Figure 3. Responses of walking flies to dynamic odor stimuli.

(A) Upwind velocity (left, top traces; average$\pm$SEM) of different groups of flies responding to a 10 s pulse of ACV at dilutions of 0.01% (n = 13 flies, 147 trials), 0.1% (n = 19 flies, 304 trials), 1% (n = 18 flies, 302 trials) and 10% (n = 75 flies, 1306 trials). Left-bottom traces show PID measurements using ethanol (max concentration 10%), normalized to maximal amplitude. Right inset: mean upwind velocity during odor (2 to 3 s) as a function of odor concentration (black; mean$\pm$SEM), and fitted Hill function (green; green dot: ${\kappa }_d$=0.072%). (B) Turn probability calculated from the same data. Right inset black traces: mean turn probability after odor (11 to 13 s). ${\kappa }_d$=0.127% for fitted Hill function (green). (C) Upwind velocity (average$\pm$SEM) in response to stimuli with off-ramps of 2.5 (n = 38 flies, 528 trials), 5 (n = 38 flies, 567 trials) and 10 (n = 35 flies, 557 trials) seconds duration. Bottom traces: PID signals of the same stimuli using ethanol. (D) Same as C, showing turn probability from the same data sets. White arrows in C and D show elevated upwind velocity and turn probability that co-occur during a slow off-ramp. Black arrow in D: peak turn probability response at the foot of the off-ramp. (E) Upwind velocity (mean$\pm$SEM; n = 31 flies, 346 trials) in response to an ascending frequency sweep stimulus. Bottom trace: PID signal of the stimulus, measured using ethanol. Right inset: average ($\pm$SEM) modulation of upwind velocity as a function of frequency in each stimulus cycle (see Materials and methods). (F) Same as E for turn probability calculated from the same data. Right inset: modulation of turn probability as a function of frequency. (G) Equivalent to E, showing responses to a descending frequency sweep (n = 33 flies, 345 trials). In the inset, the first high-frequency cycle was left out of the analysis. (H) Same as G for turn probability calculated from the same data. (I) Equivalent to G, showing responses to a simulated ‘plume walk’ (n = 30 flies, 393 trials). (J) Same as I for turn probability calculated from the same data.

Figure 3—figure supplement 1. Data processing methods.

(A–C) Segmentation of data into moving and non-moving epochs for analysis. (A) Distribution of ground speed values for all flies during trials with a 10 s odor pulse (n = 75 flies, 1306 trials; data from Figure 1). Y axis on a logarithmic scale. Note large peak close to 0 mm/s corresponding to non-moving epochs. (B) Probability of moving at greater than 1 mm/s increases during odor and remains elevated for tens of seconds after odor offset. PID measurement (top trace) and probability of movement (bottom trace) during a 10 s odor pulse (mean$\pm$SEM; n = 75 flies, 1306 trials; data from Figure 1). Thus, if non-moving periods are not omitted from computation of movement parameters such as ground speed and angular velocity, the means of these parameters are heavily influenced by the fraction of non-moving flies (i.e. the number of zeros) in each epoch. (C–D) Effects of low-pass filtering on estimates of behavioral responses to fluctuating stimuli. (C) Upwind velocity (top) and ground speed (middle) of flies in response to an ascending frequency sweep stimulus (mean$\pm$SEM; n = 31 flies, 346 trials; data from Figure 3E). Blue traces: data as it was used in Figure 3. Red traces: data processed exactly as the blue traces, except we omitted the low-pass filtering at 2.5 Hz. Note that the difference between the two sets is small and mostly shows as increased high-frequency noise in the periods before the stimulus. Bottom black trace: stimulus. (D) Same as C, showing turn probability (top) and curvature (middle) in response to the same stimulus. (E–F) Reliable modulation of behavior at high frequencies can be observed in response to valve clicks. (E) Mean ground speed (n = 31 flies, 248 trials) in response to a random train of valve clicks with a 50% probability of occurrence. Vertical gray lines: time at which the odor valves opened or closed, producing a click sound and slight vibration. Note that flies slowed their ground speed after every click. (F) Modulation of ground speed during random valve clicks (black trace; mean$\pm$SEM (absolute values); n = 31 flies, 248 trials; data in E) and during every cycle of an ascending frequency sweep stimulus (green trace; mean$\pm$SEM; n = 31 flies, 346 trials; data and analysis in Figure 3E, inset). Frequency of valve clicks ranged from 0.18 to 2 Hz and was calculated as one over the inter-click interval (responses to the first click were ignored).