Model of parafoveal chromatic and luminance temporal contrast sensitivity of humans and monkeys

This repository contains data and analysis code from Gelfand, E. C., & Horwitz, G. D. (2018). Model of parafoveal chromatic and luminance temporal contrast sensitivity of humans and monkeys. Journal of Vision, 18(12), 1-1; DOI: 10.1167/18.12.1.

The data are contained in two MATLAB .mat files. LMTF_data_A_public.mat contains data from monkey 1 and LMTF_data_U_public.mat contains data from monkey 2. Each file contains a single n x 7 matrix called data.

Each row of data represents a single threshold estimate. The columns are as follows:

• L-cone contrast at detection threshold
• M-cone contrast at detection threshold
• Temporal frequency (Hz) at detection threshold
• Threshold out of gamut (1 = yes, 0 = no). If 1, then the L- and M-cone contrasts in columns 1 and 2 are at the gamut edge and should not be taken as reasonable estimates of threshold.
• Horizontal position of the stimulus in the visual field (in tenths of degrees)
• Vertical position of the stimulus in the visual field (in tenths of degrees)
• Session identifier. Four threshold measurements were typically made in each session.

The analysis code is LMTF_generate_module_data_public.m. To execute this code, load one of the two datasets into the workspace and then run LMTF_generate_module_data_public(data).

The output of LMTF_generate_module_data_public is a structure with four fields.

• legacy This is a structure with 26 fields, three of which are more important than the other 23. These fields are:

  • legacy.mode5params An 18-element vector containing the fitted parameter values of the model described in the paper. See tf_fitterr3.m for more information. The 18 parameters (with reference to the equations in the paper) are as follows.

    • ζ₁: The transience of the non-opponent detection mechanism (Eq. 6)
    • n₁: The number of low-pass stages for the first filter of the non-opponent detection mechanism (Eq. 5)
    • n₂: The difference in the number of low-pass stages between the first and second filters that compose the non-opponent detection mechanism (Eq. 5)
    • τ₁: The log10 time constant (in s) of the first filter of the non-opponent detection mechanism (Eq. 5)
    • τ₂: The difference in log10 time constant between the first and second filters that compose the non-opponent detection mechanism (Eq. 5)
    • ζ₁: The transience of the opponent detection mechanism (Eq. 6)
    • n₁: The number of low-pass stages for the first filter of the opponent detection mechanism (Eq. 5)
    • n₂: The difference in the number of low-pass stages between the first and second filters that compose the opponent detection mechanism (Eq. 5)
    • τ₁: The log10 time constant (in s) of the first filter of the opponent detection mechanism (Eq. 5)
    • τ₂: The difference in log10 time constant between the first and second filters that compose the opponent detection mechanism (Eq. 5)
    • θ: The angle (in radians) of the non-opponent detection mechanism in the LM plane.
    • b_{0, LUM}: The gain of the non-opponent detection mechanism at the fovea (Eq. 10).
    • b_{1, LUM}: The slope of the fall-off in sensitivity of the non-opponent detection mechanism with eccentricity (Eq. 10)
    • b₂: Sensitivity anisotropy between horizontal and vertical meridians, identical for the non-opponent and the opponent mechanisms (Eq. 10).
    • b_{3, LUM}: Upper/lower visual field asymmetry for the non-opponent detection mechanism (Eq. 10).
    • b_{0, RG}: The gain of the opponent detection mechanism at the fovea (Eq. 10).
    • b_{1, RG}: The slope of the fall-off in sensitivity of the opponent detection mechanism with eccentricity (Eq. 10).
    • b₃, _RG: Upper/lower visual field asymmetry for the opponent detection mechanism (Eq. 10).

  • legacy.mode5models A 13 x m matrix of parameter values describing the model fits at each of the m visual field locations tested. See tf_fitterr2.m for more information. Parameters 1–6 control the non-opponent (LUM) detection mechanism. Parameters 7–12 control the opponent (RG) detection mechanism.

    • ξ_{1, LUM}: The gain of the non-opponent detection mechanism (Eq. 6)
    • ζ_{1, LUM}: The transience of the non-opponent detection mechanism (Eq. 6)
    • n_{1, LUM}: The number of low-pass stages for the first filter of the non-opponent detection mechanism (Eq. 5)
    • n_{2, LUM}: The difference in the number of low-pass stages between the first and second filters that compose the non-opponent detection mechanism (Eq. 5)
    • τ_{1, LUM}: The log10 time constant (in s) of the first filter of the non-opponent detection mechanism (Eq. 5)
    • τ_{2, LUM}: The difference in log10 time constant between the first and second filters that compose the non-opponent detection mechanism (Eq. 5)
    • ξ_{1, RG}: The gain of the opponent detection mechanism (Eq. 6)
    • ζ_{1, RG}: The transience of the opponent detection mechanism (Eq. 6)
    • n_{1, RG}: The number of low-pass stages for the first filter of the opponent detection mechanism (Eq. 5)
    • n_{2, RG}: The difference in the number of low-pass stages between the first and second filters that compose the opponent detection mechanism (Eq. 5)
    • τ_{1, RG}: The log10 time constant (in s) of the first filter of the opponent detection mechanism (Eq. 5)
    • τ_{2, RG}: The difference in log10 time constant between the first and second filters that compose the opponent detection mechanism (Eq. 5)
    • θ: The angle (in radians) of the non-opponent detection mechanism in the LM plane.

  • legacy.mode5fvs An m element vector containing the error of the model fit at each of the m locations tested. The other fields in data.legacy are similar but are based on models that were rejected but used in the fitting process.

• raw Contains the raw data organized by location in the visual field.
• eccs An m x 2 matrix providing a list of the visual field locations tested (in 10th of degree of visual angle).
• domain Unused.

To inspect the model fits and explore the action of the parameters values on the function fits, use LMTFBrowser.m, passing the structure produced by LMTF_generate_module_data_public.m in as the input argument to the function. LMTFBrowser.m is essentially completely undocumented, but on the off chance that any besides me ever wants to use it, I am happy to provide documentation.