How can I differentiate between syllable-initial [ɣ] and [ə] using Praat or other software?

Question

I am currently studying Amdo Tibetan. In this language the voiced velar fricative [ɣ] is reported to occur as the first sound in some syllable-initial consonant clusters. More specifically, this sound occurs only before voiced consonants. (Before voiceless consonants this phoneme surface as voiceless [χ] or [h].)

Here are some examples of phonetic transcriptions of words starting with this sound (taken from various academic papers):

• [^ɣza ^ɦda wa]
• [ɣ^əzɨm]
• [ɣgon pa]
• [ɣlab raŋ]
• [ɣgar]
• [ɣdʑak]

There are many papers written on the phonology/phonetics of Amdo Tibetan and all of them use the IPA symbol [ɣ] for this sound. This so-called "pre-initial" sound is quite common in Amdo Tibetan and I hear it quite often when listening to the recording that accompanies my Colloquial Amdo Tibetan textbook, for example.

The problem is that no matter how hard I listen, I can't seem to hear that this is a velar fricative. In fact, I don't hear it as a fricative consonant at all. Instead, it just sounds like a very short [ə], or at least some other kind of central vowel. Thus, instead of [ɣlab raŋ] I hear something like [əlab raŋ].

My native language is English and a voiced velar fricative is not a sound I can produce naturally, but I have read a description of it in "A Course in Phonetics" and I've listen to a recording of this sound on Wikipedia (see here: http://en.wikipedia.org/wiki/File:Voiced_velar_fricative.ogg). Yet to me this sound does not sound at all like the sound that Amdo Tibetan linguists are transcribing as [ɣ].

My question is: Is this [ɣ] sound actually being pronounced more like a [ə] (or some similar vowel sound-- in which case the transcriptions are "inaccurate"), or do I just hear it as [ə] because I am unaccustomed to hearing a syllable-initial [ɣ]? Or perhaps this velar fricative is being pronounced so lightly that it's fricative nature is not easily heard? Can I use software such as Pratt or WaveSurfer to figure out which sound is actually being produced?

P.S. I can upload a sound file if that would help.

Answer 1

One thing to keep in mind is that IPA is language-dependent, i.e. there is not a one-to-one correspondence between IPA symbols and speech sounds. This point is sometimes glossed over in introductory linguistics courses, where students sometimes are given the impression that the same symbol represents the same sound in all different languages. But this is far from true. Take the symbol [s] for example. What we call [s] in various languages is articulated in a variety of ways, and it is acoustically different in those languages, as well. When my Icelandic friends produce [s] I hear it as somewhere between [s] and [ʃ]; Icelanders, meanwhile, have trouble hearing the [s]/[ʃ] distinction in English because they have no such contrast phonemically (/s/ vs. /ʃ/) in their native language. Indeed, in a spectrogram, the [s] produced by my Icelandic friends shows fricative noise in a frequency range somewhere between that of my [s] and my [ʃ]. Another example comes from Hebrew, which allows many more complex onset clusters than English does. Rina Kreitman played some Hebrew clusters for non-Hebrew-speaking listeners, and those listeners sometimes reported hearing consonant-schwa-consonant sequences, presumably because their native languages didn't allow for such clusters; Japanese speakers also report "hearing" vowels between consonants in English clusters.

All of this is to make the point that, yes, you are likely being influenced by your native language when you are hearing this so-called voiced velar fricative as a vowel. English doesn't have such a fricative; it doesn't even have the approximant version of it (transcribed as [ɰ]), nor are approximants allowed in that position in onset clusters (*[wga], e.g.), so it makes sense that you would map this sound to the closest thing that your native language's phonotactics would allow in that syllabic position--a mid-central-ish vowel.

That being said, it is indeed quite possible that, articulatorily, the sound most often being transcribed as a velar fricative is being lenited to an approximant, in the sense that there is not enough of a constriction in the vocal tract to produce audible turbulent noise. This is quite common cross-linguistically; for example, the pronunciation of the Icelandic word saga is usually transcribed as [saɣa], but in reality it is often produced as [saɰa]; that is, the back of the tongue and velum approach each other for the velar consonant but they never get close enough to result in audible fricative noise.

Acoustics aren't going to help you distinguish between a vowel and an approximant (unless you have minimal pairs in the language under investigation), because the distinction between the two is largely phonological. You might be able to see evidence for deciding between the fricative [ɣ] and the approximant [ɰ], though. The former should manifest itself with some aperiodic "noise" in the spectrum--some of the otherwise regular vertical lines corresponding to the regular glottal pulses in vowels and approximants will be "smudged over" with textured noise.

Note, though, that even in the absence of visible aperiodic noise, researchers might still label this sound as a velar fricative in broader transcriptions (see the second paragraph in the Wikipedia entry for the velar approximant)!