Note to readers: Portions of this post have been revised and published in the following paper:

Dingemanse, Mark. 2019. “Ideophone” as a comparative concept. In Akita, Kimi & Pardeshi, Prashant (eds.), Ideophones, Mimetics, Expressives, 13–33. Amsterdam: John Benjamins. (doi:10.1075/ill.16.02din) (PDF)

Recently I’ve been having a conversation with Roger Blench about whether structural markedness should play a role in the definition of a useful cross-linguistic conception of ideophones. Over the last few years, Roger has been producing a steady stream of exciting new data on ideophones, often straight from the field (e.g. handouts and drafts on Nyinkyob ideophones, Ngiemboon ideophones, Kolokuma Ịjọ ideophones, and Bafut ideophones). His most recent position is staked out in a paper on Mwaghavul expressives on his website. Here is a key quote:

“Ideophones not only fall into different word classes, but also into a range of conceptual classes. They may demonstrate a characteristic phonology, morphology or canonical form, but this is absent in some languages, even where the ideas they express are conserved. To characterise this richness, it is helpful to switch to a larger class of ‘expressives’ (a characteristic Asian terminology) to encompass these ideas; ideophones would just be a subset. [emphasis MD]

In this reply (which is an edited version of a document shared with Roger some weeks ago) I raise two problems with this proposal. The first problem is one of definitional criteria: how far can we dilute before we lose substance? The second is one of descriptive choices: does the Mwaghavul case really warrant changing our conception of ideophones, or are there other ways to handle it?

Can we do without ‘structural markedness’?

The problem that prompts Blench’s proposal is that the structural markedness of ideophones appears to be absent in some languages, or in some ideophones in some languages, or specifically in a particular class of ideophone-like words in Mwaghavul. Blench proposes to solve this problem by relaxing the “markedness” criterion and going for a meaning-based approach: having an ideophone-like meaning (i.e. a meaning in the semantic domain in which ideophones are normally found) would be enough to be included in the class of “expressives” (I return to the terminology below).

A first problem that I can see with this is that if we were to drop form-based criteria entirely, this leaves us with only a semantic notion, and a very general one at that. Perhaps “Words with typical ideophone meanings.” But how do we identify the typical ideophones in the first place? Or maybe “Words with meanings in the domain of sensory imagery.” But what is to stop us from including every word that happens to have a meaning in the domain of sensory experience (broadly construed)? Of course this is not what Blench proposes; his main worry revolves around a class of “body insultatives” in Mwaghavul — words which are semantically very similar to words that are formally ideophones in related languages, but which in Mwaghavul do not appear to be particularly marked structurally.

Is Blench’s proposal the right solution? In the Mwaghavul paper, Blench justly raises a problem for definitions that identify ideophones solely on the basis of form: “identifying expressives by shape clearly does not work for many languages” (Mwaghavul expressives p. 3). His own proposal, however, seems vulnerable to exactly the inverse problem: it seems equally clear that identifying expressives solely on the basis of meaning does not work for many languages. There is something about these words that makes us feel they are different from run-of-the-mill words, and this something lies not just in meaning, not just in shape or form, but in some kind of combination of semantic/functional and structural/grammatical features. I have discussed this general problem in chapter 2 of my thesis (Dingemanse 2011a:23–8), where I proposed to define ideophones as “marked words that depict sensory imagery“. This definition of ideophones employs a combination of structural (marked), semiotic (depict) and semantic characteristics (sensory imagery), precisely because (in my view) the nature of these words cannot be attributed to any of those domains separately.

Clarifying the notion of markedness

The criterion of structural markedness1 is well-grounded cross-linguistically: the literature abounds with qualifications like “very striking” (Vidal 1852:15 on Yoruba), “phonologically peculiar” (Newman 1968:107 on Hausa and Tera, Chadic), “structurally marked” (Klamer 2002:263 on Kambera, Malayo-Polynesian). In language after language, they are described as “words of unusual phonological shape” (Childs 1988:27 on Kisi, Mel); “their phonological structure clearly differs” (Frajzyngier 1989:196 on Pero, West-Chadic); they are “distinguished by their aberrant phonology” (Kruspe 2004:102 on Semelai, Mon-Khmer); and they “stand out by their peculiar phonetic makeup” (Diffloth 1980 on Sre, Mon-Khmer).

Two clarifications about markedness are in order. First, exactly what it means to be marked in a particular language is something that belongs to the description of that language. What is marked in one system may not be marked in another. For instance, in Hausa (Newman 1968), words do not normally end in a consonant, but ideophones commonly do: this is one sense in which they are marked. In Filomeno Mata Totonac (FMT, McFarland 2009), words often end in single consonants, but ideophones one-up this by featuring final consonant clusters. So Hausa ideophones would be pretty ordinary words in FMT (at least as far as syllable structure goes); but within the Hausa system, they are marked.

