Data, stimuli, and analyses for "High-level aftereffects reveal the role of statistical features in visual shape encoding"

Data and code share. This record contains data and code (written in MATLAB) to reproduce the results shown in: Morgenstern, Y. , Storrs, K., R., Schmidt, F., Hartmann, F., Tiedemann, H., Tiedemann, H, ., Wagemans, J., & Fleming, R. (in press) . High-level aftereffects reveal the role of statistical features in visual shape coding. Current Biology Below is a summary of shared scripts that load data, run the analysis (including options for fitting model parameters or loading pre-computed fitted parameters), and plotting the results. Figure 1 Fig1C_shapespace.m: draw shape space (as in Figure 1C) Fig1EFG_plotpsychometricdata.m: fit psychometric model to pooled data and plot (as in Figure 1EFG) getExptShapesHbias.m: saves a data structure (which we call 'package') with adaptor, test, and human biases from experiment 1. (Used to fit models; e.g., see fitGabPyr2Hbais.m or fitTAEGANfit2Hbais.m) Figure 2 and 3A fig3A_modeval_expt1.m: generate figure that evaluates models in Figure 3A on how well they predict aftereffects in Experiment 1. (script located in the 'Figures 2 and 3A/models' directory). Code to fit the models, and figures that show examples of model predictions are in the model directories, and summarized below: Model: GabPyrAE note: To run GabPyr, you will likely need to recompile the .mex files in 'matlabPyrTools/mex'. Then move the recompiled files into the 'matlabPyrToos' directory fig2B_GabPyrAEVisFigs.m: produce GabPyrAE model images for example adaptor and test image ( as in Figure 2B ) figS2BC_GabPyrAEExp.m: get GabPyrAE model responses to simulated tilt aftereffect experiment using anisotropic noise, and plot model responses as in Figures S2BC. getGabPyrAEMod.m: given an adaptor and test image, this function produces the unfit GabPyrAE prediction fitGabPyr2Hbias.m: fit GabPyrAE model to best predict human baises in Experiment 1 using data structure from getExptShapesHbias.m. This is the general function that calls on MATLAB's GA algorithm to minimize the error function in fitGabPyrNormMod2Stims.m evalGabPyrAE_fitmod_aic.m: evaluate GabPyrAE fitted model on how well it predicts human biases from experiment 1. evalGabPyrAE_unfitmod_aic.m: evaluate GabPyrAE fitted model on how well it predicts human biases from experiment 1 Model: TAE fig2CD_TAEmod.m: produce TAE model images for example adaptor and test image (as in Figure 2CD) getTAEModonShape.m: given an adaptor and test image, this function produces TAE Original prediction. Input to function is adaptor and test shapes, and TAE model parameters alpha and sigma. getTAEGAN_spwt_onShape.m: given an adaptor and test image, this function produces TAEGAN prediction. Input to function is adaptor and test shapes, and TAEGAN model parameters which include a constant term, alpha and sigma, as well as Gaussian pooling parameter that determines how much TAE to incorporate on the test or mean shapes from neighbouring adaptor line segments. getTAE_spwt_onShape_nn.m: given an adaptor and test image, this function produces TAE nearest neighbour prediction. Input to function is adaptor and test shapes, and TAE nearest neighbour model parameters which include a constant term, alpha and sigma, as well as Gaussian pooling parameter that determines how much TAE to incorporate on the test or mean shapes from neighbouring adaptor line segments. fitTAEGAN2Hbias.m: fit TAEGAN model to best predict human biases in Experiment 1 using data structure from getExptShapesHbias.m. This is the general function that calls on MATLAB's GA algorithm to minimize the error function in fitTAEGANMod2Stims.m fitTAENN2Hbias.m: fit TAE nearest neighbour model to best predict human biases in Experiment 1 using data structure from getExptShapesHbias.m. This is the general function that calls on MATLAB's GA algorithm to minimize the error function in fitTAENNMod2Stims.m. evalTAEGAN_fitmod_aic.m: evaluate TAEGAN fitted model on how well it predicts human biases from experiment 1. evalTAENN_fitmod_aic.m: evaluate TAE nearest neighbour fitted model on how well it predicts human biases from experiment 1. evalTAE_unfitmod_aic.m: evaluate TAE Original model on how well it predicts human biases from experiment 1 Model: PSAE fig2EF_PSAEmod.m: produce PSAE model images for example adaptor and test image getPos_spwt_ShiftononShape_io.m: given an adaptor and test image, this function produces PSAEGAN prediction. Input to function is adaptor and test shapes, and PSAEGAN model parameters which include a constant term, alpha and sigma, as well as Gaussian pooling parameter that determines how much PSAE to incorporate on the test or mean shapes from neighbouring adaptor line segments. getPos_spwt_ShiftonShape_io_nn.m: given an adaptor and test image, this function produces PSAE nearest neighbour prediction. Input to function is adaptor and test shapes, and PSAE nearest neighbour model parameters which include a constant term, alpha and sigma, as well as Gaussian pooling parameter that determines how much PSAE to incorporate on the test or mean shapes from neighbouring adaptor line segments. fitPSAEGAN2Hbias.m: fit PSAEGAN model to best predict human biases in Experiment 1 using data structure from getExptShapesHbias.m. This is the general function that calls on MATLAB's GA algorithm to minimize the error function in fitPSAEGANMod2Stims.m fitPSAENN2Hbias.m: fit PSAE nearest neighbour model to best predict human biases in Experiment 1 using data structure from getExptShapesHbias.m. This is the general function that calls on MATLAB's GA algorithm to minimize the error function in fitPSAENNMod2Stims.m. evalPSAEGAN_fitmod_aic.m: evaluate PSAEGAN fitted model on how well it predicts human biases from experiment 1. evalPSAENN_fitmod_aic.m: evaluate PSAE nearest neighbour fitted model on how well it predicts human biases from experiment 1. Model: ShapeComp and No Adaptation eval_ShapeComp _aic.m: evaluate ShapeComp 1 parameter fitted model on how well it predicts human biases from experiment 1. eval_NoAdaptation _aic.m: evaluate model that predicts no adaptation on how well it predicts human biases from experiment 1. Figure 3BC fig3BC_Experiment2.m: load, analyze, and plot experiment 2 data (as in Figure 3B and C). figS4_Expt2_stimuli.m: show adaptors (in black) and test shapes for ShapeComp (purple), PSAE fit GAN (green), and no adaptation model (white) (as in Figure S4)