(A) A boosted decision tree classifier took n-dimensional vectors (where n is the number of simultaneously recorded neurons before or during SC inactivation) of CDh neuronal activity from individual epochs in single trials and returned a ‘classifier epoch index’ indicating the classifier’s guess regarding both which task-epoch and which cue-side condition the activity came from. (B) A confusion matrix is traditionally used to show how frequently each possible classifier epoch index (vertical axis) is applied for each true epoch index. Here we have additionally color-coded a confusion matrix to highlight three categories of classification outcome: (1) correct classifications (magenta), (2) cue-side misclassifications (blue), and (3) epoch misclassifications (green). (C) Across sessions aggregated cross-validated confusion matrix before same-side SC inactivation. Cell shading starts at white for 0% with larger values grading darker up to black at 100%. For each ‘true epoch index’ column (for example, ‘pre-cue cue-contra’ in column 1), the numeral in each row is the (rounded) percentage of times classifiers assigned that row’s class label to all inputs of that column’s input (true) class. For example, the number 81 in the 1st row and 1st column indicates that, across classifiers/sessions, when classifiers were given ‘pre-cue cue-contra’ input vectors, 81% of those vectors were labelled as ‘pre-cue cue-contra’; the number 19 in column 1, row six indicates that the remaining 19% of the time, classifiers erroneously labelled ‘pre-cue cue- contra’ activity vectors as ‘pre-cue cue-ipsi’. (D) Presentation as in (C) but for during same-side SC inactivation. (E) Difference of aggregated confusion matrixes (during minus before). Cell color mapping starts at saturated yellow for −50% with larger values grading to white at 0%, followed by increasing positive values grading up to saturated brown at +50%. (F) Left: per-session breakdown of classifier performance with classification rates divided into outcome categories shown in panel (B). Connected dots are from a single session. For each true epoch index on the horizontal axis, colored dots show the percentage of times the single-session classifier epoch index fell into one of the three classification categories defined in (B) (correct classification, epoch misclassification, cue-side misclassification). Right: per-category breakdown of mean classification rates across sessions and true epoch indexes; cc: correct classifications, ee: epoch errors, ce: cue-side errors. Shaded areas indicate 95% bootstrapped confidence intervals on mean from individual session classification rates. (G) Presentation as in panel (F), but for during same-side SC inactivations. (H) Left: per-session differences in classification categories (during minus before). Right: per-category breakdown of difference in mean classification rates across sessions and true epoch indexes. Dotted line marks 0. (I–N) Presentation as in panels (C–H) but for opposite-side SC inactivation sessions.