| Sumario: | Previously conducted research has indicated that the "between observer reliability, for highly skilled observers, by session, by scoring category for initiated behavior is acceptably high. For the self-self observer reliability to be lower, it would require essentially, that the non-systematic errors of two observers consistently be in the same direction, which is relatively unlikely. Adequate tests of self-self observer reliabilities will have to wait until series of standardized materials, such as films, are produced for scoring and then rescoring (test-retest). It would appear, thus, that researchers may plan to utilize formal scoring techniques such as Bales' Interaction Process Analysis with a reasonable confidence that training will produce observers who are reliable, or consistently arbitrary, in their scoring. Data were collected by time unit, permitting collation under several plans. The "actual behavior" of each subject consisted of four units of six minutes each for each of four sessions. The stability of subjects was examined by two plans. The plan one correlations are ordinarily lower than the plan two correlations. This may be interpreted as evidence for the hypothesis that a person will be more consistent, given the same amount of time, when interacting with the same individuals than when interacting with different individuals.
|
|---|