I observe that certain states of my body and certain of my words and actions are commonly accompanied by certain sensations and mental activities of mine. I also observe that Smith is often in similar physical states, utters similar words and performs similar actions. But I do not observe, in the same sense, that Smith's physical states and activities are correlated with feelings and thoughts on his part. He could be a mindless robot, as far as my observation goes. Nevertheless, I have no reason to think that he is a mindless robot, and his observable activities do so closely resemble the ones I engage in when I have thoughts and sensations. So Smith, so much like me in the one respect, is probably like me in the other. He probably 'has a mind'.
Such is the form of the usual analogical argument for the existence of other minds. Although an acceptable version of the argument may someday be found, w;rsions that have so far been offered seem to be vulnerable to at least two of the attacks philosophers have made on them. First, the sample is restricted. The number of cases in which I have observed a correlation between mental and physical states or events is limited, and the ratio of observed correlations to the total number of observed physical states of the relevant kind is usually small. I have observed pain and screams together on only a few occasions, for example, and ratio of this number to the total number of screams I have heard is not high. My sample would seem too fragile to bear the weight of the conclusion which is supposed to rest on it.
One can increase both the number of observed correlations and the ratio of these to the total number of observed physical states of the relevant sort, however, by choosing one's version of the analogical argument carefully. Consider this version, in which the sample class consists of a single entity: