The Non-linguistic in Poetic Language. A Generative Approach

Abstract

This article follows a generative approach to language, which argues that language is in part organized by rules and conditions which are specific to language. In Chomsky (Syntactic Structures, 1957), which is the foundational statement of generative linguistics, an example of a language specific rule was the ›transformational rule‹. This means that language may also be subject to other modes of organization, including linear concatenation, or counting, or rules of sequential ordering, which are not specific to language: these are the ›non-linguistic‹ modes of organization of language. In this article I identify some of the non-linguistic modes of organization as they appear in poetic language, and suggest that many aspects of poetic language must in fact be seen as non-linguistic in this sense. This violates a widely-stated hypothesis held by many linguists (at least since Sapir) who work on poetry, that poetic language is a development of the rules or conditions or processes of non-poetic (ordinary) language: this is the ›development hypothesis‹. I begin with verse, which is text which is divided into lines. Lines are not linguistic entities: they are non-linguistic modes of organization of language. I argue that lines are not composed by the linguistic processes which generate ordinary speech, and which start from conceptual structure and apply syntactic rules to generate utterances. Instead, I suggest that lines of verse may be composed directly by non-linguistic processes which do not necessarily begin from conceptual structure; in this sense, conceptual structure – or meaning – plays a less determinate role in the production of poetry. Direct composition of lines bypasses the normal syntactic rules; instead verse lines are composed by linear concatenation of elements, and have no overall syntax and hence have no inherent interpretation. This presents two problems: first, if lines are not composed by syntactic processes why does a poetic text generally resemble a syntactically well formed text, and second, if there is no syntax there should be no interpretation – the line should be interpreted as though it were a random jumble of words – and yet lines are perfectly easy on the whole to interpret. I propose to solve these two problems by independently generating an unspoken ›ordinary language‹ text alongside the poetic text, and using the ordinary text as a filter by which the poetic text is assessed and if necessary edited by the poet in composition. This explains some formal aspects of verse, including its resemblance to the list of words or phrases (another non-linguistic mode of organization of language), and its sometimes ›crazy‹ syntax. It also explains some of the semantic distinctiveness of verse, such the greater use of indirect meaning, and verse's potential to surprise by its meanings. I conclude this section by speculating on how this might relate to problems of individual creative difference. Next, I look at metrical rules as a constraint on verse lines, and apply the theory of Fabb and Halle (Meter in Poetry. A New Theory, 2008), which attempts to solve the problem of how metrical elements are counted (another non-linguistic phenomenon). I show that while metrical rules share some rules or principles with phonological rules, they also share rules or principles with other types of counting and non-linguistic (e. g., musical or dance) rhythmic organization. Thus, again, literature involves a way of organizing language which is ›non-linguistic‹ in the sense that it is not specific to language. One consequence of this approach is to reinforce the notion that the line is the compositional unit; this goes against the traditional metrical notions that the metrical line is composed from feet or other smaller component parts. In the next section, I distinguish two ways in which any kind of form can hold of language. One type of form is ›inherent‹ in the sense that it is given by the linguistic system, and part of the history of production of the text: being a noun or being a sentence would be a kind of inherent form in this regard. The other type of form is ›communicated‹ in the sense that it holds of a text only because the text communicates to us that it does. Genre would be an example of communicated form because we assign a genre to a text as much as we assign a meaning to a text, on the basis of the evidence provided, in the context of our knowledge (of literature), and with varying degrees of certainty. I suggest that many kinds of poetic form hold simultaneously in two ways, and constitute a pairing of two distinct kinds of form: thus, metrical form can be split into a type of form which is inherent and a type of form which is communicated, and the same is true of lineation. This dual status means that both metrical form and lineation can hold in multiple and contradictory ways of the same text: a text can be in two metres at the same time, and have multiple possible simultaneous lineations. Thus, prose poems might be inherently lineated (composed by the same processes as verse) but communicate that they are prose (present evidence on the page that they are prose). In the final section of the paper I examine the widespread formal practice of parallelism. I note that like metre and lineation it has both inherent and communicated aspects: the inherent aspects of parallelism relate to its use of semantic and formal priming in the mental lexicon, and the communicated aspects relate to the ways in which formal parallelism can create relations between previously unrelated words or notions. I recall Jakobson's influential view that parallelism draws attention to the linguistic form of the text, and thus manifests the poetic function, and I reinterpret Jakobson's account so that it becomes an account of coinciding symmetry and asymmetry, which I argue is an aesthetic effect.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 20 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.