Brain regions responsible for recalling the past overlap with those responsible for imagining the future, which may explain why our mental travels often feel so seamlessly unbroken, flowing from past to future. Moreover, recalling our past involves just as much reconstruction as imagining our future. It seems like our memories of our past may have more to do with providing us with scenarios to construct events, imagine and plan for the future with, rather than them being a perfect archive of videotapes of our past records. This ability to replay past events and imagine future possibilities is known as mental time travel.
Psychologist Michael Corballis argues that the origin of mental time travel may date back in evolutionary history, rather than being uniquely human, based on the recent behavioral studies in apes, rats, scrub jays, and even cuttlefish. In an article titled “Mental Time Travel: Animals Anticipate the Future”, the author describes Scrub jays who cache (storing away in hiding or for future use) their worms in the presence of another scrub jay, recached more worms to newer sites than jays who cached their worms in private. Moreover, recaching happens more frequently in birds that had previously stolen food cached by other scrub jays, suggesting they foresee other birds may act the same way and steal their cache. In another experiment, scrub jays were given pine nuts in compartment A on some mornings and nothing in compartment C on other mornings. During evenings, pine nuts were given in compartment B. It was observed that the jays would store most of these seeds from compartment B to compartment C (which is normally empty), thus insuring the nuts for the next morning so that they have nuts in 2 compartments, both A and C. Secondly, when given different types of food (pine nuts and dog kibble) but only one type is given each morning, jays cached the food not available the next morning in each compartment the night before, resulting in having both types of food the next morning. In another experiment with squirrel monkeys, when given a choice between 1 and 4 dates (eating more makes them thirsty) with a caveat that their water bottles would be taken away just before they made their choice. If the choice is 1 date, water is returned 30 mins later, while on the other hand, if 4 is chosen, water is returned 3 hours later. The choice of initially 4 diminishes to 1 in 80 percent of cases, suggesting they made a choice considering the future.
In rats, Hippocampal place cells record the animal’s location in space by shifting the population of active cells, creating a cognitive map. These cells are also observed to fire later when the animal is resting, grooming, and even when it’s sleeping. These reactivations of place cells, recorded as sharp wave ripples, represent past trajectories taken by the animal in the maze. Such reactivations are termed as replays and they are usually in reverse order and have a faster time scale. There are also forward replays or preplays which are observed before the animal sets out on its path. These preplays, arranged in the correct order, do not correspond to past path trajectories but rather represent a path that the animal may take in the future. Reverse Replays may play a role in memory consolidation by strengthening past memories in cortex, while Forward replays or preplays may play a role in planning future trajectories.
The article “The Role of Hippocampal Replay in Memory and Planning” explains the difficulty in ascertaining the functionality of these replays is that they sometimes correspond to trajectories that represent a part of the maze not in use or haven’t been visited recently. Also, the replays which are seen prior to the animal setting out, have shown to represent paths that the animal is supposed to avoid, thus making the trajectories anti-correlated to the behavior. It seems that the ongoing task demands shape the nature and content of these replays.
It seems like non demanding tasks like sleeping or sitting may correspond to the replays playing a role in memory consolidation while during a demanding task like the rat being at a junction making a decision towards a goal, the replays have been found to be more varied. They may correspond to predicting future paths, but also retrieving past actions if relevant, so their role can vary between task relevant and consolidative. Replays, thus may be a way to mediate retrieval of information needed to inform behavior, rather than planning specific trajectories.
There are also other cells that were discovered later: Grid cells in the medial entorhinal cortex code location by firing at regular intervals, thus forming a hexagonal grid to help the animal understand its location in space by integrating information about location, distance, and direction; Head direction cells code where the head is facing in a certain direction, Border cells are active when the animal is close to walls in a closed location, while Hippocampal time cells code relative time in which the event has occurred. These cells together create an enormous number of combinations reflecting the possible spatial context the animal is in.
In 2 Articles titled “Wandering tales: evolutionary origins of mental time travel and language” and “Language, Memory, and Mental Time Travel: An Evolutionary Perspective”, Michael further suggests that the capacity for mental time travel could have been the first step towards language. Non-human animals (including primates) seem to have limited voluntary control over their voice when compared to humans, which may suggest that the origin of communicative language may lie in visual signals rather than vocal ones. Visual signals like pantomime are a way of sharing mental time travels, as seen in apes (chimpanzee trying to show her daughter how to use a stone to crack a nut properly or orangutans requesting to have their stomachs scratched). For our ancestors, pantomime (involves the whole body) could have been further conventionalized into gestures involving just hands and arms, which were freed due to the emergence of bipedalism. Further conventionalization could have resulted in the creation of symbols and eventually towards facial expressions and voice control for speech. Conventionalization may have the cost of transparency but leads to greater efficiency. Intentional communication may have had its origin in manual hand gestures, with facial gestures being an intermediary, leading to voiced speech where speech itself can be considered as gestures comprising of larynx, velum, and tongue. Thus, Intentional communication could have evolved from manual gestures to explicit facial gestures to eventually mostly hidden gestures, all three of which are present for communication today. This is further illustrated by the fact that speech has a visual component to it for interpretation, as shown by the mcgurk effect (see here). Voluntary control of speech likely involved neural changes, including a direct connection between motor cortex and nucleus ambiguous (midbrain vocalization center), which seems to be unique to humans. Miniaturization and compartmentalization of speech output by tucking it into the mouth seem highly efficient as it adds little cost to breathing, which we have to do anyways to survive. Speech allows for communication at night or when there is no visual contact between the two parties. It also freed up hands for other tasks and may have enhanced our storytelling which was later also further enhanced by the emergence of writing and more recently by the internet.
Language can be considered as a device by which we share our mental time travels, an externalization of imagination with different languages having different ways to represent various common representations (objects, actions, events) by conventionalizing for efficiency which in turn increases exclusiveness as a result of abstraction. He goes on to suggest that given that words become part of our memory, it is possible that language may have expanded our capacity for mental time travels and storytelling, which may have affected our brain storage capacity. Thus, the relationship between language and memory may be bidirectional. The experience of past and future may go way back in evolution to animals that move – for them to know where they are in space and where they want to go next; to distinguish where they have already been and what happened. The sharing of that information came way later.
In a paper titled “Mind-wandering as spontaneous thought: a dynamic framework”, the authors suggest the lack of strong constraints on thought as a defining feature of mind wandering. Deliberate constraints are understood through cognitive control (process by which goals or plans influence behavior by supporting flexible, adaptive responses and complex goal-directed thought or a goal directed guidance imposed on thought, actioned by top-down executive processes) while automatic constraints operate outside cognitive control and are controlled by a group of mechanisms including sensory salience (attention sustained by high perceptual contrast) or affective salience (attention sustained by emotional significance of perceptions, thoughts). Deliberate constraints are very minimal during dreaming and increase as we go towards mind wandering and further towards creative thinking and they are the strongest during goal directed thought. While for automatic constraints, they are low to medium during Spontaneous thought (dreaming, mind wandering, and creative thinking) but are the strongest during rumination or obsessive thought, where thoughts stop being spontaneous.
Default network consists of DN Core (a hub for internally oriented cognition), closely linked to DNMTL (centered around medial temporal lobe) and DNSUB3 (role not clear yet). DNCore and DNSUB3 are recruited more during task unrelated and internally oriented thoughts while DNMTL seems to be active when deliberate constraints on thoughts are weak, i.e., DNMTL is active when the individual is not aware that they are experiencing task unrelated thoughts, establishing a link between DNMTL (and its central component, medial temporal lobe) and spontaneous thought. Evidence suggests spontaneous mental stimulations during rest in humans and rats were initiated by the MTL and supported by hippocampal cortical interactions. A study found that, after watching a film, the spontaneous recall of film clips was preceded by firings in the same medial temporal neurons that were activated when first watching the film. In rodents, periods of rest invoke hippocampal place cells to replay previously taken routes and preplay future ones yet to be taken, as discussed in a previous section.
In contrast, DAN (Dorsal Attention Network) is recruited for stabilizing our attention on the external world by constraining it. This can occur via deliberate cognitive control, which is linked to the Frontoparietal control network (FPCN). FPCN can link to DN and support internally focused deliberate planning or with DAN to support external focused visual and spatial planning. FPCN, thus puts deliberate constraints on thought. FPCN and Cingulo opercular control network (COCN) implement cognitive control at different timescales. FPCN has transient control with short term adjustments and task changes. COCN, on the other hand, shows temporally sustained control for maintaining a task over time.
