Maximizing the benefits of peer interaction: Form-focused instruction and peer feedback training

1 Form-focused instruction and teacher corrective feedback

Form-focused instruction is based on the observation that although interaction in the L2 ‘may result in a better ability to communicate, interaction without any attention to linguistic accuracy does not necessarily improve linguistic accuracy’ (Loewen, 2015, p. 57). Importantly, form-focused instruction occurs when learners’ attention is drawn to linguistic forms while the primary focus remains on meaning.

The majority of studies investigating form-focused instruction have focused on its effectiveness for the acquisition of grammatical structures. Many of the earlier studies compared whether form-focused grammar instruction leads to better learning outcomes than no instruction at all (e.g. Day & Shapson, 1991; Doughty, 1991; Lightbown & Spada, 1990; Lyster, 1994; White, 1991; for a meta-analysis, see also Norris & Ortega, 2000). In summary, these studies found that instruction is indeed superior to no instruction. For example, in a study with child ESL (English as a second language) learners in Canada, Lightbown and Spada (1990) observed the teachers’ use of form-focused instruction in four different classes using a macro-level and a micro-level analysis of the classroom observational data to assess learners’ accuracy on several grammatical forms. The results showed that learners in the class in which the most form-focused instruction was provided were the most accurate on the production of these grammatical forms. In contrast, learners in the class in which the least form-focused instruction was provided were the least accurate on the same forms. Nevertheless, those learners did not lag behind the other learners on overall performance, such as comprehension skills.

Many of the more recent studies, rather than comparing form-focused instruction to no instruction, have explored the effectiveness of different types of form-focused instruction, such as enhanced input vs. non-enhanced input or deductive vs. inductive instruction (e.g. Cerezo, Caras, & Leow, 2016; Lee, 2007; Lindseth, 2016). As a whole, these studies show that several types of form-focused instruction can have positive effects on the acquisition of grammatical structures.

One particular type of form-focused instruction is corrective feedback. A large body of SLA research has focused on oral corrective feedback provided by teachers, and there is strong evidence that it facilitates L2 development, as shown in several meta-analyses (Li, 2010; Lyster & Saito, 2010; among others). This is the case both in classroom contexts (e.g. Ammar & Spada, 2006; Ellis, Loewen, & Erlam, 2006; Goo, 2012; Kartchava & Ammar, 2014; Loewen & Philip, 2006; Lyddon, 2011; Lyster, 2004; Sheen, 2010; Van de Guchte et al., 2015; Yang & Lyster, 2010), laboratory contexts (e.g. Egi, 2007; Iwashita, 2003; Loewen & Nabei, 2007; Mackey & Philip, 1998; McDonough, 2005; Nassaji, 2009; Yilmaz, 2013, 2016), and a combination of both contexts (Lyster & Izquierdo, 2009).

2 Corrective feedback on different types of linguistic targets

Although there is strong overall evidence for the positive impact of teacher feedback on L2 learning, research has also shown that a number of variables mediate its effectiveness (for further discussion, see Lyster, Saito, & Sato, 2013). One of those factors is the linguistic structure targeted by the feedback. For instance, in a study on teacher corrective feedback, Yang and Lyster (2010) investigated whether prompts (an output-prompting type of feedback) and recasts (an input-providing type of feedback) differentially affect the acquisition of regular and irregular past tense forms in English. Regular past participles, which add -ed to the infinitive form (e.g. played), were classified as rule-based, whereas irregular past participles, which require learners to learn each form individually (e.g. went), were classified as item-based. The results indicated that prompts worked better than recasts for regular past tense forms. However, prompts and recasts worked equally well for irregular past tense forms. The researchers argued that, because irregular past tense forms differ substantially from the infinitive form, recasts of irregular past tense forms were more salient and therefore more noticeable than recasts of regular past tense forms.

More recently, Sato and Loewen (2018) investigated the impact of output-prompting and input-providing teacher corrective feedback and metacognitive instruction on two types of linguistic structures in English: third-person singular -s and the possessive determiners his and her. The results showed that one type of feedback, specifically output-prompting feedback, had a greater impact on the acquisition of possessive determiners than third-person singular -s. Like Yang and Lyster (2010), Sato and Loewen argued that the saliency of the target structure mediates the effects of corrective feedback. In Sato and Loewen’s study, possessive determiners were operationalized as more salient than third-person singular -s. In addition, the researchers argued that the communicative value of the target structure also impacts the effects of feedback. Whereas third-person singular -s in English is unlikely to cause a communication breakdown, possessive determiners are more communicatively valuable, as using them incorrectly can cause misunderstandings.

In sum, research on the impact of corrective feedback in relation to various linguistic structures shows that feedback is not equally effective for all structures. One important consideration is the saliency of the linguistic target. Since feedback on more salient structures is easier to notice, it is likely to be more effective than feedback on less salient structures. Another consideration as identified by previous research is the communicative value of the linguistic target: corrective feedback seems to have a greater impact on communicatively valuable structures than on structures with a lower communicative value.

