The Sure Witness: Certainty On The Stand

AJuries have always found the confident witness persuasive, and the law has rarely discouraged them. In 1977 an appeal court in the United States set out five things a judge should weigh before an identification was placed before a jury, and the certainty the witness expressed was one of them. Two decades later, when genetic testing began to be applied to old convictions, that same courtroom certainty turned up again and again in the cases that had gone wrong: witnesses who had pointed across a courtroom without a flicker of hesitation had pointed at an innocent man. The conclusion drawn at the time, and repeated ever since, was that certainty tells you nothing about accuracy. Psychologists now argue that it was drawn from the wrong moment. Whether certainty is worthless in itself, or worthless only by the time a trial gets hold of it, is what divides the field.

BHelena Marchetti, who studies memory at the Trieste Institute of Applied Cognition, holds that the trouble is one of timing rather than of certainty. Her group examined 241 convictions later overturned, and a mistaken identification had contributed to 71 per cent of them. Only 44 of the 241 files, however, preserved any record of what the witness had said at the moment of the first identification, and in most of those files the witness had been questioned repeatedly before the trial opened. Her laboratory work shows why the gap matters. Of 640 volunteers who watched a staged theft and were then shown a lineup from which the thief had been removed, those told afterwards that they had picked out the police suspect remembered having been far surer at the time than those told nothing at all. Her objection is therefore not that confidence is meaningless but that the confidence a court hears has been contaminated by everything said to the witness since; what it measures, on her account, is the confirmation the witness has been given rather than the memory underneath.

CThe argument makes a prediction, and the prediction can be tested. If certainty is corrupted after the event rather than empty from the start, then certainty recorded at the instant of the first identification, before anyone has spoken, ought to track accuracy closely. In 2019 four police districts agreed to write down, word for word, what each witness said immediately after viewing a lineup, and 348 identifications were gathered in this way. Because the truth of a real case is never known to researchers, accuracy was estimated from how often a witness chose a filler, a person the investigators knew had committed no offence, instead of the suspect. Witnesses who called themselves certain avoided the fillers in 92 per cent of cases, against 41 per cent of those who expressed any hesitation. Marchetti presents the finding as evidence about where certainty loses its value, not as a defence of the use courts make of it.

DCallum Ferris, a legal psychologist at the Waikato Centre for Evidence and Law, goes further and attaches conditions. Certainty carries information, on his account, only where the lineup has been run properly: one suspect and no more; fillers who fit the description the witness first gave; a warning that the offender may be absent; an officer running the lineup who does not know which member of it is the suspect; and a statement of confidence taken before any comment is made. Across 1,100 laboratory identifications his group found the most confident witnesses correct 97 per cent of the time when all five conditions held, and 62 per cent when the officer knew where the suspect stood. Blind administration of this kind is now ordinary in the districts he advises, although it was dismissed as unaffordable when he first proposed it. Ferris concedes two things. A figure of this sort describes a group of witnesses and cannot be read off for the one standing in front of a jury; and of the 44 archive files that preserved an initial statement, only three came from lineups run in the manner he requires.

EThe quarrel has weight outside the laboratory. Guidance issued in 2016 in several jurisdictions obliges officers to record a confidence statement in the witness's own words at the first identification, and some forces now log it on a scale from zero to a hundred as well; nothing, however, prevents a witness who wavered at that first viewing from expressing complete certainty in court three years later, and nothing obliges a jury to be told of the difference. Marchetti accepts, awkwardly for her own position, that a cue which has been contaminated is not the same as a cue which is empty, and that her archive cannot separate the two. Ferris answers that the distinction dissolves in a courtroom, since jurors have no means of telling the certainty of the identification room from the certainty of the witness box, and hear only the second.

FA settlement of a sort is taking shape, in which the two accounts are attached to different moments rather than ranked against each other. At the first identification, under the conditions Ferris sets out, certainty behaves like a usable measure, which would explain the field figures; by the time of trial it has absorbed so much of what the witness has been told that it registers the telling instead, which would explain the overturned convictions. What neither side has settled is whether the relationship survives when the original sighting was brief or badly lit, for no field study has yet recorded the viewing conditions with any reliability, and no laboratory can reproduce the fear of a real crime. Until that gap closes, courts will go on deciding what certainty is worth on grounds that are not evidence.