Figure 1
Study 1: protocol for the 2-AFC picture similarity recognition memory task. (a) Study sequence: 160 pictures were shown in random order within both buffer and target blocks, each image was displayed for 2000 ms following a fixation cross (200 ms). Targets were not identified as such to participants, who were instructed to try to pay attention to all the images; (b) Test sequence: example picture-pairs used in the test phase (all pairs were shown in random order). Note, for illustration purposes the target A picture is shown on the left and the lure on the right. For comparison, the B picture shown in the target block is also illustrated (not shown at test). Pictures illustrated are all from the category of green landscapes and trees.
Figure 2
Study 1: illustrating the distribution of trial level confidence ratings (a) by the remember response (no or yes); (b) by recognition accuracy (incorrect or correct). Box plots show median values as horizontal black bands. The associated violin plots illustrate the distribution of responses, along with mean values, shown as black squares. A straight line between the mean values is equivalent to a point biserial plot.
Table 1
Study 1: variables of interest [confidence, recognition accuracy (proportion correct), proportion remembered, vividness, and reaction time] by test-pair condition.
| TEST PICTURE PAIR CONDITION (n) | CONFIDENCE | PROPORTION CORRECT | PROPORTIONREMEMBERED | VIVIDNESS | REACTION TIME, (ms) |
|---|---|---|---|---|---|
| A–A’ (724) | 44.6 (1.09) | .63 (0.02) | 0.52 (0.02) | 50.05 (1.07) | 3178.3 (48.74) |
| A–B’ (731) | 48.65 (1.13) | .59 (0.02) | 0.56 (0.02) | 52.34 (1.08) | 2974.9 (45.74) |
| A–X (739) | 47.21 (0.65) | .65 (0.02) | 0.53 (0.02) | 50.58 (1.08) | 2923.0 (44.28) |
| ANOVA | F (2, 2191) 3.36; p = 0.04, η2 = 0.003 | F (2, 2191) 3.06; p = 0.05, η2 = 0.003 | F (2, 2191) 1.82; p = 0.16 | F (2,2191) 1.24, p = 0.29 | F (2, 2191) 8.48; p < 0.001, η2 = 0.008 |
[i] Note. Results are compared by one-way ANOVA. Effect sizes are represented by the partial η2 statistic, where η2 = 0.01 indicates a small, η2 = 0.06 a moderate, and η2 = 0 .14 a large effect.
Figure 3
Study 1: CAC plots of trial level data across all test-pair conditions, illustrating recognition accuracy (mean proportion correct) by confidence (binned in quartiles): (a) overall data; (b) the overall data split by the experience of remembering. [(** p < 0.001), (* p < 0.01). Error bars represent 95% C.I. – numbers represent trials assessed within each bin of confidence – only positive or negative bars are shown in Figure 3b for clarity; numbers represent trials assessed within each bin of confidence – bold figures represent remembered trials and italic figures represent trials associated with not remembering].
Figure 4
Study 1: CAC plots by test picture-pair condition illustrating recognition accuracy (mean proportion correct) by confidence (binned in quartiles) split by the experience of remembering, (a) when lures were perceptually similar (the A–A’ condition), (b) when lures were mnemonically similar (the A–B’ condition), (c) when lures were dissimilar to any picture shown at study (the A–X condition). Example picture-pairs are illustrated for each condition, all the A target pictures had been shown during the study phase. The B picture was shown during the target block in the study phase of the experiment. Neither A’, B’ nor X images had been shown previously in the study. [(** p < 0.001), (* p < 0.01). Error bars represent 95% C.I. – for clarity only positive or negative bars are shown; numbers represent trials assessed within each bin of confidence – bold figures represent remembered trials and italic figures represent trials associated with not remembering].
Table 2
Study 1. Results of 2 × 4 ANOVA by test-pair condition, examining trial level data for recognition accuracy (proportion correct) by remembering (no = 0, or yes = 1), across the four bins of confidence.
| STUDY 1 | MAIN EFFECT | F (df) | p VALUE | h2 | POWER | |
|---|---|---|---|---|---|---|
| Condition | A–A’ | Remembering | 4.56 (1,723) | 0.035 | 0.006 | 0.56 |
| A–B’ | Confidence bins | 5.2 (3,730) | 0.001 | 0.021 | 0.93 | |
| A–X | none | – | – | – | – | |
Figure 5
Moderation analysis. (a) path diagram: note that the hypothetical variable names (X, Y, W) are changed in the PROCESS macro for SPSS to match the variable names in the analysis (Hayes & Scharkow, 2013; Hayes, Montoya & Rockwood, 2017) where the moderator variable [(W) = remembering or memory vividness], can be dichotomous or continuous for the analysis, the independent predictor variable [(X) = accuracy], can be dichotomous or continuous for the analysis, and the dependent variable [(Y) = confidence], must be continuous for the analysis; (b) statistical model: this represents the interaction (XW) tested, see Hayes (2017).
