This paper is about empirical validation of the Abilities for Computer Assisted Learning Questionnaire II (ACALQ II). The inventory is presented to the user when she is first time creating her user profile in the digital learning material evaluation system, eValuator (Nokelainen, 2004a; 2004b). Although we are measuring quite stabile personal features such as social abilities, the user is encouraged to update her profile whenever she feels like it. Our motivation to profile each user who is evaluating learning material is two-fold: First, profiling allows us to control individual differences in order to reduce error in evaluation scores, and second, it enables personalised search functions, for example to seek for or recommend existing evaluations of 'matching' or 'same type' peer users.The ACALQ II is a 42-item five-point Likert-scale self-report instrument for adolescent and adult learners. The instrument contains five parts: 1. motivation (6 factors, 12 items), 2. metacognitive strategies (4 factors, 10 items), 3. social abilities (2 factors, 6 items), 4. serialistic-holistic approach (2 factors, 8 items) and 5. signaling (2 factors, 8 items). The item number per factor is low in the first two parts of the instrument as they are developed on the grounds of the APLQ (Ruohotie, 2000). We analysed theoretical structure of the first three parts of the instrument with two empirical samples (total n=178). The first sample (n=112) represents Finnish university students whose major is computer science or science of education. This adolescent sample includes 18 female and 28 male respondents; gender information is missing in 66 cases. Their age median is 21 years. The data was collected between October 2002 and December 2003. The second sample (n=66) represents Finnish postgraduate adult learners (47 female and 17 male respondents, two respondents' gender information is missing). The age median of the second group is 30 years. The adult data was collected between October 2002 and December 2004. The statistical procedures in this study included (1) variable selection based on descriptive statistics and correlations, (2) comparison of the group means, (3) Bayesian dependence modeling, (4) reliability analysis and (5) confirmatory factor analysis. The results of the nonparametric group difference analysis supported our assumption that the instrument is consistent within the two samples, but is still able to capture variance between them. The results of Bayesian network modeling showed that the structure of the ACALQ II is valid for both groups. Results considering the learning motivation scale showed that in both samples the scale is two-dimensional. The strongest motivational factor across all the samples is MF5 "Self-efficacy". MF1 "Intrinsic Goal Orientation" is shared with first and second sample, and MF6 "Test Anxiety" is shared with first and third sample. The analysis shows that the weakest dimensions are MF2 "Extrinsic Goal Orientation" and MF3 "Meaningfulness of Study". Results considering the learning strategy scale show for both samples only low dependencies between the items. LS1 "Metacognition in Learning" (i.e. "Planning") is the most important factor in the adolescent sample. The adult learner's network indicates that there is no difference in conceptual level between metacognition in theory and practice. The results show that the social ability scale is two-dimensional ("Interpersonal" and "Intrapersonal") in both samples. Results of the CFA showed that all the optimised solutions surpass the baseline model by both comparative fit index and Tucker-Lewis coefficient.