Vol. 2. No. 2 A-1 September 1996

Return to Table of Contents Return to Main Page

Analytic and Systemic Analyses of Computer-Supported Language Learning Environments

Joy L. Egbert
Center for English Language Training
Indiana University

Leonard M. Jessup
School of Business
Indiana University

Abstract

One essential component of a theory of CALL, and one which teachers, researchers, and technology can influence, is the learning environment. Research has made it clear that the classroom environment plays an important role in learning. Unfortunately, the definition of essential environmental characteristics and the interaction among them is not clear. In this study, learner perceptions of a "package" of salient dimensions of an ideal computer-supported language learning environment were measured. Analytic and systemic analyses of learner perceptions indicated that learners perceive their learning environments in unexpected ways and that the technology has an impact on these perceptions in that it allows the classroom to be "individualized" in ways not possible without it. Also discussed are implications for task construction and grouping and the importance of learner perceptions to an understanding of computer-supported language learning environments.

Levy (1990) observed the need for a theory of computer-assisted language learning (CALL) that would, by describing the underlying principles for CALL, provide a theoretical framework which can incorporate further technological advances and insights. He noted that "our language teaching philosophy, method, or approach needs to be broadened to encompass new technologies, and the inter- relationship between language teaching and computing needs to be carefully explored" (p. 5).

One factor that has precipitated interest in a theory of CALL is the increase in the number of computers available in schools. At the same time that the number of computers in educational institutions is increasing rapidly, changes in the purposes for which they are used and changes in the curriculum in general are barely noticeable. The lack of change raises questions that remain [-1-]largely unanswered: For example, what is being done with these computers, and why? What should schools and colleges look like and do in the high technology future that they face? What theoretically sound, vital changes in curricula can and should be made? What types of systems are needed to affect these changes? A theory of CALL may assist in answering some of these questions for language educators by describing how and which students learn with different kinds of technology, identifying factors that must be addressed in the application of the technology, and serving as a guide for research on computer-assisted language learning.

One essential component of a theory of CALL, and one which teachers, researchers, and technology can influence, is the learning environment (Spolsky, 1989). Research has made it clear that the classroom environment plays an important role in learning; in addition to being a valid predictor of learning outcomes (see, for example, Anderson, 1973; Fraser, 1986), the classroom environment acts as a mediator between the macro environment and the learner. Unfortunately, the definition of essential environmental characteristics and the interaction among them is not clear.

This study takes a first step toward defining these environmental characteristics and their interrelationships. To this end, presented below is a summary of environmental conditions that have been related specifically to achievement in English language learning. This environmental framework is then applied to natural classrooms in an effort to better understand second language learning environments and the effects of technology on perceptions of the environment. The research questions of this study are:

• How do adult community college ESL learners perceive their classroom environments?
• When computer technology is added to support a drill-and-practice environment or to create a cooperative environment, how do the learners perceive these new environments?
• To what extent and how do the patterns of perceptions and the relationships between variables change from the initial to the intervention environment?

Conditions for an Ideal Language Learning Environment

Any general theory of second language acquisition (SLA) must encompass all aspects of language acquisition, including morphology, grammar, lexis, and pragmatic knowledge (Ellis, 1986); this breadth makes the scope of investigation enormous. SLA theory in general also has a multidisciplinary focus that draws on learning theory and evidence from disciplines such as linguistics, psychology, and sociology for explanations of the language learning process. Recent theories present a complex view of language acquisition as contingent upon a host of factors; it is generally accepted that language acquisition is the result of the interplay between a cognitive mechanism(s) (which Chomsky and others claim is innate) [-2-] and the learner's environment. Whether cognitive or environmental factors have more influence is contested and the definition of both is under debate (see McLaughlin, 1987, for a discussion).

Although not all learners acquire language in the same way (Klein, 1986; Richard-Amato, 1988) or for the same purposes (Spolsky, 1989), conditions learners need for acquiring language include a propensity for language learning, language faculty, and access to the target language (see Klein, 1986, for example). Spolsky (1989), in his comprehensive general theory of second language learning, states these conditions in the form of the equation "Kf = Kp + A + M + O" (Knowledge and skills in the future are a result of knowledge in the present, abilities, motivation/ affect, and opportunity). Spolsky claims that if any of these conditions is absent, no learning will take place. In addition, the speed, structure, and end state of a learner's acquisition are thought to depend on the various strengths of these conditions; in other words, the greater the amount of each, the greater the learning.

In the SLA and English as a Second Language (ESL) literatures, research repeatedly points to four conditions which, when present in the language learning environment in one form or another, support optimal language learning (the literature does not suggest , however, interactions among the conditions or their necessary strengths). Although there may be other conditions, these four are the most widely researched and supported in the literature and make up a general model of conditions for optimal language learning environments. These conditions are summarized briefly below (for a more thorough overview, see Egbert, 1993).

Condition #1: Opportunities for learners to interact and negotiate meaning with an authentic audience

Many researchers have noted that learning is essentially interaction between students and others (see, for example, Ahmad, Corbett, Rogers, and Sussex, 1985; Kelman, 1990; Levin and Boruta, 1983). If learning is such a social process, then social interaction is necessary for learning (Vygotsky, 1978); this is especially true for second language learners striving for communicative competence. "Social interaction" implies the negotiation of meaning. Other researchers have found that language learners must be involved in purposeful interaction, which includes a real audience that is actively involved with the learners (see for example Ernst, 1994; Pica, 1987; Pica, Young, and Doughty, 1987; Webb,1982, 1985). The implication, then, is that learner involvement in authentic social interaction in the target language (TL) with a knowledgeable source (e.g., the teacher, another learner, a technology, a family doctor, or other who can negotiate in the TL) facilitates language acquisition. [-3-]

Condition #2 Learners involved in authentic tasks which promote exposure to and production of varied and creative language

Authentic, in this context, means something that students know something about and are interested in that has a real purpose behind it (for examples, see Roen, 1989; Reid, 1989). Johnson (1991) explains that an authentic task must provide learners with a reason to share ideas and information, preferably within a system that allows problem solving, as studies have found that learners interact more when working on problem solving tasks than on other activities. Grammar drills, tasks common to ma ny ESL classrooms and software, are not authentic tasks in this sense.

Spolsky (1989) claims that "whatever the language learner brings to the task, whether innate ability, a language acquisition device, attitudes, previous knowledge, and experience of languages and language learning, the outcome of language learning depends in large measure on the amount and kind of exposure to the target language" (p. 166). An authentic task alone, therefore, may not be sufficient; varied and creative language implies a diversity of tasks and sources and learner use of both receptive and productive language skills that take into account the multiple learning styles and preferences among learners (Krashen & Terrell, 1983).

Condition #3 Learners have opportunities to formulate ideas and thoughts and intentional cognition is promoted

Learners need adequate time and feedback, both of which facilitate the formulation of ideas. They must also be given the opportunity to reflect on and communicate their ideas; however, because learners are presented with opportunities to formulate ideas and thoughts does not imply that they will take these opportunities or make the best of them. During the learning process, learners must be "mindful" (Salomon, 1990), that is, they must be motivated to take the opportunities presented to them and to be cognitively engaged. A certain degree of guidance, whether from peers or others, may facilitate learning and promote mindfulness (Zellermayer, Salomon, Globerson, & Givon, 1991).

Condition#4 An atmosphere with ideal stress/anxiety level in a learner-centered classroom

Before learners can be mindfully engaged and willing to communicate and share their ideas, they must experience an optimal level of anxiety in the language learning environment; this means that feelings of worry or apprehension must be facilitative, rather than debilitative (Brown, 1987; Krashen & Terrell, 1983; Lozanov, 1978). The use of computers can affect this atmosphere in different ways. Technology use can assist in the creation of an optimal stress environment by creating a learner-centered classroom, which implies that learners have some control over their learning (see Bereiter & [-4-] Scardamalia, 1990; Kremers, 1990; Robinson, 1991, and others for this discussion). Kreeft-Peyton (1990) suggests that giving more control to the learner removes the confounds of teacher, learner, and school personalities, styles, and goals.

Strategies for Implementing the Conditions

Briefly, there are many strategies by which the four conditions presented above can be operationalized in the second language classroom, including using groupwork techniques, providing concrete opportunities to interact in English, focussing on survival skills and functions, using problem-solving, providing a multitude of media, recycling content in various ways, and providing open-ended opportunities for meaningful language use. In addition, providing adequate time on task, adequate feedback, prompting, and other assistance, adequate information and learner opportunities to choose and participate support the conditions.

The four conditions, however, do not dictate specific methods, techniques, or content; rather, each of the conditions is necessary to the creation of ideal language learning environments. Further, each of these conditions has been independently related to second language learning through discrete analysis, but the "package" of conditions has not been observed interacting in natural ESL classroom environments. Because these conditions seem to have an impact on each other and because little is known about how they act/interact in actual classrooms, research is needed to assist in the understanding of language learning environments, particularly those supported by technology.

Determining how computers can be used to support the learning and teaching of ESL involves examining how the technology fits with the conditions for ideal language learning environments. Educators in both first language (L1) and second language (L2) classrooms are employing computer technology in many ways in attempting to support these conditions. There have been some successes and some relative failures (see a study by Chapelle & Jamieson, 1986). Technology can even exacerbate classroom problems (Holdstein, 1987). (For a more complete discussion of ways in which computer technology is being and could be employed in L1 and L2 classrooms, see Egbert, 1993).

The experiment described in the next section examines how computer technology fits with the environmental conditions outlined above. It asks the following questions: How do adult community college ESL learners perceive their classroom environments? When computer technology is added to support a drill-and-practice environment or to create a cooperative environment, how do the learners perceive these new environments? To what extent and how do the patterns of perceptions and the relationships between variables change from the initial to the intervention environment? [-5-]

Method

The theoretical framework of this study is based on Salomon's (1990) proposal that educational researchers need to examine classroom situations from two complementary paradigms, the systemic (patterns of changes) and the analytic (changes in patterns). In the systemic approach, the unit of analysis is the entire classroom. This approach recognizes that classrooms are dynamic, and that individual aspects are not easily (nor needfully) isolated for testing. Using a systemic approach, the researcher can evaluate the associations among factors in the classroom, describe how the factors look and how their interrelations change, and suggest why they might change in this way. In the analytic approach, the unit of measurement is one or more discrete variables that the researcher assumes can lead to conjectures about direct relationships and causality. The present study adopts both the analytic and systemic perspectives as a step toward greater understanding of classroom language learning environments supported by technology. The model for observation in this study was the list of proposed conditions for ideal language learning environments presented earlier.

