Discrepancies in autobiographical memories — implications for the assessment of asylum seekers: repeated interviews study

Introduction

Under the terms of the 1951 United Nations convention on the status of refugees, a refugee is someone with a well founded fear of persecution on arbitrary grounds — such as ethnicity or political opinion — who cannot achieve protection in their home country.1 When they escape to a new country, their application for asylum is considered in the light of the information they can supply and any facts known about their country. There will often be little documentary evidence about the asylum seeker, and a legal decision on status by the authorities in the country of reception may rest on their credibility as a witness.

Asylum seekers sometimes give accounts of persecution that differ with each telling (that is, they contain discrepancies). A common, understandable assumption is that the case is not a credible one — that is, the account is false or exaggerated and that is why it cannot be recalled clearly. In the United Kingdom, national guidelines state that “discrepancies, exaggerated accounts, and the addition of new claims of mistreatment may affect credibility” and that such discrepancies may be used as a reason for refugees being refused asylum.2

Generally, the more detail a memory has, the more believable and convincing the account is. The gist of an autobiographical memory (central details) can be reconstructed from general (historical or schematic) knowledge, whereas details of a specific event (peripheral details) cannot. Recall of peripheral details is thus seen as a good way of distinguishing between “accurate recollection and plausible reconstruction.”3 This is presumably the principle that, in part, guides state authorities' reliance on consistent details as an indication of credibility.

However, this view has been challenged in research on witnesses' evidence. Laboratory and field studies have shown that people recall more details that are central when an event has a high level of emotional impact, such as armed robbery, than when an event is emotionally neutral. Their recall of central details is, however, at the expense of their recall of peripheral details.4 Also, peripheral detail is more susceptible to disruption after the event. For example, discussions about the event5 and the exact wording of questions6 can change the detail of the responses given by the asylum seeker.

Emotional disorders can affect the quality of material that people remember. Clinical anxiety can lead to selective attention to threat, and depressed patients are biased towards recalling negative personal memories in favour of positive ones.7 Patients with depression and with post-traumatic stress disorder can have difficulties in retrieving specific autobiographical memories. 8 9 Refugees often present with clinical depression or with post-traumatic stress disorder.10^–12

Where the experience is highly traumatic — for example, serious injury to the person — the situation may be even more complex. There may be important differences between traumatic and non-traumatic memories. For example, initial recall of traumatic events by people with post-traumatic stress disorder typically does not involve normal narrative memory.13 In a military population, inconsistent (discrepant) recall of traumatic experiences over time has been reported,14 with a notable correlation between symptoms of post-traumatic stress disorder and an increase in traumatic memories.

Our study explores empirically the consistency of recall of traumatic and non-traumatic events in a sample of refugees with no obvious motivation to fabricate accounts. Any inconsistencies found would require an alternative explanation. We hypothesised that there would be discrepancies between an individual's accounts, that the discrepancies would be exacerbated by symptoms of post-traumatic stress disorder, and that more discrepancies would occur in peripheral details than in central details.

Methods

We invited 27 Kosovan Albanians and 16 Bosnians to take part in research about memory; 23 were men and 20 were women, all aged between 18 and 64 (mean 39.5 (SD 14.5)). The research was not part of clinical treatment or medicolegal assessment. All participants had been granted leave to remain in the United Kingdom under the United Nations High Commissioner for Refugees group programmes — that is, none of the participants had given accounts of their experiences to gain their asylum status. We obtained written (translated) informed consent from all participants and the study was granted ethical approval.

One of us (JH) interviewed all participants twice (with an interpreter). The time between interviews ranged from three to 32 weeks. At the first interview we used a translated form of the post-traumatic diagnostic scale to assess the level of post-traumatic stress.15 Participants were asked to recall a traumatic event from their experiences. Fifteen predefined questions were asked about the chosen event — for example, what was the date? Similar questions were asked about a non-traumatic event. After answering each question, participants were asked to rate that particular element as central or peripheral to their experience. At the second interview participants were asked to recall the same two events. The same 15 questions were repeated and we assessed whether the participants were depressed by using a translated form of the Beck depression inventory.16

Results

Four of the Bosnian participants left the study after the first interview. All participants reported traumatic experiences. Scores of symptom severity on the post-traumatic diagnostic scale (maximum 51) ranged from 5 to 50 (27.3 (10.9)). Participants were divided into high and low scorers: scores of ≥ 26 were categorised as high (n=19); scores of ≥25 were categorised as low (n=21). The depression scores (maximum 63), standardised to the Beck depression inventory II, ranged from 7 to 52 (mean 24.2 (11.6)). Of the 39 scores, 31 indicated probable clinical depression (score >14); 21 indicated moderate or severe depression.

The mean age of the Bosnian group was greater than that of the Kosovan group (46.2 v 35.5 years; P<0.05, t test) and the average time between interviews was significantly longer for Bosnian refugees than for Kosovan refugees (159 v 29 days; P<0.0001). No significant differences in psychopathology between the two groups were seen. To reduce the limitations of these differences on interpreting the results, and to take account of the loss of four subjects from the Bosnian group, each of the hypotheses was tested on the whole sample and then on the larger of the two subsets — the 27 Kosovan participants — alone. Significant findings only are reported.

Discussion

Discrepancies (including the provision of new information) exist between autobiographical accounts of refugees given by the same individual on two occasions up to seven months apart. These findings cannot be explained on the grounds of intent to deceive. For refugees with high post-traumatic stress disorder, more discrepancies were found with longer times between interviews. (In the asylum process, there may be months or years between the original interview and an appeal hearing.) In addition, more discrepancies are found in details rated by participants as peripheral, compared with recollection of the central gist of the event. Discrepancies therefore cannot be taken as automatically implying fabrication.

Our findings cause us to question why such discrepancies might exist. Our research cannot provide a causal explanation but it can point to some possibilities. A common difficulty reported was related to the experience of repeated events that are similar. This may have led to the recall of an event similar in type but different in detail at the second interview, or to the mixing up of two or more events. The emotional state of the refugee at the time of the interview may have affected his or her responses. For example, one participant changed his description of his treatment by military police from “we were slapped around” to “we were badly beaten.” He may simply have been in a different mood state in each interview, thus giving different evaluations of his experience. In states of depressed mood, recall is biased towards negative memories.7 Further studies should consider measuring the person's state of mood at each interview.