Figure 1. Intermittent odor plume stimuli and olfactometer design.

(A) Graphical illustration of the intermittency measure. Intermittency (I) is the fraction of time an odorant concentration is above a threshold (0.1*C₀, where C₀ refers to the time-averaged source concentration). In a turbulent plume I drops as a function of distance. Hence, upstream (near the odor source) I tends to be large (here I=0.6) compared to downstream, distant from the odor source (here I=0.09). A steady signal has a high intermittency, and a sporadic signal has a low intermittency. (B) Odor delivery system used to deliver methyl valerate and 2-heptanone. Two counterbalanced proportional valves maintained constant flow rate. (C) Example of odor concentration (red) and flow rate (black) on a single trial. (D) Cross-correlation between photoionization detector (PID) measurement (odor concentration) and the command voltage driving movement of the odor proportional valve. Maximum correlation coefficient is 0.872 ± 0.119 at a lag of 160 ms (n=8643 trials). (E) Example correlation between the trial intermittency value measured from the PID reading vs the intermittency value measured from the voltage command for one session (n=64 trials). Linear regression: y=1.09x+0.023, r²=0.996, p<0.0001. (F) Example traces of odor concentration at gain 1 (darker colors) and gain 0.5 (lighter colors) for naturalistic, binary naturalistic, and square-wave stimuli. (G) Median r² of the correlation between voltage intermittency and PID intermittency for sessions of naturalistic (red), binary naturalistic (orange), and square-wave (blue) stimuli (n=48 sessions per stimulus type, naturalistic median = 0.945 interquartile range [IQR]=[0.937–0.949], binary naturalistic median = 0.998 IQR=[0.997–0.999], square-wave median = 0.987 IQR=[0.982–0.991]).

Figure 1—figure supplement 1. Additional information on intermittent odor plume stimuli and intermittency calculation.

(A) Odor intermittency as a function of distance downstream from odor source. Three odor concentration thresholds used to calculate intermittency are represented. Black line displays an exponential decrease of intermittency, based on the 0.1*C₀ threshold (C₀ is source concentration) used in this paper. The surrounding lines show similar exponential relationship when thresholds were three times lower (red dashed, bottom), or three times higher (red solid, top), but with an expected downstream shift. X-axis units are pixel number downstream from the upstream release location and can be interpreted as release distance of arbitrary unit (AU, i.e. scale-free), but were based on a flow chamber of roughly 1 m downstream size. (B) Example trial of voltage command and resulting photoionization detector (PID) reading. (C) Median slope of the correlation between voltage intermittency and PID intermittency for sessions of naturalistic (red), binary naturalistic (orange), and square-wave (blue) stimuli (n=48 sessions per stimulus type, naturalistic median = 0.96 interquartile range [IQR]=[0.94 1.06], binary naturalistic median = 1.04 IQR=[1.03 1.06], square-wave median = 1.01 IQR=[0.99 1.04]). (D) Median PID gain on 0.5 gain trials (n=48 sessions per stimulus type, naturalistic median = 0.52 IQR=[0.50 0.54], binary naturalistic median = 0.52 IQR=[0.50 0.56], square-wave median = 0.54 IQR=[0.51 0.57]). One sample t-test with Ho:μ=0.5, p>0.05.