3.2.1 Studies of particular grammatical features in academic writing

Previous linguistic analyses of academic writing have considered a wide range of individual grammatical features. Many of these are specialized features used for information packaging, such as:

• – shell nouns and associated grammatical structures (Aktas and Cortes Reference Aktas and Cortes2008; Gray and Cortes 2011; Gray Reference Gray2010; Charles Reference Charles2003)

• – passive voice (Tarone, Dwyer, Gillette and Icke Reference Tarone, Dwyer, Gillette and Icke1998; Baratta Reference Baratta2009)

• – existential there (Huckin, Pesante, and Hutz Reference Huckin and Pesante1988)

• – extraposed constructions with anticipatory it (Groom Reference Groom2005; Hewings and Hewings Reference Hewings2001)

• – conditional clauses (Warchal Reference Warchal2010; Ferguson Reference Ferguson2001)

• – citations (Hewings, Lillis, and Vladimirou Reference Hewings, Lillis and Vladimirou2010; Hyland Reference Hyland1999; Charles Reference Charles2006a,Reference Charlesb)

Surprisingly, some conversational grammatical features are also employed in academic writing for information packaging functions. For example, Swales et al. (Reference Swales, Ahmad, Chang, Chavez, Dressen and Seymour1998) and Hyland (Reference Hyland2002a) describe the use of imperatives/commands in academic writing to focus the reader’s attention on points of special importance, and Webber (Reference Webber1994) similarly describes the special functions of questions in written academic texts.

One area of research that has been especially prevalent in discussions of academic writing is the lexical and grammatical marking of stance and evaluation (e.g., Baratta Reference Baratta2009; Biber Reference Biber2006a, Reference Biber2006b; Biber and Finegan Reference 259Biber and Finegan1988, Reference Biber and Finegan1989b; Charles Reference Charles2003, Reference Charles2006b, Reference Charles2007; Crompton Reference Crompton1997; Dressen Reference Dressen2003; Grabe and Kaplan, Reference Grabe, Kaplan, Markkanen and Schroder1997; Holmes, Reference 264Holmes1986; Hunston Reference Hunston and Ghadessy1993, Reference Hunston and Coulthard1994; Hyland Reference Hyland1994, Reference Hyland1996a,Reference Hylandb Reference Hyland1998a,Reference Hylandb; Hyland and Tse Reference Hyland and Tse2005; Mauranen and Bondi Reference Mauranen and Bondi2003; Meyer Reference Meyer, Markkanen and Schroder1997; Myers Reference Myers1989, Reference Myers1990; Silver Reference Silver2003; Salager-Meyer Reference Salager-Meyer1994; Swales and Burke Reference Swales, Burke, Leistyna and Meyer2003; Tucker Reference Tucker2003; Varttala Reference Varttala, Cortese and Riley2003; see also Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999, especially chapter 12). Corpus-based approaches to stance typically focus on overt lexical and grammatical patterns that mark tentativeness, possibility, certainty, and attitudes in discourse. Stance investigations have analyzed language at a range of linguistic levels, including analyses of a single lexical item (e.g., Diani Reference Diani2008 on really as an emphasizer, Aijmer Reference Aijmer2009 on seem), lexical classes (e.g., Swales and Burke Reference Swales, Burke, Leistyna and Meyer2003 on evaluative adjectives), and particular grammatical structures (e.g., Biber and Finegan Reference 259Biber and Finegan1988 on stance adverbials; Baratta Reference Baratta2009 on passives; Charles Reference Charles2006b, Reference Charles2007, and Hyland and Tse Reference Hyland and Tse2005 on that-clauses; Hewings and Hewings Reference Hewings and Hewings2002 on extraposed clauses).

Most research on stance in academic writing has focused on lexico-grammatical patterns; that is, on patterns in which a lexical item conveys a particular stance meaning relative to a proposition in the immediate discourse context. Stance studies rarely focus on evaluative lexis alone. Several of these studies attempt to document the entire system of stance marking, accounting for the range of stance meanings (attitudinal vs. epistemic meanings; certainty vs. doubt) as well as the range of grammatical devices used to express those meanings (e.g., modal verbs, adverbials, complement clauses). The frameworks presented in Biber (Reference Biber2006a, Reference Biber2006b; Biber and Finegan Reference Biber and Finegan1989b; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999) and Hyland (Reference Hyland1996a, Reference Hyland1998a, Reference Hyland1998b) generally take this approach: they analyze a whole set of structural types of stance markers that convey a range of stance meanings.

Hyland (Reference Hyland1996a,Reference Hylandb, Reference Hyland1998b) has focused specifically on the expression of ‘hedging’ in academic research articles. For example, content-oriented hedges have two major functions: indicating the accuracy of a proposition (e.g., adverbials like generally, approximately, partially, possibly), or limiting the writer’s commitment to a proposition (e.g., the present work indicates…, the model implies…). Hyland (Reference Hyland2002c) extends this line of research by investigating the ways in which authors refer to themselves (and when they do not refer to themselves) in academic prose. One of the themes that Hyland has developed over recent years is that academic research articles are interactive, in that authors actively try to involve the reader in the communication process. Specific studies in this line of research have investigated the use of addressee features (Hyland Reference Hyland2001), questions (Hyland Reference Hyland2002b), and directives (Hyland Reference Hyland2002a) in academic research articles.

A single grammatical structure that has received considerable attention in academic writing research is the noun phrase. As early as 1960, Wells documented the ‘nominal style’ of academic writing, in contrast to the ‘verbal style’ of other varieties. The most studied characteristic of the noun phrase in academic writing is the heavy reliance on nominalizations, described as the most important manifestation of ‘grammatical metaphor’ in modern science writing (see, e.g., Halliday Reference Halliday and Ghadessy1988/Reference Halliday2004; Halliday and Martin Reference Halliday and Martin1993; Banks Reference Banks2005, Reference Banks2008). Grammatical metaphor refers to the meaning expressed by nominalized forms, whereby the description of processes and actions – usually expressed with verbs – is converted into more static representations expressed by nouns (see Halliday Reference Halliday2004 for a collection of key works on nominalization and grammatical metaphor in science writing).

