Most of the participants were members of the Institute of Active Ageing of the Hong Kong Polytechnic University. The experiment was first promoted by the institute using its internal email system and participants were recruited through phone calls. Other participants were recruited by referrals through snowball sampling. This study targeted individuals aged 60 years or above. The recruited participants were generally well-educated, physically healthy (self-reported), mentally healthy (using The Montreal Cognitive Assessment (MoCA)) and were able to read or speak fluent Chinese/Cantonese. A prior power analysis was performed and determined that a total sample size of 92 subjects was required to have 95% power for detecting an effect size of .38 (as reviewed in meta-analysis) when taking α = .05. Thus, 110 participants were drawn from the recruiting pool to take part in the study to avoid data loss.
Participants were blinded to the assignment of conditions and were randomly assigned, according to their number of participation in the trial using simple randomization, to either the neutral word priming group (control group) or the negative age stereotype priming group (experimental group). The research team generated random numbers for the random assignment procedure. Participants were given a $200 (~USD$25) supermarket coupon after completion of the experiment.
The present study was a between-subject design where participants were randomly assigned to one of two conditions.
The priming task is adopted and modified from a previous study [5, 26]. It was performed using E-prime 2.0 software . To ensure the primes flashed on screen were beyond awareness, a similar adjustment procedure was used (see ). The individualized stimulus onset asynchrony (SOA) was determined in each trial block in which a total of ten neutral words was flashed either 1 cm above or 1 cm below the cross-point (center) in each trial. The participants were required to focus on the cross-point and to respond to the computer by pressing the designated keys as quickly and accurately as possible. Patterned masks (rows of at signs: @) were used before and after each flash of a word. After the trial block, participants were asked to try to report any words viewed during each trial. Their SOA were reduced or enhanced whenever 2 words or above were correctly reported or no single word could be reported, respectively. The priming SOA for this study ranged from 32 ms to 208 ms (M = 98.70 ms, SD = 48.60).
Negative age stereotype primes were taken from Levy’s study while neutral words were adopted from the most frequent words used in the Chinese context . To ensure the words are relevant to the Hong Kong elderly population, a list of 60 words (22 words are positive, 22 words are negative and 16 words are neutral) was generated, and then 16 participants between 50 and 69 years old rated each of the words according to its relevance to themselves in a 7-point Likert type scale ranging from 1 indicating ‘very positively related to you’ to 7 ‘very negatively related to you’ while the score of 4 referred to ‘irrelevant to you’. The negative words and neutral words rated at the highest frequency were used in the priming task.
This study extended the priming procedure in which 4 blocks of 40 trials were presented to the individualized SOA. In the negative age stereotype priming condition, 40 trials consisted of (i) 12 negative age stereotype primes (repeated once); (ii) 2 highly rated negative age stereotype primes (i.e., dementia and clumsy) (repeated twice); (iii) 4 selected neutral words (besides, sentence, moreover and even) (repeated once). More specifically, 12 negative age stereotype primes and 4 neutral words were presented twice and 2 highly rated negative age stereotype primes were presented four times, which constitutes 24 trials and 8 trials, respectively, out of 40 trials in each block. In the neutral condition, 4 blocks of the 40 neutral words were flashed randomly.
The priming procedure was intensified based on Levy’s (1996) priming paradigm as we would such as to exacerbate the priming effects as well as counterbalance the number of trials in each of the blocks to avoid fatigue and the tendency of pressing the same key.
During each trial, similar to the prior individualized procedure, the participant was asked to indicate whether the stimulus was flashed above or below the cross-point. The typical flow of each trial is shown in Figure 1.
After responding to the 40th trial in each block, four emotional words (two negative and two positive) were randomly presented to the participants and they were asked to rate whether the targeted words were positive or negative. As suggested by previous studies, individuals who were negatively primed would tend to respond to the negative words in a quicker manner, which could be seen as the activation of a negative stereotype at the subliminal level . The reaction time and correct rate of clicking the flashes were presented to the participant after finishing the whole task.
Montreal Cognitive Assessment (MoCA) was used as a baseline assessment. It is a 10-min test that evaluates several cognitive domains with a total score of 30. The Hong Kong version is validated and is available at the MoCA official website .
The HKLLT is a validated tool for assessing episodic memory for aged Chinese-speaking adults . It is also used for investigating differences among older adults with normal cognitive ability and older adults with mild cognitive impairment . A random control list was used in this study. It is comprised of 16 words formed by four categories: family member (grandmother), country (Chile), furniture (wardrobe) and vegetables (cucumber). All words were in random order such that no words within the same category were presented consecutively. Three attempts were presented to the participants and the total learning score (out of 48) was computed over three trials. It also involved 10- and 30-minute delayed recalls as well as a recognition task. The recognition task required participants to indicate in list of 32 items, where half of them were targets while half of them were foils, in a yes or no manner. A discrimination score was calculated as it considers both correct hits and false alarm errors.
Sociodemographic information and a scale measuring social participation were included in a questionnaire. Self-rated health was measured using a single item ranging from 1 (‘very good’) to 5 (‘very poor’). Expenditure was captured by the perceptual item ‘Do you have enough money for daily expenditures?’ 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 (see Appendix A) . 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 a higher score indicating a higher rate of social participation.
The experiment was done in the laboratory setting and it was divided into three sections: (i) prepriming test, (ii) the priming task & (iii) memory assessment. The experiment lasted for approximately 2 hours and its research flow is presented in Figure 2. Participants first provided informed consent followed by the MoCA and a simple visual acuity test using the “Tumbling E” Eye Chart. The simple visual acuity test was chosen to ensure participants did not possess severe problems with their eye-sight so that they were able to perform the priming task on the computer screen . Then, participants entered the trial session for determining their personal SOA before the priming stage. This stage was framed as a reaction time test in which they were asked to press the appropriate key as accurately and as quickly as possible using their individualized SOA. This deceptive explanation was also put on the information sheet in order to avoid contamination of the present priming task. Participants started the priming task according to the allocated randomized treatment.
The HKLLT was implemented immediately after the priming task. All task instructions were standardized and computerized according to the manual of HKLLT by using E-prime 2.0. During the first learning trial, participants first listened to the 16 target words and were asked to record the words aloud to the microphone. No feedback was given by the examiner or the computer. The procedure was repeated in the second and third learning trials. Participants were also asked whether they had been using any strategic methods to memorize the words.
After the learning trials in the HKLLT, participants were asked to fill out a questionnaire including sociodemographic information and other scales such as frequency of social participation. Without prior notification, the participants were asked to stop filling out the questionnaire and asked to recall the list of words again after 10 minutes and then 20 minutes (i.e., 30-minute delayed recall). The recognition task was then immediately performed. The remaining time was given for completion of the questionnaire if necessary. After signing the receipt of coupon collection, the debriefing session was given in which the research flow was explained. Participants were also asked not to reveal any details of the study.
Analyses were performed using SPSS, version 25. Multiple independent samples t tests and chi-square independence tests were used to investigate differences in sociodemographic characteristics and baseline assessments. Analysis of covariance (ANCOVA) was used to test group differences in priming manipulation and memory performance.
Since the 10-minute and 30-minute delayed recall scores were highly correlated (r = .89), we computed one composite score (total delayed recall) by adding the two delayed recall scores. The total learning score, total delayed recall and discrimination score in the recognition trial were taken as dependent variables while the average score of social participation frequency was taken as a moderator for the moderation analyses using the SPSS macro PROCESS (model 1) . This is taken for analyses since it allows researchers to use continuous variables for moderators as well as for probing the interaction, if any, by using the Johnson-Neyman (JN) technique.