This paper consists of four parts. In section 1, I shall offer a strategy to bypass a counter-example which Gareth Evans (1975) offers against Quine’s Thesis of the Inscrutability of Reference. In section 2, I shall introduce a criterion recently pro-duced by Crispin Wright (1997) in terms of ‘psychological simplicity’ which threatens the perverse route offered in section 1. In section 3, I shall argue that a LOT model of human cognition could motivate Wright’s criterion. In section 4, I shall argue that if we instead model human cognition by a recurrent neural network, then Wright’s criterion is unmotivated. Thus I shall produce a Connectionist Defence of the Inscrutability Thesis.