37
How can corpora be used to explore the language of poetry and drama?
Dan McIntyre and Brian Walker
1. Poetry, drama and corpora
Recently, methods from corpus linguistics have increasingly been applied in the analysis of literary texts (see, for example, Semino and Short 2004; Mahlberg 2007; O’Halloran 2007). In this chapter we outline some techniques for using corpora to study poetry and drama and demonstrate the value of using a corpus linguistic methodology in stylistic analysis. To demonstrate how corpus linguistic techniques might be applied in the analysis of poetry and drama, we analyse two specially constructed corpora. The first contains the poems from William Blake’s Songs of Innocence and of Experience and the second is composed of approximately 200,000 words of Hollywood blockbuster film-scripts. We use these to answer a series of research questions, in order to show how corpora can shed light on particular stylistic issues.
2. Background to Songs of Innocence and of Experience
William Blake published his first collection of poems in 1783, with Songs of Innocence (SoI) following six years later in 1789. Songs of Experience (SoE) appeared in 1794, and although ‘advertised … as a separate companion volume’ (Bottrall 1970: 13) was soon bound together with Songs of Innocence and both collections were issued in one volume with the combined title of Songs of Innocence and of Experience (hereafter referred to as Songs). While the reputation of Songs was slow to develop, they have become widely admired and the subject of much critical attention. Bowra, for example, claims that ‘the Songs deserve special attention if only because they constitute one of the most remarkable collections of lyrical poems written in English’ (Bowra 1970: 136).
3. A corpus-based case study of Songs of Innocence and of Experience
SoE and SoI are said to complement each other and are very much intended to be read as a whole (Bowra 1970). According to Bowra, the primary theme of Songs is contrast. Our main research question, then, centres on how these two collections are different. We will explore the lexical and semantic differences that manifest themselves in Songs. We compare the words and semantic domains in SoI with the words and semantic domains in SoE, and vice versa. We also consider whether generalisations made by previous analysts are supported by corpus-based generalisations.
We used the software Wmatrix (Rayson 2008) to carry out the computer-based analysis of Songs, utilising its automatic semantic annotation system. Wmatrix assigns semantic tags by matching the text against a computer dictionary of semantic domains developed for use with the software (see Rayson et al. 2004 for details of this procedure). The source text used for this study was obtained from Project Gutenberg (see website) and contains forty-seven poems. Two electronic versions of Songs were created in plain text files (unformatted .txt), so that we had one file containing SoI and the other containing SoE. The texts we used were checked and amended using facsimiles of Blake’s original plates (available online). The amended .txt versions of SoI and SoE were uploaded onto Wmatrix and compared with each other at the word level, to produce a list of key words, and at the semantic level, to produce a list of key semantic domains.
The notion of key or keyness here refers to items with an unusually high or an unusually low frequency in a source text or corpus when compared, using a statistical test, to a reference corpus. Wmatrix uses the Log Likelihood (hereafter LL) statistical calculation to evaluate differences in frequencies, and the higher the LL value the more key or statistically significant an item is or the higher the likelihood that the unusually high or low occurrence of an item is not due to chance. Keyness, however, is not an indication of whether the occurrence of an item is interpretively significant (for more on keyness, see Scott, this volume). This is up to the analyst to establish by exploring in more detail each key item. Note that in our study of Blake we have only considered key items that (a) occur more in our source corpus than in the reference corpus, and (b) that have an LL critical value of 10.63 or more, which indicates 99.9 per cent confidence of significance.
By comparing SoI and SoE against each other, we can see that there are very few key words and key domains with a LL of 10.63 or over. We could say that this is our first result, that lexically and semantically the texts are actually quite similar. There could be an argument, however, that the limited number of key items is connected to the small size of the texts we are comparing: SoI contains 2,433 words; and SoE 2,919 words. But we can show that this is perhaps not entirely the case by comparing Songs with a similarly sized but very different text in terms of genre, theme, style and period. For example, comparing SoI with a twenty-first-century academic journal article produces around thirty-nine key words and thirty key domains. So we can see from this that while the two collections of songs are, as Blake states on the title page of Songs, ‘Shewing the Two Contrary States of the Human Soul’, they are doing so using similar language, lexically and semantically speaking. There are, however, some differences highlighted by the comparison and we will turn to those now.
