Development and Validation of a New Naming Task: The Southwestern Item Fluency Test (SWIFT)

Tests of confrontation naming play an important role in the clinical assessment of language. Commonly used naming tests have a variety of limitations including ceiling effects, outdated stimuli, differential item functioning across groups, and often ignore speed of word retrieval even though this is a prominent component of word-finding difficulties (WFD). The aim of the current study was to develop and validate a brief, psychometrically sound naming task that is sensitive to WFD. One-hundred items were selected from the Bank of Standardized Stimuli (BOSS) (Brodeur et al., 2014), an open-source collection of high-definition photographs, to create the Southwestern Item Fluency Test (SWIFT). All items were normed on visual characteristics and selected items were required to (1) be easily recognizable, (2) represent a range of semantic categories, (3) be high in object-name agreement, and (4) span a range of difficulty. Phase I (Test Development) -Phase I of the study involved development and analysis of the SWIFT, where it was hypothesized that (1) confrontation naming item difficulty could be predicted by psycholinguistic characteristics, and (2) using the most predictive psycholinguistic characteristics to rank SWIFT items would result in four stimulus pages with similar visual and psycholinguistic characteristics. Psycholinguistic predictors of Age of Acquisition and Lexical Frequency were found to best predict item difficulty based on previously published difficulty indices on the Boston Naming Test (BNT). SWIFT items were ranked by predicted difficulty and divided into 4 groups of 25 items that contained an equal distribution of semantic categories and level of difficulty. ANOVAs comparing the SWIFT pages on visual and psycholinguistic characteristics found no statistical differences, showing strong initial psychometric support for the SWIFT development. As a result, four SWIFT stimulus pages were created, each containing 25 items presented in 5x5 rows. Administration of the SWIFT consists of a timed portion, where participants have 30 seconds per page to name as many objects as quickly as possible, and an untimed portion that ensures each item is administered. Phase II explored the utility of the SWIFT in a clinical sample (N=53). The SWIFT was hypothesized to demonstrate: (1) a single underlying "naming ability" factor, (2) adequate internal consistency (≥.70), (3) convergent validity with other tests of word retrieval, (4) acceptable discriminant classification of individuals with impaired word-finding ability, (5) greater sensitivity to Clinician-Rated WFD (CRWFD) compared to the Boston Naming Test (BNT). Statistical Analyses included (1) Principal Components Analysis (PCA), (2) Cronbach's alpha, (3) Spearman's rho Coefficient Correlations, and (4) Receiver Operating Characteristic (ROC) curve analysis with Area Under the Curve (AUC). Patients had a variety of neurological, general medical, and psychiatric diagnoses. The PCA demonstrated that a single component best explained the SWIFT variance with very high internal consistency (Cronbach's alpha = 0.976). Both SWIFT total scores demonstrated convergent validity with word-retrieval tests, specifically the BNT, and divergent validity with visuospatial tests. It was more closely associated with speeded tasks (WAIS-IV Coding) and tests of executive functions (TMT Part A and B) compared with the BNT. Both SWIFT total scores were successful in classifying individuals with objective impairments on tasks of word-retrieval (AUC = 0.834 - 0.835) and were categorically better than the BNT (AUC = 0.646) at classifying individuals with CRWFD (AUC = 0.707 - 0.724). The SWIFT appears to be a promising new naming tool with good sensitivity to WFD in clinical samples and may prove valuable to clinicians when speed of naming is of particular interest in the assessment of naming abilities.