Testimony

4. Vulnerable assumptions

Another problem for any reductionist project concerns the fact that the plausibility of supposing that testimony-free verifications of reliability are usually available seems covertly to depend upon accepting testimony-laden checks, observations and falsifications. We have seen this problem arise in the particular case of checking for expertise, but its significance is wider than that since all sorts of apparently personal verifications will turn out to be dependent upon unchecked testimony. If I become suspicious of a job applicant’s claim to have a Ph.D., for instance, I can seek to check personally on the claim (in the sense of conducting the investigation on my own behalf) but this endeavour will invariably involve accepting the word of some people or institutions that the degree was or was not awarded. Moreover, the reliability of these people or institutions will often be either taken for granted or accepted on the basis of reputation, which is itself a complex form of hearsay, as are the documents and certificates the institutions may offer as evidence. Such often-ignored facts account for the way Hume and others persistently and unconsciously conflate individual and social readings of the term ‘observation’, a conflation that makes Hume’s project seem more manageable than it really is. When he refers to the need to base ‘our’ (that is, the sum of individual cognizers’) reliance upon testimony on ‘no other principle than our observation of the veracity of human testimony’ ([1748] 1975: 111), Hume is sliding between the two senses of ‘observation’. ‘We’ can observe the veracity of human testimony only if we already trust each other’s testimony sufficiently for the project to be no longer that of the autonomous knower.

Hume’s problems are symptomatic of the fact that the reductionist picture in which the autonomous knower figures so centrally is flawed at heart since the existence of a common language in which reports are made and accepted or rejected already carries with it a commitment to some degree of unmediated acceptance of testimony. The reductionist strategy necessarily supposes that we can fully understand what we are told in a public language and then independently investigate its truth or correctness, but the connections between linguistic comprehension and truth (or at least correct assertibility) are much closer than this picture allows. The reductionist project must concede that the autonomous knower might consistently understand correctly all that they were told but none the less find every report mistaken or deceitful. But any such ‘discovery’ would rather serve to refute the initial supposition that the reports had been correctly understood. Public languages are, after all, learnt and their use is subject to correction, so reports on usage and on some facts embodying that usage must have a degree of success that the reductionist ignores. Fixity of meaning depends upon some degree of successful reporting, though just what degree is a further question (Coady 1992: Ch. 9).