[note: this post has been updated based on suggestions from Dr. Nina Spada, one of the authors of the study]
Grammar is a divisive word. If you admit you teach grammar, you could be shunned in certain ELT circles. And even for those circles that do accept grammar, the debate still rages about whether it should be taught directly (explicitly) as it has in the past, or indirectly (implicitly), more in line with more modern or “post-modern” methods.
While the debate won’t easily be put to rest, a lot of evidence has come out against implicit grammar instruction as ineffective or less effective than explicit instruction. If you are still on the fence, this 2010 meta-analysis by Nina Spada and Yasuyo Tomita should help you make up your mind.
Spada, N., & Tomita, Y. (2010). Interactions between type of instruction and type of language feature: A Meta‐Analysis. Language learning, 60(2), 263-308.
Explicit more effective than implicit grammar instruction #ResearchBites http://wp.me/p2pCpe-1V6
Introduction and Definitions
Spada and Tomita analyzed a total of 41 separate studies (in 30 publications) between 1990 and 2006. 63% of these studies were based on implicit grammar instruction. The authors calculated effect sizes for each study and compared each group’s average effect size to come to a conclusion about which type of instruction was more effective, on what type of linguistic features, and for how long. The authors categorized studied by instruction type, complexity, and type of knowledge based on the following definitions:
- Instruction Type
- Explicit instruction was defined as any instruction that involved rule explanation, language contrasting, and metalinguistic feedback.
- Implicit instruction was defined as instruction that did not involve rules or attending to any form.
- There are numerous ways to measure complexity. The authors chose linguistic complexity based on the number of transformations a particular form had to go through.
- Simple language features were those forms that included one transformation rule and one or two transformations.
- An example would be article usage, tenses, plurals, etc.
- Complex language features were those that involved multiple transformations.
- Question formation, passive voice
- Outcome Measures
- Declarative knowledge, i.e. knowledge of rules, is knowledge measured by controlled tasks such as metalinguistic judgments (judging whether a sentence is correct or not), multiple-choice tests, scrambled sentences, and “constrained constructed responses” that ask learners to produce an utterance.
- Implicit knowledge, i.e. the spontaneous ability to use a grammatical form, was measured by free writing, oral picture descriptions, information gaps.
The results indicated that explicit instruction had was more effective for both simple and complex language features. In addition, explicit instruction led to both greater explicit* and implicit knowledge. Finally, explicit instruction was also more effective in the long term (as measured by delayed post tests). One result that surprised the authors: the largest effect size in this study was of explicit instruction of complex language features on implicit knowledge (measured by “free constructed response” tasks). Implicit instruction only showed a medium effect size (some effectiveness) for simple language features on free tasks.
The authors point to a few caveats about their findings:
- It is hard to tell if measures of implicit knowledge are really measuring spontaneous production or automatized declarative knowledge.
- If they had looked at complexity in a different way (e.g. pedagogical complexity – how difficult it is to teach a feature), the results may have turned out differently
- The number of studies they included that had delayed post tests were low. More research needs to use delayed post tests.
- As Geoff Jordan pointed out in the comments, the study does not take into consideration the context of the explicit feedback, meaning whether it was included as part of a typical coursebook’s presentation and practice exercises or done in some other way.
- Dr. Spada responded to this in an email: “To clarify, most of the studies included in the meta-analysis included instructional activities and exercises that were developed by the researchers but if I recall correctly, some of them also included exercises from ELT textbooks. In any case, whatever the source of the instructional materials they were all coded in the same way – i.e. whether the instruction involved rule explanation, language contrasting, and metalinguistic feedback or not.”
This article matches quite nicely the post I made a few weeks back about the power of being explicit. Being explicit about learning and language is clearly more beneficial than hoping learners will discover these things on their own. And just because grammar is taught explicitly doesn’t mean it is support for grammar translation, rule lectures, or grammar McNuggets. Rule explanation can come up quite organically in any class, from PPP to Dogme. And, any type of grammar can be made meaningful or fun. Implicit grammar instruction takes longer than explicit instruction to have even a medium effect. So, in my view, don’t beat around the bush. Get straight to the rules and then move on to what’s really important: using the language. As Dr. Spada pointed out to me, “…explicit attention to language form does not exclude attention to meaning/communication/content other than language. Furthermore, most of the research investigating the effects of instruction on L2 learning indicates that a combination of language-based and meaning-based instructions works better than an exclusive focus on either one.”
*Originally, I used the term “declarative knowledge”. However, Dr. Spada pointed out in an email that “declarative” is often contrasted with “procedural”. In the article, they used the term “explicit”: “While declarative knowledge is considered to be the same as explicit knowledge this is not the case with procedural and implicit knowledge. So while you are technically correct that we measured learners’ progress in terms of their declarative L2 knowledge and their implicit knowledge, we used explicit & implicit as the contrasting constructs for L2 knowledge when discussing the findings.”