Working Papers in Linguistics 13 (1993): 79-90 University of Melbourne


Aspects Of Repetition in Discourse

 Thor May


Repetition.jpg1. Introduction

It is often claimed that  language is a system for communicating information. In fact, language has a multiplicity of functions, but when it comes to information, that which is to be given significance is always framed by the known, hence repeated, elements. The organization of language is largely a matter of what is repeated, when, where, why, by whom, how and how often.

For the purposes of this analysis, I will take a much broader view of repetition than is normally found in linguistics, considering a cline from local (often idiosyncratic) repeating clauses or phrases to stable units such as lexical items which have become formal, generalized tokens in the language. This is not a paper which proposes a neat solution to some small puzzle in a linguistic model. Rather, it outlines for further study some properties of a very general phenomenon.  

2.  The concept of repetition

 Some notion of repetition can be immediately understood, even by naïve observers. On closer inspection however, the concept turns out to be rather slippery. It is worth exploring the general characteristics of non-linguistic repetition a little in order to sharpen our approach to the linguistic variety.  

If there is an event E which is repeated, then E is understood to exist only on the occasions of its repetition, not in the intervals between E1 and E2. If there is an object O which occurs repeatedly, then O is normally held to exist not only on the occasions of its appearance, but also during the intervals when it is not observed. At least, this is the working assumption of almost everyone in their daily lives, quantum quibbles about Schrodinger's cat notwithstanding. Now what about a cognitive entity such as a word. Is it an object or an event? Does it exist between spoken repetitions? If words have the status of objects, what about phrases or sentences? Do they exist between repetitions, or are they recreated from something else? The language of this thesis will often appear to presume not only the occurrence, but the duration over time of numerous entities. This is a convenient convention, no more, and exactly what we do every time an abstract noun is deployed. I remain open to argument about the ontological status of such entities. 

Interval duration itself is tricky. If X is repeated, typically we then suppose its occurrence in a time frame must have been punctuated by a period of non-X. At least two serious difficulties arise with this simple proposition. Firstly, is repetition an artefact of the observer, or of an operation? Imagine that you are watching a computer monitor. The image is steady. In fact an electron gun is retracing the image thousands of times a second. Is the assertion of repetition therefore always a statement relative to a particular frame of reference (in this case, either the observer or the electronic process)?  

The second difficulty is a paradox of inclusion. You turn off the moving cinema image, which to your eye has appeared seamless. The film, when you examine it, is a series of frames, and adjacent frames seem to be the same, repetitions in fact. Yet any two frames separated by twenty others are clearly different. At what increment do we say that the description of repetition no longer occurs? Or perhaps I know you as a person. I have a name for you, anticipate your behaviour with some success and have some confidence that you are the same person today that I met yesterday. Yet a biological treatise tells me that you are a crawling mountain of cellular organisms, millions dying off at any given moment and others being born, so that the whole squishy heap is replaced at regular intervals. Are you a repetition of some prior self ? It all seems to turn on scale (as in magnification) and the perspective from which we deal with a particular phenomenon.  

The problems of perspective and scale affecting a view of repetition are central to linguistics. Yet one might surmise that few linguists have sorted out in their own minds whether they are the person watching the computer monitor or the electronics technician, the cinema buff or the laboratory developer, the speaker's friend or the biologist.  

3. Repetition as evidence: three perspectives 

Let us suppose that two linguists are studying discourse processes. What will they consider to be evidence? One is an enthusiast for Conversational Analysis. He analyses transcripts in fine detail and develops an elaborate model of dyadic exchange. He sticks to the transcript evidence and consciously avoids speculation about cognitive processes.  

Repetition here is strictly a matter of people saying things twice, and its significance, if any, is defined as a property of the text. For example, a text may "have" cohesion, meaning repeated or anaphoric elements. The relationship of text cohesion to thought, memory, or even to understanding is none of the first linguist's business. This man is the heir (though perhaps not fully aware of it) to a tradition of scientific behaviourism which dominated Western thought for much of the 20th Century and generated an impressive body of research.

