Theoretical background

The advantage of interleaved practice over blocked practice can be explained by the discriminative contrast hypothesis (Carvalho & Goldstone, Reference Carvalho and Goldstone2014; Zulkiply & Burt, Reference Zulkiply and Burt2013), stating that interleaved practice helps learners distinguish similar concepts. When engaged in interleaved practice, learners devote their attention to the distinctions between constantly switched categories, which allows them to compare exemplars from similar categories and pay attention to subtle distinctions between these categories (e.g., comparable to the similarities between blueberries and blackberries). In other words, “when between-category discriminability is low, interleaved practice can be more beneficial than blocked practice” (Suzuki et al., Reference Suzuki, Yokosawa and Aline2022, p. 673). This phenomenon is known as interleaving effects. It is crucial to note that to produce interleaving effects, the categories or concepts need to be similar. When two categories are divergent from each other and the level of between-category discriminability is high (e.g., blueberries versus potatoes), the effectiveness of interleaved practice may become limited. For example, the advantage of interleaved practice over blocked practice was found in Eglington and Kang’s (Reference Eglington and Kang2017) study, which explored undergraduate students’ acquisition of similar organic chemical compounds. However, interleaving effects were not found in Carpenter and Muller’s (Reference Carpenter and Mueller2013) study, which examined how L2 learners of French acquired sounds with high between-category discriminability (e.g., /o/ in cadeau versus /s/ in brebis) assessed by multiple-choice questions.

Although prior studies in various fields have shown the advantage of interleaved practice over blocked practice, it is worth pointing out that these two types of practice schedules can be efficacious in different aspects. The superiority of blocked practice over interleaved practice might be found in L2 fluency development. The effectiveness of blocked practice in fluency may be explained by the speech production model (Levelt, Reference Levelt1989), which consists of three main elements: (a) the conceptualizer (e.g., generating a preverbal idea), (b) the formulator (e.g., grammatical encoding, phonological encoding, the encoding of form-function-context mappings), and (c) the articulator (e.g., producing output in an L2). According to the speech production model, when L2 learners complete a task for the first time, their attention is directed to working memory for the purpose of conceptualization, for instance, creating a preverbal idea. As a result, limited attentional resources can be allocated for formulation and articulation. However, practicing the same task consecutively (i.e., blocked practice) can familiarize learners with the task. Thus, they may free up their attentional resources for the encoding of form-function-context mappings and articulation, which might enhance the fluency of learners’ L2 pragmatic performance.

Effectiveness of blocked practice and interleaved practice

No studies in the existing literature have investigated the effects of interleaved practice and blocked practice on L2 pragmatic competence. Most of the L2 practice studies exploring the influence of interleaved practice and blocked practice have been focused on grammar (Nakata & Suzuki, Reference Nakata and Suzuki2019; Suzuki et al., Reference Suzuki, Yokosawa and Aline2022), suggesting the superiority of interleaved practice over blocked practice on grammatical accuracy. For example, Nakata and Suzuki (Reference Nakata and Suzuki2019) examined the effects of interleaved practice and blocked practice on Japanese English as a foreign language (EFL) learners’ acquisition of syntactic structures with low between-category discriminability (i.e., the simple past tense and the present perfect tense; first-conditional, second-conditional, and third-conditional constructions). The results of grammaticality judgment tests showed that the interleaved-practice group significantly outperformed the blocked-practice group on a one-week delayed posttest. Similar results were found in Suzuki et al. (Reference Suzuki, Yokosawa and Aline2022). They investigated the influence of interleaved practice and blocked practice on Japanese EFL learners’ acquisition of five similar relative-clause constructions. The results of a picture description test demonstrated that the interleaved-practice group produced more accurate relative-clause constructions than the blocked-practice group on both immediate and delayed posttests.

Only a handful of studies have examined the influence of interleaved practice and blocked practice on L2 oral fluency. Importantly, the small number of studies have revealed inconclusive findings with respect to which practice schedule is more effective in enhancing fluency. For instance, in Suzuki’s (Reference Suzuki2021) study, the results of oral narrative tasks showed that blocked practice led to greater fluency development (e.g., higher articulation rate) compared with interleaved practice. However, in Suzuki et al.’s (Reference Suzuki, Yokosawa and Aline2022) research, the results of a picture description test demonstrated slight superiority of interleaved practice over blocked practice in Japanese EFL learners’ fluency in producing five similar relative-clause constructions on the immediate posttest. Due to the mixed findings, more studies are needed to further investigate the effects of interleaved practice and blocked practice on L2 oral fluency.

