Reuven Tsur

Rhyme and Cognitive Poetics

Reuven Tsur

Rhyme and Cognitive Poetics

One of the most fascinating things that can happen to someone engaged in the systematic investigation of any subject matter is to discover that his partial researches which he began independently from one another begin to grow together into one coherent field of research. My own work, which began with sporadic close-readings of a wide variety of poems, has been gradually growing together to a coherent theory of cognitive poetics. Recently I have published a rather comprehensive and systematic exposition of this theory (Tsur, 1992b). In the course of my explorations of poetry in a cognitive framework, I had shorter or longer observations relevant to rhyme in a variety of contexts, with little or no connection between them: from the point of view of phonetic and acoustic representation, its affect on short-term and long-term memory, the gestalt rules of grouping, semantic information processing, and the generation of such poetic qualities as witty, emotional, or hypnotic. The present paper, prompted by Uzi Shavit's illuminating little book Rhyme and Reason: Studies in the Historical Poetics of Hebrew Poetry, proposes to collect some of these observations and integrate them into a comprehensive cognitive account of rhyme. It will, here and there, fill in gaps, and add an entirely new section on dactylic rhyme.

In the introductory section of his book, Shavit proffers an overview of research on rhyme, and points out that, roughly, there are two kinds of approaches to rhyme: those that focus attention on the similar and the conflicting elements in rhyme; and those that regard rhyme as an organizing principle in the prosodic domain. Shavit regards these approaches complementary rather than mutually exclusive. In a footnote he comments: "These points of view are not contradictory. Thus, for instance, Reuven Tsur, in his discussions of grammatical and anti-grammatical rhyme (Tsur, 1976; 1983a; 1983b: 53--70) makes an important contribution to an understanding of the semantic role of rhyme. He also made an important contribution to an understanding of the compositional role of rhyme in his discussion of the sonnet (Tsur, 1988b: 84-91), as well as of the quatrain (Tsur, 1988a: 27-32); in addition, he explored the euphonic character of rhyme (Tsur, 1983a; 1983b: 78-82; 1987). Thus the distinction between the various kinds of approaches is mainly methodological" (Shavit, 1993: 14). Later, in a symposium on cognitive poetics (Tel Aviv University, 30.3.1993), he made a twofold suggestion: that one should attempt to integrate these and other scattered observations on rhyme into one comprehensive view, and that one should explore the relationship between a cognitive approach to rhyme and an historical approach of intertextuality: whether they are unrelated, or complementary in some way. In the present paper I make an attempt to take up these two challenges. The main bulk of the paper will approach rhyme from a variety of directions, enumerated above, attempting to account for significant intuitions of readers about rhyme. As for the relationship between cognitive and historical explanations of poetic affects, the paper will come to the following conclusion: Alternative intuitions concerning poetic affects are determined by alternative cognitive processing of individual poems; intertextuality can reinforce or tone down one or another such intuition.

Exploiting the Phonetic Code
Cognitive poetics assumes that poetry exploits for aesthetic purposes cognitive (including linguistic) processes that were initially evolved for nonaesthetic purposes, just as in evolving linguistic ability, old cognitive and physiological mechanisms are turned to new ends. The exploitation of the phonetic coding for rhyme is a case in point.1 Speech researchers distinguish a speech mode and a nonspeech mode. In the latter, the shape of the perceived sound is similar to the shape of the auditory information. In the former, only an abstract phonetic category (such as [a] [b] [i]) is perceived; the sound information that carries it is shut out of consciousness. I have suggested that there may be a third, poetic mode, in which some of the rich precategorial auditory information may reach consciousness, strongly affecting the emotional or poetic qualities of the speech sounds. Speech consists of several parallel streams of information. At the listener's end, we translate a stream of acoustic information into a stream of phonetic representations which, in turn, we translate into a stream of semantic representations, and so forth. There is a series of conversions, of recodings, from one stream to the other. Conversion from the acoustic to the phonetic stream involves a complete restructuring of the information, resulting in a string of abstract phonetic categories, so that only very little sensory information reaches, subliminally, consciousness. Some speech sounds are more thoroughly restructured ("more encoded") than others. In the voiceless plosives /p, t, k/ almost no pre-categorical auditory information is perceived, only the abstract phonetic category, whereas in periodic continuous sounds like vowels or sonorant consonants, or in such sibilants as [s, S] some rich pre-categorical sensory information is subliminally perceived, and may become accessible even to conscious introspection. This sensory information reverberates in short-term memory for a very short period of time. We must, however, assume that even in the case of voiceless plosives auditory memory is active, albeit in an impoverished form. As Conrad (1964) found out, there is evidence for acoustic confusion even with visually presented speech sounds. A syntactic unit can be perceived as a unit only if we can hold its beginning in echoic memory while new information is still coming in.

In the poetic mode of auditory perception, the precategorical sensory information that reaches consciousness somehow becomes significant. Liberman et al. (1972) describe a series of experiments by Crowder and Morton (1969), who found that in auditory (but not visual) presentation, vowels produce a recency effect (that is, performance on the last items is better than on the preceding ones) in certain cognitive tasks, but stops do not. Part of the explanation seems to be as follows:

The special process that decodes the stops strips away all auditory information and presents to immediate perception a categorical linguistic event the listener can be aware of only as /b, d, g, p, t, or k/. Thus, there is for these segments no auditory, precategorical form that is available to consciousness for a time long enough to produce a recency effect. The relatively unencoded vowels, on the other hand, are capable of being perceived in a different way [...]. The listener can make relatively fine discriminations within phonetic classes because the auditory characteristics of the signal can be preserved for a while. [...] In the experiment by Crowder, we may suppose that these same auditory characteristics of the vowel, held for several seconds in an echoic sensory register, provide the subject with rich, precategorical information that enables him to recall the most recently presented items with relative ease.

Crowder later found that this recency effect could be drastically reduced by a "verbal suffix", that is, when a nonsense syllable such as /ba/ was used to indicate when the recall-attempt was to be started, but not when a pure tone was used (1982). The explanation he offered was that each later arrival exerted "lateral inhibition" on earlier arrivals in the neurological system, so that only the abstract phonetic category should linger on, but not the precategorical sensory information. In our foregoing terms, rhyme delays the special process that in the course of decoding the stops strips away all auditory information, presenting to immediate perception a categorical linguistic event the listener can be aware of only as /b, d, g, p, t, or k/; by the same token, it exploits the "echoic sensory register" for aesthetic ends. Occasionally it may also cause confusion in acoustic memory.

Consider the following stanza from FitzGerald's "The Rubáiyát of Omar Khayyám":

(1) Some for the Glories of this World; and some
    Sigh for the Prophet's Paradise to come;
      Ah, take the Cash and let the Credit go,
    Nor heed the rumble of a distant Drum!

Elsewhere (Tsur, 1992a: 35-37) I pointed out that this quatrain is rich with euphonic sound-patterns. Some of them are neutral as the Prophet's Paradise in line 2, or Cash and Credit in line 3 (where they are focussed on words of symbolic significance). One is conspicuously onomatopoetic: rum ble ... Drum. It is the latter to which I wish to pay here some attention. All the sounds of rumble (except for the /b/) are continuous and periodic--and relatively unencoded. The first three of them are repeated in Drum, in the same order. The /d/ of Drum too is heralded by distant, and less conspicuously by heed (in such a periodic context, the periodicity of the voicing in the stops /b, d/ is foregrounded too). Now consider a remarkable thing about the sound pattern of the last line. The reader or listener is quite likely to become aware of the rich precategorical, periodic acoustic information and relate it to the rumble of the drum. In this way, the sound becomes "an echo of the sense", or, rather, the meaning and the acoustic structure reinforce the effect of similar features in each other. This phenomenon becomes even more remarkable when contrasted to another sound pattern of which Drum is part as well: it rhymes with some and come. Most readers who are aware of the rich precategorical auditory information in Drum report that they are not aware of a similar richness in the preceding rhyme-fellows. In the rumble--Drum sound pattern in relation to which I suggested there that the auditory trace may be enhanced rather than inhibited when several conditions meet, such as (1) when the sounds involved are continuous and periodical; (2) when massive sound clusters are repeated; (3) when semantic features of the words tend to draw attention to, rather than away from, the acoustic and articulatory features of sound patterns; (4) when one of the syllables concerned is the last one in a perceptual unit (that is, no "verbal suffix" follows it).

Bruno Repp (personal communication) suggests that if a subsequent stimulus is very similar to the preceding stimulus, it may generate an enhanced response, because of their integration. Robert G. Crowder (personal communication) suggests that there would be precedent for the assumption that the total effect would be the larger for having had a repeated sound. This depends on his assumption that both inhibitory and enhancing interaction takes place within the formant energy of the words, even though they may be spoken at different pitches. Thus, such sound patterns as rhyme and alliteration not only "exploit" the working of the auditory short-term memory, but actually enhance it.

