5. noun phrases as style markers - · pdf filetive marker in a number of middle scots texts,...

24
5. Noun phrases as style markers 5.1. Introduction In chapter 3 above, I outlined three different approaches to the study of lan- guage in its social context. Two of these approaches, correlational sociolingu- istics and traditional stylistics, relate the use of relevant linguistic features to their contexts of use. At the end of that chapter, I reviewed a number of stud- ies of media language using a wide range of linguistic features. As this study is devoted to syntactic variation in newspaper language, I went on, in chapter 4, to sketch one area of the grammar of English, the noun phrase, which looks particularly promising as a locus for syntactic variation. In this chapter, I shall tie the two strands together and show in what respect the noun phrase can be used as a variable, and what results are to be expected from such an investiga- tion. Many researchers have recognised the value of the noun phrase as a style marker, both those who work in a correlational sociolinguistic framework and those who work in a traditional stylistic framework. The former concentrate on noun phrase structures that are in a paradigmatic relationship and - at least to some extent - are similar in meaning, such as different realisations of rela- tive pronouns or the opposition between the genitive and the of-construction. The latter, on the other hand, investigate for instance the complexity of the noun phrases and take this to be one of the stylistically relevant features. In the following I shall discuss the two approaches in some detail because this will provide the basis for my own analysis of the language of the British na- tional daily newspapers. 5.2. Noun phrase variables in paradigmatic relationships Over the last decade or so, many researchers have tried to apply a Labovian framework to variables beyond the immediate limits of phonology, and some of them have investigated the noun phrase as locus of variability. However, few, if any, have tried to adopt the free variation hypothesis to this level of variation. It is just too obvious that in this respect phonology differs substan- tially from other levels of linguistic description (cf. Bolinger 1977: 3; Lavandera 1978; Jacobson 1980; and Romaine 1984b). The collection of papers in

Upload: phamdang

Post on 12-Mar-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

5. Noun phrases as style markers

5 .1. Introduction

In chapter 3 above, I outlined three different approaches to the study of lan­guage in its social context. Two of these approaches, correlational sociolingu­istics and traditional stylistics, relate the use of relevant linguistic features to their contexts of use. At the end of that chapter, I reviewed a number of stud­ies of media language using a wide range of linguistic features. As this study is devoted to syntactic variation in newspaper language, I went on, in chapter 4, to sketch one area of the grammar of English, the noun phrase, which looks particularly promising as a locus for syntactic variation. In this chapter, I shall tie the two strands together and show in what respect the noun phrase can be used as a variable, and what results are to be expected from such an investiga­tion.

Many researchers have recognised the value of the noun phrase as a style marker, both those who work in a correlational sociolinguistic framework and those who work in a traditional stylistic framework. The former concentrate on noun phrase structures that are in a paradigmatic relationship and - at least to some extent - are similar in meaning, such as different realisations of rela­tive pronouns or the opposition between the genitive and the of-construction. The latter, on the other hand, investigate for instance the complexity of the noun phrases and take this to be one of the stylistically relevant features. In the following I shall discuss the two approaches in some detail because this will provide the basis for my own analysis of the language of the British na­tional daily newspapers.

5.2. Noun phrase variables in paradigmatic relationships

Over the last decade or so, many researchers have tried to apply a Labovian framework to variables beyond the immediate limits of phonology, and some of them have investigated the noun phrase as locus of variability. However, few, if any, have tried to adopt the free variation hypothesis to this level of variation. It is just too obvious that in this respect phonology differs substan­tially from other levels of linguistic description (cf. Bolinger 1977: 3; Lavandera 1978; Jacobson 1980; and Romaine 1984b ). The collection of papers in

Page 2: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

84 Noun phrases as style markers

Trudgill and Chambers ( 1991) is a clear indication of the recent increase of in­terest in syntactic variation.

The only way to salvage the Labovian framework for variables beyond phonology is to make a clear methodological distinction between the linguistic and the non-linguistic factors affecting the choice of one variant over the oth­er. If the two can be kept separate, then we are in a position to assess the ex­tent to which social or stylistic factors influence the chosen variable without the necessity to claim that there is free variation between its variants.

Two areas within the domain of the noun phrase structure lend themselves quite readily to this type of investigation. The first area is the choice of the rel­ative pronoun, and the second the alternation between the genitive construc­tion and the of-construction.

5.2.1. Relative clauses

In her important study, Romaine ( 1982) extends the methodology of correla­tional sociolinguistics in several directions. Not only does she use syntactic variables rather than the more traditional phonological ones, but she also uses a corpus that consists of written texts, and furthermore of texts representing an earlier stage in the development of the language she investigates. She analyses the occurrence of relative clauses and in particular the different types of rela­tive marker in a number of Middle Scots texts, written between 1530 and 1550.

Romaine does not use a binary variable, that is to say a variable with two variants, but a three-way opposition. The relative marker in Middle Scots, as in Modem English, could take one of three forms, WH (quhilk- which), TH (that) or 0 (instances of omitted relative markers). She meticulously distin­guishes between linguistic and extralinguistic factors affecting the choice of the relative markers. Among the former she discusses the distinction between restrictive and nonrestrictive relative clauses, the semantic and syntactic fea­tures of the modified noun phrase, and the syntactic position or function of the relative marker, whereas the latter are discussed in terms of contextual styles.

Croft (1986: 274) is of course correct when he points out about Romaine's linguistic factors that "most if not all of these factors are fundamentally extra­linguistic". It is indeed largely an extralinguistic matter whether a given noun modification is restrictive or nonrestrictive, or whether the modified noun re­fers to an animate or inanimate entity to give just two examples. But Croft really misses the point in this respect. It is an extralinguistic matter what type of antecedent, i.e. head noun phrase, is selected, but Romaine's point is that this choice will then affect, or to some extent restrict, the choice of relative

Page 3: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Noun phrase variables in paradigmatic relationships 85

markers. Hence the relative markers are to some extent dependent on the lin­guistic structures in which they are embedded and which themselves, natural­ly, depend on extralinguistic factors. But viewed in Croft's way, everything in the linguistic structure of a text will - ultimately - depend on extralinguistic factors.

