Figure 6 (blue, red, and green bars) shows results of the same/different task. Visual inspection of the plot suggests a data pattern that is very similar to that of
Experiment 1. First, the proportion of “same” responses is lower in the “other” color conditions than in the initial and final conditions, indicating that participants were not merely guessing. We do note that the proportion of “same” response in the “other” color condition was higher than it was in
Experiment 1. This might suggest that the different mask design in
Experiment 2 degrades memory representations and thereby renders the task more difficult. However, a generalized linear mixed model of the data from both
Experiments 1 and
2 (excluding the “other” color conditions) together with mask type as an additional fixed effect (this was the only difference between the two experiments; formula:
same ∼ cue type * comparison color + mask + [1/
image]) did not explain the data better than a model without mask as a fixed effect (LRT = 2.56,
p = 0.109). This indicated that the type of mask (localized box or full screen) did not influence participant responses to the initial and final comparison frames in a systematic way. Furthermore, within
Experiment 2, a generalized linear mixed model applied to a single initial or final comparison color condition, as well as the corresponding “other” condition, invariably explains the data better when it does include comparison color as a fixed effect than when it does not: postcue final and postcue other (LRT = 10.15,
p = 0.0014), postcue initial and postcue other (LRT = 10.75,
p = 0.0010), retrocue final and retrocue other (LRT = 15.66,
p < 0.001), or retrocue initial and retrocue other (LRT = 4.91,
p = 0.027). This finding again indicates that participants are not uniformly guessing but rather that their memory representations provide more evidence for the initial and final color than for the other color. In agreement with
Experiment 1,
Figure 6 additionally suggests that, for the retrocue condition, the proportion of “same” responses is higher for the final color than for the initial color. Indeed, a generalized linear mixed model applied to the retrocue initial and retrocue final condition pair explains the data significantly better when it includes comparison color as a fixed effect than when it does not (LRT = 4.55,
p = 0.0328). Also in agreement with
Experiment 1,
Figure 6 suggests that this difference between the initial and final color was absent in the postcue condition, which would suggest an interaction between cue type (retro or post) and comparison color (initial or final). The difference is indeed absent in the postcue condition (generalized linear mixed model applied to the postcue initial and postcue final condition pair yields LRT = 0.006,
p = 0.937), but this impression of an interaction is not supported by statistical analysis in
Experiment 2 as it was for
Experiment 1. This time, the full generalized linear mixed model applied to the data from the initial and final conditions and both cue conditions does not explain the data significantly better than a model without the interaction (LRT = 2.37,
p = 0.124).