Current Issues in The Teaching of Grammar: An SLA Perspective
Current Issues in The Teaching of Grammar: An SLA Perspective
Current Issues in The Teaching of Grammar: An SLA Perspective
The study of how learners acquire a second language (SLA) has helped
to shape thinking about how to teach the grammar of a second language.
There remain, however, a number of controversial issues. This paper
considers eight key questions relating to grammar pedagogy in the light
of findings from SLA. As such, this article complements Celce-Murcia’s
(1991) article on grammar teaching in the 25th anniver-sary issue of
TESOL Quarterly, which considered the role of grammar in a
communicative curriculum and drew predominantly on a linguistic theory
of grammar. These eight questions address whether grammar should be
taught and if so what grammar, when, and how. Although SLA does not
afford definitive solutions to these questions, it serves the valuable
purpose of problematising this aspect of language pedagogy. This article
concludes with a statement of my own beliefs about grammar teaching,
grounded in my own understanding of SLA.
T his article identifies and discusses a number of key issues relating to the
teaching of grammar in a second language (L2) and, by drawing on theory
and research in SLA, suggests ways to address these problems.
It points to a number of alternative solutions to each problem, indicating that
more often than not there are no clear solutions currently available. The aim,
therefore, is not to identify new solutions to existing controver-sies, nor even
to present new controversies. Rather it addresses within the compass of a
single article a whole range of issues related to grammar teaching,
problematises these issues, and by so doing, provides a counter-weight to the
advocacy of specific, but also quite limited, proposals for teaching grammar
that have originated in some SLA quarters. However, I conclude with a
statement of my own position on these issues.
The questions that will be addressed are
1. Should we teach grammar, or should we simply create the
conditions by which learners learn naturally?
2. What grammar should we teach?
84 TESOL QUARTERLY
SHOULD WE TEACH GRAMMAR?
This question was motivated by early research into naturalistic L2
acquisition, which showed that learners appeared to follow a natural order
and sequence of acquisition (i.e., they mastered different gram-matical
structures in a relatively fixed and universal order and they passed
through a sequence of stages of acquisition on route to mastering each
grammatical structure). This led researchers like Corder (1967) to suggest
that learners had their own built-in syllabus for learning gram-mar. In line
with this, Krashen (1981) argued that grammar instruction played no role
in acquisition, a view based on the conviction that learners (including
classroom learners) would automatically proceed along their built-in
syllabus as long as they had access to comprehensible input and were
sufficiently motivated. Grammar instruction could con-tribute to learning
but this was of limited value because communicative ability was
dependent on acquisition.
There followed a number of empirical studies designed to (a)
compare the order of acquisition of instructed and naturalistic learners
(e.g., Pica, 1983), (b) compare the success of instructed and
naturalistic learners (Long, 1983) and (c) examine whether attempts to
teach specific grammatical structures resulted in their acquisition (e.g.,
White, Spada, Lightbown, & Ranta, 1991). These studies showed that,
by and large, the order of acquisition was the same for instructed and
naturalis-tic learners (although there were some interesting
differences1), that instructed learners generally achieved higher levels
of grammatical competence than naturalistic learners and that
instruction was no guarantee that learners would acquire what they
had been taught. These results were interpreted as showing that the
acquisitional processes of instructed and naturalistic learning were the
same but that instructed learners progressed more rapidly and
achieved higher levels of profi-ciency. Thus, some researchers
concluded (e.g., Long, 1988) that teaching grammar was beneficial but
that to be effective grammar had to be taught in a way that was
compatible with the natural processes of acquisition.
Subsequent research, such as Norris and Ortega’s (2000) meta-
analysis of 49 studies, has borne out the overall effectiveness of grammar
teaching. Further, there is evidence that, contrary to Krashen’s (1993)
continued claims, instruction contributes to both acquired knowledge (see
Ellis, 2002a) as well as learned knowledge. There is also increasing
1 For example, Pica (1983) notes that some structures (e.g., plural–s) were used more
accurately by instructed learners and some (e.g., Verb–ing) by naturalistic learners. In other
structures (e.g., articles) there was no difference.
