Purpose: Fingerprint evidence plays an important role in forensic science. Informal observation and some research suggests that experts attain high levels of proficiency in determining whether two fingerprints match; typically this involves matching a partial or distorted image obtained from a crime scene (a latent print) with a print from a database (a tenprint). Like expertise in many other complex classification tasks, such as identifying pathology in radiographic images or classifying of birds by experienced birdwatchers, fingerprint expertise depends heavily on perceptual learning involving discovery and fluent extraction of structural features in varying contexts (Gibson, 1969; Kellman & Garrigan, 2009). Little research has examined what kinds of information are used by expert examiners to achieve advanced performance. Method: We compared the performance of expert examiners, novices, and novices who watched a short training video on a fingerprint matching task. Participants reported whether a latent and tenprint were from same or different sources. They also gave ratings of confidence in their responses and ratings of difficulty for each presented pair. Results: Experts were more accurate and had stronger correlations between difficulty and accuracy ratings than both groups of novices, whose performances were comparable. Approximately twenty image features, including mean local contrast and ridge reliability were extracted for each print pair and used as predictors of accuracy in a regression model. By comparing the multiple regression models for each group, we found that novices who watched the training video shared more predictors with the experts than the other group of novices. Conclusions: Our results suggest that level of expertise in fingerprint matching is correlated with learning to properly combine the right types visual information when making a fingerprint match. These results support the importance of perceptual discovery in attaining expertise and have implications for development of training interventions for fingerprint matching.
Meeting abstract presented at VSS 2013