Skip to main content
Dryad logo

Comparison of categorical color perception in two Estrildid finches


Caves, Eleanor et al. (2020), Comparison of categorical color perception in two Estrildid finches, Dryad, Dataset,


Sensory systems are predicted to be adapted to the perception of important stimuli, such as signals used in communication. Prior work has shown that female zebra finches perceive the carotenoid-based orange-red coloration of male beaks—a mate choice signal—categorically. Specifically, females exhibited an increased ability to discriminate between colors from opposite sides of a perceptual category boundary than equally-different colors from the same side of the boundary. The Bengalese finch, an Estrildid finch related to the zebra finch, is black, brown and white, lacking carotenoid coloration. To explore the relationship between categorical color perception and signal use, we tested Bengalese finches using the same orange-red continuum as in zebra finches, and also tested how both species discriminated among colors differing systematically in hue and brightness. Unlike in zebra finches, we found no evidence of categorical perception of an orange-red continuum in Bengalese finches. Instead, we found that the combination of chromatic distance (hue difference) and Michelson contrast (difference in brightness) strongly correlated with color discrimination ability on all tested color pairs in Bengalese finches. The pattern was different in zebra finches: this strong correlation held only when discriminating between colors from different categories, but not when discriminating between colors from within the same category. These experiments suggest that categorical perception is not a universal feature of avian, or even Estrildid finch, vision. Our findings also provide further insights into the mechanism underlying categorical perception and are consistent with the hypothesis that categorical perception is adapted for signal perception.


These behavioral data were collected by EMC, PAG, and DB. The data are from a behavioral assay in which birds were provided with a foraging grid with 12 wells. On top of six of the wells, we placed colored circular discs that were either the same color on both halves (solid) or different colors on each half (bicolor): two bicolor discs were used in each trial alongside two solid discs of each of the colors comprising the bicolor discs. We trained birds to search for food rewards beneath discs that they perceived as bicolor. A bird passed a trial if it flipped both bicolor discs before flipping any of the solid discs. The data archived here have been processed to show the mean "pass frequency," i.e. the proportion of trials out of 10 in which birds flipped both bicolor discs before flipping any solid discs. Some of the data files show pass frequency for each bird for each color combination; others show the mean pass frequency across birds (i.e. all zebra finches averaged together) for a given color combination.

The data also contain  the bird ID (a unique identifier for each bird), the number ID of the colors on each half of the disc (col1 and col2), and the chromatic distance and contrast for the color combination in that trial. Chromatic distance was calculated using the Receptor Noise-Limited model (see main text for details). Contrast is Michelson contrast (see main text for details).

Usage Notes

"Analysis R Script.R"- This file contains all the code necessary to build the training models and
do the analysis associated with them (i.e. compare them to observed results, generate figures).
It also has the script for the AIC comparisons between models that use a category vs. contrast.

"Bengalese Finch Data," "Zebra Finch Beak Data," and "Zebra Finch Luminance Training Data" support the above R script.
They are summary datasets (i.e. population mean pass frequencies for each color comparison). For example, "Bengalese Finch Data.csv" shows
mean pass frequency across our population of Bengalese finches for each color combination.
"data" are data regarding zebra finch pass frequency on the Beak Set, originally published in [removed for double blind review]. "Zebra Finch Luminance Training Data" show zebra finch pass frequencies for the Extended Color set.

"Zebra_finch_raw_beak_set," "Zebra_finch_raw_extended_set," and "Bengalese_finch_raw_beak_and_extended_sets" contain
individual mean pass frequency for each color comparison. These data file can be used to create
things like Figures 2 and 3 (labeling and discrimination)

Column descriptions for all data files include:
bird.ID :    "raw" datasets only. Individual bird ID.
cat :    Color comparison
col1 :    Color 1 (the redder color) in the color comparison
col2 :    Color 2 (the oranger color) in the color comparison
contrast :    Michelson contrast between Colors 1 and 2
chrom.dist :    Chromatic Distance (Delta S) Between Colors 1 and 2
pass.freq :    Proportion of comparison tasks that birds pass (population mean)
sd :    Standard deviation of pass frequency
n :    # of birds that completed comparisons
se :    Standard error of pass frequency
training :    "Bengalese Finch Data.csv": Indicates whether a given comparison is used in the training model (binary, 0=not included, 1=included)


Duke University Office of the Provost

Duke University Office of the Provost