Kinds of poetic form

I suggest there are various distinct kinds of poetic form.

One kind of form in a poem is its division into sections. A poem is in continuous sections; for example, it can be divided continuously into a sequence of lines. Sections can be on different layers – lower contained within higher. Thus, lines may be contained in stanzas, both of which are kinds of section. A poem is always divided into sections, at least one level of which is not determined by ordinary linguistic form or meaning.

Many poems also have forms that are systematically added and are not found systematically in ordinary language. I call these the ‘added forms’, and they include metre, rhyme, alliteration, parallelism and perhaps some other forms. The defining characteristic of what I distinguish as an added form in poetry is that it is predictably present in the text. Many of these forms can be found also in text that is not divided into lines, in which case the forms can only be emergent and occasional, not predictably present. The added forms are always defined relative to the sections of poetry, such as the line.

I suggest that each of the added forms of metre, rhyme, alliteration and parallelism should be divided into distinct component parts, which interact to produce the form. This means metre, rhyme, alliteration and parallelism should be treated as ‘modular’ forms. For example, rhyme is made of three component parts, and this can be illustrated from Keats's sonnet: (a) there is a relation of similarity between words that rhyme (‘told’, ‘gold’, ‘hold’ and ‘bold’ rhyme); (b) each rhyming word is located in a specific place relative to a poetic section (rhyme is line-final in Keats's sonnet); and (c) there is a rhyme pattern, which in Keats's sonnet is ABBAABBACDCDCD. These are independent of one another in the sense that the choice of rhyming words is unconnected to the fact that rhymes are line-final, and line-final rhymes can form many other patterns than this one. The three modules of rhyme all combine to form rhyme, but they are different kinds of form.

Poetry has kinds of form that are specialized for language, including metre, rhyme and alliteration. Though parallelism can be found outside language, parallelism that involves language in literature works in specific ways, including a special way of pairing words. But the various forms of poetry have properties that resemble the properties of other sorts of non-linguistic and aesthetic forms. I suggest that we separate these out not as subcomponents of a modular form, but rather as general properties of the forms. These properties include sameness and difference, variety, symmetry and asymmetry. Metre, rhyme, alliteration and parallelism all have these properties. In principle, these properties can hold over any size of text.

I have described the forms that hold in poems, including the division into lines and the added forms of metre, rhyme and so on. There are two ways in which a form can hold in a poem, as I now explain, making a distinction between what I call ‘generated form’ and ‘communicated form’ (Fabb Reference Fabb2002b).

A form is ‘generated’ when it holds within a text because there are specialized psychological processes that assign it to the text. The term ‘generated’ comes from generative grammar, which is primarily a theory of how sentences, words and sounds get their form, but which has also been used as a theory of poetic form. For example, a metre is a set of rules that control the number of syllables in a line, and may also control its rhythm; these rules are sufficiently complex that they have been argued to be similar to the phonological rules that hold of language and which are generated by specialized phonological processes. I suggest that at least one subcomponent of each of the added forms is generated.

A form is ‘communicated’ when the author intends the reader to recognize that the poem has the form. Form, in this sense, is a kind of content, part of what the poem means. Communicated forms are unconstrained in kind and have no special psychological status. As we will see, a form can be both generated and separately communicated. Keats's poem is in iambic pentameter, generated by special psychological processes, and hence in a generated form. But as readers, we infer that Keats intends us to recognize that the lines are in iambic pentameter (this is part of what we need to know to infer that he also intends us to interpret the poem as a sonnet), and in this sense iambic pentameter is a communicated form.

Kinds of form and working memory

We have various kinds of memory. Working memory holds a limited amount of material for immediate processing. In Chapter 7 of this book, I argue that though it is limited, the capacity of working memory is nevertheless sufficiently large to hold at least a whole line; it is large enough to hold a whole couplet, so long as the lines are not too long. Sections of text can thus be distinguished into those that could in principle fit as wholes into working memory and those that are too large to fit. For example, a four-line stanza is likely to be too large, in most cases (unless the lines are very short). However, poems are always divided into at least one level of section small enough to fit into working memory; this is what we usually call the ‘line’.

I have suggested that the various kinds of added form are modular, consisting of distinct component parts. We can distinguish between formal components that might in principle hold over sections of text held as complete units in working memory (usually lines, at most couplets), and formal components that cannot. Metrical rules hold of a small section of text, rhyme and alliteration are both located relative to a small section of text, and a parallel member is a small section of text. On the other hand, the patterns formed by rhyme or alliteration schemes can hold over stretches of text which are too large to fit as wholes into working memory, and the same is true of heterometrical patterns (where different lines are in different metres), and sometimes of parallelistic patterns. To make this point more precise, consider the rhymes in an English sonnet, and the distinction between rhyme location and rhyme pattern. The rhyme is always located at the end of a line, which is a small section of text, small enough to fit into working memory. But the rhyme pattern can hold over four lines, eight lines, or sometimes the whole poem, which is a section too large to fit into working memory. It might be asked how patterns are processed if, as is often the case, the text that supports the pattern is too large to fit as a whole into working memory. One possibility is that the establishment of the pattern requires recall only of small amounts of material from earlier in the text (e.g., previous line-final words in rhyme), and not the whole text. The other possibility is that pattern is not processed wholly in working memory, but involves interplay between working and long-term memory.

