Co-constructed discourse in small group interaction

Merril Swain has given an elegant and thought provoking lecture at the 22nd Annual Language Testing Research Colloquium held in Vancouver, British Columbia in March, 2000. The following is a sort of summary of that lecture. The title of the lecture is: 'Examining dialogue: another approach to content specification and to validating inferences drawn from test scores.'

Image from HERE

Interaction in small groups is very important for pedagogic and testing research since it is very practical to use such interactions in classroom and in testing. Swain sees it as an interface between testing and learning research in second language. Small groups do not have the asymmetry of power as in an interview where the interlocutor is usually a senior, more powerful person. 

The way Swain looks at discourse and text analysis is different from the conventional. She looks at the content and underlying strategic and cognitive processes that generate and are generated by the discourse in the task. This approach studies real dialogues between people, not monologues or recorded 'think-alouds'. 

Small group tasks are included even in high stakes tests. The reasons why small group interactions are included in many tests these days are:

    1. dissatisfaction with oral interviews as the sole means of assessing oral proficiency

    2. search for other tasks that can elicit aspects of oral proficiency

    3. an attempt to mirror teaching practices in testing

    4. economic reasons- group tasks are less expensive than one-to-one oral interviews

Since small group tasks are used in high stakes tests and tests in general to make important decisions, validation is necessary (but not much validation work is done in this field). What is the basis of interpreting group performance as evidence for individual language ability? In this light McNamara asks, "whose performance is it anyway?". Interviewer behaviour is found to influence candidate performance either positively or negatively. In other words, performance is never solo. It is always jointly constructed in an oral interaction. Fulcher states that candidates found group interaction less anxiety generating than one-to-one interaction. This means, affective responses lead to differences in performance. Barry found complex relationship between test-taker's and other participants' characteristics (in a study of extroversion and performance). 

The above survey shows us that individual performance in a group task cannot be safely interpreted as evidence for underlying language abilities of individuals. A situated performance interpretation is needed. If not, interpretation of tests might lead to unfair biases and induced biases.

Sociocultural Theory of Mind

Swain's research sees how output serves L2 learning. Stretching present stage of language through earner's need to communicate successfully creates linguistic form and meaning, leading to noticing gaps in linguistic system, and thus learning process. Learning may happen through use of a dictionary, grammar book, or asking a peer or a teacher or generating and testing hypotheses, etc. Thus through attempts to communicate successfully, learning happens. 

She sees output as dialogue, as a response to criticisms against seeing language as input and output, as mechanical system. Such dialogues serve both communicative and cognitive functions. This view was developed from Vygotskian theory and its interpretation.

Orignins of cognitive functioning is primarily social. Two ideas she discusses are:

1. Higher cognitive processes are mediated activities and their source is interaction.

Dialogues generate strategies. Strategies become strategic patterns of reasoning at cognitive level. In dialogues during problem solving tasks, these strategic processes become visible. Thus in mediated learning, strategies employed by learners are visible and can be studied. 

2. Knowledge is constructed through dialogue. Dialogue can be with self or others. Dialogue mediates construction of knowledge. Co-construction of linguistic knowledge happens in dialogue. In dialogue, successful communication is important. Collaborative dialogues build knowledge in the course of problem solving using language.

Implications for Testing

1. Dialogue provides validation evidence. In dialogue, cognitive and strategic processes are visible. So by studying dialogues, we understand how participants approach task's demands. This understanding of strategies and processes used can inform understanding of constructs being measured. Therefore, dialogues help validation. 

2. The process and outcome (both) of interaction are a joint achievement. Therefore, caution must be exercised in interpretation of group performance as evidence for individual linguistic abilities. 

Swain takes neo-Vygotskian Socio-cognitive perspective for her research on dialogue. These are the questions she asks: How do we know that there is learning in Co-constructed dialogue? Would one type of task be more useful to focus students' attention on form than another?

Swain finds more than input and output in the tasks she studied. There is collaborative co-construction of learning. There is negotiation of meaning, hypothesis formation and hypothesis testing. Collaboratively, participants reach a solution. Therefore it is right to ask whose performance it is.

She used dictogloss and jigsaw tasks in her research. Information gap was embedded in jigsaw task. Jigsaw task did not give a language model, but comprehending the story from given pictures was easy. Dictogloss task invoved listening to a story read aloud at normal speed. Dictogloss task gave a language model. But without comprehending the story delivered orally, candidates could not proceed with the task. So, processing demands of both tasks are different. Therefore, she expected different strategies and performance levels on both tasks. But the prediction went wrong. Learners performed equally well on both the tasks. Also, proficiency interacted with performance. High proficiency students wrote well on dictogloss and jigsaw tasks. Low proficiency students came up because they had vocabulary help in the dictogloss task. Generally, jigsaw students performed better than dictogloss students. Another point she emphasizes is that we cannot predict what the test-takers would focus in a task- however controlled the task is. Therefore, she suggests to make use of task discourse for materials for measurement. 

Summary 

Small group is of interest to both language learning and testing research. Testers usually measure performance in small groups. In a group, performance is jointly constructed, and distributed among participants. Dialogues in groups foreground cognitive and strategic processes. Testing therefore must look for fair means for scoring this shared or co-constructed dialogue as individual ability. Interlocutor in a pair is very important. 

Dialogues lead to learning. Measuring implies measuring how much learning has taken place too. Therefore examining content of dialogues may be useful. We can understand cognitive and strategic processes involved in performance. Qualitative validation methods like expert judgement, introspective and retrospective accounts of test takers and raters, interviews and test and discourse analysis of performance can inform us of the important characteristics of tasks, discourse and interlocutors.