Tackling cross-linguistic vowel markedness system[at]ically: features or what?

Question

I have been trying to find alternative ways of representing vowel phonemes for cross-linguistic comparisons in a unified, systematic way that would also reveal their relative (un)markedness. At the same time, the way has to be as economical as possible, so that it doesn't multiply entities unnecessarily. A question arises here whether the, now traditional, mainstream approach based on binary oppositions is the best one, or if something along the lines of dependency phonology could be better.

The problem is, for instance, that if we represent back rounded vowels as, say, [-front, +round] and front unrounded vowels as [+front, -round], it is not obvious that, inherently, they are both the unmarked classes, whereas [+front, +round] and [-front, -round] are both the marked classes (as if [+front] and [+round] were in quasi-complementary distribution in terms of markedness). Using [back], [front] and [round] as unary features, for instance, is not helpful either:

• /i/ = [front]
• /u/ = [back, round] (as if it was more marked than /i/, which it isn't)

The evident imbalance is clear when we make [round] the inherent dependent feature of [back] in the feature geometry tree, because then we would have to make [unround] the inherent dependent feature of [front]:

1.    /i/
   [front]
       |
   [unround]

2.    /u/
   [back]
       |
   [round]

In addition, the strictly binary system also fails to account for the reality of [+low] generally decreasing the (ease of) articulatory "implementability" of either [+round] or [+front] due to physiological constraints, again, as if [+low] were somehow opposed to [+front] and/or [+round], at least to some extent. The same is true of the derived, nasal subsystems in which much stronger nasalization is required in the open vowels for their nasality to be perceived than in the closed ones.

Hence my question in a somewhat different wording:

Are there theoretical approches in phonology that account better for the cross-linguistic markedness tendencies?

Relevant questions to ask might include:

• Does /ɯ/ usually arise from /u/ or from /i/ diachronically? (The example of Japanese seems to suggest the former.)
• Does /y/ usually arise from /i/ or from /u/ diachronically? (Many European languages seem to support the latter.)
• Is it correct to assume that among the close unrounded vowels, the markedness hierarchy is as follows: /i/ < /ɨ/ < /ɯ/?
• Is it correct to assume that mong the close rounded vowels, the markedness hierarchy is as follows: /u/ < /ʉ/ < /y/?

I will be grateful for any comments, suggestions, examples or references.

Answer 1

There has been a general desire to equate unmarkedness with structural simplicity, so that [i], [a] and [u] might have only one feature and [ø] would have more features. If order for this to work out, one either needs underspecification so that some values are not underlyingly present but are filled in by later rule, or are never filled in (i.e. privative features). These are assumptions shared (at least in part) by most theories, ranging from PSM, Dresher's contrast theory, Dependency Phonology and Government Phonology, UFT, and Selkirkian theory, non-exhaustively.

Assuming a version of UFT, the relevant vowel place features are [coronal], [labial], [dorsal], and either [open] or [closed] for vowel height (the former being the official Clementsian approach). Accordingly, a front unrounded vowel is formally simpler than a front rounded vowel, and a high vowel is formally simpler than a mid vowel. Unfortunately the reduction of markedness to formal simplicity doesn't progress much further in this theory, since [ɯ] would be as simple as [u] but by usual diagnostics and factual assumptions, [ɯ] is more marked than [u].

I conclude that the attempt to reduce markedness to formal simplicity is a failure, that although one can emulate that effect to some extent, attempts to fit feature theory to markedness complicates phonological theory, and ultimately fails. Instead, attention has been put on what "markedness" is, and asking what really explains those facts. The basic explanatory idea that explains away the grammatical concept "markedness" is "difficulty of distinguishing". The reason why [ɯ] is less frequent as a phoneme than [u] is that [u] employs two means of signalling low F2, and [ɯ] employs only one. So [u] and [ɯ] are too similar (hard to tell apart), and [u] more clearly / unambiguously represents the percept "low F2". The prediction then is that to the extent that a surface distinction between X and Y is difficult to maintain, that distinction will tend to be eliminated (and various phonetic and grammatical considerations will dictate what the outcome is).

I also think one should be careful about claims that are based on questionable writing practices. Very often, some language will have a vowel transcribed as "u" which is somewhere between [o] and [ʊ], but for the sake of simplifying an orthography, "u" is used. The distinctions [ɨ / ɯ] and [ʉ / y] are prime examples. If you have a contrast between the members of that opposition, then you might reasonably maintain that a language (like Norwegian) has both [ʉ] and [y]; otherwise, I would be skeptical unless there is acoustic evidence based on formant values of IPA vowels as produced by certified speakers, compared to the vowels of a given language.

[Edit]

I am usually skeptical about the notion of articulatory markedness, which would only make sense w.r.t. articulatorily-difficult gestures (e.g. intricate timing of click movement). It can't be more complicated to raise and back the tongue without protruding the lips than it can be to both raise and back the tongue and protrude the lips. The test case would be a pair of sounds, where it is physically difficult to produce the sounds differently, but when they are correctly produced, they are extremely easy to distinguish, perceptually. The problem is that we can test for ease of perceiving differences, but we have no way to quantify how hard it is to produce [q] (as opposed to [k]) or [θ] (as opposed to [s]). If you speak Czech, [θ] might be really difficult to produce, but that crazy "Czech r" which the language is famous for and I swear after a zillion tries I can't get right, that is no problem (for a Czech speaker). I'm basically willing to put all of my money on the program of reducing segmental markedness to problems of perception, and grant that cases where similar-dounding sounds can end up even more similar-sounding because the articulatory timing distinctions needed to make the phonetic distinction are difficult.

Answer 2

Patricia Donegan gives a unified and conceptually simple theory of the vowel systems of the world in her 1978 dissertation, On the Natural Phonology of Vowels. I suggest you look there.