Chapter 5. Manipulating task complexity across task types and modes

This paper reports on the impact of manipulating cognitive task complexity on L2 production across task types and across modes. Specifically, the paper presents results from two separate but interrelated studies. The first study measured the impact of manipulating the cognitive demands of tasks along the degree of displaced, past time reference in a narrative task, the number of elements in an instruction-giving map task, and the reasoning demands in a decision-making task. A repeated-measured design was used in which 42 Spanish/Catalan-speaking learners of English performed a simple and a complex version of the three task types. In the second study the same tasks were used but in a dialogic mode. Forty learners organized into 20 dyads performed the six tasks under the different conditions of complexity. In both studies, speed of production (syllables per minute in unpruned and pruned speech) and pausing behaviour (number of filled pauses) were calculated, as well as structural complexity (S-Nodes per AS-unit), lexical complexity (Guiraud’s Index of Lexical Richness), and accuracy (Number of errors per 100 words and the ratio of repaired to unrepaired errors). An affective perception questionnaire was used to measure difficulty, stress, confidence, interest, and motivation, and proficiency was used as a control variable by means of X-Lex and Y-Lex vocabulary size tests. Both correlations and non-parametric Wilcoxon’s Signed Ranks tests were used to capture the impact of proficiency and Task Complexity on production. Results show that, first, learners’ proficiency correlates with fluency, lexical complexity and accuracy in monologic tasks, whereas only weak correlations are found when tasks are performed dialogically. Secondly, task complexity has a different impact on each type of task, with learners being more accurate on the complex narrative task, less fluent and more lexically rich and accurate on the complex instruction-giving task, but task complexity has little or no impact on the decision-making task. Thirdly, each task type is shown to trigger very different types of discourse which are compared for each dimension of production. Results are discussed by specifically focusing on the similarities and differences existing between tasks and modes, and in l ight of attention theories in relation to L2 task performance (Skehan’s Limited Capacity Model and Robinson’s Cognition Hypothesis) as well as L2 speech production models (Kormos, this volume).