When we emit strings of speech sounds, we obviously have recourse to phonetic coding. It is much less obvious that phonetic coding is a major resource in the performance of a wide range of cognitive operations. We have already mentioned the Crowder and Morton experiment. Researchers at the Haskins Laboratories (e. g., Liberman and Mann, 1981: 128-129; Brady et. al., 1983: 349-355; Mann, l984: 1-10) in their research on the possible causes of some children's difficulty to learn to read have revealed a deficiency in the use of phonetic coding by poor readers; good readers, by contrast, seem to make an excellent use of it. In one experimental task, poor readers had greater difficulty than good readers in tapping once or three times in response to the number of syllables in such spoken words as pig or elephant, or once, twice or three times in response to the number of phonemes in such words as eye, pie or spy. This has been interpreted as a deficiency in the use of phonetic coding. In another task, they had to memorize groups of words--either rhymed or unrhymed, as in the following ones:

(2) chain  train brain  rain   pain
    cat    fly  score  meat  scale

Good readers did consistently better with both kinds of groups than poor readers. However, with the rhymed groups, their performance seriously deteriorated. While their reliance on phonetic representation increased their overall performance, the similar sounds of the rhyming words seems to have caused confusion in their acoustic memory. Thus, this experiment reveals an intimate relationship between rhyme and the cognitive mechanisms involved in certain memory tasks. There was no difference in IQ between the two groups (in fact, in one of the experiments the average IQ of the poor readers was insignificantly higher). In nonverbal memory tasks the poor readers were as good as the good readers (in fact, insignificantly better). It was only that the good readers made efficient use of phonetic coding. Since the poor readers made inefficient use of the acoustic information in short-term memory, they were not penalized by the similar sounds of the rhyming words. Virginia Mann advances two plausible explanations for these experimental results. "On the one hand, poor readers might not resort to phonetic representation at all, relying instead on semantic or some other modes of representation. On the other hand, they may attempt to employ phonetic representation but for some reason their representations are less effective" (1984: 8).

Crowder and Wagner (1992: 228-230) summarize an experiment by Byrne and Shea in which subjects had to take a "reading test", reading out lists of words and then, unexpectedly, were given a memory test. They were presented with the words read earlier, interspersed with a number of additional words, to which they had to respond "old" or "new". The "new" words were either phonetically or semantically related to the "old" words. "Assume the prior items were home and carpet: house and rug would be the semantically similar foils and comb and market would be the phonetically similar foils". Good readers tended to confuse both phonetically and semantically related words, poor readers semantically related words. Crowder and Wagner insist: "The fact that poor readers seem to be using 'too much' meaningful processing does not imply that they are better at top-down processing than good readers. It is just that they may be so deficient in bottom-up processing they have no other recourse".

I do not presume to arbitrate between the two possibilities advanced by Mann. But this line of investigation may have tapped a kind of individual differences in verbal strategies that may characterize very advanced readers too; I mean students and professors of literature. Poor readers who in the kindergarten or the first three grades of elementary school are unable to respond appropriately to the number of phonemes or the syllables of a word, may learn to do so in due course. Such measures as the ones applied to first graders would be insufficient to measure a literature student's relative reliance on phonetic coding. However, we may find in prosody classes students of literature who are incapable of telling which syllable of a bisyllabic word in their mother tongue is stressed, even though they can pronounce it correctly. The plain fact seems to be that even at this level of students and professors of literature there are individual differences in this respect, and there are persons who do not seem to feel at ease with phonetic representation, and seek to fall back as frequently as possible on semantic coding. This is most conspicuous in the response of various readers to what Snyder (1930) called "hypnotic poetry". In a number of poems by Poe, Coleridge, and certain other poets, many readers are inclined to "attend away" from the meaning of the words, and to become "spell-bound" by their sound. The reader feels as if he were entangled by the sounds of these poems, and tends to perceive their meaning relatively dimly. However, some other readers respond quite differently: they may find the sound effects of the same poems rather boring and unemotional. Such differences are due to a person's being high or low on the personality variable "absorption" (Glicksohn, Tsur & Goodblatt, 1991). Still other readers may ignore all in all the sound effects of the same poems and seek to account for their significance solely in terms of their meaning. More often than not, such interpretations will also ignore the possible hypnotic effect of these poems. Such an inclination to account for the significance of poems solely in terms of their meaning may also be an indication of a basic preference for a semantic rather than a phonetic coding.2

Speech consists, then, of strings of abstract phonetic categories. The precategorical acoustic information that carries them is normally shut out of consciousness. Still, the perceived poetic affects, on the one hand, and the facilitation of certain cognitive tasks, on the other, indicate that some of this information is subliminally present and active. Rhyme typically exploits this precategorical acoustic information and, actually, enhances its memory traces. In nonaesthetic memory experiments, this reliance on phonetic representation reveals two typical effects. It enables verbal material to linger for some time in short-term memory for more efficient processing, on the one hand, and tends to cause acoustic confusion, on the other. In poetic language, the verbal material is subjected to much more sophisticated processing than in other uses of language. But in some instances at least, rhyme reverberates in echoic memory more intensely and for even longer than most aspects of poetic language.3

What in the nonaesthetic memory experiments is called acoustic confusion, in an aesthetic context may be perceived as "harmonious fusion", or "musicality" ("musicality" by its very nature is more intimately associated with the the sensory aspects of speech sounds than with the abstract phonetic categories). As a result, the rhyming units are felt to be rather closely knit together, even when they are apart in time to some extent; alternatively, the rhymes may be felt to spread some sort of an inarticulate sensory net over a considerable region of a poem.

Now let us consider in light of the foregoing discussion the following issue. In some instances a fuller body of the rhyme words is perceived than in other instances. Suppose we take a random collection of pieces of rhymed verse and a random collection of readers, and ask the readers to characterize the rhymes by such pairs of adjectives as compact ~ diffuse; lean ~ plump; clicking ~ reverberating; tight ~ spacious; flat ~ plastic, or the like. We may obtain a welter of judgments suggesting--wrongly, I believe--that such judgments are a matter of purely arbitrary subjective decisions. But if we realize that three kinds of variables are involved in such judgments, they will cease to appear arbitrary. The three kinds of variables concern the phoneme structure of the rhymes, the judge's cognitive style, and the relative strength of the gestalts present in the poem, both on the prosodic and semantic levels. It is difficult, though not impossible, to sort out what is the relative share of each of these variables in the judgments.

As for the phoneme structure of the rhymes, we may expect that the relatively unencoded speech sounds (that is, the ones that are not thoroughly restructured in perception, and some of the rich precategorical sensory information does reach consciousness) would more frequently elicit judgments that use the second member of each pair of the above adjectives. Such rhymes may be felt to have a fuller body. The phoneme structure of the rhymes may affect their perceived quality in an additional way too (Tsur, 1992a: 52-88). Speech sounds that are later acquisitions of the infant have greater emotional and aesthetic potential than the earlier acquisitions; among the later acquisitions, continuous and periodical sounds tend to have a highly musical or pleasant emotional quality, whereas abrupt or aperiodic sounds may have an unpleasant or unmusical quality. In French, the nasal vowels and the sound cluster -eur consist of late acquisitions that are acoustically continuous and periodic, and relatively unencoded from the phonetic point of view. It may be expected that highly emotional and musical (e.g., Symbolist) poetry in the French language should resort to the nasal vowels and the sound cluster -eur in its rhymes more frequently than rationalist poetry, or poetry that sets great store by its idea contents. A simple counting revealed that this is indeed the case (ibid., 66-67): Baudelaire in the first poem of his Fleurs du Mal used in his rhymes nasal vowels and the sound cluster -eur two and a half times as frequently as Boileau in the first one hundred lines of his L'Art Poétique. Or, consider the first stanza of Verlaine's "Chanson d'Automne", where the rhymes are nasal consonants and vowels, and the sound cluster -eur:

(3) Les sanglots longs
Des violons
De l'automne
Blessent mon coeur
D'une langueur

As for the judge's cognitive style, the less a reader tends to rely on phonetic coding in his language processing, the more he will be inclined to use the first member of the afore-mentioned adjective pairs in characterizing their perceived affect. And this might be the case, for probably different reasons, with persons who have low tolerance for delayed categorization. Phonetic coding consists in the substitution of an abstract phonetic category for the acoustic information that transmitted it from the speaker to the hearer. The greater the delay in categorization, the richer the precategorical sensory information that becomes available to a person, and the fuller the body of the sound patterns perceived. By the same token, however, the cognitive load of the information to be handled is greater too, and the period of uncertainty resulting from the exposure to uncategorized sensory information is longer; that is why so many persons are intolerant of delayed categorization.

