The introduction of the term approximant has been a useful one for phonetics. It has come to replace the earlier terms frictionless continuant and semivowel. It is usually credited to Peter Ladefoged, in his early work A Study of West African Languages (1964), and Abercrombie recommends adoption of the term in his 1967 book Elements of General Phonetics. The basic idea is straightforward. We can take the opening section of Wikipedia’s article on Approximant as the basis for a provisional definition:
Approximants are speech sounds that involve the articulators approaching each other but not narrowly enough, nor with enough articulatory precision, to create turbulent airflow. Therefore, approximants fall between fricatives, which do produce a turbulent airstream, and vowels, which normally produce no turbulence. This class is composed of sounds like [ɹ] (as in ''rest'') and semivowels like [j] and [w] (as in ''yes'' and ''west'', respectively), as well as lateral approximants like [l].
I have no problem with this as a definition of approximant. However, a much later introduction has been the concept of voiceless approximant. An example of these often quoted is the sound [ʍ] as found in the “voiceless w” in some English speakers’ pronunciation of ‘whine’ (particularly in Scottish accents). Another is the palatal sound at the beginning of the word ‘hue’. Ladefoged and Maddieson (The Sounds of the World’s Languages, 1996, p 326) give a few examples of voiceless approximants: speaking of the distinction between ‘you’ and ‘hue’ in English, they write “the onset in the second word is normally a voiceless palatal approximant, [j̊], for which the IPA has no unitary symbol.”. Later, they write “In … Scotland, the words ‘weather’ and ‘whether’ contrast, the latter beginning with a non-fricative [ʍ]”. From the data they report, it seems extremely unlikely that any language has a phonemic contrast between a voiceless approximant and a voiceless fricative.
The problem with voiceless approximants now confronts us. We know from the definition of approximant that if an articulation is made with a narrow enough air passage to produce turbulent airflow, and hence to generate fricative noise, then it cannot be an approximant. If the airflow and articulation of an approximant have to be such that turbulence is not generated, how can a voiceless approximant be audible? If a voiceless approximant is audible, how can it be non-fricative? This dilemma is presented in a very useful discussion piece arising from disagreement among Wikipedia contributors at phonology - Do voiceless approximants exist? What is the consensus among phoneticians/phonologists? - Linguistics Stack Exchange
It seems that the only way out of this dilemma is to argue that there are two ways of generating an audible voiceless continuant sound, one being the familiar type of most fricatives, where the noise resulting from turbulent flow is produced at the point of articulation, and the other being noise produced by non-turbulent airflow and generated in the vocal tract, though not at a specific point of articulation. A detailed explanation by Shadle (“The aerodynamics of speech” in Hardcastle and Laver (eds) The Handbook of Phonetic Sciences,1997, p.37) describes the difference between turbulent and laminal flow (interestingly, though she produced valuable work on speech aerodynamics she was an engineer who had previously been working on jet engines). If we go back to the pioneering work of Kenneth Pike, in his ‘Phonetics’ (1943) we find a distinction proposed between ‘local’ friction and ‘cavity’ friction. Pike’s cavity friction presumably depends on laminar flow. On p. 71, he writes “One of the marked weaknesses in current usage of friction as a criterion for consonant vowel differentiation lies in the failure to distinguish satisfactorily between two types of friction which function very differently and have different origins, even though the border lines are not sharp between them. Generally speaking, one type retains its audibility when voiced (e.g. for sibilants) the other is a weak fricative and audible only when sounds are voiceless (e.g. most vowels and certain sonants). The first type results from stricture at a single local point; the second is due to cavity friction, that is, voiceless resonance of a chamber has a whole caused by air going through it as through an open tube.”
It does seem, then, that there is at least a theoretical possibility of a viable phonetic distinction between voiceless fricatives and voiceless approximants.
