Chapter 3 Notes: Grammar and Its Description

Grammar Definitions and Descriptions

Grammar is a system of rules governing how words and morphemes combine to form sentences.
Grammar description involves analyzing individual sentences to identify underlying rules.
The question is whether it should include more than sentence-level analysis, considering language use in communication and context.

Language Competence and Grammaticality

Debates exist regarding the inclusion of non-grammatical, contextual factors in grammar description.
Chomsky focused on grammatical competence, emphasizing phrase and clause structure rules.
Grammaticality judgment tasks assess whether sentences are grammatical (possible) or ungrammatical (not possible).
Table 3.1 presents sample items from a grammaticality judgment task.
In grammaticality judgment tasks, sentences are randomly distributed, and test-takers judge acceptability.
Native speakers complete these tasks quickly, while L2 learners show variable performance.

Grammaticality Judgment Tasks

Ungrammatical sentences in these tasks violate morphology or syntax rules.
L2 research uses grammaticality judgment tasks to assess language competence.
The Johnson and Newport study investigated the relationship between age of English learning and success.
Task performance measured language ability, with lower scores indicating less success.

Communicative Competence

Hymes' theory of communicative competence emphasizes social factors.
It includes judging sentence appropriateness in context, not just grammaticality.
Native speakers choose from vast grammatical options in every utterance.
This question was central to structural linguistics research.

Systemic-Functional Linguistics (SFL)

M.A.K. Halliday founded systemic-functional linguistics.
SFL describes language as a resource for expressing meaning in social contexts.
Grammar provides the means for expressing speech functions.

Grammar as a Resource

Grammar description involves more than morphological and syntactic rules.
Halliday saw his functional approach as complementary to Chomsky’s generative approach.
Chomsky focused on innate language abilities, while Halliday focused on external, social factors shaping language development.

Child Language Acquisition

Halliday observed children interacting with parents from birth.
Before words and grammar, children use vocal chords to express intentions and desires, called “functions.”
General functions include satisfying material needs, controlling behavior, getting along, expressing self, exploring, and communicating information.
Children learn specialized meanings within each function.
For example, within “satisfying material needs,” children request milk, food, etc.
Initially, meanings are indicated through gestures or sounds.
Children learn specific forms (vocabulary and grammar) through interaction with caregivers.
They learn which forms are appropriate for a given function in a communicative setting.
Halliday described this process as “learning how to mean.”
The child acquires a range of potential, their ‘meaning potential’.
This involves mastering elementary language functions and a range of choices in meaning within each one.
Choices expand as the functional potential of the system is reinforced by success.
This mapping from form to function occurs through interaction.
It's impossible for a child to learn what is appropriate without real-context communication.

Register

In systemic-functional linguistics, analysis starts with the communicative situation, or register.
Register is similar to Hymes’ concept of a speech event.
It's defined by situational factors like institutional setting, communication goals, roles, participant relationships (intimate or not, power equality), and mode (spoken or written).
Systemic-functional linguists use knowledge of contextual factors to explain grammatical choices.
Doing grammar means establishing and explaining the principles behind the wordings of a natural language.
In systemic-functional linguistics, principles derive from examining which situational factors determine which selections in the linguistic system.
Identifying important situational factors involves analyzing the larger register: mode of communication, participants, and purpose.
REGISTER: Defined by situations in which we use language. Includes:
- Mode (Writing or Speech)
- Purpose (Teach, Complain)
- Setting (Home, Work, School)
- Relationship (Friends, Strangers)
Registers: General or Specific
- General - Written or Spoken
- Specific - Newspapers, Academic Lectures, etc.
Although individual sentences are analyzed in a systemic-functional approach, they cannot be analyzed in isolation, as the goal of analysis is to explain how the communicative context impacts the wording of the sentence.
The unit of analysis is not a sentence, as it would be in a generative tradition, but a text.
Each grammatical feature observed within a text (e.g., the use of the past tense or personal pronouns) represents a choice made by the speaker or writer, and each of these choices can be explained by looking at the type of text in which they occurred, as each individual text (e.g., an academic essay or an informal conversation) can be linked to a larger register or social context.
Grammar is not so much a collection of sentence-level rules, but rather is a system that can be exploited by speakers and writers for the purpose of communication.
Mastering a language means mastering the ability to use language forms for particular purposes in a variety of social contexts, in a way that is not just grammatical, but appropriate and effective.

