Cameroon. This latter work has recently been confirmed and explored in greater depth by (Mve, 1997).

That these languages from different parts of the world all exhibit what have been referred to as fricative vowels, or perhaps more commonly syllabic fricatives, should not be taken to mean that identical or even similar phenomena occur in each. It is in fact the wide range of phenomena found that has lead Ladefoged and Maddieson (1990) to adopt the term `fricative vowel' (in that it is more general) in preference to `syllabic fricative'. On the other hand, the most common analysis for most of these sounds is to view these as marginal fricatives spreading to nuclear position under specifiable conditions, e.g. deletion of the vowel nucleus with compensatory lengthening of an adjacent fricative. An alternative analysis in some instances is to see these as high vowels having fricative allophones when preceded by a fricative. In this paper I look at the fricative vowels found in the Len dialect of Mambila, which appear to require a different treatment than the usual. Two primary issues are addressed: first, what is the most appropriate phonological characterization of the fricative vowels of Len - and here I hope to show that these are specifically fricative vowels and not syllabic fricatives; and second some speculation is offered as to the historical developments which might be associated with these vowels. Before looking at Len, I first provide a brief description of the fricative vowels found in some Grassfields languages, Bantoid languages which are geographical neighbours and genetic cousins to Mambila. I will then give a brief general overview of the relevant aspects of Len phonology, followed by a discussion of the phonetic characteristics of the fricative vowels found in Len which help determine their phonological characterization. This is followed by speculation as to their historical development, including the possible relation to similar phenomena found in Narrow Bantu. It would be premature at this point to pretend to offer anything resembling a definitive statement on the historical issues.

2. Fricative vowels in Bantoid

The main reports of fricative vowels or related sounds in Bantoid deal with the Grassfields languages. Elias et al. (1984) mention only briefly that "in some [Mbam-Nkam or Eastern Grassfields] languages has caused proto C1 to have reflexes with heavy frication" (p.58). There is no explicit reference to fricative vowels, but examples given show clearly that there are words with nuclear friction; e.g. > `war'. They go on to comment (p. 59) that "the main difference between reflexes of and is that the first may modify the preceding consonant [i.e. in the manner just described], whereas the second does not. It is also apparent that has a similar though perhaps less extensive effect.

Fransen (1995) in describing Limbum (part of the Mbam-Nkam group, and geographically relatively close to Mambila), reports causing labiodentalization of preceding stops, as well as , , , and , and that is realized as when it both follows a labiodental fricative and precedes . She goes on to say that in closed syllables, "the transition between the labiodentalized consonant and the final consonant is so close that one hardly hears the vowel and one is inclined to assume syllabic consonants" (p. 35). Hence `bird', for example, is realized as , or . Fransen also mentions the occurrence of an alveolar fricative following the labial stop, giving a surface contrast of `ash' vs. `give birth'; the underlying difference is said to be one of palatalization of the labial, i.e. vs. respectively.

Kelly (1974) has described a very similar phenomenon in Fang, a Bantu language spoken in Southern Cameroon and Gabon. Following Guthrie, Kelly observes labiodentalization, "contact between the upper teeth and the inside of the lower lip" (p.120) including of , (unlike in Limbum), but for these latter only open approximation is involved. Labiodentalized consonants are followed only by or "zero, in which case the labiodental element itself carries the pitch [...] and may now be regarded as `syllabic'." (p. 120). Kelly also notes that central vowels are often retracted following labiodentals. The voiced labial stop may also occur with friction, as or followed by a close front vowel with friction, and contrasts with plain followed by a non-fricative close front vowel. Mve (1997) also looked at Fang, though different dialects than that examined by Kelly. His is the only study of these languages to provide instrumental evidence of the friction discussed (which didn't occur in all of the dialects looked at), with spectrograms showing clearly that this friction may occur throughout the duration of the nuclear portion of the syllable.

3. Len-Mambila

Mambila comprises a cluster of dialects or languages straddling the Nigeria-Cameroon border. Although its precise classification is subject to verification, the most widely accepted hypothesis has it as part of North Bantoid (Blench, 1993), and as such an early offshoot of Bantoid. Consequently, while it is a relatively close relation of Narrow Bantu in the overall (i.e. Niger-Congo) picture, within Bantoid, it is by this classification quite distant from Narrow Bantu. The Len dialect is spoken in Bang, Saam and neighbouring villages in Nigeria and extends south to Kumchum and Bang Kukwum in Cameroon. There is phonetic/phonological and lexical variation within Len (correlating roughly with villages in Nigeria as opposed to those in Cameroon). Variation regarding the realization of the vowels studied here does occur, though it is not relevant to the main points presented. The data used in this paper come from speakers on the Nigerian side of the border, and consists mainly of a wordlist of some 900 items collected from three speakers of Len. Recordings of a wordlist containing the target vowels (i.e the fricative vowels and contrasting high vowels) were additionally made (at least three repetitions of each) from two other Len speakers.

