Figure 2: The CDNs integrate multimodal inputs over longer timescales as mating progresses.

(a) Termination in response to brief, strong CDN stimulation does not increase as mating progresses.

(b) Left: flies are given two strong 500-millisecond CDN stimulation pulses. Right: probability of mating termination in response to independent (top) or integrated (bottom) pulses.

(c) The response to two CDN stimulation pulses is greater than expected if each pulse acted independently (grey line = 2p-p², estimated using the single pulse response (p)). Pulses are integrated over longer timescales later in mating.

(d) Male flies become progressively more likely to stop mating in response to a 350-millisecond wind gust (Extended Data Figure 2a–c; Video 4).

(e) Silencing the CDNs with tetanus toxin (Tnt) prevents wind-induced mating termination (Video 5).

(f) Paired wind gusts (650 milliseconds at 10 minutes, 250 milliseconds at 15 minutes) are integrated over a longer timescale at 15 minutes into mating.

(g) Six-second optogenetic grooming neuron stimulation causes CDN-dependent termination with increasing propensity as mating progresses.

(h) Two pulses of grooming neuron stimulation separated by 5 seconds are integrated at 15, but not 10, minutes into mating.

(i) CDN silencing during only the first of two optogenetic grooming pulses induces the same level of termination as if only one pulse (no CDN inhibition) was delivered.

(j) Response of a model system to input pulses of varying duration. For pulses longer than the integration window, information accumulates faster and peaks higher with a greater $\tau$.

(k) Fitting behavioral data to the linear model (left) reveals a higher time constant of integration, $\tau$, at 15 minutes into mating compared to 10 minutes (right). Parameter error bars represent one standard error of the parameter fit (Supplementary Note 2, Extended Data Figures 3,4). Gray shaded regions represent model pointwise 95% coverage intervals.