Rat visual perceptual decision making: A 100-day longitudinal behavioral time series
Data files
May 19, 2026 version files 2.82 MB
-
Rat195Vectors_290426.csv
2.81 MB
-
README.md
4.54 KB
Abstract
This archive contains example behavioral data from one rat performing a 2-alternative forced-choice visual motion discrimination task over a 100-day period. It is provided for the purpose of developing and demonstrating statistical methodology for modeling non-stationary behavioral time series. The dataset contains n =42,754 sequential behavioral trials, and reports five trial-level variables: the date and time of voluntary trial initiation, the direction of visual motion, the reaction time (elapsed time between trial initiation and response), the trial outcome (correct or error), and the reward magnitude.
Dataset DOI: 10.5061/dryad.hhmgqnkxb
Description of the data and file structure
This dataset contains trial-level behavioral data from a rat performing a 2-alternative forced-choice (2-AFC) visual motion discrimination task over 100 days, for the purpose of studying both motion perception and sensory decision-making in rodents.
Other relevant details of the experiment
Subject ID: Rat195 adult male Long-Evans rat
Motion coherence: (motion signal strength) 0.85 in all trials
Reward Size Rule
The reward offered for a correct response was history-dependent, as follows:
1. In trials immediately following a correct trial, the water reward for a correct response was 160 arbitrary units (AU)
2. In trials immediately following an error trial, the reward for a correct response was 80 AU
The experimental purpose of this was to disincentivize lapse (inattentive trials). Regardless of trial history, rats received 20 AU for incorrect responses, to reward responding and encourage continuing. The ‘proposed reward’ is the amount of water reward dispensed; we did not have any way to experimentally verify that this amount of water was, in fact, ingested by the rat.
Correction Trial Rule
Selection of the direction of motion for each trial was history-dependent, as follows:
1. In trials immediately following a correct trial, motion direction was L or R with equal probability, selected randomly and independently of other trials
2. In trials immediately following an error trial, there was a 0.5 probability of entering “correction trial state.”
a. If the correction trial state was not entered, the motion direction was random
b. If the correction trial state was entered, the direction of motion was deterministically the same as that of the previous trial. Once this state was entered, the state persisted until the next correct response.
The experimental purpose of this was to discourage response bias and force the rat to use both available responses regularly to continue to earn rewards. Without correction trials, the alternative strategy of giving the same response in every trial (ignoring the stimulus) would have yielded a reward in 50% of trials.
Because of these two experimental sequential dependencies, analysis of these data in Reinagel (2013) excluded trials following error trials.
Data Preprocessing
The trials were generated by a computer using custom MATLAB code. This code saved trial-level data and metadata in a complex data structure in a MATLAB (.mat) file daily. These data were retroactively curated for this study to extract the trial-level variables of interest for a time period during which other experimental variables (such as motion coherence, reward size, reward rule, correction trial rule, penalty rule, and testing schedule) were held constant.
The raw data files contained 43,081 recorded trials in the date interval. Of these, 241 trials were excluded as invalid due to one of the following: human-interrupted trials, reaction time not recorded or recorded as 0, or reaction time exceeding 10 seconds.
Files and variables
File: Rat195Vectors_290426.csv
Description:
Variables
- trialDate: (datetime) the date and time the trial was initiated by the rat
- RT: (float) reaction time (delay from stimulus onset to response) in seconds
- correct: (Boolean) 0 if rat’s response was an error, 1 if response was correct
- dotDirection: (float) direction of the motion signal, in radians (value 0 or π)
- proposedReward: (float) amount of reward rat was given for its response (AU)
Code/software
The data are stored as plain text in a comma-separated value (csv) format. No special software is required to view the data.
Access information
Other publicly accessible locations of the data:
- none
Data was derived from the following sources:
- none
Although these data were analyzed in (Reinagel, 2013), the data were not published at that time. These data were also not included in the subsequent data archive (Reinagel and Shevinsky, 2020) because that archive excluded experiments with history-dependent reward rules.
Acknowledgements
SP and DD provided expert technical assistance with animal training. AC implemented and validated the random dot motion task and trained this rat prior to the collection of this dataset.
During this experiment the rat had access to a perceptual decision task 24h/day in his home cage, with no other source of water. Trials were self-initiated by the rat licking a request port. This triggered the onset of a random-dot motion visual stimulus on a cage-attached display screen. The apparent motion stimulus persisted until the rat indicated a response by licking one of two water ports, earning water rewards for correct responses. There was no penalty for an incorrect response, and no imposed inter-trial interval.
This experiment was previously described and these data partially analyzed in (Reinagel, 2013), where additional details may be found. All experiments were conducted with the permission and under the oversight of the Institutional Animal Care and Use Committee (IACUC) of the University of California, San Diego. Importantly, the animal’s hydration state, health, and welfare were checked daily. Occasionally the rat was taken off task for a day, in which case he was given free water.