4. Len phonological structures

4.1. Syllables

The basic syllable structures of Len (as in other Mambila lects) are simple. Only three syllable shapes are permitted, CV, CVC, and V; with the latter being restricted to affixes. With one apparent exception, any consonant phoneme may occur in onset (C1) position; the range of consonants available to coda (C2) is restricted to . The last of these occurs only rarely in the data, and never in onset position. There are no consonant clusters, though affricates and prenasalized consonants do exist. The vast majority of words in the language are monosyllabic. One of the variations found within the Len speaking area appears to be the greater frequency of CV syllables (through loss of C2) in the southern (Bang Kukwum) district.

4.2. Consonants

Table I lists the consonants which may occur stem initially in Len. These represent phonetic occurrences, though most appear to be phonemic.

Of these, the status of those in parentheses is unclear; at least one of the speakers consulted did not use these variants at all, and it is possible they should be seen as free variants, with prenasalization being the variant feature. It may represent regional variation within the Len area indicative of change in progress (similar variation appears to exist across Mambila lects). In one case - - it is not entirely clear by the available data that it is distinct from its voiced counterpart. Other than these, initial , , and b are rare in the available data, though they do appear to be contrastive, and similarly, occurs only infrequently. Finally, is normally realized as phonetically, and it may be desirable to analyze this phonemically as a sequence of consonant plus glide. These questions are left open for the present. Consonants represented in square brackets are major allophonic variants, , of and of .

4.2. Vowels

The following charts plot (a) the main phonetic vowels found in Len, and (b) a possible phonemic representation of vowel system. (As with the consonant system, some cases await additional data for resolution.)

As the phonetic inventory indicates, there are two vowels in Len with friction associated, one with palatal friction and the other with labiodental friction. One of the main questions of interest is whether either or both of these vowels should be considered phonemic, or whether they are conditioned variants of some other vowel.

4.3. Phonotactics

Other studies dealing with similar phenomena (i.e. the Grassfields and Fang work discussed above) have focused on the labiodentalization of consonants. From this perspective, there is a subset of Len consonants which may be said to have labiodentality associated with them; viz: , , , , ; there is however, only one vowel which may follow these, . is also found with other labiodental consonants ( , , , ), where it contrasts with , though there is variation across speakers in this regard. Examples of these, including instances of contrasting high vowels, are shown in Table II.

Also similar to the Fang and Grassfields reports, it is possible to speak of a palatalized voiced labial stop in Len. Only one vowel may follow this stop, the high front vowel with palatal friction. This vowel is also found to co-occur with post alveolar fricatives/affricates as opposed to alveolar fricatives/affricates, however unlike the ostensible palatalized labial, the postalveolar fricatives/affricates are not restricted to co-occurring with the fricative vowel.

Examples, including instances of contrasting high vowels, are shown in Table III:

4.4. Phonetic characteristics

4.4.1. We consider first the vowel with palatal friction, . For some speakers this has been found to have a clear front quality, [i], while for others a more centralized quality, , exists. The spectrograms in Figure 1 illustrate instances of `ask' from two different speakers (1a, b) compared with `do' as spoken by one of these speakers (1c). The following characteristics may be noted. First, the fricative noise is consistent with a fricative produced in the post-alveolar region; second, there is a clear formant structure of a vowel of a high front quality `underlying' this fricative not typical of consonantal, or marginal, fricatives. (Note F2 is lower in 1a, b compared to 1c, indicative of a somewhat retracted quality.) Third, the vowel can be seen to start with the release of the stop closure - i.e. there is no intervening period of friction. Fourth, the formant transitions into the vowel give no indication of a palatalized release to the labial consonant. And fifth, the location of the friction, or its peak intensity, is variable (in (1a) it is in the first half of the vowel, in (1b) towards the end of the vowel), though is associated with almost the entire duration of the vowel. In these examples this may be due simply to an early cessation of voicing in (1b), however it is noteworthy that when asked to do several repetitions of the same word, for all speakers the location of the friction appeared to vary somewhat, and in some instances would disappear entirely.

