Original article
Recall bias in reporting medically unexplained symptoms comes from semantic memory

https://doi.org/10.1016/j.jpsychores.2006.11.006Get rights and content

Abstract

Objective

When people report somatic complaints retrospectively, they depend on their memory. Therefore, retrospective reports can be influenced by general beliefs on sickness and health from semantic memory. We hypothesized that individuals with medically unexplained symptoms (MUS) would have recall biases stronger than those of people without complaints when reporting symptoms retrospectively, and that this effect would be a function of time between symptom experience and report.

Methods

To compare two time frames, 37 participants who were high and low on MUS reported momentary symptoms combined by daily recall and weekly recall using an electronic diary.

Results

Both groups reported more symptoms when recalling the entire week than what could be expected from average momentary reports. However, participants high on MUS also reported more symptoms when recalling a week than when recalling a day. For this group, recall bias was not associated with peak heuristic or symptoms variability.

Conclusion

Symptom reports in people high on MUS increases as time passes by, probably as a results of a shift in memory retrieval strategy from using episodic knowledge to using semantic beliefs.

Introduction

Psychologists, medical doctors, and other health practitioners often must rely on patients' reports of their somatic symptoms. When somatic symptoms cannot be explained in medical terms [i.e., medically unexplained symptoms (MUS)], it is often assumed that the illness, or unwellness, is “all in the mind” [1]. In earlier studies, psychological traits and states have been considered as predictors of somatic complaints [2]. For instance, neuroticism was found to correlate with self-rated health, but not with objective measures of health [3], [4], [5]. It has been suggested that high neuroticism may only be associated with the tendency to inflate the frequency or the severity of somatic complaints when individuals recall somatic symptoms [6], [7]. The association between neuroticism and the frequency of symptom reports could not be demonstrated in two studies in which somatic symptoms [6] or stress [8] was sampled close to their occurrence. Although this specific influence of neuroticism on recall of symptoms needs further confirmation, somatic symptom reports from memory clearly induce the likelihood of recall biases. Our aim was to assess bias in retrospective reports of individuals with MUS. Studying recall bias and the underlying mechanisms in these individuals is of importance in understanding the psychological mechanisms that influence their somatic complaints report.

Recall bias in reporting somatic symptoms has been demonstrated in children with headaches [9], adults undergoing painful procedures [10], and chronic pain patients [11] by comparing the average results of experience sampling with retrospective reports over the same time periods. Usually, when pain is recalled, estimates are higher than what could be expected from average momentary pain reports. Recall bias in reporting diffuse and nonspecific MUS has, however, less extensively been studied.

The accessibility model of emotional self-report by Robinson and Clore [12] explains the discrepancy between online and retrospective emotional self-reports as the difference between (reporting) feelings and beliefs. According to their model, emotional self-reports depend on principles of accessibility: People report from knowledge that is accessible. When experiential knowledge (i.e., feelings) is inaccessible, information comes from episodic or semantic memory. Emotional self-reports will therefore be influenced by either (a) relevant situational cues, (b) situation-specific beliefs, or, when situational cues are not available, (c) semantic knowledge, such as “identity-related beliefs” [12]. This model may apply to reports on somatic symptoms as well.

Momentary reports on pain tap from experiential knowledge (i.e., perceived somatic sensations). However, several studies have demonstrated that people have poor access to most internal physiological processes (such as accelerating heart rate) and that they usually make inferences about somatic symptoms based on nearby situational cues [4], [13], [14], [15]. It may be speculated that momentary reports of diffuse nonspecific somatic complaints are less accurate than momentary reports of specific pain sensations. It is very likely that pain sensation is more accessible than diffuse nonspecific sensations. Reports of momentary symptoms may therefore, in part, be influenced by directly available situational cues.

Like momentary reports, recall of symptoms may also be influenced by situational cues. Here situational cues are details of relevance tapped from episodic memory. For instance, Redelmeier and Kahneman [10] compared real-time pain and patients' memories while undergoing painful colonoscopy. They found that overall pain recollections immediately after the procedure largely reflected the intensity of pain at the worst part and at the final part of the experience. The results of Stone et al. [16] showed that pain patients with higher variability in their momentary pain reports rated the level of their recalled pain higher than those with low variability. Thus, pain recall biases can be the result of typical peak-end heuristic effects [10], [16]. It remains to be examined whether situational cues tapped from episodic memory may also influence the recall of nonspecific symptoms.

It may be argued that individuals who report MUS have personal semantic memories that contain many identity-related beliefs about somatic symptoms and illness-related cognition. For instance, memories of childhood experience of one's own illness and parental illness are a risk factor for unexplained symptoms in adulthood [17]. A wider time frame between symptom and report makes it increasingly likely that an episodic recall strategy is replaced by a semantic one [18]. This shift from using episodic memory, which includes specific lively details, to using semantic memory, which contains more general knowledge, introduces belief-consistent recall biases. Semantic memory processes might additionally play a larger role when symptoms do not possess clear features (e.g., chronic malaise compared to acute pain).

