Japanese learners’ spoken requests in the study abroad context: appropriateness, speech rate and response time

ABSTRACT Studies that have measured pragmatic fluency in the form of planning time and speech rate, as indicators of processing ability, have confirmed that learners’ processing ability and pragmatic knowledge are independent from each other. The present study is a longitudinal, developmental investigation which investigates whether the study abroad stay (SA) has an effect on the overall appropriateness, planning time and speech rate in high and low imposition situations of ten L2 Japanese learners’ oral requestive output during their overseas sojourn. With the exception of appropriateness which had the most positive change, (mainly with low imposition requests), the study found that the SA had minimal positive impact on the variables examined, suggesting that activation and processing of pragmatic knowledge needs enhancement from other sources, and that learners need a longer period of time to develop appropriate pragmatic output across a range of academic encounters. Results further revealed much individual variation, further highlighting the complexity of examining pragmatic performance over time.


Introduction
Unlike traditional language classrooms, a learning environment such as the study abroad (SA) context has the unique potential to provide second language learners with ample opportunities to observe and interact with native speakers of the target language (TL), and hence practise the foreign language in its natural environment, and in a variety of different social situations. As a result, in addition to the acquisition of grammar, vocabulary, and pronunciation, the SA context can also enhance learners' fluency, through repeated opportunities for spontaneous, authentic interactions in the TL community (Freed 1995;Segalowitz and Freed 2004;Taguchi 2012;Tavakoli and Wright 2020). Equally importantly, spending time abroad has been found to potentially enhance L2 learners' intercultural sensitivity (Czerwionka, Artamonova, and Barbosa 2015;Song 2020), i.e. their ability to develop 'a positive emotion towards understanding and appreciating cultural differences' (Chen 1997: 5), and aspects of their pragmatic ability (e.g. Perez Vidal and Shively 2019;Shively and Cohen 2008), i.e. their ability to use and interpret language appropriately in a given sociocultural context (Bachman and Palmer 2010;Taguchi 2012: 6).
Studies, however, have reported mixed results and much variability in the pragmatic gains (for a full review see Taguchi 2012;Xiao 2015), making it now clear that the SA context alone is not a panacea for pragmatic development (Kinginger 2008). Learners who do not make the most of the SA experience and do not engage sufficiently with NSs in the TL society may see no improvement in their pragmatic comprehension or production, and/or oral fluency. It has also been suggested that, along with intensity of contact, learners' individual differences (linked to personality-related factors), level of linguistic proficiency, length of stay and/or attitudes towards the L1 culture can additionally influence learners' development of pragmatic knowledge and processing ability in a SA context (Bataller 2010;Kinginger 2008;Li 2014). Therefore, different gains have been reported across different aspects of pragmatic competence (e.g. Barron 2003;Ren 2015;Schauer 2007Schauer , 2009Woodfield 2012) and considerable individual differences in L2 gains often stand out (Perez Vidal and Shively 2019). Such mixed research findings were also supported by studies that documented very little development over time (Barron 2006;Economidou-Kogetsidis and Halenko 2022;Iwasaki 2010).
Second language learners' pragmatic ability has traditionally been examined mainly in relation to their pragmatic performance, measuring either their pragmalinguistic knowledge (i.e. knowing what linguistic resources are available to perform specific language functions) and/or their sociopragmatic knowledge (i.e. understanding features of context and selecting the most appropriate linguistic resource in a given context) (Culpeper, Mackey, and Taguchi 2018;Leech 1983;Thomas 1983). However, pragmatic performance and development have been found to be supported not only by pragmatic knowledge, but also by the processing ability for utilising such knowledge during interaction (Kasper 2001;Li 2014;Taguchi 2012). According to Kasper (2001), pragmatic competence involves the acquisition of pragmatic knowledge and the gaining of automatic control in processing it (cited in Taguchi 2007: 117), and as Taguchi explains: ... indications of performance fluency, such as planning speed and oral fluency, which have been used in taskrelated research, could provide additional useful information from a processing perspective, that is, how rapidly [learners] can process pragmatic knowledge and convey speech intentions. (2007: 117) Studies that have measured pragmatic fluency in the form of planning time and speech rate as indicators of processing ability (e.g. Lee 2013;Li 2014;Taguchi 2007) have, in fact, confirmed that learners' processing ability and pragmatic knowledge are independent from each other. This has led to the claim that, in order to better understand how L2 pragmatic production develops during study abroad, it is important to investigate both the development of pragmatic knowledge and the learners' processing ability (Li 2014: 104) which is linked to pragmatic fluency.
Yet, within the relevant literature, only a small number of studies (e.g. Lee 2013;Li 2014;Taguchi 2007Taguchi , 2011Taguchi , 2012 have given attention to the processing dimension of speech acts, and very few have examined the potential relationship between L2 pragmatic knowledge and L2 fluency (Tavakoli and Wright 2020: 82). In addition, despite the fact that research on language learning in SA settings has generally revealed that oral fluency, unlike other areas of linguistic proficiency, consistently improves during SA (Freed 1995;Juan-Garau 2018;Wright 2018), to the best of our knowledge, only a handful of studies (e.g. Li 2014, Taguchi 2011 have so far examined the processing dimension of L2 requests from a SA perspective. The present study aims to fill this gap by attempting to enhance our understanding of the impact of the SA context not only on pragmatic knowledge but on pragmatic fluency as well. The study therefore measures two components of pragmatic performance: pragmatic knowledge (measured by overall appropriateness in pragmatic production with specific reference to requests), and pragmatic fluency (measured by planning time and speech rate) as the processing ability which demonstrates the knowledge fluently in real-time tasks.
More specifically, the study is a longitudinal, developmental investigation which investigates pragmatic fluency (though planning time and speech rate), and overall pragmatic appropriateness in relation to ten L2 Japanese learners' oral requestive output during their SA sojourn. The research questions of the present investigation are as follows: 1. Does the SA stay have an effect on the overall pragmatic appropriateness of Japanese learners' L2 spoken requests in high and low imposition situations?
2. Does the SA stay have an effect on the planning time of Japanese learners' L2 spoken requests in high and low imposition situations? 3. Does the SA stay have an effect on the speech rate of Japanese learners' L2 spoken requests in high and low imposition situations?
Within the context of our study, we approach appropriateness from a perception point of view, and define pragmatic appropriateness as the extent to which an utterance is perceived by the hearer as suitable for a particular purpose in a specific sociocultural context in such as way so that it has the 'proper level of politeness, directness and formality, in the given situations' (Taguchi 2011: 273). We also consider oral fluency from a cognitive perspective, and see it as 'the rapid, smooth, accurate, lucid and efficient translation of thought or communicative intention into language' which develops in terms of automaticity, speed, and 'effective processes in planning and constructing words and sounds in utterances and connected speech' (Tavakoli and Wright 2020: 3). Within this cognitive dimension of fluency, we define L2 pragmatic fluency as the type of fluency which allows learners to 'gain automatic control in processing linguistic information' and to rapidly 'process pragmatic knowledge and convey speech intentions' (Taguchi 2007: 117;Taguchi 2012: 9).