The second linguist has been influenced by currents of rationalist thought. She may concede the tidiness of the first linguist's model, but find its interpretation distorted or sterile. Yes, the conversational analyst, within the special selectivity of his model, may trawl up repeating elements from a text, but so what? What significance do they really have for speaker and listener? Are they inserted by design? At what level (morpheme, word, clause, topic..) does a speaker have conscious control over her language generation, including repetition? And what about the torrent of inner, unspoken, often fragmentary language that washes the edge of her consciousness day and night? Surely other people experience this? Surely public spoken language is part of the same ocean as this inner voice, and cannot properly be interpreted in isolation?  

The two linguists clearly have different perspectives, which will deeply affect the questions they ask and the answers they accept as meaningful. The behaviourist claims to be objective. The rationalist will admit subjective evidence. Both implicitly accept an ontological dualism which goes back to the mind-body debates of the European Enlightenment, and earlier.

 A third view, implicit in the Madhyamika Buddhist tradition of mindfulness-awareness, rejects the dualist position. This has been adopted fairly recently by some cognitive scientists such as Varela, Thompson & Rosch (1991), and is also making an impact in certain areas of general psychology. In this view language (and thought) is not representational, not symbolic. It is an emergent property of vast, dynamic nets of relationships among systems of neurones, and among ecological systems which include but extend beyond the individual.

The implications of the third approach are difficult to grasp by anyone who has not made some study of parallel as opposed to linear problem processing (e.g. in computers), and connectionist models of cognitive science. Phenomena comparable at this level to the processing of natural language appear to be embedded in many processes found in the natural world. Indeed, research on the mathematics of complexity and chaos theory is really an attempt to find unifying principles which underlie such emergent phenomena. Further explanation would take the present paper too far afield, but readers are urged to seek out an understanding of the basic concepts.

In an emergent-system philosophy of language, the source and meaning of repeating entities will acquire different colouring than for those who see language purely as representational and symbolic.  This third position is close to my own, but much of what follows will initially cover more familiar perspectives.

4. Repetition as evidence: the scale of magnification

The introductory section raised the question of scale. I remarked that a human being could be regarded as a squishy heap of cellular structures, endlessly replicating themselves. Our sense that our friends (indeed, ourselves) are the same people from day to day stems from viewing them as unified structures rather than as an aggregate of short-lived cells. In fact the whole is more than an aggregate of the parts, and this is true through a whole series of stages from subatomic particle, to molecule, to cell, to body part, and so on.

The point is worth making because it holds as well in the conceptual structures of language as anywhere else in nature. It is critical to what counts as explanation in a model of language.

For example, the formulaic phrase "you know" occurs 975 times in one 130,000 word corpus which I studied. The repetitions of "you" or "know" taken individually (i.e. at a lower level of organization) could not reveal the significance of the combination. Potentially conflicting explanations of this phrasal repetition might be attempted respectively at the level of discourse context, or at a more embracing level of societal context, or in the general psychological context of language repetition, or at an ecological level encompassing the interrelationships of societal, cognitive and specific discourse functions. Generative, functional-systemic, discourse and sociolinguistic models have all settled on different levels of explanation for repetitive phenomena, and not surprisingly may differ on what is seen to be significantly repetitive at all.

 5. Past studies of repetition in linguistics 

Repetition has been almost ignored (under that label) in syntactic research, although grammatical rules themselves catalogue a type of repetition, and studies of anaphora etc. approach the more regular forms of lexical repetition. Less regular repetition, and that occurring above clause level has held little interest for sentential grammarians.  

The mention of repetition has become fairly common in various kinds of text and discourse studies. Some discussion of this (Halliday & Hasan 1976, Hoey 1991, and others) occurs in a later section of this paper.  

Text linguists will mention repetition in the context of cohesion, and what they discuss generally are words with some informational salience in an utterance which can be shown to have applied at some prior place in the utterance. This is said to create a "tie" which in some way assists communication. Whether this tie enters into the consciousness of the interlocutors is moot. They may also note the reuse of a syntactic structure when it occurs in close proximity (usually greater proximity than the word ties), and prosodic repetition relating to intonation, rhythm or rhyme.  

