Controlled Conditions

Despite their origins in classroom practice, information gap tasks used in research have been implemented primarily under conditions more typical of a controlled environment than a classroom setting. Research conditions have included monitored sessions outside the classroom (Long, 1985; Pica, Holliday, Lewis, Berducci, & Newman, 1991; Pica, Holliday, Lewis, & Morgenthaler, 1989; Pica et al., 1996; Pica, Young, & Doughty, 1987), special researcher visits to actual classrooms (Pica, 1991), and breaks during regular class meetings (Doughty & Pica, 1986). Some studies have gathered data by implementing tasks in authentic classrooms, but as extracurricular activities added on to the regular classroom agenda (Newton & Kennedy, 1996; Williams & Evans, 1998). Data have also been collected during individual, small group, or interview sessions controlled by a researcher, often on a short-term basis. Such practices guarantee uniform delivery of task treatments across multiple participants, with consistent timing in task implementation. The research design can be maintained with reliability and consistency, so that task directions are followed carefully and tasks are implemented uniformly across different learners and within a comparable time frame. These approaches have shed light on the processes and outcomes of what is often referred to as instructed SLA (DeKeyser, 2003; Doughty, 2003).

Data collected through the use of information gap tasks under controlled conditions have been highly reliable and informative (particularly with respect to theoretical questions). However, concerns have been raised about the authenticity of the data with respect to the term instructed SLA (e.g., Byrnes, 2000). First, the learners who have provided the data are, first and foremost, subjects in a study and only secondarily are they students in a classroom. Additionally, teachers have questioned the usefulness and relevance of the data to actual classroom life (Pica, 1994). Also, as both DeKeyser (2003) and Doughty (2003) noted, studies implemented under controlled conditions lack the ecological validity that research during actual class time can offer.

These concerns have yet to be resolved and might better be addressed in sociolinguistic projects than in SLA research. Many interventions designed to promote SLA can be considered instructional if the treatment they offer is characterized by feedback, explanation, or pedagogical moves associated with instruction (e.g., Krashen & Seliger, 1975). These features lend authenticity to the use of such interventions in research on instructed SLA outside of the classroom context. However, it is important to think of instructed SLA in the classroom as a distinct, multifaceted experience, as learners come together to learn an L2, meeting collectively and often frequently over a period of time. Despite the familiarity such a context can offer, classroom learners are often more reluctant to respond to checks on their comprehension or to seek clarification when they interact with their teachers than when they interact with interlocutors outside of the classroom (Pica, 1987). The classroom provides a distinct context for the study of instructed SLA, and the information gap task can play a vital role in addressing questions about the interactions and interventions that occur there. The meaning-focused classroom appears especially relevant in this regard.

Meaning-Focused Contexts

Although it has been shown that much of an L2 can be learned through focus on meaning, there is increasing evidence that learning that results from a focus on meaning is incomplete and that grammatical imprecisions remain (Lightbown & Spada, 1999; Swain, 1985, 1991). As Harley (1993) and Long (1996) discussed, this is especially apparent for forms whose encodings of meaning are linguistically complex or limited in their frequency of occurrence, perceptual salience, or functional transparency. Several researchers have shown that meaning-focused content classrooms do not offer students many opportunities to produce and modify their output morphosyntactically (Swain, 1985, 1991) or to be given feedback on their grammatical accuracy (Pica, 2002; Pica & Washburn, 2003; Pica, Washburn, Evans, & Jo, 1998; Shah, 1999; Swain, 1991). Evidence to this effect has been found in observations and case studies of high school classes in math, history, and life science (Shah) and in descriptive studies of university-level English language courses on literature and culture (Boyd-Kletzander, 2000) and on film and culture (Pica; Pica & Washburn; Pica et al.).

The predominant discourse in the high school classrooms was characterized by an interactional pattern in which teachers asked questions or requested information, students supplied answers, and teachers gave follow-up responses. Typically, the teachers responded to the students by modifying their output for them; that is, they often elaborated on students' responses without requiring the students to modify their own output. In (2), taken from a high school science class, the teacher elaborated upon the student's answer of “lemon” by adding the names of other acidic foods, commenting on them, and ending with a yes/no question. The teacher's response supplied the class with a sample of L2 forms, their functions, and their meanings. However, it failed to focus attention on the single-word answer “lemon” and encourage the students to elaborate it linguistically. Very appropriately, therefore, the teacher's question was followed up with a simple response of “yes.”

