by Rod Ellis
The results of this meta-analysis provide some support for Skill-Learning Theory. Both types of instruction were beneficial for both receptive and productive knowledge but, overall, comprehension-based instruction was more effective for developing receptive knowledge in the short term, and production-based instruction more effective for productive knowledge in the long term.
All in all, however, there was no clear evidence to support the superiority of either type of instruction. The meta-analysis did not enable Pienemann’s Teachability Hypothesis to be tested. This hypothesis applies only to acquisition of productive knowledge of ‘new’ grammatical features and it was not possible to establish whether the target structures investigated in the studies were ‘new’ or already ‘partially acquired’.
Processing Instruction
I now turn to look in more detail at a sub-set of the input-based studies based on VanPatten’s (1996) Input Processing Principles, which were introduced in Chapter 8. VanPatten argued that it is difficult for learners to attend concurrently to different stimuli in the input and so they need to decide how to prioritize their attentional resources during online processing. Due to limited working memory capacity, they typically resort to a number of default processing strategies captured in Input Processing Principles such ‘Learners prefer processing lexical items to grammatical items—for example, ‘morphological markings’ (p. 14) for semantic information’ and the First Noun principle, i.e. ‘Learners tend to process the first noun or pronoun they encounter in a sentence as the subject/agent’ (p. 33). For VanPatten, two points followed. First, the purpose of instruction should be to help learners overcome these default processing strategies. Second, only grammatical features that are governed by these principles are deemed suitable targets for investigating the effects of Processing Instruction on acquisition. For example, English past tense is a good candidate for instruction because learners are likely to fail to notice the inflection on the verb if the sentence also contains an adverbial marker of past time, as in ‘Yesterday I visited an old friend in London’.
Because the problem learners face in acquiring certain grammatical features relates to perception—i.e. attending to them in input—it follows that what is needed is ‘a type of grammar instruction whose purpose is to affect the ways in which learners attend to input data’ (VanPatten 1996: 2). VanPatten argued that such instruction will prove more effective than traditional production practice—i.e. practice involving text-manipulation activities. While not dismissing production practice, VanPatten (2004a) claimed that it only assists skill development—i.e. it increases control over the use of a grammatical feature in production—and that input alone is sufficient for acquisition.
Processing Instruction, then, aims to help learners overcome their default processing strategies. It consists of three components:
explicit explanation of the target structure
explicit strategy training to enable learners to overcome the default processing strategy
structured input activities.
The most important component is structured input. Lee and VanPatten (2003) proposed the following guidelines for the development of structured input activities:
Present one thing at a time.
Keep meaning in focus.
Move from sentences to connected discourse.
Use both oral and written input.
Keep the learner’s processing strategies in mind.
Structured input consists of both referential and affective activities. The former require the learner to choose between two interpretations of a sentence—for example whether it refers to the ‘present’ or ‘past’—when responding to sentences like ‘Maria visited her sister during the holidays’. The latter require learners to indicate an opinion or belief and so allow for more than one correct answer. Thus referential activities force attention onto the target form, whereas affective activities simply provide learners with input containing the formNOTE 4.
The studies we will now consider examined position (3) above—namely, the claim that Processing Instruction is more effective than production-based instruction in helping learners acquire L2 grammar. In an early study, VanPatten and Cadierno (1993) compared the effects of two instructional treatments, one directed at manipulating learners’ output through production practice, and the other aimed at changing the way the learners perceived and processed input. Learners of Spanish in a university level course, who received input-processing practice relating to Spanish word order rules and the use of clitic pronouns (as in the sentence ‘La sigue el señor’—’Her follows the man’) performed better in comprehension tests than a group of similar learners who received production practice involving both purely form-focused and more meaning-centred activities. This input-processing group also performed at the same level as the production practice group in a production task. This study, then, suggested that Processing Instruction was superior to traditional production practice.
The instruction provided to both groups in this study also contained explicit explanation about the target. A later study (VanPatten and Oikennon 1996) investigated whether the provision of explicit explanation had contributed to the effect of the processing instruction. In this study, there were three groups: (1) received explicit information about the target structure followed by structured input activities; (2) received only explicit information; and (3) just completed the structured input activities. Acquisition was measured by means of both comprehension and production tests. In the comprehension test, significant gains were evident in groups (1) and (3), but not in (2). In the production test, group (1) did better than group (2). VanPatten and Oikennon interpreted these results as showing that it was the structured input, rather than the explicit information, that was important for acquisition. Other studies—for example, Sanz and Morgan-Short (2004) and Benati (2004)—have since replicated these results. However, more recent studies—for example, Fernández (2008); Henry, Culman, and VanPatten (2009)—have shown that explicit information plays an important role at least for some grammatical structures, such as those that are redundant and therefore less noticeable in the input.
