Implicit Relational Learning In Multiple-Object Tracking Task: Do People Really Track Objects?

Abstract

Earlier, we showed that multiple-object tracking task can be used to examine explicit relational learning. In this task, participants were instructed to track four out of eight objects and report at the end of the trial whether a single cued object was among those they tracked (yes/no task). The display also contained two bars of different width. In Informative condition, the location of the cued object predicted the correct choice. If the answer was "yes", then the object was located next to the narrower bar; otherwise, it was located next to the wider bar (or vice versa). In Random condition, the location of the object did not predict the correct choice. We found that participants in Informative condition were more accurate when instructed to track 2 out of 8 objects than when instructed to track 4 out 8 objects. This result suggests that participants in Informative condition do track the objects, even though their answer at the end of the trial is clearly influenced by the contextual background information. In the next experiment, we manipulated the availability of background information by either presenting the bars at the very end of the trial (Bars Shown condition) or removing them at the very end of the trial (Bars Removed condition). We found that performance in Bars Removed condition has deteriorated significantly, while performance in Bars Shown condition remained highly accurate, suggesting that the contextual background information is utilized when participants are making their final choice.