Okay, let's design a robust Randomized Controlled Trial (RCT) for a psychology study on a memory intervention. We'll use a hypothetical scenario: Investigating the effectiveness of a novel digital cognitive training program for improving working memory in healthy older adults.

Psychology RCT Design: Digital Cognitive Training for Working Memory

1. Define Independent and Dependent Variables Clearly

• Independent Variable (IV): Type of Cognitive Intervention.


• Levels:


• Experimental Group: Participants receive access to the Novel Digital Cognitive Training Program (e.g., a gamified online platform designed to target working memory processes like N-back tasks, digit span, spatial recall).


• Active Control Group: Participants receive access to an Active Control Program (e.g., a general knowledge quiz app, a simple puzzle game, or a digital diary/journaling app) that requires similar time commitment and screen interaction but is not designed to specifically target working memory. This helps control for the Hawthorne effect, time investment, expectation, and general engagement.


• Dependent Variables (DVs): Measures of Working Memory Performance.


• Primary Outcome Measures: A battery of standardized neuropsychological tests specifically assessing different facets of working memory.


• Verbal Working Memory:


• Digit Span Forward & Backward (WAIS-IV)


• Letter-Number Sequencing (WAIS-IV)


• Visuospatial Working Memory:


• Corsi Block-Tapping Test


• Spatial Span (WMS-IV)


• Secondary Outcome Measures:


• Everyday Memory Functioning: Self-report questionnaire (e.g., Memory Functioning Questionnaire - MFQ).


• Processing Speed: Select subtests from WAIS-IV (e.g., Symbol Search, Coding).


• Mood: Beck Depression Inventory-II (BDI-II), State-Trait Anxiety Inventory (STAI) – as potential confounds or mediating factors.

• Approach: Stratified Randomization.


• Why? Simple randomization can sometimes lead to imbalances in important baseline characteristics, especially in smaller sample sizes. For memory studies, factors like age, education, and baseline cognitive ability are crucial. Stratified randomization ensures these factors are evenly distributed across groups.


• Stratification Variables:


• Age: (e.g., 65-74 years vs. 75-85 years)


• Years of Education: (e.g., <12 years vs. 12-16 years vs. >16 years)


• Baseline Working Memory Performance: (e.g., based on a pre-screen measure or the initial composite score of the primary DVs, categorized into tertiles or quartiles).


• Procedure:


• After participants meet inclusion criteria and complete baseline assessments, they are grouped into "strata" based on their demographic and baseline cognitive characteristics.


• Within each stratum, participants are then randomly assigned (using a computer-generated random number sequence) to either the Experimental Group or the Active Control Group.


• This ensures that each group has a comparable proportion of participants across these critical variables.

• Assumptions:


• Power: 0.80 (80% chance of detecting a true effect if one exists).


• Alpha ($\alpha$): 0.05 (significance level, two-tailed).


• Effect Size (Cohen's d): This is the most crucial assumption and needs to be based on previous research or pilot data. For cognitive interventions, a medium effect size (Cohen's d = 0.50) is often a reasonable target, though some studies may aim for a smaller but still meaningful effect (e.g., d = 0.40) to be more conservative. Let's use d = 0.40 to ensure sufficient power for a potentially more subtle effect.


• Statistical Test: Independent samples t-test (comparing mean change scores between groups).


• Attrition Rate: Assume a 15-20% attrition rate over the intervention period.


• **Calculation (using G*Power or similar software for d=0.40, alpha=0.05, power=0.80, two-tailed t-test):**


• Required sample size per group is approximately 100 participants.


• Therefore, a total of 200 participants are needed before accounting for attrition.


• Adjusting for Attrition (e.g., 20%): $200 / (1 - 0.20) = 200 / 0.80 = 250$ participants.


• Final Target Sample Size: Recruit approximately 250 participants (125 per group) to ensure adequate power of 0.80 for detecting a medium-small effect size (d=0.40) at $\alpha$=0.05, even with up to 20% attrition.

• Inclusion/Exclusion Criteria:


• Inclusion: Healthy older adults (e.g., 65-85 years), able to read and understand English, access to internet/device, willing to commit to the intervention schedule.


• Exclusion: Diagnosed neurological conditions (dementia, stroke, Parkinson's), major psychiatric disorders (severe depression, schizophrenia), uncorrected sensory impairments, current use of medications known to significantly impact cognition, significant cognitive impairment (e.g., MoCA score < 26).


• Standardization of Procedures:


• Assessment Environment: All baseline and post-intervention cognitive assessments conducted in a quiet, distraction-free environment, ideally the same room for all participants.


• Test Administration: Trained and supervised assessors follow strict protocols for test administration and scoring.


• Intervention Duration/Intensity: Both groups commit to the same scheduled time commitment (e.g., 30 minutes/day, 5 days/week for 8 weeks). Software will track actual usage to monitor adherence.


• Instructions: Standardized written and verbal instructions for both intervention and control groups, emphasizing commitment and effort.


