Reuven Tsur and Yehosheva Bentov

Rhythmic and Strophic Organization in Mediaeval Hebrew Poetry

Back to abstract
Back to home page

Reuven Tsur
Tel Aviv University

Yehosheva Bentov
Tel Aviv University and The State Teachers' College Seminar Hakibbutsim

Rhythmic and Strophic Organization in Mediaeval Hebrew Poetry
(a cognitive approach)

Some Basic Concepts
The system of the so-called "classical metres" of Mediaeval Hebrew poetry in Spain was imported from Arabic poetry in the middle of the tenth century.1 It needed thorough readjustment to the constraints and possibilities of the Hebrew language. Even after that it was felt for long to be alien and arbitrary. But by the first quarter of the eleventh century, the greatest poets acquired a masterful command of it. These are some of its essentials: The metre is based on a distinction between shorter and longer events. The short events consist of schwa mobile or its allophones; the long events consist of full vowels. The schwa mobile has four allophones, pronounced as /a/, /&/, /o/, and /u/. A schwa mobile followed by a full vowel is called a "peg". Mediaeval writers considered pegs as the longer events, vowels as the shorter ones. Schirman considered this "a mistake", and proposed that it would be more parsimonious to treat schwa mobile as a short event, and all the rest as long events. "This mistake in identifying the elements of versification originates in the fact that our sages did not know at all the concept of syllable [...]. If we free ourselves of their conceptions and view the metres of the poems as they are, we shall discover in them short syllables, that is, syllables that contain schwa mobile or its allophones, as opposed to all other syllables to be regarded as long" [Schirman, 1961: 85; all translations by R.T.]. We have adopted Schirman's approach. A number of events, between two and four, are grouped into "feet". Graphically, schwa mobile is usually marked with a In most feet there is one schwa mobile; but in the so-called "metre of vowels" there is no schwa mobile at all; and in a few very rare instances there are feet with two schwas. Two or more feet make a "hemistich". The feet in a hemistich need not be of the same structure, although some recurrence of structures should be expected in every metric pattern. Two hemistichs constitute a stich. The structure of the second hemistich usually repeats (sometimes with minor rule-governed changes) the structure of the first one. These changes concern the rhymed endings, where the schwa mobile of the last foot is systematically omitted in many poems. The classical metres are intimately (but not exclusively) associated with the "equi-rhyme". In the equi-rhyme, an indefinite number, sometimes up to hundreds, of stichs rhyme with the same ending. In some instances we find "ornamental opening" where, in the first stich, the first hemistich rhymes with the second one, and every foot boundary coincides with a word boundary; otherwise, the first hemistich of every stich is unrhymed, and words may run on from one foot to another, or even from one hemistich to another.

The Problem
Traditional research of the prosody of Mediaeval Hebrew poetry discovered and formulated the rules of the classic metres. There was, as a rule, no attempt to understand the rhythmic character of this poetry. Today we have very little knowledge concerning the way the "pegs" and vowels were performed; and only very few readers are capable of putting what is known into practice. Schirman was the first to point out this problem in all its severity:

The scholars who have so far studied the metres of Mediaeval Hebrew poetry in Spain discussed them, mainly, from the optic point of view. They determined the division of these metres, enumerated in them the vowels and pegs, and constructed metrical feet of them; they sometimes argued about the best way to construct the feet of some stich, in this or that metre, or what is the subsidiary form of some metre. They believed that by this they had done their duty as if the determining of the structure of the feet provided us with knowledge about the very essence of the metre, including its sound and rhythm. We have, however, several testimonies from the Spanish era and even from later periods suggesting how much our ancestors enjoyed the sounds of the poems read. The science of Spanish-Hebrew metrics is based on the writings of Mediaeval Arab and Jewish authors. However, their capacity of understanding was rather moderate; and although the poets themselves could compose wonderful metered poems, they were unable to define correctly the essence of their metres (Schirman, 1961: 84).

The present work assumes that Cognitive Poetics is capable of proposing tools for going beyond the explicit rules mentioned above to learn something about the rhythmic character of this poetry. One of the assumptions underlying the present work is that if we can discover, with the help of elementary statistical tools, frequencies and correlations prevalent in this poetry, frequencies and correlations that cannot be accounted for by explicit rules--they may bear witness to the unconscious intuitions of the poets of the period. And if we succeed in accounting for these frequencies and correlations with the help of cognitive theories, we shall be accounting, by the same token, for the principles underlying these intuitions.

Cognitive Assumptions
In order to understand the rhythmical nature of any metre, whether quantitative or syllabo-tonic, one must borrow from Gestalt psychology the assumption that metric organisation is a system that determines the character of its parts or, more precisely, a system in which the whole and its parts determine each other's character. The sequence of the poem's stichs divides the auditory field into larger perceptual units; the hemistichs divide the stich into two equal (or almost equal) parts. One precondition for poetic rhythm to have psychological reality requires the reader to perceive not only a sequence of more or less regularly alternating longer and shorter syllables, but also a higher unit, obtaining a larger unit divided by a sequence of smaller units. "The perception of that which divides is as necessary to the fact of division as is the thing divided" [Chatman, 1965: 23-24]. In our case, this minimum requirement for rhythmicality is satisfied by the hemistich (the higher unit) and the metric feet that divide it into almost equal units.

We have suggested above that the whole and its parts determine each other's character. One of the rules of Gestalt psychology states that the two levels (the hemistich and the metric foot) cannot be perceived as equally clear-cut (with equal "Prägnanz"). When one of them is strengthened, the other one is weakened. If the reader or the performer emphasises the unity and integrity of the hemistich, the metrical feet are pushed into the back of the mind and are experienced as pleasant rhythmical recurrence. If, by contrast, attention is focussed on the foot rank, the reader will become aware of the small units and their relatively fast alternation, while the larger unit, the hemistich, is pushed to the back of the mind. Nonetheless, it will be sufficiently active in the mind to be perceived as equal in length to the other hemistichs, and for the smaller units that constitute it to be perceived as separate and congruous groups. The units of meaning (words, phrases, clauses) may have a decisive affect on the shift of attention to and fro.

The reader is inclined to bestow perceptual integrity upon precisely the hemistich, rather than on the stich rank. It would appear that the reason is to be sought in the limitations of short-term memory. Short-term memory is the mental faculty in which the exact words of a text are preserved for short periods of time; it is indispensable for the verbal material to be accessible to mechanisms of cognitive processing. Its span is the time period during which one is capable of remembering a telephone number, for instance, without rehearsal. As anyone can observe, during this period of time short-term memory works acoustically, like an echo box (we remember the sound of words rather than their meaning). For a metrical unit to be perceived as a whole, the performer must complete it before the memory trace of its beginning fades out. Since the limits of short-term memory have been fixed (Miller, 1970: see now also Chafe, 1987) at seven items (plus or minus two), the largest unit that can be perceived as a whole, that is, that can be completed before the sounds of its beginning fade out, is the hemistich. The hemistichs as perceptual units are not defined only (or necessarily) by the pause between them. The hemistich is united and isolated, usually, by an intonation contour (which is not necessarily identical with the intonation contour that cues a syntactic unit in prose language). Short-term memory is, Miller says, similar to a purse that can hold no more than 5-7 coins, irrespective of their nominal value. Consequently, if we code information in larger units (that is, if we carry along coins with greater nominal value) we can keep more information in our short-term memory. The intonation curve that articulates the hemistich is such a coding device. In order to perceive the whole stich as a perceptual unit, one must remember only two intonation curves; and since short-term memory functions in the acoustic mode, the musical ingredient of intonation may facilitate remembering. Consequently, immediate experiencing of the stich is possible only via an experiencing of the hemistichs.

As we have said above, one of the basic rules of Gestalt psychology states that the two ranks (the hemistich and the metric foot) cannot be perceived as equally clear-cut (as equal in Prägnanz): when one is strengthened, the other one is weakened. The wide range of metric patterns available in Mediaeval Hebrew poetry allows for a very wide range of relationships between the hemistich and the foot. For the purpose of our study, we propose to single out three kinds of possible relationships, in an increasing order of salience of the hemistich, and a decreasing order of the salience of the foot:

The shift of relative salience from the foot to the hemistich and back is not a mere matter of the poet's whim. According to the conception propounded here, it concerns different kinds of solutions to one perceptual problem. The perceptual problem is this: for poetic rhythm to be generated, two ranks of units are indispensable: a larger unit to be divided into smaller units, and smaller units to divide the larger unit. According to the rules of perception, as we have said, when one is strengthened, the other one is weakened. If we succeed in describing significantly the three levels of metric patterns mentioned above, and if we discover that in Mediaeval Hebrew poetry in general, or in certain well-defined parts of it, one level is significantly more frequent than the other ones, we may perhaps learn something about the unconscious rhythmic conception held by the poets of the period.

At this point we must consider three additional Gestalt principles. First, for rhythm to be generated, regular repetition is not enough; there must be also differentiation. A monotonous sequence of beats that are not differentiated in duration or amplitude or fundamental frequency does not arouse in the listener a need for perceptual (rhythmic) grouping. If such groupings occur, they are arbitrary and projective. Second, if the prevailing conditions permit, the whole will tend to be divided into parts of equal structure. If the whole is divided into units that have exactly the same structure, these will tend to become salient at the expense of the whole. Third, therefore, if the parts are to be made dependent on the whole, their similarity must be modified to some extent. In the feet of hamm@rubb& and haSSalem metres, the function of the schwas can be accounted for by the first and third principle. Had the foot consisted of four full vowels, there would have been no differentiation to require rhythmic grouping; the occurrence of the short vowel in the sequence serves to exact precisely such grouping; by the same token, it eliminates the possibility of dividing the group into equal parts. This is the reason that the feet underlying hamm@rubb& and haSSalem metres appear to be so stable perceptually. The foot underlying hammitqarev consists of three beats: a short one and two long ones; it is impossible to divide such a foot into two equal parts.

In order to understand the nature of hammitqarev let us consider the following stich by Yehuda Halevy:

Can corpses become rooms   for hearts tied to the wings of eagles?

The structure of the unit here is SHORT+LONG+LONG; the short beat introduces variation into the sequence; and as for division, it is impossible to divide three syllables into two sub-groups of equal length and structure. Consequently, the unit tends to be stable. On the other hand, the whole stich can be divided into two equal hemistichs; the hemistich, in turn, can be divided into two equal halves, and each half into two identical feet. According to the principle mentioned above, such a structure tends to cause the whole stich to disintegrate: the single feet tend to be salient at the expense of the stich. (In this verse instance this is reinforced by "ornamental opening", displaying rhymes at the end and even in the middle of the hemistichs).

A similar story can be told, mutatis mutandis, about the where the stich is divided into two identical hemistichs which, in turn, are divided into two feet of identical structure. Consequently, here too the isolated foot tends to be salient at the expense of the hemistich. This characteristic may point up, by way of contrast, the distinctive nature of the hamm@rubb& metre, which is based on the same kind of metric foot, with two differences: there are three feet in this metre, and the third foot is one-syllable shorter. Let us consider the metre called that is, one schwa, three full syllables, one schwa, three full syllables, one schwa, two full syllables). This is one of the most regular metres in the Mediaeval metre of "pegs". One way to account for the distinctive characteristic of the hamm@rubb& metre is as follows: the metric feet display considerable stability, since they cannot be divided into symmetrical parts, that is, into parts of equal length and structure. Dividing the feet into short+long \ long+long, for instance, will yield asymmetric parts; they are, therefore, integrated into the foot, and do not tend to fall apart. The first two feet in the hemistich have identical structure; had they occurred alone in the hemistich, as in the hammarnin, they would have stood out as clear-cut symmetrical units. The structure of the third foot is similar, but one full vowel is missing; as a result, the smallest units of identical structure are the hemistichs that constitute a symmetrical stich. In this way, the focus of attention is shifted to the hemistich as the most clear-cut unit, while the exact recurrence of the first two feet is perceived as rhythmical regularity in the background. 2 In light of what has been said, the shortening of the last foot serves to weaken the independence of the individual feet, and to increase the integration of the hemistich (in some variants of haSSalem metre, the last schwa may be omitted in the rhyming hemistichs, and achieve a similar integrating effect). From the point of view under discussion, the difference between hamm@rubb& and haSSalem metres is that in the former, the short vowel occurs before the first one of three long vowels, whereas in the latter it occurs before the last one of three. As will be seen below, this difference of placement has far-reaching consequences for the rhythmic nature of these two metres.

Limited Channel Capacity
In light of these assumptions, we can account for important findings with respect to the relative frequency of the various metric schemes. Two metric schemes, hamm@rubb& and the various variants of haSSalem are by far the most frequent in Schirman's (1961) authoritative anthology. The index of metres in Dov Yarden's edition of Ibn Gabirol's Secular Poetry (1975: 441) reveals that the most common metre in this volume is hamm@rubb& 139 poems. To this, one may add hammarnin 24 poems. The second most common metre is haSSalem including all its variants, 29 poems. The distribution of the other metres in this volume is between 1 to 12. The index of metres in Dov Yarden's edition of Shmuel Hannagid's Diwan (The Son of Psalms) (1966: 392-396] shows a similar preference of hamm@rubb& and hammarnin: 90 and 26 poems. The distribution of haSSalem with all its variants is 24 poems. There is only one metre that is more common in this volume: ha?arokh: 26 poems. The other metres occur in 2 to 16 poems. The emerging picture is that in the secular poetry of Ibn Gabirol and of Shmuel Hannagid there is an overwhelming tendency to prefer the hamm@rubb& metre and its cousin, hammarnin, to all other metres, and a more moderate tendency to prefer haSSalem. The most obvious explanation for this phenomenon is that these are the most regular and simplest metres--relative, of course, to the highly complex metre of pegs (the syllabo-tonic iambic, for instance, is far more regular and simple). The third (shortened) foot of hamm@rubb& seems to be perceived as basically identical with the preceding feet; it would appear that it is shortened only in order to more effectively integrate the hemistich (this point will be elaborated in relation to the principle of good continuation). The possible variations at the end of haSSalem verse line too seem to serve a similar perceptual function. When one reads a poem in hamm@rubb& or haSSalem metre, one need not check whether every syllable conforms with the rules of the metre. It is sufficient to group the syllables into groups that fit the structure of the foot: since short-term memory works like an echo box, one will hear whether the recurring units do or do not fit the metric pattern. This kind of aural inertia is called "metrical set". The classical metres may, then, reach considerable complexities. Let us consider, at the more complex end of our scale, two variants of hammitpaSSet metre, in the opening lines of one of Ibn Gabirol's liturgical poems, and of one of Shmuel Hannagid's martial poems:

