• No results found

SURFACE FEATURES 115 CLUE-MT

Surface Features

V. SURFACE FEATURES 115 CLUE-MT

T r i g g e r : t a : PUSH_TOPIC (~P —> X) t b : IS_SALIENT (A g , - X , - P -» X) t c : IS_SALIENT (A g, ~P X, ~P ^ X) t d : POP_TOPIC ( - P - » X) ( t a < t b < t d ) ( t a < t c < t d ) ( t c < t b ) U p d a t e : t n : MAKE_SALIENT (A g , C L U E ( p o l a r i t y , n e g a t i v e ) , ~P —> X) ( t c < t n < t b )

Figure 5.1 The CLUE-MT heuristic

The intersection of the POLARITY feature with another on Knott’s list, SOURCE OF COHERENCE

is also important in realising argument. In any situation where a rule is defeated (including the defeasible rule of speaker adherence to antecedent and consequent of an implication used in MT), the speaker intends

“to inform the reader that some general rule in the reader’s model of the world is defeated in the situation being described” (Knott, 1996, p i 04)

This Knott terms negative polarity with a semantic source of coherence (following (Sanders et a i,

1992)), and is contrasted with a negative polarity and pragmatic source of coherence, where “the writer’s aim is that some general rule in the world itself actually fails in the present instance” ibid.

In the latter case only, but can be replaced by the construction admittedly...but. The use of both but and

admittedly...but features in the realisation of counter-counterargumentation, and it might be expected therefore that the pragmatic/semantic distinction might play an important role.

The distinction between semantic and pragmatic negative polarity is illustrated by Knott in his examples 6.10 (pragmatic) and 6.11 (semantic), shown in Figure 5.2 (notice that Knott employs ‘# ’ to indicate a semantically unlicensed cue - rather than the more usual linguistic reading of ‘probably ill- formed’).

} they’re still bound to win. (6.10)

Mary behaving oddly. | t e r | ,he didn’t eat any of it. (6.11) She ordered a pizza, V # admittedly...but 3

Figure 5.2 Examples 6.10 and 6.11 from (Knott, 1996, pl03)

In the first example, admittedly...but and but are both acceptable; in the latter, admittedly...but is not. This, Knott claims, is a result of the defeat of a (real-world) causal rule for the former, and of a (hearer- resident) deductive rule for the latter.

Unfortunately, this seems to be something of a weak spot in Knott’s characterisation. The first problem is with (6.11), which intuitively seems rather weak, particularly as an extra sentence of context is introduced. It is unclear exactly how to distinguish the rules in the hearer’s model of the world and the world itself. In Knott’s examples, the problem is confounded by the inclusion of a much more significant difference which may, in fact, account for the but / admittedly...but distinction rather more definitively. Knott mentions that 3dmittedly...but “has a specifically argumentative flavour” (pl03). A more incisive reformulation of this claim is that the locus of attention is the conclusion - it is the conclusion which forms the functional aim of the text fragment (this is often indicated by stress patterns in an utterance, i.e. Admittedly P, but not C ). This is in clear contrast to (6.11) where the locus is the defeated rule itself - the speaker is uttering something like P but not C; Interesting, huh? This is manifest in Knott’s example by the presence of the first sentence - a conclusion based on the defeat of the inference rule. Reformulations of Knott’s two examples may serve to clarify this point.

(Mary is anorexic) i admittedly...but \ she didn’t eat any of it (5.7) She ordered a pizza I V but 3

United are renowned for Aeir p „ , , they’re still bound to win (5,8) underhand player transfers. They admittedly...but ]

have some key players injured

In (5.7) stress patterns and the parenthetic sentence contextualise the utterance to ensure the locus of attention is the conclusion (if that context were available from non-linguistic sources, neither stress nor the parenthetic phrase would be necessary). With the locus so fixed, the admittedly...but construction becomes valid. Similarly, in (5.8), the first sentence emphasises that the locus is the defeated rule. Importantly, the rule itself does not change from Knott’s original to these reformulations - i.e. the same rule is defeated in (6.10) as is defeated in (5.8) and the same rule is defeated in (6.11) as is defeated in (5.7). The rule has not shifted from the hearer’s model of the world to the world itself, nor in the opposite direction. By Knott’s analysis therefore, the licensed cues should not change - and yet, examples (5.7) and (5.8) demonstrate that they do. This works well for an analysis of the negative polarity in MT. Under Knott’s characterisation, it would be necessary to determine whether the inference rule ‘uttering an implication suggests belief in antecedent and consequent’ is part of the hearer’s model of the world, or whether it is a rule o f the world itself. There would seem to be reasonable arguments for taking either line (or a combination of both); eschewing either completely would be rather unpalatable. Instead, using a locus-based approach, it is necessary only to determine where the locus of attention lies in an MT. It is quite clear that the locus of attention is the conclusion (i.e. the consequent of the MT) and not the rule (‘uttering an implication ... etc.’). The only exception would be a highly contrived example of the form (5.9) - where, as expected, admittedly...but is prohibited. In standard cases of MT, however, uttering Admittedly P -> Q. but not Q...\s perfectly acceptable.

"'zzzzas: {rv,,

«)

. . . ^ j A • r T> V ^ admittedly...but 3

V. SURFACE FEATURES 117

A similar situation is found in counter-counterargumentation, between attacking premises of counterarguments, and attacking the support that premises have for conclusions in counterarguments - as discussed in §4.1.1, these two techniques are characterised by the AS level operators UCP and UCI respectively. In a situation where the hearer (as modelled by the speaker) maintains an argument that it is wet on the basis that it has been raining, and that the latter implies the former, the speaker might invoke either UCP or UCI to generate the arguments in Examples 5.10 and 5.11^^:

Raining makes the f admittedly.. .but \ it hasn't been raining,

ground w e t \ j so it’s not wet. ' ^

but 1that doesn’t mean the . . .

I t s been raimng | ground is wet

In both cases, the locus of attention is the conclusion, and so in both cases, either but or

admittedly...but is appropriate. Identifying the negative polarity in UCI and UCP enables the construction of two further heuristics, CLUE-UCP and CLUE-UCI, which introduce appropriate goals of saliency with these features specified. Figure 5.3 gives the definitions for the two heuristics. Note that as with CLUE-MT, CLUE-UCP requires the implication step to precede the refutation of the precedent to license the introduction of a negative polarity feature; CLUE-UCI similarly requires the step which is dissonant with hearer beliefs - this time the implication step itself - to succeed that which is consonant.

^ Without further support, of course, Example 5.10 is an example of the fallacy of denying the antecedent. Clearly, however, UCP is generally used in conjunction with other supports for the conclusion. Example 5.11 is even weaker - there is a strong expectation for information about the abnormality (that the ground is covered, sheltered, etc.)

CLUE-UCP

Related documents