There have now been a large number of studies that have investigated Processing Instruction—with and without explicit explanation. Shintani (2015) conducted a meta-analysis of 42 comparative research experiments. The main finding was that Processing Instruction had a greater effect in the receptive tests than in the productive tests, while the opposite was the case for production-based instruction.
Shintani noted that the learners who received production-based instruction fared relatively poorly in the receptive tests. To investigate why this might be the case, she considered the role played by explicit strategy training. She showed that the effectiveness of the Processing Instruction was not influenced by the presence or absence of explicit explanation plus strategy training: structured input by itself proved sufficient, as in VanPatten and Oikennon’s study. However, the presence of explicit strategy training did have an effect in the case of production-based instruction. When Shintani compared those production-based groups who had received strategy training with those that had not, she found that the former scored notably higher in the receptive tests. In other words—while strategy training played a non-significant role in the Processing Instruction—it contributed significantly to the effect that production-based instruction had on receptive knowledge. She concluded that production-based instruction could be as effective as Processing Instruction for receptive knowledge, but only if it included explicit strategy training.
Another issue concerns the durability of the effects of the two types of instruction, as seen in the results for the immediate and delayed tests. Shintani found that the effects of the Processing Instruction in both the receptive and productive tests atrophied over time, suggesting that its impact on acquisition is not long-lasting. She interpreted this result as showing that the knowledge resulting from the Processing Instruction was of the explicit rather than implicit kind, a
s explicit knowledge is more likely to atrophy. The effects of the production-based instruction also reduced over time, but only in the production tests. The receptive test scores did not decline. Shintani speculated that—because the production-based instruction did not typically involve explicit strategy training—the learners did not develop explicit knowledge and so relied on their more resilient implicit knowledge.
There is, in fact, little evidence that Processing Instruction results in the implicit knowledge (or proceduralized explicit knowledge) needed for fluent communication. Very few of the studies included measures of learning based on free oral production. Clearly, there is a need for further research to ascertain whether Processing Instruction has any effect on accuracy in free oral production.
The following are the main conclusions regarding the effectiveness of Processing Instruction:
Processing Instruction is effective. It results in development of both receptive and productive knowledge.
However, its positive effect is more evident in tests of receptive knowledge than productive knowledge.
Also, there is no clear evidence that it leads to the kind of implicit knowledge needed for communicative language use.
In the case of more complex grammatical structures, explicit explanation together with explicit strategy training may be needed, while for simpler grammatical structures, structured input by itself may suffice.
Final comments
This review of the research that has investigated the comparative effects of comprehension-based and production-based instruction lends general support to skill-learning theory—i.e. position (1) above. By and large, comprehension-based instruction facilitates the development of the knowledge needed for comprehending grammatical structures and production-based instruction the knowledge required for producing them. However, the studies also demonstrate that both types of instruction can have an effect on both comprehension and production. This is not surprising, as the two types of instruction may not be as distinct as assumed. It is quite possible that learners receiving comprehension-based instruction engage in silent production practice—i.e. what is termed private speech in sociocultural theory—while the learners receiving production-based practice will also be exposed to input when they listen to other students’ attempts to produce the target structure and to the teacher’s corrective feedback.
DeKeyser and Prieto Botana (2015) noted that the research investigating Processing Instruction is quite limited as relatively few grammatical structures have been examined. From a pedagogic perspective, however, there is sufficient evidence to suggest that structured-input activities are a useful addition to the battery of instructional devices for teaching grammar. As I pointed out in Ellis (2002a), many published grammar teaching materials make little use of such activities.
Pattern practice
Up to now, I have focused on explicit instruction of the deductive kind. Some of the Input Processing studies I considered, however, involved inductive instruction—i.e. there was no explicit explanation of the target structures. These studies indicated that instruction consisting of structured input alone was effective for some structures. I turn now to examine studies that involved inductive production-based instruction of the pattern practice kind.
