Abstract
People tend to recall more specific personal events from adolescence and early adulthood than from other lifetime periods, a finding known as the reminiscence bump. Several explanations have suggested that events from the reminiscence bump are especially emotional, important, or positive, but studies using cue words have not found support for these claims. An alternative account postulates that cognitive abilities function optimally in adolescence and early adulthood, which may cause more memories to be stored in those lifetime periods. Although other studies have previously discussed the cognitive abilities account as a possible explanation for the reminiscence bump, it was only recently shown that cognitive abilities are indeed related to autobiographical memory performance. When this recent finding is combined with previous findings that cognitive abilities as well as autobiographical memory function optimally in adolescence and early adulthood, they suggest that the cognitive abilities account is a promising explanation for the reminiscence bump in the temporal distribution of word-cued memories. However, because the account does not aim to explain the reminiscence bump in the distribution of highly significant events, it should be regarded as complementary to the existing accounts.
Keywords
Autobiographical memory
Autobiographical memory refers to the memories a person has of his or her own life experiences (Robinson, 1986). It helps people to cope with current problems and prepare for future actions, to develop and maintain emotional bonds with others, and to establish and maintain a sense of identity over time (Bluck, Alea, Habermas, & Rubin, 2005). It is taxonomically speaking a part of episodic memory (Squire, Knowlton, & Musen, 1993), where it forms a rather complex category. It not only consists of vivid memories of important events but also contains memories of everyday events, generic or repeated memories, and autobiographical facts (Brewer, 1986; Conway, 1987).
One of the most consistently observed effects in autobiographical memory is the reminiscence bump in its temporal distribution. Whereas people hardly remember any specific personal events from the first 3 or 4 years of their lives (i.e., childhood amnesia), they tend to recall many memories from the 5 to 10 most recent years (i.e., increased recall of recent events). Besides these two effects, people also tend to recall a higher proportion of specific personal events from the period in which they were between 10 and 30 years old than one would expect based on a forgetting curve, a finding known as the reminiscence bump (Rubin, Wetzler, & Nebes, 1986).
The goal of the present review is introducing the cognitive abilities account for the reminiscence bump in the temporal distribution of autobiographical memory. Although the account has been mentioned as an alternative explanation (e.g., Rubin & Schulkind, 1997a), it has until recently received little consideration. The present review will not provide conclusive support for the account, but it will expand on the tenets of the explanation and discuss some preliminary support.
The present review consists of three parts. In the first part, the main findings and existing accounts regarding the reminiscence bump will be discussed. Special attention will be given to the finding that the reminiscence bump in the distribution of word-cued memories and the reminiscence bump in the distribution of highly important personal events might be separate phenomena (Janssen, 2015a; Rubin, 2015). There does not appear to be many studies that have reported a reminiscence bump after the age of 20 years when examining the distribution of word-cued memories, whereas there similarly does not appear to be any study that has reported a reminiscence bump before the age of 15 years when examining the distribution of highly important personal events (Koppel & Berntsen, 2015).
In the second part, the cognitive abilities account, which argues that the reminiscence bump in the distribution of word-cued memories is caused by age-related changes in the ability to bind spatial and temporal information to personal experiences, will be introduced. The account argues that cognitive abilities function optimally in adolescence and early adulthood, that autobiographical memory similarly functions optimally in these lifetime periods, and that cognitive abilities and autobiographical memory are related. Attention will also be given to the basic-systems approach on which the cognitive abilities account is built.
In the final part, support and limitations of the cognitive abilities account will be discussed. Not only will this discussion draw on findings from the autobiographical memory literature, it will also draw on findings from other domains, such as the memory for public events. Special emphasis will be placed on the role of retrieval processes. In addition, future studies will be suggested to examine the cognitive account further. It will be concluded that, although the cognitive abilities account is a promising explanation for the reminiscence bump, it should be seen as complementary to the existing accounts, because it does not explain the reminiscence bump in the distribution of highly significant events.
Part 1
The reminiscence bump
The reminiscence bump in the temporal distribution of autobiographical memory is a robust phenomenon. It has been observed in memories that have been collected with a wide range of methodologies. It has been identified when participants report the most important events from their lives (e.g., Berntsen & Rubin, 2002; Fitzgerald, 1996; Holmes & Conway, 1999, Study 1; Rubin & Berntsen, 2003), when participants indicate the high and low points of their lives on a timeline (e.g., De Vries & Watt, 1996; Elnick, Margrett, Fitzgerald, & Labouvie-Vief, 1999; Schroots & Assink, 2005), when participants report personal events connected to self-concepts (e.g., Rathbone, Moulin, & Conway, 2008), when participants are given a limited amount of time (e.g., five minutes) to recall as many personal events as possible from different lifetime periods (e.g., Conway & Holmes, 2004, Exp. 1; Demiray, Gülgöz, & Bluck, 2009; Howes & Katz, 1992), and when participants are given a series of cue words and instructed to describe for each cue word the personal event that comes to mind first (e.g., Fitzgerald & Lawrence, 1984; Franklin & Holding, 1977; McCormack, 1979; Rubin et al., 1986).
The reminiscence bump has also been observed in many different populations. It has been identified in the distributions of both men and women (e.g., Niedźwieńska, 2003; Rubin, Schulkind, & Rahhal, 1999) and in the distributions of both middle-aged and older adults (e.g., Hyland & Ackerman, 1988; Jansari & Parkin, 1996). It has even been found in the distributions of young adults when their distribution is adjusted for the increased recall of recent events (e.g., Janssen, Chessa, & Murre, 2005; Janssen, Gralak, & Murre, 2011). Furthermore, besides in the distributions of autobiographical memories of many North-American and European samples, the reminiscence bump has also been identified in the distributions of Asian (e.g., Conway & Haque, 1999; Kawasaki, Janssen, & Inoue, 2011; Maki, Janssen, Uemiya, & Naka, 2013; Maki & Naka, 2006) and Caribbean (e.g., Alea, Ali, & Marcano, 2014) samples.