Second, that the lexical class of ideophones on the whole can be characterised as “marked” does not imply that we should expect every single ideophone in the system to be marked in the same way or to the same extent. The way word classes work in natural language is not by deterministic form-to-category mappings; these are stochastic processes in the end, and a word that on its own may not be (as) marked as expected may still be considered a perfectly fine member of a given class on the basis of its patterning with like words. Good reasons to consider a word a member of a given class are for instance its appearance in the same syntactic slots and the same semantic fields. This implies that not every single ideophone has to be marked in the same way or to the same extent. This echoes an observation by Gérard Diffloth on expressive phonology and prosaic phonology: “we seem to have two distinct but overlapping phonological systems” (Diffloth 1980:50).

In other words, while I do insist we need a combination of structural, semantic and semiotic criteria to define ideophones, I don’t insist on discrete boundaries. Some may counter this is fuzzy. I would agree, and I would point out that language by its very nature is fuzzy in this way because it’s a culturally transmitted system that is ever evolving. The proper response to this kind of fuzzyness is not to dilute the category to the point where it has no empirical bite anymore; nor is it to limit one’s view to the clear-cut cases only. Instead, the proper response is to define the core as sensibly as possibly, to capture its center of gravity, while aiming to precisely describe the types of ‘fuzzyness’ in the periphery. Insofar as Roger is doing just that by putting out the Mwaghavul data, we are in full agreement.

So here then is a first possible solution for the Mwaghavul data: call them ideophones if you want, but qualify their classification in the language by pointing out in which respects they are similar to ideophones (meaning, presumely), and how they are different from them (form, apparently). One reason I have proposed a cross-linguistic definition of ideophones is to allow precisely such qualification. Language-particular discussions can be made more precise if a cross-linguistic reference point is used. The cross-linguistic reference point should not be mistaken for the language-particular category. Nor is it a checklist of necessary and sufficient criteria: it rather characterises a prototype (Childs 1994), or defines a canonical type in the fashion of Corbett’s canonical typology.

An Aslian parallel

Blench’s problem with the “markedness” of ideophones seems to stem from the fact that Mwaghavul appears to have some classes of words which structurally appear to be nothing special, but conceptually (or etymologically) are similar to classes of words that in other (related?) languages have been identified as ideophones on structural+semantic grounds. If that is the case, that’s interesting, but not unheard of, and not enough reason in my view to abandon the relatively clear notion of “ideophones” for a less clear-cut and less powerful semantically-based notion. (And note by the way that although Blench mentions the South-East Asian literature in relation to this choice of terminology, his move finds no parallel there — Gérard Diffloth’s conception of “expressives” is very much a combination of structural and semantic criteria, and Nicole Kruspe in her excellent grammar specifically mentions that she does not find a reason to distinguish between expressives and ideophones.)

For a case that could be parallel to Blench’s Mwaghavul data, let’s look at two Aslian languages. Aslian is well-known for its “expressives” (Diffloth 1972; Diffloth 1976; Diffloth 1980; Kruspe 2004; Tufvesson 2011). Semai (Central-Aslian) has an impressive number of smell “expressives” (Diffloth 1976; Tufvesson 2011). Like some of Blench’s Mwaghavul words, these expressives often have transparent etymologies and can be traced back even to the common ancestor of Khmuic and Aslian (as cognate forms occur in Kammu). In Semai, these words are considered “expressives” on structural grounds, displaying special sound patterns and the gradience that is typical of depictive vocabulary (Diffloth 1976, Tufvesson 2011).

The Northern-Aslian language Jahai has a corresponding set of smell terms, cognate in many cases. However, unlike in Semai, these are not expressives (Burenhult and Majid 2011:25–6). Burenhult & Majid present some clear structural arguments for the non-expressive nature of these words in Jahai: they are “analyzed on syntactic grounds as stative verbs”, and as such, they can be “can be negated, relativized, and nominalized”: all things we don’t see happening regularly to expressives or ideophones.

Now, if Burenhult & Majid were to follow Blench’s line of argument, it would appear that they should have lumped the Jahai terms together with the Semai ones under a semantically-based notion of “expressives”. After all, the meanings are roughly in the same domain. Yet they didn’t, and we should be glad for that: we would have missed an interesting case of historical development in two different directions: the expressive/depictive direction in Semai and the stative verb direction in Jahai. By the way, it seems, from Diffloth’s comparative work, that the expressive/depictive state may be the ancestral state so that the simplest explanation would be that Jahai represents a change in the stative verb direction.

