Friday, April 14, 2006

mixer activity: unmixed

I suppose it was bound to happen. A few days ago, for test design purposes, I made a command decision and restructured my mixer exercise (for background, refer to this previous meditation on the mixer quiz): a student now signs her paper, gives it to her partner, and then keeps that partner for the duration of the activity.

The bad news: what I have is no longer a mixer.

The good news: a much-improved format for testing that eliminates many of the cheating- and confusion-related problems inherent in previous mixers.

The original mixer format is still viable for class activities, but for tests, I'll be sticking to the one-partner concept. I made two other changes as well:

1. Only one partner at a time has a test sheet. This makes it very clear to the students that the speaker must speak and not write anything. When both partners have mixer papers in front of them, both partners also have pens and pencils. Instead of talking, many partners quietly reach across to their test papers and scribble a correction, or even an entire answer.

2. Mixer tests are distributed in two versions: A and B. Yes, it means more work for me when it's time to create tests, but I think it's worth it because it keeps nearby students from eavesdropping.

Again, these changes to the mixer are purely for tests-- not for regular mixer activities, where the ambience is more relaxed and a bit of chaos is actually desirable.

What I absolutely want to avoid is the sort of exam where an English conversation student spends half a semester speaking in class, and then is suddenly confronted with a written exam. My mixer exam has a writing component, to be sure, but the person doing the writing is not the student being tested: it's the student's partner, whose sole function is to take dictation. If the partner makes a mistake in writing down an utterance, it's up to the speaker to tell her partner what correction needs to be made. The speaker can't reach for a pen or an eraser to make the correction in silence.

While the format still isn't perfect (what format is?), today proved that I've hit upon a reliable testing method. On Mondays, Wednesdays, and Fridays, I teach four Level 1 classes. The classes are of different sizes and collective temperaments, which meant, earlier today, that the newly redesigned test would face four different environments. Result: the new test worked like a charm in all four classes. Student questions about how to take the test were few to nonexistent, and now that I've graded the test papers, I'm seeing student results that dovetail fairly well with my own impression of each student's proficiency. (I had a suspicion that, with previous versions of the mixer test, grades may have been somewhat inflated.) This time around, we saw something like a bell curve: a lot more Bs and C, only a few As, and one or two Ds. No Fs, fortunately.

The test also takes less time now. A true mixer involves quite a bit of Brownian motion as students buzz around the classroom seeking partner after partner. By chaining each test-taker to a single partner, we cut down on motion, streamline the process, and introduce a lot more focus to the activity.

An additional benefit of all these changes is that I can finally address one of the most egregious flaws of the mixer format: namely, the absence of a listening component on the test. An English conversation class will inevitably cover all four macroskills: listening, speaking, reading, and writing, but a disproportionate priority will, as is only proper, be placed on the first two skills. My previous mixer exams allowed a student to speak, but because the speaker's partner was silent, and because the speaker's utterance was prompted by something written on a sheet of paper, there was no true listening involved.

I now have 10-15 extra minutes in a given testing hour, so I'm free to whip out the pre-fab audio test components that accompany our textbook. They're specifically designed for quizzing and testing. The listening comprehension section of my new test involves a dialogue (on audio CD) followed by multiple choice questions. While I normally avoid multiple choice like the plague, I think it's fine for listening comp on the test: not so different from what students normally encounter when taking TOEIC and TOEFL tests.

I might improve upon the listening section in successive versions of my increasingly streamlined exam (perhaps students will hear a question and write a short rejoinder, for instance), but for now, multiple choice listening comp is better than no listening comp at all.

In all, not a bad day for testing.


No comments: