Thus, the AR children read slightly more (two hours) than the comparison children. This is contrary to the claim made by Vollands et. al. (1999) that the comparisons read more. (All this is highly speculative. The detailed report, Vollands et. al., 1996, contains a number of diary entries. An entry for the last month of the project, March 1995, contains the following: "No information was available in terms of the average amount of time spent reading by the children in the class as the (AR) teacher did not keep a running note of this (although she had been asked to do this)." (p. 79). It is not clear whether this observation applied only to that month.)

Adding to the confusion, the AR students were sixth graders (11 year olds), while the comparisons were fifth graders (10 year olds), even though comparisons had higher pretest reading scores. Also, in the detailed report (Vollands et. al., 1996), it was noted that a thunderstorm occurred during the pretest for the AR students, which was distracting: students "were more interesting in looking out the window than taking a reading test" (p. 98). Moreover, "cheating was witnessed" during the comparison group post-test!

The results are inconsistent. Comparison students made larger gains on one test of reading comprehension (Edinburgh) but AR students made larger gains on another (Neale), with comparisons making no gains at all on the Neale (table 10), a mysterious result for a group of good readers. This inconsistency may be due to the fact that all 26 comparison students and nearly all AR students took the Edinburgh test but only a random sample of 11 AR students and 12 comparison students took the Neale comprehension test. If one considers the Edinburgh to be the valid measure, comparisons doing less reading outperformed AR students who did more.

Thus, this study could be interpreted as showing the failure of AR. The comparisons gained more, despite reading slightly less, on the test taken by the full sample. If the thunderstorm seriously affected pretest reading scores for the AR group, they actually did worse than the results indicate. But if cheating on the posttest was widespread for the comparison groups, their gains would be exaggerated. These factors, the small sample size, inconsistency of the results, and the use of incentives in both groups makes it difficult to conclude much of anything from this study.

Table 10 Volland et. al., Project B

Edinburgh reading comprehension

Neale reading comprehension

standard deviations in parentheses

Facemire (2000) also used a comparison group that was engaged in recreational reading. AR was used over a nine week period with 15 third graders in a low SES area of West Virginia. Extrinsic rewards were limited to "bookmarks and paperback books" (p. 28). The school library had a total of 8000 books but only 800 were designated by AR. AR students were given "as much time as possible, at least 20 minutes per day" (p. 35) for reading. They also were "guided to appropriate reading materials during regularly scheduled library visits, eighty minutes per week, and were given time to discuss literature" (p. 35).

11