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

partial homonymy tends to produce ambiguity only in certain contexts. 
These traditional approaches to polysemy provide a more or less successful 
descriptive analysis of what polysemy and homonymy are; what lexical items are 
homonymous or polysemous. Their major problem, however, is that they fail to address 
several fundamental issues: the reasons why these lexical items have several senses 
attached to them in the first place; how these meanings are structured: are these senses 
grouped under the same lexical item by chance or is there any motivation for the lexical 
27
Lyons defines polysemy as “the product of metaphorical creativity” (1977: 567). It will be 
argued later in the analysis in Chapter 7 that polysemy is not only produced by means of metaphor, but 
also by virtue of combinations between the different elements in the sentence in question. 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
33
item to convey specific meanings? Is the semantic content of a single lexical item 
enough to create polysemy or, on the contrary, is the interaction with the semantic 
content of the other lexical items that co-occur in the same sentence necessary? 
These issues, neglected by traditional approaches, are at the core of investigation 
in Cognitive Semantics. In the following section, I present the explanations that this 
model provides for these questions. 
1.2.2. Cognitive Semantics 
Within this framework, the main distinction between polysemy and homonymy 
is the systematic relation of meanings that takes place in polysemy (Lakoff 1987: 316; 
Johnson 1987: 193). When speaking about polysemy, the fact that we are dealing with 
multiple meanings is not the main point but the fact that those multiple meanings are 
related in a systematic and natural way.
According to Lakoff (1987), polysemy has to be understood as categorisation, 
that is to say the idea that related meanings of words form categories and that these 
meanings bear family resemblance, an idea introduced by Austin (1961). Taylor (1995: 
108) explains this family resemblance category in terms of ‘meaning chains’. A lexeme 
can convey different meanings, A, B, C, D, … A is related to B in virtue of some shared 
attribute(s) or other kind of similarity. Meaning B in turn becomes the source of a further 
extension to meaning C and so on. This ‘meaning chain’ can be represented in (4), 
where any node in a meaning chain can be the source of any number of meaning 
expressions: 
(4)
A Æ B Æ C Æ D… 
Taylor compares these ‘meaning chains’ to Lakoff’s ‘radial categories’. A 
category is structured radially with respect to a number of subcategories: there is a 
central subcategory, defined by a cluster of covering cognitive models and in addition, 
there are noncentral extensions which are not specialised instances of the central 
subcategory, but variants of it. The extensions of the central model are not random, but 
motivated by the central model plus certain general principles of extension. One of the 
advantages of this approach if compared with classical models is that it offers adequate 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
34
means of characterising the situations where one or more senses are central or more 
representative. 
ICMs are complex structured wholes or gestalts. They do not necessarily fit the 
world very precisely. There will always be some segments of society that the ICM fits 
reasonably well and some others that it will not.
Polysemy is therefore the result of the extension of ICMs to form radial 
categories
28
. Sometimes, a single ICM can be the basis for a collection of senses that 
form a single natural category. For instance, the ICM of the lexeme window
29
can take 
three meanings: ‘an opening in the wall’, ‘a frame fitting into the wall’ and ‘the glass 
filling the frame fitting into the wall’. These three senses are not unrelated; they form a 
natural category of senses, where correspondences remain physical. These 
correspondences have been explained in terms of ‘image schemata’; i.e. recurring 
structures of, or in, our perceptual interactions, bodily experiences and cognitive 
operations (Johnson 1987: 79, see Section 1.1.2 above). 
In some other cases, these correspondences do not take place within the same 
ICM, but between the ICMs of two domains. Lakoff and Johnson (1980) propose 
‘conceptual metaphor’ as one of the means for relating the different senses of a word. 
Metaphor is understood as an experientially-based mapping from an ICM in one domain 
to an ICM in another domain
30
. 
This model of polysemy has been explored mainly with respect to prepositions, 
where analyses such as Brugman (1981, 1988), Lindner (1982), Herskovits (1987) and 
Vandeloise (1991) have shown the regularities and motivation among the different 
senses that prepositions can convey. But as we will see in Chapter 4, it has also been 
applied to the study of polysemy and semantic change in other semantic fields such as 
perception verbs. Sweetser (1990) identifies this metaphorical mapping of two different 
28
Although I focus only on the radial model in this section this does not mean that Cognitive 
Semantics rejects the totality of other approaches such as inheritance models. These are accepted as long 
as they fit the facts (see Langacker 1991b: Ch. 1). For instance, in the case of the chain 
VEHICLE
-
CAR
-
SPROTS CAR
, the fact that 
CAR
inherits from 
VEHICLE
the attribute ‘means of transportation’, and 
SPORTS 
CAR
inherits all the features of 
CAR
does not create any problem for Cognitive Semantics (I owe this 
insight to Barcelona p.c.) 
29
As will be seen later, Pustejovsky (1995) calls this type of polysemy ‘complementary 
polysemy’, i.e. where the alternative readings are manifestations of the same core sense in different 
contexts. 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
35
domains in the semantic development of perception verbs. The mappings between the 
physical and the mental domains are viewed as an important influence in the historical 
development of polysemy and of cognate words in related languages. 
The way in which Cognitive Semantics tackles polysemy provides us with a 
framework that explains and shows that meanings are not grouped together under the 
same lexical item by chance. There is bodily-based motivation that causes and organises 
radial categories of meanings. These radial categories are structured by means of 
metaphor and metonymy. In this thesis I will use this model to explain why perception 
verbs have the polysemous senses that they seem to convey. 
This framework offers a good model for the explanation of why polysemous 
senses are grouped together under the same lexical item. Its major drawback, however, 
is that it does not seem to focus on how these polysemous senses are created. In other 
words, what it takes to create polysemy, the semantic content of just one lexical item, or 
the semantic content of that lexical item in conjunction with the semantic content of 
other lexical items. As stated in the introduction to this chapter, this point is central to 
my discussion in this thesis
31
. Cognitive Semantics does not provide a sound answer, 
and therefore, we need to find a different model that does. In the following section I 
sketch the main tenets of an approach that focuses precisely on this last issue: 
Pustejovsky’s Generative Lexicon.