At dawn I seek you, my rock and my stronghold,
I unfold before you my dawn and my evening

How long, hero of Jacob, my soul to you has cried out in tears
even though it almost laughed along with the cup of your salvation

In quote (2), the metric pattern is In the hamm@rubb& and haSSalem metres, the performer may repeat quite mechanically one type of foot, and only in the last foot of the hemistich he has to deviate from regularity, for reasons we have mentioned earlier. After having established the mechanical regularity, he may focus attention on the complexities of prose rhythm, and on the rhythmical solutions required to settle the conflicts arising between prose rhythm and the metric pattern. In quotes (2) and (3), the performer must pay attention to the metric pattern, in which the alternating feet are not identical in structure: in addition to the alternation of short and long vowels in a foot, two types of feet (a complex and a simple one) alternate as well. He may not rely on his metrical set.

We have propounded above a conception of how poetic rhythm is processed by short-term memory. Underlying this conception we find the hypothesis known as the limited channel capacity hypothesis. According to this hypothesis--prevalent today among psychologists--human information-processing goes on in channels that have a fixed capacity. If we increase the volume allocated to one process, it necessarily reduces the information-processing volume available for other processes at the same time. It is impossible to increase the capacity of the channels. The only way to sidestep the problem is to recode the information in a more efficient way, so that it occupies less "volume". The semantic sequence "a person who sells goods" can be recoded as "merchant"; and "a merchant who sells meat" can be recoded as "butcher". In poetry, however, where the reader is not allowed to replace words or change their order, the possibilities of recoding are severely limited. In fact, they are confined to the kinds of recoding discussed above. Thus, for instance, considerable volume can be saved, when short-term memory must remember two intonation contours rather than six or eight isolated metric feet. Quotes (2) and (3) too can be recoded in a similar way. If one groups every two adjacent feet into a larger group, the hemistich will divide into two symmetrical halves, alleviating the load on short-term memory, but at the price of creating an additional rank in the hierarchy of prosodic units. At any rate, the increase of the number of ranks in this hierarchy reduces the channel volume available for processing additional aspects of the verse line (prose rhythm, figurative language, and so on). As a result from this grouping activity, the distinction between the smaller, dividing units and the larger, divided units, is blurred. The smaller, dividing unit is created by the grouping of two kinds of dissimilar metric feet; whereas the larger, divided unit, the hemistich, falls into two identical halves, each of which consists of two metric feet of unequal length and structure. Such additional complication of the metric organisation may turn out to be "the last straw that breaks the rhythm", preventing even a qualified reader from allocating the channel volume required for the processing of a vocative phrase enclosed in the syntactic sequence (such a vocative phrase makes two more intonation contours necessary). And conversely, the enclosed vocative phrase may be "the last straw that breaks the rhythm", preventing even a qualified reader from realising in his active memory the complex metric patterns of quotes (2) and (3). In both cases, the rhythmic performance of the verse line may collapse.

The rhythmic organisation in the classical metres of Mediaeval Hebrew poetry displays, then, the varying relationships between the dividing units and the units divided, so vital for the generation of poetic rhythm. These types of organisation can be viewed as parts of a spectrum. At one end of the spectrum we find the mitqarev metre: the structure of the foot in this metre is the simplest among the classical metres; one single foot structure keeps recurring along the entire stich; and the number of feet in a hemistich is even. As a result, the regular recurrence of the dividing units is most salient, whereas the structure of the unit divided is blurred, or even its perceptual existence may become doubtful. At the other extreme of the spectrum, we find the various variants of hammitpaSSet. In these metres, there is no regular recurrence of one dividing unit at the lowest rank: there is here no foot with one structure, but two types of metric feet, with different structures. Regular recurrence may occur only at a new, intermediate rank, generated by the grouping together of two different types of metric feet, grouping that divides the hemistich into two symmetrical halves.

At the middle of the spectrum we find the hamm@rubb& and haSSalem. Here the feet are more complex than in the mitqarev (consist of an even number of syllables); the number of metric feet is odd, that is, the hemistich cannot be divided into two halves of equal length; the third metric foot is shortened, and therefore the hemistichs are the smallest units of equal length into which the stich is divisible. Consequently, the hemistich is a relatively stable unit in these two metres; and the smaller, dividing units are rather regular--relative to hammitpaSSet metre, for instance. The fact that these are the most widespread metres in Mediaeval Hebrew poetry can be interpreted (and this is, indeed, the interpretation we propose) that the precarious balance between the need for a relatively stable unit-to-be-divided and the need for a relatively simple and regular dividing unit is the main constraint within which the poets of the period proposed to realise poetic rhythm. The two extremes of the spectrum don't necessarily generate unrhythmical units, but rather units whose rhythm is relatively hard to realise. Consequently, we may expect to find, to prophesy after the event, poems composed in these metres too, but their number will be smaller than of those composed in hamm@rubb& and haSSalem metres. At this point, we may risk an additional conjecture. Poets who resort frequently to complex metres like those in quotes (2) and (3) may be assumed to indulge in a feat of manipulating metric rules, whereas poets in whose poetry hamm@rubb& and related metres abound four or five times more than the next most common metre, presumably compose their poems for the ear rather than for the intellect controlling the rules of versification. In what follows, we shall have to qualify this conjecture substantially.

To conclude the present section of our study, let us consider the opening stich of one of Levy Ibn Altabban's brief liturgical poems:

Why, my soul, in blood you are rolling yourself?
are you for your crimes very much trembling?

Some readers have noticed that this verse line, and many similar ones by this poet, are perceived as less rhythmical than some others written in similar metres. Elsewhere [Tsur, 1969: 145-153; 183-184], the lack of musicality of this poet has been discussed in some detail, as well as the source of the musical richness of poems by some other poets. Here we wish to note only one point, relevant to our discussion. The metre of this verse is One of the most conspicuous things in the prose rhythm of this stich is the vocative phrase (my soul) enclosed after the function word lamma (why), increasing complexity. At the same time we find an arbitrary reversal of word order in the sentence, putting the indirect object "b@dam" before the predicate. This increases the burden on short-term memory and makes rhythmical information-processing more difficult. This example does, indeed, draw attention to the fact that one source of the exquisite musicality of some poems by some other poets derives from the fact that they contain no such arbitrary reversals of word order (despite all that is usually said about poetic licence). It would appear that the additional load upon short-term memory in quote (4) disrupts the rhythmicality of the hemistich; on the other hand, the hemistich contains no rhetorical repetitive schemes that might strengthen the rhythmic affect of the vocative phrase enclosed in the verse line (we shall refrain here from commenting on the grammatical rhyme).