One of the merits of Romaine's book is the fact that she approaches her data from several quite different angles. In this way she can test out different sociolinguistic methodologies on the same set of data. Her first approach cor­responds to what I have called traditional stylistics even though she does not use this term herself. She establishes different measurements relating to her relative clauses, which she then applies to her data as a whole and to its sub­parts.

She finds ( 1982: 140) that WH forms are the most common markers for non-restrictive relative clauses with some TH forms but very rarely 0, where­as in restrictive relative clauses the most common marker is TH. Both WH and 0 occur also but infrequently. The animacy of the antecedent, in marked contrast to present day English, does not seem to affect the choice of relative marker in Romaine's data of Middle Scots. The type of determiner of the ante­cedent likewise does not influence the choice in any noticeable way. The syn­tactic position of the relative marker within the relative clause has some influ­ence on its form. The system differs, however, from modem English in its standard British English form. There are, for instance, a number of cases of 0 relative markers in subject position of their clauses.

The two main measurements within the broad framework of traditional sty­listics are syntactic complexity and choice of relative marker. First, she dis­cusses the accessibility of the syntactic positions to relativisation, that is to say she investigates the syntactic roles played by the relative markers within their relative clauses. She refers to Keenan and Comrie's (1977, 1979) accessibility hierarchy, which claims that 1) a language must be able to relativise subjects, 2) that if a language can relativise a position low on the accessibility hierar­chy, it can also relativise all higher positions, and 3) that languages vary as to how far down on the accessibility hierarchy syntactic positions can be relativ­ised. (Keenan and Comrie 1977: 67). These claims refer to the following ac­cessibility hierarchy ( 1977: 66):

( 1) Subject > direct object > indirect object > oblique > genitive > object of comparison

In support, Keenan and Comrie cite numerous examples from data based on about fifty languages. Many Western Malayo-Polynesian languages, for in­stance, can only relativise subjects. Welsh can relativise subjects and direct

Page 4: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

86 Noun phrases as style markers

objects but not any other positions. Basque and Tamil can both relativise indi­rect objects besides subjects and direct objects, but again no positions further down the hierarchy. Very few languages allow all positions of the hierarchy to be relativised. English is one of them if (2) is acceptable:

(2) the man who Mary is taller than

This relates directly to a further claim that there is a direct correlation between the position on the accessibility hierarchy and gramm.aticality judgements. Judgements are uncontroversial towards the top end of the hierarchy but for every individual language they become more and more disputable towards their respective cut-off points in the hierarchy. Romaine extends their claim even further. She shows that positions high on the accessibility hierarchy are more often relativised than positions in lower ranks. The second measurement within traditional stylistics is the index of relative marker deletion, that is to say the frequency of the 0 relative marker in relation to all her relative claus­es.

Romaine then goes on to analyse her data by implicational scaling. The working hypothesis behind this approach is that each text in her corpus repre­sents an isolect which is part of a polylectal grammar but is itself invariant. It varies from a neighbouring isolect with respect to the output of one single rule.

The panlectal grid consists of the totality of possible sets of rules for an arbi­trarily limited area in space and/or time. The results of the implicational scale are used as a means of checking data against the predictions of a Bai­leyan wave model .... The assumption is that if the data scale to an accepta­ble degree, then they can do so by virtue of the fact that a succession (or wave) of rule changes has spread evenly through the grammar of a commu­nity. (Romaine 1982: 1 70)

For the purpose of this framework, she distinguishes four syntactic environ­ments and notes for every text in her corpus whether the relative marker is al­ways ( 1 ), sometimes (X) or never (0) deleted in that particular environment. However, the terms "always" and "never" have to be taken with more than just a pinch of salt, they refer to more than 25 and less than five per cent dele­tion respectively, with the variable cases lying between five and 25 per cent. As a result she presents table 5.1.

Page 5: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Noun phrase variables in paradigmatic relationships 87

Table 5.1. Implicational scale for relative deletion in Middle Scots texts

Text

The Scottish Correspondence Ane Satyre The Bannatyne Manuscript Acts of the Lords Council Burgh Records Boec:e Sherif! Court Book

Syntactic position of deleted relative Temp Dir.obj. Prep.

I 1 X 1*

1 X 0

1 1 X X 1* X 0

Deviations in bold face and asterisked (Romaine 1982: 171, table 6.31 ).

Subj.

X X X 0 0 0 0

This framework is designed to plot language change in progress. A rule delet­ing the relative marker applies "categorically" (i.e. more than 25 per cent) in particular syntactic environments, it then becomes variable and finally be­comes categorical again in non-application (i.e. fewer than five per cent). However, the table does not give any information on the direction which the change is taking. It could be from categorical application to categorical non­application or the other way round. It is only truly diachronic data that can re­veal in which direction a particular change is moving. It is customary to ar­range the rows and columns in such a way that a change can be said to enter the grammar at the bottom left -hand comer of the table spreading right across environments and upwards across the isolects usually represented by individu­al speakers but here by individual texts. It must be stressed, however, that Ro­maine sets up this table mainly in order to be able to criticise this particular approach (cf. in particular her excursus, 1982: 177-182).

Finally, in a third approach, Romaine subjects her data to the Cedergren­Sankoff variable rule program. This presents some problems because this pro­gram had been set up to handle the more traditional type of variables with only two variants, whereas Romaine is dealing with a three-way distinction. What the program, as developed by Cedergren and Sankoff (1974), does, is to estimate mathematically on the basis of observed frequencies the probability with which the relevant rule applies. Frequencies are random variables that cannot be predicted with I 00 per cent accuracy, but probabilities of rule appli­cation are fixed and pan of the competence of speakers (Romaine 1982: 183). Romaine is again highly critical of this approach, and she challenges Labov 's (1975: 228) view that a probability theory in sociolinguistics affords a "dra­matic increase in power and perception":

Probability theories might be considered more powerful with respect to their domain of application because, strictly speaking, they cannot be confirmed

Page 6: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

88 Noun phrases as style markers

or disconfinned by empirical findings. If probability statements are not falsi­fiable, then they have no empirical content. Thus, they can have no explana­tory or predictive power. (Romaine 1982: 183)

As I will try to show in detail in the concluding chapter, it is unreasonable (and self-defeating) to set methodological objectives for linguistic and socio­linguistic research which are not even attainable in natural science.