During discussion—another typical meaning-focused activity—teachers mainly provide samples of L2 input, even when negative feedback that focused on students' grammatical imprecisions might have been a more helpful response. In (3), from a university English L2 course on American culture and film, the teacher's backchannels and positive feedback during discussion of a review of the film Stand and Deliver (Ebert, 1990) likely sent students the message that their production was accurate, despite their many errors. (These errors are highlighted in italics.)

These data reflect teachers' instructional objectives for content review, reporting, and information exchange, and they comply with the interactional norms of their lessons and discussions. As such, they reveal greater attention to the meaning of students' contributions than to their form. In the science lessons, students were asked to supply specific items in response to their teacher's content questions. On the basis of their answers, the teacher elaborated on the students' output, although the students themselves were not encouraged to do so.

In the film discussions, students were encouraged to elaborate. However, linguistic imprecisions were overlooked when they did not interfere with students' attempts to display their knowledge and convey their ideas. Although there was a good deal of negotiation for meaning, the focus of the discussion was on defining unfamiliar lexical items and clarifying factual information rather than on calling attention to grammatical errors. As the data revealed, articles and pronouns, modal verbs, and bound inflections for tense and aspect were seldom the focus of classroom negotiation. Numerous contexts for their suppliance were generated, as discussion topics required students to refer to multiple individuals, concepts, and conditions, to structure information into arguments, and to connect issues and concepts. However, the absence or imprecision of these forms was seldom acknowledged or addressed. When asked to reflect on their practice, the teachers said that they were guided by theories that emphasized meaningful, comprehensible input, language learning through communication (e.g., Shah, 1999), and whole-language philosophies (e.g., Boyd-Kletzander, 2000). Their responses to the students, they believed, were consistent with their theoretical perspectives on SLA.

The teachers' earnest intentions and theoretical grounding were offset by several concerns that remained unaddressed in their classroom practice. First, in terms of theory, the roles of output and form-focused feedback in SLA were not taken into account. As argued in the output hypothesis (Swain, 1985, 1991, 2005), meaningful input and communicative experiences need to be accompanied by opportunities for learners to produce and modify their output if they are to become more syntactically accurate. Additionally, learners need questions and feedback on imprecisions of form. Such signals challenge them to move beyond L2 development that is fluent and meaningful but filled with grammatical inaccuracies. As research has shown, it is not easy for learners to both attend to formal features of language and acquire new information in a content area (e.g., VanPatten, 1990; Wong, 2001). Second, in terms of classroom practice, the challenge of attending to formal linguistic features and acquiring new content knowledge can be especially acute when the content must be learned as a school requirement or is preferred for its inherent interest. It cannot be assumed, therefore, that learners will somehow pick up new forms—especially if the forms are barely perceptible—as they attend to content meaning. There is a need for activities, materials, and strategies that can be applied simultaneously to meet these needs. Information gap tasks, in their fundamental structure and purpose, show great promise in these areas.

With these concerns and needs in mind, a team of teachers and researchers has been working to develop form-focusing information gap tasks that provide data that are both reliable and useful to the study of SLA. Our work has been carried out primarily in meaning-focused classrooms of university and community programs for adult English language learners who have both academic and conversational needs and goals. However, we have also worked with teachers and researchers in heritage and foreign language classrooms that emphasize meaning and subject matter content in their curricula (Kang, 2006). An ongoing collaboration for task development has also been initiated with the university directors and teaching staff who developed the Language through Film course excerpted in (3).

The following sections describe three types of information gap tasks developed to serve two purposes: (a) as research treatments that encourage learners to engage in oral interaction and orient their attention toward forms whose relation to function and meaning are low in salience and therefore difficult to acquire and (b) as research tools for the collection and analysis of data on these processes. This is followed by an overview of a study on the attentional and interactional processes that were revealed as learners carried out tasks whose completion depended on low-salience forms. Data from the study are then presented to address the research questions and to show how the tasks served as a research method.

