Table 5
Children Beginning School at Grade 1 (NCE/Percentiles)

From: Thomas and Collier, 2002, table C-8.

With respect to concern (2), the decline, Thomas and Collier (personal communication) point out that declines are a common occurrence as students move toward middle school and a more demanding curriculum.

Thomas and Collier (personal communication) also point out that children are offered bilingual education in Texas from pre-kindergarten to grade five. Exit is not automatic when a certain level is reached. If this is so, if most students stay in the program, the exiting of high scorers is not a factor in the decline and two-way does not have an unfair advantage in this regard.

Interestingly, these drops appear to coincide with a reduction of the amount of instruction provided in the first language. They may also be related to a problem nearly all language minority children suffer from: because they also tend to be of low socio-economic class, they tend to live in neighborhoods and attend schools with little access to books (Neuman and Celano, 2001). Recreational reading may make a particularly strong contribution to reading tests at the upper grade levels.

With respect to concern (3), the early advantage of the two-way group, Thomas and Collier note that this group had the advantage of having two-way bilingual education in pre-kindergarten and kindergarten. Table 5, however, shows that their advantage was present even for those starting in grade 1.

As Thomas and Collier point out, the sample sizes in the longitudinal study are small, and one must be cautious in generalizing from them. Also, as noted earlier, the data is not strictly longitudinal. These concerns are, of course, an invitation for additional research.

Comparison with Developmental Bilingual Education

de Jong (2004) compared a two-way and a developmental bilingual program, focusing on children who stayed in the program continuously from kindergarten to grade 5. The sample size was modest (26 in the two-way and 19 in the developmental program), but results were similar for children in other cohorts not (yet) followed to grade five. On the LAS reading test, two-way children did better than those in the developmental program, but the latter group appeared to be closing the gap. Scores are also quite high, but the test used was designed for English Learners, not native speakers of English. Because of the small sample and the presence of a ceiling effect (all scores very high), table 6 includes results of the LAS writing test as well. Again the two-way children do somewhat better.

8