Background
Interlanguage research suggests that the SA context alone does not always lead to pragmatic gains for all learners (Taguchi 2014), and there is much variability in acquiring target-like pragmatic competence. This observation, however, has concerned primarily learners' pragmatic knowledge as demonstrated only through the frequency of certain pragmalinguistic strategies and/or their sociopragmatic choices, rather than through the examination of learners' processing ability together with pragmatic fluency, i.e. how learners gain automatic control in processing and planning pragmalinguistic and sociopragmatic information in order to effortlessly and successfully convey speech intentions. The following section reviews SA and non-SA investigations that have focused on the processing dimension of speech acts.
The processing dimension of speech acts As Li (2014: 104) argues 'because pragmatic performance is jointly supported by pragmatic knowledge and the processing of such knowledge (Taguchi 2012), improvement in pragmatic performance also involves the development of both knowledge and processing ability (Kasper 2001)'. A number of studies (e.g. Li 2012; Li and Taguchi 2014;Taguchi 2012) have found that pragmatic knowledge and processing ability follow different developmental patterns both in instructed and uninstructed settings.
Such studies have also viewed learners' processing ability as pragmatic fluency measured according to planning time and speech rate or response time. While planning time is associated with the conceptualisation stage of speech production, speech rate is primarily associated with the formulation stage of speech production (Yuan and Ellis 2003), and involves various grammatical and phonological encoding processes that derive from the speaker's general linguistic knowledge (Levelt 1989).
Li's (2014) SA investigation is particularly relevant to the focus of the present study. The study focused both on the development of pragmatic knowledge and processing ability in order to better understand how L2 pragmatic production develops during study abroad. Using a computerised oral discourse completion task, it examined the effects of different levels of linguistic proficiency on the requestive production of 31 American learners of Chinese during a 15-week SA sojourn in Beijing, measuring appropriateness ratings as indicators of pragmatic knowledge, planning time (amount of time taken for preparing the oral response) and speech rate (number of syllables produced per minute) as indicators of processing ability. The learners were divided into two proficiency groups (intermediate and advanced), and the appropriateness of their responses was rated by two Chinese native speakers who were recruited as raters. Li (2014) found that both proficiency groups made significant improvement in appropriateness during the SA sojourn, as they changed their request production in terms of alerters, head acts, and internal and external modification. Neither group, however, reduced planning speed by the end of the SA period, while only the advanced group was found to gain in speech rate. Taguchi's (2011) investigation was also one of only a handful of pragmatic studies that dealt with the SA context and learners' fluency. The study aimed to investigate the effect of language proficiency and SA experience on Japanese learners' requests and opinions in English. These speech acts were examined for appropriateness, pre-task planning and speech rate (in addition to grammaticality), and were tested both in low and in high-imposition situations. Contrary to Li's (2014) results, this study revealed that, unlike proficiency and situation type, the SA experience alone had little impact on the appropriateness and grammaticality of the learners' speech actsthe students who spent a year abroad did not do better than those who stayed in their home country. The study further found an effect of situation type on planning time but not on speech rateboth groups took a shorter time to prepare for low rather than high-imposition speech acts while speech rate was found to be similar for both types of situations. The study further revealed a proficiency effect on speech rate but not on planning time, indicating that 'speech rate and planning time have distinct characteristics from each other that represent different underlying cognitive and affective mechanisms' (Taguchi 2011: 286).
