Trial-Based Functional Analysis
Trial-Based Functional Analysis is a variation on the traditional FA procedures whereby experimental control is demonstrated through paired sequences of control (Motivating Operation is absent) and test (Motivating Operation is present) conditions across potential functions. The percentage of trials in which problem behavior occurred, rather than rate of responding during a 10-minute trial as in the traditional FA, is then used in order to develop a functional hypothesis for the problem behavior. For more information about this procedure, see the following research article:
Bloom, S. E., Lambert, J. M., Dayton, E., & Samaha, A. L. (2013). Teacher-conducted trial-based functional analyses as the basis for intervention. Journal of Applied Behavior Analysis, 46(1), 208-218.
Latency
Latency is a temporal measure of elapsed time from some change in the environment to the occurrence of a specific behavior. For instance, the amount of time from a teacher's prompt to begin working on an assignment to the behavior of the student picking up his pencil and starting to work on the assignment is a measure of latency.
Why Add Latency to Trial-Based FA?
I first started considering the use of latency in data collection while conducting trial-based FAs as a result of analysis of behavior that seemed to be clearly socially reinforced, but very unclear whether the most powerful contingency was related to attention, access to preferred, or escape/avoidance. I could identify antecedent conditions within each function under which problem behaviors would occur. Looking at percentage of trials alone, I thought, would not produce precise enough data to identify differences between the functions. Consider the following data as an example:
As the example data shows, the escape and tangible conditions are very similar if considering percentage of trials alone - 60% to 70%. However, consider the average latency for each condition across trials in which problem behavior occurred:
The following latency data graphed across trials clearly demonstrates that the tangible condition consistently produced problem behavior quicker than the escape condition (seconds of latency are in reverse order on the vertical axis):
Better-Informed Intervention
In the end, an analysis is only as good as the success of the interventions it produces. In this example data, the intervention choices are not only informed by the functional hypotheses, but also by the latency analysis for each function. For instance, in the tangible condition, problem behaviors were experienced relatively consistently when transitioning from access to blocked access to preferred activities or items, with an average latency of 13 seconds. This short latency suggests antecedent/teaching strategies, such as an "Accepting No" protocol (see work by Carbone), in order to prevent problem behavior prior to the transition. If the average latency were 90 seconds, it may be appropriate to use a progressive interval reinforcement schedule with access to tangibles used as reinforcement. In the first case, there is very little time to intervene after a transition but before a problem behavior, so teaching strategies and contrived practice with transitioning would be the best choice. With a longer latency, however, a team could consider reinforcement strategies contingent upon a fixed interval of time completing a non-preferred task (Premack principle).
Further Investigation Required
The addition of latency recording to trial-based functional analysis is still just a theory being researched. In addition to my own data, I've been able to find the following research on the topic:
I hope to collect analysis and intervention data over several months to function as a case study. Beyond that, I believe more research is needed to compare results of a traditional FA to results of a trial-based latency FA and determine whether or not adding a measure of latency increases the external validity over that produced by a trial-based FA without the use of latency. Finally, as stated above, analysis is only as good as the success of interventions it produces. Therefore, intervention effects following a trial-based latency FA should also be studied.
Use the following link to access a data collection form that I created to use while conducting trial-based latency functional analysis: