This page is part of the History of Animal Training and Behavior Analysis section.
Skinner described shaping indirectly in The Behavior of Organisms in 1938, but it wasn't until 1943 that he truly explicitly shaped his first pigeon and formally described the process that animal trainers had been using for centuries.
(Peterson 2004 - Skinner discovers shaping)
Shaping is:
•The answer to “what if he/she doesn’t already have the behavior in their repertoire?”
•Perhaps one of the most basic technologies in behavior analysis, but also one of the most difficult to implement or to train effectively.
Technically, Shaping is:
•Differential Reinforcement of Successive Approximations
•Changing some dimension of existing behavior (form, distribution, etc.)
•Combining existing behavior into a novel or unusual sequence or composite
•The genesis of brand new movement cycles
•Much more complicated than simple reinforcement schedules – may include:
•Dynamic and static environmental arrangements
•Using previously conditioned (or unconditioned) stimuli to make shape-able behavior more likely
•Efficient fading of above stimuli before stimulus control transfers to shaped behavior
•Arranging against or restricting alternate behaviors and setting events for those behaviors
–Contingencies placed sequentially on different dimensions
–Contingencies placed simultaneously on different dimensions
–Changing contingencies within dimensions
•Requires responsiveness on the part of the implementer
•Happens naturalistically and unintentionally all the time (Schaefer 1970)
•One of few technologies available for the generation of behavior not already in the repertoire
Points about shaping revealed or examined by behavior analysis:
•Shape by adducting or chaining components of the target behavior, and train and use useful behaviors that may make the occurrence of the first approximation more likely
•Baer, Peterson & Sherman (1967), Ferguson & Rosales-Ruiz (2001), Lovaas (1966), Kerr, Meverson & Michael (1965)
•You can go forward or backward through the target movement cycle while shaping
•Powell & Kelly (1975), Ferguson & Rosales-Ruiz (2001), Pear & Legris (1987)
•What is the “ideal” step size or change along the shaped dimension to use from approximation to approximation? Too large a step may result in extinction, too small a step may result in movement that is too slow or loss of variability
•Eckerman (1980), Powell & Kelly (1975), Pear & Legris (1987), Koegel et al. (1988), Kerr, Meverson & Michael (1965), Galbicka (1994)
•When or after how many reinforcements should one move to the next approximation?
•Eckerman (1980), Hartman & Hall (1976)
•Variability is necessary for the shaping process
•Miller & Neuringer (2000) showed that variability itself is susceptible to shaping
•When shaping, keeping a high density of reinforcement is important
•Kerr et al. (1965), Koegel (1988)
•Shape along a single dimension or multiple dimensions?
•When shaping along a single dimension, variation along unshaped dimensions can be selected in the shaping process
•Even if change is not desired along a specific dimension, a “no-change” contingency may still need to be placed on it
•Dimensions such as latency, IRT, duration, etc. are usually important
•Pear & Legris (1987), Hartman & Hall (1976), Miller & Neuringer (2000), Koegel et al. (1988)
•Sometimes an unintended approximation will be naturalistically reinforced and will compete with the desired shaping path.
•(Powell & Kelly 1975)
•Target behaviors need to be relevant
•(Wolf et al. 1964)
Always asses in the natural environment during treatment to test effectiveness of treatment and to re-assess the propriety of targets
(Fleece, et al. 1981)
How has Shaping been implemented in Behavior Analysis?
•Functional Communication Training
•Natural Language Paradigm (Laski 1988)
•Percentile/automatic programming
•Changing criterion design (Hartman & Hall 1976)
•Communication/verbalizations (Kerr Meverson & Michael 1965)
•Increasing interests using approach (Riccardi et al 2006)
References
•Should we go forward or backward through the target movement cycle?
–Powell & Kelly had success with each
–Ferguson & Rosales-Ruiz, Pear & Legris used forward methods