More recently, Song (2020) investigated the effects of a short-term SA programme on developing the intercultural competence of American students studying Korean as a foreign language and who had spent 6 weeks in Korea. The study further examined the contribution of intercultural competence on students' oral proficiency gains. Results revealed that, even though students made significant improvement in various aspects of perceived intercultural competence, they showed no improvement in fluency and pragmatic competence. It was therefore confirmed that an increase in the learners' intercultural competence did not necessarily lead to oral proficiency gains or to gains related to pragmatic ability, and that short SA period such as 'six weeks … appears to be an insufficient duration for students to make measurable changes in pragmatic competence' (2020: 18). Taguchi's (2007) earlier, non-SA investigation, examined the refusal and request production of 59 Japanese learners of English (at two different proficiency levels) for overall appropriateness, planning time and speech rate (number of words spoken per minute). The study focused on the impact that task variation has on L2 oral output in order to provide insights into what makes a task more difficult pragmatically. It therefore utilised two situation types: one where the power relationship, distance between interlocutors and degree of imposition were small, and a second type of situation where the power was unequal, and the distance and imposition were large. In line with Taguchi (2011), results from this study showed that situation type had a significant effect on appropriateness scores: high PDR (power, distance, imposition) acts were more difficult to produce than low ones regardless of the learners' proficiency level. All learners were also found to be faster in planning low-PDR speech acts as high PDR acts took significantly longer to prepare. In terms of speech rate, similarly to Taguchi's (2011) investigation again, this study also found proficiency effects on speech rate but not on planning time. Even though all learners were significantly more fluent in producing low than high-PDR speech acts, the higher proficiency group outperformed the low proficiency one in speed. Importantly, the study further revealed that 'longer planning time did not result in more appropriate speech act production or more fluent articulation of production' (Taguchi 2007: 129).
Using data from 48 Japanese learners of English at a bilingual university in Japan, Taguchi's (2012) longitudinal study also aimed to explore the relationship between knowledge and processing in pragmatic competence. Among other things, the study employed a computerised pragmatic speaking test to examine appropriateness scores, planning and speech rate in relation to learners' requests and opinions in high and low-imposition situations. Taguchi (2012: 134) found that appropriateness increased over time for the low imposition situations but that there was no gain in the later period for the high imposition speech acts. The author attributed this finding to the frequent occurrence of low imposition everyday speech acts in the bilingual campus the students were immersed in. Regarding planning speed, a constant development was revealed regardless of the situation type, and speech rate for both types of situation similarly showed a significant increase initially but this rate later became stagnant. The study therefore confirmed that there was no effect of situation type on the development of fluency as far as planning and speech rate were concerned. Following Taguchi's investigation (2007), Lee's (2013) study similarly focused on the processing dimension of speech acts by examining the impact of power and social distance on measures of processing speed and appropriateness. Using oral role play tasks and retrospective verbal reports, Lee (2013) examined fluency in the refusal production of 40 Korean learners of English in Seoul (of low and high proficiency), 20 Korean NS students, and 20 English NSs. The participants' production was analysed for appropriateness rating, planning time, speech rate and pause length, while 3 ENSs were recruited to evaluate learners' refusals using Taguchi's (2007: 121) rating scale. The study's results regarding appropriateness were in line with Taguchi's (2007). It was found that the appropriateness scores of both learner groups were lowest in situations where the listener had lower social status (power-low situations) and highest in power-equal situations, confirming that power-low refusals were most difficult to produce. Regarding planning time, high proficiency learners took approximately the same time in both power-equal and power-unequal situations. Low proficiency learners, however, took significantly longer to plan refusals for interlocutors of lower status, a pattern which mirrored the production of the English NS group. Finally, results regarding speech rate revealed that both groups of learners were fastest when it came to refusing in power-equal refusals, and slowest when refusing in power-low situations.
While there is some empirical evidence to suggest that variables such as proficiency and situation type often have an effect on the processing dimension of speech acts, more longitudinal empirical work is needed to examine the extent to which the SA context has an effect on the overall appropriateness and processing of L2 spoken requests, and on the effect of situation type on the appropriateness and pragmatic fluency of SA learners' speech acts. The present study is a step towards this direction and aims to fill some of this gap by focusing on the requests of Japanese SA learners in the UK. The section that follows describes the participants, and the methodological procedures of the present study.

Participants
Ten Japanese exchange students at a UK university in Northwest England volunteered to participate in the study, motivated by the opportunity to track their own linguistic development over the study abroad period. In this way, reciprocity of research goals was achieved (Creswell 2009) in so far as both researcher and participant benefit. The participants were enrolled on various teacher training and business communication programmes but came together to receive additional English language classes as part of the international exchange. They comprised five females and five males who ranged in age from 20 to 22. On average, they had six years of formal English language instruction in Japan, and they all met the CEFR B2 level requirement of the exchange programme. Four students reported previous study abroad stays in English speaking countries of between two weeks and two months.

Data collection
The students completed an oral computer-animated production task (CAPT) to evaluate their pragmatic growth over the SA stay. The CAPT is designed to elicit oral responses via virtual role plays, and requires a three-step read, listen, speak process. Participants first read the brief contextual information about each request scenario, watch and listen to an initial turn supplied by an animated interlocutor, then provide a situationally-appropriate oral response. The CAPT instrument, which has the benefits of being able to simultaneously capture large amounts of oral data whilst maintaining controlled conditions, has been used as a successful measure of pragmatic development before (Halenko 2021). For this specific study, the CAPT was administered in the same way, in the same computer lab at three equidistant time points during the SA sojourn (month 1-October, month 5-February, month 8-May). Each test contained identical scenarios but were ordered differently to reduce test effects. The three-month gap between test periods, and the high likelihood that the CAPT scenarios were contextually familiar to the participants even before encountering them within the CAPT, also helped minimise test effects.
The CAPT contained six request scenarios (requests for action), differentiated by level of imposition (3 high-imposition situations, 3 low-imposition situations), which were set in the familiar context of a university campus (Table 1). The four animated interlocutors (librarian, campus security guard, accommodation officer and class tutor) were considered higher status (+P) but otherwise varied on the levels of familiarity (+/−SD) or imposition of the request presented (+/−R). The sixscenario CAPT was completed in approximately fifteen minutes though no overall time limit was imposed. All responses were recorded and transcribed for analysis.
Low imposition requests were those which matched the role and status of the student e.g. requesting to collect worksheets from a tutor. Such low imposition requests would not generally require high levels of mitigation since they fall within the roles of responsibilities of academic staff. High imposition requests, on the other hand, were those where the status of the staff member is challenged e.g. negotiating a new deadline on an assignment. The academic staff in these situations would not be obliged to comply with the request and an additional burden in terms of time and effort may apply in doing so. Within the high imposition requests in this study, the interlocutor is also being asked to bend the institutional rules which would typically warrant high levels of mitigation in the form of linguistic (e.g. situationally-appropriate request strategies) or non-linguistic (e.g. brief and timely acts) status-preserving strategies (Bardovi-Harlig and Hartford 1993). These additional moves are rapport-building and help ensure the request is accomplished in a favourable way. Including an examination of low/high imposition requests helped provide a more comprehensive picture of request performance over time. Examples of (a) low and (b) high imposition request items on the CAPT were as follows: (a) Scenario 1: Book a study room at the library.
You want to find out how to book a study room.