This focus on the nominalizations as a major building block of written academic texts has been extended in other studies to include associated noun phrase structures, such as nouns and adjectives as noun pre-modifiers, and prepositional phrases and participle clauses as noun post-modifiers (see Biber Reference Biber1988; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999; Biber and Clark Reference Biber, Clark, Fanego, López-Couso and Pérez-Guerra2002; Biber and Gray Reference Biber and Gray2010; Biber, Gray, and Poonpon Reference Biber, Gray and Poonpon2011; Fang, Schleppergrell, and Cox Reference Fang, Schleppergrell and Cox2006; Vande Kopple Reference Vande Kopple1994). In Section 3.3, we present some of the major corpus findings from previous research regarding the types of noun phrase structures that are particularly prevalent in academic writing, and then we return to more detailed analyses of many of these structures in the following chapters.

3.2.2 A comprehensive survey of the distinctive grammatical features of academic writing: Contributions from The Longman Grammar of Spoken and Written English

The research studies summarized in Section 3.2.1 focus on grammatical characteristics that have special functions in academic writing. Surprisingly, though, some of these features are not actually common in that register. For example, previous studies have documented the special functions of questions, imperatives, stance adverbials, stance complement clauses, and conditional clauses in academic research writing, even though all of these grammatical features are used much more commonly in conversation than in writing. These features are salient in academic writing – maybe because they are more likely to be found in other registers – and they serve special discourse functions in academic writing. But in terms of their frequency of use, they are not typical of the grammatical style found in most written academic texts.

An alternative approach is to take a comparative corpus-based approach to identify the grammatical features that are used more commonly in academic writing than in other registers. In the present section, we summarize the findings from the most comprehensive study of this type: the corpus-based grammatical descriptions in the Longman Grammar of Spoken and Written English (LGSWE; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999).

The LGSWE is a corpus-based reference grammar of English, describing the complete range of grammatical features in English, and comparing the distributions and functions of those features across four major registers: conversation, fiction, newspapers, and academic prose. The register comparisons are based on analysis of a representative corpus of texts (the Longman Spoken and Written English Corpus) containing approximately 5 million words from each register (see LGSWE: 24–35). The LGSWE describes the structural characteristics of grammatical features in English, but at the same time, it describes the patterns of use for those features: register differences and other contextual factors that influence the patterns of variation. For this reason, the LGSWE provides relatively comprehensive descriptions of those four registers. The description of academic prose is especially relevant for our purposes here.

The academic prose sub-corpus used for the LGSWE consists of both academic books (2.65 million words) and academic research articles (2.68 million words; see LGSWE: 32–34). Texts from both books and research articles were collected from a number of academic disciplines, including biology, chemistry, medicine, sociology, education, and law/history/politics. The corpus thus represents academic prose as a general register.

The frequency findings in the LGSWE allow us to identify core grammatical features that are typical of academic writing by virtue of the fact that they occur more frequently in that register. For example, corpus analysis in the LGSWE shows that nouns are especially common in academic prose: there are about 300,000 nouns per million words in academic prose, compared to only around 150,000 per million words in conversation (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 235). Adjectives are also much more common in academic prose than other registers (about 80,000 adjectives per million words in academic prose versus 20,000 in conversation and 60,000 in fiction and newspapers; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 506). These grammatical features are not necessarily salient, and as a result, many of these patterns have not been the focus of previous investigations. Taken together, though, they provide a comprehensive description of the grammatical style typically employed in written academic discourse.

Table 3.1 catalogs the distinctive grammatical features of academic writing, based on a survey of all quantitative findings reported in the LGSWE. Specifically, Table 3.1 lists all grammatical features described in the LGSWE that occur much more frequently in academic prose than in the other registers (conversation, fiction, and newspaper prose).

The findings summarized from the LGSWE show that written ‘academic’ grammatical features come from most structural categories. However, three word classes are especially prevalent: nouns, adjectives, and prepositions. Overall, these grammatical classes are more frequent in academic prose than in other registers, and there are many related specific features that are especially characteristic of academic prose (e.g., nominalizations, noun phrases with multiple modifiers, stance noun + of-phrase). In contrast, verbs are generally much less common in academic prose than in other registers, although there are specific verb categories that are typical of academic prose (e.g., copula be, existence verbs, derived verbs, and passive voice verbs). Similarly, there are specific categories of adverbs and adverbials (e.g., linking adverbials) that are especially common in academic prose, even though adverbs overall are more common in spoken registers.

At the same time, few of these linguistic features are uniquely characteristic of academic prose. Rather, most of these characteristics occur in other registers as well, although they are much more common in academic writing. Features which are markedly more common in, but not restricted to, a particular register can be considered ‘register features’ rather than ‘register markers’ (see Biber and Conrad Reference Biber and Conrad2009: 53–55). That is, these features are ‘(1) pervasive – distributed throughout a text from the register, and (2) frequent – occurring more commonly in the target register than in most comparison registers’ (Biber and Conrad Reference Biber and Conrad2009: 53). Taken together, these features provide an overall portrait of the distinctive grammatical landscape of academic writing.

3.2.3 Studies of register variation that include academic writing

Studies of register variation generally consider a wide range of grammatical characteristics, with the goal of comparing registers with respect to their exploitation of these features. Many of these studies compare the characteristics of academic writing with other spoken and/or written registers, while others describe the patterns of variation among sub-registers within the general category of academic writing.

For example, Biber (Reference Biber2006a) describes the typical linguistic characteristics of written academic registers in American universities (textbooks and coursepacks) contrasted with spoken university registers (e.g., classroom teaching, advising sessions), and non-academic written registers found in the university (e.g., course syllabi, university catalogs). In addition to the physical mode difference (writing versus speech), these registers differ from one another with respect to interactiveness, production circumstances, communicative purpose, target audience, and so on. That study identifies a number of grammatical features that are considerably more common in academic written registers than in spoken registers or other written registers (e.g., common nouns, nominalizations, prepositional phrases, linking adverbials, passive voice, noun + that-clause constructions, and extraposed to-clause constructions).