The reminiscence bump also exists outside the domain of autobiographical memory. It has been found in the distributions of favorite books, movies, and records (e.g., Holbrook & Schindler, 1989, 1996; Janssen, Chessa, & Murre, 2007; Larsen, 1996; North & Hargreaves, 1995; Rathbone, O’Connor, & Moulin, 2017; Schulkind, Hennis, & Rubin, 1999; Sehulster, 1996; Zimprich & Wolf, 2016) and in the distributions of memory for public events (e.g., Howes & Katz, 1992; Janssen, Murre, & Meeter, 2008; Koppel & Berntsen, 2016; Rubin, Rahhal, & Poon, 1998; Schuman, Belli, & Bischoping, 1997). The reminiscence bump also affects judgments of public events and figures. Schuman and coworkers (e.g., Schuman, Akiyama, & Knäuper, 1998; Schuman & Corning, 2014; Schuman & Scott, 1989) and others (e.g., Holmes & Conway, 1999, Study 1) found that participants more frequently named public events that had happened during their adolescence and early adulthood as events that had changed the world or that were especially important. Janssen, Rubin, and Conway (2012) asked participants to name the five best soccer players and found that participants tended to name players who reached the midpoint of their career when the participants were in their late teens.
Although the reminiscence bump is arguably one of the most robust findings in autobiographical memory, there are some notable disparities when different methods are used to collect the personal events (Rubin & Schulkind, 1997a). First, the distribution of word-cued memories contains an increased recall of recent events, whereas this recency effect is absent from the distribution of highly important events. Second, the size and location of the reminiscence bump also depend on the method employed. When words are used to cue the memories, the reminiscence bump is relatively small and located in the second decade (10–20 years), whereas it is larger and located in the third decade (20–30 years) when participants are asked about the most important events from their lives. Although the study by Rubin and Schulkind (1997a) is the only study in which the two methods are directly compared, there are hardly any studies that have found a reminiscence bump after the age of 20 years when examining the distribution of word-cued memories or before the age of 15 years when examining the distribution of highly important personal events (Koppel & Berntsen, 2015). Third, there is no bias to report a higher proportion of positive memories in the reminiscence bump when memories are elicited with words (e.g., Janssen & Murre, 2008; Maki & Naka, 2006; Rubin & Schulkind, 1997a), but there is such a positivity bias when people report the most important events from their lives (e.g., Berntsen & Rubin, 2002; Haque & Hasking, 2010; Rubin & Berntsen, 2003). The reminiscence bumps elicited by the two methods are often considered manifestations of the same phenomenon, but they have different properties. These differences suggest that the two reminiscence bumps might be separate phenomena (Janssen, 2015a; Rubin, 2015).
Other methods that can be used to collect autobiographical memories have not been compared directly. However, it appears that personal events retrieved with the timeline method (e.g., De Vries & Watt, 1996; Elnick et al., 1999; Schroots & Assink, 2005), the method in which participants report personal events connected to self-concepts (e.g., Rathbone, Conway, & Moulin, 2011; Rathbone et al., 2008), and fluency tasks (e.g., Conway & Holmes, 2004, Exp. 1; Demiray et al., 2009; Howes & Katz, 1992) have similar content, distributions, and properties (e.g., importance) as personal events that are retrieved with requests for the most important events (e.g., Berntsen & Rubin, 2002; Demiray & Janssen, 2015; Fitzgerald, 1996; Holmes & Conway, 1999, Study 1; Rubin & Berntsen, 2003). Thus, the data collected using the cue-word method seem to stand out as resulting a reminiscence bump that is different from the reminiscence bump resulted from other methods.
Explanations for the reminiscence bump
Although there are some other recent accounts for the occurrence of the reminiscence bump in the distribution of autobiographical memory (e.g., Brown, Schweickart, & Svob, 2016; Demiray et al., 2009; Wolf & Zimprich, 2016), four explanations are often cited in the literature: The cognitive account, the identity-formation account, the self-narrative account, and the cultural life script account.
The cognitive account assumes that more novel and distinct events occur in adolescence and early adulthood (Pillemer, 2001; Robinson, 1992), such as the first driving lesson or the first kiss, than in other lifetime periods. These first-time experiences are stored more strongly into memory than recurring experiences, because they are distinct. They are also retrieved more frequently, because they are used as exemplars when people encounter similar events later in life.
According to the identity-formation account, people form their identity in adolescence and early adulthood (Conway, 2005; Conway & Pleydell-Pearce, 2000). Identity is formed through self-defining memories which are vivid and emotional memories of personal events that have a large impact on a person’s identity (Conway, Singer, & Tagini, 2004), and these memories often occur during the reminiscence bump period. The self-narrative account is similar to the identity-formation account but assumes that the concept of identity takes the form of a narrative about the self (Fitzgerald, 1988, 1996).
The cultural life script account argues that transitional events tend to occur during late adolescence and early adulthood (Berntsen & Rubin, 2002; Rubin & Berntsen, 2003). These transitional events (e.g., marriage, having children, first full-time job) are highly important and prevalent, and they are expected to occur at specific ages. When people are asked to tell their life story or report the most important events of their own lives, they use semantic information (e.g., Janssen & Rubin, 2011; Janssen, Uemiya, & Naka, 2014) about the general structure of life stories to recall highly significant events from their own lives (Berntsen & Rubin, 2004).
