How a-priori biases affect sequence learning in a serial reaction time task