Sentence-length and Authorship Attribution: the Case of Oliver Goldsmith

The sentence-lengths of sixteen essays by Goldsmith are examined in relation to data from ten essays (we call these 'doubtfuls') which have been attributed to him. Comparisons between the 'doubtfuls' and the known Goldsmiths are made with reference to the X 2 goodness-of-fit test, and the method of reciprocal averaging. The Goldsmith essays form a close group, with four of the 'doubtful' essays well outside, two less remote and four within the Goldsmith cluster. Comparison with fifty essays by nine of Goldsmith's contemporaries reveals the distinctiveness of his sentence-length patterns, and strengthens the probability that the four least doubtful essays are his. In the case of Goldsmith, then, sentence-length may be considered a reliable stylistic marker.