Research Treatment

Creating Conditions for Modified Interaction and Attention

Three types of information gap tasks appear to be especially well suited to serve as research treatments and instruments through which learners can be given modified input and feedback, produce modified output, and notice L2 forms. As such, the jigsaw (Doughty & Pica, 1986; Pica, 1991; Pica et al., 1989; Swain & Lapkin, 2001), spot-the-difference (Crookes & Rulon, 1988; Long, 1980, 1981), and grammar communication (Fotos, 1994; Fotos & Ellis, 1991; Loschky & Bley-Vroman, 1993) tasks have been shown to promote attentional processes and interaction among learners. Because learners proceed through the same steps to reach the goals of task reconstruction and text comparison, the tasks share a similar organizational structure and goal requirement. This consistency extends both within and across the task types, as their implementation creates conditions for learners working in pairs to modify their interaction and attend to form, function, and meaning. The relationship between task steps and attentional and interactional processes is displayed in Table 1.

Attentional and interactional processes across task steps

As shown in Table 1, the learners proceed in pairs through five steps for each task. Each step appears on a separate page and each learner has an individual packet of pages. At the bottom of each page, they are told to “Go to the next page and don't look back.” In step 1, they read a passage based on any one of the following sources: a previously read text, a prior discussion, or a meaningful experience from an earlier session. In step 2, the learners each read a slightly different version of the original passage without revealing their respective versions to each other. The two versions differ in that each sentence is either identical to the original passage or has a phrase in which a form with low salience from the original passage appears in a different order or with a slightly different encoding. There are no truly ungrammatical phrases used in either version (i.e., formations such as a books or wented), but, rather, there are formations inconsistent with passage meaning or with the original passage read by the learners.

In step 3, the learners compare their passage versions aloud as they choose between the targeted phrases or between the sentences that contain the phrases and then justify their choices. For the spot-the-difference task, they choose between nearly identical sentences. In the jigsaw task, they first reorder the sentences to match the passage and then choose between them. In the grammar communication task, they choose among four nearly identical phrases, one of which has been excerpted from the original passage sentences. Before they can make their choice, they pool the two phrase options each has been given.

In step 4, without looking back at their choices or the passages they have read, the learners work together to write their chosen forms for the low-salience items in a single cloze version of the original passage. Finally, they turn to step 5, in which they reread the original passage, compare it with their cloze version, identify any discrepancies, and posit explanations for them. Learners' participation in all five steps can activate their attentional processes for SLA. However, their participation in steps 3–5 is especially well suited to providing spoken and written data in which these processes can be identified.

Many of these attentional processes were introduced to the field of SLA by Gass (1997), Leow (1997), Robinson (1995), Schmidt, (1993), and Tomlin and Villa (1994) through the construct of noticing, a process that has been sustained, expanded, and further defined through their research and writing to date. Tomlin and Villa proposed a model of attention that included components of alertness, orientation, and detection as they pertained to the learner's access to SLA data. Alertness referred to learners' readiness to select incoming data for further processing. Their orientation directed them to particular parts of the data, and detection—referred to by others as noticing (e.g., Robinson)—registered the data in short-term memory, thereby making it available for higher levels of processing such as hypothesis formation and testing.

This view of noticing was expanded by Robinson (1995), who situated noticing in the process of awareness, through which learners encoded and retrieved L2 data for use during task-related interaction. Leow (1997) also addressed noticing in this way, defining it as the learner's awareness of new forms as they encode L2 data. Thus, according to these researchers, noticing plays a crucial role in holding on to L2 data in the short term and in making the data available for further processing over the long term. Awareness reflects a deeper understanding of its form, function, and meaning.

These perspectives on noticing and awareness are shared by Gass (1997) and Schmidt (1993, 2001), who also looked to noticing to form a link between the processing of incoming L2 data and its conversion first into input and later into intake. By extension, noticing enables learners to recognize input deviations from L2 norms or, in other words, input that differs from their current interlanguage repertoire. These occurrences, which constitute noticing the gap (see also Schmidt & Frota, 1986), can lead learners to restructure their current interlanguage system. Gass further proposed apperception as an initial step that precedes noticing. Before learners notice the gap between the L2 data and their knowledge and production capability, they must first perceive and then relate the L2 data to their existing knowledge.

The importance placed on noticing and awareness has guided both the design of the current tasks and the analysis of the interaction data from their implementation. The different ways in which the tasks can draw learners' attention to the forms that encode word, phrase, and sentence function and meaning have led to further distinctions among noticing an individual form, noticing a difference between forms, noticing a discrepancy between a deviant form and its L2 counterpart, and noticing the relationship between a form and its function or meaning. Thus, in step 3 of the tasks, the need to locate, compare, and then choose between phrases and sentences sets up conditions for noticing a form as an item unto itself as well as for noticing differences among the forms that encode function and meaning in these phrases and sentences. These experiences are consistent with the views on noticing described previously. Learners might mention a form during their reading of the task passages and their choosing between passage sentences and phrases; this is a way to alert each other that they have identified the form, which is an early step toward task completion.

Learners' noticing of form has been signaled in various ways—for example, by underlining the target form in a passage (Izumi & Bigelow, 2000), inclusion of the form in a text reconstruction activity (Izumi, 2002), reference to the target form in a learning journal (McDonough, 2005), or during think-aloud verbal reports (Rosa & O'Neill, 1999). However, because the current tasks require learners to interact, they provide an opportunity to study noticing of form and form differences through the learners' own verbalizations. The tasks also allow for the possibility that learners will notice the gap in the accuracy and appropriateness of the sentences they choose and those they reject. The ability of language learners to notice this gap has been investigated in the literature on recasts and includes the learners' ability to correctly identify the source of error that prompts the recasts (Mackey, Gass, & McDonough, 2000), their ability to recognize the corrective intent of teacher-generated recasts (Nabei & Swain, 2002), and their accuracy in immediate recall of recasts (Philp, 2003). As a result of the tasks' emphasis on interaction, as the learners deliberate over and justify their choices, they might also modify their interaction to explain and clarify their arguments and use negative feedback such as correction and recasting for what they believe are each other's incorrect choices (although none of the phrases provided to the learners were in fact ill-formed). Such verbal behavior encourages further noticing of the perceptual features of a particular form and builds awareness of its relation to function and meaning, which contributes to the internalization process. This noticing of form, function, or meaning is related to Rosa and O'Neill's awareness at the level of understanding, defined as the learners' articulation of rules that govern a targeted structure during think-aloud verbal reports. Thus, as the learners notice the gap between correct and incorrect form uses and indicate awareness of form in relation to function and meaning, they demonstrate further processing of the form and a readiness for its recall in the next task step.

In step 4, as the learners recall and write their choices in a single cloze version of the original passage, they are given opportunities for modified interaction and negative feedback because mutual comprehension and agreement are necessary. With respect to attention, this phase of the task encourages learners to recall or retrieve their earlier choices—an experience that researchers have claimed reveals further evidence for the different kinds of noticing. This is reminiscent of Robinson (2003), who determined what was noticed in terms of what the learner was able to verbally report.

Step 5, with its emphasis on comparison and explanation, provides a context for conditions such as those of step 3. However, the conditions are contingent on the degree of consistency between learners' earlier decisions about the phrases and sentences in steps 3 and 4 and the text of the original passage. If they are able to achieve a complete match, there is no need for them to do much more than acknowledge this step and conclude the task. Should discrepancies exist, the need to identify and explain them could activate interactional processes as well as attentional ones, particularly those related to noticing the gap.

The task development process is illustrated here using a review of the film Philadelphia (Renshaw, 1994); the tasks shown in this section are a subset of those used in the small-scale study of interaction and attention that will subsequently be reported. After viewing and discussion of the film, this brief passage from the longer movie review is given to learners:

“Philadelphia” opens strongly with an effective collection of city scenes. We see Andrew as a typical workaholic attorney, already living with HIV as part of his life. HIV from the outset is simply one element in Andrew's life, but not the defining element. Joe Miller is equally well-established, both at work and at home. Nevertheless, he recognizes the social similarities between being a black man and being a person with AIDS.

