We have demonstrated the robustness and generality of the
digit superiority effect in Experiments 1a and 1b by showing similar
effects when comparing digit span to span for two different
sets of words that address a number of potentially critical itemand
set-level factors that are often invoked to explain different
recall performance with different linguistic materials. The survival
of the digit span superiority effect under these conditions
provides support, therefore, for the possibility that performance
in the typical STM task – such as the span task used here – is a
function of the extent to which long-term memory for linguistic
events can be brought to bear to support short-term processing
of novel verbal sequences. If this is indeed the case, then we
would also expect the more frequent encounter with random
sequences of digits compared to words to lead to an increasing
advantage in the STM setting as sequence length increases. For
example, recall of a two-item list A–B can only benefit from such
prior experience if a person has previously encountered the
sequence A–B a sufficient number of times for it to be retained
in long-term memory. However, a three-item list A–B–C can benefit
from three potential associative encounters: A–B, B–C, and A–
B–C. In Experiment 2, we examined whether the increase in digit
superiority at increasing sequence length reported above is more
amenable to an account based on maintenance and retrieval processes,
such as redintegration, that operate at the item level, or
one reflecting the role in STM of prior encounters with particular
sequences of verbal material.