Abstract
We introduce a language competition model that is based on theAbrams-Strogatz model and incorporates the effects of memory and learning inthe language shift dynamics. On a coarse grained time scale, the effects ofmemory and learning can be expressed as thresholds on the speakers fractions ofthe competing languages. In its simplest form, the resulting model is exactlysolvable. Besides the consensus on one of the two languages, the modeldescribes additional equilibrium states that are not present in theAbrams-Strogatz model: a stable dynamical coexistence of the two languages anda frozen state coinciding with the initial state. We show numerically thatthese results are preserved for threshold functions of a more general shape.The comparison of the model predictions with historical datasets demonstratesthat while the Abrams-Strogatz model fails to describe some relevant languagecompetition situations, the proposed model provides a good fitting.