This passage contains numerous contexts for low-salience noun and verb forms, their functions, and their meanings. There is seldom a need to enrich such meaningful passages with low-salience forms, as contexts for them are abundant. However, the passages can occasionally be modified to streamline sentence complexity, reduce paragraph length, or eliminate allusions to experiences and concepts unfamiliar to learners and teachers.

Task directions begin with a purpose statement. For the first sample task, spot-the-difference, the statement tells the learners that they will become more accurate and precise in their speaking and writing in areas such as organizing, reporting, reviewing, and editing information. A table that displays and compares the directions for steps 1–5 is found in the Appendix. As shown, the directions were identical, with slight variations within step 3.

Table 2 displays the slightly different versions of the text given to the learners. The first sentence is the same as in the original passage. Differences begin with sentence 2, as highlighted. Three different sets of texts were prepared, each of which manipulated one of the following targets: articles and determiners, pronouns and connectors, or verb and modal morphology. Each pair of learners worked through only one version (i.e., targeting articles and determiners, pronouns and connectors, or verb and modal morphology) of each task (i.e., jigsaw, spot-the-difference, grammar communication). For the spot-the-difference task, for example, pairs 1 and 2 carried out the article and determiner version, whereas pairs 3 and 4 carried out the version for connectors and pairs 5 and 6 completed the version for modals and verbal morphology. During the following session, all groups completed a jigsaw task: Pairs 1 and 2 carried out the version for verbs, pairs 3 and 4 did the version for articles and determiners, and pairs 5 and 6 completed the version for connectors. As mentioned previously, there are no truly ungrammatical phrases used in either version, only formations that are inconsistent with passage meaning or with the original passage read by the learners.

Three versions of a spot-the-difference passage

Table 3 displays the passages given to the learners in steps 4 and 5. In step 4, learners are given the cloze version of the original passage and must complete the task together. In the final step, learners are given the original passage with the items that had been excerpted in the cloze passage underlined and they compare this version with the cloze passage that they completed in step 4.

Passages for steps 4 and 5 for articles and determiners

Both the jigsaw and grammar communication tasks ask learners to follow a set of directions very similar to those of the spot-the-difference task. Learners are instructed to choose between phrases that contain the forms they need to learn, justify their choices, recall them in a cloze activity, and then compare their cloze version with the original passage and explain any differences that they find. To maintain authenticity, the purpose statement given for the jigsaw task tells the learners that the task will help them organize information, whereas for the grammar communication task, they are told that they will be helped to report information accurately.

For the jigsaw task, each member of the participant pair is provided with an unordered set of individual sentences; each set contains some sentences taken directly from the original passage and others that have been modified (as illustrated for the spot-the-difference task). Using the sentences provided, learners are asked to carry out the choosing step (step 3) in two parts. First, they are to choose the order of individual sentences as they appeared in the original passage. Then they are to choose the version of each sentence that they recall from the original movie review passage, much as they did for spot-the-difference. These two components of choosing are designed to activate slightly different attentional processes. When choosing sentence order, the learner's noticing of forms, form differences, and gaps is incidental to the choice; in choosing between different sentences, such noticing is implicit but nonetheless more directly related to the choice. Versions A and B of a sample passage are shown in Table 4. Their sentences differ slightly in articles and determiners.

Jigsaw passage versions for articles and determiners

In the grammar communication task, learners again follow directions that are nearly identical to those of spot-the-difference. However, in this case they choose among phrases that contain specific forms or features for cloze reconstruction of the review passage. Table 5 illustrates Versions A and B of a grammar communication task.

Grammar communication task for articles and determiners

Developing Tools for Data Collection

In designing tasks as research instruments on learner interaction in SLA, care must be given to ensure that a large amount of data can be collected and that the data will be a valid representation of learners' interactions and their attention to the relationships of form, function, and meaning that need to be further developed. The materials as well as their instructions and maintenance need to satisfy researchers' goals for both immediate data collection and eventual application to SLA theory. If these materials are to be used longitudinally in a classroom environment, task instructions need to be straightforward, to allow students to carry them out without continuous involvement of their teacher. Teacher and researcher involvement in design, piloting, and revision of the tasks and the writing of instructions is essential to the success of this effort. Consistency across task instructions is of paramount importance if the data are to be pooled for later analysis.

