-=> Quoting Ron Mcdermott to Carl Bogardus <=-
 CB>Do we get quality graduates? No--testing does not produce 
 CB>quality, testing only finds defects.
 
 RM> You've been repeating this mantra for some time now, and 
 RM> with all due respect to Mr. Deming, I don't think that his 
 RM> theories on manufacturing necessarily carry over to 
 RM> education.  Specifically, I find it hard to accept that
 RM> having exit exams wouldn't improve the quality of our
 RM> graduates.
I think Carl's point is analogous to the reliability engineer's saying
that "you can't test quality into a product, you must design it in."
In both reliability engineering and education the saying is an
exageration to make a point.  In reliability engineering, the point
is that changes in design and manufacturing can have huge payoffs
compared to screening the final product.  Remember when "high density"
floppy disks were much more expensive the "double density?"  The
manufacturers didn't make high density disks, they made disks and then
tested them.  Early on most disks weren't good enough for high density,
so they were labeled double density.  High density disks didn't get
cheaper because they developed better tests - they got cheaper because
they tweaked the manufacturing process until a much higher fraction
passed the test.  (Still later there were too few poor disks, and
disk notchers became popular, but that's an economics story and this
is an engineering story).  Did testing produce high density disks?  In
a sense yes; the outcome of the testing process was high density disks.
But looking at it this way misdirects attention from the manufacturing
process, where the big improvements happened.
 CB>You might find graduate quality---but where do you set the test level?
 RM> Somewhere along the line, we have to decide what it is that
 RM> we want students to know (or know how to do), and test for
 RM> those things.
I've seen this approach fail too often.  The result is often "teaching
to the test."  Back in the 70's it was common for NY high school
graduates to be able to solve standard "work problems" (Ron digs a 
ditch in 6 days, William digs a ditch in 12 days...) by a tableau
method of setting up "boxes" in a memorized arrangement and then
filling the boxes by memorized rules.  But was also common that they
didn't know why it worked, and they couldn't handle twists of the
same problem (Ron in 6, Together 4, How long for William alone?).
My first job was revising a test for measuring how well telephone
managers were managing the service in older style telephone systems.
You could get pockets of poor service (frequent slow dial tone) if you
weren't careful about where you connected new phone service.  If you did
the job properly, the test did a good job of measuring how well you
did the job.  But people had discovered that you could give a small
group of customers really bad service - must worse than anyone would
get if you did the job properly but poorly - and everyone else would
have really good service and the test would come out good because only
a few people had sub-standard service.  This behavior was similar
to "teaching to the test;" it was behavior modified in unproductive
ways to maximize test scores.
So I worry that an exit test focuses the attention on testing
instead of education, with unintended deleterious side effects.
 
___ Blue Wave/QWK v2.12
--- Maximus/2 3.01
---------------