Evaluation of internal validity using modern test theory: Application to word association

Area of Science:

• Psychology
• Cognitive Science
• Psychometrics

Background:

• Word association tests (WATs) are common tools for assessing semantic memory.
• Few studies have rigorously evaluated the psychometric properties of open-ended cognitive tests like WATs.
• Item response theory (IRT) models offer advanced methods for evaluating test interpretations.

Purpose of the Study:

• To evaluate the internal validity and psychometric properties of a WAT designed to assess casual sex-related associative memory.
• To apply unidimensional IRT models to WAT data from a sample of adult drug offenders.
• To determine the test's informativeness and associations with risky sexual behavior.

Main Methods:

• A WAT was administered to 1,138 adult drug offenders using cues potentially eliciting casual sex-related responses.
• Unidimensional IRT models were fitted to the WAT data.
• Invariance across demographic variables (gender, ethnicity, age) and sex partner profiles was assessed.

Main Results:

• The WAT demonstrated strong evidence for internal validity, with all items measuring a single latent trait.
• The latent trait of casual sex-related associative memory was invariant across demographic groups.
• The WAT was highly informative, particularly at average-to-high levels of the latent trait, and correlated with risky sex behavior.

Conclusions:

• WAT score interpretations are internally valid for assessing casual sex-related associative memory in adult drug offenders.
• The study highlights the utility of WATs in at-risk populations and the benefits of IRT for test evaluation.
• The findings underscore the relevance of associative memory to understanding risky sexual behavior in this demographic.