An anonymous reviewer commented on our Ibn Altabban section as follows: "This description is implausible. 'Poetic' reversals occur in about 35% of the lines of Shakespeare's sonnets and only 12% of the lines of his blank verse. Tsur and Bentov's claim implies that Shakespeare's sonnets are less musical than his blank verse, although the opposite seems to be true". This comment gives us an excellent opportunity to clarify our theoretical position on the English reader's home ground. Our position is that neither syntactic inversion nor metric deviation increase or decrease tension in themselves. They merely increase the load on the processing system. They may generate tension if (and only if) the "unit above" (the line or hemistich) is brought within the span of short-term memory by a rhythmical performance. The greater the load on the processing system, the greater is the tension as long as the unit above is perceived as a whole, that is, as long as the unit above can be completed before the acoustic traces of its beginning fade out in short-term memory. As soon as the load on the processing system exceeds the capacity of short-term memory, the verse-line abruptly ceases to be rhythmical, and tension disappears. The reviewer's comment on Shakespeare is valid only if one tacitly assumes that the load resulting from syntactic inversion in his verse does not exceed that theoretically undefined point, at which tension disappears. Since, according to our analysis, even the simplest metres in Mediaeval Hebrew poetry are more complex (and thus constitute a greater load on the processing system) than the iambic pentameter, a relatively small increase caused by syntactic inversion may cause the verse line to exceed that theoretically undefined point. This double-edged nature of rhythmicality has precedent in English syllabo-tonic verse. Tsur (1977: 58-61) compares the relative deviance of Spenser (124 deviances per 100 lines), Thomson (147), Pope (156), Milton (180), and Shelley (191). The numerical difference between the most and least deviant poet is small: 67 deviances per one hundred lines (that is, per one thousand positions); the difference between Pope and Milton is 24. In a discussion of appreciation, Tsur points out that in Donne's "Second Satyre" there are 291 deviations per 100 lines (Tsur 1977: 225). We cannot tell where exactly is the critical point at which rhythmicality abruptly disappears. But Milton's and Shelley's verse is clearly kept within the span of short-term memory, whereas Donne's clearly exceeds it. That's why Milton and Shelley are regarded as two of the greatest masters of musicality, and Donne is not. Nonetheless, Donne's versification is not regarded as incompetent. Aristotle demands in his Poetics that a character should be consistent; and if he is inconsistent, he must be consistently so. Tsur points out that Donne also increases the load of his verse on the processing system by increasing syntactic complexity in a variety of respects, thus indicating that his irregularity is not random, and that his metre exceeds the span of short-term memory in a consistent manner. Milton and Shelley too use syntactic complexities; but some of Donne's are, so to speak, conspicuously drastical.

Beginning and End Accented Groups
There is a very significant difference between the relative frequency of hamm@rubb& and haSSalem metres. The paradigmatic feet of both metres consist of similar elements: three long beats and one short beat. They differ in the placement of the short beat: in hamm@rubb& it precedes the first long beat , while in haSSalem it precedes the last one . This difference has far-reaching cognitive and aesthetic implications. As a first step, we have to find out how this difference affects the relative naturalness of hamm@rubb& metre (the supposed cause of its overwhelmingly frequent occurrence).

One of the earliest experiments concerning the rhythmic grouping of beats was conducted by Woodrow (reported by Chatman, [1965] and many others) back in the twenties. If we listen to an endless series of tick-tocks, how do we group them perceptually: into groups beginning with ticks, or with tocks (that is, with a strong or a weak beat)? On first thought, it depends on whether we began to listen to the series with a tick or a tock. Woodrow, however, found that it depended on the acoustic cue in which the beats differed from one another: if they differ in their relative amplitude (loudness), beginning-accented groups are created; if they differ in their relative length, end-accented groups are created; distinctions in fundamental frequency (musical pitch) make no difference in this respect.

Before hastening to draw conclusions in respect of the metres of pegs, we have to consider the perception of linguistic stress too. Stressed syllables might seem to differ from unstressed ones in their relative loudness. Experiments in the phonetic laboratory, however, show that this is not quite the case. Since Schramm's (Schramm, 1935) pioneering experiments, it is known that the acoustic correlate of perceived stress is a mixture of relative amplitude, duration and pitch. In a series of well-controlled experiments in the early fifties, Fry (1958) tried to find out what the relative share of these three acoustic cues in the perception of stress is. He recorded such words as present, subject, object. The penultimate syllable of these words is stressed when they are nouns, and the last syllable when they are verbs. He manipulated these syllables electronically, in a controlled way, varying amplitude, duration, and fundamental pitch. The words thus obtained were played to listeners (students of phonetics) who had to respond "verb" or "noun". At variance with common intuition, he found that amplitude is the least effective acoustic cue for perceived stress; relative duration overrides amplitude, and relative pitch overrides duration. These findings imply that there is no contradiction between grouping favoured by linguistic stress and grouping favoured by the metre of pegs: both favour end-accented grouping. The most effective acoustic cue for stress, relative pitch, is neutral from this point of view, whereas the second most effective cue, duration, favours end-accented grouping. The same holds true of the metre of pegs, which by definition exploits differences of duration. Now the foot of hamm@rubb& metre is end-accented: the schwa (the weakest "beat") occurs before the first full vowel; whereas haSSalem is beginning-accented: the schwa occurs before the last full vowel. Consequently, haSSalem foot is beginning-accented (there are more long units in its first half), whereas the typical foot of hamm@rubb& is end-accented (there are more long units in its second half); hence, it is the more natural (unmarked) metre between the two. 3

We have to adduce some explanation for the relationship between duration differences and end-accentedness. Roman Jakobson (1960: 357) notes that "in a sequence of two coordinate names, as far as no rank problems interfere, the precedence of the shorter name suits the speaker, unaccountably for him, as a well-ordered shape of the message". Hence we prefer Marks and Spencer or Marx and Engels, for instance, to Spencer and Marks, or Engels and Marx. This intuitive preference can be explained by the limitations of short-term memory. Short-term memory must preserve the beginning of a message, while new information is still coming in and is being processed. The shorter the unit burdening memory, the greater is the space available for the mental processing of incoming new information.