Table 37.1 shows the key semantic domains highlighted from the SoI vs SoE and SoE vs SoI comparisons.
HAPPY is the most key domain from the SoI vs SoE comparison. This contrasts rather sharply with the results from the SoE vs SoI comparison, where the most key domains are FEAR/SHOCKandVIOLENT/ANGRY. This immediately provides us with a semantic contrast which we might relate to Blake’s ‘Two Contrary States of the Human Soul’.
When using Wmatrix’s semantic annotation system, it helps to see what words have been assigned to the semantic categories you are interested in. This can help to
understand what the categorisation is telling us about a text (some semantic domains are not as transparent as HAPPY). The words for HAPPY, FEAR/SHOCK and VIOLENT/ANGRY are shown in Table 37.2. Joy, merriment and laughter occur more in SoI than they do in SoE, while SoE seems to be filled with anger, wrath, fear and terror.
Moving on to the second most key domain for the SoI vs SoE comparison, SPEECH ACTS, we can see that words to do with communication are more prevalent in SoI than they are in SoE. This could be related to interaction in the poems and provides some evidence towards more spoken communication taking place in SoI when compared with SoE. This, to some extent, ties in with Simpson’s (1993) comments, that ‘Songs of Innocence tend on the whole to be dynamic – in the sense that active communication takes place’ (Simpson 1993: 195). The ‘active communication’ is, to some extent, manifest in this SPEECH ACTS category shown in Table 37.3.
Turning now to the key words produced by our comparison between Songs, Table 37.4 shows that comparing SoI with SoE produced three times more key words than the SoE vs SoI comparison.
A useful first step in any analysis is to eliminate key items that, on initial investigation, do not seem to be interpretively significant. Three items fall into this category: Lyca,
merry and merrily. Lyca is the second most key word from the SoE vs SoI comparison and is the girl’s name featured in ‘The Little Girl Lost’ and ‘The Little Girl Found’. The third and sixth key words from the SoI vs SoE comparison are merry and merrily, which we have already noticed appear in the top key domain for the SoI vs SoE comparison (HAPPY). Notice, though, that while the key word lists for SoI vs SoE contain words that are included in the HAPPY domain, the key word lists for SoE vs SoI do not contain any key words associated with FEAR/SHOCK, VIOLENT/ANGRY, thus providing a good example of an instance where key words on their own are not enough to capture certain important differences between texts.
We can now consider the top key words from our Songs comparisons: thee (LL 18.44), lamb (LL 14.50) and what (LL 13.06). Thee (second person singular object pronoun), which is now archaic but was still in use when the poems were written, occurs twentyfive times in SoI and five times in SoE, suggesting that persons, animals or objects are being addressed directly more in SoI than they are in SoE. We can see from some sample concordance lines of thee in SoI (see Table 37.5) that direct address is taking place where the persona narrating the poem addresses either an entity within the poem or the implied reader/ listener. This relates again to Simpson’s comments concerning ‘active communication’.
However, looking more closely at the distribution of thee shows that it occurs in ‘Introduction’, ‘The Lamb’, ‘A Cradle Song’, ‘Night’, ‘Infant Joy’ and ‘A Dream’; just six out of the nineteen poems (32 per cent) that make up SoI. So while thee indicates that
direct communication takes place, interaction of this type happens in only a minority of the poems.
We can also see from a concordance of thee that twelve out of the twenty-five occurrences (48 per cent) of thee are in ‘The Lamb’, thus forming a cluster which focuses our attention on that particular poem. If we look more closely at the poem, the clustering can be broken down further, with eight of the twelve instances of thee occurring in the first stanza, forming either the direct object or the indirect object of a series of questions:
The Little Lamb (1st stanza)
(our emphasis)
(The third question, which spans lines 3 to 8, could be analysed as a series of four questions with who elided.)
The remaining four instances of thee, which occur in the second stanza of the poem, are in response to the questions posed in the first stanza. The first two introduce the answer ‘Little lamb I’ll tell thee’ and the last two form a blessing bestowed on the lamb, ‘Little lamb God bless thee!’ The poem depicts a scene where (apparently) a child is talking to a lamb. It is structured in two halves, with the first half asking a series of questions and the second half answering the questions. Further analysis of this poem is possible, but we can see that the clustering of thee focused our attention on ‘The Lamb’ and that clustering helped us notice (if we had not already) that thee is involved in a question-and-answer structure.
