EFFECTS OF AVERAGING DATA DURING STIMULUS GENERALIZATION.

MIGLER (1964) looked at stimulus generalization gradients.

The team used median scores to smooth the data.

They wanted clean curves, so they averaged across birds.

The median line looked smooth and orderly.

Single-bird graphs told a different story.

Each bird had its own peaks and dips that the median hid.

Feldman et al. (1999) saw the same trap with fixed-ratio pauses.

Means masked skewed distributions and overlap across ratios.

McSweeney et al. (1993) found the same flaw in periodic-schedule activity.

Averaging made it look like motivation shifted inside the interval.

Wolfe et al. (2023) extended the warning to visual analysis.

Steep trend and high variability—exactly what averaging can hide—lower inter-rater agreement.

Together these papers repeat one lesson: look at the raw data first.

When you graph client data, plot each session before you draw a trend line.

A flat average can hide a learner’s sudden jump or drop.

Check individual paths to pick the right intervention and avoid false alarms.