What are confabulators’ memories made of? A study of subjective and objective measures of recollection in confabulation

Abstract

Confabulating patients claim to remember events that had not actually happened, suggesting a vivid subjective experience of false memories. The present study was aimed at examining the nature of subjective experience of retrieval in confabulators and its relation to the objective ability to recollect qualitative aspects of the original episode. In Experiment 1, 5 confabulators, 7 non-confabulating patients, and 12 control subjects studied words under shallow and deep encoding conditions and underwent a Remember (R)/Know (K) recognition task [Tulving, E. (1985). Memory and consciousness. Canadian Psychology, 26, 1–12]. For recognized words, they additionally indicated two qualitative features of the encoding context. Whereas subjective (i.e. R responses) and objective (i.e. source) measures of recollection were associated in normal controls and non-confabulating patients, they were dissociated in confabulators. In Experiment 2, participants explained the content of their R responses. We found that confabulators’ recollections mainly included autobiographical information related to test items, but not to the study encounter. We conclude that remembering states in confabulators may be linked to a deficit in inhibiting irrelevant memories triggered by test items during retrieval attempts.

Introduction

Confabulating patients have memories for events that have not been actually experienced. These false memories, termed confabulations (see Gilboa & Moscovitch, 2002 for a recent review), may include as many contextual and perceptual details as do memories of true events (Johnson, O’Connor, & Cantor, 1997). Confabulation is generally considered an impairment of retrieval rather than encoding processes (Burgess & Shallice, 1996; Moscovitch, 1989; Moscovitch & Melo, 1997; Schacter, Norman, & Koutstaal, 1998), consistent with the fact that it may affect memories acquired before the occurrence of brain damage (when encoding operations were well-functioning) as much as those acquired subsequent to brain damage.

Of great interest is confabulators’ subjective experience of retrieval: patients may claim they can re-live these confabulatory reports, suggestive of experiences of vivid recollection. For example, Dalla Barba (1993) described a patient, MB, who confabulated in response to questions tapping personal episodic memory (Dalla Barba, 1993). When asked to classify his memories according to the remember (R)/know (K) distinction (Tulving, 1985), that is, to indicate whether he recollected some aspects of the original episode (i.e., R response) or whether the memory was merely familiar (i.e. K response), MB frequently attributed R judgements to his confabulatory reports.

Not only confabulators claim to re-live their false memories, but they apparently do. That is, there is evidence that some confabulating patients act upon their confabulations as if they were true memories (e.g., looking for the phone number of a non-existing friend; see Schnider, 2003), or even show emotional responses congruent with the content of a confabulated event (e.g., crying for the alleged death of a work colleague). This complete adherence to confabulations may be related to the way patients subjectively experience their memories, given that people base their decisions and actions upon their own experience (Bechara, Damasio, Tranel, & Damasio, 1997; Goel & Dolan, 2003; Koriat & Goldsmith, 1996). Thus, understanding the nature of the subjective experience associated to memory retrieval in confabulators may contribute to our understanding of the mechanisms responsible for confabulatory behaviour.

The goal of the present study was to characterize the subjective experience of remembering in confabulating compared to non-confabulating patients and control participants. We attempted to achieve this goal by investigating three issues. First, we investigated the extent to which subjective measures of recollection (i.e., R responses in the R/K paradigm) behave similarly to objective behavioural indicators of recollection (i.e., memory for details surrounding the encoding context). Second, we examined the extent to which subjective measures of recollection are affected by variables that are known to affect objective measures of recollection in these three groups of participants. Third, we sought to identify the type of information that triggers remembering states in confabulators compared to the other participant groups.

To date, there exists substantial agreement that recollection is the process responsible for the retrieval of contextual details surrounding the original episode (e.g., Atkinson & Juola, 1974; Jacoby, 1991, Mandler, 1980, Tulving, 1985, Yonelinas, 1994). Recollection is distinguished from familiarity, which is the process that reflects the global strength of the memory trace without additional qualitative information (Yonelinas, 1994). Individuals are thought to be able to report on their subjective experience of recollection and to differentiate it from the experience of familiarity (Gardiner, Ramponi, & Richardson-Klavehn, 1998; Tulving, 1985). These different subjective experiences of retrieval have been operationalized in the R/K paradigm (Tulving, 1985), in which participants are asked to indicate, for each recognized item, whether they recollect some aspects of its presentation (i.e. R response) or whether the item is merely familiar (i.e. K response). Numerous behavioural (Gardiner, 1988; Gardiner & Parkin, 1990; Rajaram, 1993) and neuroimaging studies (Eldridge, Knowlton, Furmanski, Bookheimer, & Engel, 2000; Wheeler & Buckner, 2004; Woodruff, Johnson, Uncapher, & Rugg, 2005) provided evidence consistent with the hypothesis that R and K responses reflect qualitatively distinct memory processes underlying recognition performance (Yonelinas, 1994; but see Dunn, 2004, Wixted and Stretch, 2004).

If the R/K paradigm captures the distinction between recollection and familiarity, one should find that when R responses are provided individuals are more likely to correctly retrieve features of the encoding context than when K responses are provided. In other words, subjective and objective measures of recollection should be specifically associated. In line with this notion, Dudukovic and Knowlton (2006) have recently demonstrated that normal subjects were able to identify contextual details (i.e. the ink color) for remembered items, while their source accuracy for familiar items was at chance levels (but see Hicks, Marsh, & Ritschel, 2002). Similarly, Perfect and colleagues demonstrated that normal subjects were more likely to identify the temporal (i.e. the order of occurrence in the original list) and the spatial context (i.e. the quadrant of the screen in which the item had been presented) for remembered compared to familiar items (Perfect, Mayes, Downes, & Van Eijk, 1996). Moreover, ERP studies showed that the LPC, a late positive component that is typically associated to R responses (Duzel, Yonelinas, Mangun, Heinze, & Tulving, 1997), was detected only when both the source of an information and its content were retrieved (Rugg, Henson, & Robb, 2003; Senkfor & Van Petten, 1998).

