ANGLIA RUSKIN UNIVERSITY A STUDY INTO THE USE OF LINGUISTIC STRUCTURES USED INTER-GENDER AND INTRA-GENDER IN THE TV SHOW ‘FRIENDS’ DAVID AYLIFFE A dissertation in partial fulfilment of the requirements of Anglia Ruskin University for the degree of Master of Arts in Applied Linguistics and TESOL

Submitted: July 2011

Anglia Ruskin University MA Applied Linguistics with TESOL

Acknowledgements It is with gratitude that I would like to thank Dr. Sebastian Rasinger and the rest of the teaching staff at Anglia Ruskin University. This project was supervised by the aforementioned Dr. Rasinger and his advice and support have been invaluable.

I am indebted to the fans of the TV show ‘Friends’ who faithfully and painstakingly transcribed the entire catalogue of episodes. Their accurate transcription considerably accelerated the project and gave me much more time on performing a detailed analysis of the results than would have otherwise been possible. Undoubtedly without these transcribed episodes the project would not have happened.

i

Anglia Ruskin University MA Applied Linguistics with TESOL

ANGLIA RUSKIN UNIVERSITY ABSTRACT FACULTY OF ARTS, LAW AND SOCIAL SCIENCES MASTER OF ARTS A STUDY INTO THE USE OF LINGUISTIC STRUCTURES USED INTER-GENDER AND INTRA-GENDER IN THE TV SHOW ‘FRIENDS’ By David Ayliffe July 2011

In this thesis I present a corpus-based analysis of the use of various linguistic structures and a comparison of their use in inter-gender and intra-gender conversation environments in the TV show ‘Friends’. There is a great body of research which indicates that female language is more polite than males and the overarching conceptual aim of this project was to validate or invalidate that hypothesis. The results of this study have revealed a number of interesting results; the most prominent of which is not that one genders linguistic style is different than the others but that both men and women align themselves closely depending on the company. In other words the styles both men and women employ differ greatly depending on whether or not they are talking to one of their own or inter-gender. There was no support that women use more ‘women’s language’ and the view that the stereotypes of women being linguistically restricted was not upheld, although it was shown that while men generally have a broader vocabulary and use it more creatively. Despite this, from the use of intensifiers to taboo words women were found to be a country mile ahead of masculine counterparts, while in other features, from empty adjectives to the number of questions asked, colours to hedges, both genders were found to operate on remarkably similar levels.

ii

Anglia Ruskin University MA Applied Linguistics with TESOL

TABLE OF CONTENTS 1.

INTRODUCING THIS STUDY........................................................................................... 1 1.1

THE RESEARCH QUESTION ............................................................................................ 1

1.2

DEFINING POLITENESS ................................................................................................ 2

1.2.1 What is polite language?.................................................................................... 3

2.

1.3

THE NEED FOR A CORPUS -BASED STUDY OF LANGUAGE AND GENDER IN US TV .......................... 4

1.4

THE APPROACH TAKEN TO THIS STUDY............................................................................. 7

BACKGROUND INVESTIGATIONS RELEVANT TO THIS STUDY .......................................... 8 2.1

DISCOURSE-RELATED LITERATURE .................................................................................. 8 Birmingham Model ................................................................................................... 9 Systemic Functional Linguistics ................................................................................ 10

2.1.1 Speech transcription ........................................................................................ 11 2.2

SEX AND GENDER .................................................................................................... 14

2.2.1 Politeness........................................................................................................ 17 2.3 3.

CORPUS RELATED LITERATURE..................................................................................... 22

BUILDING AND USING THE CORPUS ............................................................................ 31 3.1

CREATING THE CORPUS ............................................................................................. 31

3.1.1 The database................................................................................................... 33 3.1.2 ‘Tidying’ the data............................................................................................. 34 3.1.3 Tagging the data ............................................................................................. 36 3.1.4 Determining and assigning gender.................................................................... 37 3.1.5 The integrity of the data................................................................................... 38

4.

3.2

OBTAINING THE RESULTS ........................................................................................... 40

3.3

A SUMMARY OF THE DATA AND WHAT IT LOOKS LIKE ......................................................... 44

AN ANALYSIS AND DISCUSSION OF THE RESULTS ........................................................ 47 4.1

HEDGES ................................................................................................................ 47

4.2

TAG QUESTIONS ..................................................................................................... 52

4.3

ASKING QUESTIONS ................................................................................................. 55

4.4

TABOO LANGUAGE .................................................................................................. 57

4.5

EMPTY ADJECTIVES .................................................................................................. 61

4.6

INTENSIFIERS.......................................................................................................... 63

4.7

V OCABULARY ......................................................................................................... 65

4.8

BACK CHANNELLING ................................................................................................. 70 iii

Anglia Ruskin University 5.

MA Applied Linguistics with TESOL CONCLUSIONS............................................................................................................ 71

6.

LIMITATIONS OF THE STUDY....................................................................................... 73 6.1

SEMANTIC ANNOTATION........................................................................................... 74

6.2

CONNECT BY PRIOR ................................................................................................. 75

7.

RECOMMENDATIONS FOR FURTHER STUDY ................................................................ 76

8.

BIBLIOGRAPHY........................................................................................................... 77

9.

APPENDIX .................................................................................................................... 1 9.1

QUERY FOR RETRIEVING TAG QUESTIONS ......................................................................... 1

9.2

QUERY FOR RETRIEVING THE BREAKDOWN OF TAG QUESTIONS............................................... 1

9.3

QUERY FOR RETRIEVING THE NUMBER OF QUESTIONS ASKED ................................................ 2

9.4

QUERY FOR RETRIEVING THE STATISTICS FOR HEDGES.......................................................... 2

9.5

QUERY FOR RETRIEVING THE MOST FREQUENT ADJECTIVES.................................................... 2

9.6

QUERY FOR RETRIEVING THE MOST FREQUENT NOUNS ......................................................... 3

9.7

QUERY FOR RETRIEVING THE MOST FREQUENT VERBS .......................................................... 3

9.8

QUERY FOR RETRIEVING THE AVERAGE UTTERANCE LENGTH .................................................. 3

9.9

QUERY FOR RETRIEVING ALL SINGLE SEX WORDS ................................................................ 3 QUERY FOR RETRIEVING THE ............................................................................................ 4

9.10

DIACHRONIC USE OF ‘REALLY’, ‘VERY’, ’SO’ ....................................................................... 4

Frequency Counts -Simple ......................................................................................... 4 Frequency Counts - Complex ..................................................................................... 4 9.11

QUERY FOR RETRIEVING THE NUMBER OF EMPTY ADJECTIVES ................................................ 5

9.12

QUERY FOR RETRIEVING THE NUMBER OF PRONOUN REFERENCES .......................................... 5

9.13

QUERY FOR RETRIEVING THE NUMBER OF TABOO WORDS ..................................................... 6 General Counts by gender ......................................................................................... 6 General Counts by actor............................................................................................ 6 Counts per actor per scene ........................................................................................ 7 Counts per individual taboo word .............................................................................. 7

9.14

QUERY FOR RETRIEVING THE COLOURS USED BY THE GENDERS ............................................... 8

9.15

QUERY FOR RETRIEVING THE BACK CHANNELLING VOCABULARY.............................................. 9

9.16

THE CORPUS ............................................................................................................ 9

10.