3 Peer interaction and peer corrective feedback

Peer interaction has been defined as ‘any communicative activity carried out between learners, where there is minimal or no participation from the teacher’ (Philip, Adams, & Iwashita, 2014, p. 3; italics in original). During these learner–learner interactions, corrective feedback is often withheld – sometimes deliberately because of social factors (Philip, Walter, & Basturkmen, 2010), such as face-saving or perceived inappropriateness (see Foster & Ohta, 2005; Fujii & Mackey, 2009). Although feedback can occur during peer interaction, especially when learners are engaged in negotiation for meaning (e.g. Foster & Ohta, 2005; Fujii and Mackey, 2009), Philip et al. (2014) – in reference to studies by Adams (2007), Adams et al. (2011), McDonough (2004), and Toth (2008) – claim that ‘this occurrence may not be very promising for learning’ (p. 55), since it is not as focused as teacher feedback and not necessarily perceived as corrective by learners. In addition, they point out that feedback provided by peers may be inaccurate. Most of the studies on oral peer feedback, however, are observational, and to date there is a relatively small body of experimental research on its effects on L2 development using quantitative measures and pretest–posttest designs. However, a few studies show that peer feedback can facilitate learning if teachers encourage learners to correct their peers’ mistakes and also provide them with the tools to do so (see Chu, 2013; Martin & Sippel, 2021; Sato & Lyster, 2012; Sippel & Jackson, 2015; Sippel, 2019; Toth, 2008).

For example, in Chu’s (2013) study with Taiwanese junior high school EFL students, the learners were taught different types of feedback and then instructed to use their preferred type of feedback during peer interaction. The results indicated that the learners improved from pretest to posttest on speaking tasks. However, since there was no control group and no group that engaged in peer interaction activities but received no feedback training, the study could not determine whether the improvement was a result of peer corrective feedback or of peer interaction and practice more generally.

In another study, Sato and Lyster (2012) investigated the effects of peer interaction and oral peer corrective feedback on accuracy and fluency development among advanced Japanese EFL learners. Two groups of learners engaged in peer interaction activities and were taught to provide feedback to peers, a third group engaged in peer interaction activities but was not taught to provide feedback to peers, and a fourth group did not engage in peer interaction activities at all (control group). The results showed that while the first three groups outperformed the control group on fluency measures, only the groups that had been taught to correct their peers’ errors improved significantly on grammatical accuracy measures. This study showed that peer feedback resulted in improved grammatical accuracy without impeding fluency development. The authors noted that, compared to teacher feedback, one of the strengths of peer feedback is that learners are more engaged in the learning process, since they are not only feedback receivers, but also feedback providers. Evidence from the written peer feedback literature supports this claim. For instance, Lundstrom and Baker (2009) showed that learners benefited more from revising their peers’ essays than from reading their peers’ feedback on their own essays.

While these studies provide evidence that peer feedback can be beneficial for the acquisition of grammatical structures, other studies have questioned its usefulness. For instance, in a classroom study with adult ESL learners, Adams et al. (2011) found that peer feedback did not contribute to L2 development. It is noteworthy that, in contrast to Sato and Lyster’s (2012) and Chu’s (2013) study, Adams et al.’s (2011) study did not include a training component designed to teach learners how to provide feedback to one another. In fact, Adams et al. (2011) did not ask learners to provide feedback at all, and research has shown that the degree to which learners naturally point out each other’s mistakes during peer interaction is low (e.g. Williams, 1999; Sato, 2007; Philip et al., 2010).

4 Strategy instruction and metacognitive instruction

As discussed previously, research on peer interaction – especially interaction that does not require negotiation for meaning – reveals that learning opportunities may be missed if learners are simply asked to work on tasks collaboratively, without understanding how these tasks and interactions will help them learn, and what the learners themselves can do to maximize learning opportunities during interactions with other learners. Pedagogical solutions to this problem could be strategy instruction and metacognitive instruction. The goal of strategy instruction is to teach learners to employ L2 learning strategies, defined as the steps taken on the part of learners to improve their L2 learning and/or use (Oxford, 1990). ¹ For the context of peer feedback, this would entail teaching learners how to provide corrective feedback to each other. Metacognitive instruction, on the other hand, is designed to help learners enhance their metacognition, which has been defined as ‘the monitoring and control of thought’ (Martinez, 2006, p. 696). In the field of SLA, researchers have emphasized the important role of metacognition for the development of L2 reading skills (e.g. Carrell, 1989), L2 listening skills (e.g. Goh, 2008), L2 grammar (Sato & Loewen, 2018) and collaborative interaction (e.g. Fujii, Ziegler, & Mackey, 2016; Sato, 2019). For the context of peer feedback, metacognitive instruction would entail manipulating learners’ knowledge related to peer feedback (e.g. what it is and does, how it functions, why it is useful).

While framed as a study on metacognitive instruction, one could argue that Fuji et al. (2016) employed both metacognitive instruction and strategy instruction, since they investigated how instruction can help learners become more successful interactors and better feedback providers by teaching an experimental group of learners the benefits of peer feedback and showing them how to provide it, whereas a control group of learners did not receive such instruction. Results from a pretest and a posttest showed that learners in the experimental group provided more feedback and made use of the feedback they received more frequently than learners in the control group.