Learner perceptions of the frequency of occurrence of the environmental conditions were captured via questionnaires before and immediately following computer-supported interventions. Background data and questionnaire responses were analyzed to examine the language learning environments.

Expectations

Learner perceptions change as environmental variables change; some of these changes are predictable, and others are uncertain due to the complexity of the learning environment. Because the arrangement of conditions in the two intervention environments in which participants completed the task appeared to be dissimilar according to the framework, it was expected that participants would perceive at least one of the environments as very different from the initial environment, and that they would perceive the intervention environments as dissimilar on several conditions. For example, it was expected that learners in the cooperative treatment would perceive with greater frequency than the traditional treatment the task as authentic, opportunities to interact/negotiate, and control over their learning.

Other differences in learner perceptions were expected. The cooperative group was expected to perceive task as the central condition around which interaction, control, and other variables would group tightly, with opportunities to formulate thoughts as an outlier. The traditional group was expected to perceive intentional cognition/opportunities to formulate thoughts as central, with authentic task and interaction as outlying variables. Perceptions of the traditional environment were expected to closely resemble those [-6-] of the initial environment, while perceptions of the cooperative environment were expected to be dissimilar.

Participants

Adult learners from four intermediate classes of a community college English as a Second Language program served as a convenient sample for the study. Of the 102 initial participants, 82 completed the task and both surveys. Information concerning participants' age, gender, nationality, educational background, and previous computer use was collected. The sample was composed of 59% women and 41% men. Of 17 nationalities represented, a majority was of Mexican origin (63%). Eighty-three percent of the participants had completed 12 or more years of school, and participants were almost evenly divided on age -- 53% were traditional college age and 47% older or non- traditional students.

Interventions

Participants in intervention groups worked with one of two computer applications in a computer lab; the cooperative treatment used "Culture Trivia" (McKinnon, 1989) and the traditional treatment used "Grammar Trivia". The interventions were constructed to represent different dimensions of the environmental constructs and were expected to elicit different responses to questionnaire items. The cooperative "culture" intervention required that user triads find information in any of four different external documents to help them answer questions on the screen and related questions on an external document (one per group). The assignment dealt with life in America and other relevant survival issues. The software which the traditional "grammar" group of participants used was a grammar- based application. In this application, each learner had an individual worksheet and no other sources of English or support other than the computer. This treatment is patterned on a traditional drill-and-practice learning environment.

Language Environment Survey

A high-inference measure was developed for this study using existing tools as guidelines. The four environmental conditions were divided into eight constructs:

1. opportunities for learners to interact and to negotiate meaning (INTERACTION);

2. an authentic audience (AUDIENCE);

3. authentic tasks (TASK);

4. opportunities for exposure to and production of rich and varied language (SOURCES); [-7-]

5. opportunities for learners to formulate ideas and thoughts (HELP);

6. learner intentional cognition (COGNITION);

7. an ideal-anxiety atmosphere (ATMOSPHERE), and

8. learner control (CONTROL).

Procedure

The pretest survey was conducted in the tenth week of the sixteen-week semester. Participants were randomly assigned to an intervention. Perception questionnaires were again answered by participants upon completion of the task; at this sitting, participants were asked to answer the questionnaire only in relation to the intervention which they had just completed.

Analyses

Participants' responses to the questionnaires were analyzed using both a systemic and an analytic approach. Using the analytic approach, no significant differences were found between the factor means of the four groups (classes) by treatment and class time on the initial questionnaire and the four classes were collapsed into the two treatment populations. A repeated measures multivariate analysis of variance (MANOVA) was then conducted for the eight constructs, using treatment ("cooperative" culture software v. "Traditional" grammar software) as an independent variable and gain scores from the pretest and posttest questionnaires as the dependent variable, with "time" as the within subjects factor. Gain scores were used to examine the change from pretest to posttest (see Cook & Campbell, 1979, for a detailed discussion of this technique). Age, gender, level of education, and previous computer use were used as covariates. Country of origin was not used as a covariate due to the skewness within the study population.

Using systemic analysis, construct means were used to calculate proximities, which indicate how similar or different variables are, based on correlations. Multi-dimensional Scaling (MDS) maps of the initial questionnaire responses and of the post-intervention responses were created using the similarities calculated for the two intervention groups. These maps represent the variables as points in [-8-] multidimensional space, with similar objects or variables closer together. Stresses were calculated to determine goodness of fit, and the relationships among the constructs were described. The position of a construct in the MDS maps has no direct relationship to the frequency of that construct; rather, the relationship of constructs in the maps speaks to the correlation between frequencies (in other words, as one increases, so do those that are related to it). The relationships among the variables were noted, and, when complemented with the results of the analytic approach, helped to form the basis of conclusions for this study.

Analytic Results

Pretest

                             CC                 TG
            Scale           Mean       SD      Mean       SD 
            
            Interaction     3.75      .41      3.85      .49
            Audience        3.75      .59      3.89      .51
            Task            4.22      .57      4.40      .45
            Sources         3.98      .61      4.22      .48
            Help            3.98      .57      4.23      .41
            Atmosphere      3.76      .51      3.82      .59
            Control         2.80      .73      2.88      .86
            Cognition       4.07      .58      4.30      .53

There were no significant differences in pretest scores between the treatment groups. Any inflated, yet insignificant, scores on the pretest were accounted for by the repeated measures form of MANOVA employed for analysis of gain scores.