COPYRIGHT ............................................................................................................ 10

iv

Anglia Ruskin University MA Applied Linguistics with TESOL

TABLE OF FIGURES Figure 1: Transcribing a fit of crying (Hepburn, 2004, p.252) ................................................ 12 Figure 2: Diachronic distribution of non-standard was among men and women in York (from Tagliamonte, p.182) X axis represents generation. ............................................................... 15 Figure 3: Wolfson's bulge model (1998)............................................................................... 17 Figure 4: The BNC Spoken Corpus regions (Crowdy, 1993, p.260) .......................................... 23 Figure 5: BNC Spoken Corpus ‘Conversation Log’ (Crowdy, 1993, p.260) ............................... 24 Figure 6: The TEI markup of a punchine (own material) ........................................................ 26 Figure 7: Phrase Detectives (http://www.phrasedetectives.org) ........................................... 29 Figure 8: reCAPTCHA (from Google, 2011) ........................................................................... 29 Figure 9: HTML file as downloaded. (Excerpt from 0102.html) ............................................. 32 Figure 10: Text file sans HTML mark-up (Excerpt from 0102.txt) ........................................... 32 Figure 11: Utterances before being loaded into the database ............................................... 32 Figure 12: Tidy data (not genuine utterances)...................................................................... 34 Figure 13: User interface for the BNC .................................................................................. 43 Figure 14: Custom BNC User interface at http://corpus.byu.edu/bnc/................................... 43 Figure 15: Links between data ............................................................................................ 46 Figure 16: Use of Hedges #1 ............................................................................................... 51 Figure 17: Use of Hedges #2 ............................................................................................... 51 Figure 18: Tag Questions .................................................................................................... 53 Figure 19: Tag Question Details........................................................................................... 54 Figure 20: Asking questions ................................................................................................ 56 Figure 21: Taboo language per actor 1 ................................................................................ 59 Figure 22: Taboo language per actor 2 ................................................................................ 59 Figure 23: Empty Adjectives................................................................................................ 61 Figure 24 Detailed view of negligible 'empty' adjectives ....................................................... 62 Figure 25: Diachronic results use of main elements over a 10 year period. ............................ 64 Figure 26: Vocabulary Breadth. ........................................................................................... 66 Figure 27: Mock query to interrogate both grammatical and semantic representations. ........ 74

v

Anglia Ruskin University MA Applied Linguistics with TESOL

TABLE OF TABLES Table 1: 10 most vocal actors.............................................................................................. 38 Table 2: Data counts .......................................................................................................... 39 Table 3: The first 15 rows of the data .................................................................................. 44 Table 4: Scene Data............................................................................................................ 45 Table 5: Normalising the data. Statistics of “oh” hedge........................................................ 47 Table 6: Corpus Details....................................................................................................... 47 Table 7: Hedges Results...................................................................................................... 48 Table 8: Tag Questions Results............................................................................................ 52 Table 9: Questions ............................................................................................................. 55 Table 10: Taboo words Results ........................................................................................... 57 Table 11: Who uses taboo language? ................................................................................. 57 Table 12: When is taboo language used? ............................................................................. 58 Table 13: Collocates to the left of 'boobs'............................................................................ 60 Table 14: Empty Adjectives Results ..................................................................................... 61 Table 15: Intensifiers Results .............................................................................................. 63 Table 16: Top 10 three-word clusters .................................................................................. 65 Table 17: Vocabulary Results .............................................................................................. 66 Table 18: Unique words per actor (ordered by ‘innovative’ score)......................................... 67 Table 19: Collocates with colours ........................................................................................ 68 Table 20: Gender specific vocabulary (not a definitive list) ................................................... 68 Table 21: Not back channelling ........................................................................................... 70 Table 22: Rachel back channels (scene #24)......................................................................... 70

Word count (excluding all tables, headings and quotes >= 50): 20,809 vi

Anglia Ruskin University MA Applied Linguistics with TESOL

1. INTRODUCING THIS STUDY 1.1

THE RESEARCH QUESTION

This project aims to investigate the gender differences in language use in the TV show “Friends” with an emphasis on the politeness of the language. The transcripts of the episodes will form the corpus around which the project hinges, and the show’s language will be analysed to identify how language both intra-group and intra-group.

The corpus is too large to facilitate detailed conversational discourse analysis at anything but a minute level. However, the use of politeness structures can be analysed from an abstract level. Such politeness structures include, but are not limited to, the use of hedging devices, tag questions, taboo language, the use of ‘empty’ adjectives and an analysis of so-called ‘intensifiers’.

1

Anglia Ruskin University MA Applied Linguistics with TESOL

1.2

DEFINING POLITENESS

Abundant definitions exist of what constitutes politeness, and only three of the most pertinent will be examined here. Lakoff (2004 p.64) states "politeness is developed by societies in order to reduce friction in personal interaction". Her hypothesis that “little girls are *indeed+ taught to talk like little ladies” (p57) ignited the debate between the pragmatic use of language between the genders. Setting the stage for a wealth of further research. Goffman stated that politeness was "the positive social value a person effectively claims for himself by the line others assume he was taken during a particular contact" (1955, p.213). This gave rise to Brown and Levinson’s (1978) ‘politeness theory’ and their definition of FTA’s – face threatening acts has received a great deal of attention due to their concept of both positive and negative face. Negative face is loosely translated at the inhibition of freedom of actions, a restriction on what one can do. Conversely positive face is the desire to be liked, admired, to possess a positive self-image and that crucially this self-image is appreciated and approved of.

Brown and Levinson’s politeness theory has been criticized for assuming that politeness can be expressed in a single act, and Mills (2003) has argued that it is not utterances which are polite or impolite but the context, and the communities of practices which arbitrate over whether something is offensive or not. Politeness, Mills states, needs to be analysed at a discourse level and not a sentence or phrase level. Leech’s view seems to support Mills when he proposed six politeness maxims (1983) which he used to define politeness as forms of behaviour, approaches which can be used to minimize conflict and ensure harmony among conversational participants.

2

Anglia Ruskin University MA Applied Linguistics with TESOL

1.2.1

WHAT IS POLITE LANGUAGE ?

