Behavioural response strengths. The mean strengths of responses ± s.d. of cockroaches to optomotor stimuli at different temporal frequencies (A–D) and angular periods (E–H) of the rotation stimulus under (A,E) 500 lx, (B,F) 5 lx, (C,G) 0.05 lx, (D,H) 0.005 lx. Solid bars denote the response strength during stationary controls, and hatched bars represent that during rotating stimuli. The expected control level is 0; +1 indicates the strongest positive and −1 the strongest anti-directional response. *, **, *** indicate significance between the control and rotation distributions at confidence levels of 0.05, 0.01 and 0.001, respectively (paired sample Wilcoxon signed rank test). (A–D) The responses to rotation of the stimulus with a 60 deg angular period at different temporal frequencies were attenuated and their frequency band narrowed from the brightest (500 lx) to the dimmest (0.005 lx) stimuli. (E–H) The responses to different angular periods of the stimuli were compiled from another data set by pooling all responses to temporal frequencies between 0.4 and 10 Hz (E–G) or 0.4 and 4 Hz (H), the frequencies most likely to elicit the optomotor response according to results in A–D. (E) All angular periods were able to elicit the optomotor response at 500 lx illumination. Responses to angular periods of 40–180 deg remained significant down to 0.05 lx (F,G). The lowest light intensity at which significant responses were seen was 0.005 lx, although 60 deg stimuli did not elicit a significant response (H). Sample sizes were N=20 animals, n=40 measurements (A); N=24, n=78 (B); N=23, n=66 (C); N=23, n=66 (D); and N=8–12, n=48–174 (E–H). See also Fig. 3.

The average strengths of behavioural responses ± s.d. of cockroaches to optomotor stimuli at 0.0005 lx. See Fig. 2 for key to the symbols. (A) The responses to rotation of the stimulus with a 60 deg angular period at different temporal frequencies. Both control and rotation values were scattered around 0, and the only significant difference at 4 Hz was a confounding factor of the small sample size. (B) Combined responses from temporal frequencies between 0.4 and 4 Hz at different angular periods of the stimulus. No significant differences between control and rotation values were found. Apparently, the absolute dim-light vision threshold of the American cockroach lies somewhere between 0.005 and 0.0005 lx. Sample sizes were N=11 animals, n=22 measurements (A); N=9–12, n=27–132 (B).

Examples of light-on responses and graded responses to light modulation during rotation (30–60 s) in one photoreceptor cell. Stimulus frequency is indicated above each 500 lx (A) and 5 lx (B) recording. (A) Light-on responses adapt to plateau near resting potential at approximately 800 ms (inset image in the 0.4 Hz recording). Stimulus rotation caused a membrane voltage modulation. In 2.4 and 12 Hz recordings, a slight repolarization marks the onset of rotation (inset image at 12 Hz), because this cell faced white during the control. (B) The 5 lx recordings show small, <10 mV light-on responses (left inset image at 0.4 Hz). The modulation response is overlaid with photon shot noise (inset images at 0.4 and 2.4 Hz). In the 12 Hz recording, the modulation begins to be lost under the noise, but behavioural results show that an optomotor response at this frequency can still be possible.

Examples of photoreceptor responses near the behavioural threshold. Quantum bumps at 0.05 lx (A–C) and 0.005 lx (D,E). Grey-and-white background denotes the blacks and whites of the rotating grating passing the visual field of the photoreceptor. (A) A recording with the 0.4 Hz stimulus shows bumps during control (0–30 s) and rotation (30–60 s). (B) A 10 s subset of A, with red asterisks denoting identified bumps, and template bumps produced beneath. Cumulated bump sums during successive 2500 ms stimulus cycles are shown on the right (12 cycles per 30 s); bump rate is elevated during the latter 1250 ms when the cell faces white. (C) A 3 s excerpt of a recording with a 4 Hz stimulus. The right-hand panel shows stronger bump synchronization with white than that in B. (D) A recording at 0.005 lx showing characteristically few bumps. The inset image shows the typical bump waveform. (E) A 3 s example with a 4 Hz stimulus (left) contains only one bump. A scarcity of photons leads to similar bump sums for black and white (right). Significant optomotor responses were still measured under these conditions.