This mechanism is responsible also for some intuitive preferences of poets and composers in the rhythmic organisation of their work. According to Cooper and Meyer (1960), of two parallel musical phrases the longer one comes usually last. As suggested above in footnote 2, the iambic pentameter cannot be divided into two parts of equal length and structure; equal length is achieved in a 5+5 division, while equal structure is achieved by a division into 4+6 or 6+4. Consequently, while in the iambic tetrameter or hexameter, perceptual dynamic compels a caesura after the fourth or sixth positions, respectively, in the iambic pentameter, caesura may occur after the fourth, fifth, or sixth positions. It would appear that the distribution of caesurae is not accidental, and the fourth and the sixth position are not equal in this respect: in English, Hungarian, and Hebrew iambic pentameter, caesura occurs very significantly more frequently after the fourth than after the sixth position. In the syllabic metre, too, in French, in Baudelaire's and Villon's decasyllabic verse, we find the 4+6 division one and a half times more frequently than the 6+4. In his lecture mentioned above, Roman Jakobson (1960) describes the sophisticated prosodic rules of the Serbian folk epic; for the purpose of our discussion, several of those rules boil down to one requirement: an obligatory, sharp break after the fourth syllable of the decasyllabic verse (cf. Tsur, 1977: 72-82; 1992: ch. 6). All these examples suggest one principle: in a considerable variety of languages and versification systems in poetry, as well as in music, this principle of short-term memory is active, inducing poets, composers, readers of poetry, and listeners to music to prefer intuitively the precedence of the shorter segment to the longer one. This principle may have induced Mediaeval Hebrew poets to prefer the foot to the foot.

Secular and Liturgical Poetry
The above findings concerning the relative frequency of the classical metres in equi-rhymed poems are based on the combined corpus of secular and liturgical poems. The vast majority of this corpus is in the domain of secular poetry, and only a negligible minority occurs in liturgical poetry. Owing to these proportions, it is possible that even if the distribution of classical metres in liturgical poetry is different, it would not noticeably affect the overall picture. So, the distribution of metres in liturgical poetry alone has been checked (Bentov, 1990). In about 100 equi-rhymed liturgical poems, hamm@rubb& was found only 25% of the time, whereas hammitpaSSet was found about 30% of the time. None of the other metres were found more than 10% of the time. The difference between 25% and 30% is not so big; but the reversal of proportions between secular and liturgical poems is quite remarkable, and demands some explanation. About twenty-five per cent of the liturgical poems checked were by Shlomo Ibn Gabirol. From his poetry, the following picture emerges: out of 239 secular poems, the metre of 148 is hamm@rubb& (62%), about 10% each are due to hammarnin (24) and haSSalem (29), while the metre of six poems is hammitpaSSet (2.5%). In his liturgical poetry, hammitpaSSet is dominant (24 poems out of 69, that is, 34%), while hamm@rubb& is only second to it (18 out of 69-- 26%).

We have found, then, substantial differences of preference between secular and liturgical poetry; but there is no perfect symmetry here: hammitpaSSet has no such overwhelming dominance in liturgical poetry as hamm@rubb& has in secular poetry; its preference is much more moderate. Within the theoretical framework proposed above one may conjecture that this is not an all-out cognitive reversal, merely some alleviation of the burden on the cognitive system, affording more efficient processing of poetic rhythm, even when the metre in question is one that constitutes considerable burden on it.

As we have noticed above, hammitpaSSet constitutes a relatively heavy burden upon the cognitive system in two respects: the hemistich consists of two different kinds of metrical feet; and the feet are beginning-accented. In quotes (2) and (3) above, the structure of the longer foot is , as in haSSalem: here the first half is longer than the second half. The shorter feet, by contrast, are perceptually ambiguous. The short foot in quote (2) contains two vowels only, with no schwa; here the first part is equal to the second part. In quote (3), on the other hand, a schwa is enclosed between the two vowels of the shorter foot; a three-position-long unit cannot be divided into quasi-symmetrical parts, and it is impossible to know whether it is end-accented or beginning-accented. When, however, we group together in these hemistichs--as suggested above--two pairs of feet so as to obtain two symmetrical halves, we also obtain two beginning-accented units: in each pair of metric feet, the longer and more complex foot comes first, and the shorter and simpler foot comes second. Consequently, hammitpaSSet metre constitutes a burden upon short-term memory for two reasons: it consists of two types of metric feet, and it is beginning-accented at two levels (in respect of the structure of the longer foot, and in respect of the grouping of feet together).

In the Cognitive Poetics workshop at the Katz Research Institute for Hebrew Literature two rival cognitive prosodic theories have been developed over the years, the present one in the Gestalt psychological tradition, the other in the generative linguistic tradition. David Gil's point of departure for his generative theory was in such non-canonical poetic genres as cheers at the soccer games, in political demonstrations, chants of peddlers and hawkers, of hucksters at the marketplace. He has recorded his corpus of texts in Western Europe and the US, in the Middle East and the Far East, obtaining an impressive inter-cultural sample. This theory is based on a hierarchical conception of binary "trees", in which one of each pair of units is relatively strong, and one relatively weak. For our business, only one of Gil's findings must be considered, quite puzzling in itself. Most of these rhythmical cheers have no melody; but the relatively few melodies that do occur in them, occur only in those cheers, in which the first member is stronger. In other words, there is a significant correlation between melody and beginning-accentedness. It would appear that this is an inter-cultural phenomenon. (We wish to insist that we do not use Gil's method, only his finding, and interpret it within our own theoretical framework).

Within the theoretical framework propounded here, the explanation for this finding is almost self-evident. We have said that a versification unit is perceived as a complete rhythmical whole, if it does not exceed the capacity of short-term memory; that the capacity of this memory is rigidly fixed, and the only way to increase information in it is to recode it more efficiently; that short-term memory works like an echo box, acoustically; and that owing to its limitations, there is a tendency to leave the longer units, or the ones hardest to process, to the end. "It sounds somehow better" in this way. Consequently, the cheers in which the longest or hardest to process member comes last constitute a less heavy load upon the cognitive system; cheers in which the greater weight comes first may require additional acoustic coding, so as to facilitate the acoustic functioning of short-term memory. It would appear that music serves in these instances as just such acoustic recoding devices.

If this analysis is correct with respect to rhythmical cheers, one may perhaps make also inferences as for the phenomena encountered in liturgical poetry. There, as we have seen, the metric organisation may constitute a heavy load upon the cognitive system, owing to both the beginning-accented grouping and the structure of the hemistich, that contains two different types of metric feet. Here, too, melody may serve to alleviate the acoustic load on short-term memory. For all we know today, liturgical poetry used to be sung, whereas non-liturgical poetry used to be recited. This may explain why is hammitpaSSet metre so widespread in liturgical poetry, a metre that constitutes such a heavy load upon the cognitive system and is, accordingly, so rare in secular poetry.

