| | |
Summary: Behavior Research Methods, Instruments, & Computers
2003, 35 (4), 493-503
Many experiments involve training in a task. This is
commonly done to reduce the variability that would arise
from unskilled subjects. In this case, the experimenter is
interested only in the final level of performance, often
describedby one or a few summary values (mean response
time, standard deviation,percent correct, etc.). However,
some researchers are not interested simply in a snapshot,
but in the whole dynamic of performance over training
(e.g., Logan, 1988; Rickard, 1997; Shiffrin & Schneider,
1977). Because of the large number of data involved, it
is often convenient to summarize them in a curve: the
learning curve (Heathcote, Brown, & Mewhort, 2000;
Newell & Rosenbloom, 1981).
Learning curves describe the evolutionof performance
over trials t. They are given by the following equation:
(1)
where a is the asymptote of the curve and b is the ampli-
tude. These two scaling parameters act as boundaries,
|