Access to the different task types ensures that learners can work on multiple forms, their functions, and their meanings, and it provides researchers with ample data for analysis. In a given week, learners might engage in a meaning-focused activity, such as a discussion, during their first session and then spend the following days reading a text passage based on the discussion and completing its companion spot-the-difference task. In a subsequent session, learners might read and discuss a different passage and, finally, complete its companion jigsaw or grammar communication task.

Learners often tend to work independently, and it is important to gather data on their oral interactions and displays of attention even during those times. The tasks do this in several ways. First, learners are not allowed to show each other their passage versions, so oral interaction is ensured as they make their choices and justify their answers. Second, they are asked to share the same page to complete the cloze step of the task, which again ensures against parallel private work. Finally, they are asked to write both their responses to the cloze passage and their identification of differences between their choices and those of the original passage. Although this precaution limits the speaking requirement, the written responses provide a record for further analysis.

Instrument for Data Analysis

If a task is designed to activate attention to form, function, and meaning and to require spoken and written interaction throughout its implementation, it can yield a rich source of data for analysis of the relationship of attentional and interactional processes as they bear on SLA. Such data also eliminate or greatly reduce the need to rely on follow-up interviews or introspective data for insight into attentional processes.

Method

Tasks

Each of three task types was created using passages taken from three different reviews (Brenner, n.d.; J. Ellis, n.d.; Hicks, 1994) of the movie Philadelphia (Demme, 1993). All review passages already contained numerous examples of the targeted low-salience forms. As a result, they needed only minor modification, mainly to ensure uniformity of length (13 sentences). Three versions of each passage were created to accommodate the three different form categories: one for articles and determiners, another for pronouns and connectors, and another for verb features. A second goal was to gather a large amount of data on the different forms in the passages.

Data were gathered from each of the three versions of each task type (grammar communication, spot-the-difference, and jigsaw). This range allowed the researchers to study task implementation and classroom compatibility and to track attentional and interactional processes associated with low-salience features across the tasks. Thus, for example, on day 3, all six pairs read the same passage for the grammar communication task, but pairs 1 and 2 completed the version whose passages were modified for articles and determiners, pairs 3 and 4 completed the version modified for pronouns and connectors, and pairs 5 and 6 completed the version modified for verb endings and modals. The implementation of tasks is shown in Table 6.

Task implementation matrix

Results and Discussion

Jigsaw, spot-the-difference, and grammar communication task findings for all pairs of participants are shown in Tables 7 and 8. The attentional processes (Table 7) and the interactional processes (Table 8) are broken down by specific process, task type, and step (3–5). For steps 3–5, these tables present token counts and percentages of the sentence or phrase decisions the six pairs needed to make in choosing, recalling, and comparing the forms of the 12 sentences in their task passages. Therefore, the base number used for determining the percentage scores for each step was 72 for the jigsaw and grammar communication sentences. Because only five of the pairs recorded their spot-the-difference implementation, their number of decisions was 60. The percentage distribution of attentional and interactional processes for each step are displayed in Figures 1, 2, and 3. The data were coded by the researchers for these attentional and interactional features. Interrater reliability was .95 for attentional features, .91 for interactional features, and .99 for recall scoring. Steps 1 and 2, which involved silent reading, did not allow for collection, coding, or analysis of attentional or interactional data.

Frequencies and distribution of attentional processes used in sentence and phrase decisions across task steps

Frequencies and distribution of interactional processes used in sentence and phrase decisions across task steps

Distributions of attentional processes in pairwork decisions across task steps in jigsaw (J), spot-the-difference (S), and grammar communication (G) tasks.

Distributions of awareness in pairwork decisions across task steps.

Distributions of interactional processes in pairwork decisions across task steps in jigsaw (J), spot-the-difference (S), and grammar communication (G) tasks.

