Method Comparisons: Conclusion

Date: 2015-10-07; view: 324.

Several studies should not have been included in the meta-analysis. In some, it was not at all clear that there was a genuine comparison of whole language/language experience and basal/skill-oriented methods. These include Stallings (1975), which was not a method comparison at all, Ewoldt (1976), in which many "language experience" subjects did not actually have language experience instruction, Usova and Usova (1993), which included a wide variety of activities under "whole language" and had a very small sample size, and McCanne (1966), which used language experience for second language acquirers, which may have been inappropriate because of premature production demands. Applying my criteria of using only tests of reading comprehension eliminates Morrow, O'Connor and Smith (1990) and Dahl and Freppon (1995) as well.

Of the five studies remaining, in three cases I reached conclusions opposite to those of Jeynes and Littell: I have argued that results were inaccurately reported in Carline and Hoffman (1976) and Lamb (1972) and that the direction of the effect size should be reversed in Harris and Serwer (Harris, Serwer and Gold, 1966). If my interpretations are correct, effect sizes for the still- eligible studies should be changed: for Hoffman, from -.23 to + .68; for Lamb, from -.75 to +.44; for Harris and Serwer, from -.51 to .18, for Manning et. al., from 1.21 to 1.97. The effect size for Morrow (1992) remains the same.

Only two studies in this set provide clear evidence that one group did more reading than the other, and in both cases those who read more did better on tests of reading comprehension; Morrow (1992), with an effect size of 1.23, and Harris and Serwer (1966a, 1966b), with an effect size of .18. For other studies in which groups are labeled whole language or language experience, with no clear data on amount read, effect sizes still favor whole language: Carline and Hoffman (1976), with an effect size of .68, Lamb (1972), an effect size of .44, and Manning et. al., (1989), with an effect size of 1.97. The average effect size for all five of these studies is +.90 favoring whole language/language experience. Excluding Manning et. al., the study with the largest effect size, the average effect size is +.63. Even if we include McCanne (1966) and exclude Manning et. al., the average is +.38 in favor of whole language/language experience.

Note that this conclusion is not dependent on my policy of limiting measures to tests of reading comprehension. Allowing Morrow, O'Connor and Smith (1990) (d = -.18) and Dahl and Freppon (1995) (d = .67) into the re-analysis does not change the final result very much.

Jeynes and Littell's conclusion was that although "pure" whole language students did well, basal/skill groups were in general a winner over whole language, with a mean effect size of -.65 in favor of the comparison groups. My conclusion is nearly exactly the opposite.

In my view, neither of our results should be taken as definitive. No study considered the amount of real reading done to be a central variable, and only two studies attempted to determine the amount of reading students did. In light of Harris and Serwer's finding (see also Evans and Carr, 1985) that children in skills classes actually spent more time reading than children in language experience, one must be cautious in concluding that children in any language experience or whole language class actually read more than those in traditional classes.

What is clear, however, is that Jeynes and Littell's interpretation of the research is not the only possible one.