Participant perceptions of the eight constructs can be summarized quickly to suggest an overall picture of the initial environment. In this environment, learners sometimes interacted, and, unless speaking with the teacher, interacted sometimes in a language other than English. Learners interacted relatively inoften with an authentic audience, and perceived the teacher as being a more authentic audience than other learners. The tasks in which learners participated were usually important but not necessarily [-9-] authentic. In this environment learners used all language modes and sometimes participated in different types of activities. Learners perceived that they received enough feedback but not enough time to do their work. Learners felt comfortable, but were also sometimes afraid to make mistakes or speak in front of other learners. Learner opinions were important to what happened in the class, but learners believed that they did not usually help to decide about topics or content, activities, or work configurations. Learners paid attention often.

In anecdotal accounts learners confirmed their classroom environment as one in which learners, at times, spoke languages other than English. In this environment learners participated in different activities with different sources, but some activities they considered important were neglected. Participants seemed to indicate that learners were afraid of the language, not of their teachers or peers. In addition, the teacher did not always have time to help, but other learners frequently did. In general, learners characterized their environment as satisfactory, and they suggested ways in which it could be improved.

Posttest

For the post-treatment questionnaire, participants responded based on their perceptions of the treatment environment. The mean scores for all participants indicated that participants perceived all of the factors to occur more frequently than they had in the initial classroom environment (although not all differ significantly statistically) and the frequency of all the factors was above the midpoint of the scale.

Cell means and F scores of the MANOVAs by treatment group and time are listed for each construct in Table 2. "Treat" indicates F score and significance level for differences between the treatment groups on the results of the perception questionnaire; "Time" indicates the F score and significance level for differences between the first and second administrations of the questionnaire. (Item means for the scale for each construct and discussions of such are provided in Egbert, 1993.)

      	  	  Pretest       Posttest       F
                    CC    TG      CC    TG       Treat   Time
                    
Interaction   	 3.75  3.85    4.05  4.34      5.33*   34.05***
Audience      	 3.75  3.89    4.13  4.28      1.34    36.13***
Task          	 4.22  4.40    4.45  4.61      3.39    17.65***
Sources	 	 3.98  4.22    4.17  4.40      5.65*    7.91**
Help          	 3.98  4.23    4.18  4.22      2.09     2.29
Atmosphere     	 3.76  3.82    3.79  3.86       .25      .21
Control       	 2.80  2.88    3.18  3.15       .01    13.44***
Cognition     	 4.08  4.31    4.36  4.59      7.04**  19.33***
                [ *  p < 0.05, ** = p < 0.01, *** = p < 0.001 ].[-10-]

The cognition construct was significant for treatment (p<0.01) and time (p<0.001). However, "previous computer use" covaried significantly with treatment for the cognition factor (p <.05). The source of the difference appears to lie in the distribution of previous computer use among the randomly assigned participants; on a two-point scale (1=yes, 2=no) the traditional participants had less previous computer use (20 yes, 18 no) than those in the cooperative group (30 yes, 14 no). Apparently, there is a positive relationship between cognition and previous computer use; participants who had previous computer use tended to report higher levels of cognition.

Anecdotal evidence indicates that some participants in Treatment 1 perceived the culture task to be interesting and useful. Comments from the Treatment 2 participants also indicated that these participants found their treatment interesting and helpful. Due to the few comments on the posttest, it is difficult to point out any other trends in this data. [-11-]

Systemic Results

Interpretation of MDS maps by treatment for the initial questionnaire detected no meaningful differences. Figure 1 presents the stress and r-square values along with maps of the results of the questionnaires by treatment group. The stress values of less than .15 and total variance accounted for (RSQ) indicate a good fit for the two-dimensional MDS solutions for these data.

A discussion of the relationship of the factors is provided below. Covariates were not included in the MDS analysis because of the difficulty of mixing data from two point scales with data from larger scales, even when standardized. The strength of the relationship (the correlation) between any two constructs is determined by the proximity between the constructs; the closer the distance, the stronger the relationship.

Pretest

In this initial map, Sources and Help appear to be strongly related; this implies that participants perceived as opportunities for exposure and production increased, so did the amount of time and feedback received (or vice versa). In addition, Cognition and Control are plotted in opposite quadrants, which suggests that there is little relationship perceived between the frequency with which learners pay attention and the frequency with which they exert control over the environment. On the other hand, Control and Atmosphere are related, which may indicate that the frequency of stress or nervousness participants perceived is related to the frequency with which they control their environment.

Interestingly, in the pretest map, Interaction is located in a dimension by itself, unrelated to authenticity of task or authenticity of audience as would be expected, and Task is also an outlier. This indicates that in this environment Sources, Cognition, and other variables are not related to the authenticity of the task. [-12-]

Posttest Treatment One

As in the pretest map, inner and outer circles of constructs can be found in the posttest maps. In the cooperative treatment posttest map, Audience is still a central variable; however, it has been joined by Help and Sources in the inner circle. This indicates that feedback and time, along with opportunities for exposure and production, have assumed a central position in this environment. This is a logical change, as the number of sources and opportunities for production both orally and in written form were central to this environment. Atmosphere and Control, formerly in the same quadrant, have moved to positions in opposition to each other and are located along with the remaining variables equidistant from the inner circle. If this means that the conditions in the outer circle are equally related to those on the inner, and that those on the inner are central variables, this suggests that the manipulation of Sources and Help, which are under the teacher's control, results in a change in perceptions of the environment. A judgment regarding the quality of the changes is not possible without further information about the results of the changes.