In sum, previous research shows that strategy instruction and metacognitive instruction in general have strong potential to facilitate language acquisition. Relevant to the present study, research shows that strategy instruction and metacognitive instruction could be a powerful tool to maximize learning opportunities during peer interaction. ² At the same time, it is possible that learners, especially at lower levels of proficiency, may need additional support to be able to use this instruction to their best advantage during peer interaction. One way to provide this support could be form-focused instruction, since basic linguistic knowledge is likely a prerequisite for the ability to provide feedback to another learner.

1 Participants

One-hundred and nineteen students enrolled in six sections of third-semester German language classes at a large American university were eligible to participate in the experiment. However, data from 32 participants who were absent on at least one of the treatment or testing days were excluded from the analysis reported here. As a result, the data of 87 participants were analysed. Of those 87 participants, 53 were male and 34 were female, and 78 were native speakers of English. They had various majors, but none of them was majoring in German, and 72 participants were taking German to fulfill a requirement. Of the 87 participants, 81 had never lived in a German-speaking country and 6 had lived in a German-speaking country for a limited period of time during their childhood. However, at the time of the study, exposure to the German language occurred almost exclusively through the classroom for all participants.

Two sections of the third-semester German course were assigned to the PI group (n = 27), two sections were assigned to the FFI group (n = 32), and two sections were assigned to the PFT group (n = 28). Participants’ average age across the three groups was 19.8 years and they had been studying German for 3.6 years on average. The participants were asked to self-rate their German proficiency for reading, writing, speaking, and listening on a scale from 1 (not literate/fluent) to 10 (very literate/fluent). A one-way ANOVA revealed that the participants from the three groups did not differ significantly in age (F(2, 86) = 0.24, p = .786), years of previous German study (F(2, 86) = 0.78, p = .462), or their self-rated proficiency in reading (F(2, 86) = 1.01, p = .370), writing (F(2, 86) = 1.10, p = .338), speaking (F(2, 86) = 0.67, p = .516), or listening (F(2, 86) = 0.82, p = .443).

Three instructors participated in the study. A native speaker of German with five years of teaching experience taught the PI group. This instructor is also the author of the present study. A native speaker of Swedish with native-like fluency in German and 20 years of teaching experience taught the FFI group. A native speaker of German with eight years of teaching experience taught the PFT group. Before the study was implemented in the classes, all instructors received detailed lesson plans from the researcher to ensure treatment fidelity. The instructors also participated in a 90-minute meeting with the researcher. During this meeting, the researcher briefly talked about the study, explained the lesson plans, and gave the instructors the opportunity to ask questions.

2 Target structures

The linguistic target of the study was the German present perfect tense, which is used to describe events in the past, mostly in spoken German. The present perfect tense consists of two separate components: auxiliary verb selection (a rule-based structure) and past participle formation (an item-based structure). Additionally, word order can pose challenges to learners, because the auxiliary verb (haben ‘to have’ or sein ‘to be’) is placed in the second position of a main clause, whereas the past participle is placed at the end.

(1) Die Kinder haben ihre Hausaufgaben gemacht.

 The kids   have  their homework   done.

 ‘The kids did their homework.’ ³

(2) Der Mann ist nach Hause gegangen.

 The man  is  to home  gone.

 ‘The man went home.’

(3) Das Baby ist um 6 Uhr  aufgewacht.

 The baby  is  at 6 o’clock woken up.

 ‘The baby woke up at 6 o’clock.’

(4) Die Jugendlichen haben auf der Straße getanzt.

 The teenagers   have  on the street  danced.

 ‘The teenagers danced on the street.’

Auxiliary verb selection in German can be considered a rule-based structure because transitive verbs typically select haben (1), whereas intransitive verbs that express a change of location (2) or a change of state (3) select sein. If they do not express a change of location or a change of state, intransitive verbs usually select haben (4). ⁴

German past participles can be regular, irregular, or mixed (a combination of regular and irregular). The regular past participle is formed with the circumfix ge. . .t and the verb stem (machen ‘to do’ → gemacht ‘done’). Irregular past participles are formed with the circumfix ge. . .en and the verb stem (schlafen ‘to sleep’ → geschlafen ‘slept’), or with ge. . .en and a stem-change in the vowel between the infinitive form and the past participle (gehen ‘to go’ → gegangen ‘gone’). Mixed verbs take the regular circumfix ge. . .t but change the stem (bringen ‘to bring → gebracht ‘brought). Since learners must memorize for each verb to which group they belong, German past participles can be considered an item-based structure.

3 Pre- and post-assessments

The pretest and the two posttests consisted of an oral production task and an error correction task. Although an oral production task is generally more likely to address implicit knowledge than an error correction task, which is explicit by its very nature, the oral production task used in the present study contained components that were quite explicit and therefore might have assessed explicit knowledge more so than implicit knowledge. Nevertheless, the inclusion of the two types of tasks allowed the researcher to assess the participants’ use of the target structure in two different contexts, an oral and a written context. Both tasks were administered in the classroom and the oral production task was always administered before the error correction task.

