Short test of semantic and phonological fluency: normal performance, validity and test-retest reliability

Tests of verbal fluency, whether of Semantic Fluency (SF) or Phonological Fluency (PF), are often used as a standard method of neuropsychological assessment. However, very little normative data, standardized on a UK sample, is available, and little is known about the utility of popularly used 'short-forms' of these tasks. Additionally, very little is known about verbal fluency test-retest reliability. In this report we describe the performance of N = 365 normal participants on a version of SF requiring the generation of exemplars of the category 'animals' and a version of the PF task requiring participants to generate words beginning with the letter 'B'. From this data we have derived a percentile distribution for both tasks. We also addressed the impact of sex, age, years of education and IQ upon both SF and PF performance. No sex differences were found on either the PF or the SF tasks. A significant but small correlation between age and SF, but not PF, was observed. Modest correlations between both years of education and IQ and PF and SF performance were also seen. We also report test-retest reliability scores for performance on both SF and PF tasks. Performance on the longer 'FAS' version was found to correlate highly with scores obtained using just the letter 'B'. This suggests that little additional advantage obtains from administering three-letter versions of PF. An extremely high degree of correlation between SF tasks in which participants are given 1 minute and scores obtained when participants are given 1 minute 30 seconds was also observed. Finally, in order to assist users in deciding whether a changed retest score is due to error measurement or a real effect, we calculated Standard Error of Prediction (SEP) scores.