An intensive look at intensity and language learning

The practical implications of the findings for the organization of instructional time for second language learning, as well as directionsfor future research in which variables such as ag

An Intensive Look at Intensity and

distribu-a simildistribu-ar intensive English distribu-as distribu-a second ldistribu-angudistribu-age progrdistribu-am were followedthroughout their intensive experience In one program, the 400 hours

of instruction were concentrated in a 5-month block; in the other, the

400 hours were experienced in a series of intensive exposures across thefull 10-month academic year Language development was comparedacross the two contexts four times via a battery of comprehension andproduction measures Overall, the findings showed substantial progressover time for both groups, with no clear learning advantage for eitherconcentrating or distributing the intensive experience These resultsare consistent with research comparing the effects of massed anddistributed conditions on the learning of complex skills in otherdomains The practical implications of the findings for the organization

of instructional time for second language learning, as well as directionsfor future research in which variables such as age, proficiency, andlearning targets are manipulated, are discussed

doi: 10.5054/tq.2011.240858

T his article reports findings from a longitudinal study of the effects

of different distributions of instructional time on the learning ofEnglish as a second language (ESL) The research question thatmotivated the study was how varying the degree of intensity ofinstruction—concentrated in an uninterrupted learning experienceversus distributed across a series of successive intensive experiences—would affect different aspects of language proficiency over time Thefindings have implications for theoretical accounts of the optimal

distribution of time for language learning and for the learning ofcomplex skills in general (the spaced–distributed practice pheno-menon) In addition, an assessment of the relative effectiveness of thetwo types of intensive programs may inform decisions made by educatorswho weigh a range of practical issues when implementing intensiveinstruction in the various contexts in which such courses may be offered.

THE DISTRIBUTION OF TIME IN LANGUAGE LEARNING

Within the language learning literature, it has long been recognizedthat a few hours a week of exposure to a new language, even if continuedfor several years, does not allow students to attain very high levels ofproficiency This is especially true when additional access to the targetlanguage is not readily available outside the classroom Benseler andSchulz (1979), for example, noted that, in an early issue of the ModernLanguage Journal, Hills (1919) had lamented the level of oral proficiency

in modern languages of young Americans at that time and included theintensifying of second language (L2) instruction at the postsecondarylevel among his proposed reforms Several decades later, Stern (1985)reported on similar concerns within the European context Hedescribed an initiative for improving secondary school students’ foreignlanguage skills, which involved concentrating the time available forinstruction into ‘‘compact’’ courses There is also empirical evidencedemonstrating that the distribution of small amounts of instruction oversuccessive school years leaves students with limited communicativeabilities in the L2 (e.g., Donato, Tucker, Wudthayagorn, & Igarashi,2000; Netten & Germain, 2004; Spada & Lightbown, 1989)

Although many learners of all ages continue to receive their languageinstruction under limited exposure conditions, intensive courses areincreasingly common across a range of contexts, institutions, andlanguages In general, a course or program is deemed intensive whenthe hours available for instruction are concentrated into blocks of time,giving students exposure to the L2 for several hours a day The length ofthe intensive experience varies widely, however When offered asprofessional development by an employer, which may require releasingemployees from their regular responsibilities, the total amount of timemay be quite short (e.g., 25 hours concentrated into a week-long course

to enable receptionists in a French-speaking region of Quebec to takebasic phone messages in English) At the other extreme, intensiveprograms may continue over several months as participants strive toachieve sufficient proficiency to pursue postsecondary education in theirL2 (e.g., the multilevel intensive ESL programs offered at English-medium universities around the world), to meet the language

requirements of a particular job (e.g., the months of language training

in French and English provided to Canadian federal civil servants) or tobecome citizens of an adopted country (e.g., the years of languagetraining and civics programs provided to newcomers in countries such asSweden) Among these prolonged programs are the well-documentedCanadian English and French intensive courses for children at the end

of elementary school, which typically last 5 months (Collins, Halter,Lightbown, & Spada, 1999; Lightbown & Spada, 1994; Netten &Germain, 2004)

There is evidence that an intensive language learning experience canlead to substantial progress in an L2 in a relatively short amount of time,for both children (Collins et al., 1999; Germain, Netten, & Movassat,2004; Lightbown & Spada, 1994; White & Turner, 2005) and adults(Freed, Segalowitz, & Dewey, 2004; Serrano, 2007) There is also someevidence that the benefits of an intensive course may persist over time,although this aspect has received less research attention (Dussault, 1997;Lightbown & Spada, 1991) What is less clearly understood is whetherone continues to extract maximum benefit from an intensive experiencethat persists over several months, or whether distributing the total timeavailable for instruction across a series of intensive experiences may yieldsimilar or possibly even superior results In other words, how extensiveshould an intensive language experience be? Are there advantages tospacing intensive exposure?