Some kinds of form have component modules that must hold over sections small enough to fit into working memory. This could be explained if these modules of the added forms had to be processed in working memory. Note that there is no commonsense argument to be made here, or any argument from what else we know about language: working memory does not have a special status for language processing (Gathercole Reference Gathercole and Gareth Gaskell2007), so it is not obvious that any of the poetic forms should necessarily have to be processed in working memory. The only reason to think that they are is that they are limited in the size of section over which they hold. This leads to a prediction. If a poem has added forms that must be processed in working memory, then it must be divided into sections small enough to fit into working memory. This prediction is met by all the literatures discussed in this book.

This book hypothesizes that poetry is processed section by section, by holding whole sequences of words in working memory and processing parts of their added forms while each sequence is held there. I will attempt to demonstrate this by approaching the problem from two directions. In Chapters 4, 5 and 6, I survey a range of poetic traditions and show that in each case the relevant components of the added forms hold over a poetic section small enough to fit into working memory. In Chapter 7, I draw on psychological accounts of working memory, which show that it is, in principle, possible for these small sections to be fitted there as wholes. None of this proves that this is actually happening in the mind of the reader or hearer as they process poetry; it only shows that it is, in principle, possible.

Poetic sections

This book proposes the following definition of a poem:

At least one level of poetic section in a poem involves a division of the text that is independent of modality, whether the text is spoken, written or signed. Some poetic sections are sufficiently small that they can be processed as whole units in working memory. As the definition makes explicit, there is no requirement that the sections processed in working memory are syntactically or prosodically coherent, or coherent in meaning (semantically coherent). However, nothing prevents this being the case, and there may be some advantages to having the poetic sections also syntactically or prosodically or semantically coherent.

Sections of a poem do not need to have added forms. In many kinds of poetry, there are large sections such as stanzas, quatrains, and verse paragraphs, which need not have any specific added forms. In fact, ‘free verse’ is poetry divided into sections where there are no added forms. Where a section has no added forms, there is no constraint on the size of the lines or other sections. However, nothing prevents the sections of such poems being composed in order to fit into working memory, and there may be some advantages in doing so. These advantages may include the possibility of fitting relatively small coherent semantic units into working memory as chunks, as noted in the previous section.

Generated poetic form

The distinction between generated poetic form and communicated poetic form is proposed and explored in Fabb (Reference Fabb2002b), and also discussed by Versace (Reference Versace2009). ‘Generated’ is a term taken from generative linguistics, the type of linguistics initiated by Noam Chomsky (Reference Chomsky1957) with Syntactic Structures. A form is generated when it is created by a set of rules. Generative linguists have argued that all human languages have sentences that are generated by syntactic rules, and the forms that sounds take in combination are generated by phonological rules. The rules are similar in all languages. The psychological implication is that the ways in which humans process language have component parts that are structured as generative rules. Generative linguistics was applied to the study of poetry in the generative metrics of Halle and Keyser (Reference Halle and Keyser1966, Reference Halle1971), which proposes that metrical form is generated by processes that are universally available to all humans. Not all literatures have metrical verse, but all literatures have the potential for metrical verse.

The metrical theory of Fabb and Halle (Reference Fabb2008) is a descendent of the theory of Halle and Keyser, and I use it to illustrate what I mean by a generated form. In the Fabb–Halle theory, each metre can be thought of as a test for a line of poetry; if the line passes the test, it is in that metre, which is to say that it is ‘well-formed’ in that metre. The metre is a mechanism that takes a line as input and adds a grid structure to the line as output. So, for example, the rules for English iambic pentameter might take as input the second line of Keats's poem, ‘And many goodly states and kingdoms seen’, and the rules generate as output the following bracketed grid:

The line is well formed in part because its grid has just one asterisk on the final level, which is a requirement on all metrical grids. It will be possible to generate a well-formed iambic pentameter grid from any sequence of ten syllables, and the rules will also allow a well-formed grid to be generated from a sequence, which is missing a syllable at the beginning or has an extra syllable at the end. This is as it should be, because iambic pentameter as a metre requires the line to have ten syllables, with some permitted variations. So far, all the grid does is check that the line has the right number of syllables. However, the rules for iambic pentameter also control the rhythm to some extent, and the metre in this case further requires that ‘a syllable carrying primary stress in a polysyllabic word must project to gridline 1’. An asterisk in the grid projects from the syllable directly above it, and in this grid, the gridline 1 asterisks project from even-numbered syllables, which means that even-numbered syllables will be associated with stress. In this line, three syllables come under the control of the rule: the first syllable in ‘many’, ‘goodly’ and ‘kingdoms’ and all project to gridline 1.

The grid for iambic pentameter is generated from the line by the following set of rules (where R is ‘right’ and L is ‘left’):

• Iambic pentameter

  1. (i) Project each syllable as an asterisk forming gridline 0.

  2. (ii) Gridline 0. Starting at the R edge, insert an R parenthesis, form binary groups, heads R. (Optional: skip first asterisk. Optional: final group may be incomplete.)

  3. (iii) Gridline 1. Starting at the R edge, insert an R parenthesis, form ternary groups, heads R. Final group must be incomplete (binary).

  4. (iv) Gridline 2. Starting at the L edge, insert an L parenthesis, head L.

These are iterative rules, rules that repeat and so build a periodic grid, having a repeating pattern. The rules count syllables, and they count in a periodic way that thus allows a periodic rhythm to arise.

Ours is just one theory of generative metrics. In all the different theories, a specific metre is formulated as a set of rules that operate as a check against which a line of poetry is tested: if the line passes the test, it is in the metre. In some theories, this is a test that the line just passes or fails, and so is metrical or unmetrical. In other theories, the line gets a score rather than a strict pass or fail; the higher the score, the closer it is to the metre, which is a way of theorizing gradient metricality. What all the theories have in common is that they consist of a set of rules, constraints and conditions precisely formulated in advance of testing any particular line against them; the rules are not dependent on the line itself or its context. Furthermore, the rules build structure, though not always a grid, and this structure is considered to hold as a determinate fact about the line, a fact that can be discovered, just as linguistic structure holds of any piece of language. Theories of generative metrics also all assume that the rules, conditions, constraints and structures are produced by specialized psychological mechanisms, which are below the level of consciousness, just as linguistic rules and structures are. The combination of determinate rules, determinate structure and psychological reality means that this type of metrical form is what I call a ‘generated form’.

In the present book, I propose that metre, rhyme, alliteration and parallelism all have components that are generated forms. Generated forms are limited in kind, because they are adaptations of specific psychological mechanisms. Kiparsky (Reference Kiparsky and Freeman1973) notes this, and compares verbal art with the visual and other arts that, in contrast, vary much more greatly over time and space; poetry varies less, because of its close dependence on language; he says it ‘fits’ language. All languages are underlyingly similar in their generated forms. Hence, the generated poetic forms that have developed from those languages ensure their poetries are also similar. I suggest that all the generated poetic forms are processed over sections of text, such as lines, which are held as wholes in working memory.

Communicated poetic form

The notion of ‘communicated poetic form’ is just that authors are able to give forms to texts and that we can recognize them. Most non-linguistic accounts of poetic form are actually accounts of communicated poetic form, without being theorized or categorized as such. A communicated form holds of a poetic text when the audience attributes that form to the text.

In Fabb (Reference Fabb2002b), I explain communicated form as a phenomenon by using linguistic pragmatics, formulated in Relevance Theory, this being the theory of communication developed by Sperber and Wilson (Reference Sperber and Wilson1995). The following are the three processes by which communicated form comes to hold of a text:

1. (i) It is possible for an audience to formulate a thought that is a direct observation about a text. Such a thought might be, ‘This text is titled “sonnet”’ or ‘This text has fourteen lines’. I suggest that these direct observations are analogous to the ‘literal meanings’ of a text, or what Sperber and Wilson call the explicatures of an utterance.

2. (ii) Each such formulated thought is a premise in a logical deduction. It is combined with other premises, which may similarly be drawn from observation of the text or drawn from knowledge of literary form, and the combined premises can lead to a conclusion. For example, the direct observation might be, ‘This poem has fourteen lines’, and the premise borrowed from knowledge of literary form might be, ‘If a poem has fourteen lines, it is a sonnet’, and by the logical rule of modus ponens, the conclusion would be, ‘This poem is a sonnet’.

3. (iii) The conclusion is established with a certain degree of confidence, depending on the confidence with which the premises hold. The premise, ‘If a poem has fourteen lines, it is a sonnet’, can only be held with a middling degree of confidence because though many fourteen-line poems are sonnets, not all of them are. This middling degree of confidence is inherited by the conclusion. I suggest that forms are strongly or weakly communicated depending on how certain the audience is that the forms hold. Where several strands of inference point to the same conclusion, that conclusion is strengthened. For example, other observations about the same poem, such as its being titled ‘sonnet’ or its having a specific rhyme scheme, might generate premises which lead to the same conclusion that it is a sonnet, and each of these deductions will strengthen the view that it is a sonnet.

Communicated poetic form holds of a text as the content of an inference about that text. Presumably, the audience will treat these inferences as intended by the author. This is what makes the form ‘communicated’ and hence the content of an implicature, which is an intended implication. An author may compose a text with the informative intention that thoughts about its form could be derived from it. When a poet writes in iambic pentameter, she is communicating to us the thought that it is in iambic pentameter. This is a thought that can have further inferential consequences, including thoughts about genre, about other poems in this metre, and so on. A key notion is that the form is ‘ostensively’ communicated, which means that the author makes us notice the form. A line can be communicated as a form in a text by writing it out as a distinct sequence on the page, or in speech by pausing at the end of the line: these are ostensive markings of the line boundary, and hence of the line as a communicated form. These devices do not create the lineation, but instead point to the lineation that is conceptually independent of them.

There is no limit on what can be a communicated form of a text, since a text can be arbitrarily given any kind of form, and that form recognized and named by the reader or hearer. Consider, for example, the communicated form expressed by the rule that forbids all words with the letter ‘e’, a rule that holds in the 1969 French novel La Disparition by Georges Perec. Perec was a member of Oulipo, an avant-garde movement that invented many innovative forms of this kind (Mathews and Brotchie Reference Mathews and Brotchie2005). The notion of ‘communicated form’ is less theoretically specific than the notion of generated form, since at its simplest it just says that practitioners and audiences are capable of consciously categorizing and manipulating the forms of texts.

What appears to be the same form can sometimes be analysed into two distinct forms: a generated version of the form and a communicated version of the form. Metre is a kind of generated form, but it can also be a kind of communicated form (Fabb Reference Fabb2002b: 88–135). For example, iambic pentameter can be a set of rules and conditions that operate below the level of consciousness, and this is its version as a generated form. But iambic pentameter can also be a set of explicitly known and stated rules which can be found in a poetry book, along with cultural knowledge about iambic pentameter, and this is its version as a communicated form. These two kinds of iambic pentameter are not the same, and a text might have one but not the other. In Fabb (Reference Fabb2002b: 128), I discuss an English poem by George Canning called ‘Sapphics’, which communicates that it is in a sapphic metre both in its title and by using a pattern of eleven unstressed syllables (σ) and stressed syllables (σ́) of σ́σσσ́σσ́σσ́σσ́σ, which loosely imitates a quantitative metre called sapphics. However, following Attridge (Reference Attridge1974: 216), I show that these lines are actually in iambic pentameter, choosing the option of inverting the first stress and adding a syllable at the end, and keeping these options fixed throughout the poem to conceal that it is in iambic pentameter and communicate that it is in the different metre of sapphics.

