Research ArticleNEUROSCIENCE

Parallel processing of polarization and intensity information in fiddler crab vision

See allHide authors and affiliations

Science Advances  21 Aug 2019:
Vol. 5, no. 8, eaax3572
DOI: 10.1126/sciadv.aax3572
  • Fig. 1 Hypothesized models of intensity and polarization channel integration in crustaceans.

    (A) Horizontally and vertically oriented receptor cells project to the epl1 and epl2 layers of the lamina, respectively, where they synapse with three types of descending neuron (monopolar cells M2 to M4), resulting in three channels of information per ommatidium: horizontal (H, M3) and vertical (V, M4) polarization, and intensity (I, M2) (30, 31) [redrawn from (7)]. (B) Single-channel model demonstrating a fusion of V, H, and I into a single value [intensity-polarization (IP) contrast]. (C) Parallel-channel model in which the polarization (V and H) and intensity (I) channels combine separately into two parallel measures (P contrast and I contrast).

  • Fig. 2 Intensity and polarization images of two black-headed gulls (Chroicocephalus ridibundus) viewed against a clear sky.

    (A) Original intensity and polarization images and (B) the same images showing the visual features that are resolvable by the crabs at increasing viewing distance based on the visual resolution of the region of the eye in Gelasimus vomeris [formerly Uca vomeris (20)] viewing approximately 15° to 20° above the horizon (45, 46). The polarization information is presented as a receptor activity ratio, i.e., the relative opponent output of the horizontally (H = 1) and vertically (V = −1) oriented photoreceptor channels calculated using a visual model (7). Note how the intensity contrast of a predator can vary depending on the animal’s coloration and illumination, but the polarization contrast remains the same. The Supplementary Materials provide the details on the polarization video camera used to capture these images. Photo credit: Sam Smithers, University of Bristol.

  • Fig. 3 Predictions from the IP response models and results from behavioral experiments.

    The predicted response probabilities of a simulated crab population (n = 10,000) to a range of intensity contrasts, with the addition of a set of fixed polarization contrasts (polarization distance, 0 to −0.5; gray lines in increasing lightness) using (A) the single-channel model and (B) the parallel-channel model (see the Supplementary Materials for model calculations and explanation), and actual response probabilities (i.e., the proxy for detected visual contrast) of fiddler crabs to looming stimuli based on (C) varying intensity contrasts, (D) varying polarization contrasts, and (E and F) mixed intensity and polarization contrasts. The ranges of contrasts presented in (E) and (F) are the same as in (C) and (D) but with the addition of a fixed polarization or intensity contrast, respectively. Error bars are Wilson score intervals calculated using the sample size for each point (n) and the number of responses. Vertical dashed line is the location of zero contrast between stimulus intensity [for (C) and (E)] or stimulus polarization [for (D) and (F)] and the background. The data from two separate experiments are presented in (E), each with a different range of Weber contrasts. Note that the magnitude of response to any given stimulus depended on its contrast relative to that of the other stimuli tested within the same experiment rather than its absolute contrast. This is illustrated by comparing the response to the stimuli colored blue in (D) and (E), both of which have exactly the same polarization contrast (intensity contrast is zero). n is the number of animals that contributed to the response probability measured for each contrast.

  • Fig. 4 Interactions between intensity and polarization contrasts.

    (A) Addition of a fixed polarization contrast (P) to a range of intensity only stimuli (I1 to I4). (B) Solo and combined effect of two intensity contrasts (Ia and Ib) and two polarization contrasts (Pa and Pb). Note that in (B), for clarity, each of the I- and P-only stimuli is plotted twice, once for each stimulus combination. C, control (no intensity or polarization contrast). Error bars are Wilson score intervals calculated using the sample size for each point (n) and the number of responses. Nonsignificance (ns) between the highest solo response probability and combined probability was determined using pairwise McNemar tests. Intensity and polarization contrast levels for each stimulus are plotted on the lower-most axes (blue squares, Weber contrasts; red triangles, polarization distance). n is the number of animals that contributed to the response probability measured for each contrast.

  • Fig. 5 Image processing inspired by single- and parallel-channel models.

    Intensity (A) and (B) polarization images are combined, (C) to enhance intensity contrast through the single-channel model or (D) as separate layers of contrast information using the parallel-channel model. Photo credit: Martin How, University of Bristol.

  • Fig. 6 Experimental setup and properties of the IP screen.

    (A) Treadmill apparatus and IP screen. Crabs were subjected to looming stimuli that varied independently in intensity (produced by the digital projector) and polarization (produced by the modified LCD panel). (B) Polarization (solid lines) and intensity (dashed lines) measurements of the IP screen at different intensity and polarization screen RGB values (R = G = B).

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/5/8/eaax3572/DC1

    Fig. S1. Angle of polarization (AoP) of the IP screen.

    Fig. S2. Top-view schematic of the two-channel polarization camera used to capture video of seabirds.

    Fig. S3. Simulation results from the IP response model showing the normally disputed response thresholds.

    Fig. S4. Example predictions from the IP response models.

    Movie S1. Example freeze response of a fiddler crab to a looming stimulus.

    Data file S1. Data from behavioral experiments.

    MATLAB code for running the IP response model

    Reference (47)

  • Supplementary Materials

    The PDF file includes:

    • Fig. S1. Angle of polarization (AoP) of the IP screen
    • Fig. S2. Top-view schematic of the two-channel polarization camera used to capture video of seabirds.
    • Fig. S3. Simulation results from the IP response model showing the normally disputed response thresholds.
    • Fig. S4. Example predictions from the IP response models.
    • Legend for movie S1
    • Legend for data file S1
    • Legend for matlab code
    • Reference (47)

    Download PDF

    Other Supplementary Material for this manuscript includes the following:

    • Movie S1 (.avi). Example freeze response of a fiddler crab to a looming stimulus.
    • Data file S1 (Microsoft Excel format). Data from behavioral experiments.
    • MATLAB code for running the IP response model (.m format).

    Files in this Data Supplement:

Navigate This Article