4.4.2. Like the one with palatal friction, the vowel with labiodental friction is somewhat variable in quality, but always in the range of a high central unrounded quality. With respect to articulation, while this vowel is described as having labiodental friction, it should be pointed out that the stricture at the lips and teeth is more open than that observed for the labiodental fricatives . A spectrogram of this vowel is shown in 2a, compared with spectrograms for (2b), (2c), and (2d). Again the fricative noise can be seen to co-occur with a clear formant structure, and it is typical of that associated with labiodental (or labial) fricatives in that it is more diffuse and weaker in intensity than that found for other fricatives. The formant structure is consistent with that of (cf. , 2b). Notice again that there is no intervening period of friction between the consonant and the vowel. Though 2a may appear to suggest this, 2b, c illustrate that this is maybe interpreted as a period of aspiration associated with the velar consonant. Finally, 2d allows comparison with a labiodentalized velar stop as found in a neighbouring language, the Ndung dialect of Kwanja. A clear boundary between the fricative noise and the vowel is observable.

One further point of interest may be noted: in comparing the spectrograms of and , the vowel qualities associated with the two articulations are clearly different, though both represent high vowels rather than a schwa-like quality often associated with syllabic fricatives (cf. Demolin 1993).

4.5. Summary and discussion

To summarize this section, from the distributional evidence examined it is apparent there is one vowel in Len which may be termed a fricative vowel. That is, the two surface vowels having friction are found to be in complementary distribution, one variant following a labial consonant and the other coming after non-labial consonants. The nature of the friction, but not its existence, appears related to the preceding consonant in a lingual-labial pairing. Exceptions to this generalization are the labiodental fricatives, which take labiodental friction, and the post-alveolar fricatives/affricates, which take palatal friction. It is characterized as a high central unrounded vowel ( ), as it contrasts with both front and back unrounded vowels, and its main realizations are more central than either of these two. One could conceivably argue, as Fransen does for Limbum and Kelly for Fang, that this high central vowel causes (certain) preceding consonants to labiodentalize (or palatalize, in the case of labials); for Len, this begs several questions, not least of which is why the place of articulation of the friction should oppose that of the consonant. More expected would be for the POA of friction to approximate that of the consonant, as we do see for the labiodental fricatives and post-alveolar fricatives/affricates, though Dell (1994) in fact wonders why high vowels should precipitate friction at all. More important, the phonetic evidence gives no support at all to the supposition that the preceding consonant has been fricativized; rather, it may be best considered a secondary articulation on the vowel, not the consonant. There appears to be little reason, then, to add labiodentalized or palatalized consonants to the phonetic inventory of Len, claim this is precipitated by the high central unrounded vowel, and then subsequently have to argue that this feature spreads back to the vowel, or syllable nucleus.

The fricative vowel of Len, therefore, appears to be just that - a fricative vowel and not a syllabic fricative - and to further justify Ladefoged & Maddieson's (1990) use of this term. This of course raises important issues with respect to their phonological characterization, particularly their feature specification, as well as broader questions of theory. A full treatment of these issues is beyond the scope of the present paper, however it is worth looking briefly at a proposal found in Zoll (1995), as this has some relation to issues raised in the next section.

In looking at the first degree (`superclose') vowels of many Bantu languages, and their spirantizing effect on preceding consonants, Zoll proposes that these vowels be defined as [+consonantal], in order to distinguish them from high vowels and capture their influence on preceding consonants. Without commenting on the merits of this analysis for Bantu first degree vowels, it would appear unsatisfactory for Len. First, it would require [+consonantal] elements in the syllable nucleus (which Zoll accepts in her case), and while this may be an appropriate analysis in some languages, certainly the expected case is that syllable nuclei are [-consonantal]. Second, these vowels in Len, with their clear formant structure, bear all the phonetic hallmarks of vocalic, i.e. [-consonantal] segments. Rather, it seems exploring other avenues may provide additional insight into an appropriate analysis of these vowels in Len and related Grassfields languages.

5. A Historical perspective

Among the important questions with regard to fricative vowels, particularly given their relative rarity, are `where do they come from?, where do the go? and why?' These have seldom been addressed directly in the literature, but all are obvious and interesting questions. In this section I address briefly these questions, presenting some speculations on the historical development of fricative vowels and syllabic fricatives and the relations of the Len fricative vowel. Given our current state of knowledge of the Bantoid branch, it should be clear from the outset that we are still quite some way from being able to provide answers to these historical questions with any degree of confidence.

