Constructing Balanced Sessions of Discrete Trials: Demonstration Software
Research funded by Eunice Kennedy Shriver National Institute of Child Health and Human Development P30HD004147; National Institute of Mental Health R01MH090272; National Institute on Deafness and Other Communication Disorders R42DC010365
Sessions used in assessing, training, and maintaining conditional discrimination performances typically involve many discrete trials on which instances of the discrimination are presented (e.g., for matching a circle to another circle rather than a square or matching a printed word with its corresponding picture from a set displayed). The Balanced Trials Generator software described by Gerard, Mackay, Thompson, & McIlvane (2013) rapidly and automatically constructs trials for a variety of these sessions, thus saving the experimenter or teacher extensive time typically involved in session preparation.
Conditional discrimination procedures commonly involve trials in which a sample stimulus is presented together with three or four stimuli as choices (comparison stimuli) that appear in different response positions across trials. The purpose of creating balanced sessions is to provide the student or experimental participant with conditions that do not lead to performance based on irrelevant aspects of the display or procedure (e.g., selections based on the position in which correct stimuli had appeared rather than the stimuli themselves).The software arranges the stimuli evenly across available response positions while ensuring that:
• The correct stimulus changes unsystematically across trials
• The position of successive correct stimuli changes unsystematically across trials
Demonstrations of the Balanced Trials Generator software for constructing sessions of many trials are available. It allows users to specify stimuli to create their own sessions. Demonstrations of three types of sessions can be created. Baseline sessions involving multiple trials of an already established performance, with either 3 or 4 response positions, can be constructed. The third type is a teaching session that utilizes a training technique called exclusion (McIlvane and Stoddard, 1981).
The large table below illustrates output from construction of a baseline session (labeled BA) using the Balanced Trials Generator software in which the student’s task is to match a sample stimulus (e.g., CAT in Trial 1) with the identical stimulus from a group of three comparisons (CAT, DOG, FLY). These stimulus names must be entered by the teacher/experimenter into the boxes containing alphanumeric names on the demonstration website (A1, A2, B1, B2 etc.) before clicking on the box marked “3-comparison baseline” (Note that each click produces a new session). These three stimuli then are used in arranging displays that change unsystematically across trials. Sessions involving other specific stimuli (e.g., numerals, colored patches, forms) may be constructed with names supplied by the user. The generic alphanumeric names may also be used.
The output may be printed and used by a teacher to guide the arrangements of stimuli for each trial. The same sheet may be used as a manual score sheet that shows the student’s selection on each trial. Teachers and experimenters who use a computer to display stimuli and record responses must transcribe the contents of the output table into the particular software to be used. Note that two tables, which are printed after the detailed trial-by-trial session list below, provide summaries describing
- the distribution of stimuli as correct comparisons across the available response positions
- the distribution of stimuli as incorrect comparisons across available response positions.
This is an international Beta test of the software, which can be used to teach arbitrary relations between sample and comparison stimuli (e.g., pictures as samples, printed words as comparisons) as well as identity matching. To do so, it should be noted that the number of stimuli must be larger than the number of comparisons to allow specification of the different samples.
Number of stimuli: 3
Number of comparisons: 3
Number of response positions: 3
Baseline trials (BA): 36
+ = correct stimulus
Session Constructed by Balanced Trials Generator
Distribution of correct stimuli across comparison positions:
|1 ||2 ||3 ||Total|
|CAT||4 ||4 ||4 ||12 |
|DOG||4 ||4 ||4 ||12 |
|FLY||4 ||4 ||4 ||12 |
|TOTAL||4 ||4 ||4 ||12 |
Distribution of incorrect stimuli across comparison positions:
| ||1 ||2 ||3 ||Total|
|CAT||8 ||8 ||8 ||24 |
|DOG||8 ||8 ||8 ||24 |
|FLY||8 ||8 ||8 ||24 |
|TOTAL||8 ||8 ||8 ||72 |
Gerard, C. J., Mackay, H. A., Thompson, B., & , McIlvane, W. J. (2013). Rapid generation of balanced trial distributions for discrimination learning procedures: A technical note. Journal of the Experimental Analysis of Behavior. Advance online publication. doi: 10.1002/jeab.58
McIlvane W. J. & Stoddard L. T. (1981). Acquisition of matching-to-sample performances in severe mental retardation: Learning by exclusion. Journal of Mental Deficiency Research, 25, 33 – 48.