Data analysis
To aid comparability of fluency investigations in L2 pragmatics, three types of fluency measures employed in existing literature (Li 2014;Taguchi 2007Taguchi , 2011Taguchi , 2012 were also used here to analyse the Japanese learners' request performance: (a) appropriateness ratings measured the extent of pragmatic knowledge, whilst (b) pre-task planning time, and (c) speech rate indicated pragmatic processing ability. This section provides details on each of these measures. Appropriateness was assessed using a six-point Likert scale (0-5). This study drew on Taguchi (2011) to define pragmatic appropriateness as the extent to which an utterance is perceived by the hearer as suitable for a particular purpose in a specific sociocultural context in such as way so that it has the 'proper level of politeness, directness and formality, in the given situations' (Taguchi 2011: 273). This definition was subsequently used for an adapted version of the rating scale featured in Halenko (2021). The same rating scale for this study included aspects of both pragmalinguistic and sociopragmatic features of discourse in order to gain a holistic assessment of the request responses (Table 2). Two EFL tutors from the study site (1 male and 1 female), with similar levels of teaching experience (10-15 years), agreed to rate all the oral requests following the final test stage at T3. Both tutors participated in an initial briefing session for standardisation purposes. Rating involved reading through the transcripts and providing a rating based on the scale provided in Table 2. The total number of request tokens rated was 168 scenarios (accounting for absence of data from two participants at T3). Responses which differed more than 1-point within the scale were reviewed and consensually adjusted. High levels of interrater reliability were achieved (0.85).
Planning time was operationalised on the basis of time taken to mentally prepare an appropriate oral response for each scenario on the CAPT. In this study, planning time was measured from the moment participants finished reading the prompt out loud on each slide to when they began producing an oral response. This pre-production period is referred to as offline planning (Yuan and Ellis 2003). Participants were permitted unlimited time to mentally prepare their oral responses. This natural break in speech production was considered the planning period and was measured in number of seconds. Audacity computer software allowed for an accurate measurement of planning time.
Speech rate was also captured as another measure of performance speed. Learners' speech rate was calculated based on the number of words spoken per minute (Lee 2013;Taguchi 2011). In line with previous studies (Taguchi 2007), 'pruned' speech (Lennon 1990) was used where hesitations, filled and unfilled pauses, repetitions and false starts, for instance, are excluded to obtain a more accurate measure of fluency speed. I would feel satisfied with this response because the levels of directness, politeness and formality are generally appropriate and effective. The expressions may contain several non-L2-like features, but the expression would be regarded as achieving minimal levels of appropriateness for a positive outcome, nevertheless. 2 I would not feel very satisfied with this response because the levels of directness, politeness or formality are not sufficiently appropriate or effective for the situation. Features are more inappropriate than appropriate and fail to achieve the satisfactory levels of indirectness, or the expressions contain insufficient mitigation for a positive outcome. 1 I would not feel satisfied at all with this response because the levels of directness, politeness and formality are entirely inappropriate and ineffective for the situation. It is difficult to imagine a positive outcome could be offered. 0 No response provided Since the group's pragmatic production and processing capabilities were tracked on three occasions (T1, T2 and T3), a repeated measures ANOVA was selected to measure performance across the time points.

Effects of SA on pragmatic appropriateness (RQ1)
The first research question asked whether SA stay has an effect on the overall appropriateness of learners' L2 spoken requests, and whether learners' appropriateness scores differ across the three time periods. Table 3 presents the descriptive statistics of the three measures, i.e. the mean average and the standard deviation (SD) of the appropriateness ratings in relation to the three time periods for all six situations. As can be observed from Table 3, the appropriacy mean average in T1 was 33.0, in T2 36.6 and in T3 39.3, therefore indicating an increase in appropriateness with time. Yet, it can also be observed that the SD increases with time, importantly suggesting more within group differences with time.
Repeated measures ANOVAs were run in order to examine whether this increase was statistically significant. Overall significance was shown at a p < .05 level [F(2, 14) = 7.33, p = .007, partial eta squared = .512]. Posthoc comparisons showed that no significant main effects in appropriateness ratings were present between T1 and T2 (p = .220) or between T2 and T3 (p = .355). The improvement in appropriateness ratings between T1 1 and T3, however, was significant at a p < .05 level (p = .024), suggesting that participants needed a longer period of time to develop their pragmatic competence and improve the appropriateness of their request production.

High-imposition and low-imposition requests and appropriateness
The study further examined whether the appropriateness of learners' requests changed over time according to situation type, thus the appropriateness of high-imposition and low-imposition requests was analysed separately. As far as high-imposition requests were concerned, the descriptive statistics (mean average and standard deviation) in Table 4 during T1-3 showed a moderate increase in appropriateness with time (T1: 17.5, T2: 18.6, T3: 19.5). However, the repeated measures ANOVA revealed no significant effects overall [F(2, 14) = 1.59, p = .238, partial eta squared = .186]. This confirmed that the appropriateness of the learners' high-imposition requests did not significantly improve with time.  Turning to low-imposition requests, the descriptive statistics again showed a more marked increase in appropriateness with time (Table 4) (T1:15.5, T2:18.0, T3:19.75). Importantly, this time the repeated measures ANOVA showed that this increase was statistically significant [F(2, 14) = 5.03,.p = .023, partial eta squared = .418]. Pairwise comparisons of time revealed increases from T1 to T2, and from T2 to T3 were not significant (p = .411, p = .753 respectively). Therefore, the appropriateness of the learners' low-imposition requests significantly improved only when comparing their performance at the beginning (T1) of their SA experience with the last phase (T3) of their stay (p = .021, p < .05)

Effects of SA on planning time (RQ2)
The second research question asked whether the SA stay had an effect on the planning time of learners' L2 spoken requests across the three-time SA periods, and in high and low imposition situations. Table 5 shows the average planning time (in seconds) for all situations at each time period T1-T3. As can be seen from Table 5, the mean average scores show that the amount of planning time needed for learners to plan their responses decreased after time period 1 (T1: 2.13 seconds, T2: 1.33 seconds, T3: 1.33 seconds). The SD also decreased over time (0.96, 0.73, 0.44) a result which suggests more within group differences in T1 than at T3. The ANOVA tests revealed that there was no overall significant difference in planning times when all situations were calculated together [F(2, 14) = 4.53, p = .08, partial eta squared = .393].