VAN (Ventral Attention Network) and a General Salience Network directs attention automatically towards salient (objects or things that stand out like a loud noise, a new message notification, etc) stimuli, acting as a salience detection and filtering system. Each second we are bombarded with so many thoughts, sounds, smells, etc, the salience network selects which stimuli deserve our attention. Automatic constraints on thought can be exerted by DNCore, Salience Networks (including VAN), and DAN. By contrast, FPCN exerts deliberate constraints on thought by coupling with DNCore, DAN and Salience networks to reduce their influence.
Thoughts can shift from being spontaneous and therefore with relatively weak constraints to being highly constrained automatically to then highly constrained deliberately. These fluctuations in the type of constraints seem to be associated with changing interactions between brain networks. Deliberate constraints are well characterized and linked to executive functions and control networks, automatic constraints on the other hand are more diverse and probably served by diverse neural correlates.
Within the dynamic framework, dreaming is a type of spontaneous thought characterized by being highly unconstrained, hyper associative and therefore predicted to be linked with very low or absent deliberate constraints (exception lucid dreaming), strong influence from an internal source of variability and low to medium influence of automatic constraints. Dreaming should be accompanied by strong recruitment of DNMTL, weak to medium recruitment of DNCore, and strong deactivation of the FPCN. A recent metanalysis of REM Sleep, which is associated with the highest rate of dreaming, shows similar findings.
Creativity, on the other hand, is unique among spontaneous thought as it involves dynamic shifts between 2 ends of a spectrum of constraints as it alternates between the generation of new ideas (highly spontaneous) and critical evaluation of these ideas (more likely to be associated with high automatic constraints than goal directed thought because creatives frequently use emotional, visceral reactions (gut)). FMRI Studies show DNMTL is more active during idea generation than during evaluation, FPCN and DNCore are more active during evaluation than during creation.
The authors of the paper Mind-wandering as spontaneous thought: a dynamic framework summarizes the process as follows, “DNMTL (centered around medial temporal lobe) and Sensorimotor areas can act as a source of variability in thought. Salience networks, dorsal attention network DAN and DNCore subsystem can apply automatic constraints on the output of DNMTL and sensorimotor areas, thus limiting the variability and increasing the stability of thoughts. The frontoparietal control network FPCN can exert deliberate constraints on thought by flexibly coupling with DNCore, DAN, and Salience networks and thus either reinforcing or reducing the automatic constraints. Roles mentioned here of each network is meant to be illustrative rather than comprehensive, only including those interactions which are so far well understood”, as specified by the study authors.
During spontaneous thoughts like “when I was walking, I began daydreaming about my new laptop that I ordered, then recalling the state of my previous pc to when will I ever buy a console to “do I even need one”? to “and so on. Here DNMTL exerts a relatively strong influence on thoughts with relatively low deliberate and automatic constraints.
During automatic constrained thought like “as I took a turn on the road, I began to worry about whether I will be able to finish my article before the end of the month. I haven’t completed any project in the last 2 years and have only one completed article up on my website. Will even a single person ever read this?”. Here salience networks and DNCore exert relatively strong automatic constraints on thought with relatively weak variability and deliberate constraint.
During deliberate constrained thought like “when I reached my place, I realized that these thoughts are making me feel miserable and I tried to think of something else entirely. What task am I on currently? oh yeah, I have to reach my place and start reading my book or maybe I should exercise first and then read and.” Here the FPCN exerts strong deliberate constraints on thought with relatively weak sources of variability and automatic constraints.
Clinically significant alterations in spontaneous thought can be as follows: those marked by excessive variability of thoughts over time (can prevent coherence i.e., meaningful interconnectedness between thoughts) and those marked with excessive stability (reducing the flow of thoughts), both of which if they become chronic can have unfavorable effects on one’s wellbeing.
Depression and rumination seem to be characterized by excessive stability in thought, more specifically negative thoughts, and the resulting difficulty in disengaging from such thoughts. One characteristic of depression is rumination where one remains fixated on one’s problems and feelings about them. Being largely involuntary suggests underlying automatic constraints. During depression, DN, and salience networks show greater activation while FPCN shows lower activation which suggests depression has automatic constraints on thought. In a recent metanalysis, individuals with depression showed greater connectivity in DN and reduced connectivity within the FPCN. And that FPCN shows increased coupling with DN but decreased coupling with DAN which may reflect inclinations towards internal thoughts at the cost of engaging with the external world. The overly connected DN imposes greater automatic constraints on DNMTL, therefore, leading to an overly constrained thought flow with excessive internally focused thoughts.