My criticism of this approach is based on the fact that it hinges on a highly implausible notion of speech community. It assumes that there are homogene­ous speech communities sharing a uniform grammar. The observed variation in surface frequencies of linguistic features is explained as the result of a prob­ability index appended to each rule indicating the likelihood with which it ap­plies for every speaker in every given context.

Nevertheless, Romaine's study is important for my present purposes be­cause she discusses many of the fundamental issues of sociolinguistics and be­cause she uses syntactic variables to illustrate her arguments. In the final chap­ter I shall come back to some of her points on sociolinguistic methodology. As hinted at above I shall suggest that she reaches unduly pessimistic conclusions primarily because of her unrealistic presumptions for sociolinguistics as a scientific method.

Her results are moreover important in that she uses one particular type of noun modification as a variable. The choice of the relative marker seems to be quite obvious from a sociolinguistic point of view. Even though she does not talk in terms of free variation - in fact she is fairly sceptical whether it is a useful notion for syntactic variation at all - it nevertheless seems clear that this is one of its attractions. There are linguistic and extralinguistic criteria, some of which have been summarised above, which govern to some extent the choice of one of the three types of relative marker, but given identical con­texts, there are no claims that the choice of relative marker would change the meaning or the function of the relative clause.

Cheshire ( 1982 and 1991) is another fairly recent example of a detailed analysis of variation going beyond phonology. Using a Labovian framework, she analyzes a group of adolescents frequenting two playgrounds in Reading, England. By using such a small and comparatively homogeneous group of speakers, she excludes a considerable number of non-linguistic parameters, such as age variation, regional variation and to a large extent social variation, since all her informants come from a basically working class background.

Her linguistic variables are also severely restricted. She excludes lexical and phonological variation and concentrates on morphological and syntactic features which are used in non-standard forms by her informants. One of these features is the relative pronoun. The paradigm of relative pronouns in Reading

Page 7: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Noun phrase variables in paradigmatic relationships 89

English has one variant more than Standard English, because what is also used in this function. A further difference to Standard English is the fact that the 0 relative pronoun is also used as a subject of its clause, which is not possible in Standard English.

5.2 .2. Genitive versus of-construction

Altenberg ( 1982 see also 1980) offers one of the most comprehensive recent studies of the variation between the genitive and the of-constructions, as illus­trated by the following two examples:

(3) Her father's arrival changed everything (4) The arrival of her father changed everything (Altenberg 1982: 11)

At the outset, Altenberg stresses that the two variants are not in free variation even though they "convey roughly the same meaning" ( 1982: 11 ). Like Romaine ( 1982), he concentrates on historical and written data. In his case the texts were all written in the 17th century.

He distinguishes six types of influencing factors, only one of which is con­cerned with non-linguistic features. Phonetically, a final sibilant in the modi­fying noun phrase is a significant, albeit not overriding, factor. EspeciaiJy in formal prose, it tends to increase the percentage of of-constructions. Morpho­logically, a regular plural ending in the modifying noun also tends to increase the percentage of of-constructions.

Syntactically, three main types of conditioning factor are distinguished: the premodifier weight of both the modifying and the head noun phrase postmodi­fication and other types of right branching expansions of the modifying noun phrase and the recursive expansion of the modifying noun phrase. The first of these factors has relatively little influence on the choice between a genitive or an of-construction. A right expanded modifying noun phrase favours the of­construction whereas a right expanded head noun phrase favours the genitive construction, both of which can be seen as strategies to keep modifying and head noun phrase close together. The genitive construction is less likely to have its modifying noun phrase recursively modified more than once, that is to say two genitives in a row seem to be about the limit. The of-construction, on the other hand, shows slightly more flexibility.

Lexically, the choice between the genitive and the of-construction is condi­tioned by the semantic nature of the modifying noun phrase. This appears to be one of the strongest conditioning factors. The frequency of the genitive construction is highest with modifying noun phrases denoting human individ­uals, somewhat lower for animals, very low for human collectives and abstract

Page 8: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

90 Noun phrases as style markers

nouns, and practically zero for concrete nouns. The lexical category of the head noun phrase, on the other hand, is marginal and idiosyncratic.

On the relational level, the choice is influenced by the kind of relation that obtains between the modifying and the head noun phrase. In Modem English, this allows John's betrayal to be interpreted as subjective and the betrayal of John as objective. Thus subjectiveness favours the genitive construction but the effect of the head-modifier relationship is to a great extent determined by stylistic factors.

The last area of conditioning factors deals with communicative aspects. Texts to which the status of private, intimate and personal can be ascribed tend to favour the genitive construction whereas public, distant and imperson­al texts favour the of-construction. In the same way non-specialised texts such as conversations on everyday topics tend to use the genitive construction whereas specialised texts such as religious or political texts or texts with liter­ary topics tend to use the of-construction. And finally, non-literary prose gen­res and strict verse forms prefer genitive constructions whereas literary prose genres prefer of-constructions.

This is only a selection of all the conditioning factors that, according to Al­tenberg, affect the choice between the two variants of his paradigm, and he re­peatedly stresses the problem of quantifying the influence of individual fac­tors, because in every single instance, all the relevant factors work together to determine the choice of one form over the other. But the above outline suffic­es to sketch a very sophisticated attempt at establishing the linguistic and non­linguistic factors responsible for one particular example of syntactic variation. He concludes:

Roughly speaking, the grammatical and thematic factors account for no less than 85% of the GEN/OF options in the variable prose material. Further­more, if we take into account that the stylistic choices are also highly pat­terned, and regard GEN as the unmarked choice in informal contexts and OF as the unmarked choice in most formal contexts, the percentage of "prin­cipled" variation rises to 98%. This does not necessarily mean that the re­maining 2% are in free variation or that they resist explanation altogether, only that other and more delicate instruments of analysis are needed. (Alten­berg 1982: 303)

Dahl ( 1971) writing a decade earlier than Altenberg, investigated the use of the genitive construction in journalistic style. It is generally recognised that the genitive is the preferred construction for those modifying noun phrases that are "highest on the gender scale" (Quirk et al. 1985: 1277). But Dahl points out that in journalistic style, the genitive is used in places that in other varieties would use the of-construction. According to her, the use of the geni-

Page 9: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Premodifier variation 91

tive for inanimate things is particularly common in newspaper headlines, but it has already extended its influence into the main text of newspapers ( 1971 : 141 f).