Gray (Reference Gray2015b) takes this kind of research a step further by comparing and contrasting the grammatical characteristics of academic research writing from different disciplines, while simultaneously considering the influence of different types of research (i.e., qualitative, quantitative, and theoretical research). This study found that linguistic variation across six disciplines and three research types could be associated with a number of situational characteristics. For example, some of the patterns observed in the corpus could be explained by specific disciplinary differences (e.g., the frequent use of if-conditional clauses in philosophy; the frequent use of linking adverbials in theoretical physics). Other linguistic patterns were associated with the type of research. For example, regardless of discipline, qualitative research articles utilized narrative features (e.g., past tense verbs) of language to a greater extent than quantitative or theoretical research.

‘Multi-dimensional’ (MD) analyses – a research approach developed to investigate the linguistic patterns of register variation – have often included linguistic descriptions of academic writing compared to other spoken and written registers (see, e.g., Biber Reference Biber1988, Reference Biber1995, Reference Biber2006a; Conrad and Biber Reference Conrad and Biber2001). Studies in this research tradition use large corpora of naturally-occurring texts to represent the range of spoken and written registers in a discourse domain. These registers are compared with respect to ‘dimensions’ of variation, comprising constellations of linguistic features that typically co-occur in texts.

MD analysis uses statistical factor analysis to reduce a large number of linguistic variables to a few basic parameters of linguistic variation. In MD analyses, the distribution of individual linguistic features is analyzed in a corpus of texts. Factor analysis is used to identify the systematic co-occurrence patterns among those linguistic features – the ‘dimensions’ – and then texts and registers are compared along each dimension. Each dimension comprises a group of linguistic features that usually co-occur in texts (e.g., nouns, attributive adjectives, prepositional phrases); these co-occurrence patterns are identified statistically using factor analysis. The co-occurrence patterns are then interpreted to assess their underlying situational, social, and cognitive functions.

Researchers have undertaken MD analyses of numerous languages and of many discourse domains in English. Given that each of these studies is based on a different corpus of texts, representing a different language and/or discourse domain, it is reasonable to expect that they would each identify a unique set of dimensions. However, despite these differences, there are certain striking similarities in the dimensions that are uncovered across these studies. Most importantly, in nearly all previous MD studies, there is a dimension associated with an oral-literate opposition (see Biber Reference Biber2014). Table 3.2 summarizes the linguistic composition of this oral/literate dimension in studies of particular discourse domains in English.

Linguistically, this opposition is realized as two fundamentally different ways of constructing discourse: clausal versus phrasal. That is, the oral pole of this dimension consists of verb classes (e.g., mental verbs, communication verbs), grammatical characteristics of verb phrases (e.g., present tense, progressive aspect), and modifiers of verbs and clauses (e.g., adverbs and stance adverbials). In most studies, these ‘oral’ features also include various kinds of dependent clauses that function as clausal constituents, including adverbial clauses and finite complement clauses. These clausal features usually co-occur with pronoun classes and various types of colloquial features (like discourse markers). In contrast, the ‘literate’ pole usually consists of phrasal devices that mostly function as elements of noun phrases, especially nouns, nominalizations, attributive adjectives, and prepositional phrases.

Functionally, this ‘oral-literate’ dimension is usually interpreted as distinguishing between a personal/involved focus (personal stance, interactivity, and/or real time production features) versus informational focus. And in nearly every case, this parameter is the first dimension identified by the statistical factor analysis (i.e., it is the most important factor, accounting for the greatest amount of shared variance).

For our purposes here, the most important characteristic of this first dimension is that it consistently distinguishes academic writing (and other kinds of specialist informational writing) from all other registers. Further, within the domain of academic writing, this dimension distinguishes science research writing from other disciplines and other types of academic writing. Grammatically, these registers (i.e., academic writing generally, and science research writing in particular) are distinctive in their use of nouns and phrasal modifiers of noun phrases, in contrast to a greater use of clausal and involved grammatical features in other registers.

It is perhaps not surprising that Dimension 1 in the original 1988 MD analysis was strongly associated with the oral/literate opposition, given that the corpus in that study ranged from spoken conversational texts to written expository texts. For the same reason, it is somewhat predictable that a similar dimension would have emerged from the study of spoken and written registers in World English varieties (Xiao Reference Xiao2009), the study of eighteenth century general written and speech-based registers (Biber Reference Biber, Diller and Gorlach2001), the study of university spoken and written registers (Biber Reference Biber2006a), and the study of spoken and written responses on a standard English as a second language (ESL) exam (Biber and Gray Reference Biber and Gray2013a).

However, it is more surprising that restricted comparisons of written registers would uncover a first dimension with a similar set of co-occurring linguistic features, associated with a similar opposition between oral-clausal and informational-phrasal styles (Reppen Reference Reppen, Conrad and Biber2001; Gray Reference Gray2013; Biber and Kurjian Reference Biber, Kurjian, Hundt, Nesselhauf and Biewer2007; Grieve et al. Reference 263Grieve, Biber, Friginal, Nekrasova, Mehler, Sharoff and Santini2011). In all of these cases, the linguistic composition of Dimension 1 is surprisingly similar, generally opposing verbs, dependent clauses, pronouns, and interpersonal features versus nouns and phrasal noun modifiers. Of course, the specific patterns of register variation vary, distinguishing between ‘oral’ versus ‘literate’ registers within a restricted discourse domain, and within a single mode. But the linguistic differences among those registers are strikingly similar across studies, suggesting that this is a fundamentally important parameter of register variation. In the following chapters, we undertake detailed linguistic investigations of many of the individual linguistic features from this dimension of variation.

