• The dataset is drawn from a human category learning experiment.

• The dataset is available at:
  https://github.com/crossley/crit_learn_delay

• This repository contains behavioural data from experiments investigating criterial learning under different feedback delays.

• Dataset is available within the repository at: https://github.com/crossley/crit_learn_delay/tree/main/data

• There are four folders within this data folder:

  • bin_data_v2
  • delay
  • immed
  • maip

• The raw data lives inside each of these folders in the form of .csv files.

• Data files were read in and combined using data.table in R.

• Each subject completed multiple problems; we computed the number of trials to reach criterion (t2c) for each problem and averaged across problems for each subject.

• We filtered out outliers with t2c >= 300.

• Sample sizes per condition before filtering:

##          cnd     N
##       <char> <int>
## 1:     Delay    20
## 2:  Long ITI    21
## 3: Short ITI    17

• Sample sizes per condition after filtering:

##          cnd     N
##       <fctr> <int>
## 1:     Delay    17
## 2:  Long ITI    18
## 3: Short ITI    12

• Shapiro-Wilk Test for Normality

• Normality is rejected in Long ITI and Short ITI conditions.

• Despite this, ANOVA and t-tests are generally robust to mild violations of normality.

##        Effect DFn DFd       SSn    SSd          F            p p<.05       ges
## 1 (Intercept)   1  44 698082.27 106086 289.535175 5.673951e-21     * 0.8680799
## 2         cnd   2  44  26365.99 106086   5.467751 7.571651e-03     * 0.1990608

• Significant main effect of condition on t2c.

• Proceeded with pairwise t-tests to interpret effects.

• Delay condition resulted in significantly more trials than either Long or Short ITI.

• No significant difference between Long ITI and Short ITI.

• Clear visual separation of Delay condition from the other two.

• Error bars represent standard error of the mean (SEM).

• Longer feedback delays led to slower learning (higher trials-to-criterion).

• Results support the hypothesis that delay disrupts criterial learning.

• Caution is warranted in interpreting results due to non-normality and outliers.