Abstract: Method and device for recognizing characters, in which in a learning phase, as well as in a subsequent working phase features of character patterns in a number of aspects thereof are classified in a number of groups. During the learning phase the results of these classifications are recorded in a store as statistic frequencies. During the working phase the result of the feature classification of a freshly offered character is utilized in determining, for each class of characters, (a) the probabilities of the features found in this character, and (b) the weight of the features of a character pattern to be recognized. The weight attributed to each feature depends upon the shape of the pattern, and upon the basis of the values of these weights, the features are selected. The stored statistic frequencies of the selected features are multiplied by the values of the weights, and the largest value among the results is utilized for indicating the class of the feature.