As for the relative strength of the gestalts present in the poem, we must be aware of degrees of gestalt organization: there is a spectrum of stronger and weaker gestalts, with inarticulate, gestalt-free information at the weak end. In visual perception the presence of a gestalt-free ground--say, shading--lends "plasticity", or "depth-dimension", to a strong visual shape. The same principle seems to be at work in phonetic perception in a poetic context. Here, too, the presence of inarticulate, precategorical, sensory percepts seem to lend "plasticity", or "depth-dimension", to a well-articulated abstract phonetic category. However, the inarticulate, precategorical, sensory percepts gain greater or lesser relative weight, according to the context in which they occur. This observation seems to be true at all levels, from the smallest-scale phonetic context to the widest poetic context. Rakerd (1984) adduces carefully controlled experimental evidence suggesting that vowels in consonantal context are perceived more linguistically than are isolated vowels. In plain English this means that in isolated vowels one may perceive more precategorical auditory percepts than in consonantal contexts. One possible explanation for this is that vowels in consonantal context are subject to what speech-researchers call "parallel transmission"; in other words, "the talker often coarticulates the neighboring segments of an utterance (that is, overlaps their productions) such that the acoustic signal is jointly influenced by those segments" (Rakerd, 1984: 123). This finding is corroborated by Repp (1984) who discovered some cognitive strategies by which listeners can attend at will to the abstract phonetic categories of the fricative sibilants [s, S], or switch to the underlying precategorical auditory information. "The skill involved lay in perceptually segregating the noise from its vocalic context, which then made it possible to attend to its 'pitch'. Without this segregation, the phonetic percept was dominant".

In some poems the rich precategorical auditory information seems to have an especially strong impact on their emotional, sometimes uncanny, atmosphere, as well as on their musicality. Let us return to the first stanza of Verlaine's "Chanson d'Automne" (quote No. (3)). When we perform a poem like this, we choose from the beginning a delivery-style that would allow us a sufficient margin of freedom to manipulate the nasal backvowels prolonging and segregating subliminally their crucial portions so as to allow to perceive and register their "dark" quality, as well as their rich sonorous quality, before "lateral inhibition" by later arrivals may occur, or so as to proactivate and enhance the sound trace of similar sound patterns to arrive later; or, at least, to pronounce the nasal vowels in such a way, that there should remain a sufficiently perceptible portion of their acoustic signal that is not jointly influenced by the neighboring segments. This margin of freedom is not easy to achieve. It seems to be only possible in a general state of mind in which parallel cognitive processing is encouraged, whereas in connected speech there is a tendency to proceed linearly rather than move in different directions from the central sequence. As will be suggested below, in poetry one may distinguish a convergent from a divergent style; in the latter, perceptual and conceptual gestalts are considerably weaker than in the former. In order to allow for the disruption--subliminal though--of the linear sequencing of speech sounds (i.e., for the segregation of the relevant portions of the auditory stream), the whole message must be less thoroughly organized, on all levels, as in divergent style, where the linguistic stress pattern diverges from the conventional metric pattern, and so does the syntactic unit (clause, sentence) from the prosodic unit (line). In the world stratum of divergent poems we frequently find diffuse shape-free (and sometimes thing-free) entities rather than things that have stable characteristic visual shapes. In short, the freedom to adopt the cognitive strategy of segregating or integrating the crucial portions of the sound stream, so as to move back and forth between auditory and phonetic modes of listening, is at its fullest when the cognitive system is on no level of the poem under the control of some strong shape, definite direction or patent purpose. In such a context of relaxed shapes on all levels of a poem, the greater the divergence of the repeated sound clusters from strings of arbitrary verbal signs, the more they assume the emotive effects of nonreferential sound gestures (cf. Tsur, 1992a: 72-73).

Rhyme and Gestalt

(4) The Curfew tolls the knell of parting day,
    The lowing herd wind slowly o'er the lea,
    The plowman homeward plods his weary way,
    And leaves the world to darkness and to me.

(5) The Curfew tolls the knell of parting day,
    The plowman homeward plods his weary way,
    The lowing herd wind slowly o'er the lea,
    And leaves the world to darkness and to me.

No. (4) is the first stanza of Thomas Gray's "Elegy Written in a Country Church-Yard", No. (5) is the same stanza, with its rhyme pattern re-arranged. Semantically and thematically they convey quite similar information (the rearrangement of lines interferes with certain subject-predicate relationships, but this does not affect my argument). A comparison of the two, however, reveals a rather different tone. The second version is felt to be considerably simplified, it has a more straightforward tone. If we ask readers to apply one member of the adjective pair emotional ~ witty to one version and one to the other, more likely than not, emotional will be applied to the original version, witty to the re-arranged one. Gestalt theory is the most convenient tool to relate systematically such perceived affects to the structure of the texts. The fundamental law of perception, the Law of Prägnanz, is commonly defined by gestalt psychologists as follows: "The psychological organization of any stimulus pattern will always be as good as the prevailing conditions allow", followed as a rule by a list of conditions for "good". "The 'laws of organization', as formulated by Max Wertheimer, designate the conditions which maximize our tendency to respond to groups of individual stimuli as unified 'percepts'. These conditions include proximity and similarity" (Herrnstein-Smith, 1968: 41). In quote (5), the similar endings day - way and lea - me are in greater proximity than in quote (4); in quote (5), therefore, the prevailing conditions allow a better psychological organization of the stimulus pattern than in quote (4).

The regional quality of such rhymes as in quote (5) is typically perceived as 'witty', in the sense of "characterized by sharpness usually associated with cleverness and quickness of apprehension". This quickness can easily be accounted for, if we consider the following description of the two versions. In quote (5), the second a-rhyme ("way") immediately follows the first a-rhyme ("day"); and the same is true of the b-rhymes. Each couplet is quickly completed and left off. In quote (4), by contrast, the first member of each rhyme-pair must be kept in mind for longer before it can be completed, and is perceived as somewhat "slower". That is why the a-b-a-b rhyme pattern suits better such descriptions as "The lowing herd wind slowly o'er the lea, / The plowman homeward plods his weary way", whereas the a-a-b-b rhyme pattern would suit better, e.g., Pope's witty arguments (cf. below, quotes (10-13). The a-a-b-b rhyme pattern has a simple, "good" shape, since it groups together by similarity two verse lines in close proximity. The a-b-a-b rhyme pattern has a less simple shape, since the rhyme endings grouped together by similarity are less in proximity. In fact, in order to obtain two similar parts on a higher level, two dissimilar lines must be grouped together on a lower level: if on a lower level the a+b lines are grouped together, the whole stanza divides into two symmetrical parts, of exactly the same structure, ab. It should be observed that the a-b-a-b rhyme pattern is relatively complex only as compared to the a-a-b-b rhyme pattern, but is quite simple as compared to some other rhyme patterns (compare, for instance, quotes (6) and (7) below). Consequently, (5) is apt to break up into two quickly alternating couplets, whereas (4) constitutes one stable whole. In (4), then, it takes longer and more complex processing before a complete, closed whole can be achieved. That is why (4) is perceived as slower, less witty, perhaps more emotional.4

In poetry, sometimes it is more convenient to talk of strong and weak gestalts in terms of convergent and divergent structures. When a line ending converges with a sentence ending, they yield a stronger shape than when they diverge. Likewise, when metrical strong positions converge with linguistically stressed syllables and with patterns of alliteration, they yield stronger shapes than when they diverge in one way or other. To this, one may add some observations concerning line structure. An eight-position-long or twelve-position-long iambic line may be divided into two segments of equal length (four or six positions long) and of equal structure (each beginning with a weak position and ending with a strong position.)5 A ten-position-long iambic line, too, can be divided into two segments of equal length (5+5); the structure of these segments, however, will be unequal (one beginning and ending with a weak position, the other beginning and ending with a strong position). Alternatively, the line can be divided into two segments of unequal length (4+6 or 6+4), but of similar structures (each beginning with a weak position and ending with a strong position). Consequently, while the iambic tetrameter and hexameter can be balanced by a caesura at the middle, the iambic pentameter has only a rather vague region of balance, where a caesura may occur after the fourth, fifth or sixth position (the caesura after the fourth position being the unmarked, after the sixth, the marked caesura). Consequently, the tetrameter and the hexameter tend to fall into two symmetrical halves, whereas the pentameter tends to be perceived as more integrated.

Let us consider now two more complex examples:

(6) Farewell, farewell, but this I tell
    To thee, oh wedding guest:
    He prayeth well, who loveth well
    Both Man and Bird and Beast.

    He prayeth best, who loveth best
    All things both great and small;
    For the dear God who loveth us,
    He made and loveth all.
             (Coleridge, from "The Rhyme of the Ancient Mariner")

(7)             But Patience to prevent
    That murmur, soon replies: "God does not need
    Either man's work or his own gifts; who best         10
    Bear his mild yoke, they serve him best.
His state
    Is kingly: thousands at his bidding speed,
    And post o'er land and ocean without rest;
    They also serve who only stand and wait.
                         (Milton, from "On his Blindness")

Consider the italicized sentences in (6) and (7). They have three things at least in common. They convey a similar "moral", in both the word "best" is repeated, and they have a similar syntactic structure: a complex sentence in which the subject is a rank-shifted relative clause preceded by "he". In both, the word "best" receives unusually strong emphasis at the end of the line: in quote (6), because the word converges with the caesura and the line ending, yielding an unusually well-articulated, symmetrical, stable whole; in quote (7), because an unusually strong emphasis and prolongation on the word best is required, in order to indicate the line ending behind the "fluid" run-on sentence. The italicized run-on sentence in quote (7) has exactly the length and structure of a pentameter line, "straddled" between two pentameter lines. Its "movement" is exceptionally "impetuous", because it drastically upsets the formal line's balance: it does not begin and end in the regions of balance, but very near the end of the two lines concerned, always leaving a mere two-syllable-long chunk, which has, consequently, high "requiredness", that is, satisfies a strong need for completion. I have elsewhere compared in great detail these two passages (Tsur, 1977: 203-207). I wish to repeat here only one issue concerning the differences between the italicized sentences, and add a new point.