3.3.1 Academic writing contrasted with conversation: Which is more grammatically complex?

Grammatical complexity is usually linked with elaboration and clausal embedding in linguistic theory. A ‘simple’ clause has only a subject, verb, and object or complement. A ‘simple’ noun phrase has a determiner and head noun. Additions to these structures represent elaboration, resulting in ‘complex’ grammar. In particular, there is widespread agreement among grammarians that embedded clauses are an important type of grammatical complexity (often contrasted with ‘simple’ clauses; see, e.g., Huddleston Reference Huddleston1984: 378; Willis Reference Willis2003: 192; Purpura Reference Purpura2004: 91; Carter and McCarthy Reference Carter and McCarthy2006: 489).

Academic writing is often claimed to be structurally complex in these terms. Thus, academic writing has been described as having longer sentences, longer ‘t-units’ (a main clause plus all associated dependent clauses), ‘longer and more complex clauses with embedded phrases and clauses’ (Hughes Reference Hughes1996: 34), and a greater use of subordinate clauses (see, e.g., O’Donnell et al. Reference 268O’Donnell, Griffin and Norris1967; O’Donnell Reference O’Donnell1974; Kroll Reference Kroll, Keenan and Bennett1977; Chafe Reference Chafe and Tannen1982; Brown and Yule Reference Brown and Yule1983).

In contrast, conversation has long been described as grammatically simple in these terms. Conversational participants share time and place, and they normally also share extensive personal background knowledge. As a result, pronouns and vague expressions are common, and referring expressions generally do not need to be elaborated in conversation. Because of these factors, conversational grammar is assumed to be generally not complex, employing ‘simple and short clauses, with little elaborate embedding’ (Hughes Reference Hughes1996: 33).

The following conversational excerpt illustrates many of these characteristics:

This conversational excerpt illustrates the reliance on short, simple clauses, such as it’s pretty cool, I’ve got free tickets, I’ll go with you, and it’s like five bucks. In addition, the excerpt illustrates many of the most salient characteristics of conversation, including hesitations, false starts, and short non-clausal utterances.

However, an equally important – but less often noticed – aspect of conversational grammar is that it is complex, relying heavily on dependent clauses. In fact, nearly any conversational interaction illustrates a dense use of dependent clauses. So, for example, Text Sample 3.4 is taken from the same conversation as Sample 3.3:

Corpus evidence (see below) shows that Sample 3.4 is typical of conversation generally. Thus, the stereotype of conversation as grammatically simple is misleading. In fact, dependent clauses are pervasive in conversation. For example, in Sample 3.4, almost every conversational turn includes one or more dependent clauses.

It similarly turns out that our linguistic stereotypes about academic writing are only partly correct. On the one hand, it is possible to find some academic texts with a heavy reliance on embedded clauses, as in Text Sample 3.5:

However, many other academic texts employ surprisingly few embedded clauses. These texts are composed of long, complex sentences, but the complexity is associated with dependent phrases (embedded in noun phrases), rather than dependent clauses. Thus consider the following excerpt from a biology research article. While there are a few dependent clauses in this passage, the much more important characteristic is the pervasive use of phrasal modifiers:

Detailed consideration of individual texts like Samples 3.3–3.6 help us to notice discrepancies between the preconceptions of linguists regarding grammatical complexity versus the linguistic patterns found in actual spoken and written texts. However, it is difficult to determine the general patterns of use through such an approach; we have no way of knowing whether an individual text sample is representative of the entire register. In contrast, corpus-based analysis is ideally suited for this kind of research question (see, e.g., Biber, Conrad, and Reppen Reference Biber, Conrad and Reppen1998; McEnery, Tono, and Xiao Reference McEnery, Xiao and Tono2006). By basing analyses on large, representative collections of texts, it is possible to document the linguistic patterns that are actually representative of a register. And through such analyses, we often discover patterns of use that directly contradict our preconceived notions.

Research findings relating to grammatical complexity provide a classic case study of this type, showing how corpus-based research can uncover linguistic patterns of use that directly contradict our prior beliefs and expectations. The results presented below are adapted from Biber and Gray (Reference Biber and Gray2010) and Biber, Gray, and Poonpon (Reference Biber, Gray and Poonpon2011), based on analysis of large corpora of texts from academic research writing (the 20^th Century Research Article Corpus) and natural conversation (from the Longman Spoken and Written Corpus).

In Chapter 2, we introduced the linguistic features in English that are related to grammatical complexity. As noted in Section 2.5.2, complexity features can be compared along two grammatical parameters: Parameter A – Structural Type, and Parameter B – Syntactic Function. The corpus analyses presented below show that these two grammatical parameters are especially important for understanding the preferred distribution of complexity features in conversation versus academic writing.

The first parameter – structural type – refers to the ways in which information can be conveyed and elaborated through different types of grammatical structures: finite dependent clauses, non-finite dependent clauses, and dependent phrases. These structural types differ in the amount of grammatical information that they incorporate. Finite dependent clauses mark tense and usually include a grammatical subject, which often identifies the agent of the action described by the verb. Non-finite dependent clauses can mark aspect and voice (active or passive), but they do not mark tense or modality. In most cases, non-finite clauses also do not include a grammatical subject. Both finite and non-finite dependent clauses incorporate the full set of other clause elements, including objects and adverbials, specifying the semantic relations among those clause elements. Dependent phrases are at the opposite extreme from finite dependent clauses, including almost no overt grammatical information: no tense, aspect, voice, or modality; no overt identification of an agent; and no explicit identification of objects or adverbial elements.

The second parameter – syntactic function – refers to the syntactic role that the structure serves in a clause or phrase. The major distinction here is whether a structure functions syntactically as an element in a clause or as an element in a phrase. Clause-level elements can be adverbials (modifying the main verb) or complements (filling subject or object slots in a clause). Phrase-level elements modify the head of a phrase. These are most often modifiers of a noun (e.g., hypothetical distribution, blood capillaries, the average for all groups, a decrease in opacity), but they can also serve as modifiers of an adjective (e.g., a relatively slow rate).

The analyses in the following sections show that both of these grammatical parameters are important for understanding the distribution of features across spoken and written registers. Figure 3.2 (above) summarizes the patterns of variation in conversation versus academic writing: finite dependent clauses and constituents in clauses are more common in conversation, while dependent phrases and constituents within noun phrases are preferred in academic writing.

