Fwd: [eddra] tests

[Peter Farruggio <pfarr@cal.berkeley.edu>]

Bracey's reply to a friendly academic who had asked for info about 
good tests (he seemed to have standardized tests in mind)...



> From: GERALD BRACEY <gbracey1@verizon.net>
>
>
> Tom,
>
> About 15 years ago, psychometrician Bob Mislevy wrote that it's not 
> too much of an exaggeration to say that the tests that dominate 
> education reflect 20th century statistics applied to 19th century 
> psychology.  I might argue for 18th century because I see the hand 
> of the British Empiricists--Locke, Berkeley, Hume, that bunch.  But 
> his point was and no doubt still is that tests don't reflect the way 
> people learn.  Most psychometricians don't know much about cognitive 
> psychology.  A study by Lorrie Shepard about 15 years ago indicated 
> that at the state and district level, people in charge of testing 
> had very naive theories about how learning occurs.
>
> Partly, this lack stems from the history of testing.  Testing as 
> developed by Yerkes, Terman, etc., was not about learning.  It was 
> about discriminating.  They wanted tests that would discriminate 
> among the testtakers because that was the only way they could make 
> differential predictions about those testtakers--which is what they 
> really wanted to do.
>
> My gripe with NAEP is not with the tests themselves which probably 
> are among the best.  It's with the achievement levels which are, in 
> addition to being arbitary, irrationally high.
>
> Similarly, it is the condition of high stakes testing that is the 
> problem, not the tests.  When Bob Linn took a 50 year retrospective 
> look at the use of tests to improve education he concluded that 
> little had been accomplished and the reason was that tests that were 
> adequate as monitors lost their validity when high stakes were 
> attached (Education Researcher, March 2000).  This conclusion was 
> amplified in the monograph from Sharon Nichols and Dave Berliner 
> "The Inevitable Corruption of Indicators and Educators Through High 
> Stakes Testing."  That will be out soon as a book, Berliner and 
> Nichols, Collateral Damage.
>
> Jerry