conspiracy to plagiarise, conspiracy to obscure plagiaristic forms of identical entries through deliberate
reduction of common headword entries towards mere similarities, and establish plausibility of these activities in
both language directions. It will have to perform these functions against a host of more compelling professional
and societal factors which apparently constitute far more plausible contending explanations for the same data.
93.
None of these conditions has been satisfied yet and there is no statistically sound basis for selecting plagiarism
as the only or the most probable or even a likely explanation".
[61] In his affidavit Prof Dunne also discusses the choice of random and nonrandom sampling used by Media 24. It
appears that the random choice employed by Prof Kidd was a random choice of page numbers in the works
using a computerbased spread sheet programme and then 4 locations on each of the chosen pages. Prof
Dunne expresses some reservations about this process but appears, in the main, to be satisfied with the
integrity thereof. The nonrandom sampling appears to be the work of Dr Anton Prinsloo and is the choice of
headwords commencing under the letters B, D, I and S. Prof Dunne refers to these as the "Prinsloo
similarities".
[62] Prof Dunne makes the following general observations in relation to the similarities identified by Prof Kidd and
Dr Anton Prinsloo:
"52.
The frequency of Kiddor Prinsloosimilarities does not warrant an inference of plagiarism. While plagiarism, if
and when it occurs may give rise to very strong forms of Kiddor Prinsloosimilarities, the converse, from
particular to general, does not necessarily apply.
53.
On the basis of these arguments, Prof Kidd's analysis may be internally consistent. However, its relevance to
any claim of plagiarism has to rest completely on the validity of either the KiddorPrinsloosimilarities as
substantive and conclusive evidence for that purpose of eliminating
Page 493 of [2015] 3 All SA 478 (WCC)
all explanations other than plagiarism for the shared headwords. The presence and the counts of headwords
that are unique to each dictionary will also have to be explained within an invocation of plagiarism. I deal with
this issue in the next section."
[63] Prof Dunne further called for a third dictionary to be compared as part of his analysis and was furnished with a
spread sheet analysis relating to the Longman Handboek vir die Afrikaanse Taal ("the HAT"), a far larger
monolingual Afrikaans dictionary with 14503 entries. His assessment was that there were in fact more
similarities in the Pharos than the OUP work when these works were compared to the HAT. Having assessed
the HAT also, Prof Dunne concluded as follows:
"62.
Given the larger word count of the Oxford Skoolwoordeboek, the modest frequencies of Prinsloosimilarities,
the softness of those criteria, the unspecified nature of the Kiddsimilarities, and the necessities for dictionaries
to address common word usage and common target users, there appears to be very little statistical evidence in
favour of even a loose definition of plagiarism at work in this state of affairs."
Assessing the expert evidence
[64] In the recent judgment in the National Potato CoOperative case16 the Supreme Court of Appeal revisited, inter
alia, the function of the expert witness in litigation, and the manner of evaluating such evidence. The court
stressed, with reference to Gentiruco17 and Coopers 18 that such evidence is admissible when the court is
able to avail itself of "appreciable help" from an expert where such person by reason of specialist knowledge
and skill, is better qualified to draw inferences from the proved facts than the trial judge.
[65] In Coopers, Wessels JA observed as follows at 370GH:
"There are some subjects upon which the court is usually quite incapable of forming an opinion unassisted and others
upon which it could come to some form of independent conclusion, but the help of an expert would be useful."
[66] Both Professors Kidd and Dunne are highly qualified and experienced in the field of statistics. The real purpose
of the statistical evidence in this case is not to afford the statisticians an opportunity to express a view as to
whether OPU has breached Media 24's copyright by plagiarising its work (for that is preeminently the terrain
of the lexicographer and the linguist), but rather to give the court some idea of the numerical percentage of
the incidence of such alleged copying in the instant case. After all, it cannot be expected of the court to trawl
through more than 20 000 entries (5 000 per side in each dictionary) to assess the similarity between
example sentences and senses. The task would simply be too immense. Indeed, it is so immense that neither
party has tendered evidence from a lexicographer who has done just that. The sampling of similarities by Prof
Page 494 of [2015] 3 All SA 478 (WCC)
Kidd and Anton Prinsloo, and the statistical recurrence thereof was seen by Media 24 as the most suitable
and reliable way of establishing the extent of the alleged copying in its founding papers.
[67] Inherent in the approach adopted by Media 24 to litigate by way of application is the limitation imposed on
the court through the receipt of such evidence: ordinarily a court would want to enjoy the benefits of such a
witness being crossexamined by the opposing camp and answering questions in clarification of the court's
own difficulties in adjudicating the dispute. Very often, that exercise would demonstrate whether the expert is
truly independent rather than being the proverbial "hired gun", and most importantly permit the court to
grapple , first hand as it were, with the issues it was being asked to resolve.
[68] At an early stage of these proceedings before the filing of an answering affidavit, OUP's attorneys suggested
to their opponents that the matter be sent to trial with the founding papers to stand as a simple summons
with a declaration to follow. However, Media 24 dug in its heels and insisted on the matter proceeding by way