Education Letter to the Editor

‘Reading First’ Story Shows Limits of Using Isolated Data

January 29, 2008 1 min read

To the Editor:

Your front-page article “Massive Funding Cuts to ‘Reading First’ Generate Worries for Struggling Schools” (Jan. 16, 2008) includes a discussion of the David Hill Elementary School in the Hillsboro district in Oregon. According to Hillsboro administrators, David Hill “more than doubled the proportion of its 3rd graders meeting state benchmarks—to 76 percent—from 2003 to 2005,” an improvement attributed to the Reading First program, which was implemented at the school in the fall of 2003.

The picture looks different when we consider more data. From the Oregon Department of Education’s Web site, here are the percentages of 3rd graders at David Hill meeting or exceeding grade-level achievement since the 2000-01 school year: in 2000-01, 82 percent (or 18 of 22 students); 2001-02, 65 percent (17 of 26); 2002-03, 33 percent (14 of 42); 2003-04, 79 percent (34 of 43); 2004-05, 73 percent (32 of 44); 2005-06, 67 percent (35 of 52); and 2006-07, 67 percent (24 of 36).

The number of children under discussion is quite small, a range of from 22 to 52. Also, scores are very similar over the years, except for an unusually low score in 2002-03, the year just before Reading First was introduced. They were not low before this.

In other words, the claim that Reading First worked is based on the scores of a few children on one test. If 14 more children had reached the benchmark in 2002-03, it would have looked just like the other years. We cannot judge the efficacy of a program on the basis of the scores of 14 children on one test.

A colleague alerted me to the presence of the pre-2003 data and has pointed out that data could be misused to argue against Reading First: The percentage of David Hill students meeting or exceeding the benchmark has declined almost steadily since 2003-04. Should we conclude, then, that the longer a school uses Reading First, the less effective it is? Clearly, the sample sizes are much too small to arrive at such a conclusion. And clearly, we need to look at more than a few isolated percentages to come to conclusions about the effectiveness of programs.

Stephen Krashen

Professor Emeritus

Rossier School of Education

University of Southern California

Los Angeles, Calif.

A version of this article appeared in the January 30, 2008 edition of Education Week as ‘Reading First’ Story Shows Limits of Using Isolated Data