Note that the referenced studies have primarily focused on grammatical properties with simple and decontextualized language usage (e.g., using the simple past tense on a grammaticality judgment test). The influence of interleaved practice and blocked practice on contextualized language usage (e.g., making requests or suggestions) is unclear. This topic is worth exploring because effective practice schedules facilitate the proceduralization and automatization of L2 pragmatic knowledge, which can promote accuracy and fluency of learners’ L2 pragmatic behaviors (Dekeyser, Reference DeKeyser2007). The accurate and fluent pragmatic performance produced through effective practice can further help EFL learners use English as a lingua franca to conduct intercultural communication smoothly, avoid communication misunderstandings, and establish rapport among interlocutors from various cultural backgrounds (Zhang, Reference Zhang2021, Reference Zhang2022).

Studies on practice in the field of instructional pragmatics

A myriad of (quasi-)experimental studies and meta-analyses (e.g., Plonsky & Zhuang, Reference Plonsky, Zhuang and Taguchi2019; Taguchi, Reference Taguchi2015) have reported the positive influence of instruction (e.g., consciousness-raising activities and metapragmatic explanations) on promoting L2 learners’ pragmatic competence. Specifically, consciousness-raising activities (e.g., form-comparison tasks) draw L2 learners’ attention to the target pragmatic feature and enhance its salience. The awareness that students establish from consciousness-raising activities might improve their subsequent intake of the target pragmatic feature (Schmidt, Reference Schmidt, Kasper and Blum-Kulka1993). With respect to metapragmatic explanations, they “could ensure learners’ pragmatic awareness at the level of understanding” (Li, Reference Li and Taguchi2019, p. 121) and acquaint students with L2 pragmatic knowledge (e.g., pragmalinguistic forms used for making suggestions). Learners who are equipped with pragmatic knowledge pertaining to the target pragmatic feature might be more likely to notice the feature during real-life communication in the L2 (Zhang, Reference Zhang2021). In addition to awareness-raising activities and metapragmatic explanations, both Plonsky and Zhuang (Reference Plonsky, Zhuang and Taguchi2019) and Taguchi (Reference Taguchi2015) reported that pragmatics instruction entailing practice is more efficacious in developing learners’ L2 pragmatic competence than instruction without follow-up practice.

In the line of research on L2 pragmatics practice, a small number of studies have explored the effects of the intensity/amount of practice on L2 pragmatic competence (e.g., Li, Reference Li2012, Reference Li, Taguchi and Sykes2013). For instance, in Li’s (Reference Li2012) study, a regular training group, an intensive training group (i.e., practicing the target speech act twice as much as the regular training group), and a control group (i.e., without practice) received metapragmatic instruction on making requests in Chinese. The results of oral discourse completion tasks (DCTs) showed that the intensive training group produced significantly more appropriate requests than the control group and the regular training group. In a subsequent study, Li (Reference Li, Taguchi and Sykes2013) examined the influence of the amount of practice on L2 pragmatic competence. The results of oral DCTs revealed that practicing requests four times was sufficient for enhancing the accuracy of the L2 learners’ production of requests in Chinese.

To summarize, studies that examine the effectiveness of practice in L2 pragmatic development are sparse (Li, Reference Li and Taguchi2019). The small number of studies on L2 pragmatics practice have revealed the positive effects of repeated practice on L2 pragmatic competence (Li, Reference Li2012, Reference Li, Taguchi and Sykes2013). A more intriguing question is how repeated practice should be arranged to optimize L2 pragmatics learning/teaching (e.g., blocked practice versus interleaved practice). Crucially, no studies have investigated the influence of practice schedules on L2 pragmatic competence. This strand of research is significant because practice schedules are instrumental in improving the quality of L2 pragmatics instruction. Furthermore, optimal practice schedules contribute to the proceduralization and automatization of L2 pragmatic knowledge.