The placement of the rank-shifted clause in the sentence has a decisive effect upon the respective passages. "When a prediction remains unfulfilled for too long, the sentence becomes excessively laboured" (Johnson-Laird, 1970: 263). "Laboured sentences" and weak gestalts have one thing in common: both take up relatively large amount of mental space. The gestalt of Milton's sentence is rendered fuzzier than Coleridge's. In Milton's sentence, "Who" predicts a relative clause which, in turn, predicts a main clause in which it is to fulfil the function of a noun; secondly, within the relative clause, "who" predicts a verb related to it in a subject-predicate relationship. This expectation is prolonged by the interpolated "best". In Coleridge's sentence, the two clauses come in the reverse order: an independent clause followed by a relative clause; the former predicts, syntactically, no other one. All this increases the respective convergent and divergent characters of the two passages. The transitionary quality of Milton's sentence is all the more conspicuous, since one may contrast it to a similar (yet different) structure in the last line of the same sonnet: "They also serve who only stand and wait". This sentence is built 'on the model of' Coleridge's sentence: it converges with the line; the relative clause follows the independent clause (expectations are 'released'). So, the two comparable sentences mutually reinforce, by contrast, the transitionary quality of the earlier, and the firm closural quality of the later one.
In the present context, I wish to point out one more curious phenomenon. Some readers report that the word "best" is felt to be somehow fuzzier, less sharp at the end of line 10 in quote (7), than at the end of line 5 in quote (6). I used to think that this impression was due, mainly, to the sharpness or fuzziness, the convergence or divergence, of the gestalts in which the word "best" occurred. Though I still believe that there is some such relationship, now it seems to me quite unlikely that the impression of sharpness or fuzziness is simply displaced from the wider gestalt to the key-word "best".

As I have suggested above, there may be an actual difference in our articulation of the word "best"--a sharper articulation in quote (6), a more prolonged articulation in quote (7). This would, perhaps, afford easier access to the inarticulate precategorical sensory information in quote (7) than in quote (6). In the preceding section, I have discussed in some detail the relationship between abstract phonetic categories and the rich precategorical auditory information. I have suggested that we may be influencing unwittingly this relationship by adopting one or another cognitive strategy in our performance: the freedom to adopt the cognitive strategy of segregating or integrating the crucial portions of the sound stream, so as to move back and forth between auditory and phonetic modes of listening, is at its fullest when the cognitive system is on no level of the poem under the control of some strong shape, definite direction or patent purpose. The conclusive tone of the great religious truths enhanced by the repeated superlatives do arouse, in both passages, a psychological atmosphere of definite direction and patent purpose. However, while this psychological atmosphere is reinforced by the exceptionally strong prosodic shapes in quote (6), it is considerably mitigated by the exceptionally fluid and weak prosodic shapes in quote (7). Gestalt psychology in its discussion of figure-ground relationships has discovered certain principles of colour interaction that might be extended to sound, if we substitute "gestalt-free elements" for "colour". The interaction of colours or of gestalt-free elements does not work across the boundaries of strong gestalts. The weaker the gestalts, the stronger the interaction of colours or of gestalt-free elements across their boundaries. "The ambiguity of a weak figure on a strong ground immensely increases colour interaction" (Ehrenzweig, 1970: 172). Furthermore, "as in all relationship between form and colour the reverse effect can also happen. Strong colour interaction tends to make sharp outlines seem much softer than they are; it levels down differences in tone" (ibid., 171). If my suggestion of the analogy between colour and gestalt-free elements is valid, the ambiguity of the weak figures resulting from the syntactic and prosodic shapes blurring each other in quote (7) may immensely increase the interaction of gestalt-free, inarticulate precategorical auditory information underlying the word "best". This, in turn, may work the other way as well: to make the outlines seem even softer than they are. Such increased interaction of gestalt-free elements may be responsible for the fuzzier, plumper, fuller impression of "best" in quote (7) than in quote (6).

The grouping of lines by rhyme, and the relative strength of gestalts generated by such groupings may have a decisive affect in an additional issue, upon which I shall only touch here. I have already alluded to what Snyder called "hypnotic" or "spell-weaving" poetry. Regular metre gives security to what John Crowe Ransom calls the "Platonic Censor in us" (Chatman 1965: 212). A witty quality is generated when this security is genuine, a hypnotic quality, when this security is false. Metre gives security, because it is predictable; security is genuine, when other aspects of the poetic text, too, provide predictable orders or do not violate too drastically the reader's rationality or sense of reality. In Coleridge's "Kubla Khan", in Poe's "Ulalume" and throughout Alexander Pope's work, metre is more than usually regular. Disregarding matters of contents, one may suggest that Pope's work is perceived as witty because, among other reasons, his rhymes are in close proximity and yield perfectly predictable and symmetrical patterns; its versification arouses genuine security. But in the first stanzas of "Kubla Khan" (cf. Tsur, 1987a) and "Ulalume" (cf. Tsur, 1992b: 434-445), the versification arouses false security: their regular metre is coupled with unpredictable rhyme patterns (in Coleridge even with unpredictable line lengths), and symmetries, even when established, are frequently disturbed or even overturned by later or intervening lines. Hence their strong hypnotic effect.

In some instances, on the contrary, too much predictability of the rhyme pattern may arouse that feeling of uncertainty which, coupled with the certainty of predictable rhythm, generates the "spell-weaving", or "trance-inducing" affect of hypnotic poetry. In Tsur (1988a) I have discussed such a Hebrew poem, "Bircat Shoshanim" ("Rose Greeting") by Jonathan Ratosh. On the metric level, Ratosh's poem has an obtrusive, more than usually regular, ternary metre. The length of lines is regular in this poem, and thus strictly predictable. On the grouping level, however, this poem is much less predictable than, e.g.,"Kubla Khan". But it achieves this state of uncertainty by a device that is very different from that of "Kubla Khan": all of the twenty-four lines rhyme on one group of two speech sounds [et]. This succession of lines, with their undifferentiated rhyme pattern, has no self-generated grouping principle, and their formal division into eight-line-long stanzas is purely external and arbitrary. Such perceptual grouping is prone to arouse a feeling of saturation and strong cravings for change. Thus we may refer to two opposing aspects in the prosodic organization of Ratosh's poem. On the metric level and line level, it induces a marked feeling of security, whereas on the level of line-grouping it induces an unusually strong, almost chaotic sense of insecurity (cf. also Glicksohn & al., 1991, where we report an experimental investigation of the readers' response to this poem).

Rhyme and Meaning
In his paper "One Relationship between Rhyme and Reason", Wimsatt (1964: 153-166) points out the difference between Chaucer's rhymes and Pope's. Chaucer's are "tame" rhymes, in which the same parts of speech are used in closely parallel functions, as in

(8)  And he was clad in cote and hood of green.
    A sheef of pecock arwes, bright and kene,
    Under his belt he bar full thriftily,
    Wel coude he dresse his takel yemanly ...

Not so Pope, who achieves his witty effects, among other means, by rhyming, e.g., nouns with verbs, verbs with adverbs, in different syntactic positions, as in

(9)  Blessed with each talent and each art to please,
    And born to write, converse, and live with ease,
    Should such a man, too fond to rule alone,
    Bear like a Turke, no brother near the throne ...

Such rhymes are perceived as vigorous.Tame and vigorous describe the perceived effects of these rhymes. The remark that vigorous rhymes involve different parts of speech describes their structure. The vigorous effect of "anti-grammatical" rhymes (Roman Jakobson's term) can be explained in terms of a semantic information-processing model.

According to a prevalent--though by no means the only--present day-view of semantic representation, the meaning of words can be conceived of as bundles of elements that are frequently referred to as "semantic features", or "meaning components", or "primitive concepts". Thus, the word bachelor can be analyzed into the following set of components: [+NOUN +COUNT +CONCRETE +ANIMATE +HUMAN +ADULT +MALE -MARRIED]. The change of the sign of the last but one feature on this list to [-MALE] results in a meaning structure that can be realized in the surface word spinster. Deletion of the last feature in the above list results in a meaning structure that can be realized as the word man. If we delete the last feature on the above list and change the algebraic sign of the last but one feature, we get the meaning structure of woman, and so forth.

Some cognitive and linguistic manipulations are typically focused on the lowest item in the list of features. When one says "That person is a bachelor", one usually asserts of that person only that he is unmarried, not that he is a human being, or an adult, or a male (although one specifies, by the same token, "that person" as adult and male). Similarly the negative: "That person is not a bachelor" negates only the lowest feature: saying "That person is not unmarried", while still specifying the person as adult and male. Assertion and negation appear to concern only a single--the lowest--feature of nouns.

