Introduction.
In June, 2002, the New
York Times reported that New York Regents Exams altered literature
passages on its tests according to "sensitivity guidelines." Some of the changes were ludicrous, some merely
stupid, and some would lead testtakers to the wrong answer. The state promised it would never do it again,
but as the Times' Michael
Winerip found, the censorship and bowdlerization is still going on. How the Department got caught and other aspects
of the affair were detailed in "The Twelfth Bracey Report on
the Condition of Public Education," Phi
Delta Kappan, October, 2002.
More important for this essay is a sentence in Winerip's article
that draws attention to another problem with tests--passing or failing
can turn on a single item.
---------------------
A statement buried in the 15th paragraph
of Michael Winerip's "How New York Exams Rewrite Literature (A
Sequel)"
deserves to be brought center stage.
Winerip wrote "In the world of make-or-break exams, one
question scored incorrectly can make all the difference in a student's
future." To people unfamiliar
with the technology of testing, Winerip's words probably look like
hyperbole. They are not.
We who have worked in the field of testing
have known this for some time, but the capriciousness of high-stakes
testing entered public discourse only because of Martin Swaden, a
Minnesota parent.
Minnesota requires students to pass a
test in order to obtain a high school diploma.
Swaden's daughter failed.
Swaden reasoned that the best way to help her next time round
was to look at the test and his daughter's answer sheet to see what
kinds of questions were giving her trouble. The state denied his repeated requests to see
them, yielding only when he threatened to sue (he's a lawyer, giving
the threat some credibility).
Matching his daughter's answer sheet against
the key containing the officially right answers, Swaden discovered
that the testing company, NCS Pearson, mis-scored six questions, enough
to put his daughter over the top.
She was hardly alone: 47,000 students got lower scores than
they deserved, 8000 students failed when they had should have passed
and 525 seniors who had actually passed were denied the right to walk
across the stage and pick up their diplomas at graduation ceremonies.
These 8000 falsely failed students did
not suffer mere temporary humiliation.
Some gave up jobs to attend summer school they didn't need.
Others forked out for tutors.
They suffered slights from classmates.
Some changed plans for the future and some dropped out of school. They also filed a class-action suit against
NCS.
NCS claimed that the error was a one-time
affair and couldn't happen again.
It found a scapegoat and fired him.
It hired the current or former state directors of testing from
Virginia, Washington and Texas to testify in its behalf (parents,
do not expect state officials to take the lead in ferreting out errors).
The testing director from Virginia said that such errors as
NCS made were likely.
The judge, though, rejected NCS' one-time-mistake
arguments, concluded that NCS had a long history of shoddy quality
control and had failed to hire enough employees to cure the problem.
He permitted the plaintiffs to sue for punitive damages. NCS
settled out of court. Some
students received as much as $16,000, but most got less than $2,000. The total award amounted to over $8,000,000.
Test companies fail the quality control
test far too often. CTB-McGraw
Hill errors forced thousands of New York City students to attend summer
school when they didn't need to.
Nevada officials caught a similar error in 2002.
Also in 2002, Georgia jettisoned the entire state's results
on Harcourt's SAT-9--the company mis-scored all 340,000 tests.
CTB-McGraw Hill's sloppiness so offended Arizona that the state
sought another contractor (but given the limited number of testing
companies with the capacity to handle large state contracts, had to
accept another with similar risks for errors. Basically there are only four companies: Harcourt
Educational Measurement, CTB-McGraw Hill, NCS, and Riverside. Educational Testing Service, once a major player
in the K-12 testing arena has decided to become on again).
All of these errors were made before the
Elementary and Secondary Education Act of 2001, the so-called No Child
Left Behind Act, had any impact.
That law, requiring all children be tested in grades three
through eight every year in reading and math (with science to be added
in two years) will vastly increase the amount of testing in this country
(and some think that it is no accident because the Bushes and the
McGraws who founded CTB-McGraw Hill have been vacation-together families
for 75 years). The same, short-staffed,
error-prone contractors will develop and score these tests.
We can thus safely predict the test companies
will make many more errors. We
cannot be confident that anyone will detect and report them.
Some kids will falsely flunk and some institutions will falsely
bear the label--and the consequences--of being a "failing school." The states and the testing companies, fearing
to look bad more than they fear doing wrong by children, will have
no incentive to look for the mistakes.
Indeed, one wonders how many other errors
already lie undetected in computer files
because parents from other states have not pressed to see how the
tests match their children's answer sheets.
They should. No one
else will. Swaden won without litigation in Minnesota.
A Florida parent had to sue to see her daughter's answer and.
Florida Governor Jeb Bush fought the suit claiming that letting
parents see their children's answer sheets would cost the state too
much money. He lost.
Most parents likely don't realize how
narrow and arbitrary the difference between pass and fail is. When
Virginia changed its test equating procedure for one test, the new
procedure required one less item correct to pass than the old procedure
called for (a peculiar outcome the state has not accounted for).
That one-item difference changed 5,625 flunks into passes (the
children were informed months later that they were no longer failures).
Virginia also lowered the passing score on several of its tests
in 2002, but did not apply the new standards retroactively. As a consequence, some 50,000 kids who flunked in 2000 and 2001
would have passed had they been lucky enough to take the test in 2002.
The testing industry is perhaps the largest
unregulated business in the nation. There is no FDA of testing to
stamp the tests for quality or make sure that the tests won't poison
anyone. Legislators, governors, and boards demand accountability
from teachers and school administrators, even though they only have
control over their "products" for a few hours a day, half
the year. Test companies have
total control over their products.
One wonders when they will be held accountable.
Minnesota and Florida have made starts, but only after determined
parents forced the issue. Parents in all other states should take
heed.