Figure 3.2 The distribution of complexity features in conversation and academic writing

Tables 3.3–3.5 present the results of statistical comparisons for the use of specific complexity features in conversation versus academic writing. Table 3.3 presents the findings for finite dependent clause types; Table 3.4 displays non-finite dependent clause types; and Table 3.5 shows dependent phrase types. The rate of occurrence for each grammatical feature was measured in each text, allowing us to then calculate mean scores and standard deviations for each register (See Chapter 2, Section 2.4). Tables 3.3–3.5 report the results of statistical comparisons (using ANOVA) to determine whether the mean scores for conversation versus academic writing are significantly different from one another.

The r² scores in these tables provide a measure of the importance or strength of the difference between the two registers. For most of these features, the r² value is over .3 (i.e., over 30% of the variation in the feature can be predicted by the register difference between conversation and academic writing). However, the r² values for some features are considerably larger. For example, finite complement clauses controlled by verbs (that-clauses and WH-clauses) have r² values over .55 (Table 3.3), while prepositional phrases as noun modifiers have an r² value around .90 (Table 3.5).

One especially interesting finding that emerges from these tables is that many of these complexity features are not common in academic writing. In fact, these linguistic differences are for the most part split between features that are strongly favored in conversation (and rare in writing) versus features that are strongly favored in academic writing (and rare in conversation). Figures 3.3 and 3.4 plot the most frequent features, showing the magnitude of these differences visually.

Figure 3.3 Common finite clause types functioning as clausal constituents

Figure 3.4 Common dependent phrasal types functioning as constituents in a noun phrase

In terms of the two grammatical parameters introduced above, the most strongly favored types of structural complexity in conversation are finite dependent clauses (Parameter A) functioning as constituents in other clauses (Parameter B). Figure 3.3 plots the mean scores for finite clauses functioning as adverbials (if and because) and finite clauses functioning as verb complements (that, ZERO, and WH), showing that these clause types are much more common in conversation than in academic writing. Text Sample 3.7 illustrates the dense use of these finite clausal structures typical in everyday conversation.

In contrast, these two grammatical parameters are aligned in the opposite way in academic writig, which has a preferred use of phrasal rather than clausal structures (Parameter A) functioning as constituents in noun phrases (Parameter B). The use of prepositional phrases as post-nominal modifiers is the clearest case of this type (see Figure 3.2): they are extremely common in academic writing but rare in conversation. Attributive adjectives and nouns as nominal pre-modifiers are also phrasal constituents embedded in the noun phrase, and they pattern exactly like prepositional phrases as post-modifiers. Text Sample 3.8 illustrates the dense use of all three phrasal complexity features typical of academic prose:

Some grammatical complexity features are mixed or intermediate on the two parameters (see discussion in Section 2.5.2). That is, a feature may be ‘clausal’ in terms of structural type, but a ‘phrasal constituent’ in terms of syntactic function, and vice versa. These features include:

1. a) finite clauses functioning as a constituent in a noun phrase

2. b) phrases (non-clausal) functioning as a constituent in a clause

3. c) non-finite clauses (all syntactic functions)

Figure 3.5 shows that these intermediate features are generally less frequent in absolute terms than the other two sets of features, and that the differences between conversation and academic writing are generally less extreme. The exception is simple adverbs functioning as adverbials, which are very frequent in conversation. Other ‘mixed’ complexity features, however, show smaller differences between conversation and academic writing: prepositional phrases as adverbials are only moderately more frequent in academic writing than in conversation (r² = .37); verb + to-clause constructions are only moderately more frequent in conversation (r² = .13); that relative clauses occur with nearly the same frequency in both registers (no significant difference).

Figure 3.5 Dependent structures that mix the two parameters

In addition, Figure 3.5 shows that Parameter B (syntactic function) continues to be particularly important for distinguishing between the characteristics of conversation and academic writing, even for dependent clauses. This is especially the case for dependent clauses functioning as constituents in a noun phrase, which tend to be strongly favored in academic writing (except for that relative clauses), despite being clausal in terms of their structural type. Thus, noun complement clauses (that and to), WH relative clauses, and non-finite relative clauses are all much more common in academic writing than in conversation, although none of these structures is especially frequent in absolute terms. Text Sample 3.9 is a duplication of Text Sample 3.8, with these ‘mixed’ complexity features highlighted. This text sample is typical of academic writing generally, with seven dependent clauses functioning as noun phrase constituents versus only two functioning as clause constituents.

Taken together, Figures 3.3 and 3.5 show that there are more dependent clauses in conversation than in academic writing. The difference is especially striking for finite dependent clauses functioning as clause constituents. For conversational participants, these kinds of structural elaboration do not seem complex; we generally do not even notice these structures, and they certainly do not inhibit normal communication. However, they are structurally ‘complex’ according to the definition of embedded clauses added on to simple clauses.

Our main goal in this section is not to claim that dependent clauses are unimportant in academic writing. As Figure 3.5 shows, certain types of dependent clause are actually more common in academic writing than conversation. However, the main point here is to draw attention to an often overlooked grammatical characteristic of academic writing: the dense use of phrasal modifiers. The discourse style of academic writing is dramatically different from conversation in its reliance on these phrasal modifiers. More surprisingly, we show in the following sections that academic writing is different from almost every other English register in the use of these features. This finding sets the stage for subsequent chapters, which trace the historical development of this distinctive grammatical style.

3.3.2 University textbooks contrasted with university classroom teaching: Is there a general academic style?

It is not surprising that academic writing should be different linguistically from conversation, given that the two differ situationally in just about every conceivable way (mode, production circumstances, communicative purposes, interactivity, etc.). However, the specific nature of the grammatical complexity differences between conversation and academic writing is certainly surprising, with dependent clauses being more frequent in conversation (and phrasal modifiers being much more frequent in writing).

