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

1.2.3. Lexical Semantics and the Generative Lexicon 
The ‘Generative Lexicon’ is Pustejovsky’s (1995) approach to the problem of 
lexical ambiguity, to the multiplicity of word meaning and to the question of how we are 
able to give an infinite number of senses to words using finite means. The main thesis of 
this approach is that a core set of word senses is used to generate a larger set of word 
senses when individual lexical items are combined with others in phrases and clauses. 
This system has four levels (argument structure, event structure, qualia structure and 
lexical inheritance structure) which are connected by generative devices (type coercion, 
30
For a full account of different types of ICMs, see Lakoff (1987: 281ff.). 
31
Chapter 7 is devoted entirely to this issue. 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
36
selective binding and co-composition) that provide the compositional interpretation of 
words in context. 
Pustejovsky argues that former approaches to natural language semantics have 
ignored either the problem of how words are used in novel contexts or the creation of 
such new senses on the basis of compositionality. In language, words can have more 
than one meaning, but the means in which this extension of meaning is carried out can 
vary. Based on Weinreich’s (1963, 1964) two types of ambiguity, Pustejovsky 
distinguishes between contrastive and complementary ambiguity.
‘Contrastive ambiguity’, traditionally known as homonymy, takes place when a 
lexical item accidentally takes two distinct and unrelated meanings. Pustejovsky is not 
interested in the reasons (historical, orthographical…) why this arbitrary association of 
senses occurs, as they are not relevant for the lexicon construction and the synchronic 
study of meaning (cf. Lyons 1977), but in the various processes that can disambiguate 
lexical items with this type of ambiguity. He proposes the following three processes: 
(i) Pragmatically constrained disambiguation: when the comprehension of the utterance is performed 
in a specific context. 
(ii) Priming and context setting: disambiguation by virtue of the discourse within which the sentence 
appears. 
(iii) Sortally constrained disambiguation: the knowledge of the predication relation in the sentence. 
The other type of ambiguity is ‘complementary polysemy’: when lexical senses 
are manifestations of the same basic meaning of the word as it occurs in different 
contexts. He distinguishes between those cases where the category of the lexical item 
changes and those where it does not; the latter is what he calls ‘logical polysemy’.
These multiple senses of a word have overlapping, dependent or shared meanings 
and seem to be systematically related. In the cases of nominals, Pustejovsky proposes 
seven different types of alternations: count / mass, container / containee, figure / 
ground
32
, product / producer, plant / food, process / result, place / people. The 
appropriate interpretation depends on the context. For instance, in the case of figure / 
ground reversals
33
: 
32
An interesting example for this alternation is the ‘animal grinding’ as discussed in Pelletier and 
Schubert (1986) and Copestake and Briscoe (1996). 
33
The phenomenon of figure / ground, first introduced by the Gestalt psychology, is at the very 
centre of the Cognitive Linguistics approach, see Talmy (1978), Langacker (1987, 1991a). 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
37
(5)
a. 
The window is rotting 
b.
Mary crawled through the window 
The ambiguity of window as ‘physical object’ and ‘aperture’ and the choice of 
one of the interpretations in each sentence is determined by the context in which these 
sentences have been uttered.
In the case of adjectives, the logical polysemy depends on what the adjective 
modifies; and finally, in verbs, it depends on the multiple complement types they select, 
as for example in the case of inchoative and causative verbs. 
The main difference between these types of ambiguity lies in the manner in 
which the senses are related. In contrastive ambiguity both senses are contradictory in 
nature, i.e. one is available only if every other sense is not. In complementary polysemy, 
on the other hand, there is a weaker shadowing effect, both senses are relevant for the 
interpretation of the lexical item in context, but one seems to be focused for purposes of 
a particular context. 
This is why Pustejovsky (1991, 1995: 39ff.) argues that the distinction between 
these ambiguities is necessary and rejects other approaches such as the ‘Sense 
Enumeration Lexicon’ (cf. Hirst 1987), where such a distinction is waived. The 
advantage of this type of approach is that the lexicon remains a separate and independent 
source of data. It is insufficient, however, when accounting for the creative use of words 
(new senses in novel contexts), the permeability of word senses (not atomic definitions 
but overlap and make reference to other senses of the word) and the expression of 
multiple syntactic forms (a single word sense can have multiple syntactic realisations). 
Logical polysemy has also been discussed by Briscoe and Copestake (1991) and 
Copestake and Briscoe (1996) under the name of ‘constructional polysemy’ and defined 
as one lexical item with apparent ambiguities that arise from a process of co-
composition in the syntax. This case is more apparent than real because lexically, there 
is only one sense and it is the process of syntagmatic co-composition (Pustejovsky 1991) 
that causes the sense modulation. Their approach is similar to that of Pustejovsky 
(1995), but they also provide a formal mechanism for treating the compositional 
interpretation derived from the qualia as defeasible knowledge. Copestake and Briscoe 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
38
(1996) also proposed another type of polysemy called ‘sense extension’. In this case, a 
lexical item is predictably related to two or more senses. It is not a sense modulation but 
a sense change, and therefore, they argue that it requires lexical rules that can create the 
derived senses from basic senses, together with specific conditions related to the speaker 
and context. 
Yet another type of sense extension not included in Pustejovsky’s analysis is 
‘referential transfer’, i.e. when a name of a property is mapped into a new name 
denoting a property to which it functionally corresponds. This phenomenon has been 
addressed in Fauconnier (1985) and Nunberg’s (1996) ‘predicate transfers’, which 
Nunberg argues are licensed pragmatically. 
Pustejovsky’s Generative Lexicon proposes a model that addresses the question – 
neglected by Cognitive Semantics – of how senses are created. It states that a core set of 
word senses is used to generate a larger set of word senses when individual lexical items 
are combined with others in phrases and clauses. Although Pustejovsky is mainly 
concerned with non-metaphorical meanings, it seems an appropriate model to account 
for the way in which polysemy is created. In Chapter 4, I apply this model to the 
analysis of the polysemous senses in perception verbs.