Likewise, in the word association game, according to Clark (1970: 276-277), the stimulus word man elicits woman 62 percent of the time (resulting from the change of the lowest semantic feature [+MALE]), but elicits boy only 8 per cent of the time (resulting from the change of the lowest-but-one feature [+ADULT]); it elicits girl only 3 percent of the time (resulting from changing the two lowest features). Thus, woman is obviously the unmarked response to man, whereas boy and girl are marked to varying degrees. Clark's findings can be accounted for by assuming that the responses obtained are produced according to the simplicity-of-production rule: "Perform the least change on the lowest feature, with the restriction that the result must correspond to an English word [...] Unsuccessful applications of simpler rules therefore force people to use more and more complex rules" (Clark, 1970: 280-281). Eventually, this amounts to the principle of exerting minimum effort. Anti-grammatical rhymes are produced contrary to the simplicity-of-production rule: 'Change as many features as possible, as high on the list as possible',6 resulting in the assertion of maximum effort--hence its vigorous effect.

The semantic structure of the rhymes in (9) is, then, anti-grammatical; its perceived effect is vigorous and witty. Returning now once more to the first stanza of Verlaine's "Chanson d'Automne" (quote (3)), we may notice that most of its rhymes, too, are anti-grammatical: longs is an adjective, violons a noun; l'automne is an abstract noun, whereas monotone is an adjective. Even coeur and langueur, both of which are nouns, are contrasted in a feature fairly high on the list: [+/-CONCRETE]. Here, however, the perceived effect is quite different. The rhymes enhance a certain "thing-free" atmosphere, and an intense, vague feeling rather than sharp wit. The semantic features of the anti-grammatical rhyme are well articulated in awareness, owing to the oppositions pointed up by it. In a divergent stylistic context, these features are diffused among the other diffuse elements. When the same heterogeneous semantic elements are 'yoked' together in an environment of convergence and strong shapes, such rhymes usually have a 'witty' regional quality, 'witty' used, as above, in the sense of 'sharpness' usually associated with cleverness and quickness of apprehension. This quickness, then, can be accounted for by the properties of strong shapes.

At this point, I propose to draw together my discussions of phonetic and semantic information-processing, via Wimsatt's treatment of the homoeoteleuton:

It would be only an exaggeration, not a distortion, of principle to say that the difference between prose and verse is the difference betwen homoeoteleuton and rhyme. "Non modo ad salitem ejus exstinguendam sed etiam gloriam per tales viros infringendam", says Cicero, and Quintilian quotes it as an exmple of homoeoteleuton or like endings. Here the -endam and the -endam are alike, logically and legitimately alike; each has the same meaning, or is the same morpheme, and each supports the logic of the sentence by appearing in a certain place in the structure. Stylistic parallels or forms of meaning of this sort seem to come fairly to the aid of logic; they are part of the normal framework of the prose (Wimsatt, 1954: 153-1540).

When, in prose, not only the endings but also the roots rhyme, the result is an effect of alogicality, if not of excess and artificiality. Of somewhat the same character is the cursus of metrical ending.

And if a prose writer were to reinforce a pair of parallel or antithetic clauses by making each one an iambic pentameter, we should say that this was decidedly too much, that the metrical equality was hardly interesting unless it combined with a vein of logic that ran differently (ibid, 154).

I should like to relate this discussion, and its sequel, the discussion of "tame" and "vigorous" rhymes, to the above discussion of codings used by good and poor readers. The experiment by Byrne and Shea has tapped two kinds of codings, semantic and phonetic, exploited for such cognitive activities as reading, or keeping words in active memory. Similar-meaning words reinforce the use of the semantic code; similar-sound words, the use of the phonetic code. In ordinary speech, the use of the phonetic code is "transparent"; it is exploited for the efficient use of short-term memory, but no conscious or half-conscious attention is paid to it. In literary use, some attention is shifted to it: we do acknowledge its affect on the whole, but are hard put to identify its source. Since, however, we use language as a rule in order to convey meanings rather than mere sounds, semantic coding does have a certain primacy over phonetic coding, even in literary language. Whenever possible, we tend to foreground semantic coding; only when something seems to "go wrong" with the semantic coding, we tend to shift our attention to the phonetic coding. When, in Wimsatt's quotation from Cicero, the like endings "each has the same meaning, or is the same morpheme, and each supports the logic of the sentence by appearing in a certain place in the structure", that is, "stylistic parallels or forms of meaning of this sort seem to come fairly to the aid of logic", the semantic and syntactic coding appears to be entirely satisfactory, and the readers or listeners seem to feel little need to attend to the phonetic coding. Since, however, like endings have similar sounds as well, the readers or listeners do direct increased attention to them, but use this similarity merely to reinforce the similar meanings. Not so with rhymes. The greater the difference in the meanings of the rhyming words, the more the readers or listeners are inclined to shift attention to the phonetic similarity. In anti-grammatical rhyme, the difference of meanings is, by definition, greater than in grammatical rhyme; so, both the semantic and phonetic representation are more active in perception.

In view of the foregoing analysis, we may distinguish three significant kinds of relationship between the semantic and the auditory-phonetic information in literary language. Paraphrasing a distinction in ecological acoustics, rhyme is resonant, homoeoteleuton is thunk (Graver, 1993: 297). In homoeoteleuton, semantic and logical relationships are dominant, to which the sound-information is subordinated; it is perceived as rather flat, compact, relative to the other possibilities. In rhyme, by contrast, the sound-information becomes relatively loose, both in the sense of released to some extent from its attachment to meaning, and in the sense of less closely packed together. It reverberates in echoic memory, and the whole is perceived as more spacious, more plastic, having a fuller body. In this case we have two possibilities: "when the cognitive system is on no level of the poem under the control of some strong shape", and when it is on some level(s) of the poem under the control of strong shape(s). In the latter case, in Pope's poetry, for instance, the regular metre, the symmetrical couplet form, and the parallelisms are just such strong shapes that exert rigorous control over the cognitive system, and considerably tone down the impact of the rich precategorical auditory information. As a result, the auditory information remains subordinated to the abstract phonetic categories, generating a "figure-ground" relationship, which renders the perception of the string of abstract phonetic categories more plastic, bestowing upon it, so to speak, depth-dimension. In the former case, in Verlaine's "Chanson d'Automne", for instance, the rich precategorical auditory information may get out of control, reverberate at large, and tend to assume the emotive effects of nonreferential sound gestures. By the same token, attention tends to be shifted away from the meaning to the sound of the poem. By analogy with visual perception, the pre-categorial information may be assumed to behave differently at the two levels of rhyme: as long as the abstract phonetic category is in control, the interaction between the minute, inarticulate sound-percepts is contained within the limited area of the speech sound; when the pre-categorial sensory information bursts the constraints of organising shapes, such interaction takes place over considerable areas of the poem, and at great intensity levels. This is a phonetic reason for Pope's rhymes to sound so active, but also so focussed.

In what follows, I shall contrast the above conception of the relationship between rhyme and meaning to another, rather widespread one, that seems to arise from some critics' uneasiness with the need to handle issues related to sound patterns and rhyme schemes. In many instances, critics try to improvise some ad hoc semantic feature for the specific rhyme pattern. I propose now an (admittedly unduly brief) consideration of such an instance (Jones, 1969), in order to see whether the semantic and gestalt tools proposed here can handle the instance discussed by Jones more adequately. Let us consider a few couplets from Pope (all italics are Pope's):

(10)  Our plenteous Streams a various Race supply;
      The bright-ey'd Perch with Fins of Tyrian Dye,
      The silver Eel, in shining volume roll'd,
      The yellow Carp, in Scales bedrop'd with Gold,
      Swift Trouts, diversify'd with Crimson Stains,
      And Pikes, the Tyrants of the wat'ry Plains.
                                ("Windsor Forest", 141-146)

(11)  Yet let not each gay Turn thy Rapture move,
      For Fools Admire, but Men of Sense Approve.
                                ("An Essay on Criticism", 390-391)

(12)  Others for language all their care express,
      And value books, as Women Men, for Dress.
                                ("An Essay on Criticism", 305-306)

(13)  Sole judge of Truth, in endless Error hurl'd:
      The glory, jest and riddle of the world.
                                (Pope: "An Essay on Man", II. 17-18)

(14)  Such if there be, who loves so long, so well,
      Let him our sad, our tender story tell;
                                ("Eloisa to Abelard", 363-364)

John A. Jones, in his book on Pope's couplet art, quotes (10), above, and makes the following comment on its third line:

Because the participle "roll'd" is the rhyme word, the verb quality of "rolling" is emphasized rather than adjectival or substantive quality. "Shining volumes" is more effective coming before the rhyme "roll'd" than it would be after it, for it is the climactic rolling or writhing that is highlighted. We do not always think of volumes as round but here it means "coils"; and when "roll'd" describes "volumes", the eelish quality is heightened, as the reader can easily imagine, even if he has never landed an eel (Jones, 1969: 74-75).

