A.R. Jonckheere

A long-standing pillar of the intellectual fabric of University College, London, fell in September last year. Although A. R. Jonckheere's first name, Aimable, is an apt description of his person, it was never heard; instead, the pithy appellation "Jonck" was used, both by himself and all who knew him. The brevity of this name simultaneously conjures up the direct nature of the man and the fondness that those around had for him.

Others have written of Jonck's French astronomer father (discoverer and cataloguer of binary stars), of his early education by Jesuits, of his being brought to England by his mother, leaving her divorced husband behind in France, of his schooling in Edmonton, of his first employment as an actuary, of his pacifism and consequent internment in starving Nazi-occupied Jersey and of his post-war acceptance by University College to read psychology with statistics. He  obtained a first class degree in 1949. Clearly, the flair for mathematics that Jonck showed at Edmonton (still evidenced by, for example, his notes on the use of complex variable in the solution of cubic equations) had been recognized by his psychology professor, Cyril Burt, himself a noted statistician, and by the redoubtable Flo David, his lecturer in the statistics department.

For two years after graduation, Jonck worked with Hans Eysenck and Cyril Burt, each of whom had his own agenda, but there was no animosity, perhaps because Jonck avoided the politics and kept to the statistics, which in both cases heavily involved factor analysis and the niceties of the rotation of axes.

The post-war expansion of higher education enabled Jonck, having obtained his bachelor's degree and having proved himself as a statistical assistant, to be at once accepted as a lecturer in his own department, when the opportunity arose in 1951. His brief was to teach statistics and animal psychology. It is clear that an able statistician would be much appreciated in a department that had been shaped by Spearman and Burt, but it was the new professor, Roger Russell, who had recently introduced animal psychology into the curriculum and, with it, the necessity for someone to lecture on it.

So Jonck became a member of the U.C.L. academic community and, as he said, at the hundredth anniversary celebrations of the U.C.L.  Psychology Department, it was for him a privilege at that time to have the opportunity to interact with, or attend the lectures of, so many great people, like A.J.Ayer and K.R.Popper, J.B.S.Haldane and J.Z.Young, who were members of that community.  Thus, philosophy and biology joined with psychology virtually as a single topic in Jonck's thinking. Later, Jonck undertook to give lectures on the psychology of perception to students of the Slade School of Art at U.C.L.. This brought him into contact with Ernst Gombrich, author of "Art and Illusion", broadening the scope of perception beyond the laboratory.  Eventually, everyday life itself, with its ethical, sociological, economic and emotional aspects, became, along with science, the object of analysis.

Another contact, more directly relevant to Jonck's statistical duties, was Maurice Kendall, who, until 1949 had been in the U.C.L. statistics department, but had then become professor of statistics at the L.S.E.. Jonck applied Kendall's S method of rank correlation to the case of testing a trend across k samples, giving the famous "Jonckheere trend test". It has been implied that this is merely Kendall's S repackaged, but the paper contains much mathematics demonstrating the conditions for the asymptotic approach of the statistic to normality, and this essential work is original.(Kendall, who himself refers to Jonck's paper, did not consider it to be derivative.) Another paper extended the trend test to cover repeated measures designs and yet another applied it to the case of ordered contingency tables. These are all of great utility in the analysis of psychological data. In addition to these tests, Jonck presented, with G.Bower, a number of nonparametric tests for application to learning data. Perhaps it is the technical nature of these papers (mostly written for this journal, of which he was editor from 1969 to 1975) that leads many to think that Jonck only published one statistical paper. The presentation of the tests as a practical corpus  was left to others. Chris Leach, whose "Introduction to Statistics. A nonparametric approach" (1979), gives some account of the k-sample trend test (but, strangely, not of the repeated measures test), writes of his "particular debt to A.R.Jonckheere, whose excellent lectures made me appreciate the general usefulness of Kendall's S statistic." (Leach did expound the repeated measures test, but  elsewhere, in an isolated paper.) Perhaps the most compendious exposition of the trend tests is given by G.F.Ferguson in,  "Nonparametric trend analysis: A practical guide for research workers", 1965, which extends their application to trends of higher order than monotonic.

There are two senses of "psychological statistics". The first relates to the analysis of data that are the typical product of psychological investigation, and it is to this that the trend tests were directed. The second concerns the statistical nature of behaviour itself. As an example, in his lectures on the adaptive behaviour of "lower" organisms, Jonck perforce spoke about kineses, the random changes of direction made e.g. by woodlice. This, coupled with a mechanism that slowed movement in a damp atmosphere, would, for purely statistical reasons, lead the animals to congregate in damp environments. ("It can be proved!" he asserted.)

Jonck, encouraged by Russell, was working on his thesis at a time when, in the United States, Bush and Mosteller were constructing stochastic models of learning in which the states on trial n were assumed to pass into those of trial n+1 by matrix multiplication and so were known as "linear operator models".

Even as he worked on his thesis, Jonck supervised R.J.Audley, a student of his, and one with similar gifts. Jonck's thesis was on "fixations" of rat behaviour in insoluble choice tasks, while Audley's was on the mathematical description of choice behaviour in rats. Audley's thesis used "urn models with contagious sampling" inspired by Feller's then recent book on probability theory and its applications, which gave an account of Polya's "urn schemes" in which the number of balls of a certain colour to be sampled from the urn increases or not according to the occurrence or not of some event corresponding to a reinforcement. Such a scheme leads to a gradually increasing or decreasing probability of sampling a ball of a particular colour, and this, it was hoped, mimicked the gradually increasing probability of a correct response (say) in a learning experiment. Audley's thesis passed with flying colours and Burt was moved to congratulate Jonck on the excellence of this production of his supervisee. The joint effort of this supervision led to several publications and to the addition of another class of stochastic process to the canon: the Audley-Jonckheere model of learning.

The next year Jonck spent in Geneva, collaborating with Piaget and his group. One resulting book, co-authored by Jonck, Mandelbrot and Piaget on "genetic epistemology" contained a section by Jonck  on geometry and perception (Mandelbrot's was about relations between perception and statistical induction). So this was largely a book on mathematical psychology. Another book, also mathematical, coauthored by Apostel, Jonckheere and Matalon produced a general theory of learning which embraced Hull, Tolman, Guthrie and Piaget, and, into the bargain, compared the Audley-Jonckheere model with those of Gulliksen, Bush and Mosteller  and Estes.

These are the main areas of Jonck's output, but he gave many extracurricular lectures on perception and developmental psychology, and later contributed in a mathematical way to several papers on human memory. But Jonck's main role was that of adviser in matters of experimental design and analysis to the whole department. He was merciless in exposing flaws in designs, but still people came. This critical facility (which formed the method of many of his lectures, in which he would take a well-known experiment and demolish it) has been thought to be the cause of the relative paucity of Jonck's publications, he being as critical of himself as he was of others. But quantity is not everything, and the corpus of trend tests together with the stochastic learning model represent a considerable contribution to knowledge. He continued for twenty years after retirement as an Honorary Research Fellow, advising, arguing and lecturing as before, until a blood-clot stopped him in his still-active tracks. At his thronged funeral gathering, people became aware of the enormous scope of his large library of books, covering most areas of knowledge. For every book on any topic, there was another giving the metatheory of that topic, in keeping with Jonck's habit of analysing and then analysing the analysis. So, e.g., among the tomes of mathematics is found "Mathematics: its content, methods and meaning, and among those of literary criticism is found "Principles of Literary Criticism".

J.C.Flugel wrote of one of Jonck's predecessors (S.J.F.Philpott), "I cannot recollect a single occasion on which he refused to give help or advice to me or anyone if it were in his power to do so – at whatever inconvenience to himself. … Philpott was a man whom to know was to respect and love."  Here, for "Philpott", read "Jonckheere".

Source: Obituary A. R. Jonckheere 1920 - 2005 (Editor of BJMSP 1969-75)
J. D. Valentine
Volume 59, Number 2, November 2006, pp. 475-477(3)
British Journal of Mathematical and Statistical Psychology (Used with permission)

Additional obituary links (The Independent Published: 13 October 2005 Obituary & The Guardian Published: Friday October 21 2005 are available on the contacts page.