Observational coding in family therapy process research.

Abstract

Observationally based coding systems represent one major source of information regarding the process of family therapy, the nature of change mechanisms, and the necessary components of intervention manuals. In this article, the authors introduce a two-dimensional framework for organizing the diverse approaches to observational coding that currently exist. These dimensions focus on the information used in identifying the coding unit and the degree of inference involved in assigning a code (i.e., meaning) to that unit. The goal in developing this framework is to provide a common basis on which observationally based family therapy process research initiatives can be developed, compared, and integrated.Family therapy process research has not only come of age-it is thriving! In addition to consistent representation in major journals, several major reviews and proposed models have dealt with issues of context (

Microanalytic coding versus global rating of maternal parenting behaviour

“…Gardner (2000) suggests that global ratings may provide more consistency in cross-situational settings compared to frequency counts and it is likely that a global rating system would better generalise across settings, as global coders are able to take into account the setting of the observation and devise a global impression, whereas frequency counts of behaviours in microanalytic coding are very much dependent on the actual base rates of behaviour (Alexander et al, 1995). While this means that microanalytically coded data may give a better indication of base rates of behaviour within a specific context, global ratings would be better able to pick up differences across contexts.…”

“…However, because of the detailed coding requirements, microanalytic coding is extremely labour intensive and costly, therefore restricting its use to research where considerable funding is available. Furthermore, microanalytic coding may not take into account the context as aptly as global ratings (Alexander et al, 1995), and therefore may not be able to describe whole relationships independently from the moment-to-moment observed behaviours (Carlson et al, 1987). For example, interval based coding (e.g., Sanders, 2000) does not lend itself well to sequential or dyadic analyses, although frequency based codes can allow this type of analysis (e.g., Eyberg et al, 2004).…”

“…In general, global ratings involve a summary judgment of behaviour on particular dimensions (Carlson et al, 1987) and require larger coding units and a higher level of inference (Julien et al, 1989). Microanalytic coding captures moment-to-moment behaviours by breaking data into small time units or chunks (Alexander et al, 1995;Carlson et al, 1987), and microanalytic codes range from interval based coding (e.g., Sanders, 2000) to ones which record every instance of behaviour (e.g., Eyberg et al, 2004). Microanalytic coding is suggested to be superior to global ratings due to its objectiveness and minimal requirement for judgement (Alexander et al, 1995;Carlson et al, 1987).…”

“…Microanalytic coding captures moment-to-moment behaviours by breaking data into small time units or chunks (Alexander et al, 1995;Carlson et al, 1987), and microanalytic codes range from interval based coding (e.g., Sanders, 2000) to ones which record every instance of behaviour (e.g., Eyberg et al, 2004). Microanalytic coding is suggested to be superior to global ratings due to its objectiveness and minimal requirement for judgement (Alexander et al, 1995;Carlson et al, 1987). However, because of the detailed coding requirements, microanalytic coding is extremely labour intensive and costly, therefore restricting its use to research where considerable funding is available.…”

Personality and mental health as predictors of rater bias in observational data

“…Microlevel coding systems (e.g., Hill & Stephany, 1990) assess behavior by attending to overt, observable actions on the part of the research participants (or targets); such behaviors might include, for example, head nods or eye shifts. In these systems, interactions between the participants are usually broken into discrete units, based upon predetermined time intervals or speaking turns (Alexander et al, 1995). The trained observer then classifies the types of interactions that occur within each unit.…”