Treatment Two

The inner circle in the traditional treatment posttest map consists of Help, Task, Cognition, and Audience. The other variables remain in the outer circle as in the pretest map. Rather than being related to opportunities for exposure and production (Sources) as in the initial map, time and feedback (Help) are related more closely to the frequency with which learners pay attention (Cognition) and the authenticity of task and audience. Control in this map seems to be related to Interaction rather than Atmosphere as in the pretest map. One explanation is that, because the learners were working individually and had more opportunities to work only in English, they perceived a greater frequency of control. Additional information concerning how the participants actually worked would assist in a more accurate interpretation of this data.

Treatment One v. Treatment Two

Audience and Help are central variables in both maps, indicating that feedback and time were central to these environments. It is interesting to note the centrality and the relationship of the help and audience factors in both contexts; it was expected that in the traditional environment learners would perceive a lack of authenticity in the audience because the audience was either the computer or another learner working on his/her own task. Although interaction was not central to this environment, the audience was still perceived as an authentic source of English.

The treatments differ on the cognition factor, the map for the traditional treatment group relating it to the other central [-13-] variables, and that of the cooperative group showing it as an outlier in its own domain. This suggests that in the traditional treatment learner perceptions of the frequency with which they pay attention is related to their perceptions of task authenticity and time and feedback. These perceptions may be a function of group and task construction in that learners who are responsible for completing the task individually may pay more attention than those who have group members to assume the responsibility for them.

In the cooperative treatment, the underlying structure of the group's perceptions of the environment produced a pattern with three discernible areas consisting of social, control, and instrumental constructs. The social area includes Interaction, Atmosphere, and Audience. The control area is comprised of Control and Help, and the instrumental section of the map contains Sources, Task, and Cognition. No such logical pattern is evident in the map of the traditional treatment. Future tests with a greater number of participants may provide evidence to better explain these patterns.

The MDS plots show that in both treatments interaction remained on the outer circle. It is interesting to note that it is variables that teachers can or do generally control that are central to the cooperative environment, but that in the traditional environment cognition, a variable that teachers cannot manipulate, is one of the central factors.

Summary of Systemic Results

The most surprising finding in this analysis is that Interaction was an outlier. In addition, task was not a central variable, while authentic audience was central in the pretest and both of the posttest environments. Although some of the changes in patterns by treatment were logical (e.g., Sources becoming more central to the cooperative treatment environment), others, such as the consistent centrality of Audience, were unexpected. The continuing outlier status of Interaction and, in the cooperative treatment, Task , was also a surprising finding.

Discussion and Conclusions

First, learners did not act as expected in their treatment groups. This could be due to the grouping of the participants; instead of maximizing the time that English was spoken, forming participant groups of three gave the participants in the "cooperative" treatment the opportunity to speak more of their first language, where the individual traditional participants were restricted mainly to English on the computer screen and on the handout. Anecdotal evidence from the pretest supports the claim that learners speak in their first language in the class. Additional support for this interpretation comes from the posttest [-14-] item analysis for this scale; participants in both groups perceived themselves to almost always "talk" to the computer-teacher in English. If, as proposed, the traditional treatment participants had more chances to "speak" with the computer, they would then perceive themselves to be speaking English more often than the cooperative participants. There was no direct measure of participant in teraction to measure this variable; however, these results seem to support the findings of previous studies on group work and negotiation, such as Doughty & Pica's 1986 finding that, in order to generate more interaction, tasks and groups must be struct ured so as to require participation in English.

Because learners are grouped does not indicate that they are interacting. The learners in groups of three may have actually had less opportunities to participate or interact for several reasons. One reason participants may not have been interacting (or interacting less often than in the traditional group) was because there were three people working together with the "teacher" (computer), as opposed to dyads or solos working in the traditional group. In addition, Doughty & Pica (1986) noted that it often happens that more fluent or outgoing students may take over if each student does not have required input; this could be the case in the present study. In short, because they perceived that the task was interesting and important did not indicate either that the participants were speaking English or that they were even taking part in the activity. The conclusions of Windeatt's 1986 study (reported in Meskill, 1992) suggest that at times learners in groups around computers actually work as individuals, interacting only to suggest or reject answers because they have no responsibility to the group. This suggests that individual responsibility may promote interaction more than group responsibility where one or more members can "freeride" (see, for example , Kerr & Brunn, 1981). In order to test this hypothesis, in future research of this kind, a cooperative task should include a worksheet or other assignment that gives each participant individual responsibility.

Taken together, these explanations help us to understand interaction in the posttest environments in the treatment groups. In coming to conclusions about the value of interaction in language learning environments, further research must examine a variety of tasks and what learners actually do in groups, i.e., who speaks, how much, to whom, in what language, and what kind of utterances they are making. The next iteration of this study must look at the kinds of interaction occurring in both the initial and intervention environments.

The second important finding is that the participants perceived that, in their pretest environments, the audience was authentic only sometimes. The Audience construct was central to all of the MDS maps. The maps seem to indicate that the frequency of other variables increases as this one does; if this is so, providing a [-15-] context in which learners frequently expect to learn from other learners and the "teacher" (whether computer, human, or something else), and in which learners perceive that they are interacting with others on authentic topics, is essential to forming an optimal language learning environment. The indication that learners in both treatment groups perceived that the audience was more authentic (a better source of English and more interes ted) after the intervention may be important in that all participants were working in labs with or in the presence of fellow learners rather than in a teacher-fronted context as in the initial classroom. This implies that, in these environments, participants perceived that interacting with a learner audience helps them learn English and that their partners are more interested in what they say than in the daily classroom context. Because learners in all environments seem to interact with other learners more frequently than with the teacher (i.e., they noted more help from other learners and more frequent interaction with classmates), learners must increase the frequency with which they perceive other learners as an authentic audience.

