Local Grammar and Register Variation: Explorations in Broadsheet and Tabloid Newspaper Discourse

Monika Bednarek, Department of Linguistics, University of Sydney1


While there is much linguistic research on news discourse, most studies (outside CDA) have focussed on the so-called ‘quality’ newspapers. In contrast, this study systematically compares the language of ten British ‘popular’ and ‘quality’ newspapers, on the basis of a 70,000 word comparable corpus. The comparison focuses specifically on what is variously called evaluation, appraisal or stance —- the linguistic expression of speaker opinion, applying Hunston and Sinclair’s (2000) corpus-based pattern approach to the analysis of evaluation. As these authors show, certain lexico-grammatical patterns are commonly associated with evaluation, and together make up what they call a ‘local grammar of evaluation’. An example of such a pattern is: it + link verb + adjective group + clause (e.g. It seemed important to trust her judgement). Evaluative adjective patterns of this kind are examined in the corpus, in order to find out if there are any differences between the two types of newspapers. It is suggested that the differences between broadsheet and tabloid publications lie less in the frequency of evaluative patterns than in the function of these patterns as well as in the types of adjectives used in them.

1. Introduction

The aim of this paper is to compare evaluative adjective patterns in British English broadsheet and tabloid newspapers. To illustrate what I mean by ‘evaluative adjective patterns’ let us look at a quotation from the Sydney Morning Herald’s Good Weekend magazine of April 29, 2006, which reads as follows:

There’s something chilling about the inevitability of collateral damage, which happens in every war, just or otherwise.

This is very clearly a statement of the writer’s negative opinion about, or evaluation of non-military casualties in wars. Of particular interest here is how this opinion is expressed, namely through the adjective pattern there + link verb (be) + something + adjective group + about, which is used to introduce the author’s evaluation. This is in fact a pattern commonly associated with evaluation (Hunston and Sinclair 2000:85), which becomes readily apparent by looking at a few examples from the 100-million-word British National Corpus (BNC):

there is something irritating about the way Phil sits there
there is something abnormal about their opiate receptors
there is something suspect about the property
there is something crude about sex
there is something special about complex things
there is something wonderful about make-believe
there is something different about him
there is something peculiar about Harley’s contract
there is something empty about such shots
there is something unusual about me
there is something bad about the way they look
there is something strange about Mr Hyde, something evil
there is something tragic about this
there is something odd about the expression imperator noster
there is something sexy about Sue Lawley in real life
there is something paradoxical about this aspiration

Adjectives such as irritating, abnormal, crude, tragic, odd etc that fit into the slot following something in this pattern can be described as inscribing (Martin and White 2005:61, Bednarek 2006a) evaluation, that is, the evaluation is directly expressed through attitudinal lexical items, rather than implied (invoked or evoked) by non-evaluative lexis. But even adjectives that do not normally inscribe evaluation gain a subjective-evaluative meaning in this pattern, for example nationality adjectives (Hunston and Sinclair 2000:85):

There is something very American about the National Archives collection of presidential libraries.

It is evaluative adjective patterns of this kind that are the topic of this paper, in particular the usage of such patterns in a specific register of English: newspaper discourse. I shall first give a brief introduction to the notion of evaluation, before outlining the corpus-based pattern approach to evaluation with regards to Hunston and Sinclair’s (2000) concept of a local grammar. Finally, I will compare evaluative patterns across so-called broadsheet and tabloid newspapers, with the help of a comparable corpus consisting of 100 ‘hard news’ stories from ten British newspapers2. In so doing, this study aims to contribute to (non-CDA) linguistic research on news discourse which so far has mostly focussed on quality newspapers (Bednarek 2006b).

2. The concept of evaluation

Evaluation is here defined as referring to the linguistic expression of speaker/writer opinion along a number of semantic dimensions or parameters. That is, evaluation can relate to judgments of entities/propositions as good or bad, important or unimportant, comprehensible or incomprehensible, likely or unlikely, genuine or fake, expected or unexpected etc (Bednarek 2006b, chapter 4), covering notions of affect (Besnier 1990), modality (Coates 1983, Perkins 1983, Palmer 1995) and evidentiality (Chafe and Nichols 1986, Johanson and Utas 2000).

Evaluation has increasingly fallen under the scrutiny of linguistic research in the past years (e.g. Hunston and Thompson 2000), with many studies in the field of English for Academic Purposes (EAP, e.g. Hunston 1994, Hyland 1999, Bondi and Mauranen 2003, Charles 2006). A framework for the analysis of speaker opinion has also been developed within Systemic Functional Linguistics (SFL) under the name of appraisal (with the three sub-categories of attitude, engagement and graduation) (e.g. Macken-Horarik and Martin 2003, Martin and White 2005), and there are many corpus-based analysis of stance (e.g. Biber et al. 1999, Conrad and Biber 2000)3. While analyses of evaluation in EAP are characterised by a number of different approaches (usually, but not exclusively corpus-based), studies of appraisal mostly use manual, close-reading analysis (exceptions are the corpus-based analyses of Miller 2002, and Kaltenbacher 2006). Stance analyses that work within Biber and his colleagues’ framework (distinguishing between attitudinal stance, epistemic stance and style stance) and which use large-scale corpora most often work with previously established lists of stance features, sometimes including grammatical frames. However, only a few of these approaches to evaluation look systematically at grammatical frames or patterns of evaluation, and those that do seem to focus mostly on academic discourse (e.g. Hewings and Hewings 2002, Hyland and Tse 2005, Römer 2005). In contrast, this paper applies the pattern-based notion of a local grammar of evaluation to the analysis of hard news reportage in British tabloid and broadsheet newspapers, with a focus on adjective patterns.

