Aging and intraindividual variability in performance: analyses of response time distributions.

Myerson et al. (2007) asked if older adults are really more inconsistent than younger adults. They tested 48 people on a simple computer task. Half were 18-30 years old. Half were 65-80 years old.

The team looked at every response-time, not just the average. They used math tools called ex-Gaussian models to pull apart the whole distribution. This let them see how much each person wiggled around their own speed.

Once they held response speed constant, age differences vanished. Slower responders were more variable, whether they were 20 or 75. The link was so tight that knowing a person's average speed predicted their scatter.

In plain words: slow people look inconsistent, but age itself does not add extra noise.

da Silva et al. (2020) also used fine-grained computer timing. They found that boys with Duchenne Muscular Dystrophy were slower, not less accurate, than peers. Both studies show that speed, not age or diagnosis, drives the visible difference.

Cook et al. (2020) warn that sloppy measurement can fake variability in single-case data. Joel's paper gives you the tool: always model mean speed first, then check if any extra scatter remains.

Aydin et al. (2022) push for effect-size metrics that weight clinical goals. Their PCES approach pairs well with Joel's finding—if speed explains most variability, set your performance criterion around speed targets, not around age norms.

Stop labeling older clients as "inconsistent" when they are simply slower. Plot each person's response-time distribution, control for their mean speed, then decide if extra variability is left to treat. Use speed-based goals instead of age-based ones. This keeps your expectations fair and your interventions focused on what actually changes.