As the basis of her classification, she uses a large collection of genitives for inanimate things taken out of The Times, The Daily Telegraph, the Daily Mail and The Observer. Her classes include words denoting collective communi­ties, geographical names, the genitive before superlatives, names of animals, means of locomotion and machines, planets, buildings and places, newspapers and periodicals and some groups of less importance. It is unlikely that any of the classes mentioned by Dahl is actually restricted to newspaper language, even though it may be more frequent there. But her data does not allow any comparison with other varieties. She gives the frequency of all the construc­tions she discusses but they are not set in relation to the length of the texts from which they are taken or alternatively in relation to the frequency of the of-construction. Because of this failure, it is not surprising that she could not find clear differences between the four newspapers under investigation. Her collection contains more examples from The Times and The Daily Telegraph than from the other two newspapers but this is due to the fact that "their vol­ume of articles and news reports is greater than that of the Daily Mail and The Observer" (1971: 172) and thus it does not provide any relevant information at all.

5.3. Premodifier variation

Abberton ( 1977) presents a very delicate analysis of premodifying structures in noun phrases. She distinguishes 19 different types of premodifiers that can occur between the determiner and the head of the noun phrase. She also sets up twelve form classes according to the different patterns of premodification, but she uses more general terms than those of her 19 types of premodifier. To give some typical examples, pattern one consists of a head noun with neither determiner nor premodifier pattern three of an optional determiner, an adjec­tive and a head and pattern eight of a noun phrase in genitive case as a deter­minative and an optionally premodified head. The last of her patterns is given as the following formula:

(5) ± Det + ((± Adj) +Gen. N) + ((± Adj)+(± N) + N) (Abberton 1977: 58)

The pattern is exemplified by The farmer's butter and The kind farmer's nice buttermilk. In both cases, the bracketing given by Abberton seems implausi­ble. The determiner the is clearly part of the genitive noun phrase acting as a

Page 10: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

92 Noun phrases as style markers

determinative for the entire noun phrase. The genitive noun phrase and the de­terminer can only as a unit be replaced by a name, viz. John Brown's butter but not *The John Brown's butter.

Abberton then goes on to compare the use of these patterns in eight texts drawn from the Survey of English Usage, four from the category of novels and four from scientific writing. The corpus consists of some 10,000 words, which yields 6160 noun phrases for her analysis. But several of her twelve patterns are only represented by very few tokens. Hence her conclusions are not very far reaching. Her figures are too small to reveal any clear overall pat­terns.

The general conclusion is that both types of English examined are remarka­bly similar in nominal group predmodification [sic!] structure: the vast ma­jority of nominal groups are not premodified or are premodified by only a determiner and/or one adjective whose type tends to vary but slightly with the style of English in question. ( 1977: 63, her emphasis)

The most important differences that she is able to locate through her sophisti­cated classification scheme, that "postmodified nominal groups are far more common in the science texts" (1977: 62).

It must be conceded, however, that it was presumably the classification it­self that interested Abberton in her research rather than the correlations with the two types of written British English. The corpus seemed to serve more as a data base for the exemplification of her classes.

Coates (1977) investigates sequences of premodifiers in 17 texts of the Survey of English Usage. She compares such sequences in different structural positions of the noun phrase. Her main interest lies in the prenominal position between the determiner and the head, which she contrasts to postnominal posi­tion after the head to preprenominal position before the determiner and to predicative position outside the noun phrase altogether after the copulative verb to be.

Ordering, it seems, is basically a semantic matter. A head is modified to form a new conceptual unit, which can then be modified in its turn. Modifi­cation at any level can involve several modifiers or only one. And at the point where the nominal group expresses what the speaker/writer wants to say, modification will cease. (Coates 1977: 19)

The prenominal slot turns out to be particularly important in that it very often differs considerably from premodifiers that are further removed from the head. The item in prenominal position very often forms a new head together with its head to form a "temporary compound" ( 1977: 21 ), which may itself be further modified.

Page 11: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Complexity as style marker 93

Sichel schmidt (1989), finally, investigates the order of prenominal adjec­tive sequences in German. He is mainly interested in the psycholinguistic pro­cesses that govern the production and reception of premodifying adjectives. Is it the first adjective of a sequence which stays most prominent in the memory, or is it the last one, the one nearest to the head of the noun phrase (i.e. is the principle of primacy or the principle of recency more important)? He finds that potential semantic contrasts of adjectives to other adjectives in the context are generally more important than the mere syntactic ordering.

5.4. Complexity as style marker

Aarts (1971) is one of the first to notice the stylistic significance of the com­plexity of noun phrases. His short but important paper is based on extracts to­talling 72,000 words from the Survey of English Usage at University College London. His corpus can be divided into four categories: light fiction, scien­tific writing, informal speech, and formal spoken and written English. It was his working hypothesis that the distribution of noun phrase types within the English clause structure is not random, but that the subject position is associat­ed with structurally "light" noun phrases, whereas non-subject positions are associated with structurally "heavy" noun phrases. He further hypothesised that the distribution pattern would be significantly influenced by the variety of English in which it occurs.

Aarts counts those noun phrases as "light" that consist of a simple pronoun, a name or an unmodified noun with or without a determiner. All other noun phrases are understood to be "heavy". After presenting several tables with the detailed figures he summarises his findings in the following table:

Table 5.2. Distribution of"light" and "heavy" items after Aarts

Subjects Non-subjects

"Light" items

6749 4753

Source: Aarts ( 1971: 290, table 1 0).

