Fig. 3. Results of a representative blocking experiment. (A) Pooled half-experiments with colours or patterns (with BG18 filter) during the first training period (N=53). (B) Control group with patterns (white light) during the first training period (N=27); the switch from white light to monochromatic blue or green light erases the predictive value of the patterns. (C) Control group with colours (replay) during the first training period (N=26). Cross-hatched columns training; open columns, test period; shaded areas, compound stimulus; unshaded areas, single stimulus. Values are means ± S.E.M.