Eric Brighton and Brock Kirwin, Department of Psychology
Introduction
Visual pattern separation is the ability of an individual to correctly distinguish between images that they have already seen, images that are very similar but not identical, and novel images. Over the last eight years visual pattern separation has come to be well accepted as a measure of an individual’s ability to form long-term memories. Studies have been conducted to see the effects of aging, exercise, depression and anxiety among other things on their ability to do the pattern separation task (Shelton et. al 2013) (Holden et. al 2013). This has been a very useful metric so far, but by itself has failed to account for many of the expected outcomes. For example, despite the commonly held understanding that anxiety affects the ability of an individual to form long-term memories, in a recent study conducted in Dr. Kirwan’s lab, subjects who have been diagnosed with anxiety scored similarly to a control group of healthy individuals in the visual pattern separation test. An aural test could provide a second, crucial way to understand the process of memory formation and to see how the process is affected in different populations. Other studies have shown differences in how the brain handles visual and audio stimuli (Conway et. al 2009). As of yet the pattern separation task has never been done in the aural domain.
Methodology
The subjects were first be presented with a series of 60, 5-second, 3-second, or 2-second audio clips1 followed by a two second pause during which time they were asked to classify each as typically indoor or outdoor sounds to ensure active listening. After completing that task they were given instructions to listen to another series of sounds. This time subjects were asked to classify the sound as NEW (never been heard before), OLD (exact match to a sound heard already), or SIMILAR (sound closely resembling a clip already heard). Subjects were then be presented with 60, 5-second, 3-second, or 2-second clips followed by 2 seconds to answer OLD, NEW or SIMILAR. Their responses were recorded.
Results
Figure 1 shows the baseline visual pattern separation for comparison. Figure 2 shows the average proportion of answers given during the second run for hits (exact matches correctly identified), lure correct rejections (LCR, similar sounds identified as similar), and lure false alarms (LFA, similar sounds incorrectly identified as the same).
Discussion
Discussion For the first round of testing we had only 5 second sound clips. As seen in figure 2, the results were not comparable to the visual pattern separation task, with participants doing only slightly better than chance between LCR and LFA. We hypothesized that the
length of the clip affected performance and so refactored the experiment so that some participants would hear 2 second clips, others would hear 3 second clips, and others 5 second clips. The only significant difference between the trial lengths was between the 2 second and 3 second lure false alarm types with 3 second clips performing slightly better (fewer LFAs). Besides that, trial length had no effect on the accuracy of the participant. Participant feedback indicated that many of the sound clips that were intended to be distinct could be interpreted as similar, leading to a higher rates of “similar” responses and overall ambiguity about the experiment.
Conclusion
While this experiment, thus far, has been unable to match the baseline level of accuracy from the visual pattern separation task, it has allowed us to develop a working task that we can continue to improve upon. Our next step is to regenerate a more extensive sound clip bank. We will then do a rating test where participants will listen to two similar sounds and rate them from 1 to 5 in similarity. In our subsequent tests we will then only use those with a rating of 1 or 5 (most and least similar) to help remove ambiguity. This, combined with a more rigorous categorical sound clip screening process we help to boost the accuracy levels of the auditory pattern recognition task.
