Article Figures & Data
Figures
- Figure 1.
ELL Pyramidal cell responses to conspecific signals. A, Raster plots of chirp responses on a high-frequency (100 Hz) beat. AM of the EOD stimulus is shown in black; three representative OFF cells are shown in cyan boxes, while three ON cells responses are displayed in red boxes. The same six neurons are used for all panels. B, Responses to chirps on a low-frequency (10 Hz) beat. For a detailed description of all chirps used, see Table 1. The shaded area highlights the duration of the chirp. See Extended Data Figures 1-1 and 1-2 for comments and analysis of specific response properties (adaptation time and biphasic responses to low frequencies).
- Figure 2.
Detection efficiency depends on beat frequency. A, Detection of chirps on 10 Hz beat by ON cells as a factor of neurons included in the analysis (n = 17). Error probability is the probability of an ideal observer to correctly assign a spike train as a chirp or beat response. Detection error levels for individual chirp identities are shown in gray. Red line indicates mean detection error for all chirps. ON cells can reliably detect the occurrence of all chirps. B, ON-cell performance is worse on 100 Hz beats. C, D, Mean OFF-cell performance (cyan, n = 16) is also more efficient on a 10 Hz beat.
- Figure 3.
Discrimination of chirps on high-frequency beats is poor. A, Discrimination for small chirps (chirps 5, 6, 7, 8, right) and big chirps (chirps 1, 2, 3, 4, 9, 10, left) on 10 Hz beat. Mean ON-cell chirp discrimination shown in red, OFF cells are in cyan, and discrimination for individual chirp pairs are shown in gray. B, Discrimination of small chirps on a 100 Hz beat is relatively inefficient for both ON (red) and OFF cells (cyan). Discrimination of big chirps varies with chirp identity, but is still poor. For chirp-by-chirp discrimination comparisons, see Extended Data Figure 3-1.
- Figure 4.
Temporal coding properties of pyramidal cells. A, Mean firing in response to SAM stimuli. Mean firing rate (±SE) for both ON (red, n = 19) and OFF (cyan, n = 15) cells in both stimulation configurations (global: pink and pale blue; local: red and cyan) peaks at 5 Hz. B, Phase locking to AM sinusoids is also best at low frequencies. Maximum phase locking is seen at 15 Hz, with the exception of locally stimulated OFF cells, which peak at 5 Hz. C, Mean coherence to noise stimulation is also low pass. ON-cell coherence is shown in red, OFF-cell in cyan. The upper-bound coherence measure (solid line) is based on the response–response correlations between multiple presentations of the stimulus, while the lower bound (dashed line) is based on stimulus–response correlations. Shaded areas indicate SE; darker shading indicates local stimulation. Mean global lower-bound maximums: ON cells, 23.1 Hz (±0.86 SE); OFF cells, 11.6 Hz (±0.75 SE; Wilcoxon rank-sum test, p = 0.04). D, Coding of low-frequency envelopes is poor. Mean (±SE) lower-bound coherence between responses and the low-frequency (0–20 Hz) envelope of a bandpass RAM stimulus (40–60 Hz) are displayed for both local and global stimulus configurations. Both ON and OFF cells exhibit peak envelope tuning at 10.10 Hz (±1.52 SE; Wilcoxon rank-sum test, p = 0.21). OFF cells have noticeably lower coherence to low envelopes than ON cells.
- Figure 5.
Coding of chirps by a small population of higher-pass neurons. A, The mean instantaneous firing rates of ON cells over the time course of chirp and beat stimuli. The difference in peak firing rate(FR) characterize a small population (n = 5, red) that bursts in response to A. leptorhynchus small chirps (beat responses: dashed lines; peak FR, 77.67 ± 7.11 Hz; chirp responses: solid lines; peak FR, 159.57 ± 8.68 Hz; Wilcoxon rank-sum test (p < 0.001). The majority of the ON-cell population (n = 16, black) only showed a modest increase in firing (beat response: peak FR, 62.68 ±1.27 Hz; chirp response: peak FR, 66.56 ± 4.21 Hz; Wilcoxon rank-sum test, p = 0.002). Shaded area represents SE. B, Even the bursty population fires less in response to A. albifrons smallest chirps compared with the beat (beat response: peak FR, 155.45 ± 8.82 Hz; chirp response: peak FR, 117.75 ± 9.40 Hz; Wilcoxon rank-sum test, p = 0.04). C, Mean (±SE) upper-bound and lower-bound coherences for the bursting (red) and nonbursting (black) populations. D, Example raster plots of chirp responses used for A and B. The nonbursty population (black box) responds similarly to both A. leptorhynchus chirp and beat. The bursty population (red box) bursts to A. leptorhynchus chirps, but not A. albifrons chirps. For comparable data obtained from A. leptorhynchus, see Extended Data Figure 5-1.
