Exploring a concurrent-chains paradox: decreasing preference as an initial link is shortened

Experiments with pigeons and rats on concurrent-chains schedules examined a paradoxical effect reported by R. A. Preston and E. Fantino (1991). One schedule in the concurrent chain had a variable-interval (VI) 60-s initial link, and its terminal link was a 10-s delay to food. The other schedule had an initial link that ranged from VI 60 s to VI 2 s, and its terminal link was a 20-s delay to food. The paradoxical effect--a decrease in preference for the 20-s delay as its initial link was shortened--was found in some conditions but not in others. An analysis of response-reinforcer delays suggested that the paradoxical effect occurred in conditions in which responding on the short VI schedule almost always led to the 20-s delay, eliminating the possibility of switching to the alternative with the shorter delay.