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

4.2.1. DISCUSSION 
One of the major drawbacks that Pustejovsky sees in previous compositional 
approaches is that these do not see the lexicon as an active and central component in the 
linguistic description, but either as active functors or passive arguments. In his 
Generative Lexicon, Pustejovsky argues 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. For instance, in the case of verbs, the extensions of 
meaning are achieved by the combination of the verb with its arguments. 
In Chapter 2, I presented a typology of both prototypical and non-prototypical 
meanings in sense perception verbs. In the case of the semantic field of touch, I 
proposed several extended meanings, some were still physical such as ‘to partake of 
food or drink’ and ‘to affect, physically’, and some of them were abstract or 
metaphorical as ‘to reach’ and ‘to affect, non-physically’. In this section, I offer a 
semantic characterisation of these meanings using Pustejovsky’s framework to see 
whether this framework is enough to explain how these semantic extensions are 
obtained. 
One of the meanings proposed for the sense of touch is ‘to partake of food or 
drink’, as illustrated in (6): 
(6)
John hardly touched the food 
ARGSTR = 
[ARG1 =
1
: human] 
[ARG2 =
2
: physobj (TELIC = eat)] 
[D-ARG= 
3
: how] 
EVENTSTR =
[E
1
= e
1: 
activity] 
QUALIA = [FORMAL = partake (e
1
,
1
,
2
,
3
)
[AGENTIVE = touch_act (e
1
,
1
,
2
) 
The verb touch in (6) is associated with three arguments: two TRUE 
ARGUMENTS (the subject ARG1 and the object ARG2) and a DEFAULT 
ARGUMENT, each argument is given a boxed number, called ‘tags’ (Pollard and Sag 
1994), that indicate how the information is shared between these arguments and the 
qualia. The verb is analysed as an activity, as shown in the EVENT STRUCTURE. Both 
the ARGSTR and the EVENTSTR are bound by the QUALIA. The activity is identified 
122

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
as that AGENTIVE act of ‘touching’ involving the subject ARG1 and the object ARG2. 
The FORMAL role expresses the activity of there being such an object ARG2, whose 
role and function is expressed by its TELIC quale, and a D-ARG that tells us about the 
manner in which the activity was performed. As the ARG2 in this case is the food, the 
meaning that I have represented is ‘to partake of food’; however, if we change this 
ARG2 for another one like the drink, the meaning would be ‘to partake of drink’ instead. 
It seems that Pustejovsky’s analysis accounts for this example quite neatly, as we 
can see how every member of the sentence contributes to the overall meaning of the 
sentence.
Example (7) is more complicated. There are two possible interpretations: one is 
the prototypical meaning of touch as an activity verb, and the other is the extended 
physical meaning of ‘to affect’. The characterisation of (7) is as follows: 
(7)
John touched my clothes 
ARGSTR = [ARG1 =
1
: human] 
[ARG2 =
2
: phyobj] 
EVENTSTR =
(a) 
[E
1
= e
1: 
activity] 
(b)
[E
1
= e
1: 
activity] 
[E
2 
= e
2
: result] 
RESTR = < 
α
QUALIA = [FORMAL = affected (e
2
,
2
) (Only in (b)) 
[AGENTIVE = touch_act (e
1
,
1
,
2
) (
2
not in (b)). 
In this example, the verb touch has two TRUE ARGUMENTS: the subject 
ARG1 and the object ARG2. The EVENTSTR in (7) is more complicated than that in 
the previous example, as there are two options. Option (a)
112
corresponds to the first 
interpretation, i.e. prototypical activity meaning. In (a), the verb is taken as an activity in 
the EVENTSTR, which also results in an AGENTIVE quale that involves the subject 
ARG1 and the object ARG2. Option (b) corresponds to the second interpretation, where 
there is not only a contact between the subject and the object, but also an effect on the 
object, a change of place, position in this case. In (b) then, the verb is analysed as an 
accomplishment, containing two subevents, an activity (e
1
) and a result (e
2
). The order in 
123