3. A local grammar of evaluation

What, then, is a local grammar? The concept of local grammar goes back to the finding that software for the grammatical analysis of corpora (parsers), which are based on traditional grammatical models, find it hard to deal with specific areas of language, and that there is still a lot left over (Hunston and Sinclair 2000:76). Hunston and Sinclair (2000), following Gross (1993) and Barnbrook and Sinclair (1995), have suggested the use of parsers that are based on local grammars for such cases. Local grammars are essentially descriptions of particular areas of language (rather than the language as a whole), such as dictionary definitions, newspaper headlines, the language of cause and effect or the language of evaluation. In other words, a local grammar describes one meaning only (Hunston 2002:178), even if meaning is defined rather broadly. Local grammars typically work with transparent category labels referring to functional categories that are characteristic for the area of language that is to be described (Butler 2004:158), e.g. Definiens, Definiendum for definitions, Evaluative category, Thing evaluated for evaluative language. Local grammars are then used to describe and analyse sentences. The resulting description is ‘functional’ in a different way from the tradition of functional grammars, and incorporates some valuable pragmatic parameters (Hunston and Sinclair 2000:79). (There is also some overlap with Fillmore and his colleagues’ FrameNet approach (Fillmore and Atkins 1992, http://framenet.icsi.berkeley.edu). For a comparison of the two approaches see Hunston 2003, Bednarek 2008). Areas of language that have been described with the help of local grammars include dictionary definitions (Barnbrook and Sinclair 1995), the language of difference (Woodward 2002, cited in Hunston 2003), the language of emotion (Hunston 2003, Bednarek 2008) and the language of evaluation (Hunston and Sinclair 2000).

Concerning the latter, Hunston and Sinclair (2000) propose that evaluative adjectives typically occur in patterns of a certain kind, patterns of the type mentioned at the beginning of this paper such as there + link verb + something + adjective group + about. Another example is the pattern it + link verb + adjective group + finite or non-finite clause. Here are some examples from my corpus (B = broadsheets, T = tabloids):

Evaluative category Thing evaluated
  it link verb adjective group finite or non-finite clause
B It would be difficult to find a worse build-up
B It is inconceivable that Diana would send Burrell a letter of that nature
T it was hard to imagine more despicable words
T it is hypocritical and scandalous that she has been treated in this way

In this pattern (or collection of several patterns (Hunston and Sinclair 2000:84)), the adjective group realises the kind of evaluation (Evaluative category) that is expressed by the speaker/writer, whereas the finite or non-finite clause realises the entity that is evaluated (Thing evaluated). We can now see what is meant by the usage of transparent category labels in local grammars, namely, labels that directly reflect the discourse functions of sentence elements. Hunston and Sinclair (2000) identify a large number of evaluative adjective patterns, suggesting that these make up a local grammar of evaluation (as far as adjectives are concerned). These patterns occur either exclusively or usually with evaluative or subjective adjectives and automated parsing would therefore be more or less successful, with problems occurring when patterns are shared with non-evaluative adjectives or when the parsing depends on the lexical item involved (Hunston and Sinclair 2000, Hunston 2003).

The pattern approach on which local grammar is based was essentially developed by Hunston and Francis (2000). Importantly, patterns concern relatively general grammatical ‘surface’ categories such as v (verb group), n (noun group) or adj (adjective group) (Hunston and Francis 2000:45). This also means that in the local grammar approach no distinction is made between different kinds of link verb (e.g. those involving evidential or modal meanings such as seem, appear and be). It would, of course, be possible to describe patterns in more detail, and to make a difference between ‘It SEEM adjective group clause’ and ‘It BE adjective group clause’ or to distinguish patterns with different types of evaluative adjectives (e.g. epistemic vs. attitudinal). The analyses reported on here could hence be taken as a starting point for a more detailed development of evaluative patterns and categories, which is beyond the scope of this paper (as is a discussion of the theory behind establishing patterns; see e.g. Hunston and Francis (2000), Hunston and Sinclair (2000), Hunston (2002, 2003) for further elaboration). In general, the concept of a local grammar of evaluation has hardly been applied so far in linguistics (but see Radighieri 2006 for an application to the analysis of the newspaper arts review), and I shall here use it in my analysis of evaluation in ‘popular’ and ‘quality’ newspaper discourse4.

4. Popular vs. quality newspapers

The starting point of the analysis of evaluative patterns in newspaper discourse was to find out if there are any differences in the use of evaluative patterns between the so-called popular or quality newspapers (broadsheets vs. tabloids), and if so, what these differences are, and whether they confirm our expectations about the popular press (in that we would expect them to be more evaluative and more explicitly so). In other words, should the two types of news discourse be seen as different varieties or registers of English in terms of a local grammar of evaluation?

4.1 The corpus

The corpus that was used as the basis of this comparison consists of 100 hard news stories from ten British national daily newspapers, covering the same ten topics, as visualised in table 1 (with the numbers referring to an automatic word count with Microsoft Word).


