Polysemy and metaphor in perception verbs: a cross-linguistic study

light refers directly to Mary’s coat and anaphorically to Sue’s.
(3)
Mary is wearing a light coat; so is Sue. (Cruse 1986: 62) 
The adjective light can have two interpretations ‘not heavy’ or ‘not dark’; in (3) 
it must have just one of them, in both of the instances where it is implied, directly or 
anaphorically. 
These tests seem to work well in some ambiguous cases, but unfortunately, there 
are too many counterexamples for us to be able to take these tests as a definitive way of 
resolving ambiguity (Zwicky and Sadock 1975).
Another important distinction between polysemy and homonymy is the syntactic 
behaviour of the linguistic form in question. Traditionally, homonymous words are 
thought to be represented by two different syntactic categories.
As a way around these insufficient criteria, Lyons proposes the possibility of 
circumventing the problem: the maximisation of either homonymy or polysemy; that is 
to say, either to associate a separate lexeme with every distinct meaning or group every 
distinct meaning under the same lexeme. The maximisation of homonymy is the 
alternative chosen by Kempson (1977), who proposes a ‘constant semantic value’ (1977: 
82) for each lexical item in a language. If a lexical item has more than one sense, it is 
characterised separately without taking into account the relation of this sense with the 
other meanings conveyed by that lexeme. The other possibility, the maximisation of 
26
A first version of the do so test was first discussed by Lakoff (1970). 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
32
polysemy, is taken by Cognitive Linguistics, although it is restricted as they only 
consider the senses associated with a polysemous word with the same syntactic category. 
Neither of these possibilities offers a complete solution, although for methodological 
reasons the latter is preferred to avoid an infinite number of dictionary entries.
One of the reasons that Lyons suggests for rejecting the maximised homonymy 
alternative is precisely the fact that distinctions of sense can be multiplied indefinitely as 
in examples such as mouth of the river, mouth of the tunnel, mouth of, etc. Taylor (1995: 
105) also rejects this possibility on the basis of a prototype categorisation theory. I argue 
in Chapter 7 that it is very important to bear in mind what it is that really causes the 
multiplicity of meaning: is it the possible ambiguous lexeme (mouth) or the lexeme in 
conjunction with other arguments or modifiers (of the river, of the tunnel)? 
In view of these unsuccessful possibilities, Lyons attempts to make the 
distinction between polysemy and homonymy more precise by proposing three 
necessary conditions for absolute homonymy (L
i
and L
j
are two lexical items): 
(i) L
i
≠ L
j
(Lexemic distinctness: the fact that there are two different senses involved). 
(ii) L
i
* = L
j
* (Formal identity, which subdivides homonymy into partial homonymy cases: 
homophony (two word-tokens are formally identical in the phonic medium if they have the same 
phonological representation) and homography (two word-tokens are formally identical in the graphic 
medium if they have the same orthographic representation)). 
(iii) (x 
∈ L
i
* & y 
∈ L
j
* & x = y & R (x, w
i
) & R (y, w
j
) Æ (w
i
≡ w
j
). Where R symbolises the 
relationship of realisation that holds between a form and a morphosyntactic word. (Grammatical 
equivalence: the fact that both senses belong to the same morphosyntactic category). 
Lyons’ main aim in trying to define absolute homonymy is to show that 
polysemy
27
and absolute homonymy tend to produce purely lexical ambiguities, whereas