"Heavy" items

1149 4310

Table 5.2 provides strong evidence in support of his first hypothesis. "Light" noun phrases are more frequent in subject position. 85.5 per cent of the noun phrases in subject position are classified as "light" as opposed to only 52.4 per cent of the non-subjects. On the basis of the figures for "light" and "heavy" noun phrases in the four text-groups, he can show that there is also some evi-

Page 12: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

94 Noun phrases as style markers

dence for his second hypothesis that the distribution of noun phrase types is influenced by the variety of English, but he is very cautious and does not at­tempt to draw far-reaching conclusions in this respect, even though the data as he presents it shows clearly that the association between "light" items and sub­ject is greatest in informal speech and smallest for scientific writing.

Quirk et al. (1972: 933-934, and again 1985: 1350-1352) reprint his fig­ures (without acknowledgement) in a somewhat modified table but with the same distinction between "light" and "heavy", or, as they prefer to call it, be­tween simple and complex noun phrases. Aarts' subsample of "formal spoken and written English" is merely labelled "serious talk" in the 1985 version of their grammar.

Unfortunately, the table with which Quirk et al. (1985: 1351) illustrate these findings is incomplete and it contains some blatant mistakes. Therefore I present the entire table with the corrected figures as table 5.3. Quirk et al. 's ta­ble lists the number of simple and complex noun phrases in subject and in non-subject position. For the entire sample they add the percentage figures and the totals for the noun phrases both in subject and in non-subject position. For the four subsamples they do not give the percentage figures or the totals. This saves valuable printing space but it makes it difficult to interpret the fig­ures without having recourse to a pocket calculator.

The first column of their table gives the total of each row, that is to say the total of both the simple noun phrases and the complex noun phrases either in subject position or in non-subject position. These row totals are also given for the subsamples individually. However, the figures that are given are not the correct sums of the respective rows. The names and pronouns, and the noun phrases with multiple modification are subsets of the simple and the complex noun phrases respectively and should, of course, not be added to these sums. The row totals that are given also do not add up correctly to the grand total of 16961 of the entire corpus. The correct sums of all simple and all complex noun phrases for each row, however, do add up to the stated grand total, and these corrected figures also correspond to those given by Aarts. ( 1971 : 291, table 9).

Page 13: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Complexity as style marker 95

Table 5.3. Noun phrase structure and distribution after Quirk et al.

Total all names & all multiple simple pronouns complex modification NPs NPs

whole subject 7898 6749 5821 1149 456 sample % 46.57 39.79 34.32 6.77 2.69

not subject 9063 4753 2193 4310 1777 % 53.43 28.02 12.93 25.41 10.48 total 16961 1 1502 8014 5459 2233 % 100.00 67.81 47.25 32.19 13.17

informal subject 2212* 2064 1941 148 62 speech % 52.77 49.24 46.30 3.53 1.48

not subject 1980* 1169 677 811 327 % 47.23 27.89 16.15 19.35 7.80 total 4192 3233 2618 959 389 % 100.00 77.12 62.45 22.88 9.28

fiction subject 2431* 2220 1943 211 92 % 46.45 42.41 37.12 4.03 1.76 not subject 2803* 1682 754 I 121 434 % 53.55 32.14 14.41 21.42 8.29 total 5234 3902 2697 1332 526 % I 00.00 74.55 51.53 25.45 10.05

serious subject 2088* 1745 1478 343 127 talk % 45.40 37.94 32.14 7.46 2.76

not subject 2511 * 1273 599 1238 492 % 54.60 27.68 13.02 26.92 10.70 total 4599 3018 2077 1581 619 % 100.00 65.62 45.16 34.38 13.46

scientific subject 1167* 720 459 447 175 writing % 39.75 24.52 15.63 15.22 5.96

not subject 1769* 629 163 1140 524 % 60.25 21.42 5.55 38.83 17.85 total 2936 1349 622 1587 699 % 100.00 45.95 21.19 54.05 23.81

Source: Quirk et al. (1985: 1351, table 17 .124). Asterisked figures are corrections of obvi-ous faults. Percentage figures relate to the total number of noun phrases per subsamp1e. Per-centage figures and column totals for the four different styles are my addition.

On the basis of this table, Quirk et al. observe that less than one third of the noun phrases in the entire sample are modified and less than one-eighth have multiple modification (the exact figure for multiple modification is 13.17 per

Page 14: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

96 Noun phrases as style markers

cent, which is, of course, slightly more than one-eighth rather than less). The differences between the four subsamples are very considerable in this respect. In informal speech and in fiction, as few as about one-fourth of all noun phrases are modified and only about one-tenth have multiple modification, whereas more than half of the noun phrases in scientific writing are modified and almost one-fourth have multiple modification.

Quirk et al. further observe that nearly one-half of the noun phrases in the sample are pronouns and names, but again there are considerable differences which need to be pointed out. In informal speech almost two-thirds are names and pronouns but in scientific writing they comprise only just over one-fifth.

There is a clear association of the complexity of the noun phrases and their syntactic position. In their words "the majority of simple noun phrases - and the overwhelming majority of names and pronouns - are subjects of clauses or sentences, but only rather less than a quarter of complex noun phrases are sub­jects" (1985: 1351 ). These facts are less readily apparent from the table above because the relevant percentage figures are not given. For the sample as a whole, 58.7 per cent of all simple noun phrases appear in subject position. The association is strongest in informal speech. In this style 63.8 per cent of the unmodified nouns are in subject position and only 15.4 per cent of the modi­fied nouns. This is a difference of 48.4 percentage points. The same difference is successively reduced across the four styles. It is 41.1 percentage points in fiction, 36.1 in serious talk and 25.2 in scientific writing.

On the basis of these overwhelming differences, Quirk et al. (1985: 1352) conclude that

even such coarse-grained comparisons as these make clear how sensitive is the noun phrase as an index of style and how responsive it can be to the ba­sic purpose and subject matter in varying types of discourse.

The application of a chi-square test to the four values of unmodified and modi­fied noun phrases in subject and nonsubject position respectively in the four subsamples of Quirk et al. 's table reproduced above, reveals that the differ­ence between these samples is statistically highly significant (X2 = 1093.64, df = 9, p < 0.00 I). In view of this fact, the validity of the figures given for the sample as a whole are of course somewhat reduced.