Clearly some language is considered to be more polite than others. Cutting (2002) tells us that Example 1 (below) is more polite than Example 2 due in part to the use of negatives; their use avoids imposing on the hearer and anticipates rejection which helps to make the offer easier to refuse by the hearer. Example two is more direct and hence is considered to be less polite. Polite language is considered to be orthographically longer than impolite language and indeed this was the sole metric Gibson (2009) used in her study 1) I couldn’t borrow $30, could I, if you don’t need it right now? 2) Could I borrow $30? (Cutting, 2002, p.47) “Shall we stop for a drink?” is clearly a request, and linguists (Swan, 2005) would allow us to believe that ‘shall’ is a polite form. However is it so polite that the actual meaning of the utterance is hidden? Did the speaker not want to say “I want to stop for a drink” but the context disallows them? Herein lies the conundrum, the problem of distinguishing language (semantics) from language use (pragmatics) and the search for meaning is a complicated one. In pragmatics, “the meaning is relative to a speaker of user of the language, whereas meaning in semantics is defined purely as a property of expressions in a given language in abstraction form particular situations, speakers of hearers” (Leech, 1983, p6). Therefore if an utterance and situation are inextricably bound up with one another, then the context of the situation is indispensable for the understanding of the words (Malinowski, 1923 cited in Sunderland, 2006). In essence many utterances do or can do, many things at once. Linguists (Holmes, 1983, 1986; Lakoff, 1973; G. Lakoff, 1973) almost routinely butt heads with trying to define the bounds of linguistic structures. Does the use of ‘sorry’ indicate an apology?

The transcript (CNN, 1998) from U.S. President Clinton’s ‘apology’ for his

indiscretions showed that he managed a 4 minute address to the nation to express sorrow for his misdemeanours without ever using the word ‘sorry’. Equally the receptionists’ favourite ‘I’m sorry but he’s in a meeting at the moment’ lacks any sincerity and serves its place as an expected pre-fabricated chunk of language. The pragmatic use of language, (as well as other features such as intonation) can make a semantically impolite sentence polite or the adverse equally true. At a meta-linguistic level politeness is also subjective; subjective by gender, and/or by social occasion - clearly more politeness is required when speaking with one’s peers than to a three year old child, and in this context, it is not the use of discourse markers which makes impoliteness noticeable but the lack of them. 3

Anglia Ruskin University MA Applied Linguistics with TESOL

1.3

THE NEED FOR A CORPUS -BASED STUDY OF LANGUAGE AND GENDER IN US TV

“*The+ non-spontaneous dialogue of a film script tells us little of the way people really talk, but it often represents an ideal to which speakers aspire. It allows us to see what expectations different speech communities have, and what patterns are deemed appropriate for each sex, whether or not particular individuals respect the stereotype, they are familiar with them as competent users of the language, and the ways in which people flout the conventions provide rich material for analysis" (Livia, 2004, p.145, emphasis added).

The prominence of gender and gender-related acts are an effervescent characteristic of today’s society.

The whole advertising industry is built around the premise of women

behaving like ladies while big, burly aggressive men dominate the environment. With Britons now watching more than 30 hours of television a week (Sweney, 2010) the effects of such stereotypes on the impressionable youth are wide-reaching and important. This quote from Livia succinctly highlights the need for this study, the portrayal of genders through the media is an eternal hot potato, an area rife with strong emotions and divided opinions. Many studies have concluded that the genders exemplify their stereotype, representing women as submissive and men as dominant while men rarely exhibit effeminate traits (see Casey et. al, 2008). Such preliminary findings are important given the reach and intrusion into the lounge of these TV shows. Stateside, the episode of Friends immediately following the U.S. Super Bowl attracted in excess of fifty million viewers (Ginsburg, 2004) and the final episode was the most watched [American] TV show of the [2000-2009] decade (Chicago Tribune, 2009).

From Lakoff’s (1973) intuitive assumptions about gender more rounded empirical studies have been conducted. As indicated these have tended to focus on specific areas such as gender prominence in advertisements (Allan & Coltrane, 1996) gender in music videos (SommersFlanagan et. al., 1993), gender stereotypes in children’s television (Browne, 1998), semantic fields of vocabulary in off-the-shelf women’s magazines (del-Teso-Craviotto, 2006), occupational roles (Glascock, 2001) and vocabulary breadth in TV shows (MacFadden et. al., 2009). More works have also looked at linguistically discrete areas such as the use of specific intensifiers (Tagliamonte & Roberts, 2005), swear words (Gao, 2008) or key word clusters (Demmen, 2009). Almost all of these studies (Demmen’s study of Shakespeare the exception) are American-centric and in comparison relatively few studies have been approached from a 4

Anglia Ruskin University MA Applied Linguistics with TESOL British-centric view (for additional exceptions see Furnham, 1993 & Jaworski et. al., 2003). To date, and to my knowledge no other study has been conducted to look specifically into the politeness of language on a TV series; although regarding politeness, Gibson (2009) interestingly analyzed politeness in the fast-food industry.

Importantly it is gender and not sex which is at the centre of this study and it is inter-group and intra-group variation which is being studied – that is the difference in mixed gender and same gender conversation. Intra-group comparisons are useful because they serve to highlight the differences acting almost as a benchmark against which comparisons can be made.

Cameron (2005) has argued for a move away from the “binary difference” (p.490) view of the genders detailing how the construction of gender identity between men and women today has never been more similar. Her references are to social interactions in naturally occurring situations, variables which are void in the glitz of Hollywood’s studios. Therefore this study is not, nor will attempt to be an analysis of language as used in everyday conversations, and few comparisons will be drawn between the two. As Livia has cited, the project hinges around the portrayal of stereotypes and the characters’ alignment with said stereotypes.

Lauzen and Dozier’s (1999) study of the 1995-96 TV season – one year after Friends first aired, reported that women made up just under a third of the behind-the-scenes production team. Just 1 in 5 writers (22%) were female and while this improved marginally for directors and producers (28% and 36% respectively) they were still heavily outnumbered. Female executive producers accounted for less than 20% of the total. In contrast, in Friends just 12 of the 42 writers were female (29%). Two of the 12 (17%) producers were female while just 4 out of the 18 (22%) directors were female responsible for directing only 19 of the 238 episodes (8%).

The Sapir-Whorf hypothesis theorizes that what we do and how we think are influenced by language (Kay & Kempton, 1984). If this is true in its strongest sense, it affords the sinister possibility of a culture controlled by the media. Lippi-Green (1997) studied the language of characters in Disney films and found that heroes and lead characters used received pronunciation (e.g. Robin Hood spoke with a California accent) while foreign accents, often Russian or Arabic, were assigned to those of criminals, villains, servants and country folk amongst others (e.g. King Louis the orangutan in ‘The Jungle Story’ spoke with an African American accent). 5

Anglia Ruskin University MA Applied Linguistics with TESOL Her stance that “beliefs about the way language should be used are passed down and protected in much the same way that religious beliefs are passed along and cherished” (p.15) is important and relevant to this study, it is after all the script writers (along with the odd adlib) who are influential in the language which is broadcast.

As has been identified above (1.2) the problem with defining politeness is a tricky one and any study which attempts it will be flawed in some way. The extent to which the genders “respect their stereotype” (Livia op. cit., p.145) and how or indeed if they push the margins of expected norms will be interesting to discover and this study will analyse the most documented traits of politeness providing indicative results regarding the use of politeness structures.