The second highest key word from the SoI vs SoE comparison is lamb. While also occurring in ‘Introduction’, ‘Night’, ‘The Chimney-Sweeper’ and ‘Spring’, it (not surprisingly) relates mostly to the poem of the same name. (Lambs also occurs five times in SoI, which, if included in the lamb-count, could provide evidence for a recurrent theme throughout the collection – but this would need to be explored more thoroughly). This key word, then, firmly focuses our attention on ‘The Lamb’. We will return to this key word shortly.
Turning now to what, anybody who is familiar with Blake’s Songs will probably conjecture that this key word relates to ‘The Tyger’, as this poem (famously) asks a series of questions (e.g. ‘What immortal hand or eye / Could frame thy fearful symmetry?’). This is indeed the case. While what is used in other poems, 64 per cent of occurrences are in ‘The Tyger’. The top key word, again, seems to focus our attention on a specific poem and a particular device used in that poem. Looking more closely at the questions that are asked in ‘The Tyger’ (another example, ‘And what shoulder and what art / Could twist the sinews of thy heart?’) we can see that what forms part of a complex subject, ‘what shoulder’, while the object of the questions consists of a noun phrase post modified by a prepositional phrase (‘the sinews of thy heart?’), referring to a part of a part (or parts) of the tiger.
Both ‘The Tyger’ and ‘The Lamb’ ask a series of questions, but there are some interesting differences. First, the questions in ‘The Lamb’ have who as the subject and thee as the object, while in ‘The Tyger’, what + body part forms the subject and thy + body part forms the object. The who in ‘The Lamb’ relates to a human form (God/Christ), while thee suggests that the persona narrating sees the Lamb as a whole. However, in ‘The Tyger’, the narrator does not seem to be sure what made the tiger, and the use of thy + body part suggests that the speaker sees only individual elements of the animal (a brief discussion of ‘The Tyger’ pertaining to this point can be found in Short 1996: 73). Lastly, while ‘The Lamb’ has a question/answer structure, ‘The Tyger’ only asks questions, but does not offer any answers.
Returning to the key word lamb, we can see that this word appears only once in SoE. In fact, it appears on line 20 of ‘The Tyger’ (‘Did He who made the lamb make thee?’). This line is foregrounded in the poem as it asks a who-question rather than a what-question. It also contains the word thee (whereas other lines contain thy + body part), and thus seems to form a firm link back to ‘The Lamb’.
The top key words seem to be directing us to specific poems in Songs and to specific elements within those poems that are related. These relations have not gone unnoticed by other critics and ‘The Lamb’ and ‘The Tyger’ are often seen as poems working in opposition or as contrasting elements of a connected entity. Bowra sees them as ‘symbols for two different states of the human soul’ (Bowra 1970: 158).
The final two key words to be considered are all and our and are used more in SoI than in SoE. Both words appear to encode inclusiveness, which we can illustrate with an investigation of all. All occurs in twelve out of nineteen poems. There are three instances of all being used adverbially for emphasis, while the rest of the time it is used as a (pre) determiner, either referring to the whole (eight times or 38 per cent) or functioning pronominally to refer to everyone (people) (thirteen times or 62 per cent). This contrasts with SoE where, as well as being used significantly less, all is used predominately as a (pre)determiner for non-human entities (71 per cent non-human whole, e.g. ‘All the night’, vs 29 per cent everyone/people). There is a very general sense, then, that SoI is more inclusive than SoE; SoI includes all people.
Looking more closely at the instances of all when it refers to people, we can see that in ‘Introduction’, ‘A Cradle Song’, ‘The Divine Image’ and ‘On Another’s Sorrow’ the reference is unrestricted, sitting within a religious context with some instances forming the object in clauses where the subject (for example, ‘Thy maker’ in ‘A Cradle Song) can be assumed to be Christ or God.
In ‘The Echoing Green’ and ‘The Chimney-Sweeper’ all refers to a restricted group of people; in the former all refers to ‘the old folk’ watching the children play; and in the latter the referent is ‘thousands of sweepers’. There is an interesting ambiguity at the end of this poem as the all here (‘So, if all do their duty, they need not fear harm’) could be read as being either restricted or unrestricted and it certainly seems that while the persona narrating more than likely is still referring to the ‘thousands of sweepers’, the poet is referring to everyone else and thus commenting on the social situation and the plight of the chimney sweepers at that time.
It should be apparent that in some cases our corpus analysis validates some of the subjective critical responses to Songs. Of course, key comparisons can only be a starting point. In order to fully understand the lists produced by a computer tool, we must return to the text. Quantitative analysis guides qualitative analysis, which might guide further quantitative analysis.
4. Blockbuster movie scripts as an object of study
In this section we demonstrate the possibility of constructing and analysing large corpora of dramatic texts to answer questions about genre. Our motivation for studying blockbuster movies comes from existing work on the blockbuster genre within Film Studies. Here we analyse a much larger corpus than the collection of Blake’s poems. We also make use of mark-up.
One problem with insights from Film Studies is that they tend to be rather more subjective and rather less detailed than is generally acceptable in stylistics. They also tend to avoid entirely the analysis of film dialogue. Nonetheless, a corpus stylistic approach can be used to validate some of the more subjective comments of film critics, and to provide insights into the linguistic construction of particular genres. To explore this, we constructed a 200,000-word corpus of action/thriller blockbuster film scripts.
The work of the film critics Tasker (1993), Jeffords (2004), Neale (2004) and Langford (2005) contain the following observations:
1. The action blockbuster is defined in part by an imbalance in its representation of gender, biased in favour of the white male hero.
2. The action blockbuster displays clear gendered roles and distinctions between genders.
3. Physical prowess is a trait of the white male hero.
4. The white male hero operates in the margins of society.
However, these are observations made on the basis of qualitative analysis of specific films. In order to explore the validity of the above claims for blockbusters generally, we set out to answer the following research questions:
1. Is there a difference in the amount of male and female speech in the action blockbuster?
2. Are gendered roles reflected in the key topics that male and female characters talk about?
3. Is physical strength and prowess a key semantic domain in action blockbuster screenplays?
4. Does male speech reflect an opposition to authority?
5. A corpus-based case study of blockbuster movie scripts
We took the definition of a blockbuster to refer to ‘a film which is extraordinarily successful in financial terms’ (Hall 2002: 11) as well as ‘those films which need to be this successful in order to have a chance of returning a profit on their equally extraordinary production costs’ (Hall 2002: 11). Our corpus comprises the following full film scripts:
1. | Air Force One (1997) |
2. | Alien (1979) |
3. | Armageddon (1998) |
4. | Basic Instinct (1992) |
5. | Bladerunner (1982) |
6. | Collateral Damage (2002) |
7. | Eight Legged Freaks (2002) |
8. | Fantastic Four (2005) |
9. | Ghostbusters (1984) |
10. | Indiana Jones and the Last Crusade (1989) |
11. | Jaws (1975) |
12. | Jurassic Park (1993) |
Practical constraints meant that we were not able to balance our corpus according to year of production. It is also the case that our corpus is relatively small and these issues are obvious caveats to our analysis. Nonetheless, our analytical findings at least provide some measure of objective support for the more subjective claims of Film Studies, which may be further tested against larger corpora.
We began by manually tagging our corpus to enable us to separate out dialogue from screen directions and subtitles. Table 37.6 outlines the mark-up we employed.
We then used Multilingual Corpus Toolkit (available from Scott Piao) to extract the male and female speech into two separate files. We used WordSmith Tools 4.0 (Scott 2004) to calculate word frequencies and type/token ratios, and Wmatrix to calculate key words, key semantic domains and n-grams (for more on n-grams see Greaves and Warren, this volume).
If the Hollywood blockbuster is defined partly by an imbalance in its representation of gender, with a bias in favour of the white male hero, then we might expect to see this reflected in the distribution of speech between male and female characters. An initial count of word frequencies shows this to be the case. In the corpus, there are 64,549 tokens of male speech compared to just 13,898 tokens of female speech. This is a clear
difference, even without a test for statistical significance. Male speech is clearly dominant in the corpus, which might suggest that male characters have a greater screen presence than female characters and are dominant in that sense. It is, of course, necessary to be careful in our claims. In terms of speech density, there is little difference between male and female speech, as the respective type/token ratios of 45.68 and 43.68 show (standardised type/token ratios were calculated every 1,000 words). It is not necessarily the case, then, that male characters are more linguistically complex characters than females. Nonetheless, in simple quantitative terms, there is evidence in the corpus that there is a bias towards male expression.
