Supplemental Materials
Incidental Learning of Trust: Examining the Role of Emotion and Visuomotor Fluency
by J. W. A.Strachan et al., 2016, Journal of Experimental Psychology: Learning, Memory, and Cognition
Experiment S1: Liking Ratings
To support our interpretation of the results of Experiment 3, we report the results of an additional experiment carried out using different stimuli, run by a different experimenter, and using a different subject pool. These data were collected while electroencephalographic (EEG) recordings were made, and so cannot be directly related to the main results in our article. However, a similar EEG experiment by the same experimenter using trustworthiness ratings has been shown to be effective in eliciting reliable changes in judgements (Manssuer, Roberts, & Tipper, 2015).
Methods
Participants
Twenty-two participants (16 female, mean age 20 years) from Bangor University volunteered for this study. All procedures were approved by the Bangor University ethics committee.
Stimuli, Design, and Procedure
Eight male and eight female face stimuli were taken from the Glasgow Unfamiliar Face Database (Burton, White,McNeill, 2010) and the Utrecht Face Database (pics.psych.stir.ac.uk). The faces were selected and grouped into A and B groups based on equivalence in happiness, trustworthiness, and liking judgements given by a group of 16 independent observers who had a mean age of 21 (SD = 5) and were mostly female (N = 14) and right handed (N = 15). Happiness (M = 77.5, SD= 11.8) was rated from neutral (0) to very happy (+100), trust (M = 8.9, SD = 9.6) was rated on a scale from very untrustworthy (−100) to very trustworthy (+100), and liking (M = 0.6, SD = 14.0) was rated from dislike (−100) to like (+100). As evidenced by the happiness ratings, these faces were preselected to show slight smiles (which typically show a profile of trustworthiness changes similar to the full smiles used in Experiment 2), and so this is a further distinction from our main experiments.
In all phases of the experiment, the faces were at a size of 527 × 685 pixels and the objects were 350 × 263 pixels. The experiment was run using E-Prime 1.0 (Psychology Software Tools, Inc.) with a 1,920 × 1,080 screen resolution on a 24” Samsung SyncMaster BX2431 LED display, which was 342 × 569 mm in dimensions and had a 500 Hz refresh rate.
The procedure of this experiment was different from the procedures of experiments in the main article. These differences are discussed below.
Liking ratings. Ratings of each face were collected at the beginning and the end of the experiment. Each trial began when participants pressed the spacebar, at which point a fixation cross appeared for 1,000 ms followed by a directly gazing face for 1,000 ms and then a screen containing a visual analogue rating scale asking “How likeable is this person?” At this point, a cursor was visible on the screen and participants used the mouse to click along the scale at the point that represented how trustworthy they judged that person to be. The extreme left of the scale was labeled “Very Unlikeable,” and the extreme right of the scale was labeled “Very Likeable,” whereas in the main experiment these were represented with a “−” and a “+,” respectively.
Passive viewing phases. The main difference between this experiment and those in the main article was that, owing to EEG recordings being made, participants also underwent a passive viewing phase immediately before and immediately after cueing (i.e., between liking ratings and cueing). In these phases, participants pressed the space bar to initiate each trial, and then a fixation cross was presented for 500ms. After this, a face appeared for 750ms, followed by a screen reading, “Please Relax” for 1,000ms. There were 192 trials in total, and each face was presented aroundsix times in total.
Gaze cueing. During a gaze-cueing trial, participants initiated the trial by pressing the space bar. A fixation cross then appeared for 1,500ms followed by a face maintaining direct gaze for 1,500ms. The face then changed direction, and 500ms later an object appeared in either the cued or uncued location. This object disappeared as soon as a response was made or remained until 3,000ms had elapsed (this is different from our main experiments, where faces appeared for 2,500ms regardless of participants’ responses). When a response was made, the object disappeared and the face returned to direct gaze for 2,000ms. At the end of the trial, participants saw a screen reading, “Please Relax” for 1,000ms. There were five blocks of gaze cueing in total, with 32 trials in each block, and the same response keys (H and the space bar) were used with similar counterbalancing measures to Experiment 1.
Data Analysis
RT filters were applied in the same way as in Experiments 1, 2, and 3. No participants were removed on the basis of RT filters.
Average likeability ratings were calculated for each participant both at the beginning (pre) and end (post) of the experiment for both valid and invalid faces, and these scores were analysed in a 2 × 2 (time × validity) repeated-measures ANOVA. EEG data were not relevant to the current study and so are not examined here.
Results
Gaze Cueing
Over the course of the five blocks, RTs were lower to valid than invalid trials (see Figure S1). A 2 × 5 ANOVA found a main effect of validity (F(1,21)=7.72, p=.0113, η2G =0.01) and a main effect of block where responses were faster in later blocks than earlier (using Greenhouse Geisser correction for violation of sphericity assumption: F(2.53,53.16)=15.22, p=.0000, η2G =0.09) but no interaction (F(4,84)=1.21, p=.3144, η2G =0.00).
A similar 2 × 5 ANOVA using accuracy scores (coded as percent correct in each trial type in each of the five blocks, see Table S1) found only a main effect of block (F(4,84)=3.73,p=.0077, η2G = 0.05), as participants generally committed more errors at the beginning of the experiment than the end, but there was no overall effect of cueing validity on errors (F(1,21)=0.01, p=.9219, η2G = 0.00), and no interaction of validity and block (F(4,84)=0.40, p=.8058, η2G = 0.01). These results suggest that the attention cueing effect emerged primarily in RT measures rather than error rates, and that it remained stable over time.
Likeability Ratings
The changes in likeability ratings for the faces in Experiment 3 are shown in Figure 3c. A repeated measures ANOVA with rating time (pre- and post-experiment) and cueing validity of faces (valid and invalid) as within-subjects factors found no overall effect of time (F(1,21)=1.43, p=.2447, η2G =0.02), or of cueing validity (F(1,21)=0.10, p=.7605, η2G =0.00), and no significant interaction between the two (F(1,21)=0.17,p=.6823, η2G =0.00).
Table S1. Accuracy Rates (% Correct) in Response to Valid and Invalid Faces Across Five Blocks in Experiment S1
Trial / Block 1 / Block 2 / Block 3 / Block 4 / Block 5Valid / 94.89 / 95.17 / 96.59 / 97.16 / 97.44
Invalid / 94.03 / 96.31 / 96.31 / 97.73 / 96.59
FigureS1. Line graph tracking changes in reaction times (RT; ms) across five blocks in response to valid (dotted line) and invalid (solid line) trials in Experiment S1. Error bars show standard error.
Figure S2. Line graph tracking changes in liking ratings from the beginning to the end of the experimental session for valid (dotted line) and invalid (solid line) trials in Experiment S1. Error bars show standard error.