At this point another problem arises. The explanation offered here may perhaps explain that acoustic recoding with the aid of music enables the use of metres that constitute a relatively heavy load upon the cognitive system; it does not, however, explain why the poets should want to take advantage of these new possibilities. In his Aesthetics, Beardsley (1958) speaks of three "general canons of evaluation". Aesthetic evaluation based on relationships within the aesthetic object must rely on one of the three general canons: unity, complexity, and some intensive human quality. The present article explored poetic rhythm, which is one of the intensive qualities of the poetic text. For such a quality to be generated, a precarious balance between unity and complexity must be stricken: some degree of complexity at the foot rank must be generated, without disrupting the perceptual unity at the hemistich or the stich rank. The greater the complexity at the foot rank, the greater is the danger of disintegration at the higher ranks. If we take Beardsley's position seriously, we must assume that the poet will want to achieve the greatest complexity possible in the given conditions without destroying the perceptual unity of the hemistich or the stich. If the acoustic recoding with the aid of music makes it possible to increase the unity of the higher units, one may expect that the poets will attempt to increase the complexity of the stich at the foot rank--thus heightening the aestheticalness of the poem. 4

The Girdle Poem
The "girdle poem" flourished in Mediaeval Hebrew poetry roughly at the time when the equi-rhymed poems with their classical metres flourished. It too was imported from Arabic poetry, but descended from a different poetic tradition, and has largely different versification principles. It is strophic, and only a negligible proportion of it is written in the classical or "quasi-classical" metres. It grants the poet incomparably greater freedom than the equi-rhyme. The poet is free to determine, within certain constraints, the rhyme pattern of his poem, and may choose from a wide range of possible metres--far beyond the range of classical metres. The poem begins with a "guide" containing two verses or more; they need not rhyme with one another, and each one may have a metric pattern of its own. This is followed by a strophe consisting of two sections (cf. below, quote (8)): the first section contains usually three "branches" that rhyme with one another, and have identical metric patterns. This section is followed by a "girdle", which repeats the prosodic structure of the guide: the first line of the girdle rhymes with the first line of the guide; the second line of the girdle with the second line of the guide, and so forth; the lines of the girdle also repeat the respective metric patterns of the guide. The structure of the second strophe is similar to the structure of the first strophe, only the branches rhyme on a different ending; then it ends again with a girdle, whose prosodic structure is again identical with the prosodic structure of the guide; and so forth, up to 5-7 strophes.

Concerning the girdle poems, whether liturgical or secular, there is more or less consensus that they were sung. Consequently, there arises the question whether the distribution of metres in them is in harmony with our findings in equi-rhymed verse as analysed above. It is quite difficult to answer this question, since the vast majority of girdle poems are not written in the classical metres. We shall, therefore, have to proceed in two stages. First, we shall have to check whether in those rare instances in which the girdle poem is in a classical metre, the distribution of metres is similar to the one found in equi-rhymed verse. Second, in the other cases (which are the majority) we shall have to see what can be learned from the placement of the schwa mobile in the verse line.

Out of one hundred and five liturgical poems in the girdle poem form that have been checked for the present paper, only eighteen were in foot patterns that were close to the classical metric patterns. One may, as a rule, find in the "quasi-classical" metres a combination of classical feet from a diversity of metres. Sometimes it is a variant of one of the classical metres, with a change in the order of feet, or the addition of one of its feet at the beginning or the end of the verse. When we speak of "a classical foot pattern", we mean some foot in which there is a combination of two or more units of pegs and vowels, and some regular alternation of strong and weak positions. Most feet contain four beats (three long ones and a short one, that is, two vowels and one peg) reminiscent of the classical metres. The liturgical poems whose feet are identical with those of the classical metres (those which contain two vowels and one peg, or two pegs, that is, no more than four beats) have two common characteristics:

a. There are almost no shorter than four-beat-long feet.
b. There is a clear preference for feet in which the short beat is in the middle, and not at the beginning:

This preference conflicts with the general tendency detected in Hebrew poetry in Spain. An investigation of the equi-rhymed poems (liturgical and secular) in Schirman's (1961) authoritative anthology shows a deliberate preference of metres whose feet begin with a short beat (schwa or its allophones). The poems whose metres begin with a schwa three times outnumber the poems based on the various metres beginning with a full vowel. In the liturgical girdle poems checked for the present study there is a clear tendency to prefer structures similar to haSSalem and hammitpaSSet, the most common foot being that is, the short beat preceding the last vowel.

A similar tendency has been observed in the secular girdle poems. Among the twenty examples of poems written in some variety of quasi-classical metres quoted in volume 1 of Israel Levin's (1980) study there is only one poem whose metre begins with a peg. Tova Rozen (1985) examines the "classical" ingredients in the metre of seventeen secular girdle poems by Yehuda Halevy. Such widespread metres as hamm@rubb&, ha?arokh, and hammarnin are totally absent from her table; some other, rarer metres are at the bottom of the table; the majority of examples belong to hammitpaSSet,

As for the metres that are irrelevant to the classical metres, that show no signs of regular alternation of pegs and vowels, but rather display an eclectic mixture of prosodic elements, one may find a conspicuously small number of short beats (schwa or its allophones) in relatively long sequences of vowels. In verse lines that consist of long series of vowels we find no more than three short beats--and usually only one. The order of long and short beats is, however, meticulously preserved. It is very rare that in these metres a short beat occurs at the beginning of the verse line.

If there is some truth in the belief that girdle poems, whether secular or liturgical, were meant to be sung, and that from the cognitive point of view, melody serves as an additional acoustic coding device for beginning-accented metric patterns so as to alleviate the load on short-term memory--then in the girdle poems we have found exactly what ought to be expected, on the basis of our assumptions and analyses: where elements of classical metres can be identified, there is a very clear preference of beginning-accented feet; whereas in the non-classical metres, so characteristical of the girdle poems, there is, at any rate, a tendency to refrain from beginning the verse line with a short event (schwa or its allophones).

Continuation and Return
For a meaningful discussion of rhyming one should borrow from Gestalt Psychologists two principles, which they call "The Law of Good Continuation", and "the Law of Return".

A shape or pattern will, other things being equal, tend to be continued in its initial mode of operation. Thus, "to the factor of good continuation in purely spatial organisation there corresponds the factor of the smooth curve of motion and continuous velocity in spatio-temporal organisation". [...] Actually, of course, a line or motion does not perpetuate itself. It is only a series of lifeless stimuli. What happens is that the perception of a line or motion initiates a mental process, and it is this mental process which, following the mental line of least resistance, tends to be perpetuated and continued (Meyer, 1956: 92).

The "Law of Return", on the other hand, is "the law that, other things being equal, it is better to return to any starting point whatsoever than not to return" (Meyer, 1956: 151). These two laws impose different characteristics upon different strophic organisations, each of them being considered as "good" perceptual organisation under the relevant law. Let us take, as an example one of Omar Khayyám's Rubáiyáths, in Edward Fitzgerald's famous English version:

(5)  Think, in this battered Caravanserai
    Whose Portals are alternate Night and Day,
      How Sultán after Sultán with his Pomp
    Abode his destined Hour, and went his way.

Here the second line rhymes with the first one, arousing a feeling of good continuation. According to the principle of good continuation, the reader will expect that the third and fourth lines too rhyme with the first two. When this expectation is frustrated in the third line, the reader may develop an expectation for good continuation at a different level: he may expect that the fourth line rhyme with the third one, generating good continuation at the couplet level, that is, when the second couplet is a good continuation of the first one. When the reader discovers that the fourth line rhymes with the first two, he experiences "return": he may feel that he "came home", as it were, after a feeling of frustration. The "law of good continuation" seems to have precedence, and only when its application fails, the law of return is applied. In other words, the law of return is the marked one of the two. In such a reading, the third line is perceived as disturbance, and the return to the rhyming of the first lines as the overcoming of that disturbance and as powerful closure of the poetic unit, what (Herrnstein-Smith, 1968) called "poetic closure". As we have discovered in a series of experiments, the effect depends on "mental performance" (which, in turn, may depend on cognitive style, or training): the more the reader exaggerates the effect of the "deviant" line, the greater his gratification when returning to the initial rhyme; by the same token, the power of the closure is amplified Tsur et al., 1990; 1991). The force of the closure is further amplified by what Herrnstein-Smith called "closural allusion": the mentioning of going away, that serves as a metaphor for death. Let us compare now this Rubáiyáth to one of Yehuda Halevy's "seafaring-poems":

Has the flood come, destroying the world
that no surface of dry land can be seen?
And there is no man, and there is no beast, and there is no bird
has everything ended, lying in woe?

