Schematic description of a transitive inference task

(A) Example of a 7-item ordered list of photographic stimuli. “Greater Than” symbols denote which items dominate which other items, (e.g. baked potato is correct relative to every other stimulus, but fighter plane is only correct when paired with zebra). (B) Example of a single trial. Following a start stimulus, two stimuli appear on screen. If the correct item is selected, positive feedback is provided; an incorrect choice leads to negative feedback. In every trial, there is always one correct and one incorrect answer. No more than two stimuli are ever presented at a time for the duration of the experiment. (C) The training and testing sets of pairs of stimuli. During training, only adjacent pairs (AB, BC, etc.) are presented typically in a randomized order that counterbalances how often each stimulus appears on each side of the screen. During testing, all 21 pairs are presented. The “symbolic distance” (denoted as “Dist”) refers to how many steps in the list one has to make to get from one item to another. Adjacent pairs have a symbolic distance of 1, whereas pairing both terminal items has a distance of 6. The “non-terminal pairs” are those that exclude both the first and last list item. If a non-terminal pair was also excluded from the training set, then it considered is a “critical pair” at test. (D) The frequency with which each stimulus is correct, relative to the number of pairs it appears in during that phase. During training, non-terminal items are correct 50% of the time, and consequently comparisons of the overall reward rate for non-terminal are equivocal, given only the information available during training.