These four accounts are able to explain the results of studies in which participants report the most important events from their lives, but they have difficulties explaining the results of studies in which participants retrieve specific personal events with the help of cue words. Rubin and Schulkind (1997a), Conway and Haque (1999), Maki and Naka (2006), and Janssen and Murre (2008) showed that, when using cue words, there is no difference between memories from the reminiscence bump and memories from other lifetime periods on self-rated novelty, emotionality, valence, and importance. There are not only many novel, emotional, positive, or important events in the reminiscence bump period as predicted by those accounts, but there are also many mundane, everyday events rated low on these variables (i.e., recurring, unemotional, neutral, and unimportant events) in this period. These findings support the idea that the reminiscence bumps elicited with the two methods might not be manifestations of the same phenomenon (Janssen, 2015a; Rubin, 2015).
The cultural life script account is the only explanation so far that has explicitly addressed the disparities between the two methods (Berntsen & Rubin, 2004; Rubin, Berntsen, & Hutson, 2009). It has been argued that cultural life scripts do not represent an average life, but they represent an idealized life from which many common and some important events are left out. As a result, cultural life scripts are distorted from actual lives to favor positive events and events expected to occur in early adulthood (Berntsen & Rubin, 2004). The cultural life script account can explain the disparities between the two methods when participants would employ cultural life scripts when they would be asked to report important personal events but not when they would be asked to recall autobiographical memories with the help of cue words.
Berntsen and Bohn (2010) examined whether cultural life scripts are used when young adults retrieve personal events with cue words and when they report the most important events from their lives. As predicted by the account, events elicited with cue words were less often life script events (20%) than important events (71%). Similarly, Ece and Gülgöz (2014, 2017) showed that cultural life scripts are used when people are asked to recall important personal events but not when they are asked to recall memories with the help of cue words. They listed the 10 most frequently mentioned life script events from a previous study and found that the reminiscence bump disappeared from the distribution of important personal events, either when participants were instructed to exclude events from the list or when events from the list were removed afterwards. Removal of events from the list did not affect the distribution of word-cued memories.
It is believed that, because cultural life scripts are employed when participants report important events but not when autobiographical memories are elicited with words (Berntsen & Bohn, 2010; Ece & Gülgöz, 2014, 2017), there is a shift in the location of the reminiscence bump from the second to the third decade (when comparing word-cued and highly important memories) and the reminiscence bump is present in the distribution of important positive events but absent in the distribution of important negative events. However, the cultural life script account does not offer an explanation for the reminiscence bump in the distribution of word-cued memories, and the other accounts have not addressed the findings that the location of the reminiscence bump shifts when memories are gathered with different methods or the findings that the reminiscence bump in word-cued memories also contains many unremarkable events (i.e., events rated low on importance, emotionality, or novelty).
Part 2
The cognitive abilities account
Besides these four often-cited accounts, there is a fifth explanation called the cognitive abilities account. The account, which is sometimes called the biological or maturation account, hypothesizes that the reminiscence bump is caused by age-related changes in the ability to bind spatial and temporal information to personal experiences (i.e., encoding efficiency), which cause more memories to be stored (or memories to be stored more strongly) in adolescence and early adulthood than in other lifetime periods (Janssen & Murre, 2008; Janssen et al., 2008; Rubin et al., 1998).
The account should be regarded as complementary to the other accounts. It aims to explain the results of studies in which participants retrieve specific personal events with the help of cue words (but not the results of studies in which participants only report highly significant events). These specific personal events have occurred to the person remembering the event at a certain place at a time. They therefore contain at least four pieces of information (Barsalou, 1988): what the event was about (i.e., activity), who was involved (i.e., people), where the event took place (i.e., location), and when the event took place (i.e., time). Besides these four pieces, people might also recall other information: setting and background of the event; sensations, perceptions, cognitions, and emotions experienced during the event; events that occurred directly before or after the event; and the order of smaller events within the larger event.
The account assumes that autobiographical memories are not stored in one place in the brain but distributed patterns of activation. When someone experiences an event, the experience activates multiple regions in the neocortex (e.g., visual stimuli activate the visual cortex and auditory stimuli activate the auditory cortex). The medial temporal lobe encodes experiences by binding the activation in these separate brain regions, thus forming associations between the stimuli that were presented at the same time. Subsequent retrieval of events is caused by the activation of the representation of one encoded stimulus, which leads to the activation of the representations of the other associated stimuli until the entire pattern has been activated (Danker & Anderson, 2010).
According to the account, autobiographical memories are complex reconstructions of past experiences. They can involve seeing, hearing, smelling, tasting, and touching, and they can vary greatly in spatial, temporal, emotional, and narrative content. When remembering some memories, people can see, hear, or smell these events in their mind and recall details, such as the background and setting of the events and the order of actions within the events. However, for other memories, they might not see, hear, or smell those events in their minds and not recall as many details.
This account is based on three premises. It postulates (1) that cognitive abilities function optimally in adolescence and early adulthood, (2) that autobiographical memory similarly functions optimally in adolescence and early adulthood, and (3) that autobiographical memory is related to cognitive abilities. The account argues that, because the ability to bind spatial and temporal information to personal experiences is working at an optimum in adolescence and early adulthood, not only many remarkable but also many unremarkable specific personal events are retained from these lifetime periods.
