Comment: Interjections and Expressivity

“Interjections and emotion” raises both general questions about natural semantic metalanguage (NSM), and particular ones about theoretical challenges of expressive vocabulary. Since NSM’s weaknesses as a semantic theory are well known (Geeraerts, 2010; Riemer, 2006; Wawrzyniak, 2010), I will limit myself to the second question.

Conceptualist semantic theories such as NSM risk a kind of explanatory vacuity comparable to the classic postulation of a “dormitive virtue” to account for the sleep-inducing properties of opium: since any word can be defined, the definition’s structure can simply be interpreted as the structure of the “conceptualization” supposedly constituting the word’s meaning. This is too easy: some reason is needed to believe that the alleged meaning-constituting concepts actually exist. To legitimately identify linguistic meanings with concepts, we need a language-independent theory of conceptualization. Without one, concepts are simply assumed to underlie language, without evidence or a clear idea of what they actually are. To avoid this, we have to ask when, in nonlinguistic cognition, it is explanatory to postulate concepts, and then see if these concepts have the properties required to function as meanings. Here is one possible answer (see Riemer, 2013, for details).

As traditionally understood, concepts must be invariant, in essential respects, between individuals (cf. Fodor, 1998, pp. 28–29). If we can show that there are reasons to posit invariant psychological structures across individuals to explain their behaviour, we will have effectively shown that conceptual explanation is justified, and made it plausible that language simply labels preexisting conceptual representations.

It is plausible—though not necessary (see Barsalou, 2008; Prinz, 2002, among others)—to credit agents with invariant concepts to support their interactions with the external environment. People can plausibly be credited with the same invariant concept water since actual water imposes the same behavioural constraints on us all: we drink it, but don’t try to walk on it. To explain these behavioural identities, we postulate an invariant water concept, also seeing this as the meaning of the word water. On the level at which they contribute to action and enter into communication, conceptual structures evoked by environment-referring words arguably must be invariant, because they must all accurately represent the same factual information about their referents.

Hence, as is often acknowledged, public, concrete objects are those for which invariant concepts can be most satisfactorily postulated. By contrast, studies of concepts typically founder on nonpublic objects (Barsalou, 2008; Laurence & Margolis, 1999).

Internal experiences (surprise, disgust, etc.) are precisely those for which the postulation of invariant concepts is least appropriate, in NSM or elsewhere. Here, there is no equivalent to the “accurate representation” necessary for referents in the external environment: there simply are no publicly available, normative factors that could be recorded in a factual and hence invariant concept. The referential conditions that prompt interjections, and explain the regularities of their use, are private, and hence not subject to any normative adjudication. There is no “accounting for taste”: your surprise is my disgust. This self-evident fact removes the main motivation to posit concepts at all—the idea that regularities of reference are explained by invariant psychological structures (concepts).

The private nature of emotions means that we cannot even tell whether different subjects express the same psychological state with a single interjection. We cannot therefore check that G’s (surely questionable) component “this is very good” for wow captures any real psychological regularity between speakers. Asserting it is not enough: evidence is needed. The fact that we may sometimes use an interjection in similar situations is no argument that it is supported by an identical conceptual state: conceptual identity, not use identity, is exactly the question at issue—and there just are no perceptible criteria by which we can verify that the states concerned are invariant from one instance to another.

If concepts must be invariant to play their standard explanatory role, if invariance derives from relations with the external environment, and if there is no reason to think that any invariant relation exists between interjections and the external environment, the conclusion can only be that interjections are not conceptual, but expressive. This certainly does not mean that they cannot be studied; just that regularities about them cannot be captured in the theoretical language of conceptualization.

Correctly differentiating conceptual (descriptive) and expressive content is crucial for semantics. On the present argument, any concept-based approach to immediate uses of interjections is problematized, not just NSM’s. Semanticists have assumed that there are not many expressives in the lexicon. This is probably a mistake. If semantics moved away from conceptualization towards expressivity in its explanations of meaning, it would only be following much of the rest of cognitive science in a “situated” and “embodied” direction.