Such text analyses will not often speculate on the "tie" status of a highly grammaticalized item such as “the”, probably the most repeated word in English, on the grounds that it is syntactically motivated. This would be another judgement about applying explanatory scale. We could query the judgement by noting that every word in the language is an encoded unit that has been set aside for global repetition. At what point does a lexical repetition cease to be important as a cohesive tie?

 6. Uses of repetition in this study

 This paper tries to make no a priori assumptions about the repetitive status of particular linguistic elements. It begins by recognising that repetition is a complex phenomenon with a whole range of consequences depending upon whether it is global (as with codified lexical forms), contextual (within the topic reach of a particular discourse) or proximate (within the purview of something like short term memory).  

It also attempts to account for the paradox of inclusion by treating all linguistic elements which may be unitary at one level of operation, as potentially reducible to clusters of collocating indices - in effect the contributing properties and relationships which give that element its definition. Particular collocating indices from one linguistic element (such as a word itself, or a feature like [+animate], or a discourse sub-context like [+intimate]) may repeat or assume greater or lesser weight in the balance of relationships in a succeeding linguistic element. For example, the force of you in "you should come" will have a quite different weight from you in "you know".

 The broadest notion of repetition can be articulated by a very general linguistic entity. I have called this general entity a “repeating entity” (RE in shorthand). Repeating entities are codified forms which may subsume complex associative relationships. For example, I have just argued that lexical items at the micro level are concatenations of collocating indices of various weights.  

Hypotheses about the general properties of repeating entities promise to be very productive. More discussion of RE properties will be pursued later. For the moment an initial example will give a sense of the possibilities. Some iconic imagery will assist the description here. I propose that repeating entities "migrate" across cognitive space in language generation, and that they introduce collocation drag into the generative process. Collocation drag means that An RE will carry with it its whole evolutionary history of use and meaning for both speaker and listener. This multiplies the probability on reuse of eliciting other collocations (in whole or in part) which have occurred previously in the text, or in experience.  

Take the following, from a genuine conversation:

a) I want a trannie$ mum for Christmas

b) Want a cup of coffee mum

 ($ In the pre-iPod, pre-Internet world, trannie was slang for transistor radio)

These utterances occur twenty-three lines apart in a corpus. Both are pseudo-quotations, "recalled" by the mother forty years after the event. Neither has the simple meaning we might infer out of context. The first is in the context of an inquisitive boy who takes everything electronic to pieces and reinvents it. The second introduces a trick coffee pot which "talks back" to the user. The linguistic point is that [I] want X has "migrated" for reuse in the discourse, and carried with it connotations of electronic trickery.

 Collocation drag is a symptom of very interesting cognitive processes. A probably related phenomenon is that words or phrases and/or syntactic patterns which do not usually have high density in an idiolect may occur in clusters throughout a corpus when they are used. Various items seem to have characteristic repetition cluster patterns. I have found much evidence for this, both informally and in a large corpus of Australian speech. Some of this repetition stems from concentration on a particular current topic of conversation, as might be expected. Much however, seems to derive from more fundamental cognitive processes which I have called mnemonic resonance (May 1992: Unclever Talk: Mnemonic Resonance and God Knows What). It is as if an item in recent use has a “resonance” which is easily picked up and amplified for discourse re-use, even if it is not the ideal tool for a given context.

 At this point it may lend perspective if I return to tracking some of the more conventional analyses of repetition, particularly notions of cohesion in recent traditions of British textual studies. This material pays special attention to the (cognitively) more complex phenomenon of repetition expressed through various kinds of equivalent or related substitution.  