Cognitive abilities across the life span
There is substantial support for the first premise that cognitive abilities function optimally in adolescence and early adulthood. Many studies have demonstrated that cognitive abilities are affected by age. Young adults tend to perform better than middle-aged and older adults on task measuring cognitive abilities (e.g., Cerella & Hale, 1994; Fiore, Borella, Mammarella, & De Beni, 2012; Grégoire & Van der Linden, 1997; Li et al., 2004; Park et al., 2002; Salthouse, 2004).
In addition, several large-scale studies have included children and adolescents to examine the development of cognitive abilities across the life span. One of these studies is Maylor and Logie (2010), who examined with more than 300,000 participants (between 8 and 50 years old) the development of prospective memory across the life span and found that performance peaked around the ages of 16 to 17 years for women and 18 to 19 years for men.
Another study is Reimers and Maylor (2005), who measured the costs of switching between categorizing faces on the basis of gender or on the basis of emotion with more than 5000 participants (between 10 and 66 years old). The difference in the reaction time between blocks in which participants did not have to switch tasks and blocks in which they had to switch tasks was smallest around the ages of 16 to 17 years.
A third study that examined the development of cognitive abilities across the life span is Murre, Janssen, Rouw, and Meeter (2013), who examined the results of more than 28,000 participants (between 11 and 80 years old) who took at least 1 of 10 verbal memory and visuospatial memory tests. They found that adolescents and young adults performed better on these tests than middle-aged adults, who in turn performed better than older adults did. Whereas performance on the verbal memory tasks peaked between ages 14 and 26 years, performance on the visuospatial memory tasks peaked between ages 16 and 18 years.
Similar peaks in performance in adolescence and early adulthood have been found on intelligence and memory tests. Hartsthorne and Germine (2015, Exp. 1) reanalyzed the data from nearly 2500 participants (between 16 and 89 years old) that had been used to normalize the scores on the third editions of the Wechsler Adult Intelligence Scale and the Wechsler Memory Scale. The two youngest age groups (ages 16–17 and 18–19) performed best on many subscales considered to measure fluid abilities, whereas older age groups tended to perform better on subscales intended to measure crystallized abilities.
Autobiographical memory across the life span
The initial rise and the subsequent fall of the performance on prospective memory tests (Maylor & Logie, 2010), task-switching tests (Reimers & Maylor, 2005), verbal memory and visuospatial memory tests (Murre et al., 2013), and intelligence and memory tests (Hartsthorne & Germine, 2015, Exp. 1) across the life span support the first premise that cognitive abilities function optimally in adolescence and early adulthood. In addition to these findings, there are similar age-related changes across the life span in the storage, consolidation, and retrieval of specific personal events, supporting the second premise of the cognitive abilities account that autobiographical memory functions optimally in adolescence and early adulthood.
From infancy to young adulthood, there is a gradual increase in children’s ability to recall autobiographical memories. Although there are other factors involved (Nelson & Fivush, 2004), researchers have attributed the failure to retrieve memories from early childhood to the fact that the brain was not fully developed when the events occurred (Richmond & Nelson, 2007). More specifically, the ability to recall specific personal events has been linked with the ability to bind objects with the place and time that they appeared and with other objects that appeared, because binding provides the context for the event which aids later recall (e.g., DeMaster & Ghetti, 2013). Lee, Wendelken, Bunge, and Ghetti (2016) examined the ability to bind objects with spatial (item-space) and temporal (item-time) information and other objects (item-item) in children (between 7 and 11 years old) and young adults. Although they observed no differences between children aged 10 and 11 years and young adults for the item-space task, there were still differences between children aged 11 years and young adults for the item–item and the item–time tasks, suggesting that bindings develops throughout adolescence.
Studies examining binding have not only reported age-related increases in how well people bind information from childhood to young adulthood (Lee et al., 2016), there are also studies that have reported age-related declines from young to late adulthood. Whereas older adults perform worse than young adults on tasks that require them to separately recall objects, location, or temporal order, older adults show an even worse performance when they are required to bind these different kinds of information (Chalfonte & Johnson, 1996; Kessels, Hobbel, & Postma, 2007; Mitchell, Johnson, Raye, & D’Esposito, 2000). For example, Kessels et al. (2007) found that, although older adults performed worse than young adults on all tasks, the age difference became larger when participants had to recall two pieces of information (object-order, location-order, object-location) instead of one piece of information (only object, only location), suggesting that older adults have difficulties binding information.
As people grow older, there is a decrease in the prefrontal cortex volume and changes in gray matter volume and density (Martinelli et al., 2013; Raz & Rodrigue, 2006) and a decrease in the volume of the hippocampus (Raz et al., 2005). The hippocampus is not only known to be responsible for recollection (e.g., Cabeza et al., 2004), it has also been associated to the binding of information (DeMaster & Ghetti, 2013). At two time points, Raz et al. (2005) measured the hippocampal volume of 72 participants, who were between 20 and 77 years at the first measurement. The second measurement was 5 or 6 years after the first measurement. At both time points, there was a negative correlation between volume and age. In addition, most—but not all—participants showed a decrease between the two measurements.
The inability to bind information leads to deficiencies in source monitoring, which involves the recall of the circumstances under which a memory was acquired (Johnson, Hashtroudi, & Lindsay, 1993). These circumstances include the spatial, temporal, and social context of the event and the media and modalities through which the event was perceived. Source monitoring is important, because it helps to differentiate between experienced and imagined events, remembering and knowing, and episodic and semantic memory. Research has shown that older adults experience more difficulties in determining the correct source of recently learned information than young adults (Gras, Tardieu, Piolino, & Nicolas, 2011; Hashtroudi, Johnson, & Chrosniak, 1990; Henkel, Johnson, & De Leonardis, 1998; Karpel, Hoyer, & Toglia, 2001; Koutstaal & Schacter, 1997; McDonough & Gallo, 2013; Norman & Schacter, 1997). Spencer and Raz (1995) conducted a meta-analysis of 46 studies. They compared the memory for content (e.g., words, sentences, drawings, pictures, actions) to the memory for context (e.g., font size or type, presentation modality, location on screen, imagined or real, self-generated or generated by experimenter or other participant, time of day). Memory for content showed minor age-related declines, whereas memory for context showed major age-related declines.