Another important finding in this study is that, as noted above, audience assumed a more central role than task. In this study it was found that although the task may be controlled by the teacher, perceptions of the task and the way in which the task is carried out are not necessarily under the teacher's control.

One explanation for a main effect for time is that all participants perceived themselves as learning a greater amount and more useful English at their tasks because they had more opportunities during the intervention to deal with either new cultural concepts or important grammatical ones. High interest for something new (a halo effect) such as the use of the computers and a continuing belief in the importance of whatever English they are exposed to could also account for the change over time. All of these interpretations help to explain the findings.

The MDS map for the traditional treatment suggests that the constructs Task, Help, Audience, and Cognition are related; that is, as the frequency of one increases, so does the frequency of the others. These constructs might be related in a pattern such as the following: Tasks help participants learn important English, which is more interesting when there is appropriate feedback and enough time to complete the work, and the more interesting and important it is, the more learners pay attention and try to do their best. If this is the case, this seems to be optimal because the basis for the perception of the other constructs is Task, which the teacher controls. Ironically, this pattern was expected for the cooperative treatment and not for the traditional treatment. Because the MDS maps do not enable one to assign causal connections, one can only theorize reasons why task seems to be a central variable in the traditional treatment map and a relative outlier in the cooperative one. One hint is the anecdotal comment that the culture task was [-16-] like a game, and so while it may be interesting to the participants, perhaps it did not require participants to pay more attention or cause participants to perceive that their audience were commensurably more authentic. Therefore, while the frequency of interest in and perceived importance of task may increase from the pretest to the posttest, it does not increase in relation to any of the other variables on the map of the cooperative treatment.

Much of the evidence points to the idea that working with grammar is perceived by learners as an authentic task, although the literature claims that grammar activities, particularly in the drill-and-practice format employed in the traditional intervention, are not authentic because learners do not gain fluency as a result of their use. There are several reasons to reexamine this claim: First, because there is no concrete evidence as to how learners become fluent and how they process language, the contrib utions of drills and other repetitive exercises to language acquisition cannot be determined. In addition, learners who need to use grammar every day often perceive it as the foundation of the language. Given their fright of making mistakes (noted in the analysis of the Atmosphere construct), grammar may be the most authentic task content for these learners, regardless of the task structure.

Next, the greater absolute frequency of opportunities for production and exposure perceived by the traditional group may be due to the fact that each participant in the cooperative intervention had an essential individual subtask equivalent to the individual task in the traditional intervention, rather than truly being involved in every subtask and thereby gaining exposure to all of the sources used. In other words, because in the cooperative treatment each learner was given a different task and could have been preempted by group members even from exposure to the computer, the cooperative group members perceived that they had fewer opportunities to participate than were expected. This is confirmed by the posttest item analysis; the analysis shows that participants in the cooperative treatment perceived more sources but less opportunities to actually take part in the use of those sources, whereas those in the traditional treatment perceived more opportunities to do something but less sources to use during the activity.

On the other hand, the significant main effect for treatment could indicate that the cooperative treatment group members perceived that there were opportunities for exposure and production, but the data discussed above imply that they could not or did not take these opportunities. Data on whether and why learners take the opportunities offered to them would add to this analysis. Furthermore, anecdotal comments and indirect interpretation of the results indicate that learners help each other more than the teacher helps them individually. This is logical, given the number [-17-] of learners in each class and the self-help nature of the treatment tasks. Learners must perceive that classmates, who in fact seem to help them often, are good sources of feedback and other help. A more precise measure of the frequency of feedback by helper group may shed more light on interaction in the classroom.

Oddly, participants' perceptions of the relatively high frequency of "teacher friendliness' and "comfort working with others" seem to contradict their perceptions of "being nervous" and being "afraid to make mistakes." This indicates that something other than teacher personality or comfort may underlie production stress in this environment. Working alone or in small groups without a grade or other evaluation rather than in a teacher-fronted context should have led learners to experience less evaluation apprehension (see Myers, 1987, for a discussion of the research on this and related issues). This does not seem to be the case, as learners still perceived apprehension in the treatment environments as frequently as they did in the initial environment. It is unclear what the basis for this apprehension is. It could simply be the fact of being in a new situation, perhaps with other learners with whom they do not usually interact. It could be that a number of the participants are characteristically socially anxious. Alternatively, using Myers' (1987) arguments, because the participants in the treatments worked directly with other group members, they may in fact have felt pressured (an audience effect). The data collected for this study do not allow for a more specific interpretation; this is an interesting issue for future research.

The results for the scale measuring the control construct suggest that there may be a gap in the literature. Participants perceived that their opinions often had an impact on the classroom, regardless of their perception that they did not help make dec isions about content, activities, or whether to work in groups. This finding does not support suggestions in the literature that learners need control over content or activities. Learners' perception of control may be due to the opportunity for each lear ner to complete a useful task and to receive individual or small-group feedback when desired, to make decisions continuously, and to work at his/her own pace; this is a change from the teacher-fronted classroom for all the learners. In addition, data from the other scales indicates that learners control their environment in other ways. Learners in these treatments may perceive that they are in control and are exerting their opinion, for example, by choosing when, how, and in which language to interact. A different measure of control which taps these concepts (which seem evident in the environment) should be included in the survey instrument for a more accurate analysis of learner control.