7. Repeating entities as cohesive devices

Since about the time of Halliday and Hasan's Cohesion in English (1976), cohesion has been a major area of study in British text linguistics. Earlier work concentrated on identifying a large number of cohesive categories. The following set is typical. Categories of lexical cohesion  (Hoey 1991: 8, from Hasan 1984:202):  

a) General

i)  repetition

leave, leaving, left

ii) synonymy 

leave, depart 

iii) antonymy

leave, arrive

iv) hyperonymy

travel, leave (including co-hyponyms, leave, arrive)  

v) meronymy

hand, finger (including  co-meronyms, finger, thumb)


b) Instantial       

i) equivalence

the sailor was their daddy

ii)  naming

the dog was called Toto

iii) semblance

the deck was like a pool 

Later work by Hasan in particular concluded that categories of cohesion were less significant than their collective interaction in a text (Winter 1974:79 from Hoey 1991: 16) :  

"Hasan's contribution to answering the question of the relationship between cohesion and coherence is twofold. Firstly she provides evidence that greater insight into text can be achieved if one abandons the classificatory view of cohesion in favour of an integrated approach, and secondly, she shows that it is the contribution of ties that is significant, not their occurrence in isolation. ..." (Hoey 1991:16)

 Winter (1974) chose to treat cohesion more inclusively than Hasan and others and, crucially, to relate it to repetition:

"[Winter's] interest is in how the grammar of sentences contributes to their interpretation in context. For him, therefore, it is much more important to recognise the common function of the variety of cohesive ties than to distinguish them, the common function being to repeat. As he notes:  

"A commonplace observation that everyone can make for themselves is that many clauses are repeated, either partially or (almost) entirely, in speech and in writing, the most obvious kind of repetition being the very common partially repeated structures of the clause. This repetition may however, be disguised by the grammatical form which it takes; that is, either by substitution, by deletion, or by a combination of both forms. (Winter 1979:101)."  

" ... what Winter counts as repetition is broader than, for example, Halliday & Hasan's category of reiteration. He uses the term repetition to stand for ellipsis (which he prefers to refer to as deletion), substitution (the label used by Quirk et al. 1972, to describe what Halliday and Hasan term as reference) and lexical repetition (broadly equivalent to Halliday and Hasan's reiteration)."

 These quoted models may be familiar to many readers. Their mention here is to identify a common ground which can be approached for different levels of explanation. Hoey's summary of Winter's views is also useful (Hoey 1991: 20):  

<> "If cohesion is to be interpreted correctly, it must be interpreted in the context of the sentences where it occurs.  

<> We are more likely to arrive at a satisfactory account of how cohesion works if we concentrate on the way repetition clusters in pairs of sentences.  

<> It is the common repeating function of much cohesion that is important, not the classificatory differences between types of cohesion.  

<> There is informational value to repetition, in that it provides a framework for interpreting what has changed.  

<> Relations between sentences established by repetition need not be adjacent and may be multiple."

The present study is congenial to many of Winter's and Hasan's (later) observations. However, it goes further even than Winter in treating repetition and cohesion as very general phenomena which become significant in a multiplicity of ways. There would be a difference in emphasis in interpreting some of Winter's points above. Whereas Winter, Hoey and Hasan are essentially interested in the surface properties of text, this writer is interested in the cognitive properties which coexist with those surface patterns.

 Whereas previous models have assumed repetition to occur solely in actual generated discourse, this study accepts the likelihood that what it terms repeating entities may remain latent. That is, REs can remain invisible on the surface of the text while playing an important part in the cognitive process. Ellipsis would be a non-controversial example, but it could be argued that even non-recoverable repeating entities are significant items of cohesion and coherence for both speaker and listener.  

Incoherence in spoken or written discourse is readily apparent to interlocutors, but the defining properties of overall coherence have remained much more intractable for linguists than markers of cohesion. There is a good reason for this. Traditional analyses have considered coherence a subjective quality (in the mind of the decoder), but cohesion as an objective (i.e. observable) property of generated text. As Hoey (1991: 11) says:

 "..coherence is only measurable in terms of a reader's assessment. This allows us to make a simple distinction..... We will assume that cohesion is a property of the text, and that coherence is a facet of the reader's evaluation of a text. In other words, cohesion is objective ...".

 The present study differs. I made the point earlier that it was possible to reject the subjective/objective dichotomy in favour of an emergent view of consciousness in which cognition and the apparent external world co-generate our realities, including language. For the problem on hand (coherence and cohesion) we can note that if cohesion is a significant property of text, then it is also a significant property of mind, though we may not yet have properly explained that significance. Equally, if coherence is a property of cognition (in ways not yet well understood), then it will be represented somehow in the relationship between text and cognition. One link in that relationship may well be latent repeating entities. The analysis to follow will offer some rationale for this position.