The findings regarding hippocampal volume, binding, and source monitoring suggest that the recall of semantic information about personal events should remain relatively unaffected in older age but that the recall of episodic information would be affected by aging. Piolino, Desgranges, Benali, and Eustache (2002), for example, asked middle-aged and older adults to recall episodic details of specific events that had occurred in the most recent year and in past lifetime periods. Participants also recalled semantic personal information, such as names and addresses, from these lifetime periods. As expected, episodic recall declined more over time and more with age than semantic recall. Similar results were reported by Levine, Svoboda, Hay, Winocur, and Moscovitch (2002), who found that young adults produced more episodic details for autobiographical memories than older adults (i.e., internal details), whereas semantic details were produced in equal quantities among young and older adults (i.e., external details). Furthermore, St. Jacques and Levine (2007) compared the number of internal and external details in the descriptions of emotional and neutral memories of young and older adults. Young adults recalled more internal details, whereas older adults recalled more external details; and these age effects were found for both emotional and neutral events.
Autobiographical memory is related to cognitive abilities
The findings discussed in the previous section suggest that—similar to cognitive abilities—autobiographical memory functions optimally in adolescence and early adulthood, the second premise of the cognitive abilities account. There is also considerable support for the third premise that the storage, consolidation, and retrieval of specific personal events are related to cognitive abilities.
Several studies have found strong relationships between fluid abilities, such as processing speed and working memory, and episodic memory (e.g., Clarys, Bugaiska, Tapia, & Baudouin, 2009; Hertzog, Dixon, Hultsch, & MacDonald, 2003; Park et al., 1996) and event segmentation (Kurby & Zacks, 2008), but the participants in those studies were required to recall or recognize word lists or short stories without any or with only a short delay (e.g., five minutes). However, none of these studies actually measured autobiographical memory, which involves personally relevant information and longer retention intervals (e.g., days, weeks, months, or even years and decades). One way to measure autobiographical memory performance is comparing the recall of a personal event directly after its occurrence to the recall of the event after a certain interval. This technique is frequently employed with diaries (e.g., Catal & Fitzgerald, 2004; Linton, 1975; Rubin, 1982, Exp. 5; Wagenaar, 1986). Whereas most diary studies tend to use a small number of participants who record and recall a large number of events, one can also have many participants who each record and recall a single event (i.e., Kristo, Janssen, & Murre, 2009).
Like the study of Murre et al. (2013), the study of Kristo et al. (2009) was conducted through the Internet. On the same website, participants recorded one personal event that happened in the last three days; answered a series of questions about the event (people, activity, location, etc.); and rated the event on importance, emotional intensity, emotional valence, and frequency of occurrence. Participants were subsequently contacted after 2, 7, 15, 30–31, or 45–46 days and, with the help of cues, answered the same questions. They also rated the event on social sharing and reminiscing. Autobiographical memory performance became worse as the interval between the recording and the recall of the event increased. Performance was also affected by other event variables, such as valence, frequency of occurrence, and reminiscing. These findings were replicated in a follow-up study (Murre, Kristo, & Janssen, 2014) that showed that autobiographical memory was also affected by age, with young adults performing better than middle-aged and older adults do.
To assess whether participants with stronger cognitive abilities also perform better on autobiographical memory tests, the results of participants who took at least one test of Murre et al. (2013) and who completed the diary study of Kristo et al. (2009) were combined in Janssen, Kristo, Rouw, and Murre (2015). This study showed that the performance on the verbal memory tests was directly related to the retention of the personal event. However, because autobiographical memory is a complex process to which many factors contribute (e.g., Rubin, 2005, 2006), the size of the effect was small. Because the diary study did not contain questions specifically measuring the visuospatial aspects of the personal event, no effect was found for the performance on the visuospatial memory tests.
The results of Janssen et al. (2015) suggest that the cognitive abilities account might be a viable alternative to the existing accounts for explaining the reminiscence bump in the distribution of word-cued memories. Middle-aged and older adults tend to recall more specific personal events from adolescence and early adulthood, not because the events from these lifetime periods are especially important, emotional, or positive (as predicted by other accounts) but because the memory system was operating more efficiently and stored more events in these lifetime periods.
The basic-systems approach
The cognitive abilities account builds upon the basic-systems approach (Rubin, 2005, 2006), which argues that a series of lower level cognitive processes, or “basic systems,” underlie the retrieval of personal events. Although these lower level cognitive processes interact to form autobiographical memories, each system has its own functions, processes, and neural substrates (St. Jacques, 2012): control processes (lateral pre-frontal cortex), top-down and bottom-up attention (dorsal and ventral parietal cortex), self-referential processes (medial pre-frontal cortex), recollection (hippocampus and retrosplenial cortex), emotion (amygdala), and visual imagery (occipital, cuneus, and precuneus regions).
