specify hypotheses about the relations between the internal processing steps
and observable behavior. For example, when a subject fixes his or her gaze on a
specific item in a visually presented table of information, we can infer that the
corresponding internal steps involve processing that information. On the basis
of veridical recall of the presented information after the task has been com-
pleted and the presented information is no longer available to the subject, one
can infer that that information was processed during the completion of the task.
In research on transcription typing, it is possible to determine what part of the
text the typist is looking at and what part of the text is simultaneously being
typed. The general finding is that the higher the skill level of the typist, the
farther ahead in the text the typist looks during typing. Being able to look
ahead in the text appears to be critical to the superior typing speeds of expert
typists, because when their freedom to look ahead is experimentally restricted,
their typing speeds are reduced to levels approaching those for novice typists
(Salthouse, 1984, chapter 11, this volume).
It is possible to extend our analysis beyond the processing of presented in-
formation and consider one’s access of preexisting knowledge and procedures.
In that case, atask analysisof the particular task should be performed before
the data collection. A task analysis involves specifying a number of different
sequences of processing steps that could generate the correct answer for a
specific task given the subjects’ preexisting knowledge. In well-defined task
domains, such as mental multiplication or problem solving in logic, it is rela-
tively easy to specify nearly exhaustively the different sequences of processing
steps leading to a correct answer in an efficient manner. In more complex
domains, the a priori task analysis makes explicit the pool of hypothesized
processing sequences that is being considered. On the basis of the think-aloud
verbalizations of subjects, one can determine only that the verbalized informa-
tion was accessed. A task analysis is critical for relating the verbalized infor-
mation to the underlying cognitive processes leading to its access or generation
(Ericsson & Simon, 1984).
Analysis of think-aloud verbalizations is time-consuming, and therefore
researchers in expertise using these types of data tend to collect data on many
subjects for a small number of tasks (expert–novice comparisons) or to collect
data on individual subjects for a large number of tasks (case studies).
Expert–Novice Comparisons. Comparison of think-aloud verbalizations by ex-
perts and novices is the best-known method of assessing differences in the
mediating processes as functions of the subjects’ levels of expertise: Subjects at
different levels of expertise are asked to think aloud while carrying out a small
number of representative tasks. The number of tasks usually is not sufficient for
assessing stable characteristics of individual subjects; the focus is on comparing
the groups of subjects to identify salient differences in regard to mediating
knowledge and processes.
The types of differences found in a wide range of domains of expertise are
remarkably consistent with those originally noted by de Groot (1978) in the
domain of chess. Expert performers tend to retrieve a solution method (e.g.,
next moves for a chess position) as part of the immediate comprehension of the
task, whereas less experienced subjects have to construct a representation of the
534 K. Anders Ericsson and Jacqui Smith