Task performance. Participants accuracy in the short-term memory task was better than chance in both conditions (single-first: \({t}_{48}=89.28, p<.001, d=12.75\), BF = 14600\(\times {10}^{48}\); multiple-first:\({t}_{48}=17.98, p<.001, d=2.57, BF=6819\times {10}^{14};\) Fig. 3). Performance in the single-first condition was better (\({t}_{48}=9.03, p<.001, d=1.29,BF=1404 \times {10}^{5} )\), and slower (\({t}_{48}=8.81 p<.001, d=1.26, BF= 6872\times {10}^{5})\) than in the multiple first condition. In the single-first condition, accuracy levels were higher when a familiar face was the target of the short-term memory task (repeated in the consecutive displays) than when it was an unfamiliar face, without significant differences in reaction times (see supplementary materials for further details; Figure s1).
Fixations analysis.
The duration of the first fixation was examined using repeated measures analysis of variance on the effects of memory task (encoding\retrieval) * familiarity (familiar\unfamiliar) * display (single\multiple). No significant difference was found either between encoding and retrieval (memory-task main effect: \({F}_{\text{1,47}}\)=1.53,p=.222, \({\eta }_{p}^{2}=.03, B{F}_{inclusion}= .023\)), nor between single and multiple displays (Display main effect: \({F}_{\text{1,47}}\)=.352,p=.556, \({\eta }_{p}^{2}=.007, B{F}_{inclusion}= .19\)), nor between familiar and unfamiliar faces ( Familiarity main effect: \({F}_{\text{1,47}}\)=.921,p=.342, \({\eta }_{p}^{2}=.019\), \(B{F}_{inclusion}\)= .08). Only the interaction between memory task and display was significant, but did not yield higher likelihood than the null model: (\({F}_{\text{1,47}}\)=5.93, p=.019, \({\eta }_{p}^{2}=.112\), \(B{F}_{inclusion}\) =.42). The rest of the interactions, between memory-task and familiarity (\({F}_{\text{1,47}}\)=2.63,p=.111, \({\eta }_{p}^{2}=.053, B{F}_{inclusion}=.07\)), and the display and familiarity (\({F}_{\text{1,47}}\)=1.97,p= .167, \({\eta }_{p}^{2}=.04, {BF}_{inclusion}=.05\)) were not significant. Neither was the three-way interaction between the memory-task, display and familiarity (\({F}_{\text{1,47}}\)=0.07,p=.794, \({\eta }_{p}^{2}=.001, B{F}_{inclusion}= .016\)) (see Figure s4 in supplementary materials).
Next, we examined the duration of the second fixations (see Fig. 4). All three main effects were significant: the memory-task (\({F}_{\text{1,47}}=16.67, p< .001, {\eta }_{p}^{2}=0.262, B{F}_{inclusion}=13961),\)display (\({F}_{\text{1,47}}=26.22, p< .001, , {\eta }_{p}^{2}=0.358. B{F}_{inclusion}=916419\)), and familiarity \({F}_{\text{1,47}}=5.95, p= .019\), \(, {\eta }_{p}^{2}=0.112, B{F}_{inclusion}=3.518\). The interaction between the display type and memory-task (\({F}_{\text{1,47}}=17.834, p< .001\), \(, {\eta }_{p}^{2}=0.275, B{F}_{inclusion}=501.67\)), and between the display and familiarity were significant (\({F}_{\text{1,47}}=10.44, p= .002, , {\eta }_{p}^{2}=0.182, B{F}_{inclusion}=5.94)\), but the interaction between memory and familiarity was not (\({F}_{\text{1,47}}=.726, p= .399, {\eta }_{p}^{2}=0.015, B{F}_{inclusion}=.74\)). The three-way interaction between the memory-task, display and familiarity was not significant (\({F}_{\text{1,47}}<.001, p= .978\), \(, {\eta }_{p}^{2}=<.001, B{F}_{inclusion}=.784\)).