Consider the first sentence of this quotation from Jones. The word because suggests a logical, causal relationship between its two clauses. But is there one? To justify such a statement, there must be some generalization that can be consistently maintained, such as, for instance, "When a participle occurs in the rhyme word, its verb quality is emphasized rather than adjectival or substantive quality". I am not aware of any such valid generalization. In fact, all the available grammatical and stylistic evidence suggests that the adjectival quality is emphasized in this epithet. Jones's interpretation, however, crucially seems to depend on the participle's "verb quality". Since, however, everybody feels that rhymes do something important to words, and so little is known about what they do to them, Jones quite safely resorts to the rhyme word to enlist it in the service of the "verb quality" construal of the participle. When words with certain meaning components are systematically manipulated into the rhyme, one can, perhaps, make a case for its significance; but even that cannot justify any generalization of this kind. Speech sounds are arbitrarily assigned to meanings in natural languages. Versification is, typically, an additional organization of the phonological component, irrespective of the meanings of the sounds. Whenever the critic claims that there is some interaction between sound and meaning, he must make explicit the principles on which he is relying.

Syntactic inversion; may be an effective foregrounding device, unless there are factors that tend to void it. In some cases, however, it may be reasonably supposed that the poet used his conventional right to syntactic inversion merely to make his words conform with meter and rhyme. In "shining volume roll'd", there may be room for just such 'reasonable supposition'. But, sometimes, there is more to it.

Consider (13), above. The antithesis leaves little room for doubt that Error in the first line is a word of key importance. Although hurl'd contributes such components to the image as helplessness, passive endurance and unstability, its decisive component, physical transfer, has little relevance to the thought expressed by the antithesis. Hurl'd constitutes a virtuoso rhyme with world, but this has been achieved at the double price of an "inelegant" syntactic inversion and manipulating the word of key importance out of the rhyme. Is it possible that a great master of poetic technique like Pope should be guilty of such incompetence? And how could we explain, then, that precisely this "incompetent" line constitutes one of the most famous couplets on which Pope's reputation as a major poet rests? An alternative explanation would be that the syntactic inversion, the virtuoso rhyme, and the manipulation of the key word out of the rhyme serve one common effect. As the random collection of couplets in (11-14) may suggest, such inversions, manipulating words of little importance into the rhyme, are not uncommon in Pope. It will be noticed that in (11-13) the inversion occurs at the end of the first line of the couplet, whereas in (14) it occurs at the end of the second line. I shall argue that this is quite significant.

Consider (11). Here, "move" provides the finite verb and the idea of causation only; greater interest lies in the direct object Rapture, and even greater in the subject, gay Turn. In short, the syntactic inversion maneuvers precisely the nonemphatic verb into the rhyme. At the same time, the antithesis fools ~ Men of Sense focuses attention on the verbs Admire and Approve. These two verbs are near-synonyms in that they express a positive attitude; and they are near-antonyms in that they express uncritical enthusiasm and sound judgment, respectively. This semantic and rhetorical structure is the main source of this couplet's wit. The witty effect of the verb Approve is greatly enhanced here by its high degree of requiredness at this place. This verb is required here owing to its place in the rhyme pattern, in the antithesis and in the segmentation of the line. So, in light of the foregoing gestaltist discussion, one might suggest that the syntactic inversion at the end of the first line is functional: the structure of the first line is weakened sufficiently to make the part dependent on, and therefore integrated with, its context. In short, the weaker the ending of the first line, the stronger the structural closure perceived at the end of the couplet. One of the most effective means of amplifying the closure of a piece of poetry, at various levels of organization, is to weaken the closure of the unit before, and thus make it more dependent on the whole.

Likewise, in (12) the inversion highlights the noun phrases language and all their care; the least emphatic member of the clause, the verb express, is again dislocated and placed into the rhyme. Great attention is focused, again, on Dress in the rhyme of the second line, by a variety of means: requiredness arising from segmentation, the rhetorical scheme (zeugma) and the convergence of two meanings in the word Dress: 'guise, appearance, adornment' and 'fine cloths'. A similar story can be told of the first line of (13), even though in the second line attention is not focused with such vigour on the rhyming word.

Contrary, then, to Jones's contention, in all these instances it is not the rhyme that effects the meaning of, and bestows emphasis on, the word that is manipulated into it, but rather the other way around: the relatively low semantic importance of the word manipulated into the line ending de-emphasizes the first rhyme. The process is governed by the gestalt principles discussed in the present paper, and is geared to "the conditions which maximize our tendency to respond to groups of individual stimuli as unified 'percepts'".

We may observe a very different effect in (14), where the syntactic inversion occurs at the end of the second line. This couplet is perceived as "softer" than the instances discussed so far. This effect is achieved by a variety of means; but only the syntactic inversion at the end of the couplet concerns us here. "Eloisa to Abelard" is very different from Pope's other major poems. Here, the witty effect is replaced by a pervasive emotional tone, quite unusual in Pope. This emotional tone is supported by a relatively divergent structure. The drastic weakening of the closure at the end of the couplet is an effective means of weakening the overall gestalts within the rigid constraints of Pope's poetics.

Dactylic Rhyme, Intertextuality and Cognitive Poetics
In the present section I am going to focus attention on the placement of stress in rhyme, consider at some length the perceived affects of dactylic rhyme in various stylistic contexts, and discuss the relative merits of cognitive poetics and the historical approach of intertextuality in accounting for these qualities.

Regarding the placement of stress, one may distinguish for our purpose three types of rhyme: masculine (with the stress on the last syllable of the line), feminine (with the stress on the last but one syllable), and dactylic (with the stress on the last but two syllable). The third kind is very rare in English poetry, and may be illustrated by the following examples:

In his excellent study of the dactylic rhyme, Uzi Shavit (1993, 20-21) says as follows:

Most English handbooks of prosody associate the dactylic rhyme with light and comic verse; and this is right, undoubtedly, from the statistic point of view. However, does this mean that the comic quality is inherent in the English dactylic rhyme? [...] At least one famous example contradicts this assumption [...]: Thomas Hood's "The Bridge of Sighs". This poem [...] is conspicuously touching, pathetic. Still, about one third of the poem's verse-lines end with a dactylic rhyme; this kind of rhyme is, undoubtedly, one of the poem's most outstanding characteristics. Consequently, some contradiction ought to arise between the pathetic theme and message of the poem, and the comic character of its rhyme (all translations from Shavit are mine--R.T.).

However, the opposite appears to be the case. Shavit quotes Edgar Allan Poe, who in his "The Poetic Principle" says about this poem (quotes (18), (19)): "The vigour of this poem is no less remarkable than its pathos. The versification, although carrying the fanciful to the very verge of the fantastic, is nevertheless admirably adapted to the wild insanity which is the thesis of the poem" (ibid, 21). Likewise, Emil Legouis suggests that Hood "found for these poems, and especially for the second of them ["The Bridge of Sighs"], most original metres and rhymes, which heightened their pathos" (ibid.). Shavit concludes that the relationship which the English tend to find between the dactylic rhyme and low comedy is not inherent in the English dactylic rhyme, but is derived from the relative scarcity of this kind of rhyme in English. As a result of this scarcity, there is a tendency to associate these rhymes with the central texts in which they occur; and the central text in which there are more dactylic rhymes than in any other English text is Byron's Don Juan (quotes (15), (16), (17)). There is no doubt either that in Byron, the dactylic rhymes have a pointedly comic affect.

Later, in an attempt to explain the apparent contradiction between the comic and pathetic affects of the dactylic rhyme, Shavit adopts one of my own suggestions, namely that this apparent contradiction is derived from the high markedness of the English dactylic rhyme. Markedness is a double-edged phenomenon; that is, in certain contexts it may generate some heightened witty, comic quality, whereas in other contexts--some emotional quality. The prosodist's task is, accordingly, to point out the conditions in which markedness contributes to one or another quality. According to this conception, this requires a cognitive theoretical framework, that would enable an examination of the unique perceptual characteristics of the various prosodic patterns (ibid., 25).

Whithout objecting to this line of investigation, Shavit prefers to consider the opposite affects of the dactylic rhyme within a different, historical framework of theory, insisting on intertextuality as the basic characteristic of literary texts and the source of the afore-mentioned contradiction. (Later, in a personal communication, he suggested that the two approaches should perhaps be regarded as complementary; in what follows, I shall explore this possibility.) Shavit goes on:

If we view the English dactylic rhyme in this context, the comic use made by Byron and Frere of the dactylic rhymes appears directly to be related to the peculiar coupling of the burlesque epic with the ottava rima, that is, to the Italian model of Luigi Pulci and other Italian burlesque poets from whom Frere and Byron adopted not only the ottava rima, but also the comic tone and the familiar and light style characteristical of [...] "Beppo" and Don Juan (ibid., 25).

Commenting on Hood's poem, Shavit says that it has no relation whatever to Byron's epic. The metre of Don Juan is the iambic pentameter, its strophic form is the ottava rima, whereas "The Bridge of Sighs" is written in the dactylic dimeter and an irregular strophic form. The combination of the dactylic dimeter with the dactylic rhyme, says Shavit, points in a different direction, to the angels' choirs in Faust; the following is a very convincing example from Part One (lines 801-807):

To reinforce his conception, Shavit makes two further points: that the association of the dactylic rhyme with Byron's Don Juan seems to have been established only gradually, and did not affect yet Thomas Hood or his readers and critics; and that Hood left England in 1835, staying on the Continent for five years, first in Germany, then in Belgium. In Germany he was supposed to have been exposed to Goethe's Faust.