6

Anglia Ruskin University MA Applied Linguistics with TESOL

1.4

THE APPROACH TAKEN TO THIS STUDY

This approach is largely quantitative. Given the quantity of the data comparatively little of it can be analysed at an utterance level. Linguists (Tannen, 1990, Holmes, 1995, Coates, 2004 amongst others) have demonstrated how men and women use language differently although quantifying this data remains an immense task given the syntactic variations any polite request can take and the influence the pragmatic meaning can have on an utterance.

An utterance is defined here as being synonymous with what many (Schegloff, 1968, Schegloff & Sacks, 1973, Crystal, 2009) have defined a turn. A turn it is essentially a statements by an actor. Using the Friends data, utterances are always scripted and the ambiguity of what constitutes a turn is largely a mute point as each utterance is clearly marked in the transcript. In this study the term utterance will be used instead of turn.

Traditionally, sociolinguists have approached the issue of linguistic and prosodic differences from a mostly non-computational perspective using relatively small and very focused data sets. Separately, where corpora have been used it has typically focused on a lexical view of language, identifying chunks or the most frequent vocabulary (e.g. Cheng, 2004 or MacFadden et. al., 2006). This study is a hybrid of both approaches but with a heavy bias towards the latter. While frameworks exist for annotating stretches of speech the overhead is huge and out of the bounds of this study. Therefore many of the results will be based on a very topdown methodology looking at vocabulary and phrases in isolation largely independent of the surrounding vocabulary or context.

This study is relevant in so far as it can show to what extent the writers’ preconceptions about gender influence their writing for each character. In television, I hypothesise that a stark ‘binary difference’ still exists where taboo topics are avoided by virtue of advertisers evidently (Fisher et. al., 2007) calling the shots.

7

Anglia Ruskin University MA Applied Linguistics with TESOL

2. BACKGROUND INVESTIGATIONS RELEVANT TO THIS STUDY The literature review will be divided into three distinct sections each dealing with a discrete item of the project. 1. The nature of discourse including issues surrounding the transcription of spoken data. 2. The differences in politeness between the genders. 3. Corpus design

2.1

DISCOURSE-RELATED LITERATURE

Detailed discourse analysis cannot be satisfactorily carried out without some knowledge of the situation or context in which the discourse has taken place. Views differ on which contextual features are important. Hymes (1974) for instance identifies almost a dozen variables as being necessary to adequately describe a situation. Hymes’ communicative competence is a key term which refers to the knowledge of social appropriacy regarding language use. Halliday (1985b) claims that our ability to put together well-formed spontaneous utterances effortless in a casual context is a trait of every native speaker, but this view is at odds with that of Chomsky (1965) who argues that all language is “degenerate” from the basic rules of grammar.

Discourse markers and their use are at the heart of this project so it is only apt to define what actually constitutes one. Discourse markers are “a class of expressions, each of which signals how the speaker intends the basic message that follows to relate to the prior discourse” (Fraser, 1990, p.387). Fraser’s more recent work (1999) builds upon the function of discourse markers, such that discourse markers provide commentary on a relationship between two utterances, a specification which is not conceptual, but procedural providing information on the interpretation of messages. Halliday & Hasan (1976) echo this approach forking slightly to suggest that their use helps to create a cohesion between utterances and ideas. Typically, discourse markers add little meaning and identifying them can be done by removing them from the utterance and validating that the meaning remains the same.

A handful of models exist which provide a framework for the deconstruction of dialogue, be they TV sitcoms or spontaneous conversation. These models are important to this study because they serve to underline the importance of treating a conversation as a series of 8

Anglia Ruskin University MA Applied Linguistics with TESOL interlocked turns rather than just an utterance in isolation.

One of these models the

Birmingham model.

Birmingham Model Sinclair and Coulthard’s (1975) approach to classroom discourse is just one of many which allow for the application of a framework to spoken discourse, and being a popular one it has given us a structure which has inspired others. They characterised an exchange as a series of moves from: Initiation  Response  Follow up. For example:

Initiation:

Speaker A:

What time is it?

Response:

Speaker B:

6:30

Follow Up:

Speaker A:

Thanks

They note how such exchanges are not fixed and culturally the sequence may deviate as illustrated here: I

A: Oiga, pot favor, ~ q uhio ra es?

(What time is it?)

R

B: Las cinco y media.

(5:30)

F

A: Gracias.

(Thanks)

?

B: De nada.

(Not at all)

With a little work the formula was modified to accommodate this redefining the structure as: I (R/I) R (Fn)

(Coulthard and Montgomery, 1981, p.112)

Within each move an act can be defined detailing the function or the purpose of the move, and Francis & Hunston (2002) detail 32 acts which can be applied to any one move. This is important because while a move may be a ‘follow up’ the meaning can vary depending on the context: A: What time is it? B: 6:30 A: Thanks. (stranger to stranger), or A: Good! Clever girl! (parent/teacher to child), or A: No it isn't, and you know it isn't; you're late again! (boss to late subordinate)

9

Anglia Ruskin University MA Applied Linguistics with TESOL While the I/R/F model remains intact this illustrates that we can only be “absolutely sure of the function of the initiating utterance when it is contextualised with the response it gets" (McCarthy, 2000, p.120). Such a model is convenient in this rigid two party setting or similar environments such as patient  doctor but it fails to stand up against casual conversation and all of the nuances it brings with it.

Similar models have been developed to try to

accommodate this.

Systemic Functional Linguistics Halliday’s Systemic Functional Linguistics (SFL) is another approach to the analysis of language. His view is that discourse analysis which is not based on grammar is nothing more than a “running commentary on the text” (Halliday, 1985: xvi), and the aim of SFL is simply to be able to make it possible to say sensible and useful things about texts. SFL has four main theoretical claims about language: 1. That language use is functional. 2. That its function is to make meanings. 3. That these meanings are influenced by the social and cultural context in which they are exchanged. 4. That the process of using language is a semiotic process, a process of making meanings by choosing. The functional approach SFL takes to language is two-fold. Firstly, it allows us to see how language is organised and secondly it allows us to make meanings and study how language is used. Halliday proposes that language be inspected from three different points of view 

Ideational meaning (field): i.e. The real world meaning



Interpersonal meaning (tenor): i.e. The relationship between speaker and hearer (or writer and reader)



Textual meaning (mode): i.e. The text is organized as a piece of writing or speech.

These three variables are important claims Halliday because they are the three kinds of meanings language is structured to make. Practically, the use is methodical to apply and has been used primarily in the examination of short texts, typically advertisements (both spoken and written, for example, see Simpson (2001)) the pitfall of it comes in its scalability, and much like Sinclair and Coulthard’s (1975) model the overhead needed to adequately analyse text becomes unreasonable. Hence although frameworks exist their application in corpora is restricted.

10

Anglia Ruskin University MA Applied Linguistics with TESOL

2.1.1

SPEECH TRANSCRIPTION

Labov (2006) believed that the more attention a speaker is forced to attend to their speech the more the style and register of their speech would be affected. With regard to this study, given the environment (TV studio) upon which the recording is made one must assume a high regard to the management of one’s speech and an accuracy to the script. The focus therefore moves to that of transcription. The transcription of the spoken word is key for a number of reasons: i.

The data derived from the transcriptions will be the basis for the results.

ii.

