The Vocabulary Test of the California Cognitive Assessment Battery (CCAB)

Objective. We report the design and metric properties of the vocabulary test of the California Cognitive Assessment Battery (CCAB), which adaptively delivers 24 trials from a 600-word list ordered by difficulty, and describe the demographic variables associated with vocabulary test performance. Method. 2,181 English-speaking adults aged 18-89 completed 24 four-alternative synonym items administered as a 1-up/1-down staircase over 60 difficulty levels, with step sizes that started large (4 levels) and decreased after successive performance reversals. Two scores were compared: the QuickSIN estimate over all trials (the default) and the mean of the last four difficulty levels. Retests were available at 1 day in 100 participants and at 12 months in 307; reliability was assessed with Pearson and intraclass correlations. Twenty-one demographic correlates were screened in the 2,176 participants with questionnaire data, and a seven-term comprehensive regression model was compared with age-education-gender (AEG) adjustment, each with and without race indicators. A dichotomous Rasch model was fitted to the trial-level responses to calibrate the difficulty assignments of the stimulus set. Results. Median administration time was 227 s. QuickSIN and last-four scores agreed closely (r = 0.98) and retested at 0.83 and 0.82 at 1 day and 0.81 and 0.79 at 12 months. Education correlated r = 0.43 with the vocabulary score, weekly reading hours r = 0.39, age r = 0.29, highest household income r = 0.27, and parental education r = 0.26; English as a second language correlated r = -0.19; gender did not (r = 0.003). Vocabulary scores were 0.15 levels higher per year of age across the cross-sectional sample, with higher scores at older ages in every education stratum. The comprehensive model explained 32% of the variance (10-fold cross-validated R^2 = .31; AEG, 17% in the same participants) and attenuated but did not eliminate group differences: adjusted coefficients relative to non-Hispanic White participants were -0.27 SD for Hispanic, -0.38 for non-Hispanic Asian, -0.38 for non-Hispanic "other", and -0.76 for non-Hispanic Black participants, all significant. Word difficulties calibrated separately in each group agreed with the non-Hispanic White calibration at r = 0.96. Rasch calibration gave rho = 0.95 between assigned and estimated difficulty among the 316 well-sampled words, with a median displacement of 2 levels and a within-level spread of 2.9 levels; rescoring on the calibrated scale raised 12-month retest reliability to 0.84. Calibrated difficulty correlated rho = -0.85 with Brysbaert et al. (2019) word prevalence norms. Conclusions. The CCAB vocabulary test provides a reliable four-minute estimate of verbal ability that was higher at older ages in this cross-sectional sample. The assigned difficulty levels were broadly correct, but the spread of difficulty within a level limited the resolution of level-based scores to approximately 2-3 levels.