86 TESOL QUARTERLY
Freeman’s (1999) Grammar Book. This resource is especially valuable
because it not only provides a comprehensive, clear, and pedagogically
exploitable description of English grammar but also identifies the kinds of
errors that L2 learners are known to make with different grammatical
structures. Such information is important because it helps to identify which
structures and which aspects of a structure require special attention. The
Grammar Book is also ideal in that it presents information not only about
linguistic form but also about the semantic and discoursal meanings
realised by particular forms. As VanPatten, Williams, and Rott (2004)
emphasise, establishing connections between form and meaning is a
fundamental aspect of language acquisition. Thus, any reference grammar
that fails to describe the form-meaning connections of the target language
must necessarily be inadequate. In general, then, the choice of which type
of grammar to use as a basis for teaching is not a major source of
controversy; descriptive grammars that detail the form-meaning
relationships of the language are ascendant.
In contrast, the choice of which grammatical structures to teach is
controversial. Two polar positions can be identified and various positions
in between. At one end of this continuum is Krashen’s minimalist position.
Krashen (1982) argues that grammar teaching should be limited to a few
simple and portable rules such as 3rd person–s and past tense–ed that
can be used to monitor output from the acquired system. He bases his
argument on the claim that most learners are only capable of learning
such simple rules—that more complex rules are generally not learnable or,
if they are, are beyond students’ ability to apply through monitoring.
Krashen’s claim, however, is not warranted. There is now ample evidence
that many learners are capable of mastering a wide range of explicit
grammar rules. Green and Hecht (1992), for example, found that
university-level students of English in Germany were able to produce clear
explanations for 85% of the grammatical errors they were asked to
explain, while overall the learners in their study (who included secondary
school students) managed satisfactory explanations for 46% of the errors.
Macrory and Stone (2000) reported that British compre-hensive school
students had a fairly good explicit understanding of the perfect tense in
French (e.g., they understood its function, they knew that some verbs
used avoir and some être, they were familiar with the forms required by
different pronouns, and they were aware of the need for a final accent on
the past participle). Hu (2002) found that adult Chinese learners of English
demonstrated correct metalinguistic knowl-edge of prototypical rules of six
English structures (e.g., for the definite article specific reference
constituted the prototypical rule) but were less clear about the peripheral
rules for these structures (e.g., generic reference). At the other pole is the
comprehensive position: Teach the whole of
2Of course, it is not possible to specify the whole grammar of a language. Though the grammar
of a language may be determinate, descriptions of it are certainly not. The Longman A Grammar of
Contemporary English (Quirk, Greenbaum, Leech, & Svartvik, 1972) ran to 1081 pages (excluding
index and bibliography) but doubtlessly does not account for all the known facts of English
grammar. Nevertheless, there is a recognized canon of English structures that, in the eyes of
syllabus designers and textbook writers, constitutes the grammar of English.
3 Structures like English articles that are very frequent in the input can impose
considerable learning difficulty. Structures such as English conditionals may be very useful to
learners but are also difficult to learn.
88 TESOL QUARTERLY
The first approach was, of course, the one adopted in many early
structural courses based on a contrastive analysis of the learner’s L1 and
the target language. Although the contrastive analysis hypothesis as
initially formulated is clearly not tenable (see Ellis, 1985, chapter 2), SLA
researchers still generally agree that learners transfer at least some of the
features of their L1 into the L2. For example, there is ample evidence
(Trahey & White, 1993) to show that French learners of English produce
errors of the kind Mary kissed passionately John because French permits
an adverb to be positioned between the verb and the direct object.
Nevertheless, contrastive analysis does not constitute a sound basis for
selecting grammatical structures. In many teaching contexts, the learn-ers
come from mixed language backgrounds where it would be impos-sible to
use contrastive analysis to tailor grammar teaching to the entire group
because the learners have different L1s. Also, we simply do not yet know
enough about when difference does and does not translate into learning
difficulty, and in some cases, learning difficulty arises even where there is
no difference.
