The Acquisition of Academic English by Children in Two-Way Programs: What does the Research Say?

Table 4 presents NCE scores and percentiles for English reading. Note that TBE groups appear twice in the table: Each type of "enriched" bilingual education program had its own comparison group, matching it in neighborhood and socioeconomic status.

Cross-sectional comparisons pit TBE comparison groups against each of the enriched bilingual options. There appears to be no difference in the TBE- developmental comparison, but two-way students do better than TBE comparisons. Also, two-way students clearly do better than the developmental students. Although this was not a “planned” comparison with control for socio-economic status, the result is noteworthy because two-way students performed better than TBE students, but developmental students did not.

Houston: Potential Confounds

Before concluding that this data demonstrates the superiority of two-way programs, we need to discuss some unusual aspects of the Houston data.

1.	Scores for all groups are high in grade 1. In fact, they are higher than the level typically required for reclassification. This suggests that a significant percentage of the children had considerable English knowledge before starting school.
2.	Scores then decline for all groups, as we saw before in Castillo (2001). By grade 5, in fact, scores for some groups are below typical reclassification levels. This suggests that higher scoring children are being exited and lower scoring children are entering the program late. (Note that if this were true, it would exaggerate the effectiveness of a program that kept all its students and did not allow new ones to enter late, e.g. two-way bilingual education.)
3.	The two-way advantage was present early, in grade 1. This suggests that a selection bias was present, that the two-way students were exceptional from the beginning.

Thomas and Collier (personal communication) have pointed out that that children tested at the end of grade 1 could have been in school and receiving the benefits of bilingual education for three years before testing. Houston offers bilingual pre-kindergarten and kindergarten. It is thus possible that first grade scores do not represent beginning scores.

Thomas and Collier's Table C-8 provides some helpful information: It is a quasi- longitudinal analysis of children who have been in the Houston system for several years. It is not a pure longitudinal analysis, because different numbers of children were tested at each level; thus, the same cohorts were followed, but precisely the same children were not tested each year. Sometimes, in fact, the numbers vary quite a bit; there was a large increase in the number tested in grade 4 (see table 5). A valuable aspect of this data is the fact that separate scores are presented for children who began school in pre-k, kindergarten, and grade 1.

Table 5, from Thomas and Collier’s Table C-8, presents scores of children who began in school in grade 1, who did not have pre-kindergarten or kindergarten. When tested at the end of grade 1, they had had only one year of schooling. In all cases I present the sample size in parentheses, followed by the NCE score given in Thomas and Collier (2002), followed by the corresponding percentile. Note that scores are quite high even when children start school at grade 1.