Table 1: The corpus
RIO 688  771  889  803  1088  968  927  832  1229  1244 
IDS 584  983  1204  1017  782  638  291  1011  475  689 
IRAQ 508  781  752  746  672  312  278  556  143  595 
HIV 239  575  521  479  494  815  423  892  1131  755 
BARCLAYS 269  538  831  369  642  282  334  710  467  614 
DI 340  1036  866  1362  1150  532  448  946  1200  1569 
IRA 487  810  847  1144  720  453  294  481  481  402 
POLICE 413  518  1350  906  771  801  338  836  487  676 
MADAM 539  791  639  1043  896  754  318  1659  221  814 
Total: 4.434  7.921  8.599  8.845  7.605  5.965  3.949  8.826  6.258  7.798 
  37.404  32.796 

The ‘quality’ newspaper sub-corpus consists of stories from The Financial Times, The Guardian, The Independent, The Times and The Daily Telegraph, and the ‘popular’ newspaper sub-corpus comprises stories from The Sun, The Star, The Daily Mail, The Daily Mirror and The Daily Express. The subject matter was kept constant in all ten newspapers in order to establish a truly comparable corpus, and to represent topics that are central to society as well as to avoid the influence of the topic on the analysis of evaluation. Here is a short summary of the topics represented in the corpus, which are relatively varied:

Topic (date)Content
Israel (6/10/2003)an Israeli attack on Syria and a suicide bombing in Israel.
Rio (8/10/2003)the decision of the Football Association to exclude Rio Ferdinand (then defender with Manchester United) from the English team for the 2004 European championships because Rio Ferdinand did not provide a sample to drug testers on September 23.
IDS (10/10/2003) Iain Duncan Smith’s (leader of the Conservative Party at that time) conference speech at the Conservative Party conference in Blackpool.
Iraq (13/10/2003) an attack on a hotel in Baghdad by a suicide car bomber in Iraq.
HIV (15/10/2003) the sentencing of Mohammed Dica, who was found guilty of infecting two women with the HIV virus.
Barclays (17/10/2003)Matthew Barrett’s (then chief executive of Barclays) utterances concerning the expensiveness of credit cards at a hearing with MPs.
Di (21/10/2003)a note written by Princess Diana to her former butler, Paul Burrell, published by him to coincide with the publication of his book, A Royal Duty. In the note she writes about someone planning an accident involving her car.
IRA (22/10/2003)the Northern Irish peace process and David Trimble’s (Ulster Unionist leader at that time) rejection of the destruction of weapons by the IRA because not enough details were given about this destruction by General John de Chastelain (then leader of the independent body overseeing arms decommissioning).
Police (23/10/2003)racist utterances by police recruits that were pronounced on a documentary made by Mark Daly, an undercover reporter for the BBC.
Madam (24/10/2003)the sentencing of Margaret MacDonald, a British woman who was convicted of prostitution by a court in Paris, and sentenced to four years in prison and a €150,000 (£104,000) fine.

Admittedly, the corpus is rather small, consisting of only 70,000 words. This is for two reasons: firstly, the corpus had to be newly designed for reasons outlined below (involving the manual keying in of many stories, in particular from the tabloids), and, secondly, the corpus was originally designed for a manual discourse analysis (Bednarek 2006b). The frequency of evaluative adjective patterns is still relatively high in most cases (see table 2 below). As becomes apparent in table 1 above, the corpus is also more or less balanced, but the occurrences will nevertheless be normalised to a frequency per 10,000 words.

Furthermore, there are (to my knowledge) no existing large-scale corpora that would have been suitable for the exact purposes of this study. Corpora such as LOB/FLOB contain only a mixed category (A) called press: reportage which does not distinguish between daily and Sunday newspapers, between regional and national newspapers or between the categories of political news, sports news, society news, spot news, financial news and cultural news (see ICAME Manual). Moreover, these corpora include samples rather than complete texts, and in total, category A consists of only 44 texts of roughly 2,000 words each, i.e. only marginally more than my own corpus. Other potentially suitable corpora like the British National Corpus, the Bank of English, the Reuters corpus, the Rostock Historical English Newspaper Corpus, the Longman Grammar of Spoken and Written English (LGSWE) corpus, and the Zurich English Newspaper (ZEN) corpus either consist of samples rather than complete texts, or do not include all of the ten national newspapers considered in this paper, are diachronic rather than synchronic, or do not distinguish between the category of hard news and other categories. The only other alternative, to design a large-scale corpus of broadsheet and tabloid hard news reportage myself, would have been far too time-consuming (although broadsheets are obtainable on CD-ROM, the tabloids are not).

Even though the conclusions may not always be statistically definite, and wholly representative of British newspaper discourse as a whole, I strongly believe that the analysis of the corpus does indicate certain trends in this variety of English (and I will therefore also report the results of applying significance tests to the data). Nevertheless, the analysis is perhaps best be considered as a pilot investigation, to be complemented by future, large-scale corpus research.

