Baseline Measures
Socio-demographic variables
We recruited 151 potential participants in this study, 41 of them were excluded as they were not meeting the inclusion criteria (aged below 60) or rejected to come after consultation. The remaining pool of participants was evenly and randomly assigned to either experimental condition or control condition. 5 participants in both groups were excluded in analysis as they reported that they did not take any strategy during learning trials in the memory test.
Self-rated health is measured using a single item ranging from 1 = very good to 5 = very poor. Expenditure is captured by the perceptual item ‘Do you have enough money for daily expenditure?’. Participants were asked to rate this item ranging from 1 = very insufficient to 5 = more than enough. Social participation was measured by a list of 10 activities based on the proposed inventory [34]. Participants were asked to indicate the frequency of each activity on a 6-point Likert-type scale ranging from 0 = never to 5 = always within the past month. The average score of this scale was used for analysis, with greater score indicating higher rate of social participation. No significant differences between groups was found, socio-demographic information across primed group was presented in Table 1. Age, gender, education years and health status would be taken as covariates for further analyses.
Montreal Cognitive Assessment (MoCA)
MoCA is used as a brief and potential screening tool for detecting Mild Cognitive Impairment (MCI) and Alzheimer’s Disease (AD) has been validated in Hong Kong. A sum score of less than 22 is considered as a cut-off in Hong Kong [36]. All participants passed this cut-off (M = 27.55, SD = 1.46). Results showed that there was a marginally significant difference between priming intervention group on MoCA score [t (103) =1.79, p = .076], the MoCA score was controlled as covariate for further analyses.
Priming Intervention
As the priming intervention was framed as a reaction task, the percentage of correct hits and response time will be shown at the end of the task. The individualized SOA might act as a critical factor in affecting the priming intervention and its reaction time for hitting the targets. Referring to Table 2, there were no differences found in the manipulation of personalized SOA [t (103) = -1.30, p = .198], the percentage of correct hits of target during priming [t (84.01) =.191, p = .061] as well as the reaction time for hitting each priming target [t (103) = -.30, p = .766].
Emotional Word Reaction Time
As discussed, it is reasoned that the time responding to negative words would be shortened if negative primes are activated in experimental group. Our finding of average reaction time to emotional words supported this argument after controlling baseline reaction time, MoCA score and socio-demographic variables. ANCOVA results suggested that experimental group significantly reacted faster to negative emotional words than did the control [F (1, 97) = 5.81, p =.018, ƞp2 =.06], indicating a possibility of activating negative age-stereotypes during priming intervention. More interestingly, they also rated significantly slower to positive emotional words compared to the control group [F (1, 97) = 18.52, p <.001, ƞp2 =.16].
Further analyses were conducted to investigate whether social participation would moderate the effects of priming interventions by controlling the baseline reaction time, MoCA score and other co-variates of individuals. Referring to Table 3, results revealed that there was a significant main effect of priming intervention on reaction time to positive emotional words after controlling all co-variates (B = 438.14, t =3.14, p =.003, CI =153.27, 723.00). As shown in Figure 3, the interaction effect of priming intervention and reaction time to positive emotional words by social participation was found to be significant (B = -109.62, t =2.04, p =.045, CI =-216.57, -2.66), suggesting that participants in experimental condition who were relatively less socially active showed a longer reaction time in rating positive emotional words. All the regression weights were found to be non-significant in the relationship between experimental condition and reaction time to negative emotional words as well as the moderation analysis.
Memory Performance in HKLLT
Scoring and clinical findings of the HKLLT could be retrieved from previous studies [37]. The mean scores of each learning trial, total learning scores, 10-minute and 30-minute delayed recall as well as discrimination score of recognition task are shown in Table 4. Controlling the baseline MoCA score and other co-variates, control group learnt significantly more words in total learning trials [F (1, 98) = 15.01, p <.001, ƞp2 =.13], recalled more words in 10- and 30-minute delayed recall [F (1, 98) = 46.86, p <.001, ƞp2 =.33; F (1, 98) = 48.14, p <.001, ƞp2 =.33] and performed better in recognition task than the experimental group [F (1, 98) = 14.55, p <.001, ƞp2 =.13]. Furthermore, the rate of forgetting in the first 10 minute was computed by the suggested formula as [(10-minute delayed recall – learning trial 3)/learning trial 3 x 100%] [37]. ANCOVA results indicated that experimental group made more intrusion errors [F (1, 98) = 28.64, p <.001, ƞp2 =.23] and had a significantly higher rate of forgetting [F (1, 98) = 14.66, p <.001, ƞp2 =.13] than the control group.
Moderation analyses on memory performances
Table 5 showed the results of moderation analyses between experimental manipulation and memory performances. Results revealed that the interaction effects of priming intervention on total learning and total recall by social participation were non-significant (ps > .05). However, there was an interaction effect of priming intervention by social participation found in discrimination score in recognition trial (B = 7.83, t = 2.13, p = .036, CI = .52, 15.14), suggesting that participants who were relatively less socially active performed worst compared to other counterparts. In other words, those who were negatively primed but socially active did relatively equal to the control group as shown in Figure 4. Taking Johnson-Neyman (JN) technique to probe this interaction, the output identified individuals with .71 standard deviation on social participation (above which 24.76% of our samples) would not be affected by the negative priming intervention.