Accuracy and confidence of visual short-term memory do not go hand-in-hand: behavioral and neural dissociations

Abstract

Currently influential models of working memory posit that memory content is highly accessible to conscious inspection. These models predict that metacognition of memory performance should go hand-in-hand with the accuracy of the underlying memory representation. To test this view, we investigated how visual information presented during the maintenance period affects VSTM accuracy and confidence. We used a delayed cue-target orientation discrimination task in which participants were asked to hold in memory a grating, and during the maintenance period a second memory cue could be presented. VSTM accuracy of the first memory cue was impaired when the orientation of the second memory cue was sufficiently different. However, participants' response confidence was reduced whenever the second memory cue was presented; thus VSTM accuracy and confidence were dissociated. In a second experiment, we applied transcranial direct current stimulation (tDCS) over the right dorsolateral prefrontal cortex (DLPFC) to investigate the causal role of this region in VSTM metacognition. Relative to the sham condition, anodal tDCS induced a general reduction in confidence ratings but did not affect VSTM accuracy. Overall, these results indicate that our metacognition of memory performance is influenced by factors other than the accuracy of the underlying memory representation.

Figure 1. Timeline of an experimental trial.

Participants were asked to maintain in memory the orientation of a memory cue (grating); at the end of each trial they were asked to indicate whether a test probe was tilted to the left or to the right relative to the memory cue. In addition, participants provided a confidence rating for this memory jdugment by using a scale from 1–9 (1 = not confident at all; 9 = extremely confident). On 75% of the trials, the first memory cue was followed by a second cue; this could be either identical to the first cue, or its orientation differed by 10 or 40 degrees. In the active condition, participants were asked to hold its orientation in memory. Thus in this condition, participants were required to hold the orientation of two cues on each trial. The maintenance of the 2^nd cue was assessed in the same manner as that of the 1^st cue: they were asked to indicate whether a test probe was tilted to the left or to the right relative to the memory cue. The memory judgment relating to the 2^nd cue was always made after the two responses (accuracy and confidence responses) relating to the first cue. In the passive condition, participants were not required to hold the 2^nd cue in memory. To ensure that they attended to the 2^nd cue, participants were asked to indicate at the end of the trial whether or not the 2^nd cue was presented.

Figure 2. Dissociation between VSTM accuracy and confidence in Experiment 1.

A) VSTM accuracy for the 1^st memory cue as a function of orientation difference between 1^st and 2^nd memory cue in the active and passive conditions. Relative to the baseline condition (i.e. when no 2^nd cue was presented) memory accuracy was reduced only when orientation difference between the two cues was 40 degrees; this effect was not significantly modulated by memory load (although a trend for an interaction between memory load and orientation was present). The asterisks indicate conditions which significantly differ from the BL condition. Error bars indicate ±1 SEM. B) VSTM confidence as a function of orientation difference between the 1^st and 2^nd memory cue in active and passive condition. Relative to the baseline condition (i.e. when no 2^nd cue was presented) confidence ratings were significantly reduced whenever the 2nd cue was presented; this effect was not significantly modulated by the memory load or by the orientation of the 2nd cue. The asterisks indicate conditions which significantly differ from the BL condition. Error bars indicate ±1 SEM.

Figure 3. Correlation between VSTM accuracy and confidence for each tDCS condition.

Correlation between confidence ratings and VSTM accuracy of the 1^st memory cue in anodal-tDCS conditions (panel A) and sham-tDCS conditions (panel B). Correlation between confidence ratings of 1^st memory cue and VSTM accuracy of the 2^st memory cue in anodal-tDCS conditions (Panel C) and sham-tDCS conditions (Panel D) Error bras indicate ±1 SEM.

Figure 4. Dissociation between VSTM accuracy and confidence in Experiment 2: differential effects of tDCS and orientation similarity between the memory cues.

(A) Mean (n = 15) VSTM accuracy as a function of stimulus condition for each tDCS condition. A significant main effect of stimulus condition was found, with accuracy being reduced when orientation difference between first and second memory cue was 10 or 40 deg, with largest effect found at 40 degrees. In addition a main effect of session order was found (higher performance in post-tDCS versus pre-tDCS), suggesting a slight learning effect. No other main effect or interaction was observed. Error bars indicate ±1 SEM. A similar pattern of results was observed also in reaction times analysis. (B) Mean (n = 15) confidence ratings as a function of stimulus condition for each tDCS condition. A significant interaction between tDCS condition and session order was observed, such that confidence ratings were generally lower in the post-real tDCS session. As in Experiment 1, confidence ratings were reduced whenever the 2^nd cue was presented.