High-imposition and low-imposition requests and planning time
The study further examined whether the planning time of learners' requests changed over time according to situation type. The planning time of high and low-imposition requests was again further analysed separately. Table 6 shows the mean average and SD in relation to the planning time of high-imposition requests during T1-T3. A slight decrease is observed between T1 and T2 whereas a slight increase is shown between T2 and T3. However, the repeated measures ANOVA confirmed that these differences between time points were not significant overall [F(2, 14) = 2.36, p = .131, partial eta squared = .252].
Results in relation to low imposition requests during T1-T3 showed a decrease with time (Table 6 -T1:1.24, T2: 0.65, T3: 0.52). It can also be observed that the SD also decreases with time, importantly suggesting less within group difference with time. Once again, the repeated measures ANOVA revealed that these mean differences were not significant [F(1.119, 14) = 4.01, p = .08, partial eta squared = .364].

Effects of SA on speech rate (RQ3)
The last research question of the present study asked whether the SA stay has an effect on the speech rate of learners' L2 spoken requests across the three time SA periods, and in high and low imposition situations. The descriptive statistics shown in Table 7 for all six situations indicated that the average number of words spoken per minute at the group-level in the three time periods increased overall (T1: 670.64. T2: 786.21, T3: 842.23). The repeated measures ANOVA confirmed that this increase was significant overall [F(2, 14) = 4.34, p = .034, partial eta squared = .383] but only evident at the T1-T3 time period (T1-T2, p = .295; T1-T3, p = .05).

High-imposition and low-imposition requests and speech rate
The learners' speech rate was again analysed separately for high and low imposition requests. Table 8 presents the mean average and SD in relation to high-imposition requests, and indicates an increase during the three time periods (T1:312.3, T2:389.9, T3:422.5). Importantly, the ANOVA tests confirmed that these descriptive differences were significant overall (F(2, 14) = 8.72, p = .003, partial eta squared = .555). Further statistical pairwise comparisons showed that the significance lay with T1 and T2 (p = .044 significance at a p < .05 level) and T1 to T3 (p = .005 significant at a p < .01 level). In other words, learners increased their fluency with high imposition requests at the very beginning of their sojourn and also when their initial performance was compared with that at the end of their stay.
As far as low-imposition requests were concerned, the descriptive statistics showed some increase with time in speech rate (T1:358.32, T2: 396.31, T3: 419.75) (see Table 8). It can also be observed that the SD fluctuated across time periods, importantly suggesting within group differences. Despite these descriptive differences, however, the ANOVA tests confirmed that this increase was not significant [F(2, 14) = 1.15, p = .346, partial eta squared = .141].

Discussion
This longitudinal study set out to examine the impact of SA on the appropriacy of pragmatic production and linguistic processing efficiency, using planning time and speech rate as temporal measurements. When analysing performance and processing with L2 learners, one might consider a SA period to have had a positive effect on pragmatic growth if cumulative analyses demonstrated the following: request responses are evaluated as increasingly appropriate the longer the learners are immersed in the target environment, the time learners take to plan a response pre-production appears to decrease, and the rate at which responses are articulated increases over time. If we were to map this simplistic assessment to the present study, it might suggest the SA stay was beneficial for our Japanese learners since all of these positive trends can be observed in our data to some degree. However, the picture is of course much more complex than this, as outlined in the proceeding discussion.