Support for the basic-systems approach comes from two types of research. Whereas other brain regions are involved in autobiographical memory, support for the basic-systems approach has focused on four structures and their roles in autobiographical memory retrieval: the hippocampus (e.g., Cabeza et al., 2004; Milner, Corkin, & Teuber, 1968), amygdala (e.g., Adolphs, Cahill, Schul, & Babinsky, 1997; Botzung, Rubin, Miles, Cabeza, & Labar, 2010; Brierley, Medford, Shaw, & David, 2004), visual cortex (e.g., Daselaar et al., 2008; Greenberg & Rubin, 2003; Rubin & Greenberg, 1998), and medial prefrontal cortex (e.g., D’Argembeau et al., 2007; Northoff et al., 2006; St. Jacques, Conway, Lowder, & Cabeza, 2011).
Findings from functional imaging have suggested that tasks that require certain cognitive processes showed more activity in corresponding brain regions than tasks that do not require these cognitive processes, whereas findings from neuropsychology have suggested that deficiencies in one of the basic systems lead to corresponding deficiencies in the autobiographical memories, thus supporting the basic-systems approach. However, these findings also lend support to the third premise of the cognitive abilities account that lower level processes (i.e., cognitive abilities) are linked to the retrieval of specific personal events.
Part 3
Support and criticism
In the previous sections, we have seen that the cognitive abilities account is built on the idea that the encoding, storage, and retrieval of specific personal events is a biological process, in which lower level processes, such as control processes, top-down and bottom-up attention, self-referential processes, recollection, emotion, and visual imagery, interact to form autobiographical memories (Rubin, 2005, 2006). Several of these lower level processes display age-related changes across the life span, and people tend to perform optimally during adolescence and early adulthood on tests measuring them (e.g., Hartsthorne & Germine, 2015, Exp. 1; Maylor & Logie, 2010; Murre et al., 2013; Reimers & Maylor, 2005). Because cognitive abilities are related to performance on tests measuring autobiographical memory abilities (e.g., Janssen et al., 2015), it is argued that people also perform optimally during adolescence and early adulthood on these tests (e.g., Levine et al., 2002; Piolino et al., 2002), and, more importantly, more specific personal events are stored in adolescence and early adulthood than in other lifetime periods, thus causing the reminiscence bump in the temporal distribution of autobiographical memory.
It is important to emphasize that the findings of Janssen et al. (2015) and other studies that have reported that there are age-related changes in autobiographical memory do not suggest that event characteristics, such as emotional intensity, emotional valence, importance, or frequency of occurrence, are not good predictors for memory retrieval. On the contrary, these variables were just as predictive for the retention of the personal event as the performance on the verbal memory tests, but these variables do not seem able to explain changes across the life span. There were no significant correlations between participant age and event characteristics. Older adults did not rate their recent events as more or less positive or important than younger adults did. However, the findings do seem to indicate that the relationship between cognitive abilities and autobiographical memory exists above and beyond those event variables.
Similarly, the findings of Janssen et al. (2015) do not contradict the findings that external factors, such as wars and natural disasters (e.g., Benson et al., 1992; Brown et al., 2009; Conway & Haque, 1999; Zebian & Brown, 2014) or spinal cord injuries and residential moves (e.g., Enz, Pillemer, & Johnson, 2016; Schrauf & Rubin, 2001; Uzer & Brown, 2015), might cause a higher prevalence of personal events as well. Those events can have such impact on daily life that they supersede age-related changes in encoding efficiency.
Finally, the account assumes that age-related changes in cognitive abilities are mainly caused by developmental factors (Raz & Rodrigue, 2006). However, the third premise of the account (i.e., the retention of specific personal events is related to cognitive abilities) remains unaffected when age-related changes in cognitive abilities would be (partly or completely) caused by cultural or environmental factors (Baltes, Staudinger, & Lindenberger, 1999).
The cognitive abilities account has an advantage over the other accounts by being more parsimonious. Explanations for changes across the life span in cognitive abilities, whether they are developmental, cultural, environmental, or a combination of these three factors (Baltes et al., 1999), can also be used to account for changes across the life span in autobiographical memory functioning. No new mechanisms or concepts need to be introduced to explain why more memories are stored in adolescence and early adulthood than in other lifetime periods. The account or accounts that one would use to explain why middle-aged and older adults remember fewer words than adolescents and young adults on a free-recall task with a 20-second delay can also be used to explain why people recall fewer events from the period in which they were middle-aged or older adults than from the period in which they were adolescents or young adults.
In addition, the cognitive abilities account corresponds well with findings from other psychology fields. For example, clinical research involving overgeneral memory supports the premise that cognitive abilities are linked to autobiographical memory too (e.g., Williams et al., 2007). Overgeneral memory is the inability, despite explicit instructions, to report specific personal events. Overgeneral memory has often been observed in patients diagnosed with depression (Van Vreeswijk & De Wilde, 2004). When cued with the word “happy,” instead of saying “I had a good time at Jane’s party,” they might say “I always enjoy a good party.” Several studies have associated overgeneral memory with impaired executive functioning (Dalgleish et al., 2007; Sumner, 2012), suggesting that the retrieval of specific personal events is indeed related to cognitive abilities.