Of course, we also need to examine what characters talk about. To do this we examined the key words for each character, comparing male and female speech against each other, and against the 501,953 words of spoken demographic data taken from the BNC. Here are some observations from a study of the top twenty over-used key words for which there is 99.99 per cent confidence of statistical significance (p < 0.0001; LL critical value = 15.13). These key words are shown in Tables 37.7 and 37.8.
A number of the key words are proper nouns that turn up in one file of the corpus only. Nonetheless, it is perhaps noteworthy that Ben, Johnny, Ash and Reed are all names of male characters. No female names turn up as key in either the male or female speech. We can note further aspects of male-oriented discourse in the other key words used by the characters. In Table 37.7, the top key word for male speech when compared against the BNC Spoken Demographic Sampler is the vocative sir, addressed, of course, to other male characters. Guys too is generally used in the corpus to refer to groups of male characters. It would seem, then, that there is suggestion of a gender-based distinction between the speech of the male and female characters.
We can also note a preponderance of pronouns turning up as key words in both the male and female speech. In the male speech we have plural pronouns (we, us and our) and the first-person object pronoun me. These pronouns perhaps suggest a focus on group solidarity, which is often a feature of this genre of films. What particularly stands out in the female speech is the second-person pronoun you, since this is key not just in comparison with spoken language as a whole (where it is the most significant key word), but also with the male speech in the corpus (where it is the second most significant key word). There are 649 occurrences of you in the female data, and this focus on other
characters might be described as other-directed facework (cf. self-aggrandisement, perhaps), which appears common of female characters in the corpus. Gender roles are reflected in other key words too. Honey, for instance, is a term of endearment only used by female characters. And of the twenty-four instances of sorry in the female speech, twenty-two of these are apologies. We might see this as female characters occupying a subservient position wherein they feel called upon to apologise (often for circumstances beyond their control), or alternatively we might see this as evidence that they are characters more attuned to the feelings of others. Either way, this seems to be a clear gender role.
Finally, with regard to key words, it is interesting to note that while kill is key in both the male and female speech, only the female speech contains a peripherally related opposite, life. This focus on life as well as death might be seen to further reinforce the distinction between the oppositional pair male and female. Again, this might be seen as effecting a particular gender role for female characters.
There are, of course, other conclusions that we might draw from the list of key words. The proximal deictic terms this and here in the male speech perhaps reflect a focus on the here-and-now which we might see as a genre feature of blockbuster films. These are fast-moving thrillers as opposed to sedate, reflective art-house films. The fact that the definite article is also key might be a related issue. Definiteness as opposed to indefiniteness may again be a genre feature. We can also note the interpersonal discourse markers (e.g. hey) that indicate aspects of characterisation.
We can also draw out some characteristics of male and female speech by examining the key topics that each sex talks about. These were calculated using the key semantic domain function in Wmatrix and are outlined in Tables 37.9 and 37.10.
A number of the key semantic domains listed appear to reflect features of the blockbuster genre as a whole. So we find DEAD and ALIVE and WARFARE, DEFENCE AND THE ARMY: WEAPONS turning up as key in both the male and female speech, and this is no surprise given the typical plots of blockbuster thrillers. What stands out as particularly interesting from a gender standpoint is that PEOPLE: MALE turns up as a key domain in the male speech. An extract from a concordance of words belonging to this semantic category can be seen in Figure 37.1.
Some of these terms are vocatives (gentlemen, Mr, Mister, fellahs, Herr) while others are nouns referring to other male characters. The concordance demonstrates the predominance of male-oriented discourse, as we noted when we examined the key word sir (also part of the PEOPLE: MALE domain). Nonetheless, there also seems to be some evidence of a distinction between male and female speech in those semantic domains that are key for both male and female characters. For example, Figure 37.2 and 37.3 show samples from the concordances of SPEECH ACT.
We can note that male characters use a much greater variety of speech act terms. This perhaps generates a sense of the dominance of male characters in contrast to female characters.
Figure 37.1 Concordance of PEOPLE:MALE (male speech).
Figure 37.2 Concordance of SPEECHACT (male speech).
Figure 37.3 Concordance of SPEECHACT (female speech).