Pedagogic Framework for Grammar Description

Halliday’s systemic-functional approach appealed to applied linguists in second language acquisition and pedagogy.
Hymes’ theory of communicative competence laid the groundwork for communicative language teaching.
Halliday’s functional approaches provided tools for describing grammar use.
Their theories suggested new approaches for linguistic and pedagogical grammars.
Larsen-Freeman (1989; 2003; Celce-Murcia & Larsen-Freeman, 1999) proposed a framework for a pedagogical grammar.
It could be used by second language teachers to develop students’ grammatical competence.
Larsen-Freeman’s framework has three major components, called “The Three Dimensions”: Form, Meaning, and Use (see Figure 3.1).
- Form: phonology, graphology (written symbols), semiology (signs), morphology, and syntax.
- Meaning: semantics, refers to the meaning attached to a particular form when it is presented in isolation, such as a definition of a word, or the concept of “past time” associated with the –ed ending in English.
- Use: pragmatics, refers to a speaker’s or writer’s intentions in a particular communicative context.
L2 grammar instruction must provide information about all three dimensions.
L2 students need to learn the wide range of meanings a form can express within a given function and context.
Embedded within this framework is what Larsen-Freeman (2003, p. 1) calls “the grammar of choice.”
A good grammar should explain why one form was chosen over others, even with the same meaning.
[L2 students] will need to understand that, as speakers or writers, they have choices to make, and that those choices have consequences, so that they can learn to use the language in a way that honors their intentions. Besides, students need to be able to draw inferences about the intentions of others. To the best of our ability, therefore, we should help students understand the linguistic options available.

Functional Descriptions of Grammar

To help language teachers provide more functional descriptions of grammar use, Larsen-Freeman and Marianne Celce-Murcia developed a pedagogical grammar reference titled The Grammar Book: An ESL/EFL Teacher’s Course.
Drawing on individual studies and their own linguistic analyses, Celce-Murcia and Larsen-Freeman provided descriptions of form, meaning, and use.
Celce-Murcia and Larsen-Freeman (1983, 1999, forthcoming) emphasized the importance of looking at grammar in its discourse context in L2 classrooms.
Grammars for L2 learners should discuss how register and text structure impact sentence grammar.
A pedagogical grammar should include reasons/explanations for choosing one grammar structure over another.
At the time the first edition of The Grammar Book was published, there were few comprehensive accounts of grammatical choices in various communicative contexts.
Linguistic research analyzed a small number of texts.
Observations about grammatical feature use could not be generalized.
To address the uneven nature of functional descriptions of grammar in use, Celce-Murcia (1975; see also 1990, 2000, 2002) developed methods for training teachers to do their own “contextual analysis” of spoken and written texts.
Teachers could select grammatical features to learn more about and analyze their use in relevant texts.
Contextual analysis empowered teachers to do their own research and determine that once a teacher decided what she wanted to learn, she would need to analyze at least 100 tokens of a target form or structure (complete with contextual information and context) to begin making useful generalizations about where the target form occurs (or does not occur), what it means, and why it is used (or not used) by a given speaker/writer in a given piece of discourse.
In addition to time, contextual analysis also requires a great deal of training, the type of training one might find in a TESOL or applied linguistics graduate program. And while this type of analysis allows teachers to investigate specific questions that arise in their classrooms, the findings, like those of the small-scale linguistic studies being carried out at the time, cannot be used to make general claims about how a feature is typically used by most speakers of the language.
Larsen-Freeman’s Three Dimensions and Celce-Murcia’s contextual analysis provided important frameworks and tools for L2 grammar teaching.
In the early 1980s, however, when the first edition of The Grammar Book was published, there was still much to be done in the area of functional grammar description.
What was needed, in addition to new ways of training teachers to discuss grammar in the classroom, was a comprehensive account of grammar use across a wide range of text types and registers, not only for English, but other languages as well.
This was no easy task. It required the analysis of several hundred linguistic features in thousands of texts sampled from all the major domains of daily life: home, work, school, and so on. If done by hand, it would take decades, if not centuries.
With the help of computers, however, such large-scale analyses would take place, and, in a few short years, these studies would greatly enrich our understanding of how grammar is used in spoken and written contexts.