Firstly I will have to make a comment in passing on wider issues. This paper is part of a much larger study on cognitive processing which I christened the Generative Oscillation (GO) model (May 2004: a doctoral dissertation which I later discontinued, but nevertheless placed online in draft form). The is no space here to properly explain the reach of the GO model, but central to it is the hypothesis that language is processed not as a continuous stream, but rather in more or less discrete clusters of features. These clusters, called Local-Time Processing Constellations (LTPCs) in the model, are related to but not quite identical with intonation units in speech, and less closely to clauses in text. The “migrating” repeating entities discussed earlier would in fact traverse between LTPCs. From a cognitive perspective, coherence would have much to do with relationships between successive Local-Time Processing Constellations, and cohesion with the way in which migrating repeating entities reacted with different LTPCs.

Interlocutors recognise that clauses (one output of LTPCs) conform to strict rules internally, and are able partly on that basis to find them meaningful. The perceived unity of an extended discourse however is not essentially dependent upon syntactic rules (although their violation inter-clausally can cause considerable confusion). Rather, a text is considered to be coherent if, amongst other things, it conforms to topic reference in ways that may be idiosyncratic, but which remain comprehensible to and accepted by speaker and listener. It is common for much of a topic reference to remain latent, that is, non-explicit. Hence my earlier reference to the importance of latent repeating entities.

 One easy way to grasp this argument is the extreme example of the double entendre conversation. Think of a politician talking in code, or a woman fishing for the attention of a man where convention forbids her to proposition him directly. Both may be insistent, repetitive and logically progressive within the sub-text. In fact, this sort of thing also infuses normal discourse in a multiplicity of ways, so that an eavesdropping stranger will find much of what he hears only sparsely meaningful. The linguistic analysis which deals only with the surface text will be as poorly informed as the spy with his ear to the wall.

 There is a second kind of discourse unity which is non-topic dependent, although it draws heavily on topical reference, and that is cohesion. A rough definition of cohesion in the GO model environment would be an implicit recognition by either speaker or listener that certain linguistic elements in play have been met before in some incarnation, or are about to be met again. It is not required that these visitations have any propositional connection, and their individual importance may range from trivial to consuming interest. Thus while some repeating entities may do their work beyond conscious attention, cohesion is a quality that carries direct interpretive significance for each of the communicating parties.  

Cohesive nets of repeating entities can participate selectively in building coherence, but need not. It is possible in theory to have a text with a very high cohesive index but a very low coherence index, that is, a poor integration of topics. In plain language, such a text could contain many repeated propositions, but little logical argument to sustain them.

It may be a curious fact of human psychology that many people relate positively to a high cohesive index, but negatively to an excessively coherent presentation. This is my informal observation (although some colleagues have expressed skepticism). It may be that while coherence entails a certain logical coercion in interpretation for a listener, cohesion merely offers the warm inner glow of recognising tokens, and leaves the listener to interpret them as he will. Politicians are notorious for trading on this kind of preference. Choose your own favourite example. 

8. Relations among cohesive elements

 Although cohesive nets may lack a logical connectedness sometimes, those texts which are highly coherent are also likely to have closely integrated cohesive nets. Hasan has proposed two kinds of cohesive chains which are relevant to this question:

 "Hasan reports the results of research into the relationship between coherence and cohesion in children's writing, the level of coherence being measured in terms of reader response. She shows that there is no easy correlation between the number of cohesive ties and the degree of coherence awarded a text by readers. She concludes that a better explanation of the way cohesion contributes to the recognition of coherence lies (in part) in the fact that cohesive ties form chains that interact with each other; this interaction she terms cohesive harmony." 