The issue of distributing practice has received considerable attention

in the cognitive and educational psychology literature, in which theeffects of concentrating exposure in a single, or massed, learningexperience is contrasted with two or more spaced exposures to thelearning target, with total time on task held constant The evidence insupport of the advantages to learning of spacing practice (summarized

in Dempster, 1996) is so robust and well-established that a studyinvestigating the effects of distributed practice in teaching physics began

by justifying why there was even a need for such research: ‘‘with theoverwhelming preponderance of evidence supporting distributedpractice, it is natural to question the value of yet another study on thistopic’’ (Grote, 1995 p 97) If these findings are relevant for languagelearning, they would suggest an advantage for spacing intensive practice.However, there are two aspects of the research done in this paradigmthat need to be considered when discussing the implications forlanguage learning: time and learning targets In these studies, a practicesession typically lasts a matter of minutes (even in the concentratedcondition), and the total time devoted to learning may be spread overjust a few hours or days (Willingham, 2002) This is far less exposurethan even the shortest of intensive experiences, and in fact, evennonintensive language programs provide longer and more sustained

exposure In addition, the convincing findings for the spacing effecthave been observed for discrete learning targets, such as nonsensesyllables, mathematical operations, lists of uncommon or specializedvocabulary, as well as for a range of motor skills, such as typing, ball toss,upside-down printing, and gymnastics moves (Donovan & Radosevich,1999) It is not clear whether these findings would be relevant for theacquisition of a more complex skill like effective communication in anL2, especially in light of the findings from Donovan and Radosevich’smeta-analysis of the effect sizes of 63 studies of the spacing effect Theyfound much less difference between the two types of practice when thetask involved a number of distinct behaviours and choices, and a degree

of uncertainty, such as air traffic controller simulations and musicperformance Furthermore, in the relatively few experiments in whichthe targets were a closer approximation to L2 learning conditions, such

as the retention of ideas rendered in spoken or written form in firstlanguage (L1; sentences, paragraphs, or lectures), the spacing effect wasreduced or eliminated when the exposure involved paraphrase(Dellarosa & Bourne, 1985; Glover & Corkill, 1987) or changes to thelearning context (Verkoeijen, Rikers, & Schmidt, 2004), including thesame ideas uttered by different speakers (Dellarosa & Bourne, 1985).The language program evaluation literature does not provideconclusive evidence in favour of a more spaced or distributed intensiveexperience, either In L2 classroom-based research investigating theeffects of the distribution of time on language learning outcomes, whentotal time for instruction is held constant, the findings show moreadvantages for concentrating exposure (whether in the form of half days

or full days of instruction), when comparisons are made with limitedexposure of a few hours a week (Lapkin, Hart, & Harley, 1998; Serrano &Mun˜ oz, 2007; Spada & Lightbown, 1989; White & Turner, 2005).1In thecurrent study, however, given the well-documented shortcomings oflimited exposure, we were interested in examining a different distribu-tion of more substantially concentrated amounts of instruction Thisissue has received less research attention Collins et al (1999) comparedtwo versions of a communicatively oriented intensive ESL2program forbeginner-level3 11- to 12-year-old francophone students in French-speaking regions of Quebec: one in which students’ ESL instruction wasmassed into full days over 5 months; and a more distributed version in

1 Serrano and Mun ˜ oz (2007) found no significant differences on the final outcomes among the adult English as a foreign language (EFL) learners in three distributions of 110 hours

of instruction: intensive, semi-intensive, and extensive However, post-hoc comparisons of gains scores showed differences in favour of the two intensive groups.

2 We refer to these as ESL programs, although they resemble EFL contexts, in that learners typically have little or no exposure to English outside the classroom.

3 The students are not absolute beginners, but have very limited proficiency in English at the outset of their intensive experience.

which students spent half days in ESL over the full 10 months of theschool year Although both groups made substantial progress on thevarious measures of written production and aural comprehension used(oral production was not assessed), there were small but significantdifferences in favour of the massed condition However, there was a timeconfound in this study Because of the complexities of the cyclicaltimetable used in the 10-month model of intensive ESL, students ended

up with fewer total hours of instruction than their counterparts in the month model, which may have influenced the direction of the findings

5-In addition, this study, like the intensive or limited-exposure comparisonstudies cited earlier, looked at postprogram learning outcomes only, notdevelopment over time Thus we do not know whether differentdistributions of intensity affect all aspects of proficiency (listeningcomprehension, vocabulary growth, grammatical accuracy, etc.) insimilar ways and at similar rates Furthermore, Collins et al did notinvestigate the impact of variations in intensity on oral skills Given thehigh priority assigned to this aspect of language learning in manyintensive language programs around the world (including intensive ESL

in Quebec), it is important to understand how the distribution ofinstructional time may affect communicative competence

In summary, although the cognitive and educational psychologyliterature points to advantages for spaced or distributed learningconditions, there are limitations (boundary conditions) on the effectwhen the to-be-learned targets are more closely related to the types ofskills typically associated with L2 learning Furthermore, although thelanguage program literature has shown some advantages for concentrat-ing L2 exposure, the evidence is not conclusive, because of the role ofintervening variables, the reliance on postprogram comparisons, and thepaucity of data on oral production skills

The current study was designed to investigate these issues in anaturally occurring context in Quebec, in which school boards havebeen experimenting with different distributions of intensive ESL forfrancophone children at the Grade 6 level (11–12 year olds) In the twoconditions chosen for this study, the children received the same totalnumber of hours of instruction, but the time was distributed differently

In the concentrated condition, the students were exposed to 4-5 hours

of English every day in a single 5-month intensive experience In thedistributed condition, students received a series of mini-intensivesthroughout the full 10 months of the school year; that is, 4-5 hours ofEnglish per day in blocks of 4-, 5-, and 9-day exposures We investigatedhow varying the degree of intensity (concentrated versus distributed)affected the development of different aspects of L2 proficiency over time

in this learner population

Context and Participants

The 230 Grade 6 (aged 11–12 years) francophone children whoparticipated in the study began learning ESL in Grade 3 (aged 8 years).They had spent 90–100 hours, spread out over 3 years, in a regular,limited-exposure ESL program During the period covered by this study,they were in one of two intensive ESL programs that afforded themsubstantially increased exposure to English, but which differed withrespect to how the instructional time was distributed There were fiveintact classes of 5-month concentrated intensive ESL (n 5 137), allhoused in the same school, and four intact classes of 10-monthdistributed intensive ESL (n 5107), located in two different schools inthe same neighbourhood The schools were situated in two towns, eachabout 1 hour outside of Montreal There were a number of similaritiesbetween the concentrated and distributed instructional contexts First,English could be considered a foreign, rather than a second, languagebecause there were few opportunities for exposure to English outside ofthe classroom Second, the total number of hours of ESL instruction wasthe same, approximately 400 hours Third, teachers followed a theme-based approach to lesson planning and focused on the development ofspeaking and listening skills and the expansion of the students’vocabulary A fourth similarity is that the regular French academiccurriculum was concentrated into half of the time normally allocated to

it That is, French mother tongue, math, science, and social studies werealso taught intensively, with additional hours of homework required tocomplete the grade-level objectives Finally, participation in intensiveESL in both intensive models was open to students from a range ofacademic abilities

The main difference between the contexts was the distribution ofinstructional time In the concentrated model, students had full days ofEnglish every day for 5 months, from late January through June Theyhad already completed their French academic program during the firsthalf of the year In the distributed model, students had blocks of fulldays of English alternating with blocks of full days of French for theentire 10 months of the academic year These distributed mini-intensiveswere in one of the following two cycles, such that, over a period of

18 days, students had 9 full days of English: 4 days of English, 5 days ofFrench; 5 days of English, 4 days of French; or 9 days of English, 9 days ofFrench

The classes were taught by seven trained, experienced ESL teacherswho were all proficient speakers of English They all had their ownclassrooms, which they were free to decorate as they wished Although

each ESL teacher in the concentrated model taught one class over the month period from late January to June, each ESL teacher in thedistributed model taught two groups over 10 months, alternating withthe French teacher and following the pattern just described Thus therewere five teachers for the five concentrated groups and two teachers forthe four distributed groups.