Corpus Linguistics and Language Performance

Describing grammar use in social contexts required language performance data analysis.
Chomsky argued language data was limited and didn't reveal underlying competence.
The shift away from language performance temporarily reduced text analysis activity.
Two developments renewed interest in language performance data collection:
- Growing calls for socially-oriented language theories (Hymes, Halliday, Labov).
- Advancements in computer technology made it possible to analyze word counts and retrieval automatically. Automatic methods of text analysis to compare their linguistic features.
This was of particular interest to lexicographers and showed great promise in the area of functional grammar description.
The field of corpus linguistics emerged, where “corpus” used to mean text collection.
In the 1960s, corpus linguistics became associated with electronically storing and analyzing texts to represent language use domains.
The first electronic corpus, the Brown Corpus, was compiled by Nelson Francis and Henry Kucera at Brown University.
- The corpus was organized to represent a variety of written registers, and the texts included in the corpus were sampled from texts published in the United States sometime during the year of 1961.
- Two major text type categories were created – Informative Prose and Imaginative Prose – and within these categories, further distinctions were made.
- Text types within Informative prose included newspaper reporting (which was further divided into topic-related categories like politics and sports); newspaper editorials; biographies and memoirs; and “learned” publications (sampled from a variety of academic disciplines: natural sciences, medicine, mathematics, political science, humanities, and so on).
Electronic corpus design was principled in nature - texts were sampled from register categories, balancing text types.
A corpus was designed so that it was representative of a particular domain of use.
Such an approach made it possible to both describe general patterns of grammar use and compare how specific registers (e.g., news writing and academic writing) differed in their grammatical choices.
ELECTRONIC CORPORA - Websites provide comprehensive lists of corpora:
- BYU Corpora
- Corpora4Learning
- Learner Corpora Around the World
- Non-English Corpora

Methodological Developments in Corpus Linguistics

Several corpus development projects followed, and by the 1990s there were numerous collections available, which together represented a wide range of dialects and registers in English and other languages (see Hunston, 2002; Kennedy, 1998; and McEnery & Hardie, 2011, for comprehensive reviews).
The existence of these electronic collections, together with continued advances in computer technology, made it possible to analyze language use in ways that had never been done before.
Three major methodological developments in particular would have a major impact on grammar description, including the development of pedagogic grammars for second language learners:
1. Electronic corpora could be automatically analyzed to generate counts for individual words, phrases, or grammatical structures. This made it possible for linguists to provide more comprehensive descriptions of what was typical or frequent in a particular communicative situation.
2. Electronic corpora could be searched, which allowed linguists to view hundreds of occurrences of a keyword in the context of a sentence, paragraph, or entire text. Lines of text could also be automatically sorted to reveal common phrases and syntactic patterns.
3. The distribution of linguistic features within a register could be compared with that of other registers, which opened the door for large-scale investigations of how situational factors, such as the mode and purpose of communication, impact the choices we make as writers and speakers of a language.

Frequency-Based Descriptions of Grammar Use

Computer programs can't judge acceptability but can count word/structure occurrences in corpora.
This technological advancement was in many ways another revolution in language studies.
Patterns of use could be described on a large scale, analyzing thousands of texts from various registers.
A much wider range of linguistic features could be described, aided by grammatical tagging.
GRAMMATICAL TAGGING When a text is grammatically tagged by a computer program, each word in the text is annotated to indicate its part of speech and other grammatical properties, as shown in the following example, which uses the CLAWS tagger: < http://ucrel.lancs.ac.uk/claws/trial.html >:
- Example:
- ThisDT0 sentenceNN1 hasVHZ beenVBN grammaticallyAV0 tagged VVN. _PUN
- DT0: general determiner
- NN1: singular noun
- VHZ: -s form of the verb “have”
- VBN: past participle of the verb “be”
- AV0: adverb
- VVN: past participle of a lexical verb
- PUN: punctuation
Rule descriptions in the generative tradition relied on native speaker intuitions.
Corpus-based researchers highlighted the limitations of native speaker intuition regarding typical language use.
Humans can't accurately describe their language quantitatively.
This motivated corpus linguists to expand research beyond word frequency to grammatical frequency.
Analysis of tagged texts can generate lists of frequent nouns, verbs, adjective-noun pairings, and clause types following verbs.

Corpus-Based Grammars

One of the first comprehensive, corpus-based grammars of a language, the Collins COBUILD English Grammar, was developed by John Sinclair and colleagues (see Francis, 1993; Francis & Sinclair, 1994; Hunston & Francis, 1998).
Aimed to develop a grammar of English for language learners in ways that differed from traditional grammars.
1. uses only real examples from corpora to illustrate grammatical patterns
2. descriptions of grammar use are based on analysis of corpus data, rather than intuition
3. grammar is not described as separate from lexis; lexico-grammar, or the tendency of particular words and grammatical structures to occur together in spoken and written texts, is the primary focus.
Biber, Johansson, Leech, Conrad, and Finegan (1999) compiled the Longman Grammar of Spoken and Written English (LGSWE).
The LGSWE added another important component to corpus-based grammar description: the analysis of variation across registers.
The corpus upon which the LGSWE’s findings are based was designed to represent four major domains of language use: face-to-face conversation, fiction, news, and academic prose.
As with the Brown corpus, further distinctions within these core registers were made (e.g., academic prose included journal articles and textbook samples from a variety of disciplines).
The LGSWE corpus also included two dialects of English: British and American.
Texts in the corpus were grammatically tagged so that the frequency with which particular grammatical features are used in English could be compared across registers and dialects.
The approach to grammar description taken in the LGSWE is very much a functional one, the LGSWE does not provide frequency counts simply for the sake of counting, but rather aims to provide functional explanations for why some linguistic features are more or less frequently used in particular contexts.
Biber et al. (1999), like Hymes, Halliday, and Larsen-Freeman, are most interested in describing the grammar of choice:
- Every time that we write or speak, we are faced with a myriad of choices: not only choices in what we say but in how we say it. The vocabulary and grammar that we use to communicate are influenced by a number of factors, such as the reason for the communication, the context, the people with whom we are communicating, and whether we are speaking or writing. Taken together, these choices give rise to systematic patterns of use.
Biber et al. provided functional accounts of grammar use on a scale that had never been done before.
Their corpus-based methodology helped to establish empirical means for describing the “myriad of choices” that language users make every day.
These choices could be quantified with the aid of computer programs, allowing linguists to describe “what speakers and writers typically do [versus] what they rarely do” (p. 6).
In the LGSWE, frequency counts for over 30 major grammatical features within the four core registers are provided.
The findings of Biber and his colleagues had immediate relevance for pedagogic grammar description.
L2 textbook writers are faced with a number of decisions that (ideally) require some knowledge of frequency: What vocabulary words and grammatical forms are students most likely to encounter outside of the classroom? How are these forms typically used in spoken and written contexts?