"Hasan recognises two general classes of chain: the identity chain and the similarity chain. Both types of chain, but particularly the identity chain, override the careful sub-categorizations and distinctions that make up much of the earlier work on cohesion... An identity chain is made up of cohesive ties that all share the same referent(s), whether the ties in question are pronominals, reiterations, or instantial equivalents... Similarity chains are chains of ties where issues of identity cannot arise, for example, parallel processes or descriptions. ... if three occurrences arise of someone running away, there will be a similarity chain formed between the occurrences or ran away, irrespective of whether the same person did the running on each occasion." (Hasan 1984; and with Halliday 1985, quoted by Hoey 1991: 14) 

Hoey takes the linking notion further with his proposal for cohesive nets:  

".. network already does heavy duty in systemic linguistics in a quite different sense; the term that will be adopted here to describe both the complete set of bonded sentences and any sub-set of them is, therefore, the related one of net." (Hoey 1991:92)

 The general concepts employed by Hasan and Hoey here are very instructive in their patterned outline of how texts hang together. Once again however, the surface textual analysis seems to this writer to lack some crucial insight. We are, as it were, in a mountain lookout, with a great panorama of industry spread across the plain below, but we don't really know what it means. We don't understand the forces which give rise to what we are witnessing, and have no credible way to predict how it might change and develop. I will now try to take a one tentative step down the mountain to the plain, to the level of human cognition where it is all happening.

 9. Cohesion and discourse presupposition

 A brief reference was made earlier to latent repeating entities. These LREs may play a part in connotation, that is, in the sub-text of discourse. We mentioned the double entendre conversation as an example. A more formal linguistic expression of these ideas is found in work on discourse presupposition. The notion of the latent repeating entity as an organizing device has explanatory power in this context. I will borrow Hoey's concept of a cohesive net, but cast it more widely to embrace latent repeating entities. Consider a conversation on a familiar topic between two intimate friends, perhaps husband and wife:  


Decoded Meaning

+ Tomorrow

The meeting is on tomorrow. I'll be late home

- Are you sure?

Are you sure the meeting will be on?

+ George was late again

As you know, George is only late when he goes to divisional headquarters, and that is always followed by a  meeting.

- OK. I'll talk to Wendy.

I'll talk to Wendy to put off our dinner party.

Linguists would identify the material in italics as discourse presuppositions, or in “Are you sure?” as ellipsis. Formal grammars have an insoluble problem with discourse presuppositions since they are not manifest in generated strings and may be paraphrased in any number of ways. Are they therefore non-linguistic in spite of their controlling influence on the strings that are actually surface generated ?

 In the GO model a discourse presupposition is a latent repeating entity which was in the process of becoming a Local-Time Processing Constellation. This process would normally entail the accumulation of a critical mass of collocating indices (associations) until at a certain threshold an intonation unit was generated. Again, these are processes which we can’t explore here. The significant point to make for this analysis is that in a discourse presupposition, the generative process is aborted prior to surface expression. What triggers abortion here? Informally we know that an omission (say, ellipsis) is possible because the interlocutors can both supply the missing surface language. In cognitive processing terms, we could say that the threshold for surface generation has been raised by a special executive signal that substantive indices in the LTPC do not need to be communicated.

Is the cohesive net among fully generative LTPCs (i.e. ones which produce intonation units) usually more complete than the cohesive net amongst expressively truncated LTPCs (i.e. discourse presuppositions)? On the contrary, it seems to me that the latter net is so secure that elements are permitted to remain latent. Even the most explicit language contains discourse presuppositions, and therefore has a latent cohesive net.  

The ratio between latent and manifest cohesion is surely significant however. I hypothesise that the overall cohesive index is normally highest where the latency potential in cognitive processing is greatest. In plain language this would mean that where shared knowledge between participants is greatest, cohesion will also be greatest, and that unexpressed cohesive factors will be much more important than the surface markers of cohesion.

10. Conclusion

This paper has scarcely brushed the edges of an enormous topic. It has introduced the idea of repetition as a central organizing phenomenon in nature generally, and in linguistics in particular. The questions we ask about repetition will deeply influence the development of the linguistics discipline itself. The paper has argued that linguists need to be explicit at all times about the philosophical perspective from which they approach their subject. With a topic like repetition, that includes being explicit about the scale (magnification) at which explanations are sought. Thus, for example, the study of a pattern of repeating phrases will not necessarily have explanatory power for a pattern of repeating morphemes which are embedded in the phrases (the whole exceeds the meaning of the parts, while the parts may have some meaning unexploited in the whole).

