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

1.2.1. 
 
TRADITIONAL APPROACHES TO POLYSEMY: WHAT IS 
POLYSEMY? WHAT IS HOMONYMY? 
The traditional distinction between polysemy and homonymy is based on 
whether there is one or two lexical items involved. Lyons (1977: 550) considers them as 
two types of lexical ambiguity
18
and introduces some criteria for deciding when it is 
polysemy and when it is homonymy.
One criterion is etymological information about the lexical item in question. 
Lexical items with the same origin are considered as polysemic, whereas if they have 
evolved from distinct lexemes in some earlier stage of the language then they are 
regarded as homonymous. This condition is neither satisfactory nor decisive because the 
history of the language does not always reflect its present state. For instance, in present-
16
For an account of the changing patterns in the study of polysemy from antiquity to the 20
th
century, see Nerlich and Clarke (1997). 
17
Authors such as Taylor (1995), differentiate between homonymy and monosemy, where the 
former is only restricted to those cases when unrelated meanings are attached to the same phonological 
form and the latter when the lexical item has a single sense. As will be seen later, Lyons (1977) includes 
under homonymy both cases, although he does differentiate between partial homonymy, i.e homography
and homophony, and absolute homonymy. 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
28
day English, the lexemes pupil
1
‘student’ and pupil
2
‘iris of the eye’ are not usually 
related by native speakers, but they are both derived from Latin pupillus/pupilla ‘ward, 
orphan-boy’ which is itself a diminutive of pupus ‘child’
19
. The opposite case is also 
fairly common, namely when native speakers consider two lexemes derived from 
different roots in an earlier stage of the language as related. For example, the lexemes 
ear
1
‘organ of hearing’ and ear
2
‘spike of corn’ come from two different origins: ear
1
evolves from OE êare from IE *aus- (cf. Latin auris ‘ear’) and ear
2
from OE êar (cf. 
Latin acus, aceris ‘husk’) and they merged into er(e)  in ME. However, most people 
nowadays treat these two lexemes as one polysemous word and explain their relation by 
means of metaphor. Therefore, the etymological criterion can be very misleading
20
when 
deciding between homonymy and polysemy. 
Another criterion is the unrelatedness vs. relatedness of meaning; i.e. the native 
speaker’s feeling that certain meanings are connected and that others are not. One of the 
major drawbacks that Lyons states for this criterion is that relatedness of meaning 
appears to be a matter of degree, together with the fact that sometimes native speaker’s 
intuitions
21
are far from being the true interpretation, as has been seen with the ear 
example above. Attempts to formalise this relatedness of meaning have also been made. 
Katz (1972), Katz and Fodor’s (1963) Componential Analysis proposes the 
decomposition or breakdown of the sense of a word into its minimal distinctive features, 
i.e. into semantic components which contrast with other components. These minimal 
distinctive features produce formulae called componential definitions of the type [
± 
human], [
± adult], [± male] for the description of lexemes such as man, woman, girl, boy 
in the semantic field of ‘human race’ (see Leech 1981: 96ff.).
Unfortunately, this type of approach is not sufficient for the polysemy-
homonymy problem. First, the relatedness in the different sense of a word might not be 
18
Ambiguity itself is a complicated term as well. Tuggy (1993) sees ambiguity as related to 
homonymy and polysemy more related to vagueness.
19
This example is very interesting because in Spanish the word niña also covers both meanings 
‘young girl’ and ‘eye’s pupil’. 
20
Lyons (1977: 551) further states that the etymological criterion should not take part in the 
definition of homonymy. First, speakers are not aware of such historical developments. Second, this 
information is irrelevant for synchronic analysis of languages. 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
29
expressible in terms of 
± features and also because in some cases, these features are 
present in different degrees, not in absolute terms. A classical example of this problem is 
the word bachelor (Fillmore 1977, 1982). In a simplified world, where people are 
marriageable at a certain age, mostly marry at that age and stay married to the same 
person, bachelor is just any unmarried male past marriageable age. However, outside 
this simplified world, the word bachelor does not apply. That is why we find it so odd to 
call the Pope or a twice-married divorcé bachelor, even though they both meet the 
criteria of the definition given above
22
. Secondly, as Lyons (1977: 553) points out, “the 
possibility or impossibility of decomposing the senses of lexemes into a (structured or 
unstructured) set of semantic components is irrelevant, unless we can specify just how 
many components, or alternatively what kind of components, two senses must share in 
order for them to meet the criterion of relatedness of meaning”.
An alternative solution for both problems is presented in my Property Selection 
Processes in Chapter 6. I propose that, by means of different properties present in the 
prototypical meaning of the lexeme, it is possible to describe such lexemes and then see 
how only some of those prototypical properties are present in the potential polysemous 
senses that may belong to such a lexeme. This would be a possible formal explanation 
for such relatedness of meaning among possible polysemous words. 