4 Instructional materials and procedure

The instructional intervention took place over three consecutive class meetings in the fourth week of the semester. The classes met four times a week for 50 minutes each. The theme of the unit taught during this week was the culture and history of the city of Munich. An overview of the treatment is provided in Table 1.

OPEN IN VIEWER

Table 1.

Overview of treatment.

	PI group	FFI group	PFT group
Day 1	• Pretest (15 minutes) • Exposure to target structure through reading (10 minutes) • Content-focused activities with exposure to target structure (25 minutes)	• Pretest (15 minutes) • Exposure to target structure through reading (10 minutes) • Form-focused instruction and activities (25 minutes)	• Pretest (15 minutes) • Exposure to target structure through reading (10 minutes) • Form-focused instruction and activities (25 minutes)
Day 2	• Content-focused activities with exposure to target structure (50 minutes)	• Content-focused activities with exposure to target structure (50 minutes)	• Feedback training (20 minutes) • Content-focused activities with exposure to target structure (30 minutes)
Day 3	• Content-focused activities with exposure to target structure (35 minutes) • Immediate posttest (15 minutes)	• Content-focused activities with exposure to target structure (35 minutes) • Immediate posttest (15 minutes)	• Content-focused activities with exposure to target structure (35 minutes) • Immediate posttest (15 minutes)

Notes. PI group = peer interaction group. FFI group = peer interaction and form-focused instruction group. PFT group = peer interaction, form-focused instruction, and peer feedback training group.

At the beginning of day 1, participants in all groups completed the pretest. Afterwards, participants in all groups were introduced to the topic of the unit by means of a worksheet that asked participants to match pictures (e.g. a picture of the Oktoberfest) and the corresponding activity (e.g. aufs Oktoberfest gehen ‘go to the Oktoberfest’). Then, participants in all groups read a text that contained forms of the target structures. After reading the text, participants in the FFI group and the PFT group were asked to highlight all auxiliary verbs and past participles, and to write them down in a table provided on the worksheet. The German infinitives and their English translations were provided in the table. After the instructors had compared the answers with the whole class, they explained the rules of auxiliary selection. The instructors’ explanation was supported by a handout, which contained the concepts of transitivity, change of location, and change of state, as well as some relevant exceptions (e.g. sein ‘to be’ and bleiben ‘to stay’). Afterward, the participants worked with a partner to complete a short exercise on auxiliary selection, in which they were asked to circle the appropriate auxiliary and explain why they chose this auxiliary. The instructors then explained how past participles are formed, including information about regular, irregular, and mixed past participles. Afterward, the participants worked with a partner to complete a short exercise on past participles.

The participants in the PI group read the same text, but their attention was never drawn to the forms. Rather than highlighting forms in the texts or receiving grammatical explanations from the instructor, the participants in the PI group were given questions about the text that focused solely on the content, and they discussed the questions with a partner. The questions were in the present perfect tense and required answers in the present perfect tense. However, the participants were not specifically instructed to use the present perfect tense in their responses. As a follow-up activity, the participants discussed more personal questions, which were also in the present perfect tense. Finally, the participants reported to two new partners what their old partners had told them. ⁶

On day 2, the participants in the PFT group received training on how to provide corrective feedback to their peers (i.e. strategy instruction and metacognitive instruction). Following Sato and Lyster (2012), the training occurred in three stages: modeling, practice, and use-in-context.

During the modeling stage, the instructor demonstrated how learners can correct each other by administering a mini role play in front of the class. The mini role-play contained four mistakes and four respective feedback moves, two regarding auxiliary selection and two regarding past participle formation. Afterward, the instructor discussed some of the feedback moves that had occurred using a PowerPoint presentation with excerpts from the mini role play. Rather than being taught a single feedback strategy such as either recasts or prompts (as had been the case in Sato & Lyster, 2012), the participants in the present study were taught a more open-ended feedback strategy: Specifically, whenever they noticed a mistake in their peers’ speech, they were asked to point out that mistake, for example through repetition of the erroneous part of the sentence using rising intonation (e.g. *gemachen? ‘done’) or through a metalinguistic explanation (e.g. ‘you need a regular past participle’). Only if this kind of feedback did not help, participants were asked to provide the correct answer.

During the practice stage, participants were given a role-play scenario designed to help them practice providing feedback to peers in small groups (see Sato & Lyster, 2012). They also received a list of five sentences, each of which they were asked to use during the role play. Those sentences contained mistakes regarding auxiliary verb selection and past participle formation. After a few minutes of planning time, the participants acted out the role-play in their groups. While the speakers’ role was to purposefully include the errors in their speech, the listeners’ role was to detect and point out the errors.

During the use-in-context stage, participants were encouraged to use the feedback technique in more open-ended and communicative activities, that is, in all peer interaction activities following the feedback training.

The participants in the FFI group and the PI group were not instructed to correct their peers’ mistakes. Instead, they received additional exposure to the present perfect tense and additional practice opportunities. First, they read a text about the history of Munich that contained forms in the present perfect tense. Afterwards, they completed peer interaction activities in which they discussed questions about the text with a partner. The questions were in the present perfect tense and prompted them to answer in the present perfect tense.