Three major philosophical perspectives on linguistic analysis were identified. The first was the objective/positivist bias which has been expressed through behaviourist schools of thought, early structural linguistics, much descriptive and applied linguistics, discourse, text and conversational analysis, most cognitive science until quite recently, and so on. The second philosophical approach was a fresh defence of the value of subjective-rationalist insight, most notably championed by the Chomskian school of generative grammar (with strong sources in formal logic), but also favoured in much cognitive psychology and some branches of European philosophy. A third approach was said to be both very old, from the Madhyamika Buddhist tradition of mindfulness-awareness, and also very new in cognitive science. Historically in the West, the third approach drew on a refinement of the phenomenological philosophies of Husserl, Merleau-Ponty and others. This approach rejected mind-body dualism, and argued for an emergent cognition that was co-evolutionary with and inseparable from the wider world.

The paper argued that the three philosophical approaches often dealt with data that overlapped. To this extent, linguists working within one tradition could inform the research of those working within other traditions. However, the questions asked, and the levels of explanation sought, were often quite different.

Some examples of repetitive phenomena were used as a vehicle to compare analytic methods. The paper looked at fairly recent examples of British textual analysis, notably for the treatment of cohesion. The surface analysis of text by these British linguists showed up interesting patterns, and highlighted some important general properties of repetition in cohesive features. I argued however that the level of explanation offered by such surface analysis was unsatisfying. In particular, it failed to provide any insight into the cognitive processes which coexist with surface cohesive phenomena. The beginning of an alternative analysis based on emergent-system philosophy was hinted at. Since a coherent linguistic model for such an emergent-system philosophy is not yet in the public domain, this short paper could not develop extensive argument in this direction, but it is hoped that some researchers will be encouraged to enquire further.


References (these references obviously represent the vintage of the original article, but many are still worthwhile. I have added a useful recent book by Holland on complexity theory)

Bouton, Charles (1991) Neurolinguistics : Historical & Theoretical Perspectives NY:Plenum Press

Chafe, W (1970) Meaning and the Structure of Language. Chicago: U of Chicago Press

Chafe, W (1979) The flow of thought and the flow of language, Syntax & Semantics vol.12; (ed.) Talmy Givon. NY: Academic Press; pp 159-181

Chafe, W (ed. 1980a) The Pear Stories: Cognitive; Cultural & Linguistic Aspects of Narrative Production NJ:Ablex

Chafe, W (1980b) The deployment of consciousness in the production of a narrative in Chafe 1980a

Chomsky, N. (1965) Aspects of the theory of syntax. Cambridge, Mass: MIT Press

Chomsky, N. (1975) Reflections on Language. Glasgow: Collins

Givon, T (ed 1979a) Syntax & Semantics vol.12. NY: Academic Press

Givon T (1979b) From discourse to syntax : grammar as a processing strategy in Syntax & Semantics vol.12 (ed.) Talmy Givon NY: Academic Press

Gleik, J. (1987) Chaos. London: Cardinal

Halliday, M and R. Hasan (1976) Cohesion in English. London: Longman

Hasan, R. (1984) Coherence and Cohesive Harmony in J. Flood (ed.) Understanding Reading Comprehension, 181-219. Newark International Reading Association

Hoey, M (1983) On the Surface of Discourse London: Allen & Unwin

Hoey, M (1991a) Lexical Phrases & Language Teaching UK:OUP

Hoey, M (1991b) Patterns of Lexis in Text. Oxford: OUP

Holland, John H. (2014) Complexity: A Very Short Introduction. OUP. ebook: Amazon Digital Services Inc.