Thoughts can be communicated from author to audience without being fully specified. For example, a form can be communicated even when there is no name for it. Sherzer notes that Kuna texts are treated by their authors as divided into lines because they teach them line-by-line, but there is no Kuna word for ‘line’, hence no explicit category. Nevertheless, it seems plausible that when a Kuna audience hears a poetic text, they recognize its intended form, even if they have no name for it. Thus, there is successful communication of form.

Communicated form is an essential component of Bauman's definition of verbal art (i.e. literature). Bauman (Reference Bauman1975, Reference Bauman1984) argues that verbal art is verbal behaviour that must be evaluated relative to agreed rules, including rules of form. Rules of form thus provide something that can be evaluated, and this is partly why verbal art has form. For example, Bolivian Quechua singers compose coplas, a type of poetry with specific formal rules. Solomon (Reference Solomon1994: 396) describes the close attention paid by the audience who laughed at formal mistakes, and subsequent copla singers commented on them. Part of what the poets communicate is the formal rules to which they seek to adhere, and on the basis of which they are judged. In our terms, the forms must be communicated in order for the art to be evaluated.

In this book, I offer a formal definition of poetry, which should enable us to classify any text as ‘poetry’ or ‘not poetry’. But, like any generic category, poetry is also a kind of communicated form. Poetry as a type of communicated form is a category that can be extended freely, with an awareness of its range of other meanings. It is this notion of poetry and the line which Holden (Reference Holden1988) uses when he describes how, in free verse, ‘the line alludes to and plays off our expectations regarding traditional, patterned prosodies’: poetry in this sense is a meaningful category communicated by the poet, part of the way the text is understood explicitly and not a set of formal principles fitted in a specific way to human psychology.

Communicated forms place no specific demands on psychology and do not involve specialized processes. They do not have to be calculated over sections of text that are held as whole units in working memory.

The line as generated form or communicated form

It is clear that in many poems, the line is a type of communicated form, a form we are aware of and that we know the author intends for us to notice (Fabb Reference Fabb2002b: 136–77). There is often ostensive evidence for the line boundary: a pause in speech, vocables at the boundary or layout on a page are all evidence for the line as a communicated form. These are ways of using the forms of language, or of aspects of performance, to correlate with and perhaps provide evidence for the division into sections such as lines. Similarly, the presence of the line boundary can be inferred from the metre, or the presence of rhyme, alliteration or parallelism.

However, the line is also in many cases a generated form. In metrical poetry, the metre is a type of generated form that holds over a specific stretch of text: the boundaries of the text are determinate and strictly related to the operation of the metrical rules. This suggests that the line is itself a type of generated form, perhaps partially brought into existence as a section of text by the operation of the metrical rules. Similarly, a rule that strictly places a rhyme at the end of a line therefore strictly determines where the line boundaries fall, and makes the line a type of generated form. The same kind of argument can be made for the other added forms of alliteration and parallelism: the subcomponents of these forms, which depend on a specific section of text, also depend on this section having the kind of determinate properties we associate with a generated form. This all suggests that the line, and possibly other small sections such as hemistichs and couplets, are generated forms, independent of the generated forms added to them.

In a specific text, the line can be both a generated form and a communicated form. The generated line and the communicated line need not coincide in a particular poem. Updegraff (Reference Updegraff2011: 648) cites Gioia who argues that William Carlos Williams's poem ‘The Red Wheelbarrow’ appears to be eight lines of free verse but is actually an iambic pentameter couplet, ‘a free verse poem only in its visual arrangement of sound’. This means the poem has different lineations in its communicated and generated forms: its communicated form is that it is in eight free verse lines, and its generated form is that it is in two iambic pentameter lines. Holden (Reference Holden1988) analyses two free verse poems to show that they can be relineated to produce other kinds of text, an ambiguity in lineation that is an interpretative ambiguity, not a consistent structural ambiguity. Fabb (Reference Fabb2002a) analyses Matthew Arnold's ‘Dover Beach’ as a poem ambiguous in its communicated sections. ‘Dover Beach’ has generated forms, including its division into lines, the varying-length iambic metres of the lines and the rhyme at the end of lines. When ‘Dover Beach’ is printed on the page, its lineation is ostensively marked. However, if we experimentally remove the visual lineation and present it as prose, it proves difficult for readers to work out what the generated lineation is, and this is in part because it is both tempting and easy to read its beginning as a series of iambic pentameter lines, even though this is not its generated lineation. Thus, the text has a generated form which is not iambic pentameter, but the text weakly communicates that it is in iambic pentameter. Furthermore, the ending of ‘Dover Beach’ weakly resembles a kind of truncated sonnet, and thus we might say that it weakly communicates towards the end that it is a sonnet. Though the poem has a fixed and stable generated form, its communicated forms are uncertain and unstable.