For the rest of day 2 and most of day 3, participants in all three groups completed peer interaction activities focusing primarily on content, while also encouraging participants to use the present perfect tense. Those activities included an information gap activity, a fictional story-retelling task, a retelling task of the history of Oktoberfest, a question-and-answer activity on the history of Oktoberfest, and a bingo activity. The fictional story-retelling task was collaborative and open-ended. All other tasks were collaborative but had a fixed outcome. The linguistic targets in all tasks were not task essential, that is, it was not necessary for participants to correct each other’s errors to achieve mutual understanding. Participants in all groups completed the same tasks, but only participants in the PFT group were encouraged to point out and correct their peers’ mistakes. At the end of day 3, all participants completed the immediate posttest. Two weeks later, they completed the delayed posttest.

The following section presents examples of learner–learner interactions from the treatment, which the instructors in the various conditions had recorded during an information-gap activity on treatment day 2.

5 Examples of learner–learner interactions

Learners in the PFT group responded to their peers’ errors in a variety of ways, provided they chose to respond at all. One technique they used was to point out the error and give the peer a chance to self-correct it, an output-pushing type of feedback, often referred to as a ‘prompt’ in the literature.

Another technique learners in the PFT group used was to correct the error directly, an input-providing type of feedback that has been defined as a ‘recast’.

Although many of the feedback moves consisted of prompts and recasts as illustrated above, sometimes the feedback of the learners in the PFT group involved switching to English, using grammatical terminology, or expressing their own uncertainty about the correct form.

In the FFI group and the PI group, a typical interaction between two learners did not involve feedback moves. Rather, learners tended to simply move on to the next portion of the task after an error had been made.

6 Scoring and data analysis

On the oral production task, only the six sentences that required the use of the auxiliary verb sein were included in the analysis. The six sentences that required haben were not included because haben generally functions as the default auxiliary verb for L2 learners of German, especially for learners whose native language does not have a split-auxiliary system. ⁷ In other words, instances of haben were treated as distractor items on this task, and participants only received points for instances of sein, specifically, one point for each sentence in which they correctly used sein as the auxiliary. Thus, the maximum score for the auxiliary verb on the oral production task was six. For the past participle, participants received one point for each sentence in which they used the correct past participle. Since all 12 sentences required the use of a past participle, the maximum score for the past participle on the oral production task was 12. There was no partial credit for either the auxiliary verb or the past participle, and participants did not lose points for grammatical errors unrelated to the target structure, such as subject–verb agreement.

On the error correction task, participants received one point for each of the six auxiliary mistakes and one point for each of the six past participle mistakes that they identified and corrected appropriately (e.g. the mistake was *getrinken ‘drunk’, and the participant crossed out *getrinken and changed it to the correct form getrunken ‘drunk’). No point was given if a mistake was identified but not corrected appropriately (e.g. the mistake was *getrinken ‘drunk’, and the participant crossed out *getrinken and changed it to an incorrect form, such as *getrinkt ‘drunk’). Participants did not receive points for filler items that they corrected appropriately, and they did not lose points for grammatical items that they misidentified as ungrammatical and subsequently ‘corrected’. ⁸ The maximum score on the error correction task was six for the auxiliary verb and six for the past participle.

Since pretest scores were not distributed normally according to Kolmogorov–Smirnov tests, nonparametric tests were used for the statistics analyses. However, the pattern of results did not differ when parametric tests were used, although these results are not reported here. For each task, Kruskal–Wallis tests were performed to determine whether there were significant differences between the three groups at any given testing time (pretest, immediate posttest, or delayed posttest). Subsequently, Mann–Whitney tests were used to compare the performance of two groups at a time for each testing time. Following Larson-Hall (2016), Bonferroni corrections were used for post-hoc comparisons, resulting in an adjusted significance level of p = .017 (p = .05/3 = .017).

1 Oral production task

Of the 87 participants, seven participants were not included in the analysis of the oral production task due to technical difficulties or incomplete recordings. Additionally, eight participants were excluded because they misunderstood the directions and told the story in the present tense rather than in the present perfect tense. Thus, 72 participants were included in the analysis reported here.

A Auxiliary verb choice

Table 2 shows the results of the oral production task for the auxiliary verb choice (sein-verbs only). A Kruskal–Wallis test revealed that the three groups did not differ significantly on the pretest (χ² (2) = 3.80, p = .150), but on the immediate and delayed posttest, there were significant differences between the groups (immediate posttest: χ² (2) = 8.91, p = .012; delayed posttest: χ² (2) = 7.83, p = .020).

OPEN IN VIEWER

Table 2.

Results: Oral production task (auxiliary verb): Number of correct sein utterances (maximum 6).

	Pretest		Immediate posttest		Delayed posttest
	M	SD	M	SD	M	SD
PI group (n = 18)	2.06	1.66	2.00	2.20	2.11	2.35
FFI group (n = 29)	2.34	2.11	3.69	1.91	4.07	2.07
PFT group (n = 25)	1.36	1.50	3.84	1.70	3.72	1.95

Notes. PI group = peer interaction group. FFI group = peer interaction and form-focused instruction group. PFT group = peer interaction, form-focused instruction, and peer feedback training group.