Textbook Analysis

Biber and Reppen (2002) note that textbook writers traditionally haven't had access to frequency information as used in LGSWE, these often based on the author’s gut-level impressions and anecdotal evidence of how speakers and writers use language. In other words, the authors of grammar resources for L2 students often rely on their intuitions when describing grammar use
Biber and Reppen found cases where their intuitions were correct, but in many cases, they were wrong.
Biber and Reppen (2002) compared LGSWE findings with ESL textbook treatments.
High-frequency linguistic features were often missing from L2 grammar textbooks.
Low-frequency verbs were often chosen instead of high-frequency ones such as “wear, cry, revolve, arrive, touch, travel, read, rain, shine, write, ring, drive, enjoy, study, build, rise, smoke, close, speak, grow, kiss, stay, own, taste, cause, and boil” (p. 206).
The LGSWE notes the 12 most common lexical verbs (excluding be, have, and do) are: say, get, go, know, think, see, make, come, take, want, give, and mean (Biber & Reppen, 2002, p. 205).
- Of these 12, only 5 appeared in the textbooks’ example sentences.
Biber and Reppen suspect such omissions are common; native speakers notice infrequent more than common usage.
Several researchers noted similar discrepancies between L2 textbooks and corpus-based research.
For corpus linguists, frequency is important:
- Judging typical usage is a key component of communicative competence.
- Grammar descriptions should reflect real-world grammar use (L2 textbooks should provide accurate examples).
- High-frequency items make up a large proportion of language exposure for students.

Frequency Counts

The goal of corpus-based research is to provide comprehensive descriptions of feature use in real-world contexts.
It's important how the verb get is used in combination with other words to express a variety of meanings, and for what purpose this verb is used (or sometimes avoided) in spoken and written contexts
In corpus linguistics, then, frequency counts are almost always accompanied by contextual analysis and interpretation.
Corpus linguists investigate when, why, and how writers/speakers use or avoid linguistic features, This analysis takes into account immediate discourse context and larger situational context (register).
Such functional interpretations build a more comprehensive picture of choices that language users make each day to achieve communicative goals.

Revolution in L2 Grammar Teaching?

Susan Conrad's 2000 TESOL Quarterly article asked: “Will corpus linguistics revolutionize grammar teaching in the 21st century?” predicted that corpus linguistics would have a major impact on how grammar is taught in L2 classrooms:
- Monolithic descriptions of English grammar will be replaced by register-specific descriptions.
- The teaching of grammar will become more integrated with the teaching of vocabulary.
- Emphasis will shift from structural accuracy to the appropriate conditions of use for alternative grammatical constructions.
Conrad’s predictions emphasize the grammar of choice, considering what we know about grammar use - there can be no one-grammar-fits-all approach to language description (Grammatical study needs to take place within the context of a register or by comparing registers).
Different registers exhibit different patterns of use, general lists of adverbs and rules do little to prepare students for communication
Even discourse-level activities which involve the use of authentic texts can fall short if register variation is not taken into account (For instance, having students write a newspaper or analyze newspaper articles does not necessarily provide practice in the appropriate use of linking adverbials in academic writing)
Future L2 instruction would make greater efforts to integrate grammar teaching with vocabulary teaching (corpus-based research has demonstrated that particular grammatical features tend to co-occur with particular lexical items in syntactic patterns, for example, when get is used in a copular (linking verb) pattern, it is often followed by negative adjectives).

Corpus Linguistics

Making appropriate linguistic choices involves awareness of register and how grammar and vocabulary create meaning.
Grammar instruction should emphasize not only accuracy, but also appropriateness, it should aim not only to describe every grammatical feature available to a user of a language, but also how proficient speakers and writers make decisions about which grammatical features to use when.
Conrad’s (2000) predictions echoed Larsen-Freeman’s recommendations: Teaching Form is important, of course, but Form alone is not enough to build language competence. L2 students also need an understanding of Meaning and Use.
By 2000, corpus linguistics was established, but its impact on grammar description and L2 grammar teaching was uncertain.
Cortes suggested corpus linguists are still “waiting for the revolution.”
Römer suggests that corpora and corpus tools have not been fully implemented in pedagogical contexts, and current language-teaching materials still differ considerably from actual language use as captured in corpora.
Several popular L2 grammar textbooks make no use of corpus-based research; nevertheless, they are best-sellers. Significantly altering the content and organization of a top-selling textbook runs the risk of hurting future sales.
Change in the grammar textbook industry is likely to be very slow in coming.
The strongest force for change could be a new generation of [L2] teachers who were introduced to corpus-based research in their training programs, who appreciate the scope of the work, and who have practiced conducting their own corpus investigations and designing materials based on corpus research.
The pace of the “revolution” depends on corpus linguists' ability to make research relevant and accessible to L2 teachers and learners.