Readers might be more surprised to learn that the clausal discourse style described in the last section is in fact the norm for most registers in English, characterizing all spoken registers as well as most written registers. In contrast, the phrasal discourse style introduced in the last section is restricted to written registers that have an informational purpose. In this section and the next, we present corpus findings that support these generalizations. First, we compare some of the typical grammatical characteristics of written university textbooks to the characteristics of spoken university classroom teaching, showing that classroom teaching – despite its informational focus – is much more similar to conversation than academic writing. Then, in the next section, we compare the grammatical characteristics of academic writing to two other major written registers: fictional novels and newspaper prose.

We might predict that written university textbooks would be relatively similar to university classroom teaching in their characteristic grammatical features: both registers are addressed to the same audience (students), produced by the same types of speakers (academic professionals), and have similar communicative purposes (conveying information about academic topics). University classroom teaching differs from conversation in several respects: in a typical classroom lecture, the speaker and hearer have restricted interaction, the main speaker (the instructor) has the primary purpose of conveying information, and he/she has usually pre-planned the discourse. However, conversation and classroom teaching are similar in that the discourse in both situations is produced in real-time: the speaker is constructing utterances while she/he is actually speaking. This characteristic turns out to be important in influencing the grammatical discourse style of classroom teaching, despite the ways in which classroom teaching is otherwise similar to academic writing.

Biber (Reference Biber2006a) provides a detailed linguistic description of university textbooks and classroom teaching, compared to a range of other university registers (e.g., office hours, study group interactions, course syllabi, university catalogs). The results of that study show that textbooks and classroom teaching are sharply distinguished in the use of most lexico-grammatical features.

For example, Figure 3.6 shows the patterns of use for the four major content word classes, comparing classroom teaching and textbooks to conversation and academic research articles. Textbooks are not quite as extreme as research articles in their frequent use of nouns, but these two registers are nearly identical in their frequent use of adjectives, and their relatively infrequent use of verbs (when contrasted with the two spoken registers). Classroom teaching, on the other hand, is similar to conversation in its dense use of verbs, and its comparatively infrequent use of nouns.

Figure 3.6 Major grammatical classes: academic speech and writing

Figure 3.7 plots the register distribution of three general types of finite dependent clauses: adverbial clauses (if and because), complement clauses (that and WH), and relative clauses (that and WH). Here again, classroom teaching is quite similar to conversation, with a dense use of adverbial clauses and complement clauses. Adverbial clauses are actually somewhat more common in classroom teaching, while complement clauses are more common in conversation. In contrast, both of these types of dependent clause are comparatively rare in academic writing (textbooks and research articles). Finite relative clauses show a different pattern, representing one of the few features that seems to conform to a general academic style addressed to students: very frequent in textbooks, but also frequent in classroom teaching. Interestingly, finite relative clauses are not especially frequent in academic research articles; we return to this finding in Section 3.3.4.

Figure 3.7 Finite dependent clauses: academic speech and writing

Text Sample 3.10 illustrates the typical discourse style of classroom teaching, with a heavy reliance on clausal structures, and relatively little use of phrasal modification:

Both classroom teaching and textbooks are intermediate between the extremes of conversation and academic research articles. However, the grammatical style of classroom teaching (with its heavy reliance on clausal structures) is much more similar to conversation than academic writing. In contrast, textbooks are generally similar to academic research articles in their reliance on nouns and adjectives (and a lesser use of finite dependent clauses).

In sum, these findings show that there is no general academic style. Rather, there is a fundamental divide between speech and writing, with spoken academic texts being dramatically different from written academic texts in their reliance on a clausal rather than phrasal grammatical style. This section has compared two registers with a shared informational purpose but different modes. In the following section, we shift focus, restricting our analysis to only written registers but contrasting popular versus specialist registers.

3.3.3 Grammatical features of academic writing contrasted with popular written registers

The peculiar grammatical style of academic research writing becomes even more apparent when we compare it to the typical grammatical styles of other written registers. In the present section, we focus on two other general written registers: fictional novels and newspaper prose. Like academic prose, these registers are carefully written, revised, and edited, and they are addressed to large readerships. The primary differences among them relate to communicative purpose and the extent to which readers have specialized background knowledge:

Fictional novels are written to entertain and edify, narrating a story, describing characters and places, portraying thoughts and emotions, and so on. Fiction is a complex written register, composed of prose sections interspersed with dialogue. It is not surprising that conversational dialogue in fiction is similar to natural conversation in its grammatical characteristics. In the following, however, we show that the narrative and descriptive prose of fiction is also similar to spoken discourse, especially in its reliance on verbs, pronouns, and finite clauses.

Newspaper prose is more similar to academic writing in being informational in purpose. However, there are differences between the two: newspaper prose includes more narration and reported speech; academic prose includes more explanation and argumentation in support of its informational claims and descriptions. The two registers also differ in their audiences: newspaper prose is addressed to a large general audience. The author of a news story assumes that readers have some knowledge of the culture and current events, but beyond that, little background knowledge is taken for granted. In contrast, academic prose is written for specialists who share considerable expert background knowledge with the author.

Even the most basic grammatical features differ across these three registers, associated with the situational differences described above. For example, Table 3.6 presents the results of statistical comparisons for the four major word classes plus two specialized grammatical categories (passive voice verbs and nominalizations). Figure 3.8 plots the same information graphically. These three registers pattern in a highly regular manner with respect to these features: fiction makes the most frequent use of verbs and adverbs (being similar to conversation in this respect); academic prose makes the most frequent use of nouns, adjectives, and nominalizations; newspaper prose is consistently intermediate between the two extremes.

Figure 3.8 Major grammatical classes across three major written registers

Table 3.7 presents statistical results for the use of several specific dependent clause types, and Figure 3.9 plots the distribution of the major types. For the most part, we see the same pattern here: finite dependent clauses are most frequent in fiction, following again the pattern of use shown in Section 3.3.1 for conversation; finite dependent clauses are considerably less frequent in academic prose; newspaper prose is intermediate between the two extremes. The exception here is for the use of finite relative clauses, which are much more frequent in newspaper prose than in the other two registers. Overall, non-finite relative clauses are much less frequent than the other clause types, but they are the one type of dependent clause found most often in academic prose.

