Sixty-six years ago, B.F. Skinner proposed that near-miss events might reinforce continued play on slot machines, and despite some inconsistencies in the experimental literature, belief in this “near-miss effect” has endured. The present manuscript reexamines the evidence on this topic and argues that, even though some near-miss research has produced results that appear to support the idea that near misses prolong gambling behavior, several issues render these findings problematic.
The first is that, whereas conventional chained procedures that successfully produce conditional reinforcement have a logical contingency between the putative conditional reinforcer and subsequent unconditional reinforcer, the classic slot machine provides random outcomes. For example, in the three experiments of Kassinove and Schare (2001), near-miss stimuli were manipulated so that they had either a 15%, 30%, or 45% chance of occurring for each participant, but after extinction, no group differences were observed on persistence measures or self-reports.
Another issue is that most of the recent studies on near-miss effects have examined measures other than behavioral persistence, such as physiological reactions or response latencies, and few have explicitly shown that near-miss events prolong gambling behavior. Moreover, many of the studies have not tested a key question: whether or not near-miss stimuli really do have conditionally reinforcing properties. A more direct approach to addressing this question might be to examine whether or not the same app instances that are warmed up in a staging slot can be re-deployed to production after a swap operation.