The remembering–imagining system

The problem of the future

In classical physics/mechanics, a system, a collection of objects, for example, particles, fields and waves, is deterministic and reversible (Susskind and Hrabovsky, 2013). This means that given the laws that govern the system and its changes are known, a future state can be predicted exactly. Similarly knowing the state of a system at any given time means that the state of the system at an earlier or later time can be precisely established. Supposing we knew the laws that governed the cognitive, mind/brain, and system, could we then, for any given individual, predict the exact state of the system at a future point? We suggest that we could not, (see Dudai and Edelson, 2016, for related arguments). Just as in certain areas of physics, for example, quantum mechanics, it is not possible to precisely predict a future state of a system, so with people the future is only probable. However, once a future state has come into being, it may be possible, to at least some extent, to work back to previous states. Thus, the cognitive system may be retrospectively reversible (a point interestingly made by Freud, 1955 [1920], in the case of psychological states, which given the initial conditions cannot be predicted but working back to initial conditions is at least partly possible). Nevertheless, retrospectively reversible or not, given that there are an infinite number of indeterminate possible futures, this poses a major adaptive problem for goal-driven organisms. This is particularly so as the end point of all unrealized goals lies somewhere in the future. Indeed, in order to have a goal a future state has to be anticipated and often consciously imagined (Cole and Berntsen, 2016).

For many species, the future is adapted to possibly by a general preparedness to meet environmental contingencies. This is perhaps one of the simplest ways to deal with the problem of the future (amusingly reflected, perhaps, in the motto of the Boy and Girl Scouts “Be prepared!”; Baden-Powell, 1908). No doubt all species have this general preparedness; however, in our own species, more complicated forms of cognition have arisen to anticipate possible futures and also, importantly, to constrain what the future can possibly be. Society and culture generally may be one way of dealing with a potentially limitless set of futures, and that is because they constrain possibilities (Merck et al., 2016; Wang, 2016). Just as memory, especially autobiographical memory constrains the self—it limits the universe of possible future selves (Conway, 2005). Thus, memory, society, and culture constrain the range of possible futures by providing the context in which the future will most probably occur (Welker, 2016). In the case of individuals, we have argued that there is what we conceive of as an extended form of consciousness that consists of memories of the recent past and images and expectations of the near future, and it is this form of extended consciousness that we have termed the RIS.

The RIS

We visualize the RIS as a bell curve (of consciousness) moving through time with the peak of the curve as “now” or the present moment, cf. Stern (2004). To the left of the peak is a declining curve that represents decreasing memory access for the recent past as the retention interval increases and to the right of “now” another decreasing curve that represents the specificity of imagined future events (see Figure 1). The RIS then posits that we have an awareness of the recent past and near future that declines into lowered memory accessibility as the retention interval from “now” increases. Conversely, future imagined events—episodic simulations—decrease in episodic specificity declining into the schematic/generic/stereotypical, culturally specified (Berntsen and Bohn, 2010; Stroumsa, 2016) future memories, as the distance between “now” and the future increases. The boundaries of the RIS are then constantly changing, and this is reflected in our constantly changing awareness of the near past and the near future.

OPEN IN VIEWER

Figure 1.

An idealized RIS curve for past, present, and future events.

To what extent does the idealized representation of the RIS shown in Figure 1 accurately reflect our memory for the recent past and imaginations of the near future? In a recent study, we (Loveday and Conway, 2016) had people list all the personal events they could remember for each of the past 5 days, and all the personal events that they imagined could plausibly occur on each of the next 5 days. The numbers of remembered and imagined future events for each day are shown in Figure 2. Note that we specifically avoided recall from weekends as we suspected that these might feature more distinctive events and so somewhat artificially inflate the endpoints of the 5-day periods we were sampling. Nonetheless, in the pilot work recently conducted, we have found a similar curve, somewhat more pronounced, for events remembered and imagined 1 month into the past and 1 month into the future.

OPEN IN VIEWER

Figure 2.

Recalled recent and imagined near-future events.

It can be seen in Figure 2 that the number of recent memories that can be accessed decreases rapidly over the first 3 days but then seems to stabilize and even increase slightly at a retention interval of 5 days. Also, interestingly, the number of memories recalled on the first day averaged at about nine memories and varied between 6 and 12 events over participants, but most recalled 7–9 events. The specific instruction to participants was to “recall as many events as you can from yesterday and then from the day before, going back 5 days.” What an “event” was, was left to each participant to decide. The events listed typically were of discrete series of actions with beginnings and endings marked by changes in actions, locations, and goals: for example, “meeting a friend, X, for lunch” and then “attending a lecture on Y,” with each event description containing contextual episodic details. The number of contextualizing episodic details (essentially who, what, where, and when, plus some other more idiosyncratic details) listed for each memory decreased over the 5-day retention interval. Thus, the memories not only became fewer, but they also became less specific as the retention interval increased (see also Stroumsa, 2016).

For imagined future events, the decline was from a mean of 5 events for tomorrow to a mean of 4.3 events in 5 days’ time, a reliable but less steep decline. However, the specificity of the imagined future events also decreased as the time interval increased. For example, a typical imagined event for tomorrow might be “going to the gym at 5 with X & Y tomorrow evening for our zumba class,” whereas a typical imagined event in 5 days might be more “going to the pub,” “going to the cinema,” and so on. Thus, the further away in time the imagined events were, the more typical and routine they became with less imagined episodic detail. In other recent research Loveday and Conway (2016) collected imagined future events from 1 to 5 and 5 to 10 years hence, outside the range of the RIS, the imagined events were more typical of events from “life scripts” (Berntsen and Rubin, 2004) and were low in episodic content. It seems then that the episodic content of imagined future fades into culturally determined events, such as getting married, having children, pursuing a career and retirement, as episodic future thinking turns into culturally expected future events. We see this phenomenon of fading episodic detail in memories of the past and imagining future events more as a bias in the RIS rather than a sort of episodic boundary. There are always some events from the past that can be recalled with extensive episodic detail and, similarly, some events from the future that can be imagined with rich episodic detail. We believe not that this cannot be done, but rather that it typically is not done.

Another feature of the RIS is that it is present during periods of unconsciousness, for example, during periods of sleep. Indeed it may be during sleep that recent and future memories become in some way connected, perhaps by integration/consolidation with common and current goals (see Diekelmann and Born, 2010). In order to explore this, we had individuals record their waking thoughts immediately upon awakening with the expectations that many of these would be about upcoming events and the recent past. Note that the recording took place during the period of the awakening cortisol response (ACR), when activation rises in the medial temporal lobe memory system, and of course, frontal systems that have been to some extent dysfacilitated during sleep regain their normal levels of activation. This may reflect the RIS coming back on line too. Thoughts were also sampled mid-way through the day and again in the evening. Classification of awakening thoughts is shown in Figure 3.

OPEN IN VIEWER

Figure 3.

Thoughts at different times of the day.

The two major categories of awakening thoughts are orientation thoughts, for example, “What time is it?” and thoughts about the “close/near” future, for example, “what am I doing today?” There were relatively few thoughts about the recent past. These findings suggest, perhaps, that the RIS is re-engaging with future episodic imaginings after a period of disengagement during sleep. Possibly, the ACR in part facilitates a re-engagement of the medial temporal lobe memory system with executive control systems in frontal regions and, thereby, reconnects to representations of the immediate future. Thoughts at later points in the day were also goal-oriented, but to near activities and very recent memories often of task completion (see Figure 3; Loveday and Conway, 2016).

There is some evidence then for the idealized bell-shaped curve of the RIS shown in Figure 1 and the future component of the RIS seems highly active upon awakening. However, although changes in the accessibility of memories of recent past events and episodic memories of near-future events are reflected in decreasing numbers of events recalled with increasing retention intervals, there are, as noted above, other more qualitative changes in the episodicity of what is recalled. Thus, the episodicity of both memories of past and future events decreases with increasing retention intervals (Stroumsa, 2016; Welker, 2016). An interesting question that then arises is, “What is the range of the RIS?” “Where do its boundaries lie?” We suggest that the RIS is part of the goal system, cf. Conway and Pleydell-Pearce (2000) and maintains access to memories of recent events that themselves support plausible imaginings of near-future events. Raised access to recent memories keeps us informed of specific progress with current goals and with specific goal-related activities that are upcoming in the near future. The range of episodic memories with RIS-raised accessibility may be determined by how the content of those memories continues to inform goal activities, perhaps by accurately predicting future events. To take a mundane example, for a student an episodic memory of having completed a coursework assignment, say last night, might lead to an episodic simulation of handing in the completed coursework in the department the following day. This simulation when enacted will, in turn, become a recent episodic memory that maintains a highly accessible record of achievement of a specific goal (embedded in a much more complex goal hierarchy of getting a (good) degree). It is important to note that it is not the simulation that becomes a memory but rather its enactment that leads to a new episodic memory or set of episodic memories (Conway, 2009). Thus, the episodic simulation through enactment ¹ has now become a new episodic memory that may remain highly accessible in the RIS with accessibility gradually decreasing with time, perhaps over only a few days in this example. One implication of this view of the RIS is that episodic memory is critically important to episodic future thinking, because without it a person could not generate effective episodic simulations based on recent episodic memories of goal-related actions.

Dysfunctional RIS

Related to the notion that recent episodic memories allow the formation of (effective or adaptive) episodic simulations (see Schacter et al., 2012a for a detailed review; Schacter and Madore, 2016), we recently had the opportunity to study future simulations in a densely amnesic patient (Cole et al., 2016). Patient HCM, a 66 year-old highly educated married man with four adult children, suffered a series of cardiovascular incidents leading to hospitalization in a rehabilitation center with 24-hour care. Although separated from his wife, his children visited regularly. Earliest evidence of brain damage was a cardiovascular accident in 1998 (at 51 years of age). This resulted in a left lateralized hemorrhage which required a craniotomy to alleviate pressure. He also sustained two traumatic head injuries, one in 2000 (at 53 years) when he was assaulted with a blow to the head and one in 2002 (at 55 years) when he fell down a staircase. After being hospitalized, he was admitted to a specialist neurobehavioral unit for assessment of people with acquired brain injury and aiming to remediate ongoing cognitive, behavioral, and physical deficits. In summary, the conclusion of the assessment was that HCM had multiple diffusely spread infarcts indicative of global atrophy extending to both cortical and subcortical regions.

HCM undertook a wide-ranging neuropsychological test battery examining his cognitive abilities in a number of domains, including executive processing, memory, and episodic future thinking. Although HCM was enthusiastic about taking the test battery, he required 11 testing sessions to complete the tests. This was due to HCM repeatedly stopping testing when he became aware of his inability to complete tasks to pre-injury levels. This suggests that although he had online awareness of his reduced abilities, this did not augment his global self-knowledge of his current abilities (see Toglia and Kirk, 2000). HCM’s language was spared, but his working memory was impaired. His orientation showed that HCM had difficulty in estimating current temporal information and was unable to provide an answer concerning the current day, date or month. Also, he was unable to recall when he was admitted to the rehabilitation unit. His current year estimate predated the current year by a decade. Knowledge of current location was more accurate: HCM accurately described the country, city, and building (“hospital building”) indicating that he did not have global unawareness of his current situation. Finally, he was generally unable to remember people he had recently encountered. This anterograde amnesia (AA) was strikingly evident in tests of memory: in immediate recall he could recall, to some extent, parts of stories and words he had studied; however, following even short delays prior to recall, he recalled nothing and was at floor on all tests.

HCM’s profile was that of a patient with some spared intellectual capacity, reasonably good language skills, an impaired working memory, but spared short-term memory and a severely dense AA. He had in effect lost the ability to form new episodic memories. Examination of his autobiographical memory found some recall of events from early in his life, a few from the period prior to his first stroke, and good retention of factual knowledge of his life up to early adulthood. There were, however, few specific memories from later life and none at all from the past 5 years. He could not recall the name of a single one of his carers. He had then, in addition to his AA, a temporally-graded retrograde amnesia (RA). Consequently, HCM had a severely damaged RIS, and it seemed that his ability for episodic simulation of the future may have been compromised. It was decided, therefore, to investigate his ability for future episodic simulation. In order to do this, we created a short and simple Mental Time Travel Questionnaire that required HCM and non-brain-damaged matched controls to either recall specific autobiographical memories to a series of temporal cues, for example, recall an event from last weekend, or think of specific future events, for example, think of an event that might happen next weekend. The plausibility of responses was verified in HCM’s case with one of his sons and for the controls with family members.