Finally, learners in the cooperative treatment were paying attention and trying their best 75-80% of the time, but this effort was evidently unrelated to their perception of task interest. What, [-18-] then, made them pay attention? The significant covariate, not included in the maps, indicates that learners paid attention more frequently when they had previous computer use. Almost 70% of the treatment one cooperative group had previous computer use. In data collected to explain the demographics of the participants, learners reported that those who had previous computer use had used the computer for writing, a completely different activity than the task during the intervention. However, these participants had some knowledge of computers, and were therefore possibly more interested or more able to use them (or less threatened) and so paid more attention. Those participants without previous computer use may have been less interested in the computer because they either had no previous knowledge or opinion about using this type of technology or simply did not like the idea and chose, therefore, not to use the technology. These learners may have been unsure of what to do and less able or less interested than if they were in the classroom with a human teacher. Previous computer use did not covary with any of the other variables, indicating that computers were a source of interest or non-interest but not a factor in participants' perceptions of comfort in the environment or lack of control as might be expected from learners who had no previous use.

Practical Implications

To summarize, in this study participants perceived the intervention environments in many expected and some unexpected ways. Although learners do sometimes change perceptions in keeping with expectations, it is often only in examining the environments in retrospect that these changes seem logical. This situation indicates several things:

1) There may be important differences between environments that seem at first examination to be similar, and environments that seem very different to observers may not be significantly so to the participants in those environments;

2) It is important for teachers to be aware of learners' perceptions to help them make appropriate choices and to design environments that fit their learners. Awareness includes acknowledging that the environment is a system and not a number of discrete variables, and that, therefore, one seemingly small change in the classroom environment can have unforeseen impacts. The intervention environments, more than anything else, seem to have had an impact because the computers allowed the classroom to be "individualized" (or to promote "individually appropriate instruction") in a way that is difficult, if not impossible, without the technology.

It is also important for language educators to realize that learners may differ, regardless of what current methodologies dictate, on what constitutes authentic content for a language [-19-] course, and that intentional cognition is not always based on how interesting the task is. This claim is made with a caveat; in this study, learners coming to the culture treatment from grammar-based classrooms may not have understood the pedagogical value of the culture task and were therefore inclined to treat the task as a diversion or some form of entertainment.

The results of this study indicate that process losses caused in learner groups are of great concern. It seems clear that it is not enough to give each learner in a group a subtask and expect learners to cooperate to achieve a specific outcome; for maximum exposure and language use, the task should be constructed so that each learner needs to interact with all the resources available to the group and is responsible for a portion of the outcome.

It should also be noted that giving control to learners may be as simple as providing enough time for learners to complete their work, as complex as weighing a number of control variables, or as interesting as letting learners establish their own methods and levels of control by recognizing and encouraging it when appropriate. The findings of this study also indicate that learners pay more attention during a computer-assisted task when they are familiar with the technology. This indicates that time is well-spent familiarizing learners with the computers before undertaking computer-supported tasks.

There are many problems linked to studying intact classrooms, not the least the inability to randomly select. Because the class technically contains the whole population, it is difficult to generalize to other situations; however, "situated findings" (Salomon, 1992, p. 5), Although they may apply only to local situations, also present an idea of the interactions that take place in environments. Therefore, where the patterns outlined in this study are specific for these learners in these environments, the idea that patterns can be discerned and may change in similar manners is important and universal. In addition, because the design did not include a control group, absolute cause cannot be assigned; however, this does not preclude the discussion of differential impact, or a description of the differences between treatment groups after the intervention (Cook and Campbell, 1979). Finally, Nass & Mason (1990) claim that "theories that are specified or operationalized in terms of one type of technology can never be applied to any other technology" because "we do not know what feature or characteristic of the technology has been caused or has led to the effect under study'"(p. 49). They conclude that such studies end up explaining more about the environment than about the technology; as indicated above, this is, in fact, a more desirable outcome of studies of language learning supported by technology than effects solely of the technology.

Implications for Future Research

This study was based on the premise that the study of computer- supported learning environments is a step toward developing a theory of computer-assisted language learning (CALL). However, the computer as an entity did not have a major impact on the outcomes of this study, as it should not if it is integrated into the environment as the theory suggests; it is the environmental constructs which computer use can affect which play an integral part in the outcomes. Rather than searching for a separate theory of CALL, therefore, perhaps it is more appropriate to consider as a goal the development of a cogent theory of second language acquisition that includes theories of technology-supported language learning environments and learner perceptions of these environments. This issue, and others mentioned previously, are topics for future research and can be addressed with the model of the current study as a basis.

References

Ahmad, K., Corbett, G., Rogers, M., & Sussex, R. (1985). Computers, language learning, and language teaching Cambridge: Cambridge University Press.

Anderson, G. (1973).The assessment of learning environments: A manual for the Learning Environment Inventory and the My Class Inventory ERIC Document.

Bereiter, C. & Scardamalia, M. (1987).The psychology of written composition Hillsdale, NJ: Lawrence Erlbaum Associates.

Brown, H. (1987).Principles of language learning and teaching, 2nd edition Englewood Cliffs, New Jersey: Prentice-Hall Regents.