5-Procedure

All students were pretested at the beginning of their intensive ESLprogram to establish that they had similar knowledge of English Theywere tested four more times at 100-hour intervals during their respectiveprograms: after 100, 200, 300, and 400 hours of instruction (henceforthTime 1, Time 2, Time 3, and Time 4) Language proficiency testsincluded measures of oral and written production and of aural andwritten comprehension In addition, although we had carefully selectedthe participating teachers and groups so that they were as similar aspossible with respect to the points mentioned above (including students’level of English, total instructional time, curriculum, and languageteaching approach) and had included several groups from eachcondition to mitigate teacher effects, we know as experienced languageteachers ourselves that teachers and classes can differ from each other inways that cannot all be controlled for in classroom-based research Toidentify any instructional practices that might distinguish the groups or

be important in interpreting the findings, four classes, two from the month program and two from the 10-month program, were observedfour times, for the entire school day, during each of the four testingweeks In addition, teacher and student questionnaires were adminis-tered to all participants during the final testing session.4 The datacollection schedule is presented in Table 1 The paper and pencilmeasures were administered to whole classes and took 45–60 minutes tocomplete Times for oral measures are explained below

5-Instruments

All tasks used in the study were either original or adapted versions oftasks that had been used in previous studies with students of the sameage and proficiency level

4 The teacher questionnaires also provided biographical data on the teachers’ experience and education (reported earlier), whereas the student questionnaires yielded information

on language use and language attitudes, which will be relevant in follow up studies of individual performance.

The three pretests were an aural vocabulary recognition (AVR) test, adictation test, and a cloze test The AVR required students to match 80words spoken on a tape to pictures on a series of pages in a test booklet.Scores were based on the total number of words correctly identified Nodata from students scoring above 75% on the AVR were retained for thestudy The cloze passage contained 10 blanks in a 54-word text about aschool routine Students were given credit for any word that made sense

in the context (for sample items, see Collins et al., 1999) The dictationwas a 50-word text about a vacation, which had been used previously inthe Barcelona Age Factor Project (Mun˜ oz, 2006) Students were givenone point for each correct word

Longitudinal Tests

Vocabulary knowledge A yes–no vocabulary recognition test, adaptedfrom Meara (1992), evaluated familiarity with the 1,000 most frequentwords of English It consisted of a checklist that contained 120 real wordsand 60 nonsense words A different version of the test was administered

at each of the four testing times

Narrative writing A picture-prompted written narrative task was used

at each testing session (adapted from Collins et al, 1999; see alsoLightbown, Halter, White, & Horst, 2002) Four different pictures werechosen to match learners’ developing vocabulary, based on ourknowledge of the classroom themes typically used during the intensiveprograms The first two pictures involved animals, family members, bodyparts, and occupations, while the last two allowed students to imagine arange of activities, relationships, and outcomes The prompt in each casewas Imagine what is happening now, what happened before, and what is going tohappen next Learners were given 15–20 minutes to write and wereencouraged to use their imagination and to provide as many details asthey could No dictionaries were permitted, but to encourage students towrite as much as possible, they were told that they could use a French

TABLE 1

Data Collection Schedule

Pretest Time 1 Time 2 Time 3 Time 4 Paper and pencil

Y/N Vocabulary

Y/N Vocabulary Cloze

Narrative Narrative Narrative Narrative Oral measures

(N 5 108)

Information gap

Role-play Information

gap

Role-play Note AVR 5 aural vocabulary recognition; ASC 5 aural sentence comprehension; MEQ 5 Ministry of Education of Quebec; Y/N 5 yes or no.

word if they got stuck There were two measures for this task: fluency(number of words in the text) and grammatical knowledge (use of verbinflections).

Listening skills.The students’ listening skills improve dramatically inthese intensive programs, creating the problem of potential ceilingeffects on the readministration of tasks that had been appropriate in theinitial testing times Consequently, we used a different measure at eachtesting time At Time 1, we readministered the AVR, which measured theability to match individual words to pictures At Time 2, a 20-item auralsentence comprehension task required learners to match a sentence toone of three pictures (Mun˜ oz, 2006) At Time 3, we readministered thedictation from the pretest At Time 4, a 32-item general listeningcomprehension test required the interpretation of short utterances Thistest was developed in the 1980s by the Quebec Ministry of Education forGrade 9 ESL students and has been used frequently in Grade 6 intensiveESL research in Quebec (see Collins et al., 1999 for a sample item).Oral interaction To measure communicative effectiveness, two timed,paired oral interaction measures were used with a stratified subset of 12students from each class (n 5 108), selected on the basis of pretestperformance and in consultation with the teachers Students werepaired with a different partner from their class for each of the fourtesting times (information gap at Times 1 and 3; role play at Times 2 and4), and each partner received the same score (see White & Turner, 2005,regarding co-constructed oral scores).5