HCM recalled some past events and only a few future events. His future episodic simulations were, however, highly implausible. For example, he stated that he would invite his family to his apartment for a family gathering. In fact, he owned a local apartment but was no longer able to access this due to mobility problems. Clinical staff confirmed that he had not resided at his apartment since admission and was not expected to live there in the future. Nevertheless, HCM voluntarily imagined a future scenario involving his apartment. In a similar vein, in 5–10 years, he envisioned “passing” a variety of clinical examinations and moving back into his apartment. Both of these events would be unlikely to occur, according to clinicians and his son, suggesting that his future thoughts were outdated and related not to memories of his current situation but rather to memories of the conditions that existed in his life before his first brain injury years previously. It is important to note that HCM did not confabulate those memories he was able to access and it was only in imagining future events that his memories became implausible. In general, HCM’s descriptions of his personal past and future were marked by a sparsity of episodic information, similar to the blankness of the past and future reported by patient KC (Tulving, 1985). We note here too the similarity of HCM to those anosognosic Alzheimer’s patients described as having a “petrified self” (Morgrabi et al., 2009). Plausibility is perhaps one of the major differences between images of a future that are effective and adaptive in implementing goals and those that are not. Having an intact and functioning RIS facilitates the connection of memories of the recent past to episodic simulation of the near future.

HCM is a patient not only stranded in the past but also in the future, without a record of the recent past; he cannot imagine future events that are plausible and linked to his current situation—linked to “now” (see Cole et al. (2014) for further discussion of memory plausibility, Stroumsa (2016) for an interesting and related cultural discussion, and Young (2016) for a related discussion of “absence”). We have noted previously the constraining effect of memories generally on the self, for example, Conway (2005), and how in some psychological illnesses, for example, in schizophrenic delusions, this constraining relation appears to break down. This raises the interesting possibility of having episodic future simulations that are not adaptive or effective but instead are maladaptive and dysfunctional. Consider, for example, a patient with social phobia who believed that if she went out of her house she would end up in a supermarket and somehow be dragged into one of the large freezers containing bags of frozen vegetables and pulled under and “drown” (Day et al., 2004). She had a powerful and intrusive episodic future simulation of this happening, and this image of the future event prevented her from leaving the house. Such images of future events are not uncommon in a range of psychological illnesses (MacLeod, 2016). For instance, suicide attempters often report what Holmes et al. (2007: Table 2) have termed “flash-forwards” (see also Crane et al., 2012: Table 2). The flash-forwards in these patients are to episodic simulations of their own deaths in which they “see” their dead body, for example, at the foot of the cliff, under the wheels of a train, and so on. Such intrusive future episodic images may be associated with further attempts at suicide. Episodic simulations of the future in psychological illnesses may support and even motivate dysfunctional behavior and beliefs (MacLeod, 2016).

Concluding comment

The RIS is a hypothesis about the raised accessibility and increased specificity of episodic simulations of the recent past and near future. Importantly, the RIS only exists now (cf. Stern, 2004, for an exegesis of the concept “now.”) It is an expansion of “now” to encompass records of recent and associated upcoming goal-related activities (see Johnson and Sherman, 1990, for a particularly interesting discussion of the integration of past, present, and future). Its boundaries are constantly changing as goals are completed, abandoned, and/or replaced over time. If the ability to form episodic memories is lost, as in AA, then the RIS breaks down and effective future simulations are no longer possible. It is notable, too, that when this occurs, the ability to operate adaptively and effectively in one’s life is also severely compromised. HCM, for instance, and as is the case with many AA patients required 24-hour care.

The idea that episodic future simulations should be effective and adaptive is important. After all, any future could be imagined, but only a limited number are plausible and probable. We suggest that the degree to which future episodic simulations are adaptive/effective is determined by the strength of their association to specific episodic memories of the recent past (Merck et al., 2016; Schacter and Madore, 2016). The degree to which they are not associated with recent episodic memories reflects a malfunctioning RIS, possibly due to brain damage or to the effects of a psychological illness. Nevertheless, for many individuals, there may be some highly specific future episodic simulations that are implausible because they have a low probability of occurrence, for example, winning the lottery, and these may be linked to recent episodic memories, for example, buying a ticket, indicating normally functioning RIS. Possibly, the number of plausible to implausible episodic simulations held by an individual is some sort of index of mental health or well-being more generally (MacLeod, 2016).