One point of criticism that has been raised against the cognitive abilities account is that the decrease in the proportion of word-cued memories is stronger than the decline of cognitive abilities (Rubin et al., 1998). However, this statement has not been systematically tested. When examining the literature about cognitive abilities more closely, there does not seem to be a uniform decline rate for cognitive abilities; fluid abilities tend to show a stronger and earlier peak with a relative strong decline, whereas crystallized abilities often show a weaker and later peak with a relative weak decline (Craik & Bialystok, 2006; Li et al., 2004; Salthouse, 2004). Similarly, autobiographical memory studies have reported different rates of decrease in the proportion of word-cued memories per year. The decrease is, for example, stronger when participants were instructed to avoid recent memories (Jansari & Parkin, 1996) or instructed to focus on early memories (Rubin & Schulkind, 1997a). Furthermore, the statement excludes the possibilities that small declines in cognitive abilities may cause large decreases in encoding efficiency (due to high sensitivity) or that large declines in cognitive abilities may only cause small decreases in encoding efficiency (due to compensatory mechanisms). Two related abilities do not necessarily have to have the same shape. An example of such non-linear relation is word acquisition and vocabulary size in language processing (e.g., Craik & Bialystok, 2006). Although the shapes of these two functions are different, the size of a vocabulary is considered the result of the acquisition of words.
Additional support
There is additional support for the cognitive abilities account from other findings in the reminiscence bump literature. The account predicts that the distributions of word-cued memories will be similar across cultures, because age-related changes in cognitive abilities are assumed to be mainly caused by developmental factors (Raz & Rodrigue, 2006). Two studies have reported substantial cross-cultural differences (Benson et al., 1992; Conway, Wang, Hanyu, & Haque, 2005), but these studies did not use cue words to elicit the memories. Participants in Benson et al. were asked for their 10 most vivid memories, whereas participants in Conway et al. were asked to report the first 20 specific memories that came to mind. A comparison of the lifetime distributions of word-cued memories of American, Dutch, Japanese, and Polish samples that were given the same instructions and the same cue words suggests that there are no substantial cross-cultural differences (Janssen et al., 2005; Janssen et al., 2011; Kawasaki et al., 2011).
The cognitive abilities account is also supported by findings from research outside the autobiographical memory domain that examined the memory for public events (e.g., Janssen et al., 2008; Rubin et al., 1998; Schuman et al., 1997). These studies required participants to answer questions about public events. They do not include studies in which participants had to indicate which public events had changed the world or were especially important (e.g., Holmes & Conway, 1999, Study 1; Schuman & Corning, 2012). For example, Janssen et al. (2008) presented 1334 participants (between 16 and 75 years old) 15 open-ended and 15 multiple-choice questions that were semirandomly selected from a database with 239 questions about public events that had happened in the previous 60 years. Participants performed better on questions about public events from adolescence and early adulthood (10–25 years) than on questions about public events from other lifetime periods. This effect was stronger for open-ended questions than for multiple-choice questions. Research that used cue words to elicit public events has yielded inconsistent results. Howes and Katz (1992) found no effects of age but only an increased recall of recent events. However, they used 15-year age bins, which may have masked the reminiscence bump (e.g., Janssen, Rubin, & St. Jacques, 2011). Koppel and Berntsen (2016), on the other hand, used five-year age bins and found, besides recency effects, reminiscence bumps in the distribution of word-cued memories about public events in the period in which participants had been between 5 and 19 years old.
This reminiscence bump in the memory of public events cannot be explained by event characteristics, such as importance or emotional valence. A public event might have happened when one group of participants was 15 years old and another group was 55 years old. Although both groups have experienced the same event, the recall of the event is not similar across the groups. Accounts, like the cultural life script account, have difficulties explaining the reminiscence bump in the memory for public events (Janssen, 2015b; Tekcan, Boduroglu, Mutlutürk, & Aktan Erciyes, 2017), but the cognitive abilities account can explain these findings. The account expects a small difference, because it assumes that the memory system was working more efficiently for the group of participants who were 15 years old when the event happened than for the group of participants who were 55 years old.
Retrieval processes
However, similar to the observation that the other accounts have difficulties explaining the results of studies in which participants retrieve specific personal events with cue words (e.g., Conway & Haque, 1999; Janssen & Murre, 2008; Maki & Naka, 2006; Rubin & Schulkind, 1997a), the cognitive abilities account cannot explain the results of studies in which participants only report highly significant events. No account seems able to explain all the manifestations of the reminiscence bump, possibly because the search demands in each of the two types of methods are different (Maki et al., 2013).
The qualities of the memory, such as importance or vividness, are essential for requests for highly important events or especially vivid memories, but the content of the memory is crucial for the retrieval of memories related to certain words. Requests for highly important events therefore allow participants to restrict their memory search to only significant events, whereas the use of cue words requires participants to put fewer or no restrictions on their memory search. For this reason, cultural life scripts may help with memory searches for important events, but they appear less effective when searching for memories related to certain words (e.g., Berntsen & Bohn, 2010; Ece & Gülgöz, 2014, 2017).
Maki et al. (2013) gave participants neutral (e.g., book, city, flower), emotional (e.g., fear, happiness, love), and emotion-provoking words (e.g., friend, health, mother). The memories elicited by emotional and emotion-provoking words were not only more emotional than the memories elicited by neutral words, but their distributions also tended to have a later peak, suggesting that different processes underlie the retrieval of memories cued with neutral and the retrieval of memories cued with emotional and emotion-provoking words. However, Maki et al. had not asked participants to describe their memories and they could therefore not link the content of the memories elicited by emotional and emotion-provoking word to event categories from the cultural life script.
Besides cue words, autobiographical memories can be elicited by other memories (e.g., Brown & Schopflocher, 1998; Fitzgerald, 1980; Sato, 2002; Wright & Nunn, 2000) or by pictures, sounds, or odors (e.g., Chu & Downes, 2000, 2002; de Bruijn & Bender, 2018; Goddard, Pring, & Felmingham, 2005; Herz & Schooler, 2002; Miles & Berntsen, 2011; Rubin, Groth, & Goldsmith, 1984; Willander & Larsson, 2006, 2007). Memories cued by odors, although they tend to be less frequently thought of, are generally more emotionally intense, more positive, and accompanied with stronger feelings of recollective experience than memories cued by words or pictures (Hackländer, Janssen, & Bermeitinger, 2019). Furthermore, the distribution of memories evoked by odors has an earlier reminiscence bump than the distribution of memories evoked by words or pictures; odors tend to elicit memories from the first decade (0–10 years), whereas cue words and pictures tend to elicit memories from the second decade (10–20 years).