Planned contrasts to compare the effect of familiarity when multiple faces were displayed were not significant for either encoding (\(\psi =4.16, {t}_{175}=0.44,p=.658, BF= .086\)) or retrieval (\(\psi =4.81, {t}_{175}=0.51,p=.609, BF= .268\)). On the other hand, when a single face was displayed, significantly longer fixation durations were observed on the familiar face (Encoding: \(\psi =19.14, {t}_{175}=2.04,p=.043, BF=3.44\), retrieval: \(\psi =28.63, {t}_{175}=3.05,p=.003, BF=1.97;\) see Fig. 4).
Next, we examined the mean fixation duration across all fixations on the stimulus (Figure s5 in the supplementary materials). The main effects of memory task (encoding\retrieval): (\({F}_{\text{1,47}}\)=8.11,p = .007\(,{\eta }_{p}^{2}=.157, B{F}_{inclusion}=60.1\)) and display (multiple\multiple: \({F}_{\text{1,47}}\)=24.57,p = < .001, \({\eta }_{p}^{2}=.343, B{F}_{inclusion}=15683.49\)) were significant, but not familiarity (familiar\unfamiliar: \({F}_{\text{1,47}}\)=3.725,p=.06, \({\eta }_{p}^{2}=.07, B{F}_{inclusion}=.37\)). The interaction between memory task and display was significant: (\({F}_{\text{1,47}}\)=10.93,p=.002, \({\eta }_{p}^{2}=.19, B{F}_{inclusion}=38.3\)) but the rest of the interactions were not (memory-task x familiarity \({F}_{\text{1,47}}\)=1.83,p=.201, \({\eta }_{p}^{2}=.03, B{F}_{inclusion}=.36\); display x familiarity \({F}_{\text{1,47}}\)=0.22,p= .64, \({\eta }_{p}^{2}=.005, B{F}_{inclusion}=.24\)). Neither did the three-way interaction between the memory task, display and familiarity (\({F}_{\text{1,47}}\)=1.02,p=.318, \({\eta }_{p}^{2}=.02, B{F}_{inclusion}=.17\)).
Dwell time analysis.
The fixation duration analysis shows that second fixations were elongated on familiar relative to unfamiliar faces only when a single face was displayed, supporting the less exploratory need hypothesis. If the effect of fixation duration disappears in the multiple display because there are other, unfamiliar, faces to explore, this should be evident also in the overall dwell time gaze is directed to the faces. We expected to find shorter overall dwell times spent on the familiar face during encoding as there is higher need to encode the other unfamiliar faces in order to succeed in the short-term memory task. In addition, we expect the same phenomena but with a lesser effect during retrieval too, as there is a greater need to explore the other faces in the display which are unfamiliar.
The main effects of familiarity (\({F}_{\text{1,47}}=0.004, p=.948, {\eta }_{p}^{2}<.001, B{F}_{inclusion}= <.001)\) and memory task (\({F}_{\text{1,47}}=2.242, p=0.141, {\eta }_{p}^{2}=.04, B{F}_{inclusion}= <.001)\) were insignificant, however the interaction between the two was significant (\({F}_{\text{1,47}}=89.576, p<.001, {\eta }_{p}^{2}=.65, B{F}_{inclusion}=1407\times {10}^{10};\) see Fig. 6). Direct contrasts revealed significant differences between the total dwell time on familiar and unfamiliar faces within each memory task: (Encoding: \(\psi =170.9, {t}_{90}=6.\)02\(,p<.001\), BF = \(2341\times {10}^{4}\), retrieval: \(\psi =189.9, {t}_{90}=5.92,p<.001, \text{B}\text{F} = 1348.9\)). This was evident also when excluding the time following the key press (see supplementary material Figure s3).
A similar pattern was reflected also in the number of fixations. Fewer fixations during encoding of a familiar face were observed, in comparison to unfamiliar faces, as shown by the three-way interaction between the memory-task, display and familiarity was significant (\({F}_{\text{1,47}}=44.04, p< .001, {\eta }_{p}^{2}=.484, B{F}_{inclusion}=5521\times {10}^{6}\)). For further details on the number of fixations analysis see supplementary materials (Figure s2).