• No results found

Bisociation, frames, and scripts

4 Incongruity and its resolution

4.2 Bisociation, frames, and scripts

The proposal of Koestler (1970), that humour is created by bisociation (see Section 4.1 above), is a widely cited version of the incongruity theory.

Even where Koestler is not referenced, the essence of his ideas recurs. As Norrick points out (citing commentary by Johnson (1976) and by Wilson (1979) ): ‘many other writers both before and since Koestler (1964) have advanced bisociation theories of humour in various guises and in various names’ (Norrick 1986: 227). The very prominent SSTH can be viewed as a bisociation-based approach (see Chapter 6).

If there is such a wide consensus that bisociation embodies the appro-priate notion of incongruity, the question arises: why do we need to study the specific mechanisms of jokes? There are several reasons why the Koestler conjecture does not close down the debate even on incongruity-based theories of humour. We will now review three of the main reasons.

Undefined terms

There is still not a clear formal definition of Koestler’s terminology (‘frames’, ‘perceive in’, ‘habitually incompatible’) which would allow researchers to predict whether particular stimuli (e.g. specific texts) would count as manifesting bisociation or not. This is essential if bisociation theory is to become a predictive, falsifiable scientific theory, or if it is to be amenable to computer testing. Proponents of bisociation or its later versions might well argue that formal definitions have been provided, by the introduction to artificial intelligence of frames (Minsky 1975) and its various developments which use the term scripts (Schank and Abelson 1977), (Raskin 1985), concepts which are intuitively similar to a Koestler frame in more than just name. Minsky proposed the frame as a useful and psycho-logically plausible form of knowledge representation for a wide variety of subject matters and a range of tasks. Computer programs were written (e.g. Bobrow et al. (1977) ) that purported to exemplify the frame-based approach. It would seem, therefore, that there must be a formal defini-tion of a frame. In fact, there was no single accepted definidefini-tion, and what the various implementations of frames (e.g. Roberts and Goldstein (1977), Bobrow and Winograd (1977) ) had in common was that they were general-purpose data structures, generalizations of data types in more mundane programming languages such as Pascal. Indeed, some of the devices that were debated in the mid-1970s as ‘knowledge representation’ (e.g. default values, slot restrictions) are now routine parts of the general programming language Common Lisp (Steele 1990). A program which uses Common Lisp structures, or is written in an object-oriented programming language like Java (Arnold et al. 2000), could be argued to be frame-based. That is, the frame data structure can be seen to be so general that it makes very 11

11

11

111

111

111

11

little claim about how the knowledge is structured. The answer to ‘what is a frame?’ is ‘(virtually) anything’.

Frames versus general inference

A reasonable way to view Minsky’s central insight was that he was drawing attention to the effectiveness, and ubiquity, of knowledge which was both specialized (for the subject or task at hand) and precomputed (rather than being inferred on every occasion from general principles). There was also the use of default information in the absence of data to the contrary. (The justi-fication for scripts (Schank and Abelson 1977) was similar, but emphasized sequences of events.) The choice of a data structure to represent such knowledge is something of a separate issue. In the mid-1970s, there was an immediate response to the frame-based approach from those who argued that some suitable form of logical inference would be capable of gaining the same effects as a frame representation (Hayes 1979). The central points were that the reasoning phenomena pointed to by Minsky (specialization, precomputation, defaults) are not incompatible with the use of logical inference. This case has become stronger as logics have become more sophisticated, particularly default logics which allow non-monotonic or default reasoning (Reiter 1978, 1980), Turner (1984: Ch 5).

An analogous argument arises in the case of bisociation theories. Suppose it is proposed that a particular joke involves the bisociation of two frames, on the grounds that there is a conflict between two perspectives on the information in the text. Is this different from, or the same as, saying that the text supplies information which leads to conflicting inferences?

(27) An ancient, wizened man accosts a lady of the night and inquires as to her rates. She replies, ‘$5 on the floor, $10 on the couch and $15 in bed.’ As he hands the hooker $15, she remarks, ‘Okay, once in bed,’

to which he objects, ‘No, three times on the floor.’