Now let us imagine a present day native speaker of English who spends much time in reading nineteenth century English poetry, but does not know that in 1835 Hood spent some time in Germany; knows no German or Italian, and has never read Faust or Italian burlesque epics in the original languages or in English translation. What affect may we suppose him to perceive in English dactylic rhyme? More likely than not, he will perceive a comic affect in Byron's and a passionate affect in Hood's rhymes. I am not trying to extol an unsophisticated person as my prototypical reader. Rather, I wish to point out the weaknesses of the historical approach by bringing its argument ad absurdum. If we can credit a reader such as the one described above with an ability to perceive a comic affect in Byron's and a passionate affect in Hood's rhymes, we need not necessarily assume that a more sophisticated reader must be exposed to Luigi Pulci and to Goethe, in order to be capable of perceiving the respective qualities in Byron and Hood.

If we can grant as much, we must look elsewhere for an explanation; and a cognitive explanation seems to me as good as any other. But first, if we can grant as much, we must postulate one of three possibilities: that the comic or affectionate affect of the poetic passage is determined by the contents, and the dactylic rhyme has no influence whatever on it; or, that the dactylic rhyme amplifies any perceived affect in the text generated by the contents; or, that the dactylic rhyme has some comic or passionate potentials, which may be realized in certain stylistic contexts, the explicit contents being part of these contexts. It is assumed here that evidence for the second possibility supersedes the first one; and evidence for the third possibility supersedes the first two. In what follows, I shall explore the third possibility, within a cognitive frame of reference.

The dactylic rhyme is highly marked in English poetry, for three different reasons. One reason is that in the English language it is more difficult to produce it than in some other languages, hence its rarity. The other two reasons seem to be less language-dependent. The second reason has to do with the grouping of unstressed syllables with stressed ones. For unstressed syllables to achieve a sense of stability, they must be grouped together with an adjacent stressed syllable in a metrical strong position, and it is easier to group them forward than backward. Backward grouping is always perceived as marked. In the dactylic rhyme, the last two unstressed syllables must be grouped backward since, by definition, they are not followed by a stressed syllable in a strong position. What is more, when the dactylic rhyme occurs in a ternary metre (dactyl, anapaest or amphibrach), two weak positions are grouped forward, and two backward, with the same (last) stressed syllable in a strong position (as opposed to the binary metres, in which between any two strong positions there is only one weak position). This may generate a kind of uneasy feeling of perceptual ambiguity and confused directions. The third reason has to do with the acoustic cues for grouping and stress. According to Woodrow's experiments (Meyer, 1956: 106-107; Chatman, 1965: 26-27; Tsur, 1977: 88-89), non-linguistic sound stimuli are perceptually grouped into end-accented groups when there are differences of duration, and into beginning-accented groups when there are differences of amplitude. Differences of pitch do not affect the direction of grouping. There is some indication that this may be the case with linguistic stimuli too. Furthermore, at variance with the layman's intuition, linguistic stress is cued by a mixture of pitch, duration and amplitude (loudness), in a decreasing order of effectiveness (Frye, 1958). In natural speech in languages that are not syllable-timed, there are considerable differences of duration between syllables; this would render beginning-accented prosodic units more marked than end-accented ones, and dactylic rhymes especially marked. In syllabo-tonic metre in particular, beginning accented units (cued by amplitude) would be more marked than end-accented ones (cued by duration).

In view of the foregoing observations, the binary, end-accented, iambic metre would be perceived as less marked, more flexible, than either the binary, beginning-accented trochaic, or the various ternary metres (in which two weak positions are grouped with an adjacent strong one). Among the ternary metres, the beginning-accented dactylic would be perceived as more marked, less flexible than the rest. Accordingly, the middle-accented amphibrach is the only metre in which pitch differences, the most effective acoustic cue for stress, can be the dominant one, and should be perceived as the most easily flowing metre. As a matter of fact, many critics of many ages and languages have commented on the perceptual qualities of the various metres in this vein, though they offered quite different explanations for them (cf. Tsur, 1977: 83-96). One more observation concerning line length is needed here. From this point of view, the iambic pentameter line seems to be the most flexible one: shorter lines are perceived as more rigid; and longer lines tend to break up into two or more smaller (rather rigid) segments.

When comparing, then, the metres of Byron's Don Juan and Hood's "The Bridge of Sighs", one should note, first, that the former's metre is the least marked one of available metres, the latter's the most marked one; and second, that in respect of line length, too, Hood's verse is more marked than Byron's.

We have observed that the dactylic rhyme is a highly marked prosodic form and, as a result, it tends to have a witty or comic perceived affect. This is its "unmarked" affect. But it may have a more marked affect as well, which may be described as pathetic, passionate, or even hypnotic. As I have suggested above, regular metre gives security to the "Platonic Censor in us". A witty quality is generated when this security is genuine; a hypnotic quality, when false. In Byron's epic, the marked dactylic rhyme is embedded in a relatively unmarked prosodic environment that has considerable flexibility. In Hood's verse we are confronted with a conspicuous instance of "false security", so characteristic of hypnotic poetry. The sense of more than usual security is generated by the rigid dactylic dimeter; the sense of more than usual insecurity is generated by the unstable dactylic rhyme, on the one hand, and the unpredictable sequence of rhymes, on the other. If Poe is right in discerning what he calls "the wild insanity which is the thesis of the poem", this may be another ingredient that contributes to its sense of insecurity.

That is why a reader who is not equipped with wide erudition in literary history may intuit, in spite of his ignorance, a comic quality in Byron's dactylic rhyme, and a pathetic or hypnotic quality in Hood's. This capability of "switching" from one set of aspects to another of the dactylic rhyme corresponds to the ability suggested by Wittgenstein (1976: 214e) to "understand the request to pronounce the word 'till' and to mean it as a verb" as opposed to, say, an adverb). It is this mental ability that underlies our capability of offering a variety of more or less legitimate interpretations of the same text, or of associating a variety of perceived qualities with the same prosodic structures. Intertextuality can reinforce one or another such intuition that arises from the cognitive processing of individual poems. Where there are no such intuitions based on cognitive processing, intertextuality remains a mere phantom, and merely transfers the mystery from one place to another.

To conclude this section on dactylic rhyme, I wish to comment on an additional issue raised by Shavit. One of the most impressive appearances of the dactylic rhyme in modern Hebrew poetry (in the Ashkenazi pronunciation) is in a Zionist ode "The Blessing of the People" () by the great Hebrew poet Bialik.

As Shavit convincingly demonstrates, this poem became a model for a long line of rightly-forgotten odes, all displaying a feast of dactylic rhymes. Shavit (ibid., 32-33) quotes the literary historian Lachover on this poem:

Our poet too built his hymn on a very artistic construction that arouses security and pronounces vigorous commands but, at the same time, arouses cravings and longings. Its rhymes in the even-numbered lines are short and vigorous; in the odd-numbered lines, by contrast, they are long and drawn out [...] This makes them most drawn out and most expressive of the innermost hope and hidden cravings.

This passage, which I have drastically abbreviated, has something of the impressionistic chatter about it; and the prosodist Y. Bakoon objects: "this is a result of the impression from the poem's contents" (ibid.), rather than from the rhythmic structure of its rhymes. However, "the cravings and longings" suggested by Lachover may have some structural sources too: when attempting to account for the emotional affect of weak shapes in his brilliant application of gestalt theory to music, Leonard B. Meyer comments: "The lack of distinct and tangible shapes and of well-articulated modes of progression is capable of arousing powerful desires for, and expectations of, clarification and improvement" (Meyer, 1956: 160). According to our foregoing analysis, the dactylic rhyme is characterized by just such a "lack of well-articulated modes of progression" that "is capable of arousing powerful desires for clarification and improvement". The quote from Lachover may also suggest, perhaps in spite of the writer, the reason why such an ode has more of an assertive ceremonial character than a pathetic or hypnotic one. Only the dactylic rhyme is left to arouse a feeling of uncertainty, whereas from the contents' point of view, the "artistic construction arouses security and pronounces vigorous commands".

My handling of Bialik's dactylic rhyme and Lachover's comment on it, and my appeal to Meyer's treatment of weak shapes can be taken as an instance of how Cognitive Poetics offers cognitive hypotheses to relate systematically "the particular regularities that occur in literary texts" and "the specific effects of poetry". Lachover describes the poem's rhyme structure, that includes a systematic use of dactylic rhyme; he also suggests that this makes the rhymes "most expressive of the innermost hope and hidden cravings". Bakoon objects that these cravings and longings have little to do with poetic structure. My cognitive analysis of dactylic rhyme and Meyer's characterization of weak gestalts are cognitive hypotheses to relate systematically the particular regularities that occur in literary texts and the specific effects of poetry.

I have tried to give a comprehensive cognitive view of rhyme. Rhyme is one of the most powerful resources of poetic language. Readers and critics have strong intuitions about it, but have difficulties in talking about it and its contribution to the total affect of a poetic passage. So, very frequently, critics ignore rhymes in their writings, or merely point out the rhyme pattern of a poem. Other critics have recourse to impressionistic effusion in talking of rhymes. Still other critics rely on some ad hoc semantic analysis, hoping that the reader would discover the relevance of such an analysis to the poem's perceived affect; or resort to some argument based on intertextuality, sometimes merely transferring the mystery from one place to another. The present paper has attempted to reveal the sources of the possible affects of rhyme, and to offer critical tools to talk about rhyme in a meaningful way, and to relate systematically its affects to its structure.

