Autobiographical memory (AM) and metacognition are two important research fields in psychology, and yet their intersection has been surprisingly overlooked by scientists, despite theoretical statements which point to a role for metacognition in AM retrieval. AM models assume that control processes are deployed during memory retrieval. Such processes are supposed to examine the veracity of the memory and its correspondence to the individual’s goals[1, 2]. When control processes are satisfied, the memory reaches consciousness, when they are not, a new retrieval is undertaken (e.g.[3]). Other accounts of autobiographical retrieval propose that memory experiences and feelings metacognitively guide the construction of past events from the elaboration of a cue (e.g.[4]). Even though such processes have been suggested, the extent to which we can readily monitor what we retrieve from AM has not been documented. The aim of this project was to be a piece that fills this gap in the literature. Specifically, it aimed to investigate whether individuals were able to judge the veracity of what they retrieved from their AM, that is, metacognitive abilities related to AM.
Metacognition research distinguishes two types of task performance[5]. The first order performance is the individual performance at the task (e.g., the percentage of correct answers). The second order performance considers how accurately someone evaluates their first order performance, i.e.metacognitive monitoring (e.g., the percentage of correct answers estimated by the participant). A typical way to assess the second order performance in a memory task is to investigate the confidence participants have in their right and wrong answers. Appropriate metacognitive monitoring is shown when participants distinguish their right and their wrong answers with their confidence judgements.
Several researchers have considered the basis for such second order performance[6]. Confidence can be influenced by several characteristics, notably first order fluency. Fluency is the subjective experience of ease or difficulty with which we are able to process information[7]. It is often measured using response time in experimental contexts. A common finding is that the more an answer is fluent, the more the confidence is high[8]. Fluency is in turn influenced by task difficulty, that is, the more the task is difficult, the more the feeling of difficulty is strong, the more confidence is low[9].
The aim of this study was to assess metacognitive monitoring in an AM task. In metacognitive studies, participants’ first order performance is assessed with reference to an objectively correct answer (e.g., in a lexical decision task or word learning task; see[10]). As such, the confidence participants gave to correct and incorrect answers can be compared. AM studies encounter an issue here: AM is personal and subjective, thus researchers can not easily judge the veracity of answers in an AM task - and it is not usual to classify autobiographical retrievals as ‘right’ or ‘wrong’ even though some aspects of accuracy and specificity are typically measured. Our solution was to consider the subjective organisation of memories, and have people give confidence judgements for ordering of events from their personal past in a laboratory task. Note that, if someone indicates that their wedding occurred before their graduation, we can not be sure of the true order of these events, but we can assume that they should be coherently and consistently reported as occurring in the same order.. Moreover, in line with the notion of AM being highly reconstructive[11], research has shown that AM chronology does not perfectly match with real life chronology. For example, Burt et al. (2000)[12] asked participants to keep a personal diary for a few months. Participants had to record events and their date of occurrence. After a delay, they had to indicate the chronology of the events they wrote in the diary and interestingly -- given the aim of this paper -- to indicate the confidence they had in their answers. Results showed that the chronology given by the participants differed from the diary chronology. Burt et al. found no relationship between confidence and performance in this ordering task, when gauged objectively - i.e. compared to the diary.
We argue that to better evaluate metacognition in AM, we should consider its relationship with subjective organisation, hence our focus on coherence. Indeed, AM does not necessarily correspond to reality, but is organised in order to support the individual’s goals[13]. Here we are interested in assessing AM subjective chronology. In order to do this, we asked twice the order in which events occurred, and for the purposes of evaluating metacognitive accuracy suppose that the consistent responses are the ‘correct’ answers. This gives us an evaluation of the first order performance taking in consideration AM subjective organisation, and hence the possibility to assess second order performance.
Our task assessed autobiographical organisation according to the Self-Memory System[14]. This model postulates that AM has an hierarchical structure (Fig. 1). The highest level is the life story. It is a representation of one’s life and underlines the goals of the individual over the time[15]. The life story is divided in lifetime periods representing important periods of the individual’s life such as the different scholar periods[16]. These periods contain general events which regroup thematically or temporally close autobiographical memories. Autobiographical memories are supposed to follow a stable chronology that underlines the goals of the individual over their life[17]. This chronology can contradict the objective chronology to support the individual’s goals[13].
Insert Fig. 1 About Here
Skowronski et al. (2003; 2007)[18, 19] managed to investigate AM chronology with respect to its subjectivity. They asked participants to generate memories from two lifetime periods: their high school and their college. Generated memories were regrouped by pairs. Two types of pairs were created: Intraperiod pairs of memories, containing two memories from the same lifetime period, and interperiod pairs of memories, containing two memories from different lifetime periods. During the judgement of order task, pairs of memories were presented one at a time on a screen, and participants had to indicate either which event of the pair occurred first, or which event of the pair occurred last. Next, participants had to give the date of occurrence of every memory they generated. Each judgement of order was categorised as coherent or noncoherent with the chronology given by the dates (e.g., if the participant indicated that their graduation, 2012, occurred before their wedding, 2014, the judgement was coherent). Results showed that judgements of order were more successful when realised with interperiod pairs of memories than with intraperiod pairs of memories, that is, variations in task difficulty. Moreover, judgements of order with interperiod pairs were performed faster than judgements of order with intraperiod pairs. In this study, judgements of order were categorised as right or wrong through dates given for the events. Because dates in memory are associated with many biases and lack of precision, a judgement of order is a better way to assess AM chronology[20, 21], thus we proposed a new design based on multiple judgements of order.
Participants first generated autobiographical memories from two lifetime periods. These memories were organised into pairs (e.g. my bachelor graduation day - the wedding of my best friend). In a second phase, each pair of memories was subject to two judgements of order of occurrence in two separate blocks. These two judgments allowed us to distinguish two types of pairs of memories. A coherent pair of memories was a pair in which the judgements of order were consistent across blocks (e.g., if a participant indicated in one block that their graduation occurred before their wedding, and in the other block that their wedding occurred after their graduation). A noncoherent pair was a pair in which the two judgements were contradictory across blocks indicating that one judgement of order conflicted with the life story chronology (e.g., if a participant indicated in one block that their graduation occurred before their wedding, and in the other block that their graduation occurred after their wedding). Every judgement of order was followed by a confidence rating on a 1–7 scale with seven being the highest confidence. We thus categorised each pair as coherent or noncoherent, and hypothesised a greater confidence for coherent pairs of memories than in noncoherent pairs.
We also tested several exploratory hypotheses. We also asked participants to date their memories and we categorised every judgement of order as right or wrong the same way Skowronski et al. (2007)[19] did. Indeed, if a judgement of order matched with date chronology, it was categorised as date-coherent judgement of order. If a judgement of order conflicted with dates chronology, it was categorised as a date noncoherent judgement of order. Next, we tried to replicate Skowronski et al.[18, 19] results showing that intraperiod judgements of order (i.e., judgements of order with two memories from the same lifetime period) were less successful and slower than interperiod judgement of order (i.e., judgements of order with two memories from different lifetime periods). Finally, we explored the role of fluency in metacognitive ratings, that is, the effect of task difficulty on participants’ confidence, and the link between response time and confidence.