Follow-up Mann–Whitney tests revealed no significant difference between the PI and the FFI group on the pretest (U = 247.00, z = −0.31, p = .755, r = −.05), but there was a significant difference between those two groups on the immediate posttest (U = 144.50, z = −2.58, p = .010, r = −.38) and delayed posttest (U = 147.50, z = −2.56, p = .011, r = −.37), because the FFI group outperformed the PI group.

The difference between the PI and the PFT group was not significant on the pretest (U = 166.50, z = −1.49, p = .136, r = −.23), but there was a significant difference between those two groups on the immediate posttest (U = 115.50, z = −0.42, p = .006, r = −.81) because the PFT group outperformed the PI group and delayed posttest. The difference on the delayed posttest approached significance (U = 134.00, z = −2.29, p = .022, r = −.35), with the PFT group still performing better than the PI group.

There was no significant difference between the FFI and the PFT group at any testing time (pretest: U = 262.00, z = −1.81, p = .071, r = −.25; immediate posttest: U = 356.00, z = −0.12, p = .908, r = −.02; delayed posttest: U = 318.00, z = −0.79, p = .428, r = −.11.

B Past participle

Table 3 shows the results of the oral production task for the past participle. A Kruskal–Wallis test revealed that the three groups did not differ significantly on the pretest (χ² (2) = .27, p = .875). There was a significant difference between the groups on the immediate posttest (χ² (2) = 12.06, p = .002), but not on the delayed posttest (χ² (2) = 2.49, p = .288).

OPEN IN VIEWER

Table 3.

Results: Oral production task (past participle): Number of correct past participles (maximum 12).

	Pretest		Immediate posttest		Delayed posttest
	M	SD	M	SD	M	SD
PI group (n = 18)	5.28	3.21	7.61	2.00	7.28	3.18
FFI group (n = 29)	5.24	2.71	8.07	2.36	7.69	2.77
PFT group (n = 25)	5.64	2.84	9.68	1.68	8.24	2.91

Notes. PI group = peer interaction group. FFI group = peer interaction and form-focused instruction group. PFT group = peer interaction, form-focused instruction, and peer feedback training group.

Follow-up Mann–Whitney tests revealed no significant differences between the PI and the FFI group at any testing time (pretest: U = 244.50, z = −0.37, p = .715, r = −.05; immediate posttest: U = 216.50, z = −0.99, p = .325, r = −.14; delayed posttest: U = 232.50, z = −0.63, p = .529, r = −.09).

There was no significant difference between the PI and the PFT group on the pretest (U = 217.50, z = −0.19, p = .852, r = −.03), but there was a significant difference between those two groups on the immediate posttest (U = 92.50, z = −3.32, p = .001, r = −.51), because the PFT group outperformed the PI group. There was no significant difference on the delayed posttest (U = 168.50, z = −1.40, p = .161, r = −.21).

There was no significant difference between the FFI and the PFT group on the pretest (U = 336.00, z = −0.47, p = .640, r = −.06), but there was a significant difference between those two groups on the immediate posttest (U = 218.00, z = −2.54, p = .011, r = −.35), because the PFT group outperformed the FFI group. There was no significant difference on the delayed posttest (U = 295.00, z = −1.19, p = .236, r = −.16).

2 Error correction task

Of the 87 participants, one participant was not included in the analysis of the error correction task because this participant already performed at ceiling, defined as 100% accuracy, on both the auxiliary and the past participle at the time of the pretest. ⁹ Thus, 86 participants were included in the analysis reported here.

A Auxiliary verb choice

Table 4 shows the results of the error correction task for the auxiliary verb choice. A Kruskal–Wallis test revealed that the three groups did not differ significantly on the pretest (χ² (2) = .18, p = .916), but on the immediate and delayed posttest, there were significant differences between the groups (immediate posttest: χ² (2) = 13.75, p = .001; delayed posttest: χ² (2) = 12.58, p = .002).

OPEN IN VIEWER

Table 4.

Results: Error correction task (auxiliary verb): Number of corrected sein errors (maximum 6).

	Pretest		Immediate posttest		Delayed posttest
	M	SD	M	SD	M	SD
PI group (n = 27)	1.33	1.71	1.89	2.17	1.67	2.13
FFI group (n = 31)	1.45	2.08	3.42	2.08	3.06	2.24
PFT group (n = 28)	1.25	1.82	4.14	1.82	3.79	2.15

Notes. PI group = peer interaction group. FFI group = peer interaction and form-focused instruction group. PFT group = peer interaction, form-focused instruction, and peer feedback training group.

Follow-up Mann–Whitney tests revealed no significant difference between the PI and the FFI group on the pretest (U = 404.00, z = −0.25, p = .806, r = −.03), but there was a significant difference between those two groups on the immediate posttest (U = 256.00, z = −2.58, p = .010, r = −.34), because the FFI group outperformed the PI group. The difference between the two groups approached significance on the delayed posttest (U = 273.50, z = −2.35, p = .019, r = −.31).