A closer look at some of the key semantic domains in the female speech also reveals gender-related characterisation. POLITE is a key domain, for instance, comprised entirely of thanking expressions, and its absence in the male speech perhaps suggests a genderrelated character distinction. In the PERSONAL NAMES category in the female speech, what is striking is that of the 285 instances that comprise the category, 212 are first names. This perhaps reflects a difference in interpersonal relationships expressed by male and female characters. Ervin-Tripp’s classic study of sociolinguistic rules of address (1972) suggests that use of first-names indicates social closeness, and it is interesting to note that this seems to be practised by female characters to a statistically significant level but not by male characters. We might hypothesise from this that male characters in blockbusters are emotionally more distant than female characters. Further support for this conceptualisation of male and female characters can be found in the fact that a positive key word in female speech is the endearment term honey, whereas such terms are not reflected in the male key words. It also seems noteworthy that RELATIONSHIP: INTIMACY AND SEX is a domain that is key only in the female speech. An extract from the concordance for this domain is presented in Figure 37.4, and what it perhaps suggests is that only female characters in blockbusters are explicitly characterised as intimate and/or sexual beings. Seen in tandem with the findings from the PERSONAL NAMES analysis, this supports the view that females in blockbusters are characterised as strongly emotional whereas males are not.
Finally, we can briefly comment on those key domains revealed when comparing male and female speech against each other. At the top of the list of key domains for female speech compared against male speech we find PRONOUNS, the vast majority of which are personal pronouns. There seems to be some evidence here of female characters
Figure 37.4 Concordance of RELATIONSHIP:INTIMACYANDSEX (female speech).
Figure 37.5 Concordance of INPOWER (male speech).
working harder at interpersonal relationships than their male counterparts. In contrast, the top key domain for male speech compared against female speech is IN POWER. Interestingly, while physical strength and prowess is not explicitly reflected in a key semantic domain (as our research question supposes it might), the concordance extract in Figure 37.5 of IN POWER demonstrates the predominance of power as a theme of male speech. This may be seen as tangentially related to the notion of strength and prowess.
We can now turn to clusters of words, and here we can also observe a number of interesting issues concerning n-grams in male and female speech. The largest clusters in the male speech are 5-grams, seen in Table 37.11.
The first 5-gram is an interrogative that perhaps reflects the element of surprise typical of blockbuster plots. The other 5-grams are all grammatical negatives. These seem to relate to the strong emphasis that blockbusters place on the ‘complicating action’ in plot terms, that the main protagonists then have to resolve. In the female speech there is just one 5-gram that occurs five times: I don’t want to. Like the male 5-grams, it is a grammatical negative, but unlike the male examples it expresses a lack of desire to act, which might again be related to a distinction in gender roles. The 4-grams in the female speech are as in Table 37.12.
We can note that the most frequent 4-gram utilises the same four initial words as the most frequent male 5-gram, which again we might see as relating to the complicating action of the plot and the difficulties these cause for the characters. With regard to the other 4-grams, again we see a concentration on expressing a lack of desire to act (I don’t want, don’t want to) as well as an apology (I’m sorry I) and more interrogatives than we find in the male speech.
Our analysis so far has begun to reveal some potential distinguishing elements of male and female speech in the corpus. From these results we might hypothesise further and say that this gender distinction is a feature of the blockbuster movie as a genre. Nonetheless, our analysis is of a relatively small corpus and one problem concerns its representativeness. In recent years, for example, there has been an increase in blockbuster movies with strong female lead characters (e.g. the Lara Croft: Tomb Raider films) and it might be argued that our corpus is skewed because it does not include sufficient numbers of these particular kinds of blockbusters. One possibility, then, would be to compare a sub-set of such films against the other films in our corpus.
Nonetheless, our findings perhaps allow for some tentative answers to the research questions we posed above. We found that there is indeed a difference in the amount of male and female speech in the blockbuster movie, and that male speech clearly dominates (question 1). Gendered roles also appear to be reflected in the key topic that male and female characters talk about, with a greater concentration on intimacy and the mechanics of personal relations in the female speech (question 2). Physical strength and prowess is not a key semantic domain for male speech per se (question 3), though this element of the male character might be said to be reflected in the IN POWER domain that is significant in the male speech. With regard to question 4, there does not seem to be any obvious reflection of a male opposition to authority; indeed, the contents of the PEOPLE: MALE domain suggest a clear recognition of hierarchical roles (evident in such vocatives as sir). However, further qualitative analysis of the corpus may well reveal more insights with regard to this question. Such analysis would also be likely to reveal further linguistic features of dialogue and screen directions that might constitute defining features of the action/thriller blockbuster genre generally.