Target pragmatic features: Suggestions and requests

The speech act of suggestions refers to “a language function whereby one of the interlocutors may express his/her belief that a course of action could be profitably taken up by either or both of them” (Marmaridou, Reference Marmaridou and Fisiak1990, p. 565). The speech act of requests is defined as the behavior of the speaker who expresses his/her “expectation of the hearer with regards to prospective action” (Blum-Kulka et al., Reference Blum-Kulka, House and Kasper1989, p. 11). One of the reasons why these two pragmatic features were selected as the foci of the current study is that they are ubiquitous but face-threatening (Brown & Levinson, Reference Brown and Levinson1987). As far as hearers are concerned, making a request or suggestion may impose on the negative face of the addressee who wants to be undisturbed and the positive face of the addressee who strives for agreement (Brown & Levinson, Reference Brown and Levinson1987; Taguchi, Reference Taguchi2022). Realizing these two speech acts can also be face-threatening for speakers, who might hesitate to make a suggestion or request “for fear of exposing a need or risking the hearer’s loss of face” (Blum-Kulka et al., Reference Blum-Kulka, House and Kasper1989, p. 11). The face-threatening characteristics of these two speech acts may pose a great challenge for L2 learners to appropriately perform these two pragmatic features in the target language. Thus, it is important to equip L2 learners with pragmatic knowledge and engage them in activities to practice making requests and suggestions.

Another reason for selecting these two pragmatic features is based on the discriminative contrast hypothesis (Carvalho & Goldstone, Reference Carvalho and Goldstone2014; Zulkiply & Burt, Reference Zulkiply and Burt2013). The low between-category discriminability of making requests and suggestions can be reflected in two respects: (a) pragmalinguistic similarities and (b) sociopragmatic similarities (Thomas, Reference Thomas1983). Pragmalinguistic similarities refer to adopting similar linguistic forms to perform communicative acts, whereas sociopragmatic similarities refer to examining similar contextual factors (e.g., power, social distance) when making pragmatic choices. Specifically, with respect to pragmalinguistic similarities, the linguistic forms used for providing suggestions in Martínez-Flor and Alcón Soler’s (Reference Martínez-Flor and Alcón Soler2007) study and the linguistic devices used for making requests in Taguchi et al.’s (Reference Taguchi, Naganuma and Budding2015) study show that these two speech acts employ similar pragmalinguistic forms, entailing head acts (i.e., the core parts that are used for realizing the target act independently of other components), internal modifiers, and external modifiers. Table 1 presents examples of making requests and suggestions with similar head acts, internal modifiers, and external modifiers.

Table 1. Pragmalinguistic similarities between making requests and making suggestions

In addition to pragmalinguistic similarities, these two target pragmatic features share sociopragmatic similarities. When speakers make requests or suggestions, similar contextual variables, such as the power differences between interlocutors (P), their social distance (D), and the degrees of imposition (R), need to be considered (Brown & Levinson, Reference Brown and Levinson1987). For instance, when one makes a high-imposition request or suggestion to the addressee who has more power and longer social distance (e.g., a student asks a professor for a letter of recommendation, an employee makes a suggestion to his/her employer), both speech acts necessitate examining the influence of contextual factors on their pragmatic choices. More politeness is required in PDR-high situations to save the addressee’s face by adopting indirect strategies (e.g., using internal or external modifiers).Footnote ¹

Participants

Sixty-three sophomores (35 females, 28 males) from EFL classes at a public university in China participated in this study. Their ages ranged from 19.5 to 22.6 years (M _age = 20.7). They were intermediate English learners based on their scores (M = 84.39, SD = 3.95) on the TOEFL (Test of English as a Foreign Language), which they took three months prior to this study.Footnote ² Because the assessment tasks in this study involved speaking, the learners’ scores in the speaking and listening segments of TOEFL were also reported. Their speaking and listening skills were at an intermediate level (speaking: M = 20.35, SD = 2.49; listening: M = 20.51, SD = .12).Footnote ³ A background survey showed that no students had lived outside of China and all of them intended to obtain their master’s or doctoral degrees in the United States after graduating from college. Moreover, the students’ responses in the survey and the interviews with their EFL instructors showed that their focused instruction did cover the metapragmatic explanations of the target speech acts.

