The Dairy Robbery Identification
At 9.40 pm a dairy in a small town is robbed. The sole attendant, aged 19, is threatened with what she believes is a firearm held at chest height, about 1.2 metres from her face. The whole encounter lasts between four and six seconds. The lighting is a single fluorescent tube behind the offender, so his face is partly backlit. The offender is of a different ethnicity from the attendant. That night she gives a description to police that closely matches a man police already suspect. The next afternoon an officer shows her a photo montage of eight photographs, with the suspect at position 3, and says 'take your time, have a good look' while resting his finger on photograph 3. She identifies photograph 3, saying 'I think that's him, it looks like him'. Six weeks later a second officer shows her a different montage containing the same suspect. She picks him immediately and says she is '100% certain'. The case goes to trial and the defence argues the identification should be excluded.
1. Distinguish estimator variables from system variables, identify which ones are operating in this case, and evaluate their likely effect on the accuracy of the identification.
Read the worked answer
Estimator variables are features of the crime and the witness that the justice system cannot control after the event; they estimate how good the memory is likely to be. System variables are features of the investigation that the system can control.
Estimator variables present here. (i) Weapon focus: a firearm at chest height, close to the face, reliably narrows attention to the weapon at the expense of the face, and the effect is stronger when the weapon is believed real. (ii) Very brief exposure: four to six seconds is short for encoding a novel face; most face encoding that supports later recognition needs longer. (iii) Arousal: high stress at encoding can impair hippocampal-dependent memory while strengthening amygdala-dependent memory for the gist - she may remember a gun clearly and a face poorly. (iv) Cross-race effect: recognition accuracy is consistently poorer for faces of another ethnicity, and this is a robust finding, not a stereotype. (v) Poor lighting and backlighting: the face was silhouetted, removing the detail that discrimination depends on. (vi) Retention interval: six weeks is enough for trace decay and for contamination.
Overall, the estimator variables collectively predict a weak memory trace.
System variables present. The montage appears to be a 'suspect-present' array in which the suspect may not match the description as well as the fillers; the officer knows who the suspect is and is therefore not blind; his finger on photograph 3 is an inadvertent cue; the instruction to 'take your time' is not a neutral 'may or may not be present' warning; and the procedure was repeated with the same suspect six weeks later. Each of these can produce a positive identification from a weak trace by shifting the witness from an absolute judgement to a relative one - choosing whoever looks most like the offender rather than deciding whether the offender is there at all.
2. Explain why the witness's '100% certain' statement six weeks later is weak evidence, using what is known about how memory works.
Read the worked answer
Certainty feels like evidence because we assume it tracks accuracy. In fact the confidence-accuracy relationship for eyewitness identification is weak, and it becomes weaker still under exactly the conditions present here.
Memory is reconstructive, not a recording. Retrieval is a reconstruction from fragments plus expectations and prior exposure, not playback. Each time the witness thinks about the event - including during the first identification - she is re-encoding a version of it. The face seen in the montage becomes blended with the face seen in the dairy. This is a source-monitoring error: at the second identification she is recognising a face she has seen before, but the source of that familiarity is the first montage, not the robbery. The fluency of that recognition is then experienced as certainty.
Post-identification feedback inflates confidence without improving accuracy. Choosing a face is itself confirming - it tells the witness 'you remembered something'. The officer's behaviour, the fact of being shown the same suspect again, and the knowledge that police have a suspect all act as confirming feedback. Experimental work shows that confirming feedback substantially raises witnesses' confidence and their recollection of how good the viewing conditions were, while accuracy stays the same. So the '100% certain' claim is likely to reflect the confirming procedure rather than a stronger memory.
Time does not repair a weak trace. Six weeks does not improve discrimination; it gives more opportunity for contamination. Her first statement - 'I think', 'it looks like him' - is actually the more informative data point, because it was made before the confidence-inflating feedback. It also suggests she was making a relative judgement, not an absolute one.
3. Recommend two specific procedural changes police should make, and justify each in terms of the cognitive mechanism it addresses.
Read the worked answer
Change 1: use double-blind administration with neutral instructions. The officer running the procedure should not know which photograph is the suspect, and where that is impossible, a blinded folder-shuffling procedure or an automated system should be used. Instructions should state explicitly that the offender may or may not be present and that the investigation will continue either way.
Mechanism: this removes the two sources of expectancy - the officer's inadvertent cueing and the witness's inference that the suspect must be in the array. The 'may or may not be present' instruction licenses a 'not there' response and shifts the witness from a relative judgement (who is closest?) to an absolute one (is this the person?). Relative judgements are the main route to false identification from weak traces.
Change 2: do not re-test the same suspect, and take a verbatim confidence statement immediately. Record the witness's own words, ideally on video, before any discussion, and do not run a second identification procedure on the same suspect.
Mechanism: repeated testing does not add independent evidence - it adds contaminated evidence, because the second test is influenced by the first via source-monitoring error. A second identification gives the illusion of corroboration when it is really one weak memory tested twice. The immediate, uncoached confidence statement protects the one piece of information that is diagnostically useful: confidence recorded before feedback. Once feedback has occurred, that information is permanently lost.
Also worth noting for the trial. The appropriate remedy is not necessarily exclusion of the whole identification, but the jury should hear how the procedure was conducted. The defence's argument is strongest on the second identification, which should carry almost no independent weight.