Human Sequence Learning: Can Associations Explain Everything?
Rainer Spiegel, I.P.L. McLaren · eScholarship (California Digital Library) · 2001
It will be shown that whilst a popular connectionist model, the simple recurrent network (SRN) as introduced by Elman (1990), is a very good first approximation in modeling human sequence learning, it is not, in itself, sufficient.At CogSci 2000, all five papers referring to the SRN tried to provide evidence that it is an adequate model of human performance.We will take on a more moderate position.The results of a human experiment followed by a structured interview reveal that human sequence learning is not always the kind of statistical process captured by the SRN alone.