^{page 266 note 1} Kenny, A., A Stylometric Study of the New Testament (Oxford: Clarendon, 1986)Google Scholar; Mealand, D. L., ‘Computers in New Testament Research: an Interim Report’, JSNT 33 (1988) 97–115Google Scholar; Morton, A. Q., Literary Detection (Bowker, 1978).Google Scholar

^{page 266 note 2} Kenny, , Stylometric Study, 100.Google Scholar

^{page 266 note 3} Morton, , Literary Detection, 168, 183Google Scholar; see also Michaelson, S. and Morton, A. Q., ‘The Spaces in Between’, LASLA Revue 1 (1972) 23–77 esp. p. 45.Google Scholar See n. 1, page 273 below.

^{page 266 note 4} See n. 1 above.

^{page 267 note 1} Kiley, M., Colossians as Pseudepigraphy (Sheffield: JSOT, 1986) esp. 51–9Google Scholar, Bujard, W., Stilanalytische Untersuchungen zum Kolosserbrief (Göttingen, 1973).Google Scholar

^{page 268 note 1} Morton, A. Q., Literary Detection, 110–14, 180–3Google Scholar, Michaelson, S. and A. Morton, Q., ‘Positional Stylometry’ in The Computer and Literary Studies, ed. Aitken, A. J. et al. (Edinburgh: EUP, 1973) 69–93.Google Scholar

^{page 268 note 2} See below for the definition of a c-sentence. For more discussion of the use of tests involving sentence length directly or indirectly see Kenny, , Stylometric Study, 101–15Google Scholar, and Morton, A. Q., Literary Detection, passim esp. 99–101.Google Scholar

^{page 269 note 1} See the works on stylometry cited in my ‘Computers in N.T. Research’ nn. 13–22.Google Scholar also Kenny, A., The Aristotelian Ethics (OUP 1978)CrossRefGoogle Scholar and the discussion on the effects of changes of genre in Morton, , Literary Detection, 127–9.Google Scholar

^{page 270 note 1} The figures for Isocrates are published in Michaelson, S. and Morton, A. Q., ‘Things Aint at they used to be: a study of Chronological Change in a Greek Writer’ in The Computer in Literary and Linguistic Studies, ed. Jones, A. and Churchhouse, R. F. (Cardiff: University of Wales Press, 1976) 78–84, esp. 81.Google Scholar The machine readable text from which the figures were derived is still available in Edinburgh.

^{page 270 note 2} McArthur, H. K., ‘KAI Frequency in Greek Letters’, NTS 15 (1968–1969). 339–49.CrossRefGoogle Scholar gives evidence which tends to favour the four epistle theory. I did, however, run some tests on the controls in classical literature (on 19th May 1987) and obtained the following preliminary results. In six examples from Diodorus Siculus the variation in the use of καί is marginally significant at the 5% level. In seven samples from Herodotus the variation is significant at the 5% level and almost so at the 2.5% level. In twelve works of Demosthenes the variation was extremely significant and reached the 0.1% level. I tested Demosthenes works 1–9.and 14–16.using Morton's figures. The result was χ² at 50.16 for 11df, giving p as considerably below 0.001. These are preliminary results but they do suggest that greater caution may be needed in interpreting tests based on the frequency of καί in Greek literature.

^{page 271 note 1} See n. 1, page 266 above.

^{page 271 note 2} Recent work on the stylometry of English literature includes Burrows, J. F., Computation into Criticism: A Study of Jane Austen's Novels and an Experiment in Method (Oxford: Clarendon, 1987).Google Scholar This is an important work for its analysis of speech by separate characters depicted by the author. He finds significant variation between the idiolects of the author's characters. Another significant study is the recent thesis written in the Dept. of Computer Science by T. B. Horton on the stylometry of English literature and accepted for the degree of Ph.D. by the University of Edinburgh.

^{page 271 note 3} See n. 1, page 268 and n. 1, page 270 above.

^{page 271 note 4} Kenny, p. 113Google Scholar; he argues that ‘to have used statistically significant differences as a rigid criterion for variation in authorship would have ruled out central works of undoubted Aristotelian provenance’.

^{page 273 note 1} See Morton, A. Q., ‘The Spaces in Between’, LASLA Revue 1 (1972) 23–77Google Scholar (published by the International Organisation for Ancient Languages Analysis by Computer, Liège).

^{page 273 note 2} Kenny, , Stylometric Study, 15.Google Scholar

^{page 274 note 1} I am very grateful to D. A. Carson and to D. de Lacey for their help in the early stages of my use of GRAMCORD. For further details of the system see my article cited in n. 1, page 266 above.

^{page 274 note 2} Kenny, , The Computation of Style (Pergamon, 1982) 169.Google Scholar There is also a programme which runs on an ibm pc which will calculate exact p values.

^{page 275 note 1} Ibid. 111–12.

^{page 275 note 2} Kenny, A., The Computation of Style, 169.Google Scholar

^{page 276 note 1} Ibid., 105–19. for a fuller account of the procedure being used here.

^{page 282 note 1} Michaelson, and Morton, , ‘Things Aint’, 83. See n. 1, page 270 above.Google Scholar

^{page 282 note 2} Kenny, Aristotelian Ethics, 100–3. idem, Stylometric Study, 113.

^{page 282 note 3} Morton, , Literary Detection, 129.Google Scholar

^{page 283 note 1} It is, however, worth noting that in the larger group it is Or. 18 which produces the largest contribution to the χ² total. For this reason I rechecked the count for g2 in Or. 18. Of course, if someone can cite other grounds for regarding Or. 18 as of doubtful attribution, that would affect the argument here. (Or. 17 also produces a sizeable contribution to the total.)

^{page 283 note 2} I note in passing that although the p values for g2 in the groups of four, seven, and ten epistles are all in the ‘extremely significant’ category, the p value for the group of ten is even lower than that for the seven or the four. Presumably we conclude that the diversity of the ten is even more extremely significant.

^{page 283 note 3} In the course of this paper I have sometimes argued for conclusions which differ from those of Morton. Having worked through quite a number of his published papers, I would like to express appreciation of several important elements in them. His work is of importance for seeing at so early a stage the potential of literary computing, the need to establish it as a proper scientific discipline, the need for controls from classical and other literature, the potential of a whole range of new criteria, the complexity of the factors involved such as genre and chronological change, and the need to test for these rather than just allude to them. One does not assess scientific originality by whether all of a scholar's conclusions always gain acceptance, but by the capacity to develop new methods, to formulate interesting hypotheses and to devise means of testing them in detail. In this connection it is especially worth noting the attention paid to variables such as genre and chronology in the works cited above.