Table 2.
Wasserman & Castro (2005) results and best-fitting predictions.
| Test after Phase 2 | |||||
|---|---|---|---|---|---|
| Group | Cue | Obs | RWM | WCM | CEM |
| Uninformed | A | 8.77 | 10.32 | 10.37 | 8.99 |
| B | 5.95 | 5.65 | 5.68 | 6.27 | |
| C | 1.08 | 1.83 | 1.90 | 1.55 | |
| D | 1.08 | 0.00 | 0.00 | 0.00 | |
| W | 5.31 | 5.65 | 4.82 | 5.30 | |
| X | 6.24 | 5.65 | 5.68 | 6.27 | |
| Y | 6.97 | 5.65 | 6.37 | 6.54 | |
| Z | 2.31 | 0.00 | 0.00 | 0.00 | |
| Informed | A | 8.52 | 10.32 | 10.37 | 8.99 |
| B | 5.02 | 5.65 | 5.68 | 6.27 | |
| C | 2.10 | 1.83 | 1.90 | 1.55 | |
| D | 1.84 | 0.00 | 0.00 | 0.00 | |
| W | 2.89 | 5.65 | 3.49 | 3.62 | |
| X | 4.71 | 5.65 | 5.68 | 6.27 | |
| Y | 6.76 | 5.65 | 7.44 | 6.56 | |
| Z | 1.89 | 0.00 | 0.00 | 0.00 | |
Note: In a human contingency learning task, all subjects received 30 AW+, BX+, CY+, and DZ- trials in Phase 1 and 30 A+, C-, and D- in Phase 2. In Group Informed, subjects were informed that the target cue was absent during elemental trials. In Group Uninformed, information about the absent target cue was omitted during Phase 2 elemental training. In simulations of this experiment, each model used a separate absent beta parameter for the two groups. In addition to the data above, models were fit to the mean ratings of cues after Phase 1. Obs = observed food cue’s effectiveness. RWM = Rescorla-Wagner (1972) model. WCM = Within Compound Model (as described in Witnauer & Miller, 2011). CEM = Conjoint Error Model. Cells with fonts in bold font represent critical comparisons for testing retrospective revaluation effects (i.e., W vs. X represents backward blocking and Y vs. X represents recovery from overshadowing).