Figure 3.9 Dependent clause types across three major written registers

Fictional novels are surprisingly similar to conversation in the use of these grammatical features, despite the fact that they are monologic and produced in writing. It turns out that the clausal discourse style of fiction is typical of narrative prose as well as fictional dialogue. Thus consider the high density of verbs and dependent clauses in the following narrative text sample:

We have claimed in the preceding sections that the most distinctive grammatical characteristic of academic prose is its reliance on phrasal noun modifiers (contrasting with the rare use of clausal features). Table 3.8 and Figure 3.10 further support this generalization in comparison to the discourse styles of fiction and newspaper prose. Attributive adjectives are more common in fiction than in conversation (compare Figure 3.4), but they are only half as frequent as in academic prose. Nouns as pre-modifiers are rare in fiction but extremely common in academic prose. And the same pattern holds for of-phrases and other prepositional phrases as post-nominal modifiers. Newspaper reportage is again intermediate in the use of these features, using phrasal modifiers more than many other registers of English, but not nearly to the same extent as in academic prose.

Figure 3.10 Noun phrase modifiers across three major written registers

Newspaper prose combines narrative and informational communicative purposes. As a result, it employs many of the same grammatical devices as academic writing, although it uses those devices less commonly. For example, noun-noun sequences are prevalent in newspaper prose in comparison to written registers like fiction, as in Text Sample 3.12:

Prepositional phrases as noun modifiers are also prevalent in many news reports, although again not as frequent as in academic prose texts.

In summary, we have illustrated in this section how communicative purpose and audience have a major influence on grammatical discourse style within the written mode. Thus, popular written registers with non-informational purposes (e.g., fiction) use clausal linguistic styles, similar to spoken registers. Academic written registers, with informational purposes and specialist audiences, have developed a distinctive grammatical style, employing a dense use of nouns and phrasal modifiers rather than verbs and clauses. Newspaper prose is intermediate in this regard, using both clausal modifiers as well as phrasal modifiers, reflecting the fact that newspapers are written for a large, generalist readership, but for relatively informational purposes.

For our purposes in the present book, the main point of the present section has been to highlight the distinctive grammatical characteristics of academic prose. That is, even in comparison to an informational written register like newspaper prose, academic prose is distinctive in its infrequent use of clausal modifiers and its extremely frequent use of phrasal modifiers.

3.3.4 Grammatical variation among written sub-registers from different academic disciplines

Academic writing can be analyzed at many different levels of generality. For example, the descriptions in Section 3.2.2 (surveying corpus findings from the LGSWE) were based on analysis of a large general corpus of academic writing, including books and research articles from many different disciplines (see Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 32–34). At the other extreme, studies like Gray (Reference Gray2013, Reference Gray, Cortes and Csomay2015a, Reference Gray2015b) compare different disciplines, even distinguishing between sub-registers within those disciplines (i.e., qualitative, quantitative, and theoretical research articles).

In the present section, we illustrate the ways in which specific academic sub-registers vary in their use of grammatical features. To do so, we focus on differences among specialist research articles from three major discipline families: humanities, social science, and natural science (based on analysis of the 20^th Century Research Article Corpus; see Chapter 2). In addition to disciplinary differences, we also consider the influence of audience, by comparing science research articles written for specialists to science writing published in multi-disciplinary venues addressed to a wider multi-disciplinary audience (i.e., the ‘Specialist Science’ sub-corpus of the 20^th Century Research Article Corpus vs. recent articles published in the Philosophical Transactions and in Science; see Table 2.2 in Chapter 2).

Tables 3.9–3.11 present the results of statistical comparisons for specific grammatical features across these sub-registers of academic research writing. As the tables show, all of these grammatical features are associated with significant differences across the disciplines, and there are actually large differences in the use of some of these features (shown by the large r² scores).

Table 3.9 displays descriptive statistics for the major lexical word classes across disciplines, while Figure 3.11 presents the same information graphically. Verbs and adverbs are both important features here, shown by the r² scores greater than 0.4. Both of these features follow the same pattern: most common in humanities research writing; somewhat less common in popular science and the social sciences; and least frequent in specialist science research articles.

Figure 3.11 Major grammatical classes across sub-disciplines of academic writing

Common nouns have the opposite distribution, and the differences are even larger (reflected in the r² score of 0.62). In this case, the four registers follow a steady cline: least frequent in humanities writing; somewhat more frequent in popular science articles; more frequent again in social science articles; and by far the most frequent in specialist science writing.

As noted in Section 3.2.1, some researchers have focused on nominalizations as an especially important characteristic of academic writing (see Halliday Reference Halliday and Ghadessy1988, Reference Halliday2004; Banks Reference Banks2008). However, Figure 3.11 shows that nominalizations are not especially frequent in absolute terms, and that there are only minor differences among academic sub-registers (r² = .09). Surprisingly, popular science employs more nominalizations than any other sub-register, while specialist science research articles make the least use of these devices. Similar to nominalizations, adjectives (including both attributive and predicative functions) are somewhat more common in popular science than in the other registers.

Table 3.10 focuses on the use of dependent clauses in these registers, presenting detailed information on several specific clause types. Figure 3.12 plots the distribution of five major dependent clause types. Although non-finite relative clauses are exceptional (see below), the other four dependent clause types are distributed in very similar ways: all four clause types are frequent in humanities writing, and most types are also frequent in popular science. These dependent clause types have intermediate frequencies in social science. At the other extreme, though, all four dependent clause types are much less common in specialist science research articles.

