Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)

First Advisor

Paul Jarley


This study examined the personnel selection technique of biographical information (biodata) in terms of theory, criterion-related validity, and adverse impact. First, the construct validity of biodata was examined to determine if biodata theory was useful in explaining biodata's strong criterion validity. Items from an existing biodata inventory were mapped onto construct domains drawn from Mumford, Stokes, and Owens' (1990) ecology model. Relationships between subjects' biodata responses and training performance was examined for consistency with the model's predictions in an organizational sample. The ecology model did not fit the data well. Follow up exploratory analyses did yield good fit when the model was extended by grouping construct domains within developmental time periods. Second, biodata was examined in terms of simple and incremental criterion-related validity relative to a general cognitive ability test. The biodata instrument was also investigated in terms of incremental criterion validity of biodata predictor scales used in combination with a general cognitive ability, or "g," test. Predictor scales consisted of all biodata response options, "g-loaded" response options, and "non-g-loaded" response options, respectively. The biodata scale (including all biodata items) outperformed the general cognitive ability test both individually and incrementally (both before and after correcting for the effect of range restriction due to selection on g). The biodata g and non-g item sub-scales slightly outperformed the test of general cognitive ability. Finally, biodata adverse impact was assessed in two ways. First, individual biodata response options were examined for possible adverse impact. Second, separate biodata scales including and excluding adverse impact response options and a test of general cognitive ability were compared in terms of adverse impact. Eliminating response options that violated the four-fifths rule resulted in a relatively large decline in the standardized mean difference between subgroups, no appreciable decrease in biodata criterion-related validity, and minimal adverse impact relative to both the biodata scale containing all response options and the general cognitive ability measure. Research findings are discussed and implications for theory, future research, and practice are offered.