Human Signal-detection Performance: Effects Of Signal Presentation Probabilities And Reinforcer Distributions.

The team ran a signal-detection game with college students. Rare and common signals paid the same points.

They asked: will equal rewards erase the bias to shout "rare" too often?

Even with equal pay, people kept calling out the rare signal. Extra points or mild penalties did not fix it.

The bias stayed strong across every test.

Turkkan (1994) saw the same pattern two years earlier. In that study, rare success events made students say they "failed" less than they really did. Together, the papers show the bias is not about vision or memory—it is about how we weigh unlikely events.

Emmelkamp et al. (1986) showed that, with clear cues and instructions, human choice can follow the matching law. V et al. found a stubborn bias even when reinforcement is equal. The two results sit side-by-side: matching works when the contingencies are obvious, but signal-probability bias sneaks in when the task is detection, not choice.

Fox et al. (2001) found that groups of humans allocate choices to match reward ratios. V et al. show that single humans mis-detect rare signals despite equal rewards. Group choice obeys the numbers; individual perception does not.

When you teach a learner to label objects, spot safety signals, or report private events, remember: rare cues will be over-reported even if praise or tokens are equal. Add extra teaching trials for the rare labels, use richer reinforcement for correct detection of common items, or give clear rules like "only say red if you are sure." Check your data—if accuracy is low on the rare stimulus, do more teaching, not more nagging.