Implicit learning of spatiotemporal contingencies in spatial cueing.

We investigated the role of implicit spatiotemporal learning in the Posner spatial cueing of attention task. During initial training, the proportion of different trial types was altered to produce a complex pattern of spatiotemporal contingencies between cues and targets. For example, in the short invalid and long valid condition, targets reliably appeared either at an uncued location after a short stimulus onset asynchrony (SOA; 100 ms) or at a cued location after a long SOA (350 ms). As revealed by postexperiment questioning, most participants were unaware of these manipulations. Whereas prior studies have examined reaction times during training, the current study examined the long-term effect of training on subsequent testing that removed these contingencies. An initial experiment found training effects only for the long SOAs that typically produce inhibition of return (IOR) effects. For instance, after short invalid and long valid training, there was a benefit at long SOAs rather than an IOR effect. A 2nd experiment ruled out target-cue overlap as an explanation of the difference between learning for long versus short SOAs. Rather than a mix of perfectly predictable spatiotemporal contingencies, Experiment 3 used only short SOA trials during training with a probabilistic spatial contingency. There was a smaller but reliable training effect in subsequent testing. These results demonstrate that implicit learning for specific combinations of location and SOA can affect behavior in spatial cueing paradigms, which is a necessary result if more generally spatial cueing reflects learned spatiotemporal regularities.