John Palmer
University Hospital Zürich
Zürich, Switzerland
This paper reports the results of 2 experiments originally intended to study implicit sequence learning (ISL). Participants (Ps) were asked to identify in which of 4 directions (up, down, left, right) an arrow would be pointing that they would see immediately after their response (trial-by-trial feedback). In Experiment 1, 35 male Ps received 1 100-trial run with random targets followed by 2 100-trial runs with biased targets. The bias was defined as the target for trial t+1 being displaced 90 degrees clockwise (CW) or counterclockwise (CCW) from the target for trial t. The order of the 2 biases was counterbalanced across Ps. Pro-bias targets appeared in 46.5% of the trials, counterbias targets in 10.1%, and each orthogonal alternative in 25.2%. 18 Ps were extreme believers in the paranormal (sheep) and 17 were extreme skeptics (goats). Half of each group received a levodopa pill (dopamine) before the test session and the other half a placebo. The dependent variable, relative pro-bias responding (rPBR), was the difference between pro-bias and counter-bias responses (e.g., CCW responses in CW-biased runs). A succession of post-hoc analyses intended to clarify a marked CW response bias in the 1st half of the 1st biased run among levodopa Ps responding to CW-biased targets revealed a suggestive tendency for levodopa Ps to "anticipate" in the random run the target bias (CW or CCW) they would receive in their 1st biased run. As Ps at this time had been told nothing about the targets in the biased runs, this suggestive finding was called an anomalous anticipation effect (AAE). To determine if the AAE might be present elsewhere in the data, the random run was analyzed using the same ANOVA that had been used to test for ISL in the biased runs, with the target bias defined as that which Ps would receive in their 1st biased run. The ANOVA revealed a significant belief x half-run interaction, in which sheep demonstrated an increase in rPBR from the 1st to 2nd half-run and goats a corresponding decrease. This finding was interpreted as correct and incorrect anticipation by goats and sheep respectively in the 1st half of the random run; the 1st-2nd half differences were interpreted as changes in strategy due to non-reinforcing feedback during the run. The sample for Experiment 2 was 40 females. The main procedural changes were no levodopa condition and a between-P manipulation of target bias, with each P receiving 2 200-trial biased runs. The ANOVA of the random run revealed a significant main effect for belief, with goats anticipating correctly and sheep incorrectly. In both experiments, skeptics scored significantly higher than believers in the 1st half of the random run. The reversal of the traditional sheep-goat effect was speculatively attributed to goats being more comfortable than sheep in the test situation, a circumstance created by the fact that, in contrast to most sheep-goat experiments, both experimenters were goats.