In strict statistical terms it is in fact inadmissible to combine samples that differ from each other at a level of statistic significance. (cf. Butler 1985: 112-123; Hatch and Farhady 1982: 165-172; Woods et al. 1986: 132-153). And it is easy to see why this is so. The four subsamples vary considerably in size, ranging from 17.3 to 30.1 per cent of the entire corpus, and thus they contribute rather differently to the overall result. However, it is not at all clear

Page 15: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Complexity as style marker 97

whether the four subsamples are really coherent samples, there may be, and presumably are, systematic variations within each of them.

This is an example of top down analysis. Generalisations are made about a large corpus as a whole, and subsequently steps are taken to single out subsets within the sample. The subsamples themselves may also tum out to have inter­nal variations at a level of statistic significance. If such variations are found, the validity of the generalisations at the level above are of course reduced in their validity, but not, I believe, rendered invalid, particularly so because they provided a necessary step in establishing the newly found variability.

Varantola (1984) deserves some considerable attention because in some re­spects her aims are similar to mine. She investigates the noun phrase as a style marker in a corpus of engineering English. For her this is an example of a reg­ister or a language for special purposes (LSP). In her methodology ( 1984: 51) she refers explicitly to Enkvist ( 1964, 1973, 1978). She compares her corpus of engineering English with a "reference corpus, a ' norm' that is used as a yardstick" (1984: 51). Varantola' s corpus of engineering texts consists of arti­cles and news items published in professional engineering journals that are ad­dressed to the "general specialist" ( 1984: 53) rather than to a highly special­ised audience, on the one hand, or to lay readers, on the other. The "norm" corpus consists of articles and news items published in The Economist, New Society, The Observer and Newsweek. Both corpora are predominantly but not exclusively British English. It may be regretted that American English texts were included at all because it is unclear if and to what extent this inclusion skews the results.

As a general overview of her data she presents the figures that allow a com­parison with Quirk eta/. 's data. She bases her comparison on the 1972 version because her work of course predates the 1985 version.

Table5.4. Distribution of NP types in engineering English as compared with a "norm"

Total all simple names & all multiple NPs pronouns complex modification

NPs

Engineering N 2547 628 96 1919 1259 texts % 100.0 24.7 3.8 75.3 49.4 General N 1756 767 179 989 610 texts % 100.0 43.7 10.2 56.3 34.7

Source: Varantola (1984: 89, table 5, and 90}, but transformed into Quirk et al.'s (1985: 135 J) format used above. The subcategory of simple NPs does not include names.

Page 16: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

98 Noun phrases as style markers

There are some quite remarkable differences between her data and Quirk et al.'s data. Even her "norm" corpus of journalistic texts has a higher percent­age of modified nouns than any of Quirk et al. 's categories. Well over half of all its nouns are modified. In her main corpus of engineering English just over three quarters are modified and almost half of all the nouns have multiple modification. The degree of pronominalisation, on the other hand, is minimal. In the "norm" corpus about ten per cent of all noun phrases are pronouns, whereas in the main corpus that figure is less than four per cent. However, it is difficult to compare this last figure with Quirk eta!.' s results because Varanto­la does not give the figures for names but only for pronouns. Some caution in interpreting her results may be advisable in any case. She concedes that her method of calculation "may be slightly different" (1984: 89), and moreover the category of complex noun phrases given in the table mentioned above does not include the noun phrases with multiple modification. Thus "com­plex" for her really means "one modifier only". In table 5.4 above I have re­calculated the figures in such a way that the category of multiple modification is a subcategory of all the complex noun phrases. But even so, the totals of complex noun phrases for the two corpora yield different figures than those given elsewhere in her book (e.g. 1984: 59 or 91). Nevertheless, the main thrust of her argument is certainly correct when she ( 1984: 89t) concludes:

Modification of nouns is thus a major means of information cramming and space saving in journalism. In engineering journalism its level easily be­comes extremely high because of both the need for conciseness and the in­herent need to label and specify.

As a fundamental difference between her corpus of general journalism and her corpus of engineering journalism, she finds that the former relies more heavily on premodification whereas the latter prefers postmodification.

The cause may be found in the projected readership of non-specialist texts. General journalism has to be intelligible to a wide range of readers with var­ying backgrounds and thus more explicit in expression whereas specialist journals can expect a very high degree of expertise from their readers and thus use language that is not overtly self-explanatory. (Varantola 1984: 91)

V arantola discusses premodifiers and postrnodifiers in tum. She notes that one of the main functions of premodifiers is to label and to categorise, but they are generally less specific than postmodifiers, in which grammatical rela­tions are often made explicit which in premodifiers are left to be inferred by the reader.

Page 17: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Complexity as style marker 99

She repeatedly stresses the fact that mere length of either premodification or postmodification is not a very reliable indication of complexity. In fact short and compact premodification may be far more complex and require a higher degree of inferencing power on the part of the reader than a long but fairly simple relative clause in postmodification.

One notable difference between her two corpora in the sphere of premodifi­cation is the fact that in the general language corpus, adjectives are by far the most common premodifier in the slot closest to the head whereas in her main corpus nouns were considerably more frequent in this position.

Varantola 's approach is in many respects similar to the one chosen for the analysis of newspaper English to be presented in the second part of this study. However, there are also considerable differences. She works with a fairly small corpus of some 2000 noun phrases, which she subjects to a very thor­ough and detailed analysis. The corpus itself is compared to a " norm" corpus of about half its size. Thus the genre of engineering English is seen as a devia-

Table 5.5. Noun phrase structure and distribution in headlines of The Times and the Daily Mirror

Total all simple names & all multiple NPs pronouns complex modification

NPs

Both subject 1380 832 396 548 samples % 50.15 30.23 14.39 19.91

not subject 1372 666 177 706 % 49.85 24.20 6.43 25.65 total 2752 1498 573 1254 % 100.00 54.43 20.82 45.57

Times subject 1105 636 12 + 247 469 % 49.73 28.62 11.66 21.11 not subject 1117 503 4 + 117 614 % 50.27 22.64 5.45 27.63 total 2222 1139 16 + 364 1083 % 100.00 51.26 17.10 48.74

Daily subject 275 196 53+ 84 79 Mirror % 51.89 36.98 25.85 14.91

not subject 255 163 30+26 92 % 48.11 30.75 10.57 17.36 total 530 359 193 171 % 100.00 67.74 36.42 32.26

Adapted from Mlirdh (1980: 109, 110, tables 17, 19).