Situation-specific beliefs are general, despite being situation-related (e.g., “This is a sick building”). Semantic knowledge includes illness and health-related beliefs in general (e.g., “You should avoid stress”) or beliefs related to one's identity (e.g., “I have migraines all the time”). Thus, with longer time frames between symptom occurrence and report, “personal semantic memories” win influence according to the accessibility model. Personal semantic memories are coherent stories about oneself, which render the past with one's current self-conceptions and play an active role in how events are encoded and retrieved [12]. Dimensions of personality, such as neuroticism, can be viewed as a source of beliefs that differentially contribute to self-report. This suggests that, with moderate time intervals between symptom and report, individuals suffering from MUS, who generally score high on neuroticism, may give more biased reports because of influences by beliefs from semantic memory.

Our first aim was to study the influence of two different time frames on recall bias in reporting medically unexplained somatic symptoms. Individuals high and low on unexplained symptoms were submitted to a 1-week experience sampling of momentary symptoms using an electronic diary, combined by retrospective evaluations after each day and at the end of the week. It was hypothesized that individuals who generally report more symptoms would show stronger recall bias than individuals who report less symptoms. It was further hypothesized that this effect will be more pronounced for a longer time frame (i.e., week recall compared to day recall). Our second aim was to explore the recall of diffuse nonspecific symptoms compared to pain. This was performed because most previous work on recall bias has been solely focused on pain, while the instruments regularly used for the assessment of MUS [e.g., the somatization subscale of the Symptom Check List (SCL-90-R)] include items that assess salient pain levels but also items that assess transient, diffuse, and nonspecific sensations (e.g., feeling weak). The latter may be more susceptible to biases from semantic memory. Our final aim was to explore the influence of peak heuristic and symptoms variability.

Although recall-bias effects have been demonstrated before, the importance of the present study is that we have focused simultaneously on multiple variables that may influence recall of somatic symptoms: the influence of two time frames, group differences in unexplained symptoms, recall of pain versus nonspecific symptoms, and peak-salience heuristic and variability heuristic.

Section snippets

Participants and selection questionnaire

Participants high on recently experienced symptoms and controls (low on symptoms) were recruited from a sample of 405 undergraduate female students from the faculty of social sciences in Utrecht, The Netherlands. Only female participants were selected because males high on symptoms were underrepresented (i.e., counterbalancing the groups on gender was not possible). Participants were selected using a Dutch translation of the somatic subscale of the SCL-90-R [19]. Distress ratings of a list of

Compliance

On average, 51.4 alarms were generated for a participant, of which 39.5 momentary diary responses were provided. The mean number of momentary diary responses during a day was 5.6 (HMUS: mean=5.5; LMUS: mean=5.7). The mean compliance was 77% (HMUS: mean=76, range=63–89; LMUS: mean=78, range=57–90). Two participants had compliance of <70%. Excluding these two participants did not alter the results. Three participants had missing values in one day-recall response. All participants provided a

Discussion

All participants reported more somatic symptoms retrospectively than what could be expected from average momentary reports over the corresponding time frame. When the time frame over which both groups had to recollect symptoms increased from a day to a week, the symptoms recalled after a day by the low-symptom group were comparable to the symptoms they recalled after a week. In contrast, the group high on symptoms gave even more biased reports after a week than they did after a day. Thus, for

Acknowledgments

This study was funded by a grant from The Netherlands Organization for Scientific Research (NWO no. 452-02-011). The authors gratefully acknowledge the aid of Emma de Leng, Marjolein Raaijmaakers, Elske Stolte, and Corinne Stoop.

References (23)

  • CA Marco et al.

    Daily stress and the trajectory of mood: spillover, response assimilation, contrast, and chronic negative affectivity

    J Pers Soc Psychol

    (1993)
  • Cited by (54)

    • Retrospective memory for symptoms in patients with medically unexplained symptoms

      2018, Journal of Psychosomatic Research
      Citation Excerpt :

      The combined effect of less detailed processing of sensory-perceptual aspects of a somatic episode and exaggerated affective responses to it may make persons with MUS particularly vulnerable to retrospective memory distortions. This fits with findings in non-clinical groups showing that retrospective ratings of daily symptoms and experimentally induced dyspnea increase over time in high compared to low habitual symptom reporters [12,18] and that this increase is mediated by affective responses to the somatic event [12,19]. The distinction between sensory-perceptual and affective-motivational components in symptom reporting is in accordance with neurobiological [20], behavioral [21], and psychometric research [22–24] and may be highly relevant to understand MUS and biases in retrospective symptom memory.

    • Explanatory Style and Health

      2015, International Encyclopedia of the Social &amp; Behavioral Sciences: Second Edition
    View all citing articles on Scopus
    View full text