The IPA is so immensely useful for linguistics that nobody questions its use, even though it sometimes creates misunderstandings (e.g. the misconception that the characters defined in the IPA represent the set of possible sounds in the world’s languages).
But so far, nobody has suggested that there could be something analogous in the domain of morphosyntax. Morphosyntactic concepts like “subordinate clause”, “genitive case”, “second declension”, “suppletion”, “ditransitive construction”, or “aorist” seem to be too idiosyncratic across languages, or maybe simply too complex and subtle to allow for a standardization.
However, I would argue that the similarities between the IPA characters and our morphosyntactic terms are much more striking than their differences. First of all, both kinds of notations are used in two ways. On the one hand, they are used as language-particular descriptive terms: For example, the English phoneme /t/, and the Turkish Genitive case are language-particular categories that cannot be equated with similar categories in other languages (say, the French phoneme /t/ and the Russian Genitive case). But the terms are also used for comparative concepts that help us compare the languages: Thus, both these phonemes are realized by the sound [t], and in this regard they can be said to be quite similar. And likewise, the Russian Genitive and the Turkish Genitive are both genitives in the sense that they are grammatical markers occurring on adnominal possessor nominals. The slash notation for phonemes is thus the counterpart of the capitalization of language-particular terms (as discussed in Haspelmath 2010: §5).
Now one might argue that in morphosyntax, there are simply too many different comparative concepts, as has recently been suggested by Matthew Dryer (on the Lingtyp list):
“There are thousands of perfectly reasonable comparative concepts and it is completely impractical to invent a new label for each one. For any familiar notion like affix, there are dozens of perfectly reasonable comparative concepts, varying depending on one’s interests and the type of research one is engaged in.”
Dryer is of course right that the set of comparative concepts is completely open-ended. Typologists have worked not only with category-like comparative concepts, but also with token-based comparative concepts (e.g. visual stimuli, or even the individual passages of the Pear Story), and it would not make any sense to try to find technical terms for these.
However, we have many other terms in linguistics that we use every day as comparative concepts, even if we only work on a single language. These are terms like plural, 1st person, incorporation, applicative, infix, ergative, participle, associated motion (as well as those mentioned above, except second declension and perhaps aorist).
For example, at the ALT conference next week, there will be many conversations of the following type, among linguists who primarily work on one language (it seems that about one third of ALT abstracts concentrate on a single language):
– (Talking about benefactive applicatives:) Does your language have an instrumental applicative as well?
– My language has two rather different associated motion constructions.
– These two languages express ditransitives quite differently, even though they are spoken next to each other.
– This construction is phrasal, not a complex word, because a clitic can come between the two elements.
In such contexts, these terms are used as comparative concepts: They are understood to have a meaning that goes beyond the structures of individual languages (otherwise these linguists would not go to an ALT conference).
And of course, it would be completely impractical to specify each time when using a term in what sense exactly one uses the term. These are basic-level terms of our everyday lives as linguists, and we treat them much like the symbols of IPA. We expect that textbooks explain what they mean, and we expect that our interlocutors understand them when we use them.
But unfortunately, in a significant number of cases, communication fails – because our textbooks do not explain the terms properly. (I spent a whole chapter in “Understanding morphology” trying to explain the difference between clitics and affixes, but failed to provide proper definitions.)
So for these kinds of terms, one may well discuss whether it would be desirable to have some mechanism for uniform term usage – just as the IPA would have to be invented if it hadn’t been invented already.
This does not mean that everyone would have to use the terms in the agreed-upon sense – after all, many linguists deviate from the IPA in some cases (especially when they use practical orthographies, but also in other cases when there are good reasons). Nevertheless, it is very practical to have a default understanding of a term, so that when one has no reason to deviate from standard usage, one can trust that one’s colleagues understand what is meant.
It seems that the GOLD ontology was created with similar intentions, as well as some other concept registries (which did not have any impact beyond computational linguistics so far).
Reference
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in cross-linguistic studies. Language 86(3). 663-687.
A question: if I understand correctly, phonemes are not comparative concepts, because each phoneme is a language-internal grouping of individual phones. It seems to me that you argue that it is not appropriate to compare such language internal groupings. Does that mean that in historical linguistics, when researchers try to find regular sound correspondences across languages by comparing the phonemic form of words in the languages, (which seems to be the standard method of looking for sound correspondences in historical linguistics), any resulting research is not reliable? If so, that’s a major problem for much research in historical linguistics. Would it be necessary to only compare *phones* across languages instead?
This is an excellent question, thanks! I have not thought about it sufficiently, but I would think that by the logic that I have been pursuing, the answer must be yes: We should not compare phonemes, but phones. I know there was a lot of diachronic phonological research in the 1950s and 1960s (and also partly earlier, and of course also later) that tried to get some mileage out of abstract “phonemes”. Perhaps Martinet’s “Économie des changements” is the best-known representatives of this stream of work. It is my impression that most of what is value in this work can be reinterpreted in “phone” terms. But as I said, I haven’t thought about it very deeply.
Thanks so much for responding! I’ve been really interested in understanding your ideas about “descriptive concepts” and “comparative concepts”, so I’m really glad you’re still looking at comments on these older posts. 🙂
As far as reinterpreting phonemes in “phone” terms: would that involve coming up with some definition of a concept similar to the phoneme (as in, something which allows you to combine individual phones within a language into a “meaningful” grouping) which can nevertheless be used as a comparative concept by having a clear operationalization?