The larger sections of text that are not specific to poetry, including the verse paragraph or the book of an epic, are too large to be generated forms and do not interact with generated forms. They can only be communicated forms, just as paragraphs or chapters in prose are communicated forms (unregulated by the generative syntax). There are no covert and complex aspects to them as forms, they are not limited in size or structure, and they are overtly evidenced, and so they have none of the characteristics of generated forms. Häublein (Reference Häublein1978: 47) notes that Romantic and post-Romantic British poets tended to have characteristic stanza openings that may include apostrophes, exclamations, expletives, interjections and so on; his observation fits with the stanza being a type of communicated form, which is ostensively evidenced. This has the consequence that the larger sectionings of poetry can be ambiguous, as we saw for the sectioning of Keats's sonnet.

The modularity of poetic form

I earlier proposed that each of the added forms of metre, rhyme, alliteration and parallelism is a composite of distinct kinds of form. I call this a modular approach to poetic form because it treats a form as a combination of formal modules, each with their own principles of operation.

To illustrate modularity, consider the metrical form that holds of Keats's poem. The metricality of this poem has the following two distinct components or modules:

1. (i) The metre, called iambic pentameter, controls the number of syllables in the line and the rhythm of the line. I suggest that this is a generated form because it holds of a limited section of text, it has access to covert linguistic form and it operates in a complex way.

2. (ii) Every line of the poem is in iambic pentameter. This is the metrical pattern of the poem (an isometrical pattern).

The metre of a line and the pattern of metrical lines are two distinct components, which may be unrelated to one another even though they are both part of the larger metre of the poem. The fact that a line is in iambic pentameter tells us nothing in principle about what the metre of an adjacent line will be; iambic pentameter lines can combine in English poems with various other types of metrical lines, and these combinatorial possibilities are not determined by the component metres but instead are a free combination of metres. Some other poem may be in an alternating metrical pattern of ABAB, where odd lines are in the same metre and even lines in another metre, but the possibility of such a pattern is unrelated to what actual metres are used.

There is a complication, though. Pattern is, in some traditions, strictly controlled, and we will see traditions such as Welsh or Thai or Javanese in which named ‘metres’ control the patterns of whole stanzas. West (Reference West1973: 184) notes that heterometrical patterns fall into characteristic stanza types in Indo-European languages: for example, where a longer line is followed by a shorter line and this pattern is repeated to form a quatrain, or where three longer lines are followed by a shortened line, or where two shorter lines are followed by a longer line. In Fabb (Reference Fabb2002b: 178–214), I argue that the metrical patterns forming groups of lines are communicated form. I discuss whether pattern might be generated by rules similar to those of the line's metre and conclude that ‘there is not obviously a need to appeal to some special generative mechanism, with the ontological and cognitive costs this entails, in order to count lines into a line group’ (Fabb Reference Fabb2002b: 203). Thus, I propose to distinguish between the metre of the line, which is a type of generated form and is subject to universal principles, and the combination of lines into the stanza, which is a type of communicated form and probably not subject to universal principles. The relative freedom of pattern can be seen in the fact that patterns can be fixed to produce visual images on the page, in pattern poems.

An important difference between these two modules of metre is that the metrical rules apply to a section short enough to fit as a whole into working memory, while the pattern can hold of any size of section.

We can decompose rhyme and alliteration each into three modules:

1. (i) The first module of form puts a word into a specific rhyme class or a specific alliteration class, such that a word will, in principle, alliterate or rhyme with other words of the same class. For example, in Old English the three words scrifen, scrād and gescād are classed as belonging to the same alliterating type because their stressed syllables begin with sc, phonetically [š]. Hence, they can all alliterate. The principles by which words are put into a class can be complex. For example, in some types of Czech rhyme, words can rhyme only if their endings are grammatically distinct (Worth Reference Worth, Armstrong and Schooneveld1977). Many generativists have argued that this is a type of generated form, in part because it involves access to underlying linguistic form, which is itself a type of generated form (Fabb Reference Fabb1997: 111–36). The organization of words into alliterating or rhyming classes has little consequence for the hypothesis of this book about the chunking of lines in working memory, since the formation of rhyme or alliteration classes can be done independently of the text and outside working memory.

2. (ii) The second module of form locates each alliterating or rhyming word. Determining where a word falls does not usually refer to the alliteration or rhyme multi-line pattern, so the location is independent of other forms. For example, in many poems a rhyming word must be at the end of the line. I suggest that locations for alliteration and rhyme are always determined relative to a small section of text, at most a long line or couplet, and that this relatively small section can be held as a whole in working memory. Location is always relative to a small section of text and it may be that location is a type of generated form. Furthermore, in many traditions, the location of a rhyming or alliterating word is sensitive to metrical structure: a specific numbered syllable may need to be the rhyming syllable, for example, and this interaction with another generated form suggests that rhyme location or alliteration location is also a generated form.

3. (iii) The third module of form is the pattern of alliteration or rhyme. The pattern might extend over a section of text, or over a stretch of text that is not a coherent section, or over the whole text. For example, Häublein (Reference Häublein1978: 36) describes English stanzas that have an ambiguous structure, arising largely because of the use of rhyme; thus the nine-line Spenserian stanza has eight iambic pentameter lines and one final iambic hexameter line and can be divided into lines rhyming ABAB-BCBC-C, isolating the long last line, or ABAB-BCB-CC with a final couplet. This is an ambiguity arising from the possibility of reading a pattern in different ways, and suggests that at least in this case the pattern is a communicated form, as is the ambigous division into sections it implies.

There is no principled dependency between these three aspects of rhyme, and this justifies separating them out into distinct modules. We will see that there are sometimes dependencies in specific traditions, where it is impossible to separate the location from the pattern; in these traditions, it may be that the pattern is now a kind of strict added form, and perhaps a generated form. Such patterns always hold over small sections of text, small enough to fit into working memory capacity, and may be different in kind from the more diffuse and larger patterns seen in the Keats sonnet.