All learners received explicit instruction germane to how to make pragmalinguistically and sociopragmatically appropriate requests and suggestions in English. After the pragmatics instruction, all students took the pretest (Details regarding the pretest are discussed in the section “Procedure”). Then, the 63 learners were randomly assigned to two groups to practice the two target pragmatic features. The interleaved-practice (IP) group consisted of 31 learners, whereas the blocked-practice (BP) group consisted of 32 learners.

Procedure

The present study conducted a pretest, an immediate posttest, and a delayed posttest. The timeline of this study is illustrated in Figure 1.

Figure 1. Timeline of the current study.

During Week 1, all participants completed a background questionnaire. During Week 2, all students received two hours of pragmatics instruction on making requests and suggestions delivered by the researcher. The pedagogical design of the lecture was adapted from Li’s (Reference Li, Taguchi and Sykes2013) research. The instruction started with an awareness-raising task where the participants were asked to circle the appropriate requests or suggestions in English (Details are discussed in the section “Materials for Instruction”). The instructor then provided the learners with explicit metapragmatic explanations with respect to how to make pragmalinguistically and sociopragmatically proper requests or suggestions in English (i.e., declarative pragmatic knowledge; details are provided in the section “Materials for Instruction”). After the instruction, the participants were expected to master the declarative knowledge, which was reflected by the pragmalinguistic and sociopragmatic accuracy of their requests and suggestions. An example of a pragmalinguistically and sociopragmatically accurate request in a PDR-high situation is “I was wondering if you could write a letter of recommendation for me.” An example of a pragmalinguistically and sociopragmatically accurate suggestion in a PDR-high situation is “I would suggest that we have a biweekly meeting.”

Then, all participants completed multimedia discourse completion tasks (MMDCTs) as the pretest (Details of MMDCTs are discussed in the section “Instrument”). The design of administering the pretest after the instruction was inspired by Li’s (Reference Li2012, Reference Li, Taguchi and Sykes2013) studies and Li and Taguchi’s (Reference Li and Taguchi2014) research to “show the effects of practice only” (Li & Taguchi, Reference Li and Taguchi2014, p. 801) rather than the influence of metapragmatic instruction combined with practice. Having the pretest after the instruction can also avoid the potential influence of confounding variables, for instance, the variation in L2 learners’ prior declarative knowledge of making requests or suggestions. Administering a pretest after the instruction controlled for this variable because the instruction acquainted all learners with the same declarative knowledge regarding the target speech acts. Another purpose of the pretest was to assess the students’ initial knowledge regarding making requests and suggestions and to ensure that they successfully acquired the declarative knowledge (i.e., reflected by their high level of pragmatic accuracy) before moving on to the practice stage (Li, Reference Li, Taguchi and Sykes2013). After the pretest, the learners were randomly assigned to an IP group or a BP group.

Inspired by the design of the practice schedules in Suzuki’s (Reference Suzuki2021) study, at the practice stage of the current study, the IP group partook in intermixing practice (i.e., a-b-a-b-a-b …). For example, if the current practice task in the IP condition focuses on making a request, the subsequent practice involves making a suggestion. This pattern (i.e., request→ suggestion) was repeated 10 times in total. In contrast, the BP group first engaged in 10 practice tasks pertinent to making requests and then participated in 10 practice tasks involving making suggestions. In other words, if the current practice task in the BP condition pertains to making a request, the subsequent practice also describes a request-making scenario. It took approximately three minutes for the participants to complete each practice item. The entire practice session (i.e., 20 practice items) lasted one hour (More details are discussed in the section “Materials for Practice”). One thing worth mentioning is that the results from a series of pilot studies showed no order effects of the practice tasks. No significant difference was found between practicing making a request first and practicing making a suggestion first in either the BP (p = .356) or the IP condition (p = .291). The two groups’ practice schedules in the current study are presented in Table 2.

Table 2. Practice schedules

After the practice session, both groups completed an immediate posttest. Moreover, to explore the retention of the benefits of the interleaved or blocked practice, both groups completed a delayed posttest five weeks later.

Materials for instruction

At the stage of the pragmatics instruction, two sets of handouts were provided for all students; one set of handouts was focused on making requests, and the other set was related to making suggestions. Specifically, regarding the set of handouts concerning requests, the participants took part in a consciousness-raising activity involving two pairs of dialogues pertinent to making requests (see Appendix A). Each pair of dialogues (e.g., Scenario 1A versus Scenario 1B) consisted of a pragmalinguistically and sociopragmatically appropriate request and a pragmalinguistically and sociopragmatically inappropriate request in English. The purpose of this handout was to raise the learners’ awareness of the target speech act. The second handout provided a summary of the target pragmalinguistic forms and pragmatic strategies used for making requests (see Appendix B). The objective of the second handout was to acquaint the participants with explicit metapragmatic information with respect to making requests.

Similarly, the other set of handouts consisted of a consciousness-raising exercise related to making suggestions (see Appendix C) and a summary of the target pragmalinguistic forms and pragmatic strategies used for making pragmalinguistically and sociopragmatically appropriate suggestions in English (see Appendix D).

Materials for practice

Each target speech act entailed 10 practice tasks via computerized DCTs. Each DCT was designed based on the Michigan Corpus of Academic Spoken English (MICASE; Simpson et al., Reference Simpson, Briggs, Ovens and Swales2002). MICASE contains more than 190 hours of authentic real-life data from an array of speech events (e.g., conversations between friends in a PDR-low context, conversations between a professor and a student in a PDR-high situation) with 1.8 million words from 152 transcripts. This corpus encompasses various pragmatic features, including making requests and suggestions. The development of the materials for practice (i.e., corpus-based DCTs) consisted of three steps: (a) summarizing target pragmalinguistic forms used for making requests and suggestions based on previous studies (Martínez-Flor & Alcón Soler, Reference Martínez-Flor and Alcón Soler2007; Taguchi et al., Reference Taguchi, Naganuma and Budding2015) (see Appendices B and D), (b) searching request-making and suggestion-making data in PDR-high and PDR-low situations from MICASE (Here is a request-making example in a PDR-high context extracted from Transcript ID: COL999MG053 in MICASE. A student, a first language [L1] speaker of English, makes a request to a professor by asking, “I was wondering if you could comment on the differences of teaching at a small school versus a research university?”), and (c) developing corpus-based DCTs, including designing the prompts and background information based on the contexts of the examples from MICASE and creating exemplar responses by modifying the authentic data (e.g., replacing terminology with common words used in everyday life, removing repetitions and filler words).

For each corpus-based DCT, the learners first read a prompt describing the context and interlocutors. After 20 seconds, the participants were asked to provide their oral responses (i.e., requests or suggestions), which were recorded through the computer’s built-in microphone (see Figure 2). Then, an exemplar response with a metapragmatic explanation of the target speech act was provided as feedback for 20 seconds (see Figure 3). In total, the practice session included 20 tasks (i.e., 10 practice tasks × 2 target speech acts).

Figure 2. Sample practice task about making a request.

Figure 3. Sample feedback.

Note. The metapragmatic explanation translated into English is “Because borrowing a stapler from your friend Charlie is a PDR-low situation, adopting this direct strategy to make the request is appropriate.”

Instrument: Multimedia discourse completion tasks

In spite of being one of the most adopted tools to measure L2 learners’ pragmatic performance, DCTs have many shortcomings (Nguyen, Reference Nguyen and Taguchi2019). For instance, many scholars have pointed out that data collected from DCTs usually do not reflect real-life exchanges because DCTs often lack contextual cues (Golato, Reference Golato2003; Sykes & González-Lloret, Reference Sykes and González-Lloret2020). Despite the downsides, the advantages of DCTs are manifold. For example, they are effective in eliciting data directly germane to the target pragmatic feature while controlling for extraneous variables (e.g., social distance, degrees of imposition)(Golato, Reference Golato2003).