For example, “a Bla is a group of segments all of which exist within a language, and which native speakers of that language cannot consistently distinguish in a laboratory setting without prior training”. (This definition assumes that we have some way of defining “segments” across languages.) So for example, in English, aspirated and unaspirated pairs of voiceless consonants would each form a Bla*. In contrast, in Mandarin Chinese, each such pairs constitutes two separate Blas, because Mandarin Chinese speakers can consistently distinguish them. I think that this idea is at least somewhat useful, because a lot of times, phonologists argue about whether two segments belong to one phoneme or not, and their arguments are generally based on the distribution of the sounds and are generally dependent on abstract analyses. For example, there is controversy on whether in Russian, [ɨ] is merely an allophone of /i/, or a phoneme in its own right, and it is difficult to decide which analysis is “better”; in contrast, with my definition, you could say that [ɨ] and [i] clearly belong to separate Blas, because Russian speakers distinguish them very clearly, and the inability to distinguish them is one of the most obvious markers of a non-native speaker.
*For me this is a conjecture, because I can’t think of any studies off the top of my head that show English speakers generally can’t hear the difference between aspirated and unaspirated consonants, but I have a lot of anecdotal experience which suggests this to be true. The point is that this could be tested empirically be studying the ability of English speakers to distinguish aspirated/unaspirated pairs
This proposal makes a lot of sense to me. My questions are: what form should such a beast take, and how could we go about creating it?
I discussed the GOLD ontology in some detail in my 2011 paper on “Ontologies in Language Documentation” (Language Documentation and Description 9) and argued that what field linguists and typologists need is not an ontology in the strict (GOLD) sense of the term, but something more resembling a *thesaurus. An ontology is a tool for knowledge *representation* and has to conform to certain very strict criteria (Aristotelian categories, strict hierarchical subclass relationships, etc.) in order to be amenable to software agents that perform inferencing and the like. While this may be useful in the field of computational linguistics, it is not suitable for language documentation.
On the other hand, the GOLD ontology could serve as a useful starting point for a knowledge *organisation* tool, such as a thesaurus. I would be interested in collaborating on such a “glottosaurus”!
https://www.academia.edu/3792041/
One problem is that constructions, unlike sounds, are bipolar. An IPA symbol represents a speech sound (or a complex articulatory gesture), and that’s that. But a construction in a language is a bipolar sign, which can be categorized according to function, structure, or both. But comparative concepts may be comparing quite different things. TAM categories like PERFECT are semantic, hence functional; “the perfect” can be expressed in different languages by very different structures. Syntactic categories are different – an Applicative in a given language can be defined purely structurally, and so can the comparative concept. So we can’t get by with one unitary set of definitions analogous to the IPA, we’ll need at least two.
As for MORPHOLOGY cf. the work of Igor Mel’čuk, and for LEXICAL SEMANTICS – the Moscow Semantic School /Jurij Apresjan et al./.
This reminds me of a parallel suggestion, from Anna Wierzbicka (2013:2) “The NSM [Natural Semantic Metalanguage] table of universal semantic primes… provides a semantic analogue of the International Phonetic Alphabet”. Anna went on to claim that “the key difference is that the IPA was invented whereas NSM was gradually discovered”.
I took issue with this (2014:99), and responded that “the articulatory distinctions encoded in the IPA are precisely ‘the result of empirical cross-linguistic investigations’. The IPA justifies each distinctive symbol by a phonological opposition known to operate in some natural language, and characterises it in well-founded anatomical terms.”
The disagreement here can be related to the observation of Matthew Dryer’s (LINGTYP Mon Dec 4 21:48:04 EST 2017) that you (Martin) have responded to
“since there is no set of natural kinds that comparative concepts correspond to, there is no reason to expect different linguists to use the same comparative concepts”.
So, are the units of the IPA natural kinds? and are semantic primitives? Well, these, and morphosyntactic concepts too, do seem to be ‘kinds’ involved in human cognition and its embodiment.
Bob Ladd has argued forcefully that it is a big mistake to regard the IPA categories as natural kinds. But at the same time, they are of course not completely arbitrary, because the sounds of languages do cluster in such a way that one can say that the clusters are discoveries. But the IPA categories are not defined in terms of these clusters, so in this sense they are inventions – I thus think that both Wierzbicka and Nash are right here. (Likewise, I think Östen Dahl is right when he emphasizes that tense-aspect gram-types like “perfect” are discoveries; but again, the comparative concept “perfect” is not defined in terms of the clusters that Dahl (1985) discovered, but semantically.)
However, the way I understand NSM is that it is indeed claimed that the NSM primitives are natural kinds, just as Chomsky & Halle (1968) claimed that their phonological features were natural kinds (innate categories of UG). I think that in phonology, the evidence does not favour the natural-kind view (simply too much diversity, as seen in Phoible.org, and great studies such as Mielke’s book), but I’m not sure about semantics.
The analogy between phonetic IPA and grammatical construction IPA falters because grammatical constructions have form and meaning, but the acoustic
signals represented by IPA do not.
Note also that there is an IPA for sign language (“the Hamburg system”) but it is **not** commonly used, hardly even used at all.
But as David Nash commented, there could be an “IPA” for meanings as well.