For the analysis of the corpus I used Wordsmith (Scott 1999), in particular the concordancer (Concord) that lists all occurrences of certain words or phrases in the corpus. I used Concord to look for the link verbs be, look, appear, seem, remain, leave and the general nouns thing, point, kind and sort (for reasons that will become apparent later). With respect to these lexical items I searched for all evaluative adjective patterns identified by Hunston and Sinclair (2000), apart from pseudo-clefts (because of time-constraints) and attributive adjective patterns (because my corpus was not parsed). (All non-evaluative adjectives occurring with these link verbs and general nouns as well as some modal adjectives/past participles (able to, bound to, allowed to, set to, liable to, eligible for, due to, entitled to, about to, capable of) were excluded from the analysis, though it must be pointed out that there is gradedness between non-evaluative and evaluative adjectives5.

4.2 The patterns

The patterns that were analysed can be grouped into eight pattern groups. Apart from using the patterns identified by Hunston and Sinclair (2000) I have added a few new realizations and some new patterns found in my data. I will at first look at patterns in general in this section6, before shifting the focus to particular adjective categories in 4.3. As will be seen later, the latter are in fact more important than the former for distinguishing broadsheets from tabloids.

Patterns 1 and 2: Patterns with anticipatory it and there

Since patterns 1 and 2 have already been discussed above, they will only be listed here with examples from the corpus. They involve anticipatory it and there preceding the adjective group:

Pattern 1: it + link verb + adjective group + finite/non-finite clause
  Evaluative categoryThing evaluated
 itlink verb adjective groupfinite or non-finite clause
BItwould bedifficultto find a worse build-up
BItisinconceivablethat Diana would send Burrell a letter of that nature

Pattern 2: there + link verb + something/anything/nothing + adj group + about/in + noun group/-ing clause
 HingeEvaluative categoryHingeThing evaluated
 therelink verb something/ anything/ nothingadjective groupabout/ in/ withnoun group or ing clause or wh- clause
BThereisnothingiffyaboutwhat happened to her
TThere’snothingwrongwithgoing out shopping

Pattern 3: Patterns with link verb and to-infinitive clause

Patterns in group 3 all include the pattern ‘noun group + link verb + adjective group + to-infinitive clause’, with different functional mappings (the to-infinitive clause realizing a Restriction or a Thing evaluated, the adjective group realizing an Evaluative category or an Evaluating response):

Pattern 3: (i): Thing evaluated + Evaluative category + Restriction
 Thing evaluatedHingeEvaluative categoryRestriction on evaluation
 noun grouplink verbadjective groupto-infinitive clause
Tthey (the police)will bepowerlessto stop the woman they have dubbed La Madame Anglaise

Pattern 3: (ii): Evaluation carrier + Evaluative category + Restriction
 Evaluation carrierHingeEvaluative categoryThing evaluated
 noun grouplink verbadjective groupto-infinitive clause
BMr Blunkettwaswrongto attack [the documentary]
THe (Dica)islikelyto receive high-quality medical care

Pattern 3: (iii): Evaluator + Evaluating response + Thing evaluated
 EvaluatorHingeEvaluating responseThing evaluated
 noun grouplink verbadjective groupto-infinitive clause
BMark Palios, the FA’s new chief executiveisanxiousto set football a new disciplinary agenda
Thewasproudto be a cop

Pattern 4: Patterns with link verb and that-clause

Pattern 4 consists of the evaluative pattern ‘noun group + link verb + adjective group + that clause’, with the noun group realizing the Evaluator and the that-clause realizing the Thing evaluated7:

Pattern 4: link verb + adjective group + that-clause
 EvaluatorHingeEvaluating responseThing evaluated
 noun grouplink verbadjective groupthat-clause
BIamdisappointedthat it has taken Burrell six years to reveal this extraordinary
BThe FAisconfidentit can defend any action under the strict rules governing doping
BThe Football Associationwasadamantthat it was not prejudging Ferdinand
TIamgladhe didn’t get away with it
TDetectivesareconvincedthe evil romeo passed the illness to other women too

Pattern 5: Patterns with general nouns

Patterns in group 5 all concern general nouns. These are all new patterns: none of the patterns identified by Hunston and Sinclair (2000:90-91) could be found in the corpus. Why should such patterns involving nouns, rather than adjectives be included here? This is because saying this is a bad thing or this is the most important thing is very much like saying this is bad or this is the most important (with differences in endophoric emphasis). Three patterns with general nouns were found:

Pattern 5: (i): noun group + link verb + noun group with thing
 Thing evaluatedHingeEvaluative category
 noun grouplink verbnoun group with thing
Bthatis(not necessarily) a bad thing
Bitisthe last thing we need
Tthatisthe most important thing

In pattern 5:i the first noun group realises the Thing evaluated and the second noun group (with thing) expresses the type of evaluation that is expressed on the part of the speaker.

Pattern 5: (ii): noun group with thing + PP + link verb + noun group
 Evaluative categoryEvaluating contextHingeThing evaluated
 noun group with thingPPlink verbnoun group
Tthe only thing extravagantabout himwasthe cocktail of lies he showered me with every day

Pattern 5:ii is particularly interesting because it seems very similar to the pseudo-cleft adjective patterns identified by Hunston and Sinclair (2000:89):

What’s very good about this play is that it broadens people’s view
What’s interesting is the tone of the statement.

