Data from: View invariant representations in ancestral cortex
Data files
Oct 22, 2025 version files 27.44 MB
-
README.md
2.36 KB
-
SpatiallySelectiveViewInvariance-1.0.0.zip
27.43 MB
Abstract
A multilayered, thalamo-recipient visual cortex emerged ~320 MYA in stem amniotes. Despite its importance for understanding the evolution of cortical computation, its function remains unknown. We recorded visually evoked responses in the dorsal cortex of behaving turtles, considered a mammalian neocortex homolog. Using a spatial oddball paradigm, we found tuning to stimuli in deviant positions alongside adaptation to standard positions within the visual field. Eye tracking demonstrated that responses remained spatially selective despite gaze shifts altering retinal stimulus position. Thus, the turtle cortex encodes unexpected visual stimuli using computations invariant to retinal position, a property previously observed only in higher mammalian cortices. These results indicate that invariance computations preceded the evolution of local filtering computations in mammalian primary cortices, pointing to a new function for ancestral cortices. They also challenge hierarchical models of invariance computations, which assume that invariance is built from low-level features across multiple processing steps.
Title of dataset:
View invariant representations in ancestral cortex
Access this dataset on Dryad. DOI: 10.5061/dryad.k98sf7mm6
Description of the data and file structure:
Code for reproducing figures for Becker et al, "View invariant representations in ancestral cortex"
The pre-processed electrophysiological data, as well as head and eye angles of the experimental animals, are included to reproduce the specific data shown in our publication. The code provided loads the data and creates the subplots of the figures, visualizing and explaining the data provided. The code is supposed to be run in Matlab 2024b, no further packages are necessary.
Each Figure/subfigure has a script and corresponding metadata that will be loaded in the script. It is recommended to download all files into a folder and make it the current directory in matlab. Then each script can be opened, and the metadata will automatically be loaded from the same folder. Running each script will give one or several figure outputs in the end. The name of the script indicates which figure from the publication it refers to, e.g. “plot_Fig1g.m” will load the data and recreate the figure 1g. If a piece of code reproduces several figures from the same data it is indicated in the filename divided by underscores (e.g. “plot_Fig1i_j.m” plots figure 1i as well as 1j). The names of the data files for each code are mentioned and loaded in the beginning of each script. The content and structure of the data files vary from figure to figure but is explained inside the code and by the figure outputs.
The main variables of the data are:
LFP – downsampled electrophysiological data for specified periods
PC1 – first principal component of the LFP data
nPC1 – z-scored mean PC1 per recording session
Headangles/eyeangles – the estimated head and eye angles of the animal
viewangle - combined head+eye angles of the animal
Headspeed/eyespeed – the estimated speeds of the head and the eye of the animal
recname – the individual identifier per session
Sharing/Access information
Future code updates and packages can be found at:
https://github.com/EvolutionaryNeuralCodingLab/SpatiallySelectiveViewInvariance