Jackendoff, R (1987) Consciousness and the Computational Mind Cambridge, Mass: MIT Press

Krashen, S (1982) Internal processing: the Monitor in M. Burt and S. Krashen (eds.) Language Two 46-72. New York: OUP

May, Thor (1992)  Unclever Talk: Mnemonic Resonance and God Knows What online at  and also at

May, Thor (1994)  Generative Oscillation -- A Cognitive Model for the Emergence of Language (discontinued PhD thesis from the University of Melbourne). online at  and also at

Pawley, A & Frances Syder (1983) Two puzzles for linguistic theory : nativelike fluency and nativelike selection in Language & Communication; (eds) J.Richards & R Schmidt UK: Longman; pp 191-228

Pawley, A (1985a) Lexicalization in Tannen D & J Alatis (eds) Languages & Linguistics : The Interdependence Of Theory; Data And Application 1985:98-120 Georgetown U. Round Table on Languages & Linguistics

Pawley, Andrew (1985b) On speech formulas and linguistic competence Lenguas Modernas 12:84-104; Universidad de Chile

Phillips, M (1985) Aspects of Text Structure: An Investigation of the Lexical Organization of Text. Amsterdam: North-Holland

Quirk, R, S. Greenbaum, G. Leech and J.Svartvik (1972) A Grammar of Contemporary English. London: Longman

Rietman, E (1988) Experiments in Artificial Neural Networks. Blue Ridge Summit, PA: TAB Books

Rosch E; Mervis C; Gray W; Johnson D; Boyes-Braem P (1976) Basic Objects in Natural Categories Cognitive Psychology; 1976: (8) 382-439

Rumelhart, D (1975) Notes on a schema for stories in D Bobros & A Collins (eds)  Representation & Understanding : Studies in Cognitive Science. NY: Academic Press

Rumelhart, D 1977 Understanding & summarizing brief stories in LaBorge D & Samuels S (eds)  Basic Processes in Reading : Perception & Comprehension.  NJ:Erlbaum

Varela J, E Thompson & E Rosch (1991) The Embodied Mind : Cognitive Science and Human Experience. Cambridge, Mass: MIT Press

Winter E.O. (1974) Replacement As A Function Of Repetition: A Study Of Some Of Its Principal Features In The Clause Relations Of Contemporary English, PhD thesis, University of London

Winter, E (1979) Replacement as a fundamental function of the sentence in context. Forum Linguisticum 4: 95-133


This paper has been extracted from a discontinued doctoral dissertation at the University of Melbourne in the mid 1990s. The entire dissertation content itself has already been put online: “Generative Oscillation - A Cognitive Model for the Emergence of Language” (  and also ). It seems worthwhile to place the chapter on discourse repetition into the public domain independently since the approach adopted has not, to my knowledge, received wide attention, yet seems to hold quite a lot of promise. Note that originally it was also a conference presentation, and published as part of a set of the University of Melbourne’s Working Papers in Linguistics 13 (1993): 79-90.


Professional bio: Thor May has a core professional interest in cognitive linguistics, at which he has rarely succeeded in making a living. He has also, perhaps fatally in a career sense, cultivated an interest in how things work – people, brains, systems, countries, machines, whatever… In the world of daily employment he has mostly taught English as a foreign language, a stimulating activity though rarely regarded as a profession by the world at large.

Thor’s eventually awarded PhD dissertation, Language Tangle, dealt with language teaching productivity. Language Tangle (2010) is aimed at professional educators and their institutional keepers, and accordingly adopts a generally more discursive style than the Aspects of Repetition in Discourse analysis. Thor’s first shot at a PhD dissertation was on Grammatical Agency in the 1980s, based on the generative syntax models of the time which he eventually rejected and withdrew from the candidature after publishing some professional papers. Also in cyberspace (representing even more lost years!) is yet another sprawling, unfinished PhD dissertation draft in cognitive linguistics from the university of Melbourne in the early 1990s, parts of which can be seen in the repository as The Generative Oscillation Model, Postsupposition and Pastiche Talk, this piece on Discourse Repetition and a couple of other papers.

Thor has been teaching English to non-native speakers, training teachers and lecturing linguistics, since 1976. This work has taken him to seven countries in Oceania and East Asia, mostly with tertiary students, but with a couple of detours to teach secondary students and young children. He has trained teachers in Australia, Fiji and South Korea. In an earlier life, prior to becoming a teacher, he had a decade of finding his way out of working class origins, through unskilled jobs in Australia, New Zealand and finally England (after backpacking across Asia in 1972).


academic repository: at  


Aspects Of Repetition in Discourse   © Thor May 1993-2015