My approach to Cognitive Poetics is, then, eclectic. I would like to illustrate this from my recent work on metrics. I may describe an issue in generative terms borrowed from Halle and Keyser, then observe a problem and "diagnose" its nature in terms of analytic philosophy, and eventually offer a solution that requires a cognitive perspective. Consider, for instance, the following two groups of verse lines:
Group I:
1. To a gréen thóught in a gréen sháde (Marvell)
2. To the wíde wómb of uncreated night (Milton)
3. Of the wíde wórld I stand alone and think (Keats)
4. Of the wíde wórld dreaming of things to come (Shakespeare)
Group II.
5. In profúse stráins of unpremeditated Art (Shelley, "To a Skylark")
6. Another clipped his profúse lócks, and threw (Shelley, "Adonais," 93)
7. His extréme wáy to her dím dwélling place (Shelley, "Adonais," 68)
8. Upon the supréme théme of Art and Song (Yeats, "After long silence" )
9. And sat as princes, whom the supréme Kíng (Paradise Lost, I. 735)
10. Supposed as forfeit to a confíned Dóom (Shakespeare, Sonnet 107.4)
11. No, let the candied tongue lick absúrd pómp (Shakespeare, Hamlet)
12. O golden-tongued Romance with seréne lúte (Keats, "On sitting down...")
13. Shall behóld Gód, and never tast Deaths woe (Donne, Holy Sonnet 7.8)
In both groups we find in each line a stressed syllable in a weak position followed by a stressed syllable in a strong position. In the first group, the first stressed syllable constitutes a monosyllabic word in a weak position ("green thought," "green shade," "wide womb," etc.); in the second group it constitutes the last syllable of a polysyllabic. Such outstanding theoreticians as Wimsatt and Beardsley, and Halle and Keyser consider these two types as equally acceptable (a stressed syllable may occur in a weak position provided that it is followed by a stressed syllable in a strong position). Magnusson and Ryder and Kiprsky rule the first group metrical, the second group unmetrical (adding a second rule: a stressed syllable may not occur in a weak position if an unstressed syllable of the same word occurs in a strong position). Marina Tarlinskaja tells me that in Russian prosody both groups would be unmetrical: a stressed syllable in a weak position is metrical only if preceded and followed by a stressed syllable in a strong position. Magnusson and Ryder claim that their prosodic theory is superior to that of Halle and Keyser (both generative theories) in that it rules out such lines as the ones included in group II, while they are metrical under Halle and Keyser's stress-maxima theory. Halle and Keyser could, of course, reply to them: our theory is superior, because it admits both groups of lines and, indeed, both types are metrical. One can't help asking the question: if Shakespeare, Donne, Milton, Shelley, Keats and Yeats don't conform with Magnusson and Ryder's or Kiparsky's rulings, how can we tell who is right and who is wrong: the poets, or the theoreticians. And if the latter are wrong, does anything go?
I handle such a situation with tools adopted from analytical philosophy. I adopted from Beryl Lake's "On the Irrefutability of Two Aesthetic Theories" a logical notion according to which such rulings of "metricalness" would be "irrefutable." Halle and Keyser might claim that their theory is more adequate because it takes account of verse lines written by some of the greatest English poets. Magnusson and Ryder, in turn, may admit the existence of such instances, but claim that they don't refute their theory: they are, in fact, unmetrical lines (indeed, Halle and Keyser do rule some other verse structures as "unmetrical," by criteria that are no less irrefutable). From Wittgenstein I have adopted the notion of "open concepts" (which recently has been reinvented as "fuzzy concepts") suggesting that it is a matter of arbitrary decision where we draw the limit between "metrical" and "unmetrical" -- just as between "hills" and "mountains" - - before or after group II. "Arbitrary" does not suggest here "abusing unlimited power," but that the "arbiter" had some freedom of judgement in making that particular choice, in view of the particular cisrcumstances.
According to my cognitive conception one could construct, regarding excerpt 13, a mounting scale of complexity (or unnaturalness), from (a) "Shall sée óld Gód," through (b) "Shall be óld Gód," to (c) "Shall behóld Gód." In Russian prosody only the first of these constructions would be ruled "metrical"; English prosodists accept (b) but are, as we have seen, in disagreement regarding (c). According to my conception, the various theoreticians draw the utmost limit of "metricalness" at a different point on the same scale of complexity: Russian prosodists after (a), Magnusson and Ryder and Kiparsky after (b), whereas Wimsatt and Beardsley and Halle and Keyser admit (c) too (but exclude some other possibilities). Each later item requires greater effort for its rhythmical performance. If the reader is successful in offering a rhythmical performance to it, each later verse line displays greater tension than the preceding one(s); if not, the verse line falls into chaos. Consequently, Wimsatt and Beardsley and Halle and Keyser are right in claiming that both groups I and II are acceptable; whereas Magnusson and Ryder and Kiparsky are right in claiming that there is a difference between groups I and II -- but not in kind, as they say, only in degree. As I have already indicated, such a conception requires a psychological definition of rhythmical performance," from which operational predictions can be derived.
Let us survey what I have done so far. I started out with an issue in which the state of affairs appears to be unsatisfactory. I described it in terms derived from generative metrics. There is a controversy between leading prosodists concerning the metricalness of a certain metric figure. None of them can give arguments that would refute his opponents' position. To sort out this problem, I started with tools borrowed from analytic philosophy. The question at this stage is not cognitive at all: whether "metricalness" is an a priori or an a posteriori concept; whether it is refutable or not; whether there is or is not a permanent boundary between "metrical"and "unmetrical"; and on what authority do we draw this boundary, permanent or not?
As Wittgenstein succinctly put it, "We draw a boundary -- with a special purpose." The dead end in the aforementioned controversy could perhaps be avoided by explicitly asking the question "with what purpose do we draw the boundary" (Wimsatt and Beardsley do rely on performance in accounting for the acceptability of these lines, but ignore degrees of difficulty). The present approach draws the boundary in light of the reader's ability or willingness to perform the verse instance rhythmically. It has constructed a mounting scale of complexity that can be described in linguistic terms; it just happens to be the case that the purpose "rhythmicalness" is best described in cognitive terms: it is constrained by the limitations of short-term memory, by the Gestalt principles of grouping and by the relationship between clear-cut articulation and the saving of mental processing space.