Figure 3.12 Dependent clause types across sub-disciplines of academic writing

Beyond those general patterns, each type of dependent clause has a slightly different distribution across sub-registers. Finite adverbial clauses are not very common overall in any of these registers, but they are even less common in specialist science. Complement clauses – both finite and non-finite – are frequent in both humanities and popular science writing. These clause types are less common in social science, but they are especially rare in specialist science. The majority of the complement clauses in academic prose are controlled by verbs, often used for reporting previous research claims (e.g., Jones suggests that …). However, Table 3.10 shows that complement clauses controlled by nouns follow the same distributional pattern: that-clauses controlled by nouns (e.g., the fact that…) and to-clauses controlled by nouns (e.g., the attempt to…) are most common in humanities; intermediate in popular science and social science; and rare in specialist science. Finite relative clauses are also comparatively rare in specialist science, but relatively frequent in humanities writing. These differences are especially large for WH-relative clauses (see Table 3.10), but the same patterns hold for that-relative clauses. Both popular science and social science are intermediate in the use of those features.

As noted above, non-finite relative clauses are the only type of dependent clause to follow a different distributional pattern: nearly as frequent in specialist science as they are in social science and popular science, but notably less common in humanities writing. Linguistically, this dependent clause type is relatively similar to the phrasal modifiers favored in ‘literate’ discourse: it is minimally clausal (incorporating only a non-finite verb) and it functions syntactically as a noun modifier. It is thus noteworthy that this is the only dependent clause type to be strongly favored in specialist science writing.

Figure 3.13 presents the patterns of use for two features that function as pre-nominal modifiers: attributive adjectives and pre-nominal nouns (see also Table 3.11). Attributive adjectives are slightly more common in popular science articles than the other registers, but the differences are not statistically important (r² = .07). In contrast, there are major statistical differences (r² = .67) across disciplines in the use of nouns as nominal pre-modifiers: rare in the humanities; relatively common in popular science and social science; and most frequent in specialist science.

Figure 3.13 Noun phrase modifiers across sub-disciplines of academic writing

As Figure 3.14 shows, there are also major differences across academic sub-registers in the use of prepositional phrases (see also Table 3.11). Three grammatical types/functions are distinguished here: prepositional phrases functioning as adverbials (or as part of prepositional verbs), of-phrases functioning as noun modifiers, and other prepositional phrases functioning as noun modifiers. Prepositional phrases functioning as adverbials are considerably more common in humanities prose than in the other disciplines; they are least common in popular science writing. Prepositional phrases as noun modifiers are considerably less common. Humanities academic prose and popular science articles pattern alike in the use of these features: of-phrases functioning as noun modifiers are relatively common, while other prepositional phrases functioning as noun modifiers are considerably less frequent. Text Sample 3.13 illustrates the dense use of prepositional phrases as adverbials or verb complements (shown in bold underline) and genitive of-phrases (shown in BOLD CAPS) in humanities research writing:

Specialist science and social science research articles show a quite different pattern from humanities writing: of-phrases are considerably less common in science writing than in humanities, but other prepositional phrases functioning as noun modifiers are considerably more common in science articles than in humanities writing. Thus, of-phrases and other prepositional phrases functioning as noun modifiers occur with roughly the same frequency in science research writing (see Figure 3.14). Two prepositions are notably common in this role: in and for. Prepositional phrases headed by in are especially common in specialist science writing, while prepositional phrases headed by for are most common in specialist social science writing; for example:

Appositive noun phrases are another phrasal device used for post-nominal modification. These structures are familiar to all readers of modern newspaper prose, usually expressed as two noun phrases separated by a comma: NP, NP, as in:

Figure 3.15 shows that NP, NP appositives are considerably more common in newspaper prose than in academic research writing. The most common functions of these structures are to describe the role or importance of an individual person, place, or institution; for example:

While appositives with the format of NP, NP are not especially common in academic prose, NP (NP) appositives are very frequent. However, as Figure 3.15 shows, there is considerable variability among academic sub-registers in the use of appositives. Humanities research writing makes moderate use of NP, NP appositives, but to a considerably lesser extent than in newspaper reportage; NP (NP) appositives are relatively rare in humanities prose. In contrast, NP (NP) appositives are extremely frequent in science research writing. These structures can present many different types of information. For example:

We return to a more detailed discussion of the meanings and discourse functions of appositive noun phrases in Chapter 5.

Figure 3.14 Prepositional phrases across sub-disciplines of academic writing

Figure 3.15 Appositive noun phrases in newspapers versus two sub-disciplines of academic writing

Finally, Table 3.11 compares academic sub-registers for their use of a grammatical device that has been mostly disregarded in previous research: noun-participle compounds functioning as nominal pre-modifiers. These structures consist of either a pre-modifying noun combined with an –ing participle (e.g., filament forming ability) or a pre-modifying noun combined with an –ed participle (e.g., Cdkl-induced vimentin-Ser55 phosphorylation). The use of these structures enables an extremely dense packaging of information, as in:

As Table 3.11 and Figure 3.16 show, these structures are not frequent in absolute numbers, but they are considerably more common in science research articles than in the other registers. In particular, constructions with -ed participles are especially common in science research writing, as in:

Surprisingly, many of the grammatical features described in this chapter can be associated with a loss of explicitness. That is, as grammatical structures become more compressed, there is less overt expression of the meaning relationships among constituents.

Figure 3.16 Noun + participle as nominal pre-modifiers across sub-disciplines of academic writing

One clear example of this pattern involves two alternative ways of indicating the logical relationship between clauses: linking adverbials (e.g., however, therefore) versus colons. Linking adverbials might be considered maximally explicit, in that they express the specific logical relationship between two clauses (e.g., contrast or consequence). In contrast, colons (and semicolons) are maximally inexplicit; they simply identify the existence of a logical relationship between two clauses, with no overt indication of what that relationship is.

Interestingly, the academic sub-registers differ in their reliance on these two clause-connecting strategies, paralleling the differences shown for other grammatical complexity features. That is, as Figure 3.17 shows, overt linking adverbials are relatively common in humanities, social sciences, and popular science articles, but relatively rare in specialist science articles. In contrast, colons as clause connectors are much more common in social science and specialist science articles than in humanities or popular science. This pattern conforms to the overall trend in specialist science writing (and social science writing) to prefer maximally compressed structures, despite the associated loss in explicitness of meaning. These semantic relationships are discussed in detail in Chapter 6.

Figure 3.17 Explicit versus implicit signals of logical relations across sub-disciplines of academic writing