The information gap task was presented as a game similar to thoseused in intensive classes Each student in the pair described five itemsthat were missing in his or her partner’s picture (a school at Time 1, ahouse at Time 3), so that the student’s partner could draw the objects inthe correct place A screen prevented students from seeing each other’spicture and forced them to rely on words, rather than gestures, to locateobjects in the different rooms They had five minutes to complete thetask The task was audio-recorded, and the drawings were collected.Following the scoring procedure developed by White and Turner(2005), three points were allocated for each item: one point if thestudent drew the correct object, one point if it was in the correct room,and one point if the object was in the correct location The total possiblescore was 30 (10 items 6 3 points) One point was subtracted if L1-French was used to describe the object, room, or location

The role-play task was administered at Times 2 and 4 It was modeled

on one developed for the Barcelona Age Factor Project (Mun˜ oz, 2006).There were two versions of the task, which was completed in pairs,

5 Because of time constraints, only one oral task was administered at each testing session The information gap task was more appropriate than the role play task at Time 1, because

it did not require extensive discourse.

recorded, and transcribed In each, one learner playing a childnegotiated with another learner playing a parent over having a party(Time 2) or getting a pet (Time 4) Learners were required to assumetheir roles without the intervention of the task supervisor, and they wereexpected to stay on topic and respond to each other in English, todiscuss different aspects of the situation, to elaborate and justify theirpoints of view They were given 2 minutes to negotiate a solution.Because there were no established guidelines for scoring a role-play taskcarried out by two students, we developed a five-level global ratingscheme for this task in collaboration with our Barcelona colleagues SeeTable A1 in Appendix A for a description of the rating scale.

The information gap and role-play tasks are typical of the ones used inintensive classes, where the primary focus is on the development of oralinteraction skills Students often engage in pair or small group problem-solving activities and skits, working cooperatively with differentclassmates of varying linguistic abilities, and making the most ofwhatever language they have to accomplish the task The tasks thushave high ecological validity They also have construct validity, in thatlearners demonstrate the type of competence targeted by the intensiveprogram, namely oral communicative effectiveness in terms of successfulmutual comprehension rather than grammatical accuracy

Classroom observations

Members of the research team spent a full day observing andvideotaping in the classrooms of two teachers in each program the sameweek that longitudinal data were collected, that is, Times 1, 2, 3, and 4.The observations were later coded following an adapted version of theCommunicative Orientation of Language Teaching (COLT) scheme(Spada & Fro¨ hlich, 1995)

Questionnaires

At Time 4, questionnaires were administered to students andteachers The student questionnaire contained 12 items asking studentstheir attitudes toward English and the intensive experience, what theycould and could not do in English, and their opportunities for exposure

to English outside the classroom The 19 items of the teacherquestionnaire asked teachers for information about the students intheir class, the ESL materials and activities they used, their focus on skillsand language features, their teaching experience and training, and theirown language learning experiences

ANALYSES AND FINDINGS

This section is divided into three parts The first reports results fromthe pretests The second reports results from the longitudinal tests ofdifferent aspects of learning (vocabulary knowledge, listening skills, oralinteraction, and narrative writing skills) In the final part, relevantinformation about the instructional contexts yielded from the classroomobservations and teacher questionnaires is summarized

Pretests

Independent t-tests were run on each of the three pretest measures,the AVR, cloze, and dictation To maximize the probability of detectingdifferences between groups, the alpha level was set at 0.0.05 Theanalyses reveal no significant differences between the program groups(concentrated and distributed) on any of the three pretest measures: theAVR, t (236) 5 3.69, p 5 0.712; the cloze, t (236) 5 2.46, p 5 0.806; orthe dictation, t (236) 5 1.710, p 5 0.089 Thus students in the twogroups started their intensive experience with similarly limited knowl-edge of English

Longitudinal Measures

The findings for the measures of different aspects of L2 knowledgeand performance over time are reported by proficiency area A total offive analysis of variance (ANOVA) tests and four t-tests were carried out

To avoid a type 1 error inherent in multiple comparisons, the alpha levelfor all main effects and interactions was set at 0.05 and adjusted for thenine statistical tests to 0.005 For any post-hoc pairwise comparisons,alpha was set at 0.05 The Bonferroni adjustment for the number ofpairwise comparisons was applied to the significance levels of thecomparisons Partial eta squared (gp2) was used to estimate the effectsize.6

Vocabulary Recognition (Yes/No)

A mixed between-within ANOVA revealed a significant effect for time,F(3, 642) 5 126.350, p , 0.005, and a significant interaction between

6 A reviewer asked whether multiple ANOVA (MANOVA) would be a more appropriate procedure Using MANOVA would indeed protect from the possibility of making a type 1 error inherent in using separate ANOVA tests or t-tests for each dependent variable However, because the tests that make up the dependent variables are not all administered

at the same time or in the same number of repetitions (for reasons described earlier), it is impossible to enter all of them into one MANOVA To avoid making a type 1 error, the alpha was adjusted for the number of analyses carried out.

time and group F(3, 642) 5 5.311, p , 0.005 The effect size for time wasmoderate (gp2 5 0.371) and very small for the time–group interaction(gp 5 0.024) As can be seen in Figure 1, the performance over time ofthe two groups was very similar; the only significant between-groupdifference occurred at Time 4 (p , 0.05) and was very small, both interms of actual score difference (5%) and effect size (gp 5 0.029).There was also little difference in rate of learning within groups acrosstime With the exception of the Time 1 and Time 2 comparison in theconcentrated intensive group and Time 2 and Time 3 in the distributedgroup, all other time comparisons were significant for both groups (seeTables 3 and 4 in Appendix B for a complete summary of the within-group comparisons) Overall, the recognition of common English wordssteadily improved in both groups, such that, by the end of the intensiveexperience, it can be estimated that students were familiar with roughly75% of the 1,000 most frequent words of the English language.

Listening Skills

As explained earlier, listening skills were measured by a different task

at each of the four times They were thus analyzed through four separatet-tests The findings for each of the tasks are illustrated in Figures 2–5 AtTime 1 there was no significant difference between the groups on theAVR, t (235) 5 1.473, p 5 0.142 At Time 2, there was also no significantdifference between the groups on the aural sentence comprehensiontask, t(232) 5 2.800, p 5 0.006 However, at Time 3, there was a small butsignificant difference between the groups on the dictation task in favour

of the concentrated group, t (233) 5 3.056, p , 0.005 At Time 4, on theMEQ test of general listening comprehension, the significant differencewas again in favour of the concentrated group, t (232) 5 7.599, p ,

FIGURE 1 Vocabulary recognition (*p , 0.05).

0.005 In summary, small but significant differences in listening skills infavour of the concentrated intensity group appeared midway throughthe program (at Time 3) and persisted through to the end of theprogram (at Time 4).

Oral Interaction

As explained in the Methodology section, a subset of 108 students wasfollowed for the paired tasks over the four testing times Interraterreliability for both oral interaction tasks was high: 95% agreement forinformation gap and 90% for role-play

Information Gap

The findings for the information gap task are displayed in Figure 6 Amixed between-within ANOVA revealed a significant effect for time,F(1,52) 5 76.551, p , 0.005 with a moderate effect size (gp 5 0.595).There was no significant difference for group, F(1, 52) 5 7.537, p 5 0.008,and no interaction between time and group, F(1, 52) 5 0.639, p 5 0.428 Thus both groups improved significantly over time and at similar rates

Role Play

The role-play findings are displayed in Figure 7 A mixed within ANOVA revealed a significant effect for time only, F(1, 47)7 551.951, p , 0.005, with a moderate effect size (gp25 0.525) There was

between-no significant difference for group, F(1, 47) 5 0.118, p 5 0.773, and between-nointeraction between time and group, F(1, 47) 5 3.252, p 5 0.078 Thusboth groups showed similar performance on the role-play at Times 2 and

FIGURE 2 Listening comprehension: aural vocabulary recognition task.

7 The number of pairs for the role-play is slightly lower than for the information gap, because an equipment malfunction resulted in the loss of some of the role play data.

4, and both experienced a significant and comparable improvement onthis task over time.

Narrative Skills (Writing)

Each student produced four written narratives, one at each of thetesting times Mixed between-within ANOVAs were performed on boththe length of the narratives in words (a fluency measure) and the use ofverbal morphology on a four-point scale (a grammatical knowledgemeasure) The findings are displayed in Figures 8 and 9

Length of narratives

There was a main effect for time F(3, 630) 5 78.755, p , 0.005, andgroup F(1, 210) 5 10.716, p , 0.005 There was also an interactionbetween time and group, F(3, 630) 5 24.390, p , 0.005, with a moderateeffect size for the time difference (gp2 5 0.467) and a small effect sizefor the interaction (gp2 5 0.187) The between-group pairwise

FIGURE 3 Listening comprehension: aural sentence comprehension task.

FIGURE 4 Listening comprehension: dictation task (*p , 0.005).