How to read articles which depend on statistics

Abstract

In ALLC Journal Vol 1 No1, A Q Morton described some experiments in which he claimed that the counting of once-occurring words in different positions offers a simple and reliable test of authorship In Vol 2 No 3 M W A Smith refutes this claim, and in 8 pages of detailed criticism and reassessment of Morton's data concludes that there is no evidence to recommend Morton's procedures. Lay (i.e non-statistical) readers may well be bemused by argument over what should presumably be indisputable numerical facts The fact that such a dispute can occur raises questions of how articles whose substance depends on the reported outcome of statistical analysis should be read. This article discusses some of these issues in general terms and uses the two articles in question for the purposes of illustration, rather than in a spirit of criticism per se. © 1989 Oxford University Press.