An associative memory model, which can selectively recall an associative item, is discussed. An external input is injected as a context and helps to select a preferred associative item. Two models, which have different kinds of the external input, are compared by calculating critical similarity between the external input and the associative item. We also investigate how the critical similarity depends on noise level of input pattern and the ratio of the dimension of key vector to the one of associative vector.
keywords: Associative Memory, Contextually Modulated Recalling, Lexical Ambiguities, Statistical Neurodynamics