Fig 1.
The four phases of the experiment.
First, a truck is observed (a) and children are required to touch which pattern of the four possible patterns in the box matches the car. The response of the first pattern on the car stays on the screen (right box in (b)) and the car moves behind a screen with an iconic representation of its operation. Children are then asked to predict the outcome with a new set of four choices on the left (b). In training trials (b), children answer until they are correct and then see the screen lift to reveal the car with the new pattern (c). The critical response in test trials is shown in (d), with the car occluded after passing behind two screens. No feedback is provided on these trials.
Fig 2.
Mean response accuracies to single screens which performed change (CH) or identity (ID) operations.
A chance rate of 25% is shown by the dotted red line.
Fig 3.
Mean response accuracies to double screens, each of which performed change (CH) or identity (ID) operations.
Error bars show confidence intervals. A chance rate of 25% is shown by the dotted red line. The blue dots correspond to the accuracy predicted under independent application of the single function accuracies in Fig 2.
Fig 4.
Coefficients in a mixed-effect logistic regression predicting accuracy on test (two-screen) trials from a number of predictors.
There intercept here, representing the mean accuracy, should be compared to the chance rate of 1/4 (red dot, at logit(0.25) on this scale), the overall chance guessing rate for 4 options. All other coefficients should be compared to the x = 0 line, showing whether they had a statistically significant influence on response accuracy. Effects significant at p < 0.05) are shown with gray bars.