In another theoretical discussion, Catford (Fundamental Problems in Phonetics, 1977, pp. 120-124) suggests a three-way distinction between fricative, approximant and resonant. “In fricative articulation the articulatory channel is very small and the flow through it is always turbulent.” Of approximants he writes “they have non-turbulent flow when voiced; but the flow becomes turbulent when they are made voiceless”. Of resonant, he writes “In voiceless resonants … there must always be turbulent flow through the glottis; if there were not, there would simply be silence.” In the case of Catford’s resonant, therefore, the noise excitation in the vocal tract is the same as for [h], being made by turbulent flow through the glottis. A similar position is found in Laver’s Principles of Phonetics (p. 269): “All resonants have a stricture of open approximation. A necessary condition for the performance of resonant is that the airstream passes through the vocal tract in a smoothly laminar flow, with no audible local frication.” The idea of voiceless resonant is necessary for understanding how voiceless vowels are audible.
I haven’t said anything about the acoustic characteristics of voiceless approximants, mainly because I have not yet had time to analyze a controlled set of recordings of comparative fricatives and voiced approximants. What I expect to find in spectrograms (and have found in a few trials) is that in a voiceless approximant made without local friction as far as I can manage, the formants corresponding to those of the voiced equivalent are visible in the noise spectrum generated by laminar flow. In the case of voiceless continuants with local friction, what is most visible is regions of high-frequency energy similar to those of other voiceless fricatives.
Now we need to look at how the concept of voiceless approximants works in practice rather than in theory. Pike’s work was aimed at establishing a pure theory of phonetics removed from the influence of phonology and of phonemic function. But however fine the distinctions he considered between phonetically different articulations, distinctions were only of interest if they were above the threshold of perception (Pike, p. 151). This is a relevant condition in the present issue.
To keep things simple, we can confine the discussion to the voiceless [j] and [w] sounds of English, (which I will call ‘hue’ and ‘whine’), and leave aside sounds such as [h] and [ɬ]. Accounts of voiceless approximants in the context of European languages seems to be always found in discussing voiceless counterparts of (voiced) approximant phonemes: these are often (but not always) voiceless allophones of voiced approximant phonemes. Sounds classed as voiceless approximants appear to be always made with the tongue close to the roof of the mouth, though the place of constriction is of course variable. They are related to a number of other similar sounds. The relationships can be set out like this:
(1) Voiceless approximants and voiceless fricatives: if the articulation produces local friction with turbulent flow, we get a palatal fricative [ç] (though some writers on English phonology prefer the symbolization [hj]) for ‘hue’. For ‘whine’ the IPA symbol is [ʍ] in the case of a voiceless fricative, but this is also used for the suggested voiceless approximant.
(2) Voiceless approximants and devoiced fricatives: the latter can be shown with the alternative of a devoicing diacritic [ ̥ or ̊ ] appended to the symbol for a voiced fricative. Thus ‘hue’ is [j̊] and ‘whine’ is [w̥].
(3) Voiceless approximants and voiceless fricatives: a lesser degree of articulatory constriction can be shown by means of an “open” diacritic [ ̨ ], which could be taken to imply absence of local friction. Thus [x̨], [ʍ̨].
(4) Voiceless approximants and devoiced vowels: we could in theory show ‘hue’ as [i̥] and ‘whine’ as [u̥].
If the voiceless approximant is to be acceptable as a viable phonetic category, it must fulfil certain conditions:
- No sound can be classed as a voiceless approximant if it contains audible local friction.
- Phoneticians must be able to make a reliable auditory discrimination between voiceless approximants and voiceless fricatives. The auditory difference between local friction and cavity friction must be sufficient to enable the distinction to be made.
- It must be possible for a phonetician to produce and demonstrate clearly distinct voiceless fricatives and voiceless approximants.
- Languages must exist where voiceless approximants (as defined in (1)) occur regularly in speech.
I do not believe that the conditions listed above are met. On the other hand, it seems clear that at least some phonetics researchers feel the need for the category voiceless approximant. I believe that the best solution is to treat the label voiceless approximant as a “cover term”, i.e. one which does not map on to a single phonetic entity, but on to a collection of related entities. I would use a form of words like this: “The term voiceless approximant denotes a voiceless sound with an articulation similar to a voiced approximant, which may consist of cavity friction resulting from laminal airflow in the vocal tract, or local friction resulting from turbulent airflow through one or more constrictions, or both”.