They note:

In the first example, it is debatable whether what is evaluated as good is the play itself or the fact that ‘it broadens people’s view’. We regard the second of these alternatives as being the case, on the grounds that if the second example were altered to read What’s interesting about this statement is its tone, the thing evaluated would still be ‘the tone’ rather than ‘the statement’. We therefore introduce the term evaluative context to classify the prepositional phrase beginning with about. (Hunston and Sinclair 2000:89-90)

Arguably, the same is the case with pattern 5 (ii): what is evaluated is the ‘cocktail of lies’ rather than ‘him’, since we could alter the example to read What was extravagant about him was the cocktail of lies he showered me with every day, turning it from a general noun pattern to an adjective pattern (but losing the introduction of only in this rephrasing).

Finally, pattern 5 (iii) is perhaps on the borderline of being a ‘true’ pattern, involving the idiomatic phrase DO the decent thing as realizing the Evaluative category. Only three examples of this pattern were found in the corpus:

Pattern 5: (iii): do the decent thing
 Evaluation carrier(Hinge)Evaluative categoryHingeThing evaluated
 noun group(‘volition’)DO the decent thing‘elaborator’finite or non-finite clause
BHewon’tdo the decent thingandresign
BHeshoulddo the decent thingandresign
THeshoulddo the decent thingandresign

However, examples of this pattern can also be extracted from the BNC (with the ‘volition’ slot not always expressed, and the ‘elaborator’ realised by different linguistic structures including punctuation):

He did the decent thing and dropped them from his speech
the fishmonger has done the decent thing — removed the guts and gills
Robert had done the decent thing. Given Anwar alpha double plus…
We called upon Frank Gray to do the decent thing ie resign
a chap will do the decent thing that is bluff it in a Brighton hotel

There is also a variation of this pattern in the BNC with the Evaluated thing expressed by a PP with in: he had done the decent thing in abandoning Suzi (BNC).

This Evaluative category may be bound specifically to British culture, as suggested by an example from the BNC:

If you meet people from other countries, they understand what you mean if you say to do the legal thing, or the just thing, or the right thing, in certain ways, but it seems to me to be a slightly soggy, but nevertheless very important, British concept of doing the decent thing, which may be just, may be legal, may not be either of those two, but the British man has a very clear sense of what it entails.

However, a web search of Irish, Australian, Canadian and Indian websites also came up with hits for DO the decent thing, though admittedly all of these countries have a strong association with Great Britain in their past. More cross-linguistic research would be necessary to investigate the cultural specificity of this evaluative category in more detail.

Pattern 6: Adjective complementation

Patterns in group 6 concern adjective complementation, which is typical for evaluative adjectives (Hunston and Sinclair 2000:96). There are two types of pattern here:

Pattern 6: (i) Evaluator + Evaluating response + Thing evaluated
 EvaluatorHingeEvaluating responseThing evaluated
 noun grouplink verbadjective groupPP
BIamsurprised and saddenedat the behaviour and comments
BThe clubisdeeply troubledby the FA’s handling of the situation
Bweareopposedto any use or threat of force
Bshewashappywith the situation
TThe MPsarefuriousabout various ruses
TMacDonaldlookeddistraughtat the verdict
TStevenswasappalledby the documentary
Tshewasconvincedof the plot to mastermind an accident
Ttheywerefuriouswith the judgment

Pattern 6: (ii): Thing evaluated + Evaluative Category + Restriction
 Thing evaluatedHingeEvaluative CategoryRestriction on evaluation
 noun grouplink verbadjective groupPP
Bthe timeisrightfor a full public inquiry
Bheisguiltyof a dreadful lapse of professionalism
Bwhose viewsarevitalto modern Tory kingmaking
Tthe decisioniswrongfor the player and for the country
TItisso unfair and hardon her
Tthese notesarecrucialto the truths that enshrine her memory

Pattern 7: Graded adjectives

A number of patterns (group 7) concern graded adjectives, i.e. adjectives with too or enough (not good enough, important enough), comparative (larger, cheaper) or superlative (the most corrupt) adjective groups. Since there are so many variations of these patterns (and since most of these are discussed in Hunston and Sinclair 2000), I have listed the proposed parsing of such patterns in the appendix and will only give some examples from the corpus here:

With too/enough:

This is not good enough to allow people into the government of Northern Ireland (B)
…its contents were important enough to merit re-opening their investigation (T)
…one witness was too scared to give his name (B)
…the game against Turkey will be difficult enough (B)
…credit cards are too expensive (T)

Comparative adjective groups:

…an act of putting arms beyond use that was larger than the first two (B)
…which is cheaper than anything he can get through his own company (T)
…it is better to try to find a way through rather than walk away (T)
…the bookies remained less impressed (B)
…things were looking much brighter (T)
The material could not have been more damning and disturbing (B)

Superlative adjective groups:

…the Blair Government being the most corrupt, dishonest and incompetent of modern times (B)
…this particular phase is most dangerous (B)

However, one pattern involving comparison will be commented on in more detail: this is a new pattern that was found only once in my corpus, but there are also examples in the BNC:

…it would be something as huge as this (B)
…this can be something as simple as turning the back on the past (BNC)
…it is something as simple as beans on toast or a sandwich (BNC)

These examples can be parsed as:

 Thing evaluated 1HingeHingeEvaluative categoryHingeThing evaluated 2
 noun grouplink verbsomething asadjective groupasnoun group/-ing clause
Bit(would) besomething ashugeasthis

In general, it seems as if the pattern ‘something as + adjective group + as + noun group’ is linked to evaluation. Here are some examples of adjectives from the BNC that occur in the adjective slot of this pattern: arbitrary (2), bad, basic (5), physical, beautiful (2), big (3), bizarre, complex (4), cruel, cuddly, curious (2), daft, desirable, dismal, drastic, dry, easy and fluid, eccentric, effervescent and ethereal, elemental, ephemeral, evocative, exceptional, exciting, extreme, fatuous, frightening, fundamental, general, good, gruelling, hard and painful, horribly inane, important (8), innocent, minor, natty and camp, necessary (2), obvious, ordinary, petty, prosaic, ridiculous (2), risky, sensitive (2), serious, simple (22), soft and comfortable, solemn, sophisticated, special (2), crazy, straightforward (2), strange (3), trivial, unappetizing and unpromising, undignified, vital, wonderful.

