Presentation on theme: "David P. Ellis University of Maryland"— Presentation transcript:
1David P. Ellis University of Maryland email@example.com The Relationship between Task Complexity and Linguistic Complexity: An Analysis of L1 Speaker ProductionDavid P. EllisUniversity of Maryland
2Research MotivationNeed for a reliable, transparent means of grading and sequencing tasks for both TBLT and TBAResearch findings to date on the linguistic influences of task complexity are ambiguousBelief that the underlying assumption of both existing task complexity models is misguided
3Model 1 - Skehan (1998) Three components Prediction Code Complexity (language required)Cognitive Complexity (thinking required)Cognitive FamiliarityCognitive ProcessingCommunicative Stress (performance conditions)PredictionAttentional resources are finite, so an increase in one dimension comes at the expense of the other two (e.g., an increase in linguistic complexity will result in a decrease in fluency and accuracy)
4Model 2 – Robinson (2007) Triadic Componential Framework Prediction Task Complexity (cognitive factors)Resource-directing (developmental)Resource-dispersing (performative)Task Conditions (interactional factors)Task Difficulty (learner factors)PredictionAn increase in task complexity along resource-directing dimensions decreases fluency, but increases both linguistic complexity and accuracy.
5Research QuestionsHow does task complexity influence the linguistic complexity of speaker output?Should linguistic complexity be the primary dependent variable of task complexity studies?
6Method Participants: Tasks: 24 native speakers of English (12 males, 12 females)Tasks:Map directions task (adapted from Robinson, 2001)Car accident reporting task (Lee, YG, unpublished)
7Procedure Repeated measures design with full counterbalancing Participant responses:Recorded digitally and uploaded to PCTranscribed into MS Word using Express ScribeParsed by clause and input into MS ExcelCoded for Linguistic ComplexityClauses per C-unit (Robinson, 2001)Clauses per AS-unit (Foster et al, 2000)Lexical “complexity” not measuredTwo independent coders; discrepancies resolved 100%Data analyzed using SPSS
9Discussion Possible reasons for null findings: Common measures of linguistic complexity are not appropriate operationalizations of the constructLinguistic complexity does correlate well with task complexity, but task complexity is not operationalized wellLinguistic complexity simply does not correlate with task complexity, at least within tasks
10Discussion (cont’d)Reasons to abandon linguistic complexity as the dependent variable of interest:Not a reliable correlate of task complexity within tasksEven if it were, it would say nothing about L2 development, only L2 productionTeachers/Raters cannot reliably assess the linguistic complexity of a speaker’s performance, especially in real time
11Discussion (cont’d) Alternative to linguistic complexity as the DV: Amount of production (i.e., word count)
12Preliminary Conclusion Increase the complexity of tasks to induce more linguistic output, not necessarily more linguistically complex output.RationaleDoing so will not only create more opportunities for learners to consolidate, reorganize, and improve access to existing L2 knowledge, thereby improving fluency, but also spur L2 development by inducing a greater number of gaps (both lexical and morpho-syntactic), thereby improving accuracy and potentially complexity.
13David P. Ellis University of Maryland firstname.lastname@example.org Thank youDavid P. EllisUniversity of Maryland