Transcription is a selective and subjective process.

iii.

The level of detail in a transcript will determine, to a certain extent, the acuteness of the results and the ability of those results to be reproduced by othe rs.

“A transcript should not have too much information” Ochs remarks in her seminal paper on transcribing the spoken form (1979, p.44) one which does is cumbersome, too difficult to reliably assess and can leave the researcher with the task of “sorting out the forest from the trees” (op. cit. p.56). Cumbersome notations such as ‘she leaves the room in disgust’ are subjective and imply the observer’s biased (Bucholtz, 2000, DuBois, 1991) or political (Green, et. al., 1997, p.172) opinion of the actor/actress. Such “cooking” (Trappes-Lomax, 2004, p.141) of the data is undesirable when the objective is to be neutral and analyse the data dispassionately without bringing any preconception to the task nor mentally filling in the blanks or jumping to conclusions. Illegibility can also be a problem; the non-verbal context changes an almost finite number of times in a thirty minute broadcast and Ochs (op. cit.) documented a paragraph of tugs, grimaces, frowns, pushes, turns from a three year old upon the utterance of ‘yeah, I’m gonna make a car’. This is also illustrated here: 1 Caller 2: 2 3 4 CPO: 5 6. 7 8 9 10 Caller 2: 11 12 13 14 Caller 1: 15 CPO: 16 Caller 1:

Um:: (0.6) Yeah I’m ok(h)ay. Y’okay ab’t- al:right then. (0.1) HHH so::=um:: (0.1) okay.=so Kathryn was just sayin abou::t (0.2) [ye know th-] [ AHH HH ]Hk iuHHhh uhh (.) I ca(h)n t ta(hh)lk. (1.2) Hello:? Hello::? I’m sorry she’s just li:ke >broke out in 11

Anglia Ruskin University 17 18 CPO:

MA Applied Linguistics with TESOL tears< she can’t spea:k. Ri:::ght.

Figure 1: Transcribing a fit of crying (Hepburn, 2004, p.252)

Standard orthography also forces a literal interpretation of utterances that otherwise may be simple objects of “phonological manipulations” (Ochs, op. cit. p.45). This morphological approach assumes that language is being used for information and not for sound play, and it may be that the researcher is intent on examining the language used, however the language choice may be influences by social factors. Such is the genre of the corpus in this study that assimilations of speech and alliterations are functions used to exhibit humour; a corpus void of such phonological transcriptions makes such analysis [of this area of research] virtually impossible, a point which extends to any corpus void of this annotation. Form and function have to be separated to understand what is happening in discourse. There is no 1-1 relationship between what is being said and the communicative function We are required to look at not just the grammatical form but also the context in order to understand the function of the language. Also a return to the floor by the speaker does not necessarily denote a response and that response may or may not be verbal. An exchange has to be initiated and a response to a previous utterance is indicated by a great number of non-verbal clues - eye contact, gestures, etc.. Without these non-verbal clues there exists an ambiguity as to whether the latest utterance is a semantically related adjacent utterance to the one it follows or if it is ‘off-topic’, an issue which is exemplified with more than one speaker.

While we (generally) express our intentions verbally extended annotation would be suited to dialogue where the communicative load of the spoken word is low relative to the context of the situation. Several agree (Green et. al., 1997) that there are two strands of transcription: 1)

interpretive and 2)representational. Although the distinction between the two becomes

difficult as it is applied. Roberts (1997) alludes that, transcription is still a subjective process and one where transcribers “call up the social roles and relations constituted in language and rely on their own social evaluations of speech in deciding how to write it…. transcribers bring their own language ideology to the task” (Roberts, 1997, pp.167-168). Labov and Fanshel (1977) add that currently it is impossible to “provide a context-free set of interpretations of prosodic clues” (p.46), and many would agree that this is still a significant issue thirty years after they made this statement.

12

Anglia Ruskin University MA Applied Linguistics with TESOL There are three questions (Crowdy, 1995) which need to be asked when designing the criteria for any transcription scheme: 1. Who is the transcription for? 2. How will it be used? 3. What are the important features? Much has been written about the importance of intonation (Coultard, 1995), and loosely defined, intonation, stress and rhythm are the key elements which compose prosody. Itself being the ability to express oneself using linguistic elements aside of grammar or vocabulary. Effects such as sarcasm and irony are readily expressed prosaically and their use in humour has a long history. The intonation and stress pattern of an utterance heavily influence the reply which is expected and recognizing intonational features and stress is a relatively straightforward process; however the issue of transcribing such elements is far from trivial. Jefferson whose transcription system has been widely adopted allows for ‘continuing intonation’, a ‘stopping fall’ and three types of stress. Coultard (op. cit.) laments that a recording of the utterances accompanying the transcription is mandatory as (again) the application of these features is subjective and that there is little consensus on when and when not to apply certain markers. While this is a valid point it must be noted that transcription is done for the purpose of analysis and if it is being analysed for some prosodic purpose then any human-derived level of transcription is beneficial, while features are subjective presumably it is being analysed subjectively too. I would also concur with Coulthard (op. cit.) that an accompanying recording would always be always helpful.

Some estimates put word-level transcription at ~10 minutes for every 1 minute of audio (Crowdy, 1995) while to produce a detailed transcript, complete with overlaps, stress markers and changes in tempo/pitch/speed this increases to ~20:1 (Ervin-Tripp, 2000).

13

Anglia Ruskin University MA Applied Linguistics with TESOL

2.2

SEX AND GENDER

While sex is defined as a biological characteristic, gender is somewhat more subjective. Gender has been defined as “a collection of traits or characteristics which a culture associates with a either men or women” (Kiesling, 2004, p.229, emphasis added) hence de Beauvoir’s adage ‘one is not born a woman, but becomes one’, everyone having been conditioned into one of the gender roles by cultural conventions. Western culture is geared towards the polarization of sex to the extent that biological sexual variance can be surgically corrected to ‘fit’ the binary categories; breast augmentations are on the rise and men strive to keep hold of however much hair they have left both sexes keen to conform to as close to the ideal as possible.. While Cameron’s views (2005) on the linguistic capabilities of the genders becoming closer aligned, holds some water societal factors still heavily underpin and to a certain extent accentuate the typical gender stereotypes. Importantly it is gender and not sex which is under review in this study.

The term women's language is problematic because much like ‘child’s talk’ it draws a comparison again a norm - in the case of gender this is an androcentric norm. Kaye in her two articles (1989a, 1989b) highlights just some of the linguistic biases of the English language, and the verb ‘gossip’ and the heavy female connotation it carries reinforces societies perceptions that women’s talk is ‘idle’ in contrast to men’s talk which is real and important.

1a.

S_r:

A respectful term for addressing males.

1b.

M___m:

A woman in charge of a brothel.

