A comparison of the differential effects of three different baseline conditions within an ABA1B1 experimental design

The use of the ABA₁B₁ design over recent years in applied behavioural research has prompted many researchers to comment on its strengths and weaknesses (Kazdin, 1973; Kazdin and Bootzin, 1972; Liberman, 1972; Peck and Thorpe, 1971). One major problem is that many different types of non-contingent or no-reinforcement conditions have been used and that different types of baseline conditions may produce different effects. Despite the volume of research using a derivative of this design, surprisingly little is known about the effects of different baseline conditions.The baseline conditions most frequently used include the following: 1. No reinforcement (O'Leary and Becker, 1967; Hall et al., 1972; Aitchison and Green, 1974). 2. Threats (Phillips et al., 1971). 3. Reinforcing other types of behaviour (Ayllon and Azrin, 1965). 4. Equivalent amounts of reinforcement given in baseline and experimental conditions with all baseline reinforcement given non-contingently at the beginning or end of the phase (Wincze et al, 1972; Fernadez et al., 1973; Burchard, 1967). 5. Equivalent amounts of reinforcement in baseline and experimental conditions with baseline reinforcement given non-contingently at regular intervals or randomly throughout the phase (Foxx and Azrin, 1973; Baer and Wolf, 1970).Baer and Wolf (1970) have stated that: ‘Non-contingent reinforcement as a method of extinction has certain characteristics which may make it the method of choice for some experimental designs: especially those designs in which the adult offers reinforcers to a child’. They feel that this method is to be preferred because the total amount of reinforcement can be held constant over all conditions and thus will not be a confounding variable between conditions. However, as Skinner (1948) demonstrated, rate of behaviour can be changed by non-contingent reinforcement when ‘superstitious’ causal relations appear between behaviour and presentation of reinforcement. In effect the rate of behaviour increases because it has accidentally occurred before the presentation of reinforcement and so a connection is established whereby the probability of the behaviour is increased despite the fact that it is not causally related to reinforcement. Morse and Skinner (1957) have said about superstitious conditioning ‘Such effects must always be allowed for in designing experiments on complex behaviour’. Catania and Cutts (1963) and Yelen (1971) have shown that such superstitious conditioning can occur with human subjects.The present experiment employed three groups each trained under the same conditions but with different baseline conditions—1, 4 and 5 above. A multiple baseline design was used in which all possible responses were recorded in a four choice situation so that any accidental contingencies established would be evident.