Abstract
In tasks that measure serial-order memory, it is common to observe a "fill-in tendency"-when a person skips an item, the next item they report is more likely to be the skipped item (a fill-in response) than the next list item (an infill response). They tend to "fill in" the blank they skipped. The fill-in tendency has informed the modeling of serial-order memory-it presents strong evidence against associative chaining accounts because they predict more infill responses than fill-in responses. Despite the failures of associative chaining theories, evidence grows for the use of chaining-like item-dependent cues in serial-order memory. In this paper, we analyzed fill-in and infill responses from nine serial learning experiments (one new experiment and eight previously published experiments) that used variants of the spin list procedure and found strong evidence of item-dependent retrieval cues in serial-order memory. The current analyses revealed a fill-in tendency in all lists-even in those in which item-dependent cues were suspected to have been used. However, in those lists the likelihood of infill responses was higher, and consequently, the fill-in tendency was weaker. Our results expose a flaw in the conventional understanding of fill-in and infill responses. That is, the presence (or absence) of the fill-in tendency is not a strong test of item-dependent cues. Instead, changes in the magnitude of the fill-in tendency-more specifically, an increase in the likelihood of infill responses-across task conditions seem to better indicate the use of item-dependent cues.