(Kaye, 1989a, p.185) The linguistic patterns women use has a long history, and women have been found to use a higher proportion of prestige forms than their male counterparts; where prestige is defined as “linguistic forms normally used by the social group with the highest social status” (Coates, 2004, p.47). Trudgill’s (1972) Norwich survey validated this hypothesis and women were found to use the /ng/ variant more than the stigmatised /n/ alternative; true against all social classes. Trudgill found that in formal contexts where the possibility for hypercorrection was greatest, women were more sensitive to their speech than men while interestingly the lowermiddle-class women were found to upgrade their pronunciation to that of their peers one class above. Macaulay’s (1978) study into /i/ revealed similar, interestingly, he also found that lower-middle class males identified more with the working class than with the upper-middle class. Overwhelmingly women’s pronunciation was found to be closer to the prestige model 14

Anglia Ruskin University MA Applied Linguistics with TESOL than men’s and where women’s speech didn’t match that of the prestige variant women were prone (Trudgill op. cit.) to report that they did while under-reporting for men was the norm.

In Labov’s 1966 (2006) study in New York he found that Macy’s e mployees (gender not given) stratified themselves with that of the higher-class Saks’s employees; the sales people borrowing the prestige from their customers. There was a distinction that middle-class employees tried to associate with the class above while in this study those at the bottom rung of the ladder (working for Klein’s) seemed resigned to keeping their working-class vernacular possibly at an attempt to identify with their customer base. Much like Trudgill’s study (op. cit) it was the lower middle class who were prone to hypercorrect, possibly more conscious of the prestige forms and the emulation thereof. The hypothesis that women themselves believe that they are perceived by how they appear and say as opposed to what they do ensures they try to elevate their social status by monitoring the use of their language. Their need to maintain this prestige form may influence the number of risks women take with the language, producing fewer truly unique forms and pushing their linguistic boundaries less than their male counterparts, a test which can be validated in this study. As a rule the belief is that it is women who keep within the bounds of their linguistic parameters, but recent research has shown that when it comes to language change it is the women, and not the men who lead the way. This is an important area of research and the linguistic behaviour of the genders is a key component in this study. One area which has been extensively studied is the variation in the non-standard forms of ‘was’ and ‘were’.

Figure 2: Diachronic distribution of non-standard was among men and women in York (from Tagliamonte, p.182) X axis represents generation.

Tagliamonte (1998) investigated. ‘was/were’ variation in the city of York. She found a plethora of grammatical violations (see Figure 2) specifically in relation to existential constructions and

(2)

(1)

non-standard was in

non-standard were in negative tags leading to such

malformations as “They said there were nothing wrong” and “He was shorter and stockier, 15

Anglia Ruskin University MA Applied Linguistics with TESOL weren't he?” (respectively) among others (p.155). Rather worryingly she warns “…if the males adopt the change, the morphological form was in existential constructions may become the norm, at least in the spoken language” (p.184). Interesting to note is how the males are assumed to be the tipping point.

Eisikovits’ (1991) Inner Sydney study also revealed some interesting results with the use of non-standard past tenses but they appear to contradict Tagliamonte’s study. While Eisikovits found that girls did use more non-standard past tense forms than boys, she also found that this was only true of young girls (aged 13 as an average) whereas by adole scence (average age 16) the linguistic patterns of these girls had changed and aligned closely with the boys’ usage which had been relatively stable throughout.

Eckert’s (1989) Detroit study was not related to was/were and did not find a polarisation between the genders per se. In fact Eckert was very specific in the need to account for societal factors as underpinning her results. Her ‘jocks’ and ‘burnouts’ study has received a great deal of attention and highlights Cameron’s (2005) need to move away from a view of gender as we view sex. Her study did not highlight any significant breakaway phonetic mutations but she did find that girls were more experimental using “more variation than the boys” (p.263). Eckert showed that the most extreme users of phonological variables were by those who have to do the greatest amount of work to affirm their membership in groups or communities. Aligning themselves with their peers is seen as integral and this linguistic ‘Community of Practice’ is one way to achieve that.

In brief these studies show that women use more prestige forms, adolescents experiment more – with girls leading the way; Tagliamonte’s diachronic anomalies (see Figure 2) were explained away with generational shifts. Ultimately males were found to be linguistically content although even Trudgill’s study (op. cit.) found that it while it was true that women used more prestige forms it was men who led the way in the use of non-standard forms apparently unconcerned with their low-prestige forms. An issue must be taken with any study which bases its results around adolescent teens – an issue which will come up again shortly. This identify forming period raises its own questions although the finding that gi rls experiment more linguistically remains valid.

16

Anglia Ruskin University MA Applied Linguistics with TESOL

2.2.1

POLITENESS

Figure 3: Wolfson's bulge model (1998)

Wolfson’s bulge model proposes that the greatest level of politeness is given to those whom we know where the relationship is dynamic and open to negotiation (Wolfson, 1998), she hypothesised that when there is a constant need to reaffirm one’s relationship more polite language is used. The contradiction which immediately comes to mind is that of groups of men engaged in team sports. Here the language is frequently coarse and standoffish; insults fly and slang is de rigor. The function of language here, however is one of aligning oneself with a group, they are, by all accounts, linguistically inclusive devices used to express inclusion and camaraderie with a group (a view also echoed by Pilkington, 1992). Here one would expect to find very low frequency counts of any politeness device however the language is still be used to express positive politeness or friendliness. Moreover the sitcom ‘Friends’ is based around 6 relatively well-to-do twenty-somethings (later becoming thirty-somethings) a brother and sister pair and four of the six are referenced as having schooled together – essentially friends who are intimate with one another whom altogether have a low social distance from one another. The show frequently centres around two apartments in which 4 of the stars cohabit and given Wolfson’s model one would come to expect an exceedingly low level of politeness regardless of gender within conversations confined to 1)the 6 characters 2)the four who live in the two apartments and 3)those who actively live together. Interruptions are a device which are often analysed with the view to making hard-and-fast conclusions. Studies such as Zimmerman & West’s (1975) suggested that they are another way in which men maintain their ‘conversational dominance’.

Sacks et al. (1974) model of

conversation defines simultaneous speech as violation of conversational politeness ; their model documents conversations as orderly and predictable events where the participants quick-step in time to one another harmoniously transferring a metaphorical baton between themselves at each legitimate completion point. Tannen (1994) had narrowed the definition slightly to “when a second speaker usurps another speaker’s right to continue speaking by taking the conversational floor in the absence of any evidence that the other speaker intended to relinquish the turn.” (p.33). Essentially interruptions are a device used to impose ones power on the conversation, demeaning the speakers right to talk. 17

Anglia Ruskin University MA Applied Linguistics with TESOL Traditionally men have been seen as the power brokers of society and hence the inference has been that men are more prone to interrupt. To an extent this hypothesis has been validated; Zimmerman and West’s (op. cit.) study bore this out with a resounding 96%/4% ratio and Drass (1986) observed that men were more likely to interrupt when their listening time far exceeded their talk time. Drass graded 91 undergraduates according to their ‘maleness’ (see Drass, op. cit. and/or Burke & Tully, 1977 for the details), and in a nutshell concluded that sex per se was irrelevant but that the gender variable proved influential: “...the more maleness *the person+… the greater the hazard that the person will initiate an overlap or interruption during a conversation” (p.299). This could be because men are simply conditioned to be the ones dominating the conversation, Drass concluded or as Tannen declares the parameters are just different. Tannen (1990) documents how in an hour-long radio phone-in about abortion – a key women’s topic - “all the callers except two were men” (p.88), she goes on to draw attention to how females loathe “‘catapulting’ themselves onto centre stage” (p. 88) or being the centre of attention – should women choose to interrupt in an public environment this is exactly what they would be doing.

