Fig 1.
A schematic of ways in which individuals may alter behavior during social interaction.
(A) A hypothetical ethogram and behavioral density for a lone fly (blue). (B) Ethograms and behavioral frequency in the presence of a social partner when the two flies synchronize their behaviors in time. (C) The presence of the second fly alters the frequency of the different behaviors without synchronization. (D) A new social behavior (#4) arises in the presence of the partner.
Table 1.
Number of movies recorded, total recording time, and number of frames analyzed are listed for each behavioral context.
All analyses were performed on all data described here.
Table 2.
Parameters and values used in the behavioral analysis pipeline.
For more details see the supplement of [40].
Table 3.
Eight coarse behavioral clusters were used to describe 116 fine-grained behaviors obtained from the behavioral map.
These broad descriptions were used to sort behaviors and label the behavioral map.
Fig 2.
Distance and angle within the arena and to partner by behavioral context.
Histograms of distance to the center of the arena, distance to partner, and heading to partner are plotted where applicable. a) The fraction of time individuals from each context reside in ten equal-area concentric circles radiating from the center of the arena outwards. b) Histogram of the distance between paired individuals for all individuals in a given context. The red line indicates the expected distribution of distances between random points within the arena. c) 2-dimensional histogram of residence time in each paired context broken down by distance from the center of the arena (x-axis) and distance to the interaction partner (y-axis). d) Radial histogram, p(θ), of the heading of an individual given a context in relation to the centroid of the arena partner where θ = 0 refers to heading directly toward the interaction partner. Bins each represent a 20° angular range. The value of the outer ring is labeled in each radial histogram as they are scaled differently.
Fig 3.
Behavioral densities across paired and lone contexts.
a) Representative images from each of four steps used to visualize behavioral frequency. In sequence the images describe the progression of time-points on the two-dimensional behavioral space, the Gaussian filtered behavioral density, boundaries for coarse-grained regions based on hand-labeling, and finally the bubble plot summarizing the relative densities and transitions between coarse behavior regions. The size of each bubble represents the relative amount of time spent in that behavior and black lines show the transition probabilities between behaviors with line thicknesses proportional to the probability and right-handed curvature representing the direction of transition. b) Coarse behavioral densities for each of the six categories of experiment. c) We compare the densities across the six conditions by visualizing the JS divergence (Eq 1) across all groups, ordering the rows and columns based on similarity. d) Manually assigned labels describe the eight broad categories of behavior exhibited during the behavioral movies with color and location corresponding to the circle density plots.
Fig 4.
Fly behavior by radial position in the arena across context.
a) Illustration of distances dp and dc given a pair of flies in the arena. b) Stacked behavioral density plots show the frequency of each coarse behavior given distance to the interaction partner dp. A sliding window of 1 mm was used to calculate density at.1 mm increments with centers ranging from.1 to 18 mm. A stacked plot is shown for each paired condition. Color corresponds to the coarse behavior labels. c) Stacked behavioral density plots show the frequency of each coarse behavior binned in 1 mm radial windows centered at the distance dc from .5 to 7 mm. A stacked plot is shown for each experimental condition, where color corresponds to behavioral frequency at each radial position in the experimental arena. The dashed line indicates a radial distance of 7 mm.
Fig 5.
Changes in overall behavior as a function of distance to the arena center and the partner.
a) Comparison of behavioral density maps generated for each context given 1 mm bins of distance to the center of the arena. A behavioral map is generated for each context given the behavior of individuals at a specific 1 mm window of distance to the center of the arena. The JS divergence is computed between each map within a context, and the upper right section of the resulting matrix is displayed. The dashed line in each matrix indicates a radial distance of 7 mm b) A comparison of density maps for overlapping 2 mm bins of distance to partner is generated in the same manner as in (a) and the JS divergence between each map in a given context is displayed. The dashed line in each matrix indicates a partner distance of 4 mm.
Fig 6.
Correlation between behaviors performed by individuals within a pair.
The correlation coefficient is displayed across each pairing for each possible pair of coarse-grained behaviors.
Fig 7.
Partial mutual information between behaviors in paired animals.
a) The partial mutual information for all data from each behavioral pairing in a paired context is calculated and positive values are displayed. b) The partial mutual information is calculated as in part (a) independently for each pairing in a given context, and the mean value is shown for each set of paired coarse behaviors. Only positive values are displayed. An enrichment on the diagonal of these matrices demonstrates synchronization of similar behaviors in time.
Table 4.
The sums across partial mutual information are calculated for each context from the matrices displayed in Fig 7 and 7b and from simulated data generated using a one-step MM with probability densities and transition data derived from behavioral sequences for each condition.
All values are in bits.
Fig 8.
Sample ethograms and behavioral densities for fly pairs during simultaneous behavior.
Ethograms for pairs of flies in each of the paired conditions are shown along with coarse behavioral densities for each individual. Colors and ordering used correspond to the coarse behavioral labels introduced in Fig 3 and replicated at the top for reference. Samples were chosen from experiments in the top quartile of synchronization. The switch to song behaviors in the male in the courtship example is representative of courtship samples where copulation was reached before the end of recording.
Fig 9.
Location in the arena a behavior is performed depends on context.
For each of eight behaviors (columns) for each paired context (rows), the 2-dimensional histogram of residence time in each paired context is broken down by distance from the center of the arena (x-axis) and distance to the interaction partner (y-axis). Each heat map summarizes where an individual was in the arena and relation to the interaction partner when performing a particular behavior.