Lexico-syntactic parallelism is found regularly in some types of poetry and can be separated into three modules:

1. (i) One modular form is the grouping of two words into a set, where one word is in one line and its partner in another line. This is often found in parallelistic poetry. I suggest that this is a type of generated form, because it resembles other linguistic processes that pair words, such as idioms, compounds and so on, and can have specific semantic implications, as I show in Chapter 6. The pairing of two words is a little like the formation of a rhyming or alliterating set, and can take place outside working memory; the fact that pairs are remembered and reapplied in different texts suggests that it is best understood in terms of long-term memory.

2. (ii) Another modular form is the constitution of a parallel member, which is a syntactically specific sequence parallel to another syntactically similar sequence. I suggest that this is a type of generated form because it controls syntactic structure, which is a type of generated form. It also holds of a small section of text, as we would expect for a generated form.

3. (iii) The third type of modular form in parallelism is the parallel pattern. The pattern is the relation between two parallel members (and paired parallel words). In principle, this is a type of communicated form, like all patterns, though in practice the parallel patterns are often adjacent lines within a small couplet and so in principle might be held as wholes in working memory. It is possible that in order to establish a parallel pattern, it is necessary to hold the verbatim forms of the pair of lines both in working memory at the same time; in other words, the parallel pattern might require something different from the text than required by the rhyme, alliteration and metrical patterns. This might correlate with the fact that parallelism is often in small couplets, which means that it is often possible to fit a whole parallel couplet into working memory at the same time. On the other hand, there are also examples of parallel patterns that hold over longer or discontinuous sections of text. It may be that some types of parallel pattern are best understood as generated forms holding over small sections and possibly exploiting generative rules found also in the syntax, involving parallel structures. In contrast, other types of parallelistic pattern are communicated forms holding over larger stretches of text.

The important point to take away from this discussion is that a form can have some component parts, which by hypothesis must be processed in working memory, and other component parts, which need not. This is the advantage of taking a modular approach to poetic form.

Properties of forms

All the added forms are characterized by general properties, such as sameness and difference, variety, symmetry and asymmetry. The idea of calling these ‘properties of forms’ is intended to bring out the sense that they can characterize not only poetic form but also musical or visual form. Properties of forms can be generated or be communicated. In some cases, the properties may be processed in specialized psychological ways, and in some cases they seem to be restricted to small sections. In other cases, we may attribute the properties to the text overtly, as communicated forms.

I illustrate here the properties of forms by examining how they hold of metricality as a modular kind of form. Metricality has two component forms: the metrical rules that hold of the line, which is a type of generated form, and the pattern formed by combining metrical lines, which I suggest is generally only a type of communicated form. The line may be internally characterized by sameness and difference, and particularly by rhythmic variety or variation. Iambic pentameter is a metre that implies a regular rhythm not always realized in actual lines, because the metre underspecifies the rhythm rather than strictly defining it. The uncommonness of fully regular iambic lines was noted by Abbott (Reference Abbott1765: 17–22), who proposed that the sameness of the metre in each line is combined with the differentiated rhythms of the different lines to avoid monotony without destroying the musicality of the verse. Many metres underdetermine a strict rhythm and allow variation. How underdetermined the rhythm is can vary depending on location within the line. The beginning of the line tends to be looser and the end of the line stricter, and this is a patterning producing looseness as a kind of difference and producing strictness as a kind of sameness, which together produce a gradual asymmetry or difference as the line progresses. This strict ending is often called ‘cadence’. It is a theoretical question whether cadence is determined by the specific rules that generate a metre or whether it is some general property of sequential structures of different kinds, including non-linguistic structures. I suggest that it is a property that holds of forms, because cadence can operate at different levels and involve different kinds of form.

Variety is another property of forms. It seems that in some cases, variety is a specific property of sections of text small enough to fit into working memory. Golston (Reference Golston2009) argues that in the Old English poem Beowulf, adjacent hemistichs within a long line are always rhythmically different, but adjacent hemistichs across a line boundary can be different or the same; thus, variety is specific to the long line. Turpin and Laughren (Reference Turpin, Laughren, Gawne and Vaughan2013: 409) identify a similar constraint in Warlpiri songs, that ‘a line made up of identical text in both hemistichs must not have an identically repeated rhythm’. Allen (Reference Allen1973: 106) says of the Greek dramatist Euripides that in his early iambic verse, no more than one pair of light syllables can be resolved into one heavy syllable in each line, which is another type of variation scaled to the line. In these cases, the limitation on the size and nature of section suggests that variety may be a property of the generated metrical form.

