Polysemic Perception

The fascination of the day is polysemy in perception verbs. To illustrate what I’m talking about, let’s take the example of Maaka, a Chadic language, from The Grammar of Knowledge: A Cross-Linguistic Typology. The perception verb sòl covers smell, hearing, feeling, and experiencing, depending on the context:


The Europeans hear news.

Experience, Feels

You experience a famine.


He smells the cow.

Polysemous perception verbs are not unusual, but research suggests that there’s quite a strong structure to the divisions between words. The most frequently cited work is a 1983-1984 paper by Viberg which studied the structure of perception words in a sample of around 50 languages. Unfortunately I can’t find a copy, but I have found enough quoted bits to get an idea.

Viberg seems to have established some kind of ranking along the following lines:

see > hear > touch > taste, smell

The hierarchy seems to serve as both a constraint on diachronic directions of development and which meanings can be combined into a single perception word. For example, a sight verb can generalise to hearing and become a see+hear verb, but a hear verb can’t develop to cover sight, and a sight verb can’t jump straight to covering smell as well without going through hear and touch.

I’ve also seen this reproduced as a similar looking diagram, although unfortunately the diagram says something slightly different to the linear scale:

Diachronic development of perception polysemy

This diagram, unlike the linear scale, suggests that vision to touch or taste is directly possible, as in hearing to smelling. Other data seems to support this diagram over the linear scale as a more accurate model (see below).

It also seems true that hearing and sight are statistically different to the rest. Sight especially is separated from other perception in the vast majority of languages. In the Maaka example above, the one sense that sòl can’t cover is sight. Hearing is more likely than sight to be expressed by a polysemous verb, but it’s clearly the second in line to be split off. The other senses are more likely to be merged under a shared verb. Of the languages showing polysemy in Viberg’s sample (presumably in basic terms, since there are always ways to disambiguate):

  • 14 were see / hear / other
  • 7 were see / other
  • 1 was hear / other
  • 1 was see+touch / hear+smell+taste
  • 1 was fully merged (a “perceive” verb)

So in all but 1 of the 53 languages in the sample, seeing and hearing were expressed by different verbs. In 43/53 = 81%, see and hear were expressed by non-polysemous verbs, with any polysemy limited to the other senses.

Taking this a step further, perception verbs may express other distinctions, and the same hierarchy drives where those distinctions appears. Viberg apparently looked at 3 subtypes for each sense:

  • Volitional activity (look at, listen to)
  • Passive experience (see, hear)
  • Impersonal (look, sound)

English distinguishes all three for sight and hearing, either by a separate verb or the combination of verb+preposition, but merges the three into a single verb for the other senses:

Volitional activityPassive experienceImpersonal
I felt it (deliberately)I felt it (accidentally)It felt slimy (impersonal)
I tasted it (deliberately)I tasted it (accidentally)It tasted good (impersonal)
I smelled it (deliberately)I smelled it (accidentally)It smelled good (impersonal)
Lack of distinctions in feel, taste and smell

Although English is also a borderline counterexample to sight having more splits than hearing, since the three hearing verbs are clearly distinct, whereas sight has “see” and “look”, and only differentiates the activity and impersonal senses via an additional preposition.

Interesting, anyway.