Subjective estimates of recollection have also been compared to those obtained with the process dissociation procedure (Jacoby, 1991), in which recollection is measured as the ability to retrieve a specific aspect of the study event (e.g., when or where an item was presented) and to use this as a basis for controlled responding. The two procedures led to almost identical estimates of recollection, which were also similarly affected by encoding manipulations (Yonelinas, 2001). These results suggest that phenomenological experiences of recognition memory, that is R and K responses, function similarly to objective measures of recollection and familiarity in normal subjects.

Of importance, subjective and objective estimates of recollection also converge in some memory-impaired populations. Compared to normal controls, amnesic patients typically show a reduction of R responses in recognition tasks (Dalla Barba, 1997, Knowlton and Squire, 1995; Schacter, Verfaellie, & Pradere, 1996), and are also markedly impaired in objective measures of recollection, such as those derived from the process dissociation procedure (Verfaellie & Treadwell, 1993; Verfaellie, 1994), or the ROC analysis (Yonelinas, Kroll, Dobbins, Lazzara, & Knight, 1998; Yonelinas et al., 2002). Recently, however, dissociations between subjective and objective indicators of recollection have been observed. Duarte and colleagues found that patients with focal lesions in the dorsolateral prefrontal cortex (DLPFC) are impaired in recollecting the context in which items had been encountered, yet show seemingly unaltered remembering states (Duarte, Ranganath, & Knight, 2005).

The question of interest for the current research concerns the extent to which subjective and objective measures converge in confabulating patients. On the one hand, the frequent creation of false memories and attributions of true memories to wrong contexts are clear indicators of impaired (objective) recollection. On the other hand, the vividness with which these faulty memories appear to be experienced suggests that these patients may still have remembering states. It is possible, then, that the subjective experience of remembering and the objective ability to recollect qualitative features of the study context are not as tied in confabulating patients as they are in normal subjects. We have previously argued that whereas in normal subjects remembering states are thought to reflect the retrieval of contextual features of the learning episode, in confabulators remembering states may be triggered by vivid thoughts related to test items but independent of the context in which these items were originally encountered (Ciaramelli, Ghetti, Frattarelli, & Làdavas, 2006). If this is the case, the association between subjective and objective measures of recollection in confabulating patients should be lower than that observed in non-confabulating patients and normal controls. The question of whether or not subjective and objective measures of recollection converge in confabulating patients motivated Experiment 1.

One additional way to gain insight on the behaviour of subjective measures of recollection in confabulators compared to other participant groups is the investigation of how subjective and objective measures are affected by experimental manipulations that typically result in changes in recollection. The literature is rich of examples of such conditions (see Yonelinas, 2002 for a review). For example, the study of experimental manipulations at time of encoding has demonstrated that processing the meaning of the stimulus versus its perceptual features increases recollection more than familiarity (e.g., Gregg & Gardiner, 1994; Yonelinas, 2002). Further, divided versus full attention reduces recollection more than familiarity (e.g., Jacoby & Kelley, 1991; Yonelinas, 2001). Similar dissociations are also found following retrieval manipulations (e.g., Dodson & Johnson, 1996; Yonelinas & Jacoby, 1994).

If subjective measures of recollection rely on different processes than objective measures of recollection in confabulating patients compared to other groups, then conditions that support recollection should differentially affect subjective and objective measures in confabulating patients, but not in other groups. This question was also addressed in Experiment 1.

As a final issue, we sought to begin to shed light into the cause of the dissociation between objective and subjective measures of recollection in confabulation. As we described earlier, a dissociation between subjective and objective measures of recollection has been recently reported in DLPFC patients (Duarte et al., 2005). In this study, however, little emphasis was placed on the neuropsychological mechanism responsible for the dissociation. In this respect, a case-study by Curran and colleagues is quite revealing (Curran, Schacter, Norman, & Galluccio, 1997). They described a patient with a right DLPFC lesion, BG, whose R responses predominantly included associations made to the test words (e.g., for the word DISEASE: “I remember this word because one of the reasons I’m going to the dentist this afternoon is some gum disease”; Curran et al., 1997, p. 1041), rather than contextual aspects of the learning episode. These results suggest that BG's subjective experience of retrieval did not necessarily reflect any re-access to the study phase. We note, however, that BG (Curran et al., 1997), as well as patients studied by Duarte and colleagues (Duarte et al., 2005), did not suffer from confabulation, thus the mechanism underlying the dissociation between objective and subjective recollection in these patients may differ from the one operating in confabulating patients.

Our hypothesis is that confabulating patients still experience remembering states in the face of clearly impaired retrieval processes because remembering states in these patients reflect something different from the accessibility of information about the encoding context. As we discussed earlier, the experience of remembering in confabulators might be triggered by an excessive processing of context-irrelevant information related to test items during retrieval (Ciaramelli, Ghetti, Frattarelli et al., 2006). To test this hypothesis, we conducted Experiment 2, in which we asked participants to describe the contents of their R responses. Whereas R responses would be associated with the description of contextual features of the learning episode in other participant groups, such responses would be likely associated with extra-list contexts in confabulating patients (Gardiner et al., 1998).