SA effects on appropriateness
If we first consider SA effects on the appropriateness of requests made, our findings broadly indicate positive SA effects. This is shown in the statistically significant changes in evaluations of the requests between the start (T1; mean average score 15.5) and end (T3; mean average score 19.75) of the tenmonth SA period. In other words, the raters were more satisfied with the levels of directness, politeness and formality produced in the requests towards the end of the SA stay than at the beginning. That only smaller, non-significant increases in appropriateness were identified in the interim test periods (T1-T2 and T2-T3) suggests the learners needed a longer time to develop appropriate pragmatic output across the range of on-campus encounters. This corroborates previous pragmatics research indicating longer SA periods of nine months or more facilitate greater pragmatic growth than shorter stays (Félix-Brasdefer 2004;Matsumura 2003Matsumura , 2007Schauer 2009).
Examining the data at the imposition level allowed a more nuanced understanding of the pragmatic growth exhibited within the group. Considering situation type, significant gains between the start and end SA period were only noted in the less pragmatically-demanding, low imposition requests. In the high imposition requests, which demand more linguistic and pragmatic effort to get right, only moderate, non-significant gains were evident. Perhaps as expected, this suggests the intermediate learners were slower at developing the requisite skills to produce these upward (+P), non-congruent (+R) requests. This is because conceptualising, formulating and articulating (Levelt 1989) requests at the highest levels of politeness, formality and linguistic complexity, tests cognitive fluency (Segalowitz 2010) to the max, meaning appropriate execution is more challenging.
What is of interest is that the L2 environment seems to have had some impact for the low imposition requests but minimal impact on developing the skills needed for higher-level requests. This is despite SA affording the learners opportunities to interact with all the interlocutors (tutor, librarian and security guard) present in the CAPT role plays. Taguchi (2011) also reported no significant effect of SA on high-imposition request and opinion speech acts but instead found proficiency to be a deciding factor. In line with previous research documenting only slow or modest pragmatic development as a result of SA abroad (Barron 2003;Schauer 2006Schauer , 2009, the author suggested a degree of 'linguistic sophistication is a prerequisite for a successful high-imposition speech act ' (2011: 285) which may explain the challenges our intermediate learners faced. In a related qualitative study examining the linguistic features of the same request data produced by this participant group, (Economidou-Kogetsidis and Halenko 2022) found the learners often drew on their pre-established L1 systems, even by the end of the SA stay, regardless of high or low request type. Learners typically overused and relied on more direct request strategies, included little internal modification and instead incorporated non-L2-like external modification to express politeness.
A final point to note is that pragmatic growth is subject to much individual variation within the group. The Standard Deviations, illustrating distance from the mean average group score, indicate participants appropriacy scores increase at difference rates across time points (T1 = 2.33; T2 = 3.42; T3 = 4.46). In other words, some, but not all, participants benefit from the SA experience and do so with varying degrees of success. This finding resonates with Woodfield (2012) and Schauer (2006Schauer ( , 2009 who also reported mixed successes as a result of SA. For instance, Woodfield's (2012) SA study with eight graduate students in a British university over a period of eight months, similarly reported evidence of important individual variation (2012,41), despite the fact that all her learners acquired new forms of internal modification during their SA. Retrospective interviews with the learners revealed possible effects of formal instruction, and intensity of interaction in the TL community (2012: 41).

SA effects on planning time
In this study, planning time refers to the offline planning moments at the pre-task stage. The data show that overall planning time only marginally decreased between the start (T1; 2.13 seconds) and end (T3; 1.33 seconds) of the SA stay. Whilst this shows learners used less time to plan their responses as their SA stay unfolded, this reduction was not found to be significant. Echoing findings in Li (2014) and Taguchi (2011), learners continued to take advantage of the unrestricted time given to consider and plan their responses at each time point.
With regards to situation type, unsurprisingly, there were no SA effects to be reported here either. High imposition requests showed a mixed picture of T1-T2 decreases, T2-T3 increases, but an overall T1-T3 marginal decrease. This chimes with the earlier finding on appropriateness that showed no-tominimal SA effects with high imposition requests. Low imposition requests evidenced a more obvious decrease over time, indicating these were less challenging to produce, but decreases were not significant between any test points. Different to appropriateness, SA and non-SA studies have found no proficiency effect on planning time across different speech acts and target languages (Li 2014;Taguchi 2007Taguchi , 2011. It may be that our intermediate learners used the planning period for different purposes, rather than language level being a main factor. This line of questioning continues in the next section. Although evidence suggests that pre-task planning is an important stage for learners to activate and retrieve the appropriate linguistic knowledge to convey their communicative intent (Ortega 2005;Tavakoli and Wright 2020), what our quantitative data cannot reveal is how learners used these unlimited planning moments and were these opportunities used in the same way across each time period? Taguchi (2011) speculated that her learners may have maximised their planning time to consider the sociopragmatic aspects of their language choices which demand careful consideration of speaker relationship, and the degree of formality and imposition, for instance. Li (2014) and Taguchi (2007) also suggest affective variables (e.g. anxiety, tension) and personality factors (e.g. willingness to communicate) may play a role. All of these are plausible explanations for the consistent amounts of planning time used by our participant group across the entire SA period, and are pending further research.
Where our study's findings are at odds with existing literature is the proposal that planning time often leads to greater linguistic complexity i.e. the 'ambition' of the language produced (Skehan 1996) (e.g. Yuan andEllis 2003) and higher quality output more generally (Tavakoli and Wright 2020;Yuan and Ellis 2003). The exception in this study's findings specifically relates to the high imposition scenarios which did not evidence greater complexity in this study, nor did the requests improve in terms of their linguistic appropriateness during the SA period. Both of these variables are linked. Returning to the qualitative evidence in Economidou-Kogetsidis and Halenko (2022) reporting a reliance on and overuse of basic request sequences regardless of situation type, the non-significant changes in appropriateness found in the high imposition requests could be traced back to this lack of linguistic ambition or elaborateness (Skehan 1996). This finding might also be considered in light of Yuan and Ellis' (2003) suggestion of 'trade-off effects' between complexity, accuracy and fluency at the planning stage where learners must prioritise what they attend to because of limited processing capacity, even though, in this study, no time pressure was imposed for either the pre-task planning or on-task (online) planning stages which may actually offer additional processing capacity.

