In the behavioral sciences, significant emphasis is placed on observation. A behavioral scientist aims to confirm their hypothesis with behavioral evidence. Given this observational imperative, a research study typically requires capturing phenomenal streams (video recordings) of the behavior of the study's participants, which must be observed and annotated in accordance with a rigorous specification of observational criteria.
- phenomenal streams (continuous / unstructured)
- observations / annotations (segmented / structured)
In other words, the data collected in the course of a study consists in part of annotated streams of participant behavior. For cohort studies, these annotated streams are often called sessions. This is typically supplemented with various attributes and measures collected for each participant. These attributes and measures permit various participant and session groupings.
Note that the captured video of partipant behavior is an information channel that requires filtering, segmentation, and decoding on the part of the researcher, whose observations are re-encoded in the form of annotations. The annotations supervene over the raw-data stream.
Let's also note the following ...
Observations made about human behavior are always matters of judgement and interpretation. These assessments can always be re-evaluated. Was this subject producing an iconic gesture or merely aborting a reflex movement prior to a sneeze?
Further, these assessments can always be refined. For example, a researcher may want to code the verbal argument roles of utterances to see how they co-occur with certain gesture types after noticing some interesting patterns across instances.
Once a behavioral stream has been annotated (and perhaps re-evaluated and refined), the researcher can then use those annotations to query, filter, and transform particular slices of observed behaviors.
The data collection and annotation process required to undertake a systematic study of some form of human behavior is laborious and expensive.
This may all seem very obvious to the behavioral researcher, but I think it's important to highlight how behavioral research data is actually collected and annotated, because this is what makes behavioral datasets unique.