Have you ever wondered why "tasty" means that something tastes good while "smelly" means that it doesn't smell so good? What's up with that?
Manfred Krifka, eminent semanticist in Berlin, Germany, has a solution, explained in his article "A Note on an Asymmetry in the Hedonic Implicatures of Olfactory and Gustatory Terms" (2010. In Susanne Fuchs, Philip Hoole, Christine Mooshammer & Marzena Zygis (eds.), Between the regular and the particular in speech and language, 235–245. Frankfurt: Peter Lang), an impressive sounding title for a delightful little paper. It was published last year but Manfred just made it available on his website, hence this post now. Do give his paper a read!
Krifka starts his paper with a famous anecdote about Samuel Johnson:
As Samuel Johnson paused to rest on a London park bench one hot summer's day, his profusely sweating bulk caused a young woman sitting next to him to accuse him of smelling. “No, Madam,” he replied. “You smell, I stink.”
Johnson is pedantically objecting to the intransitive use of "smell" (I don't share that particular peeve) but what's interesting is that in that use, "smell" doesn't simply mean "give off a smell" but suggests that the smell is bad in some way.
What Krifka tries to understand is an asymmetry in the normal interpretation of unspecific smell-predicates versus taste-predicates. When we say that something smells, we are interpreted as saying that it doesn't smell good, whereas if we talk about it being tasty, we are interpreted as saying that it tastes good. Apart from the "tasty-smelly" pair, English doesn't allow the construction of minimal pairs. German is more productive in that respect and thus provides most of the data in Krifka's article.
In the case of the adjectives "smelly – tasty" in English, the negative and positive components seem to be part of the lexical meaning. But for the intransitive verb "smell" in English and the intransitive verbs "riechen" and "schmecken" in German, the evaluative component comes and goes. When left unspecific, they carry the evaluative component, but they are compatible with the opposite specification. "You smell nice" is perfectly fine, as is its German counterpart "Du riechst gut". And while German "das schmeckt" means "that tastes good", it is also OK to say "das schmeckt nicht so gut" = "that doesn't taste so good".
Krifka concludes that the evaluative component of these intransitive verbs is an implicature, a pragmatic enrichment that can be defused in more explicit contexts. But why should "smell" regularly be enriched to refer to bad smells, while "taste" is enriched to refer to yumminess?
Krifka's explanation is this:
The […] explanation that I would like to offer is based on the assumption that the things that we taste and the things that we smell differ systematically in how pleasant and unpleasant they are. The reason is that we have generally more control over what we put into our mouth than what enters our nose. If one guiding principle of our behavior is the maximization of pleasure, and if there are roughly equally many pleasant and unpleasant smells and tastes available, then we should draw more pleasure out of the sense that we can control, than out of the sense that we cannot control as easily. Consequently, what we taste will be more likely pleasant than what we smell. This is the condition that is known to lead to so-called I-implicatures, the enrichment of meanings to the prototypical case (cf. Levinson 2000). There are many examples that are structurally similar. For example, nurse is typically understood as ‘female nurse’ because most nurses are female, but this is not part of the lexical meaning – there is no contradiction in male nurse […]. For another example, He opened the door is understood to mean that he opened the door in a regular way, but not by smashing the door with an axe, even though we can say He opened the door with an axe. I-Implicatures are rooted in the conversational maxim called the I-Principle, which states that speakers should say as little as possible to achieve understanding. The corresponding rule for recipients is to add information that is uncontroversial, conventional and stereotypical.
This has a bit of the smell of a just-so story to me, but it appears that these are not just quirks of German and English. Apart from regaling us with tidbits such as the Mexican language Totonak being reported to have about twenty descriptive smell terms, Krifka claims that the tendency for smell predicates to be biased towards bad smells is well-known (he cites Claude Boisson. 1997. La dénomination des odeurs: variations et régularités linguistiques. Intellectica 29–49). The opposite tendency for taste predicates to be biased towards positive interpretation is discussed for the first time here in Krifka's paper. I wonder whether it holds up cross-linguistically as well.
Is it really true that most of the things we taste, taste good, since we have control over what we taste? Perhaps, but for Krifka's story to work, it also has to be true that most of the things we smell, smell bad. I'm not so sure about that. But maybe that was true during the crucial period where these meanings coalesced: after all, one hears that well into modern times, the world was much stinkier than it is right now right here in the western suburbs of Boston on a gorgeous summer's day.