Page 50 - 2017Winter

Target-Ranging Theories
Delay Lines for Target Ranging
In Figure 4, A and B, the receiver incorporates a fixed tem- plate of the broadcast that is reversed in time. Because the template is specific to each transmitted signal’s waveform and it is hardwired or programmed into the receiver, changing the broadcast requires making a different receiver for each new broadcast, which might be practical if there is a small number of potential transmitted signals in the repertoire, but it is cumbersome if there are many possible broadcasts that have to be accommodated. The spectrograms in Figure 2 show that the big brown bat changes its broadcasts virtually continuously along the dimension of duration, which means FM sweep rate and frequency. To accommodate these chang- es, the bat must listen to each transmitted sound as it is sent out and create a new template “on the fly.”
Figure 4C illustrates an auditory version of the spectrogram theory that takes each new broadcast into account for deter- mining echo delay (Suga et al., 1990; Simmons et al., 1996; Wiegrebe, 2008). The bat transmits an FM sound and receives it at the inner ear at the moment it is sent out. Reception of the broadcast stimulates auditory receptors tuned to differ- ent frequencies along the sweep in succession, which evoke a single well-synchronized spike in each auditory neuron serv- ing that receptor. Although secondary spikes are likely to oc- cur in auditory nerve fibers when this excitation occurs, the cochlear nucleus must strip them out of subsequent stages of the auditory pathway because the single-spike phasic-on re- sponse pattern dominates responses recorded from all high- er levels if the stimuli are brief FM chirps that mimic broad- casts or echoes. In the big brown bat’s inferior colliculus, response latencies in different neurons spread densely over a range from about 4-5 ms (the minimum required for excit- atory responses to ascend through the auditory brainstem) to about 30-35 ms and then more sparsely to longer latencies of 50 ms or more. Each of these neurons has a characteristic latency as well as a tuned frequency. Its single-spike response registers the corresponding time-frequency sampling point in the FM sweep, but, in addition, it retains it for the length of its latency.
Taken together, the ensemble of inferior colliculus neurons tuned to an individual frequency has a spectrum of laten- cies that resembles a frequency-tuned delay line. Each cell’s particular response latency is created by a combination of inhibitory and excitatory inputs that first prevent a response from occurring until the inhibition has ceased and then trig- gers the response at a latency that is characteristic of that cell.
The spread of inhibitory durations determines the spread of neural delays. Figure 4C illustrates how this system of fre- quency-tuned delay lines responds in succession to the arriv- al of different frequencies in the broadcast and then repeats the initial registration of each time-frequency point over and over from one cell to the next. Across neurons tuned to dif- ferent frequencies, the shape of the FM sweep is preserved as a traveling template analogous to the fixed templates in Figure 4, A and B, but it is updated for each new broadcast.
Registration of Echo Delay by Spectrogram Correlation
In Figure 4C, when an echo is received, it evokes a new pat- tern of single-spike responses in neurons tuned to succes- sive frequencies in its FM sweep. This pattern resembles the pattern already in progress from the original broadcast and moving step-by-step to longer latencies (repeated occur- rences of the same template). Responses to the echo coincide in time with one of the templates composed of ongoing re- sponses to the broadcast. In each frequency channel (spec- trogram delays in Figure 4C), the temporal coincidence of long-latency responses to the broadcast and short-latency responses to the echo is registered by neurons at higher levels of processing, in the medial geniculate and auditory cortex. Each such coincidence evokes a time-locked spike in neurons that receive inputs from neurons that have different characteristic response latencies along one of the delay lines. Coincidence-detecting neurons are “tuned” to the particular echo delay calibrated by the difference in time between the longer latency response to the broadcast and the shorter la- tency response to the echo. In big brown bats, delay-tuned neurons in the auditory cortex cover delays from 2 to 28 ms (Simmons, 2012), which fits well with the span of latencies found along the delay lines in the inferior colliculus. Figure 5A illustrates delay-tuning curves recorded from neurons in the auditory cortex of the big brown bat (Sanderson and Sim- mons, 2003). These delay-tuned cells encode a target range from about 30 cm to 5 m. A separate system of delay-tuned neurons is found in the superior colliculus (Wohlgemuth et al., 2016). It is important to note, however, that the fact of de- lay tuning does not constitute an explanation of a perceived target range because delay tuning is too imprecise to support the bat’s behavioral delay acuity (Pollak, 1980).
Unlike the fixed template designs in Figure 4, A and B, the delay line structure in Figure 4C can adapt to changes in the broadcasts and still determine the delay of echoes. Successive broadcasts can shorten in duration and thus become steeper in FM sweep rate, but the separation of delay computations
48 | Acoustics Today | Winter 2017