BackHome
Refresh: OffView Personal Schedule
Individual Submission Summary
Share...

Direct link:

The Computerized Comprehension Task: Cross-Language Evaluation of Psychometric Properties and Prediction to Developmental Outcomes

Fri, March 22, 1:00 to 2:30pm, Hilton Baltimore, Floor: Level 2, Key 2

Integrative Statement

Word learning develops rapidly in the first two years and is a building block for subsequent development. Initially weak word-referent relations are strengthened through the iterative application of domain general learning mechanisms leading to the expectation that early vocabulary should be stable, across languages, with later abilities built on similar mechanisms.

By 24 months of age, parent reported vocabulary predicts later language, literacy and reading, and academic performance in English-, Dutch-, and Danish-speaking children but accounts for small to modest variance and individual prediction is weak. This may reflect a tendency, in some vocabulary assessments, to count both weak and strong word-referent associations as known. Weak associations are fragile and context-bound whereas strong associations are stable and decontextualized. We expect decontextualized associations to predict downstream language and cognition.

We begin with psychometric data on the Computerized Comprehension Task (CCT) measure of early, decontextualized vocabulary in English, French, and Spanish from 12 samples across five countries (see Table 1). We follow with two experiments examining prediction to general language ability at age 3 and kindergarten readiness at age 4.

Experiment 1: Forty-four English and 66 French monolingual children participated at 16, 23, and 36 months of age. In the first two visits, children completed the CCT and parents completed the MCDI. In the third visit, children completed the PPVT, a sentence repetition task, and 15 minutes of free play from which an estimate of MLU was extracted. A general Language Factor accounted for 60.54% of the variance in 36-month scores. CCT scores at 23 months were the most parsimonious predictor of this factor (Akaike Information Criterion) and uniquely identified children at risk for language delay. An ROC analysis yielded good sensitivity and specificity: 85% of children at risk at 36 months had low CCT scores at 23 months and 76% of children who were typically developing had typical scores at 23 months.

Experiment 2: Forty-eight English monolingual, 58 French monolingual, and 34 French-English bilingual children participated at 22 and 48 months of age. At 23 months, children completed the CCT and parents completed the MCDI. At 48 months, children completed the PPVT and the Lollipop Test of School Readiness. Relative weights analyses were conducted with CCT and MCDI as predictors and child sex and maternal education as controls. For both PPVT and kindergarten readiness, the CCT was the only significant predictor (rescaled weights = 69.36 and 14.02 for the CCT and MCDI, respectively). A series of regression analyses assessed mediation of prediction from CCT scores to kindergarten readiness by 48-month vocabulary. Whereas CCT significantly predicted both PPVT and kindergarten readiness scores (β=.49, p=.002, and β=.43, p=.004, respectively), 48-month vocabulary did not significantly predict kindergarten readiness when controlling for demographic variables and 2nd year vocabulary.

Finally, we report preliminary data on the extension of the CCT to large-scale testing (N=3000) in Denmark (in which the psychometric properties of the Danish CCT largely mirror those observed in other languages) and to small-scale testing in an endangered language, Unangam Tunuu, in the Aleutian Islands.

Authors