Michael E. J. Masson
Colin M. MacLeod

About 25 years ago, the prevalent view of memory began to change. Prior
to that time, the predominant theories were framed in terms of engrams
or traces that were stored in memory at the time of encoding and recovered
from memory at the time of retrieval, much as money is deposited in and
withdrawn from a bank. This can be seen early on in the writings of Semon
( 1904 / 1921), but this style of theory prevailed into the 1960s (see textbooks
of that time, e.g., Kintsch, 1970). Encoding processes laid down traces in
memory and retrieval processes searched for them. Traces were products
of encoding and objects of retrieval. Clearly, processes were involved in
both encoding and retrieval, but what was stored in memory was the conse-
quences of initial processing, not records of the processing itself. This trace
view is well captured by the best known models of that period, the two-
state or "buffer" models of memory ( Atkinson & Shiffrin, 1968; Waugh &
Norman, 1965).

Then, in the early 1970s, a new perspective emerged. In large part, this
stemmed from research carried out at the University of Toronto. It began
with two key elements: the renewed emphasis on processing that formed
the basis of the levels of processing framework ( Craik & Lockhart, 1972);
and the increased stress on the interplay between encoding and retrieval

-155-