The second approach, however, is also problematic. Markedness has
been defined in terms of whether a grammatical structure is in some
sense frequent, natural, and basic or infrequent, unnatural, and deviant
from a regular pattern (Richards, Platt, & Weber, 1985). Thus, the use of
an infinitive without to following make, as in He made me follow him can
be considered marked because make is one the few verbs in English that
takes this kind of complement and because this pattern occurs only
infrequently. The general idea is that we should teach the marked features
and leave the learners to learn the unmarked forms naturally by
themselves. The problem is that, as the definition suggests, markedness
remains a somewhat opaque concept, so that it is often difficult to apply
with the precision needed to determine which structures to teach.
The selection of grammatical content, then, remains very problem-
atic. One solution to the kinds of problems I have mentioned is to base
selection on the known errors produced by learners. In this respect,
lists of common learner errors such as those available in Turton and
Heaton’s (1996) Longman Dictionary of Common Errors and Swan and
Smith’s (2001) Learner English: A Teacher’s Guide to Interference and
Other Problems are helpful.
The problems of selection probably explain why grammatical sylla-
buses are so similar and have changed so little over the years; it is
safer to follow what has been done before. Of course the selection of
what to teach will also depend on the learner’s stage of development.
The problems that the learner’s stage of development involve are
discussed in subsequent sections.
90 TESOL QUARTERLY
(e.g., clauses where the relative pronoun functions as subject and the
clause is attached to a noun phrase following the verb). There is ample
evidence to show that learners can and do learn a good deal of grammar
without being taught it. This being so, why bother to teach what can be
learned naturally? A second reason for delaying grammar teaching to later
stages of development is that early interlanguage is typically agrammatical
(Ellis, 1984; Perdue & Klein, 1993). That is, learners rely on a memory-
based system of lexical sequences, constructing utterances either by
accessing ready-made chunks or by simply concatenating lexical items
into simple strings. Ellis (1984) gives examples of such utterances in the
early speech of three classroom learners:
Collins and colleagues then report their own study of three intensive ESL
programmes in Canada, one (the distributed programme) taught over the
full 10 months of one school year, one (the massed programme)
concentrated into 5 months but taught only to above average students,
and the third (the massed plus programme) concentrated into 5 months,
supplemented with out of class opportunities to use English and taught to
students of mixed ability levels. The main finding was that the massed and
especially the massed-plus students outperformed the distributed
programme students on most of the measures of learning, including some
measures of grammatical ability, although this finding might in part be
explained by the fact that the massed programmes provided more overall
instructional time.
Collins et al.’s study points to the need for further research, especially
through studies that compare massed and distributed instruction di-rected
at specific grammatical structures. Ideally such a study would compare
short periods of instruction in a particular structure spread over several
days with the same amount of instruction compressed into one or two
lessons.4 Received wisdom is that a cyclical approach to grammar
teaching (Howatt, 1974) is to be preferred because it allows for the kind of
gradual acquisition of grammar that is compatible with what is known
about interlanguage development. However, the results of
4 Given the problems that arise in controlling extraneous variables in evaluations of entire
programmes, it might prove much easier to conduct rigorous studies of massed and
distributed learning when these are focused on specific grammatical structures.
92 TESOL QUARTERLY
Collins et al.’s study suggest, at the very least, that such a position
needs to be investigated empirically. Here, then, is an issue about
which nothing definitive can be said at the moment.
94 TESOL QUARTERLY
between the preterite and imparfait past tenses after hours of exposure
(and presumably some corrective feedback) in an immersion
programme but were able to improve their accuracy in using these two
tenses after intensive instruction. However, intensive instruction is time
consuming (in Harley’s study the targeted structures were taught over
a 6-month period), and thus, time will constrain how many structures
can be addressed. Extensive grammar instruction, on the other hand,
affords the opportunity to attend to large numbers of grammatical
structures. Also, more likely than not, many of the structures will be
addressed repeatedly over a period of time. Further, because this kind
of instruc-tion involves a response to the errors each learner makes, it
is individu-alized and affords the skilled teacher real-time opportunities
for the kind of contextual analysis that Celce-Murcia (2002)
recommends as basis for grammar teaching. However, it is not
possible to attend to those structures that learners do not attempt to
use (i.e., extensive instruction cannot deal effectively with avoidance).
Also, of course, it does not provide the in-depth practise that some
structures may require before they can be fully acquired.
Arguably, grammar teaching needs to be conceived of in terms of both
approaches. Therefore, grammar teaching needs to be reconceptualised
in teacher handbooks to include the kind of extensive treatment of
grammar that arises naturally through corrective feedback.
96 TESOL QUARTERLY
explicit and implicit memories are neurologically separate (Paradis, 1994).
The interface position argues the exact opposite. Drawing on skill-learning
theory, DeKeyser (1998) argues that explicit knowledge be-comes implicit
knowledge if learners have the opportunity for plentiful communicative
practice. The weak interface position (Ellis, 1993) claims that explicit
knowledge can convert into implicit knowledge if the learner is ready to
acquire the targeted feature and that this conversion occurs by priming a
number of key acquisitional processes, in particular noticing and noticing
the gap (Schmidt, 1990). That is, explicit knowledge of a grammatical
structure makes it more likely that learners will attend to the structure in
the input and carry out the cognitive comparison between what they
observe in the input and their own output. These positions continue to be
argued at a theoretical level. Although there is plentiful evidence that
explicit instruction is effective in promoting L2 learning (e.g., Norris &
Ortega, 2000) no published study has directly tested whether explicit
knowledge converts directly into implicit knowl-edge or simply facilitates its
development. One reason for the lack of research is the problem of
measurement, that is, the difficulty of ascertaining which type of
knowledge learners employ when they per-form a language task or test.
98 TESOL QUARTERLY
out of social interaction which scaffolds learners’ attempts to produce
new grammatical structures (Ohta, 2001). A number of studies have
compared the relative effectiveness of input-based and production-
based instruction, with mixed results, resulting in ongoing debate about
the relative merits of these two options (VanPatten, 2002; DeKeyser,
Salaberry, Robinson, & Harrington, 2002). It may be that, in
classrooms, this comparison is ultimately meaningless because, in
practise, both options are likely to involve input-processing and
production. For example, it is quite conceivable that in an input-based
approach, individual students silently produce the target structure,
while in a production-based ap-proach, an utterance produced by one
student serves as input for another. It is, therefore, not surprising that
both options have been shown to result in acquisition. 6
There is a rich descriptive literature on corrective feedback (i.e., teacher
responses to learner errors) but remarkably few studies have investigated
the relative effects of different types of feedback on acquisi-tion. Key
options are (a) whether the feedback is implicit or explicit and
(b) whether the feedback is input or output based. Implicit feedback
occurs when the corrective force of the response to learner error is
masked, for example, a recast, which reformulates a deviant utterance
correcting it while keeping the same meaning:
6 There is also controversy regarding how to measure the effectiveness of these two (and
other) instructional options. Norris and Ortega (2000) have shown that the effectiveness of
instruction varies depending on whether it is measured using metalinguistic judgements,
selected response, constrained constructed response, or free constructed response. Most
SLA researchers (and teachers, too, perhaps) would consider the last of these the most valid
measure. Ellis (2002a) reviewed a number of studies that examined the effects of different
kinds of instruction on learners’ free constructed responses, reporting that instruction can
have an effect on this type of language use.
CONCLUSION
Grammar has held and continues to hold a central place in language
teaching. The zero grammar approach was flirted with but never really
THE AUTHOR
Rod Ellis is a professor in the Department of Applied Language Studies and
Linguistics at the University of Auckland, New Zealand. He has published widely in
the field of SLA. His latest books are Analyzing Language Learning and Planning
and Task Performance in a Second Language.