A third way of attempting to establish polysemy is to search for a central or core 
meaning. Based on the classical definition of a category as a set of necessary and 
sufficient conditions for membership, Allerton (1979) proposes that when different 
senses of a lexeme share a core meaning, they are polysemous. On the other hand, cases 
when the core meaning cannot be extracted are to be considered as homonymous. For 
instance, the word paper can mean ‘newspaper’, ‘document’ and ‘academic lecture’; all 
these senses share the core meaning of ‘important written or printed material’. 
According to Palmer (1981: 105), this is possible when we have cases of metaphors and 
the other senses have been transferred from that core meaning. The disadvantage of this 
21
In Leech’s (1981: 229) view, the native speaker’s intuitions are valid as long as the speaker 
expresses the relation between meanings in terms of lexical rules. These lexical rules have psychological 
reality to the extent that they are part of the native speaker’s linguistic competence. 
22
Similar examples are discussed in Lakoff (1987) (mother), Coleman and Kay (1981) and 
Sweetser (1987) (lie). 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
30
criterion is again to decide what the core meaning is. As will be seen later in the 
discussion, under the Cognitive Linguistics approach, neither the core meaning approach 
nor Palmer’s acceptance of it in metaphorical cases is accepted. The reason is the fact 
that metaphor is understood as a motivated transfer between two different domains and 
this core meaning approach totally defeats any attempt to show a motivated account of 
semantic extension. The alternative to this approach within Cognitive Linguistics is the 
‘family resemblance model’ (Taylor 1995: 106) or what Lakoff (1987: Ch. 6, 1996: Ch. 
1) calls ‘radial categories’.
Finally, there have been attempts to test the ambiguity
23
of lexical items. Cruse 
(1986: 54ff.) makes a distinction between ‘indirect’ and ‘direct’ ambiguity tests. 
‘Indirect’ tests are designed to find two occurrences of a word form with different 
relations of meaning; these relations can be paradigmatic (e.g. synonymy) or paronymic 
(i.e. identity of root but different syntactic category). According to Cruse (1986), these 
tests are invalid since nothing can be reliably inferred from the fact that a word form has 
different meaning relations in different contexts. 
For the ‘direct’ ambiguity tests Cruse (1986) offers three criteria:
(i) Contextual modulation: an ambiguous form should not in every case be totally conditioned by its 
contexts. In a disambiguating context, the word may carry more information that can be accounted for 
in terms of interaction between the context-independent meaning of the word and the semantic 
information of the context itself, whereas in cases of contextual modulation all information is derived 
from the context.
(ii) Independently maximisable separate senses: under certain conditions, the application of certain 
terms must be maximised within the current universe of discourse, even if it creates some oddness
24
. 
(iii) Antagonism of independent senses: cases where senses cannot arise simultaneously without 
causing oddity
25
. This antagonism of senses can be tested by the ‘co-ordination test’; it is argued that 
sentences with an ambiguous lexeme cannot have both meanings at the same time. For instance, in a 
sentence like (1), the word light cannot mean both that the room is bright and that the furniture is not 
heavy (Palmer 1981: 106). 
23
For an extended discussion of ambiguity, see Kempson (1977), Zwicky and Sadock (1975) and 
Cruse (1986). 
24
The two contrastive examples Cruse (1986: 60) gives are: 
(1) ? 
Mary likes mares better than horses 
(2)
John prefers bitches to dogs 
The oddity of (1) and the acceptance of (2) can be explained in terms of prototype theory. Mare 
and horse do not correspond to the same level of categorisation: mare is subordinate level category, 
whereas horse is a basic level category. If instead of horse, it had been stallion, the respective subordinate 
level word, the sentence would have been correct. What happens in (2) is that dog represents both the 
subordinate and the basic level categories, and therefore, the suitable category for this particular case, i.e. 
subordinate level one, has been chosen. 
25
The variety of anomaly brought by this simultaneous link of independent senses is allowed in 
some contexts and is traditionally labelled as ‘zeugma’. 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
31
(1)
The room and the furniture were light. 
Another test is that proposed by Kempson
26
(1977: 129). She argues that to 
distinguish ambiguous sentences, it is necessary to turn to anaphoric processes such as 
the insertion of an expression like to do so too, where the anaphoric expression demands 
identity of meaning of the two verb phrases in question. 
(2)
John went to the bank and Will did so too. 
In this case, bank has to refer to the same entity, either the financial institution or 
the side of the river in both VPs, but not to a different entity in each one. 
A third test is the so-called ‘identity test’. In a sentence like (3), the adjective