Speech sounds are abstract categories, from which the rich precategorical sensory information is typically stripped away. Some of this information does reach, nevertheless, the cognitive system, reverberating for a very short period of time in short-term memory, and facilitating certain cognitive tasks with reference to verbal material. Rhyme exploits and enhances this sensory information. There is some experimental evidence that the memory traces of two words considerably apart in time may be fused and perceived as if simultaneously present. In certain conditions, such fusion of auditory information may be perceived as if spread over the intervening section of the poetic passage. We all are familiar with figure-ground relationships in visual perception, as well as in music. In certain circumstances well-defined by gestalt psychology, inarticulate ground may bestow depth-dimension and plasticity upon good shapes. Gestalt psychology has discovered, on the other hand, that when gestalts are considerably weakened, the interaction of colour and other gestalt-free sensory elements may be enhanced across their boundaries. It has been suggested above that similar processes may take place in the relationship between the phonetic categories and the underlying acoustic information. Some cognitive mechanisms related to the performance of poetry were suggested, that may enhance these processes (e.g., Repp's discovery concerning the segregation or integration of the fricative sibilants [s, S], or generating an attitude in which the cognitive system is on no level of the poem under the control of some strong shape, definite direction or patent purpose). Also, with reference to, e.g., Omar Khayyám's rumble of a distant Drum, suggestions were made concerning the possible interaction of semantic or thematic features with acoustic information underlying speech sounds. Some of those gestalt laws were considered that designate the conditions which maximize our tendency to respond to groups of individual stimuli as unified 'percepts' (e.g., grouping by similarity or proximity, or the need to weaken the part in order to make it dependent on the whole) and which, by the same token, may systematically account for the perceived qualities regularly associated with certain rhyme patterns. Considerable attention has been devoted here to the relatively rare dactylic rhyme, in an attempt to account for the conflicting perceived affects regularly associated with it.

The general canons for the evaluation of a poetic passage are unity, complexity, and some intense human quality pervading it. The greater the unity, complexity, or intense human quality of a poem, the greater its aesthetic worth is said to be. As for complexity, rhyme superimposes an organizing pattern upon a string of words, in addition to the semantic, syntactic, and metric patterns. By the same token, this additional organizing pattern enhances unity; and the very raison d'être of the gestalt principles considered above is to maximize our tendency to respond to groups of individual stimuli as unified 'percepts'. At the same time, by turning the phonetic and acoustic resources of rhyme along with the grouping gestalts to aesthetic ends, we shift the focus of our attention from the effiency of performance in certain cognitive tasks to such intense human qualities as, e.g., "witty" or "emotional".

Finally, I wish to end with a caveat. There is no way to predict the perceived affects of a certain rhyme pattern (or of any other poetic structure). The affects must be accepted as given, as directly perceived by the reader. It is only after the event that one may account for them with structural features of the text which, to use Sibley's (1962) term, "typically count toward them".


1. The present section massively draws upon chapters 1 and 2 of my book (Tsur, 1992 a).

2. In the present state of knowledge it is not clear at all whether there is any correlation between low absorption and a preference for a semantic rather than phonetic code.

3. Another versification device depends even more crucially on this characteristic of short-term memory. I mean the verse line, which typically is experienced as a simultaneously presented unified percept. We know, however, that the words of a verse line are presented not simultaneously, but one after the other. As I have elsewhere argued at great length (Tsur, 1977), one indispensable feature of the rhythmical performance of a verse line is that the performer must manipulate the verbal material in such a way that the unit should be completed before the memory trace of its beginning fades out in echoic memory.

4. For further discussions of rhyme patterns and gestalts see Tsur et al., 1991; Tsur, 1992 b: 111-131.

5. While the linguistic units of syllabo-tonic metre are unstressed and stressed syllables, the abstract verse line consists of regularly alternating weak and strong positions.

6. I have elsewhere (Tsur, 1983b) discussed at much greater length this conception of rhyme and semantic information-processing.


Brady, Susan, Donald Shankweiler & Virginia Mann
1983 "Speech Perception and Memory Coding in Relation to Reading Ability". Journal of Experimental Child Psychology 35: 345-367.

Chatman, Seymour
1965 A Theory of Meter. The Hague: Mouton, Janua Linguarum.

Clark, Herbert H.
1970 "word associations and Linguistic Theory", in John Lyons (ed.), New Horizons in Linguistics. Harmondsworth: Pelican. 271-286.

Conrad, R.
1964 "Acoustic Confusions in Immediate Memory". BJPsych 55: 75-84.

Crowder, Robert G.
1982 "A Common Basis For Auditory Sensory Storage in Perception and Immediate Memory." Perception & Psychophysics 31: 477-483.

Crowder, Robert G, and Richard K. Wagner
1992 The Psychology of Reading. New York and Oxford: Oxford UP.

Ehrenzweig, Anton
1970 The Hidden Order of Art. London: Paladin.

Frye, D. B.
1958 "Experiments in the Perception of Speech". Language and Speech 1: 126-151.

Glicksohn, J., R. Tsur and Ch. Goodblatt
1991 "Absorption and Trance-Inducing Poetry". Empirical Studies of the Arts 9/2: 115-122.

Graver, William W.
1993 "How Do We Hear in the World?: Explorations in Ecological Acoustics". Ecological Psychology 5: 285-313.

Herrnstein-Smith, Barbara
1968 Poetic Closure. Chicago: Chicago UP.

Johnson-Laird, P. N.
1970 "The Perception and Memory for Sentences", in John Lyons (ed.), New Horizons in Linguistics. Harmondsworth: Pelican. 261-270.

Jones, John A.
1969 Pope's Couplet Art. Athens: Ohio UP.

Liberman, Isabelle Y., and Virginia A. Mann
1981 "Should Reading Instruction and Remediation Vary with the Sex of the Child?" Status Report on Speech Research SR-65: Haskins Laboratories.

Mann, Virginia A.
l984 "Reading Skill and Language Skill". Developmental Review 4: 1-15.

Meyer, Leonard B.
1956 Emotion and Meaning in Music. Chicago: Chicago UP.

Rakerd, Brad
1984 "Vowels in Consonantal Context Are More Linguistically Perceived than Isolated Vowels: Evidence from an Individual Differences Scaling Study". Perception & Psychophysics 35: 123-136.

Repp, Bruno H.
l984 "Categorical Perception: Issues, Methods, Findings", in N. J. Lass (ed.), Speech and Language: Advances in Basic Research and Practice, vol. 10. New York: Academic Press. 243-335.

Shavit, Uzi
1993 Rhyme and Reason: Studies in the Historical Poetics of Hebrew Poetry. Jerusalem: Bialik Institute (in Hebrew).

Sibley, Frank
1962 "Aesthetic Qualities", In Joseph Margolis (ed.), Philosophy Looks at the Arts: Contemporary Readings in Aesthetics. New York: Scribner. 63-88.

Snyder, E. D.
1930 Hypnotic Poetry: A Study of Trance-Inducing Techniques in Certain Poems and its Literary Significance. Philadelphia: University of Pennsylvania Press.

Tsur, Reuven
1976 "Antigrammatical Rhymes and Linguistic Competence" (in Hebrew). Hassifrut 22: 59-62.

1977 A Perception-Oriented Theory of Metre. Tel Aviv: The Porter Israeli Institute for Poetics and Semiotics.

1983 a Poetic Structure, Information-Processing and Perceived Effects: Rhyme and Poetic Competence (in English). Tel Aviv: The Katz Research Institute for Hebrew Literature.

1983 b Meaning and Emotion in Poetry. Tel Aviv: The Katz Research Institute for Hebrew Literature (in Hebrew).

1987 a The Road to "Kubla Khan". Jerusalem: Israel Science Publishers.

1988 a Hebrew Hypnotic Poetry. Tel Aviv: The Katz Research Institute for Hebrew Literature (in Hebrew).

1988b "Epilogue", in The Golden Fleece: Sonnet Translations. Tel Aviv: Hakibutz Hameukhad Publishers. 84-95 (in Hebrew).

1992 a What Makes Sound Patterns Expressive?The Poetic Mode of Speech Perception. Durham, N.C.: Duke UP.

1992 b Toward a Theory of Cognitive Poetics. Amsterdam: Elsevier (North Holland) Science Publishers.

Tsur, R., Glicksohn, J., & Goodblatt, C.
1991 "Gestalt Qualities in Poetry and the Reader's Absorption Style". Journal of Pragmatics 16 (5): 487-504.

Wimsatt, W. K.
1954 The Verbal Icon. New York: Noonday.

Wittgenstein, Ludwig
1976 Philosophical Investigations, G. E. M. Anscombe (trans.). Oxford: Blackwell.

Original file name: Rhyme/Gestalt 2 - converted on Friday, 28 November 1997, 14:10

This page was created using TextToHTML. TextToHTML is a free software for Macintosh and is (c) 1995,1996 by Kris Coppieters