Pattern 8: Predicative adjective patterns

Finally, patterns in the last group (8) concern predicative adjective patterns: ‘noun group + link verb + adjective group’. In some cases the Evaluator can be added with a PP (8:ii) and the two main patterns in this group (8:i and 8:iii) are distinguished by their semantic mapping:

Pattern 8: (i): Thing evaluated + Evaluative Category
 Thing evaluatedHingeEvaluative Category
 noun group/clauselink verbadjective group
Bthe whole choreographyisfarcical
Tthe womanisbrilliant

Pattern 8: (ii): Thing evaluated + Evaluative Category + Evaluator
 Thing evaluatedHingeEvaluative CategoryEvaluator
 noun grouplink verbadjective groupPP with to
Ba game whichisso importantnot only to myself, but also the team and the England fans
Ta game whichisso importantnot only to myself, but also the team and the England fans

Pattern 8: (iii): Evaluator + Evaluating response
 EvaluatorHingeEvaluating response
 noun grouplink verbadjective group
TIamabsolutely devastated

4.3 Evaluative category vs. Evaluating response

With many patterns, two main types of semantic mapping are possible, those including an Evaluative category and those including an Evaluating response:

Horses [Thing evaluated] are pretty [Evaluative category] to look at [Restriction]

Benjamin [Evaluator] had been rather overawed [Evaluating Response] to meet one of the Billington family [Thing evaluated]
(Hunston and Sinclair 2000:87-88)

In the first example, the adjective indicates a quality of the thing that is evaluated by the speaker, whereas in the second example, the adjective implies a personal emotional or mental reaction of someone who may or may not be the speaker. Hunston and Sinclair primarily relate these two categories to the difference between attributed and averred evaluation (2000:87-88, 97), but it must be pointed out that both patterns can be used for attributing and averring opinion. Patterns with evaluative categories can be quoted (He said: “Horses are pretty to look at” / He said that horses are pretty to look at etc) and patterns with evaluating responses are averred if the Evaluator is the Self (I had been rather overawed to meet one of the Billington family). The distinction seems rather one between ‘evaluation proper’ and ‘affect proper’ (or between the parameters of emotivity and mental state in terms of the framework of Bednarek 2006b).

Additionally, an intermediate type of category can be found:

…it was very frustrating (B)
…to be left out of the squad at this crucial stage is devastating (T)

These examples share similarities with Evaluative categories in that the noun group or clause at the beginning of the pattern realises the Evaluated Thing (rather than the Evaluator) but are similar to Evaluating responses in that the adjective (frustrating, devastating) indicates an emotional or mental response rather than an evaluative quality. (Note also that it is examples of this kind that present problems to appraisal analyses of appreciation: reaction and affect (Martin and White 2005:57-58). Hunston (2003) discusses the difference between Everyone in the school is distressed to hear of this tragedy and …after the distressing events of 1887 in terms of a distinction between ‘reflective’ and ‘constitutive’ affect, arguing that adjectives such as distressing do indicate a quality of the evaluated entity but in treating them both as affect recognises the emotional response meaning present in such adjectives.) To sum up, the distinction between the semantic mapping involving Evaluative Categories and those involving Evaluating Responses seems valid and relatively easy to apply (see table 2 below), whereas the status of an intermediate category is more problematic, and will not be discussed further here.

4.4 The results

The results of the analysis of all the above-mentioned evaluative adjective patterns are listed in table 2:

Table 2: The Results
Clause typeBroadsheetsTabloids
 raw freqper 10,000raw freqper 10,000
pattern 1: it + link verb + adjective group + finite/non-finite clause
to-infinitive14 6 
that-clause10 5 
whether/if-clause 2 
pattern 2: there + link verb + something/anything/nothing + adj group + about/in + noun group/-ing clause
nothing1 1 
pattern 3: link verb + adjective group + to-infinitive clause
(i) Evaluative category I (restriction) 2 
(ii) Evaluative category II (carrier)11 6 
(iii) Evaluating response10 7 
pattern 4: link verb + adjective group + that-clause
(i) Evaluating response16 15 
(ii) Evaluative category  
pattern 5: general nouns
(i) Evaluative category I2 1 
(ii) Evaluative category II (context) 1 
(iii) Doing the decent thing3 2 
pattern 6: link verb + adjective group + PP
(i) Evaluating response24 38 
(ii) Evaluative category15 7 
pattern 7: graded adjectives8
(i) too/enough (category)5 1 
(ii) too/enough (response)1  
too/enough (unrestricted)10 16 
(iii) superlative1  
superlative (unrestricted)1  
(iv) Comparative2 2 
Comparative (p. of reference implicit)7 1 
Comparative (it link verb to inf clause) 2 
Comparative (other)1  
(v) Negative (p. of reference implicit)1  
(vi) Something as adj group as1  
pattern 8: predicative adjectives
(i) Evaluative category (+/- evaluator)58 48 
(ii) Evaluating response24 29 
(iii) Eval. category/response7 10 
Grand total22760.720261.6