There was no significant difference between the PI and the PFT group on the pretest (U = 355.00, z = −0.42, p = .674, r = −.06), but there was a significant difference between those two groups on the immediate posttest (U = 171.00, z = −3.55, p < .001, r = −.48) and delayed posttest (U = 178.00, z = −3.45, p = .001, r = −.47), because the PFT group outperformed the PI group.

There were no significant differences between the FFI and the PFT group at any testing time (pretest: U = 424.50, z = −0.16, p = .872, r = −.02; immediate posttest: U = 347.00, z = −1.34, p = .180, r = −.17; delayed posttest: U = 344.00, z = −1.38, p = .165, r = −.18).

B Past participle

Table 5 shows the results of the error correction task for the past participle. A Kruskal–Wallis test revealed that the three groups did not differ significantly on the pretest (χ² (2) = 3.83, p = .148). There was a significant difference between the groups on the immediate posttest (χ² (2) = 8.83, p = .012), but not on the delayed posttest (χ² (2) = 2.66, p = .265).

OPEN IN VIEWER

Table 5.

Results: Error correction task (past participle): Number of corrected past participles (maximum 6).

	Pretest		Immediate posttest		Delayed posttest
	M	SD	M	SD	M	SD
PI group (n = 27)	1.81	1.69	2.22	1.67	2.26	1.35
FFI group (n = 31)	1.23	1.65	3.06	1.81	2.55	1.84
PFT group (n = 28)	1.82	1.49	3.57	1.48	3.07	1.72

Notes. PI group = peer interaction group. FFI group = peer interaction and form-focused instruction group. PFT group = peer interaction, form-focused instruction, and peer feedback training group.

Follow-up Mann–Whitney tests revealed no significant differences between the PI and the FFI group at any testing time (pretest: U = 321.50, z = −1.58, p = .113, r = −.21; immediate posttest: U = 308.50, z = −1.75, p = .080, r = −.23; delayed posttest: U = 375.00, z = −0.69, p = .490, r = −.09).

There was no significant difference between the PI and the PFT group on the pretest (U = 368.50, z = −0.16, p = .870, r = −.02), but there was a significant difference between those two groups on the immediate posttest (U = 205.00, z = −2.98, p = .003, r = −.40), because the PFT group outperformed the PI group. There was no significant difference between the two groups on the delayed posttest (U = 286.00, z = −1.58, p = .113, r = −.21).

There were no significant differences between the FFI and the PFT group at any testing time (pretest: U = 321.50, z = −1.78, p = .075, r = −.23; immediate posttest: U = 354.50, z = −1.24, p = .216, r = −.16; delayed posttest: U = 366.50, z = −1.04, p = .299, r = −.14).

1 Overview of findings

In summary, the PFT group significantly outperformed the FFI group on only one of eight posttest measures. In addition, the PFT group significantly outperformed the PI group on five measures whereas the FFI group significantly outperformed the PI group only on three measures, indicating an advantageous effect of form-focused instruction, and a slight indirect edge of the PFT group over the FFI group. Overall, these results show that form-focused instruction alone as well as form-focused instruction combined with peer feedback training were most effective for the development of grammatical accuracy.

2 The impact of peer interaction and form-focused instruction

Regarding the impact of peer interaction, the results indicate that peer interaction as used in the present study did little to promote the acquisition of the present perfect tense in third-semester learners of German. Importantly, however, the linguistic structures targeted in the present study were not task essential, that is, the accuracy of the auxiliary verbs and the past participles was not essential for learners to understand each other and complete the activities. It is possible that peer interaction leads to greater development of linguistic accuracy when the target forms are task essential. With respect to form-focused instruction, the present study contributes to the large body of research showing that this type of instruction facilitates the acquisition of grammatical structures (e.g. Cerezo et al., 2016; Day & Shapson, 1991; Doughty, 1991; Lee, 2007; Lightbown & Spada, 1990; Lindseth, 2016; Lyster, 1994; Norris & Ortega, 2000; White, 1991). Overall, the findings are consistent with Sato and Lyster (2012), who also found that peer interaction alone was not effective for the development of grammatical accuracy, whereas peer interaction combined with an accuracy-focused component (which in their study was peer corrective feedback without form-focused instruction) was beneficial.

3 The impact of peer feedback training on L2 development

With respect to the efficacy of peer feedback training, the present study shows that peer interaction combined with form-focused instruction and peer feedback was only slightly more effective than peer interaction combined with form-focused instruction, but without peer feedback. Although the PFT group significantly outperformed the FFI group only on one posttest measure (the past participle measure on immediate posttest of the oral production task), there were two additional measures on which the PFT group performed better than the FFI group on a descriptive level, specifically, the auxiliary verb measures on both the immediate and delayed posttest of the error correction task. While this must remain speculative, it is possible that significance could have been reached on additional measures if there had been less participant attrition, especially on the oral production task, or if a larger number of target items had been included in the tests. Furthermore, the finding that the PFT group significantly outperformed the PI group on five posttest measures whereas the FFI group significantly outperformed the PI group on three measures indicates a slight additional (albeit indirect) advantage of the PFT condition over the FFI condition.