Pattern practice is closely associated with the Audiolingual Method. This was based on behaviourist learning theories, which view L2 learning as the development of habits, that become ingrained so that learners can automatically use the L2 correctly. Habits are developed by systematically engaging learners in drills of various kinds—for example, repetition, substitution, and transformation—which present learners with stimuli and carefully control their responses to ensure they do not make errors. Mimicry and memorization of these stimuli play a major role.
The effectiveness of the Audiolingual Method was investigated in a number of large-scale studies that compared it with deductive methods involving explicit presentation. The best of these studies was the Gothenberg Project (Levin 1972). In this project, different groups of learners engaged in pattern practice either with or without grammatical explanations. No significant differences between the inductive and deductive groups of school learners were found. In fact, very little learning occurred in either. In the case of an older group of high school students, however, a clearer advantage was found for the deductive instruction. Also, older learners benefitted most from the deductive method. This project, like the other global method comparisons, failed to show that one method was superior to another in terms of overall language proficiency. Short-term differences were sometimes evident, but these disappeared over the long term. A general conclusion is that pattern practice—even when accompanied by explicit explanation—contributes little to L2 learning.
Other studies have investigated whether there is any relationship between the amount of practice that individual learners engage in and L2 learning—i.e. whether learners who engage in more practice learn more. The results are very mixed. In some studies, learners who practised more learnt more. In others, the opposite proved to be the case. In still others, there was no significant relationship. Again, then, there is no clear evidence to suggest that pattern practice is effective. A reasonable conclusion is that practice of the more controlled kind does not make perfect. In fact, sometimes it can even have a deleterious effect on learning by interfering with natural acquisition processes and causing learners to overuse the structure they have practised intensively (Lightbown 1983).
There is a strong theoretical reason for doubting the effectiveness of pattern practice. The Transfer Appropriate Processing Hypothesis predicts that there is a direct relationship between the type of practice and the kind of learning that takes place. Pattern practice of the audiolingual kind positions learners as responders to stimuli and thus is unlikely to prepare them to participate in spontaneous face-to-face communication where they will need to play an initiating as well as a responding role.
Consciousness-raising Instruction
All the types of explicit instruction we have examined so far involve consciousness-raising in the sense that they aim—deductively or inductively—to make learners aware of the properties of the feature that is the target of the instruction. However, in a series of publications (e.g. Ellis 1991; 1993; 2012) I have used the term Consciousness-raising Instruction to refer to instruction based on tasks designed to help learners to construct their own explicit rules about structural features. This type of explicit instruction differs from other types in that it does not include any comprehension or production-practice activities.
The theoretical justification for Consciousness-raising Instruction rests on the claim that explicit instruction of any kind is unlikely to affect implicit knowledge because of developmental constraints (see Chapter 4), but that it can contribute to explicit knowledge, which is not subject to such constraints. The weak-interface position makes two claims: (1) implicit and explicit knowledge are neurolinguistically distinct (Paradis 2009) making the transformation of explicit into implicit knowledge impossible; and (2) explicit knowledge of linguistic features can facilitate the processes involved in the development of implicit knowledge (e.g. noticing and noticing-the gap—see Chapter 8). It follows, then, that instruction should focus on the development of explicit rather than implicit knowledge and that this is justified by the role that explicit knowledge plays in the subsequent, natural development of implicit knowledge. Consciousness-raising Instruction requires learners to engage in languaging (see Chapter 9).
Consciousness-raising Instruction makes use of consciousness-raising tasks—i.e. activities for helping learners construct an explicit representation of a target feature. They consist of data illustrating the use of the target feature and require one or more operations—for example, identification of the feature and rule-formation—that guide learners in analysing the data in order to enable them to arrive at the explicit representation. They do not require production of the feature and—unlike structured-input activities—they do no
t provide practice in input processing. The focus is entirely on representation.
A number of studies have investigated consciousness-raising tasks. A good example is Eckerth (2008). This study investigated university-level learners of German who completed two tasks—a text-reconstruction task and a text-repair task—where the learners worked in pairs to agree on a correct version of a text given to them. Although not framed by sociocultural theory, this study demonstrated the importance of the ‘learner-learner scaffolding’ (p. 102) that occurred while learners performed the tasks as they attended to the ways in which form, meaning, function, and context interrelated. Eckerth also investigated the effects of the tasks in tests designed to provide measures of explicit knowledge of the target features. He reported significant gains between both the pre-test and the immediate post-test, and also between the immediate and delayed post-tests. In other words the learners showed incremental gains across the period of the study.