Most of these studies have looked at the polarization of gender and assumed that each participant is on a level playing field to start with an equal knowledge of the conversation. However this is rarely the case.

Televised political interviews are a rich source of

conversational analysis and in 1982 Beattie analysed two interviews, one with a male interviewer and the leader of the opposition - Jim Callaghan and a second with a different male interviewer and the then Prime Minister - Margret Thatcher. Despite her powerful position, Margret Thatcher was interrupted significantly more frequently than Jim Callaghan. While Thatcher herself interrupted less than 40% of the time compared to Callaghan’s 55%. The conclusion here was not necessarily one of either power or gender but intelligence. Rim (1977 in Beattie, 1982) states it is the less intelligent subjects who interrupt more frequently than the more intelligent subjects. Broad sweeping statements are foolish, however the evidence suggests that there is a complex relationship between interruptions and dominance and that they are affected by many variables.

English is a flexible language where the lines between polite and impolite are blurred and subjective. North-East Asian languages with their honorific forms are very different and Japanese Women have long been the subjects of research (Smith, 1992) as to how women in power address their [male] subordinates. Politeness here is not marked by the use of inference or intonation but instead by the unidirectional use of ‘keigo’ a ‘honorific langu age’ 18

Anglia Ruskin University MA Applied Linguistics with TESOL used by a wife to her husband, from youngsters to their elders and so on. Such forms uniquely sum up the powerless status females in this society have.

a) “Yobu.”

[Translation: Call (imperative).]

b) “Yobimasu.”

[Translation: Call – polite variant. e.g. 'I will call him'.]

(from Smith, 1992, p.60)

Intriguingly, Smith studied the “linguistic conflict” (p.79) women in authority find themselves in when being forced to chose between a traditional powerless-heavy form or a more appropriate level of speech. A solution she discovered was that far from mimicking the speech of males, the empowered women got creative; they employed linguistic forms mothers use when talking to their children and created “new and powerful strategies” (p.79) where they acted passively but confidently omitted honorific verbs completely or substituted them. At no point did they ‘tread on the toes’ of their male counterparts and use the same language they would use.

On the back of work done by Lakoff (1973), O’Barr & Atkins (1997) embarked on a project to define women’s language, their highly popular paper concluded that it is not so much women’s language as a powerless language and that anyone is a powerless position will use more of these devices. “’Woman’s language’ is not fundamentally about gender but more basically about the lack of power” states Bucholtz (2004, p.6). O’Barr & Atkins’ courtroom study found that similar linguistic patterns were observed between a testifying housewife and a testifying male ambulance driver (both relatively powerless) while well educated professionals exhibited inverse features regardless of gender. Deuchar (1988) also argued that it is the powerless who are more likely to be linguistically articulate than those with power.

O’ Barr & Atkins

determined that power “derive*d+ from either social standing in larger society” (p. 384) is the key determinate and not gender. However they still upheld the assertion that “women tend to be high in women’s language features while more men tend to be low in these same features” (p.385)

Tannen asserts that women use language to achieve create and maintain relationships, defining ’rapport talk’. For women, "talk is the glue that holds relationships together," (1990, p.85) and so conversations are "negotiations for closeness in which people try to seek and give confirmation and support, and to reach consensus." (p.25). Men, on the other hand, use language to convey information, resulting in what Tannen calls ‘report talk’. Because men maintain relationships through other activities, conversation for them becomes a negotiation 19

Anglia Ruskin University MA Applied Linguistics with TESOL for status in which each participant attempts to establish or improve his place in a hierarchical social order. On top of this, the genders also view politeness differently where “what is perceived as rude, impolite and disruptive by women may be acceptable and normal in male interaction” (Holmes, 1995, p.53).

Cameron’s views (2005) on a spectrum of genders aside momentarily; there is no doubt that men and women are fundamentally different. A toy company, tested a playhouse which was being considered as a gender-neutral toy. After tests with children they found that girls and boys did not interact with the structure in the same way; amusingly “…the girls dressed the dolls, kissed them, and played house …the boys catapulted the toy baby carriage from the roof” (Sommers, 2001, p.27). Boys and girls are different concluded the company. Not only are our use of toys different but also our use of language and such distinctions are visible from an early age. Many (Sauntson, 2007 in England & Farris, 2000 in Taiwan) have shown that school boys and girls use language differently; boys are more direct and insistent while girls combine direct and indirect strategies softening their language. Farris concluded that girls were imitating their mothers while Sauntson concluded that girls “tend to contribute towards the accomplishment of collaboration and consensus within the group” ( 2007, p.322) opposed to boys who use language for the “negotiation of status and hierarchy within the group” (2007, p.322)

Therefore, if politeness is the domain of the women is the conclusion that men are impolite, rude and abrasive? Is impoliteness a gendered monopoly, the domain of only one gender? No-one would reasonable suggest it is but does one gender ‘do’ impoliteness better than the other? Consider the following: The BBC2 TV show “The Weakest Link” features a sharptongued host - Anne Robinson dubbed the “Queen of Mean” (The Sun, 2011). The show originated here in the UK however since then it has been exported to a further 42 countries (Wikipedia, 2011). Out of the 43 countries producing the show the production agency found it appropriate to install a female presenter as the host for all but 10 countries – giving women a near 80% stronghold on the role of playing “a cross between Cruella de Vil and Hitler's mother” or any one of a handful of playful comparisons (BBC, 2001) This raises a very intriguing question - Why did production companies the world over pick a woman for this role when they could have as equally picked a male counterpart? I propose no answer to such a question however the facts are interesting and have broad social implications.

20

Anglia Ruskin University MA Applied Linguistics with TESOL Definitions of what constrains a ‘taboo word’ vary and research into this area has faulted at this very hurdle; however the general negotiated consensus is that they are governed by social and intrapersonal parameters; heavily dependent on contextual variables - predominantly ones company. Taboo language has long been associated with males despite their having been little in the way of hard evidence to support such claims (Coates, 1986). Men are perceived as using more swear words than their female counterparts and they are perceived to use them more often. However recent work has cast doubt on this. Gao (2008) has shown that women have the more active vocabulary when it comes to swear words (47 vs. 31). His study of taboo language in another popular prime-time TV series ‘Sex and the City’ is important although it is worth noting that the professions the female actors are engaged in range from an independent businesswoman and a career-minded lawyer to a newspaper columnist and straight A finance student - the epitome of empowered positions. This is in line with Trudgill’s (1983, p.30) assertion that swearwords "occur very frequently in the speech of some sections of the community. This is largely because taboo-words are frequently used as swear-words, which is in turn because they are powerful." Other studies have had interesting results, for example Hughes (1992), found that both men and women swore more in the company of their own sex and that males’ use of swear words dropped significantly in mixed-sex conversations.