To address the drawbacks of traditional DCTs, MMDCTs (i.e., encompassing text-formatted prompts as well as dialogues presented aurally and visually via pictures) were designed for the current study. Specifically, to provide contextual cues, each MMDCT started with a prompt supplying contextual information (e.g., the relationship between conversationalists, background information). Additionally, a relevant image was incorporated below the prompt to provide more contextualization (Brown & Abeywickrama, Reference Brown and Abeywickrama2018). The prompt, image, and incomplete dialogue disappeared after 30 seconds. The learners were then instructed to speak to the computers. Their answers regarding requests or suggestions were recorded through the computers’ built-in microphones. After completing each MMDCT, the participants clicked on an arrow to proceed to the subsequent test item. Different from the practice stage, no feedback was provided at the testing stage. Each test lasted 50 minutes. Figure 4 and Figure 5 present a sample MMDCT involving making a request or suggestion, respectively.

Figure 4. Sample MMDCT of making a request.

Figure 5. Sample MMDCT of making a suggestion.

Moreover, three equivalent versions of the tests were designed for the pretest, immediate posttest, and delayed posttest to avoid practice effects. The three tests were counterbalanced across the three testing stages between the IP group and the BP group. The internal reliability of each of the three tests as indexed by Cronbach’s alpha coefficient was high for both the measure of pragmatic accuracy and the measure of fluency (i.e., pragmatic accuracy: pretest = .879, immediate posttest = .891, and delayed posttest = .893; fluency: pretest = .878, immediate posttest = .885, and delayed posttest = .896). Each test consisted of 32 items; 16 of them were target test items (i.e., eight items about requests and eight items about suggestions), and 16 test items were distractors (e.g., complimenting, congratulating). The sequence of the test items was randomized. Additionally, the target test items were piloted with 25 L1 speakers of American English and 11 proficient L2 speakers to make sure that the scenarios successfully elicited requests and suggestions. The target test items included in the three tests elicited requests and suggestions from 100% of the 36 L1 and L2 speakers.

With respect to the design of the target test items, first, each of them was designed based on the authentic data related to making requests or suggestions from MICASE (Simpson et al., Reference Simpson, Briggs, Ovens and Swales2002). Second, in accordance with the three contextual factors in Brown and Levinson’s (Reference Brown and Levinson1987) politeness theory (i.e., power, social distance, and the degree of imposition), the 16 target test items consisted of four PDR-high requests, four PDR-low requests, four PDR-high suggestions, and four PDR-low suggestions. The distinctions between the PDR-high and PDR-low scenarios were confirmed through a pilot study. A questionnaire with a five-point Likert-type scale entailing the target test items was conducted with 22 L1 speakers of American English and 10 proficient L2 speakers. The questionnaire adapted Taguchi’s (Reference Taguchi2022) survey and asked the 32 L1 and L2 speakers to rate how easy or challenging it was to perform the target pragmatic features (1 = very easy, 5 = very challenging). The results showed a significant difference between the PDR-high scenarios (M = 4.10, SD = .29) and the PDR-low scenarios (M = 2.18, SD = .61), t = 16.336, p < .001, d = 2.888 (large), suggesting that the target speech acts in the PDR-high scenarios were perceived as more challenging to perform compared with the target pragmatic features in the PDR-low situations. In addition, the scenarios in the target test items are different from the situations in the corpus-based DCTs at the practice stage.

Data analysis

The first RQ asked whether interleaved corpus-based practice is more effective than blocked corpus-based practice in improving Chinese EFL learners’ pragmatic accuracy and fluency when they make requests and suggestions.

To assess pragmatic accuracy (i.e., declarative pragmatic knowledge), based on the speech act theory (Austin, Reference Austin1962), politeness theory (Brown & Levinson, Reference Brown and Levinson1987), and the rating scales in former studies (e.g., Martínez-Flor & Alcón Soler, Reference Martínez-Flor and Alcón Soler2007), a four-level rubric (see Table 3) was adopted to measure the two groups’ speech acts of making requests and suggestions, which share pragmalinguistic and sociopragmatic similarities as discussed in the section “Target Pragmatic Features: Suggestions and Requests.” This rubric was also discussed with two EFL instructors who were L1 speakers of American English and have taught EFL for more than 10 years. With 16 target test items on the pretest, immediate posttest, and delayed posttest, respectively, the scores of each test ranged from 0 to 48. Three trained raters with PhD degrees in second language studies scored the data independently. High interrater reliability was achieved among the three raters for all three tests: pretest = .863, immediate posttest = .892, and delayed posttest = .920. The average scores among the raters were used for the data analyses.