(Norrick 1986: 239) Commenting on (27), Norrick suggests that a schema set up by the early part of the text ‘conflicts’ with a schema supplied by the punchline. These observations could equally well be phrased as saying that the joke depends on the hearer making a default inference that the old man will be sexu-ally weak, and that this (defeasible) inference is then overturned by subsequent information. It is not at all clear that these analyses differ in substance, and it is hard to envisage data which would distinguish them empirically. If that is the case, then using the terminology of frames is just a handy shorthand, with no substantive theoretical import. The crucial issue is: how does viewing something as frame-conflict tell us anything we would not glean from viewing it as some other form of logical clash (such 1

1

1

11

11

11

as the overriding of a default inference), or even just informally saying that there has been a clash between two perspectives?

Humour knowledge versus general knowledge

Articles advocating bisociation are not always explicit about the status of their frames with respect to general knowledge. Are frames entities which participate solely in the processing of humorous phenomena, or are they more general information structures used in other cognitive activities?

Opting for frames being peculiar to the process of humour interpretation would raise certain questions. How is it that the joke-audience has a language-understanding mechanism (frames) which is used solely during jokes, and is different from the language-understanding methods used for other texts? This question is not restricted to language-understanding narrowly viewed, as sometimes the frame-clash occurs in propositions inferred from the meaning of the text; hence, this would entail a humour-specific reasoning mechanism. The postulation of a completely separate cognitive apparatus for perception and reasoning during humorous stimuli is hugely implausible. This is not to argue that we treat joke texts as if they were non-joke texts in every way; the point here is that it is bizarre to propose duplicating all our linguistic and world knowledge, and perhaps even introducing a different formal mechanism, for the processing of jokes.

That position is mentioned here merely for completeness, but is almost certainly a straw man (although the suggestion by Raskin (1985: 128) that having a sense of humour involves having more scripts available seems to hint at such a position). It is much more likely that proponents of bisoci-ation assume that frame structures are suitable for general perception and cognition (as did the 1970s advocates of frames). In that case, we have to assess the Koestler formula from the point of view of necessity and sufficiency for humour. In what way do humorous bisociations differ from other align-ments or clashes between frames that might occur in, for example, analogies, misunderstandings, or poetic figures of speech (metaphor, simile, etc.)? The following observations seem rather bisociation-like, but concern the semantic patterns underlying poetic devices:

. . . two powerful but unrelated images are presented to us individu-ally and we are forced to discover their relation.

. . . juxtaposition seems to promise coherence and thus impels us to try to construct a coherence.

(Hobbs 1990: 129) For each of these other types of phenomena, one might expect that either they (all) involve the relating of two ‘habitually incompatible’ frames, or they are based on a different formal mechanism. But some misunderstandings 11

11

11

111

111

111

11

are humorous, others are not; some analogies may be amusing, most are not; some metaphors or similes may be amusing. If, on the other hand, Koestler and his successors are not proposing bisociation as a necessary and sufficient condition for humour, it is unclear what they are suggesting.

None of this is to suggest that the idea of bisociation has no merits, or that it is wholly incorrect. It is an interesting way of looking at humorous phenomena, being particularly convincing for humour where a collection of preconceived ideas about some situation is played off against another cluster of expectations. For example, BBC TV’s Monty Python’s Flying Circus presented a sketch depicting a football match between teams of famous philosophers, with the perspective of the commentator veering back and forth between a sporting contest and a philosophical debate.

With suitable development, bisociation might lead to a useful and insightful theory of humour. The point here is that this development is yet to be done. To offer bisociation theory as a theory of humour is to pass off as an oak tree something which is believed to be an acorn.

The most prominent bisociation theory of recent decades, Raskin (1985)’s SSTH, is one of the few which even starts to address the issues raised above. Raskin tries to make the notion of ‘frame’ more precise (as his version of the script), puts his structures and their associated operations forward as a general theory of language interpretation and inference, and elaborates on which kinds of frame-clash (script opposition) count as humorous. The SSTH therefore goes much further than work which merely asserts that particular jokes exemplify bisociation. However, we will argue in Chapter 6 that the SSTH is still disappointingly vague.