• Active Control Group: Crucial for controlling non-specific effects such as expectation of improvement, attention from researchers, time spent on a cognitive task, and basic digital engagement.


• Measurement of Potential Confounds:


• Baseline Assessments: Collect data on mood (BDI-II, STAI), sleep quality (Pittsburgh Sleep Quality Index), physical activity, nutritional habits, medication use, and self-reported health. These can be used as covariates in analyses or to check for group imbalances.


• Expectancy: Ask participants at baseline how much they expect to improve their memory.


• Engagement/Adherence: Monitor log-in data, completion rates, and duration of use for both intervention types.


• Statistical Control: Any baseline variables found to be unbalanced between groups despite randomization, or those theoretically linked to memory outcomes (e.g., baseline mood, processing speed), can be included as covariates in ANCOVA or regression models.

• Decision: Between-Subjects Design.


• Why?


• Irreversibility of Intervention Effects: Memory training, if effective, is designed to produce lasting changes. It's difficult to "undo" or remove the effects of cognitive training to then expose the same individual to a control condition without contamination.


• Carryover Effects: If participants experienced both conditions, the training from the first condition would likely carry over to the second, making it impossible to isolate the effect of the second condition.


• Practice Effects: While pre-post testing always has practice effects, a within-subjects design for this type of intervention would heavily confound results with repeated exposure to the nature of the tasks.


• Demand Characteristics: Participants exposed to both conditions would almost certainly guess the study's hypothesis, leading to stronger demand characteristics.


• Design Structure:


• Pre-test: All participants complete baseline cognitive assessments.


• Intervention Period: Participants are randomized to either the Experimental Group or the Active Control Group and engage with their assigned program for the specified duration (e.g., 8 weeks).


• Post-test: All participants complete the same cognitive assessments again.


• Data Analysis: Compare the change scores (Post-test - Pre-test) between the two groups.

• Goal: Maximize blinding to minimize bias. Aim for Double-Blinding as much as possible.


• Participant Blinding (Single Blinding):


• Participants will be informed that they are part of a study investigating "the effects of digital cognitive activities on brain function and cognitive abilities in older adults."


• They will not be told which specific program is hypothesized to be more effective, or if one is considered a "control." Both programs will be presented as potentially beneficial.


• The active control program will be designed to feel credible and engaging, similar to the experimental program in its digital interface and time commitment, so participants do not easily deduce their group assignment.


• Experimenter/Assessor Blinding:


• Outcome Assessors: These individuals must be blind to the participants' group assignment. They should not know whether a participant received the experimental or active control intervention. This is critical to prevent observer bias during test administration and scoring.


• Implementation: Assessors should not have access to randomization lists or participant group labels. Participants should be instructed not to discuss their assigned activity with the assessors.


• Intervention Administrators/Support Staff: If there are staff who set up the programs or provide technical support, they might be aware of the participant's group. If so, they should be different individuals from the outcome assessors and strictly instructed to avoid discussing the intervention type with participants or assessors.


• Data Analyst Blinding:


• The data analyst should be blinded to the group labels (e.g., data will be coded as "Group A" and "Group B" rather than "Experimental" and "Control") until primary analyses are complete.

• Demand Characteristics: Participants might infer the study's hypothesis (e.g., "this training should make my memory better") and unconsciously or consciously alter their behavior (e.g., trying harder on memory tests, reporting greater subjective improvement) to confirm the hypothesis.


• Minimization Strategies:


• Active Control Group (as discussed): This is the single most effective strategy. Both groups are engaged in a "cognitive activity," reducing the feeling of being in a "sham" or "nothing" group. This balances expectations.


• Blinding (as discussed): Preventing participants from knowing their specific group assignment or the exact hypothesis helps reduce the tendency to "help" the experimenter.


• Neutral Framing of Study Purpose: Frame the study broadly (e.g., "studying different types of digital activities and their impact on cognition") rather than focusing solely on "memory improvement."


• Standardized Instructions: Deliver all instructions, pre-test motivation, and post-test encouragement neutrally and identically to both groups. Avoid language that suggests one condition is superior.


• Anonymity and Confidentiality: Emphasize that individual performance will be kept confidential and will not be shared, which can reduce pressure to perform.


• Debriefing Questions (Manipulation Check): After post-testing, include questions like: "What do you think the purpose of this study was?" "Which group do you think you were assigned to?" "Did you feel that your activity was designed to improve memory?" This helps identify if participants guessed the hypothesis and allows for post-hoc analysis if a significant number did.


• Measure Expectancy: Include a baseline questionnaire about participants' expectations for memory improvement. If expectations differ significantly between groups, it can be controlled for statistically.


• Outcome Measures: Rely primarily on objective, standardized cognitive tests rather than subjective self-report, as objective tests are less susceptible to conscious bias from demand characteristics.

This comprehensive design incorporates best practices for RCTs in psychology, aiming for high internal validity and reliable findings regarding the efficacy of the memory intervention.