Empiricism

Empiricism makes claims inspectable

An empirical claim identifies the evidence behind it. Instead of “the plan is working,” report the measured response, observation period, comparison, implementation level, and the person’s assessment of fit.

The BACB BCBA Test Content Outline, 6th edition lists empiricism among philosophical assumptions underlying behavior analysis. Its other domains cover measurement, data display, experimental design, assessment, intervention monitoring, and evidence-based decisions.

ABAI’s overview of behavior analysis describes behavior analysis as a natural science that studies how biological, pharmacological, and experiential factors influence the behavior of humans and other animals. This broad framing supports gathering evidence across relevant domains.

Observation needs a defined method

State what counts, when observation starts and ends, who records it, and how missing data are handled. Select count, rate, duration, latency, magnitude, product, sampling, interview, or another method according to the question.

Representative sampling matters. Observing only the easiest staff member, quietest hour, or most successful setting can produce a polished but distorted result. Plan observations across conditions that affect the decision.

A fictional coaching example

Theo is a fictional new employee learning to prepare a communication system before home sessions. The practice defines readiness as the device charged, vocabulary page open, backup board present, and access method checked before the client arrives.

Across eight initial observed sessions, all four steps are ready in 3 of 8. After modeling, practice, and feedback, all four are ready in 7 of 10 later sessions. A second observer independently scores four later sessions and agrees on every component in 15 of 16 component judgments.

The result shows higher observed readiness during the later period and substantial agreement in a small sample. It leaves several explanations open, including training, practice, supervisor presence, and changes in scheduling. It also says nothing about whether the client could use the system effectively, which needs separate evidence.

Client report is evidence

Empiricism does not restrict evidence to outside observers. The person is the primary source for pain, comfort, preference, meaning, and many private experiences. Record their words or accessible communication and the conditions under which the report was gathered.

Keep self-report, caregiver report, clinician observation, device data, and record review distinct. Differences between sources can guide better questions. Averaging them into one score may erase useful disagreement.

Measurement quality affects conclusions

Reliability concerns the consistency of measurement. Validity concerns whether the measure represents the event or construct relevant to the decision. Two observers can agree perfectly on a definition that misses the person’s actual goal.

Check procedural integrity as well. A poor outcome under incomplete implementation offers weak evidence about the planned procedure. A favorable outcome with low integrity may point to another active variable.

Calibrate observers on the same examples and sample agreement across ordinary conditions. Agreement calculated only during easy sessions can exaggerate confidence. Report the number of sampled events, the calculation method, and any systematic disagreements.

Measurement systems can create incentives. A productivity target may encourage staff to record easy opportunities, delay difficult cases, or omit environmental barriers. Audit the distribution of observations and invite staff and client feedback about how collection changes the setting.

Empiricism reaches beyond one case

Clinical data show what happened in a particular context. Peer-reviewed research can inform likely mechanisms, risks, comparisons, and boundary conditions. Neither replaces the other.

A group result may have limited fit for one person. A striking case result may have limited generality. Integrate the best available research, case evidence, clinical expertise, context, and client values while keeping the sources visible.

Avoid data theater

More numbers do not automatically produce stronger evidence. Repeatedly measuring a convenient response can distract from a meaningful outcome. Dashboards can hide missing observations, changing definitions, and selected denominators.

Use a data dictionary, version definitions, preserve raw observations, and report excluded opportunities. Match collection burden to the decision. Stop measures that add work without improving care or understanding.

Negative, flat, and mixed results deserve the same preservation as favorable results. Keep corrections visible, explain deviations, and avoid selecting only graphs that support the preferred account. A transparent null result can prevent repeated burden and guide a better question.

Let findings change the plan

Empiricism requires responsiveness to disconfirming evidence. Predefine expected patterns and review points. When results differ, examine measurement, implementation, contextual change, competing explanations, and the person’s experience.

Document the decision and its evidence. A plan that continues unchanged despite clear contrary findings is data-rich but weakly empirical.