We aim to have shown in this chapter how corpora of poetry and drama can be used to validate or invalidate the more subjective analyses of literary critics, and – especially in the case of drama – how a corpus linguistic methodology makes possible the kind of analysis that would not be achievable through manual qualitative analysis. In so doing, we hope to have shown how, for the stylistician, a corpus-based methodology provides a way of achieving the more objective kind of analysis that is the hallmark of the stylistic approach to criticism.
Acknowledgement
We are grateful to Kelly Stanger for her assistance in constructing and tagging the film corpus.
Further reading
Hoey, M., Mahlberg, M., Stubbs, M. and Teubert, W. (eds) (2007) Text, Discourse and Corpora. Theory and Analysis. London: Continuum. (In particular Chapter 7 ‘Corpus Stylistics: Bridging the Gap between Linguistic and Literary Studies’ by Michaela Mahlberg discusses some of the issues covered in this chapter.)
O’Halloran, K. A. (2007) ‘The Subconscious in James Joyce’s “Eveline”: A Corpus Stylistic Analysis which Chews on the “Fish Hook”’, Language and Literature 16(3): 227–44. (This article is an excellent example of corpus stylistics in practice that also deals with some of the theoretical issues surrounding the corpus stylistic methodology.)
References
Bottrall, M. (ed.) (1970) Songs of Innocence and Experience: A Casebook. London: Macmillan.
Bowra, C. M. (1970) ‘Songs of Innocence and Experience’, in M. Bottrall (ed.) Songs of Innocence and Experience: A Casebook. London: Macmillan, pp. 136–59.
Ervin-Tripp, S. M. (1972) ‘Sociolinguistic Rules of Address’, in J. B. Pride and J. Holmes (eds) Sociolinguistics. London: Penguin, pp. 225–40.
Hall, S. (2002) ‘Tall Revenue Features: The Genealogy of the Modern Blockbuster’, in S. Neale (ed.) Genre and Contemporary Hollywood. London: British Film Institute, pp. 11–26.
Jeffords, S. (2004) ‘Breakdown: White Masculinity, Class and US Action-adventure Films’, in Y. Tasker (ed.) Action and Adventure Cinema. London: Routledge, pp. 219–34.
Langford, B. (2005) Film Genre: Hollywood and Beyond. Edinburgh: Edinburgh University Press.
Mahlberg, M. (2007) ‘A Corpus Stylistic Perspective on Dickens’ Great Expectations’, in M. Lambrou and P. Stockwell (eds) Contemporary Stylistics. London: Continuum, pp. 19–31.
Neale, S. (2004) ‘Action-adventure as Hollywood Genre’, in Y. Tasker (ed.) Action and Adventure Cinema. London: Routledge, pp. 71–83.
O’Halloran, K. A. (2007) ‘Corpus-assisted Literary Evaluation’, Corpora 2(1): 33–63.
Rayson, P. (2008) ‘Wmatrix: A Web-based Corpus Processing Environment’, Computing Department, Lancaster University, available at http://ucrel.lancs.ac.uk/wmatrix/
Rayson, P., Archer, D., Piao, S. and McEnery, T. (2004) ‘The UCREL Semantic Analysis System’,in Proceedings of the Workshop on Beyond Named Entity Recognition Semantic Labelling for NLP Tasks, in Association with the 4th International Conference on Language Resources and Evaluation (LREC 2004). Lisbon: LREC, pp. 7–12.
Scott, M. (2004) WordSmith Tools 4.0. Oxford: Oxford University Press.
Semino, E. and Short, M. (2004) Corpus Stylistics: Speech, Writing and Thought Presentation in a Corpus of English Writing. London: Routledge.
Short, M. (1996) Exploring the Language of Poems, Plays and Prose. London: Longman.
Simpson, M. (1993) ‘Blake’s Songs of Innocence and Experience’, in J. Lucas (ed.) William Blake. New York: Longman, pp. 189–200.
Tasker, Y. (1993) Spectacular Bodies: Gender, Genre and the Action Cinema. London: Routledge.