This, then, would be a second possible solution for the Mwaghavul case: don’t treat the words as ideophones; state the grounds for this decision by pointing out the similarities and the differences relative to to “typical” ideophones in the language and cross-linguistically. If possible, bring comparative and historical data to bear on the question.

Final thoughts

I have laid out two possible solutions here. Which one is the most optimal is not for me to decide (more data would be needed), but either would be an improvement, I think, over Blench’s initial proposal, which was to broaden (or dilute) the class of ideophones to a meaning-based category. I have argued why I don’t think a semantically based notion, broader than ideophones and less structurally grounded, is going to hold much water. It would lose much of its power because it wouldn’t uniquely pick out ideophones or ideophone-like words, and it would fail to capture one of the most salient characteristics of these words cross-linguistically. It also would run the risk of obscuring interesting variation, as shown by the Aslian case. Yet such variation is precisely what can help us to better understand ideophone typology.

Most importantly, what thorny cases like the Mwaghavul data show is that classifying words as ideophones or non-ideophones is not the most important task at hand. What we actually need is more fine-grained descriptions (preferably in quantifiable measures) of all the features and variables that make us think words are ideophonic or ordinary. The characterisation of ideophones as “marked words that depict sensory imagery” is only the beginning. More cross-linguistic data will help us to refine the notion and to spell the different ways in which words can be marked, depictive, and evoking sensory imagery.

References cited

1. Blench, Roger. 2010. ‘The Sensory World: Ideophones in Africa and Elsewhere’. In Perception of the Invisible: Religion, Historical Semantics and the Role of Perceptive Verbs, ed. Anne Storch, 275–296. Sprache Und Geschichte in Afrika 21. Cologne: Köppe.
2. Blench, Roger. 2011. ‘Mwaghavul Expressives’ (November 7).
3. Burenhult, Niclas, and Asifa Majid. 2011. ‘Olfaction in Aslian Ideology and Language’. The Senses and Society 6 (1): 19–29.
4. Diffloth, Gérard. 1972. ‘Notes on Expressive Meaning’. Chicago Linguistic Society 8: 440–447.
5. ———. 1976. ‘Expressives in Semai’. Oceanic Linguistics Special Publications (13): 249–264.
6. ———. 1980. ‘Expressive Phonology and Prosaic Phonology in Mon-Khmer’. In Studies in Mon-Khmer and Thai Phonology and Phonetics in Honor of E. Henderson, ed. Theraphan L. Thongkum, 49–59. Bangkok: Chulalongkorn University Press.
7. Dingemanse, Mark. 2011. ‘The Meaning and Use of Ideophones in Siwu’. PhD dissertation, Nijmegen: Radboud University. http://thesis.ideophone.org/.
8. Haspelmath, Martin. 2006. ‘Against Markedness (and What to Replace It With)’. Journal of Linguistics 42 (01): 25–70. doi:10.1017/S0022226705003683.
9. Kruspe, Nicole. 2004. A Grammar of Semelai. Cambridge: Cambridge University Press.
10. Lonfo, Etienne, Marieke Martin, and Roger Blench. 2010. ‘Ngiemboon Ideophones’.
11. McFarland, Teresa. 2010. ‘Ideophones and Templatic Morphology in Totonac’. In Rara & Rarissima: Documenting the Fringes of Linguistic Diversity, ed. Jan Wohlgemuth, 235–246. Empirical Approaches to Language Typology, 46. Berlin, Germany: Mouton de Gruyter. xiii.
12. Newman, Paul. 1968. ‘Ideophones from a Syntactic Point of View’. Journal of West African Languages 5: 107–117.
13. Tufvesson, Sylvia. 2011. ‘Analogy-making in the Semai Sensory World’. The Senses and Society 6 (1): 86–95. doi:10.2752/174589311X12893982233876.

1. On a sidenote, in relation to the term “marked”, many people will think of a stimulating paper by Martin Haspelmath titled “Against markedness (and what to replace it with)”. The sense in which I am using ‘marked’ is closest to Haspelmath’s (2006) sense 3 (“overt coding” in the sense of “signaled”, “coded”, “indicated”). While Haspelmath recommends doing away with the term “marked” altogether, the alternative he suggests for this particular sense, “overt coding”, could be taken to imply a particular “marker”, prototypically a piece of morphology, which is not necessarily the case for ideophones. Hence, I stick to my qualified use of the term “marked” in the sense of “signaled”, “indicated”. [↩]