Experimental Results on the Centipede Game in Normal Form: An Investigation on Learning

We analyze behavior of an experiment on the centipede game played in the reduced normal form. In this game two players decide simultaneously when to split a pie which increases over time. The subjects repeat this game 100 times against randomly chosen opponents. We compare several static models and quantitative learning models, among them a quantal response, model reinforcement models and fictitious play. Furthermore, we structure behavior from period to period according to a simple cognitive process, called learning direction theory. We show that there is a significant difference in behavior from period to period whether a player has decided to split the pie before or after the opponent. Copyright 1998 Academic Press.