In a single large-scale study, we demonstrate that verbal sequence learning as studied using the classic Hebb repetition effect (Hebb, 1961)--the improvement in the serial recall of a repeating sequence compared to nonrepeated sequences--is resilient to both wide and irregular spacing between sequence repetitions. Learning of a repeated sequence of letters was evident to a comparable degree with three, five, and eight intervening nonrepeated sequences and regardless of whether the spacing between repetitions was regular or irregular. Importantly, this resilience of verbal sequence learning was observed despite complete item-set overlap between repeated and nonrepeated sequences. The findings are consistent with the conceptualization of the Hebb repetition effect as a laboratory analogue of natural phonological word-form learning. The results also have implications for the two leading models of Hebb sequence learning: Whereas the results are incompatible with the model of Page and Norris (2009), they can be handled readily by the model of Burgess and Hitch (2006) through the abandonment of its assumption of long-term (across-trial level) decay.