Page 18: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

100 Noun phrases as style markers

tion from, or rather a stylistically marked form of, normal English. In my ap­proach, however, newspaper language is not seen as a coherent variety. On the contrary, its internal variation is at the very centre of my interest. For this rea­son I do not compare my corpus of newspaper language to any kind of "norm", but I split it up into subsamples, which hopefully will show up the ex­pected variation. The fact that I am comparing not just two but a large number of potentially different styles necessitates a far larger corpus, and it makes it impossible - for practical reasons - to achieve the same level of delicacy in the analysis of the individual noun phrases.

MArdh ( 1980), too, uses the framework initiated by A arts ( 1971) and made more generally known by Quirk et a/. (1972, 1985) in part of her study on newspaper headlines. This application is particularly pertinent for my purpos­es because it is also devoted exclusively to British newspapers, and it com­pares two different categories of newspapers, one represented by the Daily Mirror and the other by The Times. For the comparison with Quirk et al.'s data, she analyses 2222 noun phrases of verbal headlines in The Times, and 530 in the Daily Mirror. She presents the results in two different tables ac­cording to the Quirk et a/. (1972: 933) format. I have transformed her data into one table corresponding to the more explicit Quirk et al. (1985: 1351) for­mat, which I have used above.

From this data (see table 5.5) the following facts emerge: There are fewer simple noun phrases in MArdh' s headlines than in Quirk eta/.' s aggregate cor­pus, but in subject position there are actually more complex noun phrases in her data than in theirs. And in The Times there are even more complex noun phrases in subject position than in the Daily Mirror. This, according to Ml\rdh ( 1980: Ill), "indicates that initial position, which is generally the position of the subject, is attractive for the headline writer .... The sooner the reader's in­terest is caught the better." In non-subject position, too, there are more com­plex noun phrases in the headlines, even though in this instance the difference is much smaller.

As I have pointed out above, the proportion of pronouns and names in sub­ject position varies very considerably across Quirk et al.'s four styles. They constitute a very large proportion in informal speech and in fiction (87. 7 and 79.9 per cent respectively), and a comparatively small proportion in scientific writing (39.3 per cent). In the headlines of The Times, pronouns and names in subject position constitute a smaller proportion than in scientific writing (23.4 per cent) and in the Daily Mirror a larger proportion (49.8 per cent). Not sur­prisingly the number of pronouns is very much smaller in the headlines than in Quirk et al. 's corpus. Pronouns often depend on a previous mention of the ref­erent. This is not normally possible in headlines. On this basis it may actually

Page 19: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Complexity as style marker lO I

surprise that as many as 19.3 per cent of all subject noun phrases of the Daily Mirror are pronouns as compared to 1.1 per cent in The Times.

Raumolin-Brunberg ( 1991) is one of the most recent and most detailed studies of English noun phrase structures. Her corpus consists of Early Mod­em English, but she does not use it diachronicly. She concentrates on the writ­ings of Sir Thomas More (1478-1535). The first part of her excellent study is devoted to a detailed description of all the noun phrase structures attested in her corpus. In the second part she explores the noun phrase variability across the different text types of her corpus.

Her corpus, too, shows some intertextual variation but it is not as dramatic as Quirk et al.'s . It is again the most informal text samples which show the highest percentage of simple nouns. There are 78.4 per cent of simple nouns in a historical text that contains many passages of direct speech. More's private letters are another sample with a high percentage of simple nouns (73.9%). At the other end of the scale there are his polemics (Apology) (58.2%) and his prayers (42.5%) (Raumolin-Brunberg 1991: 123 and 303). But the variation does not always follow the genre distribution. Texts of the same genre are not necessarily at the same end of this scale.

She also tests the connection between simple noun phrases and subject po­sition on the one hand, and between complex noun phrases and nonsubject po­sition on the other. Table 5.6 summarises these results for her corpus as a whole.

Table 5.6. Noun phrase structure in the writings of Sir Thomas More (1478-1535)

Total all simple names & all multiple NPs pronouns complex modification

NPs

Sir Thomas subject 969 823 146 More % 36.55 31.04 5.51

not subject 1682 892 790 % 63.45 33.65 29.80 total 2651 1715 936 % 100.00 64.69 35.31

Adapted from Raumolin-Brunberg (1991: 118, figure 6.3).

The percentages of simple noun phrases in subject position (31.04) and of complex noun phrases in nonsubject position (29.80) are only marginally dif­ferent. This is an interesting contrast to Aarts' (1971) results given as Quirk et a/.'s table above. However, her figures may not be directly comparable to Quirk eta/. 's because it is not quite clear how those noun phrases were count-

Page 20: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

102 Noun phrases as style markers

ed that is to say it is not clear whether noun phrases that are themselves modi­fiers are included in Quirk et al.'s figures or not (cf. Raumolin-Brunberg 1991: 68-70).

However, the association of simple noun phrases with subject position and complex with nonsubject position does not show any stratifying power at all in her corpus. With the exception of the prayers all samples have roughly 50 per cent of simple noun phrases in subject position.

In spite of the fact that her study deals with Early Modem English data, Raumolin-Brunberg's research has got many things in common with mine. She also uses the noun phrase as a style marker in order to explore the stylistic differences between texts of a fairly coherent and restricted corpus in her case the writings of one single man, and in mine a small segment of British media language. However, there are also considerable differences. Her emphasis lies on a detailed description of the attested noun phrase structures, which she takes to be representative not just for Sir Thomas More's writing but more generally for Early Modem English. The analysis of noun phrase complexity as a style marker, on the other hand, takes second rank.

5.5. Conclusions and hypotheses

At this point it seems advisable to survey what has emerged from the review of the approaches undertaken in this and the previous chapter. On this basis it will be possible to formulate precise hypotheses about the variation that can be expected to exist in my data of newspaper language.