In other cases, variety is a property of whole texts, in which case it can only be part of the communicated form of the text. For example, Virgil ensured that no line in his Latin poem Aeneid was repeated; this cannot be determined by any specific psychological process but can only be a general high-level judgement over the text. In Fabb (Reference Fabb2013b), I examined a number of Dylan Thomas's recorded performances of his own and others’ English poetry. In order to assess how his spoken sections corresponded to lines, I looked at whether he pauses at line endings. Sometimes, Thomas pauses at a line boundary and sometimes not; sometimes he pauses midway through a line and not at the end. This partly reflects his punctuation, but not completely, and different performances can vary in pausing. Consider, for example, his performance of his own poem ‘In my craft or sullen art’ (recorded on the LP Dylan Thomas Reading Complete Recorded Poetry). This is a poem of 20 seven-syllable lines with line-final rhyme. Lines tend to have three or four stressed syllables, but not in a consistent pattern; though it is primarily a syllable-counting poem, there is perhaps a second loosely defined rhythmic metre in operation, making this a polymetric text (Fabb and Halle Reference 199Fabb and Torrego2012c: 81). In Dylan Thomas's performance, thirteen of the lines begin and end with pauses, and the other seven do not; some lines are run together without a pause and sometimes there are pauses in the middle of the line. If we characterize each sequence ending in a pause as a rhythmically defined section, then we can make a comparison between the poem as actually performed and the poem as it might have been performed if Thomas had treated every line as a rhythmic section by pausing at the end of every line. In this alternative performance, half of the lines would have involved a repeating rhythmic pattern. But in his actual performance of the poem, almost all the repeating rhythmic sections are broken up, and in his performance there is only one repeated rhythmic section. For example, the lines in (i) each have rhythmic patterns that are found in other lines, but they are performed as the rhythmic sections in (ii), thereby producing rhythmically unique units:

1. (i) But for the lovers, their arms

   Round the griefs of the ages,

2. (ii) But for the lovers,

   their arms round the griefs of the ages,

Thomas thus restructures the lineation into different sections in performance, but remarkably, every restructuring has the effect of creating rhythmic variety, measured over the whole of a twenty-line poem. In other performances of other poems that I have examined, Thomas always increases rhythmic variety by sometimes not pausing at line end and sometimes pausing mid-line, but this is the only poem in which every restructuring works towards producing total variety. Form is here manipulated to produce variety, something we will see elsewhere in poetry. Consider the status of rhythm relative to the line on the one hand and to the pause-bounded rhythmic section on the other. Relative to the line, rhythm is controlled by the metre, counting three or four stressed syllables per line. Relative to the rhythmic unit, rhythm is controlled by a constraint on variety holding over the whole poem. Some forms such as metre hold of small poetic units such as the line. Rhythmic variety here holds over a whole poem, which is a much larger unit, and this suggests that variety is here part of the communicated form of the poem; it is difficult to see how variety over a whole poem could be assessed by some specialized psychological process of the kind found in generated forms.

Variety is a characteristic found everywhere in poetry. In oral literatures, the same underlying work can be performed in variant ways, so that an oral epic may be different in different performances. In many traditions, some variation is encouraged. Fagborun (Reference Fagborun1990: 168) discusses Yoruba counting poems, which have a fixed introductory line but can then vary, ‘and there is credit for an artist who can deviate reasonably from what s/he was taught’. The term ‘multiformity’ is used in Oral Theory to describe ‘an elastic variability [which] lurked at the very heart of the oral epic singing technique’ (Tate Reference Tate2011: 343). A relevant issue is whether the kinds of variety found in poetry are essentially the same property as the kinds of variety found in non-linguistic media. For example, Wade (Reference Wade2009: 9) discusses principles of variety in music, citing for example Irish musicians who ‘practice and discuss the aesthetic of never playing a tune the same way twice, keeping its melodic identity intact, but playing subtle variations each time around. They embellish their playing with ornaments and small melodic or rhythmic variations that make each repetition of a tune slightly different’.

Symmetry and asymmetry (Fabb Reference Fabb2009c) are properties of forms that are found in all media, including painting, sculpture and music, not just in language. Symmetry and asymmetry can be properties of a whole poem. For example, Japanese haiku consist of a short line, followed by a long line, followed by a short line. Many pattern poems are governed by principles of symmetry and asymmetry, with bilateral symmetry particularly common where the left and right halves of the pattern are the same. Here symmetry and asymmetry are visual properties and at the same time properties governing the heterometrical structure of the verse. The common property of rhythmic looseness at the beginning and strictness at the end of a line is a type of asymmetry. Another asymmetry found in poetry and other uses of language is manifested by the law of increasing numbers. The law of increasing numbers is the tendency for longer material to come later and is a property both of the poetic line and other linguistic sections. In some languages, the law is a characteristic not just of poetry but more generally of elegant speech (Allen Reference Allen1973: 119, citing Behagel Reference Behagel1909: 139). Peck (Reference Peck1884: 64) says that in Latin poetry and prose, ‘[i]f the words in alliteration are unequal in length, the shorter one usually precedes’. Allen (Reference Allen1973: 119) cites Demetrius, who says that ‘in compound sentences the final colon should be the longer’. It can be seen in word formation: the Sanskrit grammarian Pāṇini noted that in copulative dvandva compounds the element with fewer syllables comes first. In poetry, the law of increasing numbers can be seen in traditions in which shorter words are earlier and longer words later in the line, as in the Finnish and relatedKalevala-metre poetry (Leino Reference Leino1986). In rhopalic verse, words are ordered such that each word is one syllable longer than the word that precedes it (Hartman Reference Hartman1980: 18). In general, symmetry, asymmetry and subkinds of asymmetry such as the law of increasing numbers are not restricted to relatively small sections and do not obviously depend on specialized psychological processes. In general, they might be best understood as part of the communicated form of the text.

Shattuck-Hufnagel and Turk (Reference Shattuck-Hufnagel and Turk1996: 203, 234), citing Grosjean and others, say that symmetry is one of a group of factors that can influence parsing into prosodic constituents. A prosodic constituent is a section of speech with its own distinctive characteristics and may, for example, be preceded and followed by a pause. They note findings from Gee and Grosjean (Reference Gee and Grosjean1983) and Grosjean, Grosjean and Lane (Reference Grosjean, Grosjean and Lane1979) consistent with the idea that syntactic constituents may be divided into parts by prosodic boundaries when these boundaries create a more equal partitioning of the utterance. However, while there is a tendency towards symmetry in the prosodic constituency of speech, a different tendency towards asymmetry sometimes operates in poetry. A caesura rule specifies that a word boundary must fall at a specific place within a line, and this can be a way of dividing the lines into parts. In some traditions, the caesura falls in the middle of the line, dividing it into two symmetrical parts, as in the Frenchalexandrin. However, in other cases, the caesura divides the line into two asymmetrical parts, as in the French décasyllabe where it falls after the fourth of ten syllables, or in Homer's Greek dactylic hexameter, where a caesura rule prevents a word ending in the exact metrical centre of the line between the third and fourth feet, instead requiring it to fall just next to the centre. Caesura placement rules are probably examples of generated form (and hold over small sections of text) and are thus an example of a generated form that has the property of asymmetry.

Symmetry is an important characteristic of sign language poetry, where the use of two-handed signs is a way of producing symmetry in three-dimensional space. Russo, Giuranna and Pizzuto (Reference Russo, Giuranna and Pizzuto2001: 100) show that two-handed signs are favoured in poetic LIS (Italian Sign Language) in comparison with ordinary LIS. Sutton-Spence and Kaneko (Reference Sutton-Spence and Kaneko2007) discuss various ways in which symmetry is used in sign language poetry. Some of the symmetries resemble those in written pattern poems, and include ASL (American Sign Language) poems that follow manual alphabets or numerical sequences. In other poems, such as an SLN (Sign Language of the Netherlands) poem by Wim Emmerik, symmetry is maintained throughout the poem, except at the end, to iconically represent aspects of the meaning of the poem, which here is about the fall from the Garden of Eden. In some sign language poems, symmetry fits with lineation, where it might also be interpreted as a type of parallelism, but it is also possible for poetic symmetry to hold overall in a text, and not to be constrained by structure. Symmetry in sign language poetry appears sometimes to be a property of the generated forms of the text, and sometimes a property of its communicated forms.

Combinations of forms

A poem can have various forms, some generated and some communicated. There are some poetic traditions that combine many kinds of form. Such poetry may be difficult to compose, and its value comes partly from the skill with which the formal rules are obeyed. Examples include the Chinese regulated poem with its tonal metre combined with rhyme and parallelism, or the Thai khlong with a tonal metre and rhyme, or Icelandic dróttkvætt with a metre that is sensitive to stress and quantity as well as controlling rhyme and alliteration, or Irish metres that have a syllable-counting metre with cadence, rhyme and alliteration. Are there any limits on what forms can coincide in a poem? This is a difficult question because there might be stricter limits on generated forms than on communicated forms, and it may be that forms can co-occur to a greater extent if they do not interact. The survey of metre, rhyme, alliteration and parallelism in Chapters 4, 5 and 6 will suggest that there are no principled limits on what forms can co-occur. Some co-occurrences are more common than others, and some may be unattested, but this may be historical accident rather than revealing any underlying principle.

In Fabb (Reference Fabb2009b), I suggested that in a tradition with three or more types of added form, there is usually one and only one dependency between distinct forms, where a form must be defined relative to another form. For example, the location of rhyme might be dependent on the metrical form of a text, but if it is, then the location of alliteration will not be dependent on the metrical form (instead, the alliterating words will be distributed within the line without regard for the metrical structure). This suggestion relates to Berlyne's (Reference Berlyne1960, Reference Berlyne1974) proposal that there is a general preference for a middle level of complexity in aesthetic objects. Berlyne developed a psychological account of aesthetic experience in which he claimed that complexity stimulated an aesthetic emotion. He argued that aesthetic objects are at their aesthetically most effective when they are at a middle level of complexity – that is, neither too simple nor too complex. He developed Wilhelm Wundt's idea that people prefer stimuli with medium arousal potential over stimuli with either low or high arousal potential. This finding, however, is contested by Martindale, Moore and Borkum (Reference Martindale, Moore and Borkum1990), who review experimental work that bears on Berlyne's proposals as part of their psychological aesthetics which promotes prototypicality as a central part of aesthetic experience instead of Berlyne's complexity and other variables. They argue that preference is related to various discrete stimuli in a variety of ways, and that in some cases the results are the opposite of those predicted by Berlyne, such that people prefer small or large complexity over medium complexity, or in some cases prefer increasingly complex texts with no upper limit. They also claim that complexity is relatively insignificant in determining preference when compared with a preference for meaningfulness. However, Berlyne's notion that there is a preference for a middle level of complexity continues to be a source of research. Thus, Reber, Schwarz and Winkielman (Reference 208Reber, Schwarz and Winkielman2004: 373) connect this to the aesthetic effects of fluency of processing, and suggest that with low levels of complexity, the source of the fluency is very salient and reduces the effect of fluency. They say that with high levels of complexity there is not enough fluency of processing to produce the aesthetic effects.

Summary

This chapter discussed a range of questions relating to poetic form and proposed some distinct ways of thinking about form. The key claim is that poetic forms are modular, and that some of the modules of a form must be processed over small sections of text, and perhaps are constrained to be processed in working memory. These types of form can only be generated forms, as communicated forms are not processed in any special way. Poetic forms share formal properties such as sameness and difference, symmetry, asymmetry and variety with other kinds of form. In general, these formal properties need not be processed in working memory, but there are some cases where they appear to be restricted to small sections and perhaps involve working memory.