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
which these subevents occur is restricted by the relation of ‘exhaustive ordered part of’ 
represented by < 
α
, where e
1
and e
2
are temporally ordered such that the first (e
1
) 
precedes (e
2
), each is a logical part of the event and there is no other event that is part of 
it (see Pustejovsky 1995: 69). In this case, the AGENTIVE act of ‘touching’ involves 
only the subject ARG1. This AGENTIVE role is bound to e
1
(activity). The FORMAL 
quale expresses the result of that activity, e
2
, and the object ARG2, which has been 
affected by such an activity upon which the result has been inferred. 
Although in (7), there are two possible interpretations inferred, it is possible to 
account for both using Pustejovsky’s framework. The disambiguation is carried out by a 
different characterisation of the EVENTSTR (Options (a) and (b)). 
A problematic example, however, is (8) where two unpredictable interpretations 
are simultaneously inferred. 
(8)
John touched Mary 
ARGSTR = [ARG1 =
1
: human] 
[ARG2 =
2
: human] 
EVENTSTR =
(a) 
[E
1
= e
1: 
state] 
(b) [E
1
= e
1: 
activity] 
(c)
[E
1
= e
1: 
activity] 
[E
2 
= e
2
: result] 
RESTR = < 
α
QUALIA = [FORMAL = affected (e
2
,
2
) (Only in (c)) 
[AGENTIVE = touch_act (e
1
,
1
,
2
) (
2
not in (c)). 
In this example, the verb touch has only two TRUE ARGUMENTS: the subject 
ARG1 and the object ARG2. The EVENTSTR in (8) is more complicated than that in 
the previous example, as there are three options. In (a), the verb is understood as a state, 
in which case, the AGENTIVE act involves the subject ARG1 and the object ARG2. 
This interpretation can be paraphrased as John being next to Mary. Options (b) and (c) 
are explained in the same way, as (a) and (b) in the previous example (7). In (b), the verb 
is taken as an activity in the EVENTSTR, which also results into an AGENTIVE quale 
that involves the subject ARG1 and the object ARG2. This option is the prototypical 
124
112
The possibility of having several different options in the EVENTSTR is my modification of 
Pustejovsky’s model. 

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
activity meaning of touch, where only contact is implied. And in (c), the verb is analysed 
as an accomplishment, containing two subevents, an activity (e
1
) and a result (e
2
), also 
restricted by the relation of ‘exhaustive ordered part of’ represented by < 
α
. Again in (c), 
John does not only have some physical contact with Mary, but he causes an effect on 
her. 
However, in (8) there is an added difficulty. In this example it is not possible to 
characterise using Pustejovsky’s model, the difference in meaning when we are referring 
to the physical or to the abstract/metaphorical touching. This ambiguity is not present if 
the EVENTSTR is (a), namely when E
1
is a state and it is understood that John is 
adjacent to Mary; but in those cases when the EVENTSTR is either (b), i.e. John can 
touch Mary physically or metaphorically, or (c) i.e. John can affect her, physically 
(making her shiver for instance) or metaphorically (feelings). The semantic 
characterisation for both arguments and the verb touch under this model is the same 
either if the reading is physical or abstract; however, it is very clear that they are two 
different interpretations that must be accounted for separately. Unfortunately, this model 
does not seem to have a solution for either explaining how these two meanings are 
inferred or solving this ambiguity by means of semantic characterisations. It could be 
argued that the possible metaphorical readings could be listed somewhere in the 
QUALIA, but it is not clear how this would work in cases like (c), where the same 
arguments and event type produce two unpredictable interpretations. 
If we look at other metaphorical meanings in this sense, for instance the meaning 
‘to reach’ illustrated in (9), the same problem seems to be present.
(9)
John touched the highest point in his career 
For example, ‘to reach’ in (9) can be represented as follows: 
ARGSTR = 
[ARG1 =
1
: human] 
[ARG2 =
2
: obj
4
: phys or non-phys (TELIC: location)] 
[D-ARG= 
3
: where] 
EVENTSTR =
[E
1
= e
1: 
activity,
4
: phys or non-phys
] 
QUALIA = [FORMAL = reach (e
1
,
1
,
2
,
3
) ] 
[AGENTIVE = touch_act (e
1
,
1
,
2
) ] 
125

