October 4, 2014
Midwest Association of Language Testers Conference
Ann Arbor, MI
CaMLA is organizing the 2014 MwALT conference and pre-conference workshop in Ann Arbor. I am serving on the program committee for the conference, and I will also be presenting a poster on my recent work developing the MET Writing test for CaMLA.
This work reports the task design process of the Michigan English Test (MET) writing section. The MET targets the A2–C1 levels of the CEFR (Council of Europe, 2001), but it is challenging to assess writing across this range with a single task. An essay task alone cannot differentiate levels below B1, since A2-level writers can only produce "phrases and sentences linked with simple connectors” (Council of Europe, 2001: 61). Conversely, the task must simultaneously target low- and high-proficiency writing goals to “[allow] better writers to demonstrate their best writing” (Weigle, 2002: 90).
Two tasks, each combined with a traditional essay, were piloted. Task A asked test-takers to recall a life experience, express an opinion about it, and elaborate on the situation in chunks of 1–3 sentences. Task B presented two photos and asked test-takers to compare them and state which they prefer. We show that responses to task B demonstrated far less syntactic complexity and propositional content, even for writers who demonstrated high ability on the essay task. Since these flaws in Task B show no correlation with writer L1 or overall ability, Task A was selected.
Task A elicits sentence-level language reflecting the writer’s ability, from beginner to advanced. The elicited language — despite being produced in stages — reads like a single, connected text, even for low-level writers. Finally, Task A prompts are simple to produce and can cover a wide range of topics familiar to test-takers of diverse backgrounds.