If power is the key to the use of swear words then one would expect a correlation where the social status of peers is the same – for example pupils at school. De Klerk (1992) examined exactly this using different groups from different schools (consistently year 6 and year 9 pupils) she found that the hypothesis that females did use taboo words was supported by her data, lamenting “female[s] were never at a loss for derogatory words” (p.10), age correlated more than gender she found. While a study during the identity-forming adolescent period raises its own questions the conclusion that girls were not striving for prestige speech but instead that which their peers use muddies the water somewhat.

21

Anglia Ruskin University MA Applied Linguistics with TESOL

2.3

CORPUS RELATED LITERATURE

The growth of corpus linguistics has bridged the gap between Chomsky’s view (1965) that only ‘competence’, and not ‘performance’, was worthy of study. His view that corpus linguistics “is almost useless… for linguistic analysis of any but the most superficial kind” ( 1961, p.131) has been well and truly extinguished; as Baker et. al. (2000, p.421) proclaim, “corpora have become the touchstone of credibility”.

Building a corpus is a complicated affair but they are needed because they show us how language is used - descriptive accounts rather than prescriptive. Owen (1993) and Francis & Sinclair (1994) engaged in a tête à tête on the merits and wherewithals on corpus linguis tics each vehemently arguing their corner. New insights into language are afforded by large corpora insist Francis & Sinclair while “grammars are founded on their authors' linguistic beliefs” and the “availability of information does not guarantee *an+ accurate description” insists Owen. Chomsky (1962) has argued that, whatever is found in a corpus is restricted to what is in the corpus and is not necessarily representative of the entire potential of a given language; his generative grammar theory has long been fraught with issues and copious amounts of real-world data only serves to reinforce such scepticism. In corpus analysis we are interested in language use as it is used (frequent collocates, idioms etc.), and while the Cartesian Product of syntactically valid sentences given only a handful of nouns and verbs is vast, we fail to exercise this creative extent opting instead for a tiny subset of what is grammatically correct (Pawley & Syder, 1983).

Representiveness is a key term which repeatedly crops up in corpus design. When it comes to corpora it is the holy grail (Biber, 1993). It entails having confidence that the sample of data selected adequately represents the genre of data it is selected to represent, achieving it would somewhat allay Chomsky’s (op. cit.) fears. With reference to this study representiveness is not a large issue since the corpus includes all available data providing a “thorough definition of the target population” (Biber, 1993, p.245) although a discussion is still merited. Having too few texts in a corpus can of course lead to skewed results and any generalisations can be misleading. Biber et. al. (2006) cite how 'burrow’ is the third most common collocate with the adjective ‘great’ (behind ‘deal’ and ‘man’) according to the Longman-Lancaster Corpus. Their explanation for this anomaly was the result of the inclusion of the text from ‘Watership Down’ where ‘the great burrow’ was the central meeting place for the rabbits. 22

Anglia Ruskin University MA Applied Linguistics with TESOL Also of interest is how different genres of text can influence results – words are used differently in different contexts. ‘See’ collocates highly with ‘paragraph’ and ‘chapter’ in the 2.2 million word British National Corpus (BNC) Law subset but with ‘patients’ and ‘table’ (as in on a page) in the 1.4 million word strong BNC Medical subset.

The BNC in their strive for representiveness somewhat arbitrarily divided the UK up into 12 zones in three supra-regions: The North, The Midlands and The South (see Figure 4 below). Most strikingly is the size of each of these areas - the heterogeneously rich West and the East Midlands represented by only two distinct areas. The linguists then set about collecting data from each of these regions; interestingly, within each region demographic sampling was used using “random location sampling procedures” (Crowdy, 1995, p.225). This is not a unique phenomena and subsequent approaches (Baker et. al., 2000) have also had to decide just how many nominal forms their corpus represents.

Figure 4: The BNC Spoken Corpus regions (Crowdy, 1993, p.260)

With written data, Atkins et. al. (1992) cited 25 different variables which must be considered in the design of a corpus; they range from cataloguing the age of the author to categorising the 23

Anglia Ruskin University MA Applied Linguistics with TESOL text status as either original or an adapted reprint. For spoken data the BNC devised a conversation log (see Figure 5 below). The audio was recorded on cumbersome cassette players, hidden from view essentially carried around in people’s pockets (Crowdy, 1993) . Such issues have long gone however as Leech et. al. mention, a new set of problems are now being dealt with.

Figure 5: BNC Spoken Corpus ‘Conversation Log’ (Crowdy, 1993, p.260)

“The use of the computer, in spite of the many advantages and new possibilities which it opens up, does not resolve the problems of the relationship between the original speech event and the transcription, nor does it obviate the problem of representing spoken language in written form.” (Leech et. al., 1995, p15)

At least one subsequent attempt (Baker et. al., 2004) to the BNC tried to replicate this method of concealed recordings only for it to end in abject failure. The architects of the project ended up using recordings from BBC broadcasts after their attempt to get spontaneous spoken data from fourteen South Asian languages all non-indigenous UK minority languages proved problematic. Issues regarding an active resistance to being recorded led to tiny amount of spoken data being recorded and the linguistics ended up harvesting the spoken data from the 2-hour nightly broadcasts from the BBC’s Asian Network – language which not as stifled and 24

Anglia Ruskin University MA Applied Linguistics with TESOL scripted as my Friends corpus still poses significant questions about its reflection of true spontaneous speech an observation also concluded by Baker et. al. (2000). Serendipitously they cited both clearer sound quality and a more distinct sentence structure to th eir advantage.

Resorting to BBC broadcasts did not alleviate their problems. Employed transcribers actively refused to transcribe some dialects declaring that linguists should study “classical… texts and not the bastardized slang” (p.516) the immigrants used. There were also issues over code switching and objections by the transcribers over the non-standard and non-prestige forms being studied. The result of these tribulations was that the 93 million word total was made up of a little more than 2 ½ million spoken words with the rest being harvested straight off the internet. Most of their text from daily news websites, exploiting their archive s of material. Far from just logistical reasons – although this was the overriding force for their selection – the web sites also provided a range of genres from sports to politics with a range of authors another fortuitous side-effect.

XML (eXtensible Markup Language), the widely-used successor to SGML (Standard General Markup Language) are both used for the storage and processing of texts in electronic form (although the latter could be considered obsolete). The big advantage they both hold is that they are interoperable between devices and between operating systems unlike proprietary systems such as an Oracle database or Microsoft’s .doc format. This is important given the variable nature the data will be used for. Essentially the structure of the data is a hierarchical tree (see Figure 6). The BNC used to index their data using a standard mark-up language (SGML) although this has since changed to the more interoperable XML. The reason for the change was simply a decrease in the monetary cost of storage, a key issue in the late 20th century.

XML, despite its name, is not a mark-up language in that it doesn’t mandate any tags (in contrast HTML mandates codes such as , ,