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
In this case, there are also two TRUE ARGUMENTS and a DEFAULT 
ARGUMENT. The verb is analysed as an activity in the EVENSTR. The activity is 
identified as an act of ‘touching’ involving the subject ARG1 and the object ARG2. The 
word point fills in ARG2 in this example and we know by the information provided in 
the TELIC quale, that the function of this ARG2 is to give reference to a location. The 
problem in this example is twofold. On the one hand, we see that although the FORMAL 
quale is ‘reach’, it does not seem to be very clear how this meaning is derived from the 
interaction between the QUALIA and the members of the ARGSTR. The other problem 
is that in this case, point does not refer to a physical point, but to an imaginary one. The 
point that John has reached is a level in his career. This problem is solved by specifiying 
in the ARG2 that the object can be either physical or non-physical (phys or non-phys). 
This information is bounded to the EVENTSTR in a way that, whenever the object in 
ARG2 is physical then, the activity referred to is also physical, and whenever the object 
selected in ARG2 is non-physical, then the activity in the EVENTSTR is also non-
physical. By adding this information to the description of the ARG2, it seems that we 
can account for the metaphorical meaning of (9).
However, in an example like (10), where the meaning inferred is ‘to deal with’, 
we do not need to specify whether ARG2 is physical or non-physical, as the ARG2 is a 
concept, i.e. non-physical entity. 
(10) John 
touched on the topic
ARGSTR = 
[ARG1 =
1
: human] 
[ARG2 =
2
: concept (CONST = abstract/count)] 
[D-ARG= 
3
: where] 
EVENTSTR =
[E
1
= e
1: 
activity] 
QUALIA = [FORMAL = dealt with (e
1
,
1
,
2
,
3
)
[AGENTIVE = touch_act (e
1
,
1
,
2
) 
In this example the FORMAL quale is ‘dealt with’. However, as pointed out in 
the previous examples how this meaning is obtained is not absolutely clear. It could be 
argued that the fact that ARG2 is a concept might result in this reading as ‘deal with’. 
However, the fact that ARG2 is a concept does not seem to explain this meaning. As we 
126

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
can see in (11), ARG2 is also characterised as a concept, but the sentence is not 
felicitous. 
(11) *Mary 
touched the joke 
ARGSTR = 
[ARG1 =
1
: human] 
[ARG2 =
2
: concept (CONST = abstract/count)] 
EVENTSTR =
[E
1
= e
1: 
activity] 
QUALIA = [AGENTIVE = touch_act (e
1
,
1
,
2
) 
To sum up, we see that it is possible to account for some metaphorical meanings 
under Pustejovsky’s approach, either by adding more information in the ARG2, physical 
or non-physical as in (9) or by specifying that the ARG2 is a concept as in (10). 
However, it seems that this model is unable to explain how and why restrictions occur; 
that is to say why (10) is acceptable, but (11) is infelicitous.
As will be explained in more detail in Chapter 7, the reason why this example is 
not felicitous lies in the fact that ARG2 is not a ‘touchable’ type of concept, i.e. a joke 
cannot be touched in any abstract possible way as a topic is in the previous example. 
From a Cognitive Linguistics point of view, the fact that the joke is not licensed with the 
verb touch stems from the way we experience this sense in our lives, in the human 
embodiment of this sense (Johnson 1987).
In other words, the bodily basis of the tactile sense, which I will characterise in 
terms of properties in the following chapter, clashes with the way in which we seem to 
understand and experience a joke. In order for this sentence to be felicitous, the 
properties of the ARG2 are required to preserve the properties that characterise the sense 
of touch.
This requirement, that I call ‘verb property requirement’ (see Chapter 7), seems 
to be acknowledged by Pustejovsky in the example when he discusses read and book. In 
order for the representation of reading to be considered well-formed, it is necessary to 
have a complement not only containing information but also being propositional, i.e. a 
complement that is ‘readable’. Pustejovsky solves this requirement by making reference 
to the dotted argument x.y in the qualia, apart from the FORMAL definition of the 
relation between arguments (1995: 96). I will propose, however, that this requirement 
should be present in all cases, even in those examples where the semantic content of the 
127

---

B. Iraide Ibarretxe-Antuñano 
Polysemy and metaphor in perception verbs 
128
arguments is more important in the extension of meaning than that of the verb itself (cf. 
‘to partake of food or drink’).