(Margoliash and Fortune 1992), the memory of TCS neurons was observed
to extend up to approximately 300–500 msec. The long integration periods
and high degree of stimulus specificity place the TCS cells described in
songbirds among the most complex auditory neurons known.
Originally, a simple model of the TCS response in birds was proposed
whereby release from inhibition resulted in the nonlinear component of the
response (Margoliash 1983). Data from intracellular studies found evidence
for that model but also for other interacting subthreshold mechanisms
similar to those described in the bat IC, along with additional threshold
nonlinearities (Lewicki and Konishi 1995). Whereas axonal delay lines may
account for the shortest-time-scale TCS responses in bats (Kuwabara and
Suga 1993), the much longer time scales of the TCS response in songbirds
cannot be accounted for by axonal delays. In songbirds, recent data show
that different subthreshold responses are associated with two distinct populations of projection neurons and a population of interneurons in the songcontrol nucleus HVc (Mooney 2000). TCS responses are thought to be
present for at least some neurons in both classes of projection neurons and
for the interneurons. Thus, as in bats, multiple mechanisms in birds may be
responsible for the TCS response in different classes of neurons. One difference between TCS neurons in CF-FM bats and birds is that, in the latter,
the neurons are predominately found in vocal-control areas—areas that
also participate in generating motor output. This suggests the hypothesis
that a relatively undifferentiated auditory input to song-system nuclei might
be patterned in interaction with central pattern generators for song that are
shaped by song learning (Margoliash et al. 1994). TCS responses are apparently rare in the ascending auditory system of birds, whereas they are
common in the ascending auditory system of bats.
2.2. Input Constraints on Representational Systems
The search for specializations in the auditory system related to vocal behavior has often resulted in identification of complex, nonlinear neuralresponse properties such as the combination-sensitive neurons described
above. However, practical limitations on stimulus power, stimulus and
recording duration, and stationarity, coupled with the high-order nonlinearities central neurons typically exhibit, have handicapped quantitative
approaches to nonlinear analysis (e.g., white-noise analysis). As a result,
generalized procedures to characterize nonlinear neuronal responses have
not been established, and linear techniques to analyze complex, especially
natural, stimuli are being developed (e.g., spike-triggered receptive fields:
Klein et al. 2000; Theunissen et al. 2000; spike-based stimulus reconstruction: Rieke et al. 1996). Information-theoretic approaches can describe nonlinear statistical properties of neurons, but these often require more data
than is practical to collect. Furthermore, the ultimate utility of informationtheoretic descriptions is not yet resolved. From this perspective, the use
340
T.Q. Gentner and D. Margoliash
to extend up to approximately 300–500 msec. The long integration periods
and high degree of stimulus specificity place the TCS cells described in
songbirds among the most complex auditory neurons known.
Originally, a simple model of the TCS response in birds was proposed
whereby release from inhibition resulted in the nonlinear component of the
response (Margoliash 1983). Data from intracellular studies found evidence
for that model but also for other interacting subthreshold mechanisms
similar to those described in the bat IC, along with additional threshold
nonlinearities (Lewicki and Konishi 1995). Whereas axonal delay lines may
account for the shortest-time-scale TCS responses in bats (Kuwabara and
Suga 1993), the much longer time scales of the TCS response in songbirds
cannot be accounted for by axonal delays. In songbirds, recent data show
that different subthreshold responses are associated with two distinct populations of projection neurons and a population of interneurons in the songcontrol nucleus HVc (Mooney 2000). TCS responses are thought to be
present for at least some neurons in both classes of projection neurons and
for the interneurons. Thus, as in bats, multiple mechanisms in birds may be
responsible for the TCS response in different classes of neurons. One difference between TCS neurons in CF-FM bats and birds is that, in the latter,
the neurons are predominately found in vocal-control areas—areas that
also participate in generating motor output. This suggests the hypothesis
that a relatively undifferentiated auditory input to song-system nuclei might
be patterned in interaction with central pattern generators for song that are
shaped by song learning (Margoliash et al. 1994). TCS responses are apparently rare in the ascending auditory system of birds, whereas they are
common in the ascending auditory system of bats.
2.2. Input Constraints on Representational Systems
The search for specializations in the auditory system related to vocal behavior has often resulted in identification of complex, nonlinear neuralresponse properties such as the combination-sensitive neurons described
above. However, practical limitations on stimulus power, stimulus and
recording duration, and stationarity, coupled with the high-order nonlinearities central neurons typically exhibit, have handicapped quantitative
approaches to nonlinear analysis (e.g., white-noise analysis). As a result,
generalized procedures to characterize nonlinear neuronal responses have
not been established, and linear techniques to analyze complex, especially
natural, stimuli are being developed (e.g., spike-triggered receptive fields:
Klein et al. 2000; Theunissen et al. 2000; spike-based stimulus reconstruction: Rieke et al. 1996). Information-theoretic approaches can describe nonlinear statistical properties of neurons, but these often require more data
than is practical to collect. Furthermore, the ultimate utility of informationtheoretic descriptions is not yet resolved. From this perspective, the use
340
T.Q. Gentner and D. Margoliash