Language has two primary functions: first, to communicate information from one person/group to another, and secondly, to form interpersonal relationships between individuals/groups, not necessarily in that order of importance. Spoken interaction may be ‘the fabric of friendship’ (Miller and Weinert Reference Miller and Weinert1998, Chafe Reference Chafe1994), but how far do speakers go to show closeness and solidarity to fellow conversationalists?

An investigation of approximately 200,000 words of conversational data from the Wellington Corpus of Spoken New Zealand English (Holmes, Vine & Johnson Reference Holmes, Vine and Johnson1998) shows that speakers may go as far as emulating each other's grammar. The usage patterns of cleft constructions were analysed in 80 informal conversations. The conversations are relatively similar in terms of participants, topic and environment, in that they are conducted in the participants’ own homes, without an interviewer present, or a given topic. In most excerpts, participants discuss day-to-day issues, such as gossip or plans for the day, and it is clear that they know each other well (they are typically living together).

Clefts such as it is just one day that I might be away for, and what he wants to do is to sell the house as soon as possible constitute a grammatical means for highlighting or advancing the most important element of a sentence. In spoken language, they often have a discourse management role (see Calude Reference Calude2008). Example (1) below illustrates a typical occurrence of clefts throughout one of the conversation excerpts analysed.Footnote ²

1. (1) WSC#DPC066: (Four speakers: DN, BT, CH, AL)

Each conversation was tagged for the total number of cleftsFootnote ³ uttered by each speaker. The data show significant tendencies of contagious behaviour. That is, either (almost) everyone uses clefts, or no one uses them, so that independently of personal style, conversationalists tend to mimic each other's behaviour with respect to this construction.

A correlation analysis (shown in figure 1) looking at the proportion of words that were clefts used by one speaker versus the proportion of clefts used by the other speakers in that conversation shows that one speaker's production of a cleft is positively correlated (0.229) with other speakers doing this as well, whether using the same cleft type or not. (For X = the total number of words, and Y = the total number of clefts used, the proportion is Y/X.)

Figure 1. Plot of the frequency of cleft usage per 1,000 words by a single speaker in a conversation (vertical axis) versus the average frequency of cleft usage per 1,000 words by other participants in the same conversation (horizontal axis). The linear trend is indicated.

The smoothed trend of the plot is indicated in figure 1. Pearson's correlation is 0.229, suggesting a positive association between an individual's behaviour and that of the rest of the group. A permutation test, performed by mixing up the speakers across conversations 1,000 times, shows that only 6 times was the correlation factor (i.e. 0.229) higher than that found for the original data. The results suggest that the observed pattern of cleft usage would be unlikely to occur by chance. This provides very strong evidence against the hypothesis that individuals’ cleft usage is independent of the usage of clefts by others in that conversation (p = 0.006). There are symmetries within the plot due to conversations involving only two speakers. These will not adversely affect the analysis presented.

Put another way, there is very strong evidence that there is a correlation between the proportion of clefts used in a conversation by everybody, and the proportion of clefts used by a particular speaker in that conversation. That is, the more clefts are uttered in a conversation, the more clefts a particular speaker taking part in it is likely to use.

In addition, the cleft constructions appear to be sprinkled throughout the conversation excerpts (as shown in the example included in (1)), rather than occurring sequentially in the interaction. However, a more detailed analysis (also perhaps of longer conversations) is called for in order to validate this trend.Footnote ⁴

The idea of contagious behaviour in conversation is a well-documented phenomenon. As observed in studies of nonlinguistic behaviour (Lakin & Chartrand Reference Lakin and Chartrand2003), humans mimic the movements, gestures and body language of others as a way of establishing rapport with them. Similarly, linguistically, there is a large body of work on Accommodation Theory (Coupland Reference Coupland, Verschueren, Ostma and Blommaert1995, Giles Reference Giles1973, Meyerhoff Reference Meyerhoff1998, Trudgill Reference Trudgill1986 inter alia) and Audience Design (Bell Reference Bell1984, Reference Bell, Eckert and Rickford2001) investigating how, when and why speakers emulate each other's behaviour in terms of phonological and lexical choices.

Claims have also been made about convergence at the syntactic level. However, convergence at this level is not all that simple to ascertain. For instance, a study by Schenkin (Reference Schenkin and Butterworth1980) claims syntactic convergence, while the data can in fact be explained by simple repetition, that is, speakers using the exact same wording, for example, A: Do you know what we are having for dinner tonight? B: I don't know what we are having for dinner tonight. Hence, this is not a case of true syntactic convergence. Priming experiments have, however, indeed shown evidence of syntactic convergence (Bock Reference Bock1986, Potter & Lombardi Reference Potter and Lombardi1998, Branigan, Pickering & Cleland Reference Branigan, Pickering and Cleland2000).

Garrod and colleagues assert that participants ‘align’ each other's semantic, syntactic and phonological usage, mimicking the same referring expressions, syntax, tone, intonation and stress (Garrod & Doherty Reference Garrod and Doherty1994, Garrod & Pickering Reference Garrod and Pickering2004). Thus, the process of ‘interactive alignment’ works across all different linguistic levels (Garrod & Pickering Reference Garrod and Pickering2004). It is this phenomenon of participants operating on common representations and reusing each other's material that makes conversation ‘easy’, according to Garrod & Doherty (Reference Garrod and Doherty1994).

Significantly, the corpus data analysed indicate that syntactic convergence appears also to be present in naturally occurring language (as opposed to elicited, experimental data). In other words, conversationalists go as far as mimicking individual constructions used by others. That is, not only do they use the same sound patterns, referring expressions and lexical items, but in fact, they use similar patterns of speech and schemas (i.e. constructions).

The contagious role of clefts in spontaneous, unplanned spoken conversation might be better understood within a larger study, where more variables can be taken into account (such as the position of the cleft within topic structure, the specific function of the cleft within the topic frame, the gender of the person uttering the cleft, and so on). However, the present analysis is indicative of the fact that, like words and body language, grammar also presents speakers with choices which reflect the communicative purposes that drive interaction.