The study investigates
the simulation of sequential learning based on a simple connectionist model
- a sequential Hopfield-net with higher order interactions. We based the
development of the model on empirical constraints (effects of stimuli's
position, memory span and implicit learning). Experimental validation is
accomplished in a study on implicit learning using two groups of
subjects (n = 40), learning regular or irregular sequences of letters, which
were also presented to the model. Simulations show qualitative results of
primacy, memory span and implicit representation of simple regularities.
Experimental validation confirms the simulation results: Subjects who had to
learn regular sequences of letters acquired the sequences faster than subjects
with irregular stimulus material. Basic features of the model are
discussed as explanations for primacy and recency effect, memory span
and implicit learning.

\underline {Key Words:} neural nets -- sequence learning -- implicit learning
-- hopfield-net -- sigma-pi units -- higher order interactions