If we chi-square the totals of all formally-defined patterns (where total > 5 in each of the sub-corpora), it becomes evident that the differences between the two sub-corpora are in fact neither large nor statistically significant. In terms of the question asked at the beginning of this paper (whether broadsheet and tabloid newspapers are different varieties in terms of a local grammar of evaluation), the answer must thus be no: they both seem to use the same range of formally defined evaluative adjective patterns. For example, the pattern ‘link verb + adjective group + that-clause’ occurs 4.3 per 10,000 words in the broadsheets, and 4.6 per 10,000 words in the tabloids; complementation patterns occur 10.3 per 10,000 words in the broadsheet corpus compared to 13.7 per 10,000 words in the tabloid corpus. This similarity may perhaps be explained by the common discourse function of hard news stories and by the fact that much news discourse is recycled talk (Bell 1991), i.e. makes reference to the same sources.

However, if we compare the difference between those patterns that involve Evaluative Categories and those involving Evaluating Responses, and chi-square those results, the differences here are statistically significant:

 raw freqper 10,000raw freqper 10,000
Evaluative category116318124.7
Evaluating response (emotional/mental)7419.88927.1

In other words, the tabloid newspapers prefer to use evaluative adjective patterns for reporting emotional or mental states, whereas broadsheet newspapers prefer to use them for evaluation proper. This confirms our expectations about the ‘popular’ and the ‘quality’ press to a certain extent.

Moving beyond pattern analysis to a brief analysis of some of the lexis involved, further differences become apparent. Taking into account only adjective types (not tokens) we find that a number of adjectives occur only in one of the sub-corpora:

Occurring only in broadsheet sub-corpusOccurring only in tabloid sub-corpus
62 ‘evaluative’ adjectives32 ‘evaluative’ adjectives
14 emotional/mental adjectives31 emotional/mental
2 evaluative/emotional1 evaluative/emotional

Let us compare these with respect to three factors that we might expect to be of relevance concerning the difference between quality and popular news stories:

The results seem to suggest that tabloid newspapers use more ‘intense’ and ‘metaphorical’ adjectives than the broadsheets, whereas the findings for formality are inconclusive (based on the small amount of data available). The assumption that tabloids use more informal, casual language than the broadsheets (Bednarek 2006b) is not confirmed as far as the evaluative adjectives in the patterns analysed here are concerned:

‘intense’anxious, jubilant, baffleddelighted, distraught, euphoric, incandescent, incensed, livid, puzzled
metaphoricaldead inside, destroyed, shattered, (incandescent, incensed)
informalfed-up with, stressedcool with
formalincandescent, saddened

However, it must be pointed out that this analysis of lexis constitutes a pilot study which needs to be followed up by more systematic research based on bigger comparable corpora.

More research into patterns that involve nouns and verbs is also necessary to complement the studies reported on in this paper (see e.g. Hyland and Tse (2005) on evaluative that-structures following verbs and nouns and Hewings and Hewings (2002) on verb patterns with anticipatory it). Furthermore, the distinction between types of adjectives should be further developed. For instance, adjectives labelled ‘evaluating response’ could be sub-divided according to the affect types proposed by Martin and White (2005) and/or according to the types of mental states suggested by Bednarek (2006b). Adjectives realizing ‘evaluative categories’ could be described in more detail according to suggestions e.g. by Hunston and Sinclair (2000), Hewings and Hewings (2002) or Bednarek (2006b) with the main distinction presumably between attitudinal and epistemic evaluative stance (Hyland and Tse 2005:47). Whereas epistemic evaluative adjectives express notions such as certainty, doubt and evidentiality, it has been proposed that attitudinal adjectives can refer to evaluative categories such as comprehensibility, emotivity (‘good-bad’), importance, expectedness, genuineness, and necessity (see Bednarek 2006b for an overview).

5. Concluding remarks

To sum up, broadsheet and tabloid newspapers are distinguished more by the function of evaluative adjective patterns and the types of adjectives used in these patterns than by the frequency of the patterns themselves. The analysis has demonstrated that local grammars, as suggested by Hunston can fruitfully be used to [quantify] ways of expressing meanings in different registers (Hunston 2002:178). Whereas the proposed analyses by Hunston (2002:180-181) only compare the frequency of patterns (e.g. patterns with propositions or verb-pattern combinations) in two different corpora, the real advantage of such an approach seems to lie in the fact that the resulting frequencies of patterns can be related to the meanings or functions of these patterns. Such descriptions are also potentially interesting for language teaching purposes. Patterns, as Hunston (2002:173-174) notes, can be taught to improve both fluency and accuracy. The particular advantage of a local grammar approach to patterns lies in the transparent category labels of local grammars, i.e. semantic labels such as Evaluator, Evaluative Category and Evaluating Response that directly reflect the discourse function of sentence elements (which the traditional categories of subject, verb etc do not) (Hunston and Sinclair 2000:79). Such transparent category labels are arguably very ‘user-friendly’. Local grammars also provide a helpful organisation of one particular area of meaning (e.g. evaluation, emotion, difference). The results of the findings above can also provide the basis for introducing more text work on the basis of ‘popular’ publications into the English classroom, which would offer two key advantages: the topics that are discussed might appeal to the students to a greater extent, and it allows for an instruction in critical media literacy (Fairclough 1995:201) concerning the overt expression of bias, and the variation between different types of (news) discourse.