5.1. Development of fricative vowels/syllabic fricatives

The literature on fricative vowels or syllabic fricatives seems to suggest that for the most part, they are surface phenomena; ready examples come from Chinese, Japanese, or even English in certain dialects and under certain circumstances (see Abercrombie, 1967; Laver, 1994). As such, questions to do with their history don't arise. In those few studies where a diachronic interpretation has been required or attempted, it has usually been one of vowel deletion followed by compensatory lengthening; i.e. it mirrors what has been proposed as a synchronic analysis. In the rare cases where phonetic evidence has been brought to bear on the examination of fricative vowels (e.g. Coleman, 1996; Demolin, 1993), a different analysis has been arrived at, indicating the vowel has not been lost, but rather has been masked through increasing overlap of vowel and consonant gestures. The acoustic evidence presented earlier is consistent with such an analysis for Len; i.e. the data clearly show a vowel is present simultaneous with the fricative. However, there is no direct comparative evidence to support such an analysis for Len diachronically - i.e. there are no C1 fricatives historically that might have spread to the nucleus. It is not inconceivable, though, that the fricative component of the vowel, characterized earlier as a secondary articulation, is the reflex or remnant of an earlier consonant, and that earlier CVCV (or even CVC) structures have, through various reductive processes, resulted in the present forms exhibiting fricative vowels. That such types of change are possible is uncontroversial, the evidence for gestural overlap is now well documented; see, for example Browman & Goldstein (1991) regarding historical change. More specifically, Williamson (this volume) examines developments in Igboid (though not within a gestural framework), some of which can handily be referred to as compression, and which have in some cases resulted in a fricative in C1. It is also clear (Connell, 1996) that earlier CVCV structures in Mambila have in some lects reduced to CVC and CV.

5.2. Correspondences with Len fricative vowels

It is an interesting fact that there is a high degree of correspondence between the Len fricative vowels, those found in various Grassfields languages and in Fang, and the first degree vowels of Proto-Bantu. Examples illustrating the correspondence are given in the Appendix. In the majority of Len words containing a fricative vowel where there is a cognate in PB, that cognate form contains a first degree vowel. Of the few exceptions, e.g. `short', the PB form can only tentatively be considered cognate when comparison with other forms in the set is taken into account. Conversely, there are considerably fewer PB forms where there is a first degree vowel but no corresponding fricative vowel in the Len cognate (`excrement', which may be borrowed (but cf. fn. 3) in Len from other Mambila lects, and `vomit', = PB `spit' are two examples given here).[1]

This high degree of correspondence requires explanation, and suggests the phonetic nature of the Len (and Grassfields) vowels may have implications for our understanding of PB vowels. In considering the possibilities, however, the assumed genetic relationship between PB and Len-Mambila must be borne in mind; that is, Len represents not a descendent or daughter of PB but, at closest, a development of a sister of PB[2], and conceivably may be even further removed. Consequently, any possible relationships or correspondences between Len and PB (as well as Grassfields) must be considered in terms of a common parent (or even `grandparent') language. The immediate implication of this is that it would be premature to conclude that the first degree vowels of PB were themselves fricative vowels like those of Len.

Nonetheless, the notion that the first degree vowels of Proto-Bantu may have been fricative vowels is not new, though it does remain controversial and has never been adequately defended. It presumably came about partly because of the apparent spirantizing influence of these vowels on preceding consonants and the fact that they have been referred to as `superclose' vowels; and while most Bantuists would agree that spirantization was triggered by the first degree vowels (cf. (Schadeberg, 1994/5) , few if any would accept the notion that Proto-Bantu first degree vowels were themselves `fricative'. Meinhof (1932), however, evidently did consider his Ur-Bantu and to have friction associated with them.[3]

The interesting question for this aspect of Bantu historical linguistics then is what was it about the PB first degree vowels that triggered spirantization. A symmetrical system based on four degrees of height is usually claimed for PB; others (e.g. Stewart, cited in Schadeberg 1994/5) suggest the distinction between PB , and , was based on the feature [+/-ATR], the first degree or `superclose' vowels being [+ATR].[4] Meinhof considered PB to have seven vowels, but there is some indication that for him it was a system with three degrees of height, rather than four, with and represented in his chart (p. 27) as high central vowels; interestingly, this accords with their place in Len-Mambila and Eastern Grassfields. Also of interest, in light of the discussion above on contraction and reduction processes, is that Meinhof assumed these vowels ( and ) came about as a result of contraction of and . And while it must be made clear that Meinhof's view of contraction here is not very explicit, and few if any Bantuists now accept this explanation if PB vowels, it did receive considerable support from Bourquin (1955). It should also be made clear that Meinhof's view of contraction does not match that presented above, as it apparently involved only contraction of vowels, whereas the view here is that there was presumably an interaction with a consonant.

A second possible implication of the correspondence found between the PB vowels and those of Len, is that the latter may somehow reflect an intermediate stage in the spirantization process found in Bantu. This, too, must be treated with caution: though it may prove to be on the right track, it must be pointed out that, with the exception of Fang, such a stage is not found in any Narrow Bantu language. Nevertheless, a connection between the Len vowels and those of PB must be admitted; an adequate understanding of this relationship and the processes involved will ultimately require greater understanding of both the historical relations between the various languages involved and the phonetic processes involved in historical change.

Acknowledgments: