Page 51 - 2017Winter

Figure 5. Comparison of neural delay tuning with behavioral tests of echo delay acuity in big brown bats. A: Delay-tuning curves measured in the auditory cortex of big brown bats. B: Performance (in percent errors) in two-choice delay discrimination tests with electroni- cally generated echoes (bat “virtual reality”; top), and performance in tests of delay-jitter detection with alternating electronic echoes in the same phase (0°/0°; middle) or inverting phases (0°/180°; bottom). Neural delay-tuning sharpness is measured in milliseconds or tens of milliseconds, whereas behavioral discrimination is measured in tens of microsec- onds and delay-jitter detection involves changes of only a few microseconds.
of two flat targets, one to its left and the other to its right (Simmons, 1973). The bat’s task was to move toward the nearer of the two targets. The bats were able to perceive changes in distance of about 1-2 cm. When the physical targets were re- placed by an electronic echo-generating system, a “virtual reality” version of the same task, big brown bats could perceive changes of about 50-70 μs. Figure 5B, top, shows the mean performance of two big brown bats discriminating differences in echo delay during such a virtual reality experiment (Simmons, 1973). The abil- ity to discriminate changes of about 1 cm or 60 μs has been found to be ubiquitous across different species of bats (Simmons and Grinnell, 1988).
Two-choice range or delay discrimina- tion experiments involve the bat scanning with the aim of its head and broadcast- ing to the left and right during each trial. These movements introduced changes in the apparent distance to the targets of up
into parallel frequency channels before registering the final overall delay allows the system to accommodate changes in the broadcasts. This process is spectrogram correlation, the first stage in a two-stage computational model of target rang- ing and target shape perception (Simmons et al., 1996, 2017). The performance of auditory spectrogram correlation has been evaluated along with the performance of a matched- filter receiver in a large simulation (Sanderson et al., 2003). The process of spectrogram correlation achieved the same performance levels as a matched-filter receiver across a wide span of echo signal-to-noise ratios.
Perception of Target Range
At this last juncture, the critical question is whether echo- locating bats can perceive a target range using the delay of echoes as the principal acoustic cue. The regular decrease in broadcast duration with declining distance to the target (Figure 2) is very suggestive, but what is needed is psycho- physical evidence that the distance to a target and the de- lay of echoes are equivalent auditory dimensions. The first demonstration of target range perception was a series of discrimination experiments in which bats of several species (including big brown bats) were trained to sit on an elevated platform and broadcast echolocation sounds in the direction
to about 1 cm or the delay of the echoes up to about 60 ms. Because the bat’s discrimination thresholds are of similar size, the bat might be able to perceive small changes in delay that these experiments would not be able to determine.
A new procedure trained bats to detect small changes in de- lay that alternate or jitter from one broadcast-echo pair to the next. In the jitter detection task, big brown bats easily determined that the delay was changing by about 1 ms, and under some conditions, they could detect changes of a small fraction of a microsecond (Simmons et al., 1995, 1996). This ultimate sensitivity roughly matches what can be achieved by a matched filter using the same signals at the same signal-to- noise ratios (Sanderson et al., 2003).
Figure 5B shows the mean performance of four big brown bats detecting jitter in echo delay accompanied by no phase jitter (0°/0°) or by 180° phase jitter (0°/180°). These results are compared with delay tuning in neurons of the big brown bat’s auditory cortex (Figure 5A). The discrepancy between delay tuning at the neural level and delay acuity in behavioral tests is striking. It is barely plausible that the ~60-μs accuracy obtained in two-choice tests can be explained by neural tun- ing, whereas delay acuity and phase sensitivity obtained in jittered-echo tests is well beyond plausible accounting. The
Winter 2017 | Acoustics Today | 49