visual imagery in working memory to simulate tying her shoes (5.2.3). As she
simulates each step, she encodes it linguistically and then expresses it verbally
to answer your question. Unless the declarative script for a solution is
processed frequently, it becomes increasingly unavailable in memory. An
exception is teachers, who often retain these declarative representations in
order to describe solutions explicitly to their students.

As performance evolves from novel problem solving, to mechanized
problem solving, to automatized problem solving, the characteristics of
novel problem solving disappear: People become increasingly fast at solving
problems, they rarely fail or need to backtrack, and their solutions become
increasingly optimal. In addition, people can perform secondary tasks while
problem solving (e.g., talking), because automatization frees strategic re-
sources (4.2.2). Whereas watching a novice is often painful, watching an
expert is usually a pleasure.

As we have seen, the three themes that we considered initially--similarity,
availability, and framing--pervade human thought. All three themes appear
in every form of thought that we have considered, including decision making,
induction, deduction, and problem solving. Each theme reflects a funda-
mental mechanism in the architecture of human cognition: Similarity reflects
comparison processes in working memory, availability reflects the retrieval
properties of long-term memory, and framing reflects the omnipresence of
frames in organizing experience. Clearly, the cognitive architecture leaves its
signature across the spectrum of human thought.

Understanding human thought is of great scientific interest in its own
right, given its unique and powerful properties. However, understanding
human thought has always held much promise for more practical concerns as
well. For this reason, researchers have often studied thought as it occurs in
natural domains ( Chi, Glaser, & Farr, 1988; Greeno & Simon, 1988). Tradi-
tionally, work on decision making has addressed important issues in eco-
nomic and political arenas. More recently, researchers have begun to explore
induction, deduction, and problem solving in educational domains, such as
geometry and physics, and in occupational domains, such as medicine and
electronic troubleshooting. To some extent, much of this work can be viewed
as applied cognitive science, yet this work also makes major contributions to
basic science, through the discovery of new methods and mechanisms.
Because of the great potential that the technical era holds for education and
job performance, and because of the serious challenges that we face socially
and economically, these developments hold much promise.

-339-