SA effects on speech rate
Turning to the second measure of processing efficiency, speech rate data measured the fluency of the request responses over the SA period. In practice, L2 fluency is 'the degree to which speech flows, and to what extent that flow is interrupted by pauses, hesitations, false starts, and so on' (Derwing 2017: 246). In this study, the speech rate data pointed to an increase in the mean average words produced per minute across the three time periods and these were found to be significant but only with the high imposition requests. Further analysis showed the request responses contained fewer markers of dysfluent speech such as pauses, hesitations, repetitions or reformulations. It is possible online planning, described as the careful production and monitoring of speech during performance (Yuan and Ellis 2003), became easier over time as a result of repeated use or practice, or the learners grew in confidence. A more likely explanation, considering the (in)appropriateness of the requests discussed earlier, is that the speed of access and control in mobilising the preferred basic request formulations, which evidence little change over time, becomes stored as procedural knowledge, leading to increased automaticity (Kormos 2006).
Whether the learners are aware of the appropriateness of their requests or not is unclear, but these simple request formulas, appropriate or not, become an established part of their interlanguage over time. This is most evident in the high imposition request scenarios which show a statistically significant increase in speech rate between the beginning and end of the SA period. Learners make fewer attempts to monitor their output and differentiate between high/low situation types unlike in the earlier SA periods. This is despite the cognitive demands of online planning being reduced somewhat in this study, given the CAPT only requires a single turn response to the animated stimulus. By the end of the SA stay, request production is smoother and more fluid (characteristics of more fluent speech), there are fewer instances of learners monitoring their output through reformulation, for example, or buying additional time through pauses or hesitations. At the surface level, learners appear more efficient in managing linguistic resources, but, as discussed earlier, this is at the expense of producing output appropriate to situation type. With regards to the low imposition scenarios, small increases in speech rate are also evident here but these are not significant. Similar to Taguchi's (2011) findings, learners appear more adept at producing low imposition requests across each time point, whilst simultaneously managing the cognitive demands of online processing. It is likely that students have encountered low level requests more frequently and so repetition and practice have improved performance.

Conclusion
The ten-month SA period in the UK effectuated change in request production and processing in different ways in each of the three variables examined in this study (appropriateness of pragmatic production, planning time and speech rate). Overall, however, SA seems to have had minimal positive impact suggesting activation and processing of pragmatic knowledge needs enhancement from other sources.
From a linguistic perspective, the data suggest SA had the most positive effect on appropriateness of request responses though this is more apparent with low level requests and much less so with high level requests. Learners were able to mobilise and improve their linguistic resources for low imposition requests but consistently struggled with upward, non-congruent requests. Another study would be needed to ascertain the cause(s) of the struggles e.g. conscious moves towards risk-averse behaviour, if learners were simply unaware of the quality of their output, or learners' pragmatic attention was not primed to notice pragmatic differences due to limited social contact or encounters with +P/+R situations.
From a processing perspective, there are possible signs that immersion may help with processing efficiency, based on the data in this study, but this does not necessarily lead to better quality output. Where task demands are greater and increase the cognitive load, such as producing high level requests, learners need additional scaffolding to support pragmatic growth, which is not available from the immersion context alone. Due to the sensitive nature of conveying pragmatic intentions, learners may be disadvantaged by the lack of explicit correction, feedback or modelling which limits growth in certain areas of language production. Directing SA learners to notice pragmatic differences through classroom intervention or tasks involving more out-of-class interaction with the L2 context may provide the much-needed triggers to stimulate pragmatic growth.
This study adds to the evidence of selective SA effects on pragmatic growth (Barron 2003;Schauer 2009;Taguchi 2011) but is not without its limitations. Besides the small sample size, the study would have benefitted from qualitative learner testimonials at each time point to help explain underlying factors behind some of the quantitative findings. Since the data also evidenced individual variation within all three variables examined, we can only draw tentative conclusions at the group level. More detailed accounts of individual performances, as well as the nature and extent of dysfluent speech, which is often overlooked in pragmatics research, are pending further inquiry. Still, this study contributes to a more comprehensive understanding of how SA affects L2 request development and how pragmatic knowledge may be mobilised in communication.