Statistical inference for individual organism research: mixed blessing or curse?
Ditch the p-value—strong visual control tells the real story in single-case work.
01Research in Context
What this study did
Michael (1974) wrote a position paper, not an experiment. The author looked at how behavior analysts use p-values and said, 'These numbers can hurt us.'
The paper warned that t-tests and ANOVAs hide the very patterns we try to see in one-person charts.
What they found
The author found that statistical inference is a 'mixed blessing' for single-case work. Strong visual proof of control beats any p-value.
In short: skip the stats, trust your eyes and the steady line on the graph.
How this fits with other research
Revusky (1967) tried the opposite. That paper showed a way to reach p < 0.05 with only four subjects. Michael (1974) later said, 'Do not do this.'
Meyer (1999) and Iversen (2021) kept the fight alive. Both later pieces still argue against p-values, showing the field has not moved on.
DeHart et al. (2019) and Barnard-Brak et al. (2020) offer a middle path. They use mixed-effects or Bayesian N-of-1 models that keep each client’s data separate yet still give a number. These new tools answer J’s worry without going back to group stats.
Why it matters
When you write a report or plan a study, ask: 'Does a p-value add anything my graph already shows?' If the answer is no, leave it out. If reviewers demand a number, try the newer single-case friendly models instead of classic t-tests. Your data stay personal, your story stays clear, and you honor the 1974 warning while still looking modern.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Remove any t-test from your current draft and add one extra phase change line to show control visually.
02At a glance
03Original abstract
Descriptive and inferential statistics are described as judgemental aids, stimuli to which the scientist can more easily react than to his raw experimental results. The increasing emphasis on the significance test as the main judgemental aid utilized in experimental psychology is credited with several harmful effects on experimental practice. The area known as "the experimental analysis of behavior" has so far escaped most of these harmful effects, but now we see an increased interest in the development of appropriate significance tests for individual organism research. This interest is based on the view that it is not possible to effect adequate levels of experimental control with much human applied research, and that in such cases a significance test would be quite valuable as a judgemental aid, both of which points are considered to be essentially incorrect, and if accepted, potentially harmful.
Journal of applied behavior analysis, 1974 · doi:10.1901/jaba.1974.7-647