This poem consists of seven stichs (only two of them quoted here). Owing to ornamental opening, its first hemistich rhymes with the second one, in fact, with all the equi-rhymed stichs. Had we written these verses in four lines, it would have become conspicuous that its rhyme scheme is a-a-b-a, just as in the Rubáiyáth. Nevertheless, the impression received is very different. This is due to the cumulative effect of several factors. First, the greatest number of closural devices is concentrated in the seventh, last, stich of the poem (Tsur, 1969: 178). Even if the first hemistich of the second stich is perceived as a unit that disrupts good continuation, and even if the rhyme at the end of the second stich is perceived as "return" to the rhymes of the first stich, all this becomes part of a longer sequence that constitutes good continuation: in each one of stichs 2-7, the end of the first hemistich disrupts the regularity established in the first stich, while the rhyme at the end of the second hemistich constitutes a return to it. Third, if the reader wishes to preserve good continuation in this respect, he must adopt a "mental performance" (Tsur, 1992: 29-41) in the third hemistich that is very different from the mental performance suggested by the third line of the Rubáiyáth. In the Rubáiyáth, the reader may discover that the more he exaggerates the effect of the deviant, "disturbing" line, the greater is the gratification afforded by the return to the rhyme of the first lines and, by the same token, the stronger is the closure and the sense of a "punch line" at the end. On the other hand, if the reader wishes to resort to the same mental performance with reference to the third hemistich in Yehuda Halevy's poem, he may discover that the more he exaggerates the effect of the deviant, "disturbing" hemistich, the stronger is his sense of return and closure at the end of the second stich; by the same token, this disrupts the sense of good continuation leading to the powerful closure at the end of the seventh stich. And conversely, the more he moderates the effect of deviation at the end of the third hemistich, the more he enhances good continuation throughout the poem. Furthermore, in light of all that has been said in the present study, it may well be that end-accentedness, pointed out in relation to hamm@rubb& metre, may further contribute to the sense of good continuation, of uninterrupted proceeding toward the end.

At this point, we wish to comment on good continuation, in connection with the cognitive structure of hamm@rubb& metre. The second foot of hamm@rubb& metre makes good continuation after the first foot, in being identical to it in structure; the third foot apparently disrupts good continuation. Actually, it goes on for some time with good continuation: the schwa and the ensuing two vowels still repeat the beginning of the preceding feet, and it is only when the third vowel fails to turn up that good continuation is disrupted. This failure to turn up is quite functional: it indicates to the reader that the sequence has come to an end at the hemistich rank, and from now on, he may expect only good continuation in repeating the same structure in the ensuing hemistichs. In other words, the omission as it were of the last vowel contributes to the perceptual articulation of the hemistich.

Equi-rhymed poems tend, then, to draw heavily on the principle of good continuation. This we have shown at length with reference to equi-rhymed verse. The girdle poem, by contrast, is based on the principle of return. The "girdle poem" begins with a "guide" consisting of one, two, or more consecutive lines that do not necessarily rhyme with one another, nor do they necessarily resemble one another in their metric structure. The ensuing strophes consist, each, of an indefinite number of "branches", followed by a "girdle". The branches rhyme with each other (but on different endings in every strophe), and have the same underlying metric pattern. The girdle exactly repeats the prosodic structure of the guide. In reading these poems, there is sometimes a very pronounced feeling that the groups of branches with their varying specific rhymes intrude upon the smooth flow of verses, so that each time when the girdle arrives, it is perceived as a return home, granting the reader an intense feeling of gratification.

In the present paper we have considered two prosodic phenomena, one at the metre level, and one at the level of the grouping of lines effected by rhyme. At the former level, we have distinguished "beginning-accented" and "end-accented' patterns; at the latter level we have distinguished between the principles of "good continuation", and of "return". Both end-accentedness and good continuation focus cognitive weight upon the relationship between what is "now" and what is "to come". Both beginning-accentedness and return focus cognitive weight upon the relationship between what is "now" and what preceded it. This may account for the very significant statistical correlation between the end-accented hamm@rubb& metre and equi-rhymed verse, as well as that between beginning-accented metres and the girdle poems. On the former correlation we have commented during our foregoing discussion. In what follows, we propose to make a few comments on the latter one.

This relationship we find sometimes both on a concrete level, and on a level of wider principles. A convenient way to bring into relief the concrete level is to point out the similarity between the structure of the foot (which is the most common beginning-accented metric foot), and the rhyme structure of the Rubáiyáth (which is one of the most condensed manifestations of the principle of return): each of them consists of four units, the first two embody some identical structural element, the third one is different, whereas the fourth one returns to the structural element of the first two units. At the level of the wider principle we have found that the beginning-accented metres constitute a heavier load on short-term memory than end-accented ones. The same is true of the principle of return. Here the reader must keep in mind the principle to which he is to return after the deviation, whereas in connection to the principle of good continuation, he only must focus attention on the sequence of incoming units, without "skipping" back and forth.

The present theoretical approach may account for additional poetic conventions in our comparison between equi-rhymed verse and girdle poems, where the musical-acoustical coding enables recourse to more complex prosodic phenomena, the processing of which constitutes a heavier load on short-term memory. According to the view propounded here, these conventions are fossils of cognitive preferences in aesthetic contexts. We shall mention two of them. First, there are no sentences run on from stich to stich in equi-rhymed verse, whereas sentences run on from branch to branch is quite common in girdle poems. Second, while in equi-rhymed verse we find only internal rhymes that contribute to good continuation, in girdle poems we find quite frequently also internal rhymes that require reliance on the principle of return. Consider the following stich from an equi-rhymed liturgical poem by Shlomo Ibn Gabirol:

To you I coo with a thirsty heart, resembling
a pauper begging at my door and threshold

Here, the internal rhymes occur in one hemistich, and point up the parallel entities of metric feet and the semantic units that overlap with them enhancing, by the same token, good continuation. There is a similar use in quote (1) above. These are the typical internal rhymes in equi-rhymed verse. While in girdle poems there are such internal rhymes, a second kind of internal rhymes abounds in them too, which require, again, reliance on the principle of return:

[The soul]
descended to a house / of clay, at the assembling / of body and form
And was hidden / there unwillingly / but detained
Not by hand captured, / even without money / she is sold
To labour the soil / of the body, imposing on it / grandeur and awe
As law, setting it apart / by the faculty of thought / from the beast

Here, internal rhyme does not direct attention to "rhyme-fellows" in a linear direction in the same verse line, but backward, to additional members of the same rhyme in the preceding verse line, in the same position. In this fashion, a complex network of directions is created, constituting an enormous burden on the cognitive system. It is very plausible that most readers cannot realise by the ear this complex network in the course of reading; its realisation remains for them an intellectual exercise of following the rhyme patterns, in isolation from the textual sequence. It is also quite plausible that if anybody is capable of realising this complex network in the course of reading in the auditory mode, he will be in need of all the aid of acoustic coding he can get.