The first research question asked about the ways in which the learners' task implementation assisted their attention to forms with low salience. First, as revealed in Table 7 and displayed in Figure 1, each task engaged the pairs in the three noticing processes that were totaled to give a score for total noticing. As shown in the “Total” column, there were 116, 128, and 108 decisions that involved the passage sentences and phrases in which the forms were noticed for the jigsaw, spot-the-difference, and grammar communication tasks, respectively. This reflects the fact that making a decision in choosing, recalling, or comparing sentences often featured two or more noticing processes. As displayed in Figure 1, noticing was especially prominent during step 3, the step that required a choice. Thus, 85% of the pairs' total noticing occurred during step 3 for the jigsaw task, 96% for the spot-the-difference task, and 92% for the grammar communication tasks.²

The consistency across percentages for step 3 was offset somewhat by distributional differences in attentional processes. These differences are illustrated in Figure 1. During the jigsaw task, 58% of the sentences were noticed with respect to their differences and 53% were noticed for their gaps. These percentages—at 93% and 83%—were also high on the spot-the-difference task. The figures for noticing during step 3 were much lower for the grammar communication task, as the pairs displayed noticing of differences for only 15% of their sentences and noticing the gap for 33%. On the other hand, simple noticing, which occurred with 89% of the sentences on the grammar communication task, was much more apparent.

Differences in the distribution of noticing in the grammar communication task compared to the jigsaw and spot-the-difference tasks might have been related to the number and format of options for choosing. During the grammar communication task, pair members had four phrases from which to choose. They tended to present each other with the phrases and announce their decisions. In the jigsaw and spot-the-difference tasks, the pair members had to choose between two sentences. This set up a basis for comparison, commentary, and evaluation.

Figure 2, which displays the distribution of awareness across task steps (labeled “reveal awareness” in Table 7), shows that all six pairs revealed awareness of the forms, their functions, and their meanings during at least one of the steps. They made 12 sentence decisions in which they revealed awareness during step 3 on the jigsaw task. Although this means that learners showed awareness on only 16% of all sentences in this task, it is notable that this constituted 60% of the pairs' awareness that was distributed across steps 3–5. During the recall step (step 4), 35% of the total evidence for awareness was found; this was revealed as the pairs deliberated over the items that they had chosen. However, instances of awareness were negligible during the comparison step. This pattern of awareness was consistent across the spot-the-difference task as well, with a distribution of 45% of all examples of awareness at step 3, 40% at step 4, and 5% at step 5. For the grammar communication task, awareness was revealed primarily during the choose step (step 3), as the pairs commented on relationships of form, function, and meaning for 32% of their sentences decisions. However, it was negligible during the recall and comparison steps.

Across the tasks, all pairs were able to recall phrases with the targeted forms from the sentences of passages they had read in steps 1 and 2 and chosen in step 3. This was revealed during their spoken decisions and written responses to the cloze passage of step 4. On the jigsaw and spot-the-difference tasks, the pairs were able to recall 89% and 97% of the phrases with nonsalient forms. For the grammar communication task, this figure was 82%. There was a strong relationship between noticing a form as encoded in the passages of step 3 and recalling it while completing a cloze version of these passages in step 4. Thus, of the 64 phrases that were recalled during the cloze portion of the jigsaw task, 81% had been noticed during step 3. These figures were 95% for the phrases recalled for spot-the-difference and 81% for those of grammar communication. Although some of the noticed forms were not recalled, their frequency and percentages were much lower than those that were recalled. Thus, on the jigsaw task, seven of the forms that had been noticed in step 3 were not recalled in step 4. This accounted for only 8% of the noticed forms, however. Similar patterns were found for the spot-the-difference task. Only two of the forms that had been noticed in step 3 were not recalled in step 4. This figure was somewhat higher for the grammar communication task, as 13% or 18% of the noticed forms were not recalled in step 4.

The second research question asked about the ways in which task implementation promoted interactional processes that have been shown to assist SLA. Findings are displayed in Figure 3 and Table 8. During the jigsaw task, all six pairs of participants often modified their interactions through negotiation signals and responses of modified output, doing so as they made 76% of their decisions during the choose step and as they recalled 42% of them during the cloze step. Modified interaction occurred throughout the steps of the grammar communication task as well. However, there was less overall modified interaction than in the jigsaw task and slightly more modified interaction during the recall step than during the choose step. These figures were 39% and 46%, respectively.