Many explanations have been suggested to account for the differences between odor-evoked and other memories (Hackländer et al., 2019). Whereas some of these accounts focused on encoding or storage processes, most focused on retrieval processes. According to Conway and Pleydell-Pearce (2000), memories can be retrieved in two ways. Generative (or indirect) retrieval is a slow and always voluntary process, in which people first elaborate on the cue and then conduct a hierarchical search. Direct retrieval, on the other hand, is a fast and often involuntary process, in which the memory instantly comes to mind and the hierarchical search is bypassed. Larsson, Willander, Karlsson, and Arshamian (2014) argued that, whereas the retrieval of memories evoked by words or pictures tends to be generative, the retrieval of memories evoked by odors is often direct. Because directly retrieved memories are often involuntary, they tend to be specific (Berntsen, 1996, 1998), which causes their retrieval to be accompanied with more physical reactions and emotional impact (Berntsen & Hall, 2004).
The cognitive abilities account emphasizes the role of encoding processes, but research examining different types of cues highlight the role of retrieval processes (e.g., Hackländer et al., 2019; Maki et al., 2013). Because adults may have a few thousand of autobiographical memories (e.g., Rubin & Schulkind, 1997b; Wagenaar, 1986), the ability to restrict the search would make the retrieval of events from memory easier when people recall memories voluntarily. One strategy that participants can employ to restrict the memory search is using cultural life scripts. They can use this semantic knowledge when they are instructed to recall the most important events of their lives or when they are instructed to recall personal experiences during which they felt certain emotions (e.g., Berntsen & Rubin, 2002; Rubin & Berntsen, 2003). The use of words and other types of cues in the laboratory is intended to mimic the direct retrieval of involuntary memories in daily life, but memories retrieved in the laboratory with the help of cue words are by definition voluntary and therefore often retrieved generatively. Therefore, besides relying on direct retrieval when asked to recall personal events with the help of cue words, participants can also use cultural life scripts—but only when they are able to associate the cue words to transitional events. Alternative retrieval strategies that participants may employ are identifying recent or early memories that are associated with the cue words.
Different methods for gathering autobiographical memories have led to memories with different temporal distributions (Hackländer et al., 2019; Koppel & Berntsen, 2015; Maki et al., 2013; Rubin & Schulkind, 1997a). Whereas the other explanations for the reminiscence bump appear to be able to explain the findings of studies in which participants report only highly significant events, the cognitive abilities account aims to explain the findings of studies in which participants retrieve specific personal events with the help of (neutral) cue words. The cognitive abilities account should therefore be considered to be complementary to the other accounts.
Future studies and conclusions
Although the support for the cognitive abilities account is promising, additional research needs to be conducted. To extend the findings of Berntsen and Bohn (2010), Ece and Gülgöz (2014, 2017), and Maki et al. (2013), a study should be conducted in which participants would be given cue words that can and cannot be connected to transitional events from the cultural life script. It would be expected that the distribution of memories elicited by words that can be linked to event categories from the cultural life script would have a later reminiscence bump than the distribution of memories elicited by words that cannot be linked to event categories from the cultural life script. For example, the word “army” can be linked to military service or conscription, the word “engine” to obtaining driver’s license, and the word “kiss” to falling in love and first kiss. The memories cued with these words would therefore have a reminiscence bump in the third decade, whereas memories cued with other words would have a reminiscence bump in the second decade. In addition, memories cued with cue words that cannot be connected to event categories from the cultural life script would have a larger increased recall of recent events, because participants might focus on recent memories.
Furthermore, because most of the support for the cognitive abilities account has come from cross-sectional studies, the account would ideally be tested with a longitudinal study in which participants’ cognitive abilities would be tested twice. The first test would be when participants would be around 20 years old and cognitive abilities tend to peak, whereas the second test would be 30 years later when they would be around 50 years old and cognitive abilities are declining. At the second test, participants would also retrieve autobiographical memories with the help of cue words. Participants who showed the strongest cognitive decline over the 30-year period are predicted to recall relatively more personal events from adolescence and early adulthood (i.e., a strong reminiscence bump), whereas participants who showed the weakest cognitive decline are predicted to recall relatively more personal events from the last 5 to 10 years (i.e., a strong increased recall of recent events).
Many studies have demonstrated age-related changes across the life span in cognitive abilities. According to Rubin’s basic-systems approach, many of these cognitive abilities support autobiographical memory. Although each ability has its own functions, processes, and neural substrates, they interact to form autobiographical memories. In the present review, preliminary support was discussed for the cognitive abilities account, which argues that the reminiscence bump in the temporal distribution of autobiographical memory is caused by age-related changes across the life span in the ability to bind spatial and temporal information (encoding efficiency). Because memory is operating at an optimum during adolescence and early adulthood, memories from those lifetime periods are stored more strongly and therefore have a higher likelihood to be retrieved at later ages. Although additional research that examines the cognitive abilities account needs to be conducted, the preliminary support discussed in the present review is promising, warranting a more serious consideration for the account than it has so far been given in the literature. However, because the account does not aim to explain the reminiscence bump in the distribution of highly significant events, it should be seen as complementary to the existing accounts.