- Figure 6.
Coding of AM by bursts. A, Sample of noise stimulus (0–60 Hz, gray) and representative spike train (black) from an OFF cell during local stimulation. B, Example of ISI distribution used to determine burst threshold (dashed line). C, ON cells (red) burst more than OFF cells (blue; ANOVA, p = 0.02), and local stimulation produced more bursting than global stimulation (ANOVA, p = 0.01). Error bars show the SE. D, Mean burst-triggered averages (red/blue) and single spike-triggered averages (black) from ON and OFF cells show that bursts are triggered by wider (lower-frequency) stimulus features than single spikes. E, Feature detection performance for burst and isolated spikes. In both ON- and OFF-cell bursts (blue/red-filled bars) tend to have lower error rates (percentage of events signaling false positives or false negatives) in detecting optimal stimulus features than single spikes (gray-filled bars), but this trend is not significant (ANOVA, p = 0.10).
- Figure 7.
Chirping behavior in freely swimming pairs (n = 28). A, The number of chirps produced during interaction is correlated with difference in EODf (r 2 = 0.1093). B, Chirping does not differ by absolute EODf, EODf pair type, or relative EODf. The mean chirp rate for low-frequency fish (<1100 Hz) and high-frequency fish (>1100) was similar (Student’s t test, p = 0.86). Mean chirps per trial based on the EODf of pairing [Low:Low (L:L); Low:High (L:H); High:High (H:H); ANOVA, p = 0.47] and by the relative frequency of individuals within the pairing (Student’s t test, p = 0.55) were all extremely low and similar in all groupings. Error bars indicate SD. C, Interchirp intervals between pairs binned by time fall show no echoing chirp exchanges. Inset, Enlarged section shows that very few chirps occur within 2 s of each other. D, Interchirp interval distribution for individual fish follow a Poisson distribution (R 2 = 0.9).
Tables
Chirp ID Frequency rise (Hz) Duration (ms) Other 1 200 100 α Shape 2 200 200 α Shape 3 350 100 α Shape 4 350 200 α Shape 5 50 50 α Shape 6 100 50 α Shape 7 50 50 Antiphase to 5 8 100 50 Antiphase to 6 9 350 200 Two frequency peaks 10 350 200 Ramp Shaped 11 60 10 A. leptorhynchus small chirp 12 122 15 A. leptorhynchus small chirp “Other” indicates differences in chirp shape not due to peak frequency or duration, such as the shape of frequency rise and fall, and the phase of the beat on which the chirp occurred.
Figure 1-1
Cancellation observed in an LS ON cell. In some recordings, a biphasic response to stimuli is observed. While this could indicate an overly strong stimulus, it could also indicate cancellation via feedback from the midbrain (Bastian et al., 2004). Cancellation is less prominent in the LS of A. leptorhynchus, but still occurs. The biphasic responses to global SAM stimuli but not to local stimulation indicate that our observed responses likely represent an increased cancellation response in the LS as a result of low-frequency tuning. Download Figure 1-1, EPS file.
Figure 1-2
A. albifrons adaptation to step stimulation. A, B, The responses of both ON and OFF cells decreased throughout the 100 ms stimulus without plateauing (A), in sharp contrast with B. A. leptorhynchus responses to global stimuli (S2; Krahe et al., 2008, their Fig. 8C,F) that adapt within 40–50 ms and then plateau. Download Figure 1-2, EPS file.
Figure 3-1
Detailed discrimination traces indicating the results of testing specific chirps against each other. For ease of analysis, chirps were grouped by duration frequency and duration as either big or small. Duration appears to be the feature that is most discriminable, so big and small chirps were not directly compared. Descriptions of chirp properties are located in Table 1. A, Chirp discrimination on a 10 Hz beat. B, Chirp discrimination on a 100 Hz beat. Download Figure 3-1, EPS file.
Figure 5-1
A. leptorhynchus data for stimulation protocols matched to those used in the main text. These data are comparable to what has already been extensively published in the study by Krahe et al. (2008), and thus not highlighted in the main text, but indicate that our methods reproduce previously published results and indicate differences between A. albifrons and A. leptorhynchus. A, Coherence to 0–60 Hz RAM stimulation. Mean coherence to noise stimulation by neuron type. ON-cell coherence is shown in red, OFF-cell coherence in cyan. The upper-bound coherence measure is shown with solid lines, the lower-bound coherence in dotted lines. B, Phase locking to sinusoids across 0–60 Hz is best at 20–30 Hz, while the firing rate is relatively constant across frequencies. C, Coherence to low-frequency envelopes. Both ON and OFF cells perform well at coding low-frequency stimulus components. Download Figure 5-1, EPS file.
In this issue