This approach may also shed some light on the way poetic conventions come into being. The metric system of pegs and vowels was introduced into Hebrew poetry by a conscious and intentional effort of Dunash Ben Labrat, and its reception process involved violent ideological and other conflicts. But the overwhelming dominance of hamm@rubb& metre within this system was the result of a long and unintentional process of natural selection: the metric pattern that had the best fit to the natural constraints of cognitive economy was the one that had a better chance to prevail and multiply.

In our study we have assumed that the phenomena discussed are governed by statistical tendencies and not by determinism. We do believe in the freedom of spirit, and that poets are able to feel their way in their domain, to make new combinations and try out their effects--while the results are tested by the ear. The finished product of the processes discussed above need not, therefore, surprise us. The forms that have a good fit to the natural constraints of the cognitive system are the most common ones; those that fit less well--are not absent from the corpus, they are only relatively rare. Nor should we assume that the forms that fit less well sounded less pleasing to the poets and their audience, but, in order to make them too sound pleasing to the ear, they require greater skill or more aid in their acoustic coding for their rhythmical realisation.


1. In the hard-copy paper we adapted a combination of the IPA and the American Transcription System for the transcription of Hebrew. In large chunks of quotations we have preserved this transcription. In isolated words within the text, however, we made an adaptation of the MRPA (Machine Readable Phonatic Alphabet). Thus, for schwa (as in English "ago") we used @; for segol (as in English "bet") we used &; for tsere (as in "made") we used e; for shin: S; for tsadik: ts; for the glottal stops aleph: ?, and ayin: $.

2. This principle of achieving unity at a higher rank by disrupting regularity at a lower rank is not confined to the metre of pegs. As it has been argued elsewhere with reference to syllabo-tonic metres, one peculiarity of the iambic pentameter is, precisely, that it cannot be divided into two parts of equal length and equal structure: if it is divided into 5+5 positions, the first part begins and ends with a weak position, whereas the second part begins and ends with a strong position. In order to obtain two units each of which begins with a weak position and ends with a strong position, the line must be divided into 4+6 or 6+4 (Tsur, 1972; 1977).

3. Woodrow's and Fry's experimental findings have been used to account for the relative naturalness of the syllabo-tonic iambic as compared to the trochaic, noted by many scholars and critics (cf. Tsur, 1977: 83-96).

4. At this point we should, perhaps, point out that we have found so far only one metre that consists of two different types of foot that occurs with any significant frequency in secular poetry: ha?arokh metre, in Shmuel Hannagid's Diwan (see above), where it is somewhat more common than haSSalem, but much less than hamm@rubb&. The structure of the hemistich in this metre is . Consequently, in spite of its complexity, its relative abundance can be explained by the fact that it is end-accented at all ranks: at the foot rank, the short event (the schwa) occurs at the beginning; and at the rank of grouping feet into equal segments: here, in contrast to hammitpaSSet, in each segment, the longer foot comes last.


Bahat, Avner (1985). "Ibn Gabirol's Liturgical Poems as Sung by Yemenite Jews", in Zvi Malachi ed., Studies in the Work of Shlomo Ibn-Gabirol. Tel Aviv: The Katz Research Institute for Hebrew Literature (in Hebrew). 179-213.

Bentov, Yehosheva (1990). "The Relationship between Hebrew Liturgical and Secular Poetry in Mediaeval Spain". Unpublished Tel Aviv University Dissertation (in Hebrew).

Beardsley, Monroe C. (1958). Aesthetics: Problems in the Philosophy of Criticism. New York & Burlingame: Harcourt, Brace & World.

Chatman, Seymour (1965). A Theory of Meter. The Hague: Mouton.

Cooper, C. W. and L. B. Meyer (1960). The Rhythmic Structure of Music. Chicago: Chicago UP.

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

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

Jakobson, Roman (1960). "Closing Statement: Linguistics and Poetics", in Thomas A. Sebeok (ed.), Style in Language. Cambridge, Mass.: MIT. 350-377.

Levin, Israel (1980) The Embroidered Coat. Tel Aviv: The Katz Research Institute for Hebrew Literature and Hakibbutz Hameuchad Publishing House (in Hebrew).

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

Miller, George A. (1970). The Psychology of Communication. Harmondsworth: Pelican.

Rozen Tova (1985). Mediaeval Hebrew Girdle Poems. Haifa University Press (in Hebrew).

Schirman, Jefim (1961). Hebrew Poetry in Spain and Provence. 4 vols. Jerusalem and Tel Aviv: Bialik Institute and Dvir (in Hebrew).

Schirman, Jefim (1961). "How should one Recite the Metred Verse of our Poets in Spain?" in Studies in the History of Hebrew Poetry and Drama. vol. 1. Jerusalem: Bialik Institute (in Hebrew). 84-89.

Schramm, Wilbur L. (1935). Approaches to the Science of English Verse. Iowa City: Iowa University.
Tsur, Reuven (1969). Studies in Mediaeval Hebrew Poetry. Tel Aviv: Daga (in Hebrew).

Tsur, Reuven (1972). "Articulateness and Requiredness in Iambic Verse". Style 6: 123-148.

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

Tsur, Reuven (1987). Mediaeval Hebrew Poetry in a Double Perspective. Tel Aviv: University Publishing Projects (in Hebrew).

Tsur, Reuven (1988). Hebrew Hypnotic Poetry Preliminary Observations. Tel Aviv: The Katz Research Institute for Hebrew Literature (in Hebrew).

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

Tsur, R., Glicksohn, J., & Goodblatt, C. (1990). "Perceptual Organization, Absorption and Aesthetic Qualities of Poetry", in Halász László (ed.), Proceedings of the 11th International Congress on Empirical Aesthetics. Budapest: Institute for Psychology of the Hungarian Academy of Sciences. 301-304.

Tsur, R., Glicksohn, J., & Goodblatt, C. (1991). "Gestalt Qualities in Poetry and the Reader's Absorption Style". Journal of Pragmatics 16/4: 193-206.

Direct reprint requests to:

Reuven Tsur
The Katz Research Institute for Hebrew Literature
Tel Aviv University
Ramat Aviv 69978, Israel

This paper was originally published as

Reuven Tsur and Yehosheva Bentov (1996) "The Rhythmic and Strophic Organization of Mediaeval Hebrew Poetry (A Cognitive Approach)", in Empirical Studies of the Arts 14: 183-206.

Back to abstract
Back to home page

Original file name: Tsur/Bentov/Eng. b2+o/ooÚ.~ - converted on Thursday, 20 November 1997, 21:19

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