Figure 6
Study 1. Column plots of results from the two moderation analyses (error bars represent 95% C.I.): (a) recognition accuracy (incorrect or correct) with confidence by remembering (no or yes); (b) recognition accuracy (incorrect or correct) with confidence by vividness at the mean ± 1 S.D., (low, mean, and high = 22, 51, and 80 respectively). The interaction (moderation) effect can be seen by comparing the columns for incorrect versus correct recognition in each case, i.e., representing the three-way interaction (a) between remembering, confidence, and recognition accuracy and (b) between remembered vividness, confidence, and recognition accuracy.
Figure 7
Study 1. Scatterplot of trial level data illustrating the correlation between the vividness of memory for the image recognised as old and confidence in the accuracy of the decision, overlaid is the linear regression (best fit) line.
Figure 8
Study 2. Example natural scene colour photographs used for both study and test phases, comprising pictures from each subject category of flowers, green landscapes and trees, mountains, and water, used for: (a) Experiment 1, in full colour – images were counterbalanced for subject category; (b) Experiment 2, half in colour and half partially desaturated for colour – images were counterbalanced for subject category and colour saturation.
Figure 9
Study 2. Protocol for Experiments 1 & 2: (a) Study phase. Relational encoding is required during trials – photographs are presented following a location shown as a cross on a circle; (b) Test phase. Photographs are re-presented, and the location of the associated cross must be recollected, judgments of vividness, accuracy and confidence are required during trials; (c) Source memory accuracy, measured based on radial error: the distance in ± degrees between the participant’s response (shown as a red circle) and the correct target location (shown as a cross on the circle).
Table 3
Study 2. Comparison of variables of interest: data are compared by experiment and assessed using unstandardized B coefficient and R2 change (N = number of trials).
| EXPERIMENT (N) | VARIABLE | M (SE) | F (df) | UNSTANDARDIZED B COEFFICIENT (SE) | p-VALUE | R2 CHANGE |
|---|---|---|---|---|---|---|
| Experiment 1 (1920) Experiment 2 (3120) | Vividness | 67.4 (0.6) 55.3 (0.4) | 187.1 (1,5038) | –12.31(0.86) | < 0.001 | 3.9% |
| Experiment 1 (1920) Experiment 2 (3120) | Confidence | 54.9 (0.7) 42.6 (0.6) | 205.5 (1,5038) | –12.3 (0.90) | < 0.001 | 3.6% |
| Experiment 1 (1920) Experiment 2 (3120) | Accuracy | 74.1 (0.6) 67.6 (0.6) | 56.67 (1,5038) | –6.55 (0.87) | < 0.001 | 1.1% |
Figure 10
Study 2, Experiments 1 & 2, box plots illustrating median values (horizontal black bands) and the associated violin plots of responses illustrating mean values (black squares) reflecting the effect of changing the visual vividness of the pictures on (a) remembered vividness, median = 74 and 63, respectively; (b) the distribution of confidence in the memory’s accuracy, median 60.5 and 38, respectively; (c) source accuracy, median = 88.3 and 80.6, respectively; (d) location error response frequency plots showing source memory performance for each degree of error from the target (0° to ± 180°, normalised by number of trials and number of participants).
Figure 11
Study 2. (a) combined trial level data from Experiments 1 & 2 illustrating the association between source accuracy and confidence as a calibration (CAC) plot of percentage accuracy averaged for each bin of confidence. The red line superimposed on the plot reproduces the linear regression line from the scatterplot (c) of confidence and accuracy; (b) trial level data from Experiments 1 & 2 illustrating the association between source accuracy and confidence by experiment as a calibration (CAC) plot of percentage accuracy averaged for each bin of confidence. (Error bars represent 95% C.I., numbers represent total trials for each bin of confidence); (c) combined trial level data from Experiments 1 & 2 illustrating the association between confidence and source accuracy as a scatterplot; (d) combined trial level data from Experiments 1 & 2 illustrating the association between vividness and source accuracy as a scatterplot; (e) combined trial level data from Experiments 1 & 2 illustrating the association between vividness and confidence as a scatterplot. (The best fit linear regression lines are superimposed).
Table 4
Study 2. Correlations between variables of interest (across all trials) within each Experiment. Data are compared using Pearson’s correlation coefficient (r) and variance (R2).
| EXPERIMENT | I.V. | D.V. | N | PEARSON’S COEFFICIENT (r) | p-VALUE 2-TAILED | R2 VARIANCE |
|---|---|---|---|---|---|---|
| Experiment 1 | Accuracy | Confidence | 1920 | 0.46 | < 0.001 | 20.8% |
| Experiment 2 | Accuracy | Confidence | 3120 | 0.48 | < 0.001 | 23.3% |
| Experiment 1 | Vividness | Confidence | 1920 | 0.65 | < 0.001 | 42.3% |
| Experiment 2 | Vividness | Confidence | 3120 | 0.76 | < 0.001 | 57.3% |
| Experiment 1 | Vividness | Accuracy | 1920 | 0.33 | < 0.001 | 10.7% |
| Experiment 2 | Vividness | Accuracy | 3120 | 0.39 | < 0.001 | 15.9% |
Figure 12
Study 2. Experiments 1 & 2 combined: results of the moderation analysis illustrated by simple slopes plots (of the interaction effect) showing the relationship between source memory accuracy and confidence, by vividness at the mean ± 1 S.D., (high, mean, and low = 89.9, 59.7, and 29.5 respectively).