Illusion of Control, Gambler’s Fallacy, or Neither? Testing Key Models of Gambling Psychology by Extending Langer and Roth (1975)

Cognitive biases about randomness, such as the illusion of control and the gambler’s fallacy, are central to the widely-used cognitive models of disordered gambling. In a highly-cited foundational paper, Langer and Roth (1975) found that participants thought they were better at predicting coin-flips when they were randomised to receive more wins at the start rather than the end of a 30-flip sequence. This replicable effect, which has been attributed to the illusion of control, could also be driven by biases that do not involve active involvement, such as the gambler’s fallacy. We therefore designed a large-scale ( N = 1,594) online 2 × 2 experiment, where the sequence manipulation was crossed with a manipulation of active involvement. In the “self” condition, participants predicted and performed the coin-flips personally, whereas in the “other” condition, participants observed another participant predicting and performing the flips. The original effect of sequence replicated, occurring across both “self” and “other” conditions. However, participants in the “other” condition consistently rated the participants they observed as being better at predicting coin-flips than participants in the “self” condition rated their own abilities, an effect consistent with neither the illusion of control nor the gambler’s fallacy. Furthermore, participants’ Problem Gambling Severity Index (PGSI) scores did not correlate with their ratings of coin-flip predictions. We conclude that while Langer and Roth’s sequence effect is replicable, it is not attributable to the illusion of control or the gambler’s fallacy, and may be a poor analogue of chance-based naturalistic gambling formats due to its lack of association with PGSI scores.