Cook, T, & Campbell, D. (1979).Quasi-experimentation: Design & analysis issues for field settings Boston: Houghton-Mifflin Company.

Doughty, C., & Pica, T. (1986). "Information gap" tasks: Do they facilitate second language acquisition?TESOL Quarterly 20(2) 305-325.

Egbert, J. (1993).Learner perceptions of computer-supported language learning environments: Analytic and systemic analyses Unpublished doctoral dissertation. Tucson: University of Arizona.

Ellis, R. (1986).Understanding second language acquisition Oxford: Oxford University Press.

Fraser, B. (1986).Classroom environment London: Croom Helm. [-21-]

Johnson, D. (1991). Second language and content learning with computers: Research in the role of social factors. In P. Dunkel (Ed.),Computer-assisted language learning and testing: Research issues and practice(pp. 61-83). New York: Newbury House.

Kelman, P. (1990). Alternatives to Integrated Instructional Systems.CUE Newsletter 13(2) 7-9.

Kerr, N., & Brunn, S. (1981). Ringelman revisited: Alternative explanations for the social loafing effect.Personality and Social Psychology Bulletin 7 224-231.

Klein, W. (1986).Second language acquisition Cambridge:Cambridge University Press.

Krashen, S. & Terrell, T. (1983).The Natural Approach:Language acquisition in the classroom Hayward, CA: The Alemany Press.

Kreeft-Peyton, J. (1990, Special Issue). Technological innovation meets institution: Birth of creativity or murder of a great idea?Computers and Composition 7 15-32.

Kremers, M. (1990). Sharing authority in a synchronous network: The case for riding the beast.Computers and Composition 7 33-44.

Levin, J. & Boruta, M. (1983). Writing with computers in classrooms: "You get exactly the right amount of space!".Theory into Practice 22 291-295.

Levy, M. (1990). Towards a theory of CALL.CAELL Journal 1(4) 5-7.

Lozanov, G. (1978).Suggestology and outlines of suggestopedy New York: Gordon and Breach.

McKinnon, K. (1989).MacTrivia TESOL CALL Interest Section Shareware. Alexandria, VA: TESOL.

McLaughlin, B. (1987).Theories of second-language learning London: Edward Arnold.

Meskill, C. (1992). Off-screen talk and CALL: Role of the machine/participant.CAELL 3(1) 2-9.

Myers, D. (1987).Social psychology New York: McGraw-Hill. Nass, C. & Mason, L. (1990). On the study of technology and task: A variable-based approach. In J. Fulk & C.Steinfield (Eds.), _Organizations and Communication Technology(pp. 46-67). Newbury Park, CA: Sage Publications. [-22-]

Pica, T. (1987). Second language acquisition, social interaction, and the classroom. Applied Linguistics 8(1) 3-21.

Pica, T. & Doughty, C. (1985). Input and interaction in the communicative language classroom: A comparison of teacher-fronted and group activities. In S. Gass & C. Madden (Eds.),Input and second language acquisition(pp. 115-132). New York: Newbury House.

Reid, J. (1989). English as a Second Language in higher education: The expectations of the academic audience. In D. Johnson & D. Roen (Eds.), Richness in writing: Empowering ESL students (pp. 220-234). White Plains, NY: Longman.

Richard-Amato, P. (1988).Making it happen: Interaction in the second language classroom - From theory to practice New York: Longman.

Robinson, G. (1991). Effective feedback strategies in CALL: Learning theory and empirical research. In P. Dunkel (Ed.), Computer-assisted language learning and testing: Research issues and practice(pp.155-168). New York: Newbury House.

Roen, D. (1989). Developing effective assignments for second language writers. In D. Johnson & D. Roen (Eds.),Richness in writing: Empowering ESL students(pp. 193-206). New York: Longman.

Salomon, G. (1992).New challenges for educational research: Studying the individual within learning environments Presidential Address, South Western Educational Research Association, Houston, January 30.

Salomon, G. (1990). Cognitive effects with and of technology. Communications Research 17(1) 26-44.

Spolsky, B. (1989).Conditions for second language learning Oxford: Oxford University Press.

Vygotsky, L. (1978).Mind in society Cambridge: Harvard University Press.

Webb, N. (1982). Student interaction and learning in small groups.Review of Educational Research 52 421 - 445.

Webb, N. (1985). Verbal interaction and learning in peer- directed groups.Theory into Practice 24(1), 32 - 39.

Zellermayer, M., Salomon, G., Globerson, T., & Givon, H. (1991). Enhancing writing-related metacognitions through a computerized writing partner.American Educational Research Journal 28 372-392. [-23-]

Joy Egbert, PhD, is an award winning teacher, researcher, and materials developer. She is the Technology Coordinator for the Intensive English Program in the Center for English Language Training at Indiana University and a lecturer in the School of Education. Her virtual CALL class is under development at http://ezinfo.ucs.indiana.edu/~jegbert/callpage.html.

Leonard M. Jessup, PhD, is an associate professor of Information Systems in the School of Business at Indiana University. He teaches in various areas of management and management information systems. He has published and presented widely on computer-supported collaborative interaction, electronic commerce, and the uses of technology for teaching and learning. His WWW home page is at http://ezinfo.ucs.indiana.edu/~ljessup.

Joy Egbert
Center for English Language Training
Indiana University
jegbert@indiana.edu

Leonard M. Jessup
School of Business
Indiana University
ljessup@indiana.edu