These differences in the choose step appeared to be related to its different demands. The fill-in format in each sentence and the four phrase options of the grammar communication task might have enabled the pairs to hone in on a preferred form. The pairs were not required to either reorder sentences or hunt down subtle differences between the two versions of each sentence, as they were asked to do in the jigsaw task. A further reason for this difference in interactional processes is that the jigsaw task gave the pairs two ways to choose. They were asked to choose the order of their sentences and, after that, to choose the sentence that they thought was better. However, as the pairs chose their order, they would often stop and choose between the two versions, reminding each other that this step was to be taken after the ordering but, nonetheless, pausing to choose between the versions. An example of this sort of modified interaction is shown in (4), in which a L1 Chinese learner (A) and a L1 Korean learner (B) work on the task:

During the spot-the-difference task, the pairs engaged in modified interaction for 25% of their sentences during their choose step, but only 4% during the recall step. On the other hand, it was during the choosing step that they recasted 28 of their sentence choices as they read their passage versions to each other. These read recasts were 47% of their choices. Yet another seven of their sentence choices were produced during pair-generated responses. There was minimal recasting by the pairs on the two other tasks. The emphasis on recasting in spot-the-difference might have been task-specific. The layout of the sentences during step 3 of this task was a paragraph, whereas the sentences were listed in a scrambled order for the jigsaw task, and they were interrupted by blank lines with choices beneath in the grammar communication task. The paragraph format might have lent itself more to reading as a way for the pairs to keep track of their choices or by association with the passage reading they had just completed for steps 1 and 2.

Correction, although low in frequency of occurrence, was provided by all six pairs during only two steps of the task, and mainly during the recall step. The low amount of correction might have been related to an absence of truly ungrammatical phrases in either of their versions. As mentioned previously, formations such as a books or wented were not used to encode erroneous phrases, but, rather, forms were employed that were inconsistent with passage meaning or with the original passage read by the learners.

The third research question asked about attentional and interactional relationships. As noted throughout the previous findings, most of the attentional and interactional activity occurred during the choose step. Although noticing processes, modified interaction, and recasting differed in their distribution, overall, for this step, there was a strong relationship between noticing forms with low salience and interactional processes claimed to assist SLA. The correlation was .62 using Pearson's r (p < .05). Despite the overall correlation across the three tasks, relationships between noticing and interaction were more prominent during the jigsaw and spot-the-difference tasks. Thus, of the 72 sentence choices that the pairs made for jigsaw, 81% of those that revealed simple noticing or noticing of differences or gaps were encoded in negotiation signals, modified responses, corrections, or recasts. These interactional features constituted 82% of the 60 decisions made during step 3 for the spot-the-difference task

However, these noticing and interactional processes were revealed together in only 45% of the 72 decisions for the grammar communication task. This might have been yet another reflection of task format, as the straightforward, familiar, fill-in-the-blank layout of the task drew the pairs to notice the phrases placed below each blank. This might also have reduced their need to modify their interaction toward mutual understanding. Additionally, task sequence might have contributed to this pattern. Grammar communication was the first of the three tasks in which the pairs engaged. It is possible that they might not have anticipated the precision that the subsequent cloze step required.

These noticing through interaction connections appeared to assist the pairs' accurate recall of phrases from step 3. Thus, 74% of the phrases recalled for the jigsaw task and 80% of those for the spot-the-difference task had been both noticed and encoded in an interactional process during the preceding choose step. On the other hand, only 35% of the recalled grammar communication choices had been noticed and encoded in this way. Of particular note was the finding that those choices made within the context of recasts—which predominated in the spot-the-difference task—were recalled more accurately than those that were accompanied by modified interaction—the dominant mode of the jigsaw task—and that both interactional processes were connected with high recall percentage scores. Thus, scores of 89% on the jigsaw and 97% on the spot-the-difference tasks suggest that the pairs applied interaction moves judiciously as needed during their task implementation.

The pairs' success with recall for the grammar communication task appeared to rely on simple noticing of forms in phrases, attesting again to the relative simplicity of the task layout and presentation of phrase options. The other two tasks activated a greater frequency and variety of interactional and attentional processes. Not only did the pairs achieve higher recall scores, but they also had more opportunities to notice crucial perceptual differences in the low-salience forms with which they clearly needed help.