How do forensic scientists make fingerprint identifications?

Fingerprints contain remarkable structure. The dynamics of the development of prints in utero dictate that ridges maintain a similar separation. This provides the kind of regularity that could enable perceptual learning processes to develop and improve the extraction of features from prints.

To address this question, we designed an experiment that extracted out what we thought were the essential elements of an identification. We tested both experts and novices. We first created pairs of print fragments:

We then presented one fragment for subjects to view for about 1 second and then showed a mask for either 200 milliseconds or 5 seconds. Then we tested the observer with two choices and asked them to pick the one they saw original. Here is a diagram of the sequence of events:

Fingerprints are often corrupted by visual noise and sometimes only part of a print is visible. To simulate these effects, we sometimes added visual noise or partially masked the fragments at test. Here are examples of these manipulations:

What we found

The basic question we're asking is whether experts perform better than novices, and if so, under what conditions. The bottom line is that experts performed better than novices is all conditions, and perform especially well when full fragments are embedded in noise. We were able to tie this to a particular brain process called configural processing, and we use a brain recording technique called electroencephalography (EEG) to identify this particular mechanism. We relied on a known component of the EEG signal called the N170, and the fact that it is delayed when a face (known to be processed configurally) is inverted. The inversion disrupts the configural process, causing the delay in the N170.

It turns out that fingerprints have an orientation, and that fingerprint experts almost always view fingerprints in an upright orientation. In fact, they will sometimes invert a particularly difficult print in order to gain perspective on the pattern.

To test to see whether configural processing was active when fingerprint experts viewed upright fingerprints, we showed upright and inverted fingerprints to novices and experts. We also showed upright and inverted faces. We expected both groups to show the delayed N170 component when faces were inverted, but this effect should only occur in fingerprint experts for the fingerprints.

Here is the data:

The dashed lines come from faces, and the solid lines from fingerprints. The x-axis is time, with 0 the time that the stimulus (face or fingerprint) comes on. The y-axis is amplitude recorded from the back-right part of the head, which records mainly amplitude from the occipital/parietal region of the brain.

As predicted, the experts had a difference for both faces and fingerprints, but the novices had a difference only for faces in the N170 component.

This result allows us to attribute the configural processing we observed in the behavioral experiment to changes in perceptual processing that occur very early on in the visual stream.

Summary

To summarize, the results of the two experiments provide converging evidence that the visual expertise gained by fingerprint examiners has profound effects on the earliest stages of visual processing, which has lots of potential to help them recover fingerprint detail from noisy or degraded images.

For More Information

For more information about this research, you can download the article from here. To participate in future on-line (web-based) experiments, contact Tom Busey at busey@indiana.edu. Please specify if you have fingerprint experience.