Patterns with graded adjectives

Pattern 7: (i): Evaluative category
 Thing evaluatedHingeEvaluative categoryRestriction on evaluation
 noun grouplink verbadjective group with too or enoughto-infinitive or PP with for
BThisisnot good enoughto allow people into the government of Northern Ireland
Tits contentswereimportant enoughto merit re-opening their investigation

Pattern 7: (ii): Evaluative response
 EvaluatorHingeEvaluative responseRestriction on response
 noun grouplink verbadjective group with tooto-infinitive
Bone witnesswastoo scaredto give his name

Without Restriction:
 Thing evaluatedHingeEvaluative category
 noun grouplink verbadjective group with too or enough
Btheyaretoo dear
Bthe game against Turkeywill bedifficult enough
Tcredit cardsaretoo expensive

Pattern 7: (iii): Superlatives
 Thing evaluatedHingeEvaluative categoryRestriction on evaluation
 noun grouplink verbsuperlative adjective groupPP
Bthe Blair Governmentbeingthe most corrupt, dishonest and incompetentof modern times

Without Restriction:
 Thing evaluatedHingeEvaluative category
 noun grouplink verbsuperlative adjective group
Bthis particular phaseismost dangerous

Pattern 7: (iv) Comparatives
 Thing evaluated 1HingeEvaluative categoryHingeThing evaluated 2
 noun grouplink verbcomparative adjective groupas or thanNG/PP/AdvG
Ban act of putting arms beyond use thatwaslargerthanthe first two
Twhichischeaperthananything he can get through his own company

 Evaluative categoryEvaluated thing 1HingeEvaluated thing 2
 itlink verbcomparative adjective groupto-infinitive clausethannon-finite clause
Titisbetterto try to find a way throughrather thanwalk away

 Thing evaluatedHingeEvaluative category
 noun grouplink verbcomparative adjective group
Bthe bookiesremainedless impressed
Tthingswere lookingmuch brighter

Pattern 7: (v) Comparatives with negative Hinge
HingeEvaluative categoryHingeThing evaluatedRestriction on evaluation
negativecomparative adj groupthannoun groupto-infinitive clause
There’s nothingbetterthannatural lightto bring out the colour of paintings

Without Restriction:
 HingeEvaluative category
 negativecomparative adj group
BThe material could not have beenmore damning and disturbing

  1. The studies reported in this paper were first presented at the linguistics research seminar at the University of Sydney, and at the Australian Association of Applied Linguistics conference in Brisbane in 2006. I wish to thank both audiences for their helpful comments. I also want to express my gratitude to Wolfram Bublitz and an anonymous reviewer for their valuable comments on an earlier version of this paper. 

  2. Hard news concerns reports of accidents, conflicts, crimes, announcements, discoveries and other events which have occurred or come to light since the previous issue of [a] paper (Bell 1991:14). 

  3. To this we might add studies of ideology in Critical Discourse Analysis (CDA, e.g. Fairclough 1995), which often deals with issues similar to evaluation. 

  4. I agree with Jucker (1992:47) that the terms popular and quality are problematic because they imply certain value judgements about these types of newspapers. I nevertheless employ these terms simply to have readily available synonyms for the broadsheets and the tabloids. Strictly speaking the term tabloid refers to the small format of the respective newspapers. However, nowadays even some quality papers have adopted the tabloid format (The Independent, The Times) so that this strict definition does not apply any more. In this extended sense, tabloid rather refers to a type of newspaper that is characterised by a variety of features such as particular lexis, layout, target audience, etc. 

  5. More detailed comments on the methodological decisions made in the analysis of the data can be downloaded from http://www.monikabednarek.com/10.html

  6. The category labels used to describe these patterns are explained in detail by Hunston and Sinclair (2000), so a few comments on them shall suffice:

    Evaluator:the person responsible for the evaluation
    Thing evaluated:what is evaluated
    Evaluative category:the type of evaluation that is expressed
    Evaluating response:a personal response to the Evaluated Thing
    Evaluation carrier:a noun group that is not directly evaluated but does carry some evaluation
    Restriction on evaluation:what the evaluation relates to
    Hinge:providing connections between parts of patterns

  7. Hunston and Sinclair also note an additional variant involving patterns with a that-clause (They were lucky that we scored when we did), which, however, did not occur in the corpus data. 

  8. With the patterns for graded adjectives, all adjectives were counted in the analysis, even those that do not primarily inscribe evaluation (large). This is because comparison can in itself be regarded as subjective and evaluative (Thompson and Hunston 2000:18): gradedness indicates comparison, and comparison with a norm or scale is often a matter of subjectivity (Hunston and Sinclair 2000:92). Most adjectives that occur in these patterns are evaluative in any case; only a few are not (large, long, late). 


Monika Bednarek (2007) “Local Grammar and Register Variation: Explorations in Broadsheet and Tabloid Newspaper Discourse”, ELR Journal, 1 (1).