Table B2.
Experiment 3. Comparison of four models via the Akaike Information Criterion
| Observer | Model | AICc | DiffAIC | Akaike wt |
|---|---|---|---|---|
| SAW | 1 | 240.02 | 0.00 | 1 |
| 2 | 623.34 | 383.32 | 0 | |
| 3 | 784.86 | 544.84 | 0 | |
| 4 | 859.54 | 619.52 | 0 | |
| DHB | 1 | 432.81 | 0.00 | 1 |
| 2 | 918.40 | 485.59 | 0 | |
| 3 | 1315.05 | 882.24 | 0 | |
| 4 | 1250.47 | 817.66 | 0 | |
| ASB | 1 | 238.96 | 0.00 | 1 |
| 2 | 1400.17 | 1161.21 | 0 | |
| 3 | 346.60 | 107.64 | 0 | |
| 4 | 1414.79 | 1175.83 | 0 | |
| RH | 1 | 266.45 | 0.00 | 1 |
| 2 | 1168.47 | 902.02 | 0 | |
| 3 | 510.97 | 244.53 | 0 | |
| 4 | 1218.33 | 951.88 | 0 |
To combine evidence for a given model over the three contrast ratios, we summed the deviance values for each of the three fits, then computed AICc accordingly, with a combined total of 12 free parameters over 81 data points. This was done separately for each observer. Model 1 (fusion only for same polarity) was uniquely favoured, for all four observers.