February 01, 2005

Test taking as audience design?

In my post on the questions that the SAT calls "Identifying Sentence Errors", I complained that the "decision about how to answer becomes a judgment about the linguistic ideology of the College Board, not a judgment about English grammar and style". In response, Mike Albaugh wrote

Long ago and far away, I took a California Civil Service Exam, with an eye toward getting a summer job as a computer operator or some such. By the fourth or fifth question, it was clear that the test had not been updated in quite a while. OK by me, as I had taken my first C.S. classes at a bit of a backwater community college, and so, in 1970, could answer the test "correctly" for 1960.

The wheels of Civil Service ground too slowly to match me up with a job before the end of that summer, but did grind finely enough to offer me a position at the then-princely sum of $14K/year. I pondered a while before deciding that continuing on to college would be a better long-term prospect, although it was nearly 1978 before I matched that salary in inflation-adjusted dollars.

The situation arises often. When I am interviewing for a job, or presenting some technical information, I try very hard to determine the level of expertise of my questioner, lest I give the "wrong" answer for that level of expertise.

This "theory of mind" reasoning is familiar to all of us, and not just in the context of a job interview or a technical presentation, because it's at the core of all human communication. Another relevant term of art is "audience design", which was used by Clark and Murphy (1983) in reference to reference, and by Bell (1984) in reference to style. As Peter Patrick puts it in his online lecture notes about the sociolinguistics of style, Bell's idea was that

speakers adjust their speech ... towards that of their audience in order to express solidarity or intimacy with them, or conversely away from their audience’s speech in order to express distance.

Recent work continues to emphasize a communicative view of style, in which speakers and writers make stylistic choices partly in view of their relationship with an audience.

A certain amount of "theory of mind" and "audience design" reasoning in test taking is inevitable, but it seems to me that we ought to minimize such factors in tests like the SAT. To evaluate students on knowledge of the norms of standard American English puts that kind of English in a privileged position, and rewards those who are familiar with it, but I agree with those who think it's right and proper to do so. I suppose that it's even plausible to grade students on their understanding of dubious made-up "rules" such the prohibitions about "singular their" and "split infinitives", since they will need to deal with people who believe such things. (I'd want the question to make the context clear, however -- "...this sentence contains what some people think is an error in number agreement, despite the fact that many respected writers and grammarians..."). I can see no value at all, however, in ranking students on how well they can predict whether or not ETS graders put credence in such phony "rules". At best this adds noise to the measurement; at worst, it measures an educationally irrelevant kind of cultural affinity.

Clark, H. H., & Murphy, G. L. (1983). "Audience design in meaning and reference." In J. F. LeNy & W. Kintsch (Eds.), Language and comprehension (pp. 287-299).
Bell, Allan (1984). "Language style as audience design." Language in Society 13(2): 145-204.


Posted by Mark Liberman at February 1, 2005 12:26 AM