The review of the studies by Romaine (1982), Cheshire (1982), and Allen­berg (1982) has stressed more the shortcomings of their approaches than their undoubted merits. I have suggested that they are modelled too closely on a correlational methodology that had been developed for phonological variables. All three of them are very clear in the difference between phonological vari­ables and their own syntactic variables but nevertheless they chose paradig­matic variables that- in spite of all the denials- stand in some sort of "free" variation. The variation is "free" in as far as the different realisations that are compared share a common core of meaning. It is not "free" to the same extent that phonological variation might well be argued to be free because both lin­guistic constraints and non-linguistic considerations govern the choice of one variant in favour of the other(s).

However, the inventory of variables that suggest themselves for this kind of analysis seems to be severely limited. Beyond the relative pronoun and the

Page 21: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Conclusions and hypotheses l 03

genitive versus of-construction not very many plausible candidates suggest themselves, with the exception of noun phrase name appositions.

The noun phrase name appositions appear to be the most promising variable with a systematic variation across my corpus of British English newspaper lan­guage. For this reason I shall devote all of chapter 9 to this construction.

Chapters 6 to 8 will follow a more traditional stylistic approach. As a start­ing point I shall apply Quirk et al.'s methodology introduced above to my data, and then, on the basis of a cluster analysis, I will break up my corpus into six different styles, which will be discussed and extensively illustrated in chapters 7 and 8.

The following list enumerates the hypotheses that emerged in the course of the above reviews of the relevant literature and in the course of the discus­sion of the structural properties of noun phrases in the previous chapter. The first two points concern the ordering of the data. They are the conditions that must obtain if claims are to be made about the significance of individual variables.

• In aJl cases of interesting statistical differences, the three newspaper catego­ries will be linearly ordered, that is to say the up-market papers will have the highest or lowest average number, or the highest or lowest proportional share of some category, and vice versa for the down-market papers, while the mid-market papers will have a value lying between the two.

• It is implausible to expect the pattern to be perfect for every variable, but generally there will be little overlap between the three categories, that is to say for one variable all the individual values of the up-market papers will be higher, or lower, than all the corresponding values of the mid-market pa­pers, which in tum will be higher, or lower, than the same values of the down-market papers.

The first of these two points is an absolute condition. If it does not apply for any given variable, this variable cannot be said to stratify according to the newspaper categories. The second point, on the other hand, is not absolute. Variables can be claimed to stratify significantly even if there is some limited overlap between the categories. The decision whether some variable is rele­vant or not will ultimately be taken on the basis of statistical tests (chi-square or ANOV A). It is highly unlikely that the statistics indicate relevant variation in cases where the above conditions do not hold in the stated way.

The subsequent hypotheses will be stated for the extreme categories only. Thus a claim that the up-market papers have a higher value for some variable x always implies that the average value of this variable for all the up-market papers is higher than the average value for the mid-markets, which in tum is

Page 22: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

104 Noun phrases as style markers

higher than the average value for the down-markets. The references in brack­ets indicate where the reasoning for these hypotheses has been given

• The proportion of premodifiers versus postmodifiers will vary significantly, even though no hypothesis can be made as to the direction the variation will take. Postmodifiers are more explicit than premodifiers. It will depend on the degree to which papers try to be explicit and specific (which favours postmodifications) and the degree to which they can leave relations be­tween head and modifier unexpressed on the assumption that the readers will be able to infer them (which favours premodification). The proportion of premodification versus postmodification will also be influenced by the fact that postmodifiers are more complex than premodifiers (section 4.4 and 4.5).

• The up-markets have a higher share of adjectives in prehead position whereas the down-markets have a higher share of nouns in prehead position (section 4.4).

• In whatever way the proportion between premodification and postmodifica­tion turns out to vary, if in any at all, it can be expected that the up-market papers show a higher degree of complexity. Noun phrase complexity can be measured in various ways leading to the following subhypotheses: • The total number of modifiers will be higher in the up-market papers. • The number of noun phrases with concatenated modifiers will be higher

in the up-markets, and among the noun phrases with concatenated modi­fiers, the proportion of those with more than two modifiers will also be higher in the up-markets (sections 4.4 and 5.3).

• The number of modifier embeddings will likewise be higher, and the embeddings witJ be deeper in the up-market papers (section 4.4).

• The proportion of structurally complex modifiers will be higher in the up-market papers (section 4.4 ).

• The association of structurally light noun phrases with subject position and structurally heavy noun phrases with non-subject position wiJI be highest in the down-market papers (section 5.4).

• The proportion of all noun phrases that are pronouns and names wiJl be higher in the down-market papers (section 5.4).

• The up-market papers have a higher share of classifying, defining and objective premodifiers in prehead position whereas the down-markets have a higher share of describing, subjective and evaluative premodifi­ers, which are gradable and appear in central position (section 4.4).

Page 23: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative

Conclusions and hypotheses 105

The last hypothesis does not concern the complexity of the noun phrase as such but a construction that is often said to be typical of journalese (e.g. Quirk et al. 1985: 1305).

• The noun phrase name appositions promise to be a particularly significant stratifying feature of newspaper language. The format Opposition Leader Neil Kinnock will have a high share in the down-markets and a very small share in the up-markets (section 2.6.6 this hypothesis will be further eluci­dated below before it is empirically tested against my data in chapter 9).

All the above hypotheses are formulated in terms of variation across the three newspaper categories. In the same way it is expected that there is considerable variation across the five different newspaper sections. However, at this stage it is impossible to hypothesise on the direction which this variation will take. It is unclear which section will tum out to be the most or the least complex. The general procedure will therefore be to establish the relevant variables for the newspaper categories and then test the variation of the same variables across the sections within the newspaper categories and across the entire data. The difference is of course that for the newspaper categories the directionality of the variation is a precondition. If a variable turns out to be highest in the mid­market papers, it will not be considered to be relevant. For the newspaper sec­tions, on the other hand, the ordering of the samples will be considered to be a result rather than a precondition.

Page 24: 5. Noun phrases as style markers - · PDF filetive marker in a number of Middle Scots texts, written between 1530 and 1550. ... tactic position of the relative marker within the relative