Do some school-age children have no language? Some problems of construct validity in the pre-LAS Espanol

Bilingual Research Journal, Summer 2002 by MacSwan, Jeff, Rolstad, Kellie, Glass, Gene V

Analysis of the Pre-LAS Espanol from External Evidence

In this section we explore the relationship between the subparts of the Pre-LAS Espanol and the body of empirical and theoretical work devoted to the study of language structure and acquisition. We will pay especially close attention to those sections of the test most responsible for the "non-" or "limited" category, namely, Parts 5 and 6.

As mentioned, the Pre-LAS Espanol (Duncan & DeAvila, 1986a, 1986b) consists of six parts. Part I is a receptive test of children's ability to understand commands in Spanish; it has 10 items. In Part 2, the examiner asks the child to point to 10 specific objects inside a picture of a house. In Part 3 of the test, the child is asked to identity one of two pictures that corresponds to a spoken utterance given by the examiner (e.g., "Hay dos animales, " "There are two animals"). Subscales 1-3 of the Pre-LAS Espanol, in which each test item is scored as correct or incorrect, may reasonably be said to identify Spanish language proficiency.

In Part 4 of the test, the examiner reads 10 sentences that the child must repeat. An item is scored as incorrect if the underlined portion of the item, which is intended to target a particular grammatical construction, is omitted or "transformed." Psycholinguists who use elicited imitation as a research tool caution that the pre-training sentences in such tasks should be of the same approximate length as the test sentences, that the test battery should begin only after the examinees show that they understand the task, that all sentences in the test battery be of nearly identical length, and that a whole test battery target a single grammatical structure while differing only in word choice (Lust, Flynn, & Foley, 1996). However, the Pre-LAS Espanol follows none of these cautions; thus, we do not know if the child's incorrect responses result from failing to understand the (pragmatically anomalous) task, unexpected processing load, lexical bias, or perhaps limitations on short-term memory (Smith & van Kleeck, 1986). Hence, while it is very likely that children who do well on this part of the test are Spanish proficient, we do not know whether children who perform poorly on this section know Spanish or not. Thus, by virtue of the design of the task, we might expect that Part 4 would yield some number of false negatives.

As was shown in Table 4, Parts 5 and 6 account for an extremely large proportion of the variance of results, with Part 5 correlating .744 and Part 6 correlating .836 with the total score (.577 and .768, respectively, with the overlapping parts removed). Because of the importance of these sections in determining the overall score, very careful attention should be given to their design and underlying theoretical assumptions.

Part 5 of the test requires the child to finish five incomplete sentences. Like Part 4, this task invokes a linguistically unnatural situation that will be strange to many children. More importantly, however, the examiner has a great deal of discretion in awarding a score from zero to three for each item. Table 8 shows the scoring rubric for Part 5 of the Pre-LAS Espanol.