Historical phonetics vs. phonology

[Daniel]

I've just noticed that while I see a clear distinction in synchronic linguistics and modern languages between the domains of phonetics and phonology, I have trouble identifying any distinction in diachronic linguistics much of the time. Further, it isn't really phonology (since that's a grammatical system in one speaker's mind), nor is it really phonetics because we're not really working with phonetic data. But I guess it's more like phonetics in that we're approximating broad phonetic descriptions.

Any comments on this?

[Guijarro]

I don't know whether this is relevant or not, for this field has never attracted me too much.

I remember Halliday making a distinction between PHONETICS and PHONEMICS, telling us a long story about the EMIC/ ETIC distinction which I have unluckily forgotten.

Maybe phonemics is just another term for phonology? No idea.

But I mention it just in case it has any relevance and you want to find out.

[freknu]

I don't know whether this is relevant or not, for this field has never attracted me too much.

I remember Halliday making a distinction between PHONETICS and PHONEMICS, telling us a long story about the EMIC/ ETIC distinction which I have unluckily forgotten.

Maybe phonemics is just another term for phonology? No idea.

But I mention it just in case it has any relevance and you want to find out.

I don't know how linguists usually treat the distinction — I rather frequently end up in arguments over IPA, linguistics, and other stuff (not just on this forum; because I'm somewhat of a pedantic, nitpicking git that can't stand ambiguities and half-hearted, wand-waving science) — but I make a stone hard distinction.

• Phonetics is low-level (narrow) study centred on acoustics — the sounds we produce, hear, process, understand, etc.
• Phonemics is high-level (broad) study centred on abstraction — how these sounds are distributed, contrasted, interpreted, represented, etc.
• Phonology is the combined study of phonetics and phonemics — the linguistic study of sounds.

As for the question of synchronic versus diachronic... I'm not quite sure I understand the issue...?

Do you mean reconstructed versus attested phonology? After all, if you are reconstructing the pronunciation of a language you cannot accurately study the phonetics of that language the same way you study an attested language; but phonemics is much easier to reconstruct and study since it need not bother with narrow acoustics, but rather only deal with abstractive contrast and distribution, which does not require phonetic data the same way that phonetics does.

Or do you mean something else?

[Daniel]

Speaking of the synchronic study of languages, the distinction is basically what you wrote, freknu, and what you hinted at, Guijarro.

Phonetics: the study of sound -- articulation, acoustics, and perception.
Phonology: the grammatical system that combines sounds into morphemes/words/etc., including language-internal contrastive categories (phonemes).


Freknu, the reason I'm wondering is this: I'm teaching a historical linguistics class and my students just submitted proposals for their term papers. The class isn't too technical yet so we're not really concerned with things like phonetics vs. phonology, but from my perspective, just reading through the proposals, I realized I had no idea which they belonged to, when they were talking about historical sound change and so forth. Is that phonology? Is that phonetics?

Let's say someone wants to study Grimm's Law (or, really, anything else). Is that phonetics or phonology?

[panini]

Grimm's Law would probably be in the realm of phonetic change, until you get into paradigmatic questions about verb roots ending in */g, gh/ being followed by */-to/. There were probably not many alternations involving GL, so at some point, words just got restructured. Verner's Law, on the other hand, has more potential for being phonological, given paradigmatic accent shifts.

An example that I think clarifies the difference between phonetic change and phonological change is the change in the inflectional pattern of geminate and (for some dialects) glide-medial verbs in Arabic, plus the merger of y-final and w-final roots, starting from Classical Arabic. In a nutshell, glides deleted intervocalically but could form diphthong-like sequences [ay, aw] before a consonant. These then became [ee, oo]. It became impossible to tell what glide the so-called lame verbs ended with and they were reanalyzed as CVCV, i.e. diconsonantal, and you have alternations like [bana, bana-t, ban-ee-t, ban-ee-na] from earlier [banay-a → banaa, banayat → banat, banay-ti, banay-na]. Geminate verbs e.g. madda also look like they are diconsonantal. What happened is that the rule inserting [ee] in /bana/ got generalized to /madda/, so you get [madd-eet] rather than [madad-ti]. This is a change in the phonological system, not a phonetic change.

[Daniel]

So contextually dependent sound change is phonological while changes in the realization of phonemes in every environment are phonetic? That makes some sense.

Intervocalic lenition would be phonological. Chain shifts would be phonetic.

But the distinction (even though that seems well applied, at least in some sense) doesn't feel very meaningful to me, in contrast to in synchronic research where it is quite important.

[panini]

But the distinction (even though that seems well applied, at least in some sense) doesn't feel very meaningful to me, in contrast to in synchronic research where it is quite important.

Well, what I think is most important is to understand the difference between the pre-grammatical stages of sound change that are simply about figuring out what sound you heard, and the more abstract inferences that have to be made about underlying forms and the phonological rules that generate the output. I'm making a plug for Hale's distinction, which I think is essential in understanding both language change and synchronic grammar.

[Daniel]

But surely those lines aren't as easy to draw as with synchronic work right? You wouldn't get a purely phonetic or phonological researcher in historical linguistics. You might sometimes make a distinction but in passing, I'd think.

[panini]

I don't see that there is any inherent difference in difficulty in distinguishing phonetics vs. phonology either synchronically or diachronically -- confusion and clarity are freely available to all parties. But perhaps you're referring to observed behavior of historical linguists vs. synchronic phonologists / phoneticians, which reflects the sociology of the fields, rather than in logic of the questions that those areas address. And then, my observation is that it's the synchronic P people who have the hardest time drawing bright lines between phonetics and phonology.

OTOH I don't think that it is important for your class purposes to categorize topics as phonology vs. phonetics vs. morphology vs. syntax. It is important for a student to understand the meaning of and consequences of phonologization of phonetic processes, likewise the morphologization of phonological rules. But those are advanced research questions.

The main sociological reason why you don't find many pure-phonetics researchers in historical linguistics is that the experimental subjects all died millenia ago. And the rest call themselves socio-phoneticians.

[Daniel]

There is certainly a clear distinction synchronically even if perhaps idealized: phonology is grammar, and phonetics is physics/physical. As much as we can separate morphology and syntax (etc.) we can separate those.

[panini]

phonetics is physics/physical.

Aha. Well, that is, indeed, the SPE position, but I think that that position leads to serious problems in grammatical theory. The issue is that languages vary substantially on how segments are actually pronounced (and often there is contextual variation within a language). Their way to dealing with that is to put everything in the phonology, except for anything that is an automatic and inviolable consequence of physics and human physiology -- anything that is language-specific is dealt with in the phonology. Since the pronunciation of [ i ] can vary hugely across languages, this is accounted for by assigning different integer values to pluses and minuses on standard features. Thus the difference between German <bieten> and English <beaten> is in the numeric value of +high, where German has a 2 and English has a 3.

But IMO the integer theory of feature implementation turned out to be extremely problematic, for the simple reason that we never could figure out what the correct integer was. If you look at the literature on phonetic tone implementation, where people did actually come up with numeric rules to generate downdrift and downstep patterns, plus if you look at some of the rules of stress computation in SPE, the idea of adding numeric operations to the arsenal of phonological rule devices imbued phonology with immense yet essentially unnecessary power.

The most compelling argument against integer-manipulative phonology is that even if it might, imaginarily, allow one to describe the "amount" of a phonetic property in a segment, it still can't handle the continuum of "whens" that are also necessary to describe e.g. the time-dependent increasing lip protrusion function of the onset in "strength", or time-dependent changes in nasalization in vowels next to nasals (see Cohn's dissertation and the contrast between English and French). The only way to phonologize this is to re-define a segment so that it is not just an ordered element is the representation, it is a continuous set of time points.

The solution to this quandary is to have a grammatical entity, "the phonetics", which is responsible for physically interpreting the categorial output of the phonology, and because languages differ in how they implement abstract segments, there will be language-specific rules of phonetics. It's hard to say what such rules would look like, I guess because most phoneticians don't seem to be particularly interested in doing detailed descriptions of multiple languages that would allow comparison and thus parametrizing differences between languages -- but it could be done.

An example of a potentially ambiguous case is aspiration in English -- is it phonetic or phonological? If you hold constant your assumptions about syllabification, you can (apparently) either treat aspiration as part of (English-specific) phonetic implementation, or as the result of a phonological rule of aspiration. There is no compelling, direct phonological fact that tells you whether the /t/ in "top" is phonologically [+spread glottis] and the /t/ in "stop" is [-s.g.]. The closest I've seen is the analogy to h-deletion, where /h/ is only possible foot-initially, in which case, if you assume a rule of de-aspiration, it could be stated as deletion of foot-non-initial [+s.g.] (so, prohíbit: prò(h)ibítion::èditʰórial:éditor). But, what in my mind takes the air out of that argument is that while /h/ deletes at least optionally in phrasal collocations creating foot-medial /h/, as in "scary (h)allucination", you don't de-aspirate word-initial phrasally foot-medial /p,t,d/ in "scary [pʰ]odiatrist". Thus, h-deletion and (de)-aspiration are actually separate processes.

Anyhow, the point is that stop-aspiration in English can be equally-well treated as a phonological process adding (predictable) [+s.g.] in a certain phonological context, or as a rule of phonetic implementation. There may, then, be an element of coin tossing in acquiring aspiration in English, which underlies the process of phonologization.

So I conclude that while phonetics and phonology qua components of language are ontologically quite distinct,  they are not always epistemologically distinct from the POV of acquisition.

[Copernicus]

I just posted a lengthy comment in the "impetus" thread that may have relevance here, so I won't repeat it.  From a historical perspective, you ought to go back to Baudouin de Courtenay, the progenitor of modern phonological theory.  What is interesting about him is that he was at the forefront of the struggle to get linguists to be interested in the synchronic study of language in the 19th century.  So right there is the dividing line between synchronic and historical linguistics.  Of course, Saussure was fighting that battle, as well, and that is why he got along so well with Baudouin's seminal concept of the phoneme. 

The crux of the matter here has to do with the fact that language exists as two distinct phenomena--a social phenomenon and a psychological phenomenon.  That is, a grammatical description can either represent something that goes on in a speech community or something that goes on in the head of an individual speaker.  We have far less access to the heads of dead speakers, so historical linguistics does not have much to say about the psychological aspect of language.  This is not to say that psychological "grammars" are irrelevant to historical change.   Quite the opposite, as "imperfect learning" was recognized even in the 19th century as a driver in linguistic change.  However, synchronic studies can address either the social or the psychological aspect of language, because we do have access to the heads of modern speakers (within limits, of course .

One thing that I find fascinating about generative linguistic theory is that it is a purely psychological approach to linguistic systems.  So it doesn't have much to say about sociolinguistics.  Indeed, Chomsky was notoriously dismissive of sociolinguistic theory and also had little interest in diachronic studies.  He even tended to confuse Saussure's social concept of Langue with his own psychological concept of competence.