6.3.1 Data and methods
One strength of the present study lies in its basis in naturally-occurring data. Other linguists take different approaches, shaped by their purposes and theoretical starting-points. As mentioned previously, Cook (2000) and Crystal (1998), despite real differences in their treatments of language play,1 share a concern with prepared and recited language play such as jokes and nursery rhymes, rather than spontaneous context-bound instances emerging through interaction. As such, their data is limited largely to published or well-known sources and, in Crystal‘s case, anecdotal evidence. In contrast, Tannen (1989) draws on recorded conversations between herself and groups of friends and acquaintances. Social characteristics of her participants are, however, limited in range and the type of interaction largely small, friendly private gatherings. Use of CANCODE to support his arguments is therefore one of the greatest strengths of Carter‘s (2004) work (Swann 2006: 9). As well as the benefits of size, CANCODE is organised by context type and interaction type, labelled where necessary in terms of the conversation‘s main activity. Unlike Tannen, Carter (2004b) is able to explore creative functions across social and generic contexts, involving various speakers and interaction types.2 As Swann (2006) notes, data in CANCODE is mainly limited to informal talk in non-institutional settings although Carter (2004: 163-4) includes workplace and business contexts from outside CANCODE.
The current study draws in a similar fashion on a corpus.3 In terms of interaction types within CorTxt, the purposes to which text messages are put are chiefly, according to sociological and linguistic research, informal, interpersonal, and typical of spoken
1
Crystal‘s is a populist account of the significance and prevalence of language play; Cook‘s a theoretical discussion situated in the wider context of play.
2 However, again in contrast to Tannen, although it is clear as Swann (2006) points out that interactions between both women and men are included, Carter gives no information about the gender balance in his sample, nor about other social factors such as class and ethnicity. It is not therefore possible to consider social, cultural and contextual factors that might affect the types of creativity people engage in and their interactional effects, and this could be seen as limiting the generalisability of Carter‘s claims.
3 Although no attempt is made to suggest that CorTxt and CANCODE are comparable in size, range or in
terms of the sophistication of their annotation: nor that the achievements of the studies based on them are comparable.
interaction between intimates.1 Although CorTxt is not divided by context or interaction type, it would be possible through analysis of text message content to identify main activities within exchanges and thus attempt to match uses of creativity to specific communicative functions and contexts. For the purposes of the current study, however, the corpus is explored as a whole and creative features matched where possible to functions and affordances of the medium.
Identifying and classifying creativity remains a contentious issue, as Carter observes. Given the impossibility of retrieving creative features automatically through surface formal features (Carter 2004b: 148),2 creativity in CANCODE is retrieved by systematically dividing the corpus into cells and by ‗reading‘ rather than frequency counts and concordance lines.3 Limitations include the impossibility of quantification, despite Carter‘s claim to have combined qualitative with quantitative methods. Although he suggests that creativity ‗leaps out at researchers from almost every transcript of the data‘ (Carter 2004b: 90), selected examples can only be, as he acknowledges, ‗taken on trust‘ and there are unavoidably few replicable figures.4 However, given the continuing relative lack of access to ‗intimate discourse‘, as highlighted by Cook (1994), the fact that such examples are identified at all is of great value and there is no inevitable need for studies of creativity to be quantitative. The great advantage of Carter‘s text-driven approach, in comparison to corpus-based methods, is the sense that Carter‘s investigations are led as much by data as by preconceived theory; and, in comparison to corpus-driven methods, that features he uncovers are not retrievable through quantitative methods.
Methods used to identify creativity in CorTxt are similarly qualitative, and examples of creativity accessed by reading. The fact that creative instances were primarily identified through formal characteristics can also be defended on the practical grounds that creativity
1
See Chapter 2: Research into Text Messaging. 2
My attempt to find creativity automatically from CorTxt through searching for metalanguage (literally, I mean), intensifiers (indeed, absolutely), vague language (chiefly, a bit) and even conjunctions (and) was largely unproductive.
3 An approach described in Chapter 3 of this thesis as a text-driven approach. 4
The same also applies to Tannen (1989). Tannen supports each observation she makes with an example from her own data, but whether it is one of many such examples or an isolated occurrence is difficult to ascertain. The nearest we get to an overview of Tannen‘s larger bank of data is her statement that ‗not all transcripts show a high percentage of repeated words but many do, and all show some‘. Other studies, such as Carter (2004: 79-80) and Crystal (2003: 413) reproduce Tannen‘s oft-quoted example of a discussion on ‗eating‘ in which phrases repeat, but the fact that one stretch of discourse fits the pattern can hardly be very convincing.
in texting is arguably less easily identifiable than in speech through sociocultural clues:1 responses such as laughter or groaning may occur verbally yet be omitted from text messages themselves. At the same time, the (limited) contextual knowledge of this researcher also serves to identify puns or shared jokes. As with other studies of creativity, identification and interpretation therefore relies on speculation, albeit cautious and informed by CorTxt and by previous research into everyday creativity.
6.3.3 Outline of the investigation
This investigation is structured according to categories of creative language identified by Carter (2004b). Creative patterns of various kinds, Carter (2004: 101-9) observes, potentially form and can be traced throughout everyday language, and choices motivating these patterns fall into two categories: pattern-forming structures such as repetition and
pattern re-forming choices as seen in puns. Although the two categories may combine, the
following investigation begins by exploring pattern-forming choices, before moving on to pattern re-forming choices. For each, their use and function in spoken interaction is explored first (Carter 2004b; Tannen 1989; North 2007), before turning to investigation of CorTxt.