This paper investigated different types of errors in serial learning, via a further analysis of the data obtained by Harcum, Pschirrer, and Coppage (1968) with continuous presentation of 10-trigram lists. The purpose was to discover whether the particular distributions of extralist intruding errors (ELI), intralist intruding errors (ILI), and failures to respond (FTR), which have been obtained using the conventional temporal gap between successive trials, are found when trials are continuous. Similar distributions would indicate that the relevant determinants are intervening cognitive factors, rather than stimulus variables. The results with continuous trials were similar to those for conventional pacing; FTR errors showed both asymmetrical and symmetrical (bowing) components, the ILI distribution was more nearly symmetrical with the minimum at the cognitively first item, and ELI errors, which were infrequent, varied little among serial positions. These results support the conclusion that an asymmetrical component of the serial-position curve is produced by the cognitive selection of one item to be learned first, to become the anchor for the later learning, and a symmetrical component is the result of associative or positional confusion.