ASSESSING THE CULTURAL SENSITIVITY OF THE BOSTON NAMING TEST-2: IS THERE EVIDENCE FOR DIFFERENTIAL ITEM FUNCTIONING BETWEEN AFRICAN AMERICAN AND CAUCASIAN ADULTS AND BETWEEN MEN AND WOMEN?

The Boston Naming Test – 2 (BNT-2) is one of the most widely used neuropsychological measures for the detection of naming deficits. However, there are few studies that have investigated its item-level psychometric properties, particularly in ethnic minorities, including African Americans. The present study examined the BNT-2 for the presence of ethnicity and genderbased differential item functioning (DIF) in a cognitively diverse sample of 744 African American and Caucasian adults recruited from a neuropsychology outpatient clinic in Chicago, Illinois. Using a two-parameter Item Response Theory (IRT) framework, all 60 items of the BNT-2 were analyzed, with difficulty and discriminability estimates generated for items 16 through 60 for ethnicity, and for items 11 through 60 for gender. Ethnicity-based DIF was detected for 10 items (i.e., “globe,” “beaver,” “dominoes,” “unicorn,” “accordion,” “latch,” “tripod,” “yoke,” “trellis,” and “palette”) and genderbased DIF for 6 items (i.e., “stethoscope,” “pyramid,” “latch,” “sphinx,” “yoke,” and “palette”), indicating that these items do not function equivalently between groups, despite being matched on naming ability. Post hoc hierarchical logistic regression analyses examined age, years of completed education, and literacy (measured as word reading ability) as potential explanatory factors for the observed DIF. Results indicated that the three variables provided partial explanations for the DIF detected in the 16 ethnicity and gender items, though none could entirely account for the observed findings. IRT-estimated difficulty parameters also allowed for the examination of item ranking on the BNT-2. For both ethnicity and gender, the IRT-estimated rank ordering of items between groups was similar, as well as being comparable (though not identical) to those initially assigned by Kaplan, Goodglass, and Weintraub (2001). Results further indicated that correlations between BNT-2 scale scores and IRT-based scale scores were very high, suggesting that the overall scoring of the BNT-2 was not subtantially biased by ethnicity or gender, despite the identified DIF and small fluctuations in item rankings. Overall, this implies that the BNT-2 is still a valid measure for use in African Americans, although future consideration should be given to revision or omission of those items identified as having DIF.Ph.D. in Psychology, July 201