Table 3. Rubric for rating pragmatic accuracy of the target speech acts

To measure fluency (i.e., procedural pragmatic knowledge) of the learners’ target speech acts, the current study focused on speed fluency assessed by speech rate. Even though speech rate is usually regarded as a global measure of fluency (Huensch & Tracy-Ventura, Reference Huensch and Tracy-Ventura2017), “it provides adequate indication of the overall speed at which a person speaks” (Bui et al., Reference Bui, Ahmadian and Hunter2019, p. 5). Moreover, this temporal measure can extend the traditional analyses of target pragmatic features with a sole focus on accuracy and provide a complementary approach to dissecting L2 pragmatic production by assessing fluency (an indicator of processing), which is usually neglected in L2 pragmatics research. Specifically, speech rate in this study was measured by the average number of uttered English syllables per second when the participants made a request or suggestion, excluding false starts, repetitions, and repairs (Derwing et al., Reference Derwing, Munro, Thomson and Rossiter2009). Three raters manually calculated the speech rate independently. High interrater reliability was achieved among the raters for all three tests (pretest = .882, immediate posttest = .891, and delayed posttest = .895).

SPSS 26 was used to conduct the statistical analyses in the current study. Before running the statistical analyses, the researcher checked the data to make sure that the assumptions for t-tests were met. First, the assumption of a continuous dependent variable has been met. In this study, the EFL learners received scores on a measurement scale, and there was an equal distance between each point on this scale. For instance, the difference in pragmatic accuracy between scoring 45 and 46 points on a 48-point test is the same as the difference between scoring 30 and 31 points, that is, one point. Second, according to the results of Shapiro–Wilk tests, the two groups’ scores on the three tests were normally distributed (see Table 4). Third, the results of Levene’s tests showed that the assumption of homogeneity of variances has been met (see Table 5).

Table 4. Results regarding normality

Table 5. Results regarding homogeneity

To answer RQ 1 investigating whether interleaved corpus-based practice is more effective than blocked corpus-based practice in improving Chinese EFL learners’ pragmatic accuracy and fluency when they make requests and suggestions, an independent samples t-test was carried out to explore the between-groups differences on the immediate posttest. Furthermore, the effect size of the between-groups contrast was assessed (small = .40, medium = .70, large = 1.00; Plonsky & Oswald, Reference Plonsky and Oswald2014). To address RQ 2 regarding the long-lasting effectiveness of interleaved corpus-based practice or blocked corpus-based practice in Chinese EFL learners’ pragmatic accuracy and fluency over time, a paired-samples t-test was conducted to examine the within-groups differences in the pretest data and the delayed posttest data. Furthermore, the effect sizes of the within-groups distinctions were evaluated (small = .60, medium = 1.00, large = 1.40; Plonsky & Oswald, Reference Plonsky and Oswald2014). In addition, an independent samples t-test was conducted to investigate the between-groups differences on the delayed posttest. It is worth mentioning that running multiple statistical analyses on the same data may enhance the possibility of committing a Type I error (Brown, Reference Brown1990). One of the approaches to attenuating the adverse influence is using a Bonferroni correction, which generates a more conservative alpha level and decreases the likelihood of making a Type I error in such cases (Loewen & Plonsky, Reference Loewen and Plonsky2016). Therefore, in the current study, an alpha level of .05 was reduced to .017 due to conducting the t-tests three times on the data set (α = .05/3). This new alpha level was used to examine the significance of the p values in this study.Footnote ⁴

Prior to answering the two RQs, the researcher examined the pretest data to explore whether there were preexisting differences between the IP group and the BP group on the pretest in terms of pragmatic accuracy and fluency. The results from an independent samples t-test demonstrated no noticeable difference in pragmatic accuracy of the target speech acts on the pretest between the IP group (M = 39.98, SD = 3.71) and the BP group (M = 39.01, SD = 4.03), t(61) = .991, p = .326, d = .250. Moreover, no significant difference in fluency of the target pragmatic features was found on the pretest between the IP group (M = 2.69, SD = .32) and the BP group (M = 2.63, SD = .35), t(61) = .786, p = .435, d = .198.