Overall, the present study supports the findings of previous studies that examined peer corrective feedback and reported that it was beneficial for the acquisition of grammatical structures when feedback training was provided (e.g. Chu, 2013; Sato & Lyster, 2012; Sippel & Jackson, 2015). Moreover, the present study adds a new dimension to the existing peer corrective feedback literature by including form-focused instruction as an additional variable. Previous studies have compared peer interaction with peer feedback training to peer interaction without any kind of focus on form (e.g. Sato and Lyster, 2012), or compared form-focused instruction combined with peer feedback to form-focused instruction combined with teacher feedback (Sippel & Jackson, 2015), but not to form-focused instruction without corrective feedback. Other studies investigated the effects of peer interaction combined with peer feedback training, but without including additional experimental groups or control groups (Chu, 2013).

In summary, the present study found that (1) peer feedback training promoted learners’ grammatical accuracy, (2) form-focused instruction also promoted learners’ grammatical accuracy, and (3) peer interaction alone had minimal effects on the development of grammatical accuracy. However, when interpreting the findings of the present study, it is important to remember that the treatment lasted only three class periods. It is likely that a longer intervention, which would have provided more opportunities for giving and receiving feedback in the PFT group, would have led to a greater impact of the intervention. While the slight advantage of the PFT group over the FFI group in the present study could be attributed to the peer feedback that occurred (as a result of the strategy instruction that was provided), it is also possible that the metacognitive instruction itself benefited learners in the PFT group. Specifically, it is possible that this instruction heightened learners’ metacognitive knowledge of peer interaction and led them to self-monitor their use of the target structures during peer interaction (and, potentially, the use of the target structures by their classmates as well), which may have contributed to the positive developmental outcomes in the PFT group.

4 Auxiliary verb selection vs. past participle formation

The results revealed somewhat differential effects of the three interventions on the acquisition of the rule-based target structure (auxiliary verb selection) and the item-based target structure (past participle formation). For example, the FFI group significantly outperformed the PI group on three of four auxiliary measures, but on none of the four past participle measures. Similarly, the PFT group also outperformed the PI group on three auxiliary measures (and significance was approached on the fourth auxiliary measure), but only on two of the four past participle measures. Thus, form-focused instruction was more effective for the rule-based structure than for the item-based structure, suggesting that explicit grammar instruction alone may be most useful for linguistic structures based on concrete rules.

With respect to peer feedback, however, the only measure on which the PFT group significantly outperformed the FFI group was a past participle measure. This finding is consistent with previous research on the differential effects of corrective feedback on more than one linguistic target. Both Yang and Lyster (2010) and Sato and Loewen (2018) showed that teacher corrective feedback was most effective for the more salient grammatical structure. In the present study, past participles were the more salient structure since they involve discrete lexical items. Perhaps more importantly, however, past participles have a significantly higher communicative value than auxiliary verbs. Whereas productions with an incorrect auxiliary verb or even without an auxiliary verb can generally be understood without difficulty, the past participle is the component of the German present perfect tense that assigns meaning to a sentence, specifically the action of the sentence. This explanation for the advantageous effect of peer feedback for the past participle is in line with Sato’s and Loewen’s (2018) finding that corrective feedback has a more substantial impact on communicatively valuable linguistic structures. Finally, an additional factor could be that past participles in German appear at the end of a main clause, whereas the auxiliary verb is in the second position. It may be the case that learners in the PFT group focused more on the most recent item of the sentence when comparing their peers’ speech to their own interlanguage. In other words, it is possible that feedback providers simply forgot to correct an incorrect auxiliary verb in their peers’ speech, since their attention may have shifted from the auxiliary verb to the past participle. In addition, feedback providers may have been distracted by other elements of the sentence that were placed between the auxiliary verb and the past participle.

The PFT group also performed descriptively better than the FFI group on two auxiliary measures (the immediate and delayed posttest of the error correction task). While statistical analyses revealed a significant difference between those two groups only for a past participle measure, descriptive results suggest that peer feedback may be useful for different types of structures, although presumably not equally useful. Since previous research on the effects of corrective feedback on different linguistic targets has focused on teacher feedback, future studies could further address this question specifically for the context of peer feedback with a variety of linguistic targets in different languages.

5 Theoretical account for the effectiveness of peer feedback training

From a theoretical perspective, the findings of the present study support Sato’s (2017) dual model of peer corrective feedback, which accounts for the unique context of peer interaction, in which learners are both receivers and providers of feedback. Like Levelt’s (1983) perceptual theory of monitoring, Sato’s model also emphasizes the importance of speech monitoring. In the present study, learners who received strategy instruction and metacognitive instruction likely benefited from comparing their peers’ productions to their own interlanguage, detecting discrepancies, and monitoring their own corrective feedback both internally and externally. These processes might also have pushed learners to monitor their own speech more carefully. In addition, learners likely benefited from noticing corrections provided by peers and subsequently modifying their own productions.