ADHD, under this framework, would be characterized by excessive variability and FMRI studies have indicated it is associated with reduced FPCN and DAN activation while failing to deactivate regions of DN leading to a reduction in both automatic and deliberate constraints on thought and a possible increase in DNMTL derived source of variability.
In an opinion paper by Joshua Shepherd, he proposes that the reason unintentional mind wandering is initiated and sustained could be due to the expected value of the current task being considered as too low; which initiates a search for a better task/goal. We already do this explicitly when looking for better goals/task with higher rewards but this proposal asks the question of whether this also happens implicitly i.e., without intention by the cognitive control system.
In a paper titled “Why the mind wanders: How spontaneous thought’ s default variability may support episodic efficiency and semantic optimization”, the authors describe that Variability in thought could be an adaptive mechanism that facilitates efficient encoding of separate episodic memories (as specified by the episodic efficiency hypothesis) and to integrate and transform episodic memories to semantic knowledge (semantic optimization hypothesis)
There are 2 pivotal episodic memory mechanisms – Pattern separation and pattern completion. Pattern separation produces distinct separate episodic memories through decorrelating (making distinct) memory patterns in hippocampus and neocortex (Neocortex can be compared to a library where information is stored while hippocampus is a librarian who refers to where the information is stored, as described by Yassa and Reagh (2013)), thus creating distinct neural representations for individual episodic events. Episodic efficiency hypothesis states that the function of variability in spontaneous thought over time may directly support the process of pattern separation through reactivations or novel recombination of mental stimulations by providing dissimilarities in consecutive representations over time and thus decorrelating distinct memories from one another. Pattern completion strengthens the representation of separate memories via multiple similar re-instantiations of the same memories, triggered by internal or external cues.
Spontaneous thoughts are often about recent memories, past events, or future plans. Content variability inherent to spontaneous thought contributes to episodic memory separation in the following ways: Spontaneous reactivations can create a framework for initiating competition between multiple instantiations of given memory, ultimately preserving the recurring (similar) features of the memory. Each spontaneous thought is reencoded as a new memory trace which competes with previously encoded memories in such a way that overlapping features that are present between them are strengthened and retained for later recall, while non overlapping features are potentially lost over time. Secondly, the role of variability in decorrelating memories by acting as a time buffer between overlapping memories. If enough time passes between similar memory traces, such that activation of one memory dies down before the other related memory can be activated results in avoiding association between those traces.
For example… if there is low variability between thoughts like “waiting for a train” “getting on the train” and “conversing on the train”, these become grouped together just based on their temporal continuity. However, due to content variability in spontaneous thought, our mind usually has a lot more going on. Like “waiting on the bus” “have my to-do list to review” “plan to write” “getting on the train” “meeting at 2” “that episode yesterday from that tv show” etc. Here content variability creates enough time between temporally contingent memories that these are not grouped together and are not strengthened. Thus, spontaneous mental stimulations play a role in separating episodic memories that are important to one’s life such as experiences that need to be distinguished from others and to avoid them being grouped with others due to temporal proximity.
Pattern completion on the other hand is a process of generating complete memory of events or experiences from the recall of its parts, that is partial cues can trigger and reactivate original episodic memory including the context and emotional tone of the experience, thus resulting in a vivid reexperience of that memory. Memories recalled during pattern completion may provide partial cues that may trigger further pattern completions resulting in a continuous chain of mental contents. For example, seeing a protein shake may become a cue to completing a memory of working out, where the workout may lead to completing a memory about the contents of an article you read about nutrition, which might complete a memory about eating broccoli soup and so on. This may explain why our minds keep moving and how new mental content emerges.
So, if there is a propensity for consecutive mental stimulations to be decorrelated by pattern separation, partial cues from them are likely to trigger patterns somewhat dissimilar to immediately preceding patterns; which may be one of the reasons why spontaneous thought exhibits heightened variability over time while also allowing for thematic associations between those consecutive mental states.
Moving on to the Semantic optimization hypothesis, Variability in spontaneous thought also supports the organization of semantic memory by conditions required for effective episodic to semantic transformations, as in similarities across representations through multiple exposures facilitate regularities within a category, while dissimilarities provide contrasting evidence to help to form a category’s boundary. For example, if a child only sees one breed of dog, he may not realize that another breed is also a dog. Or if a child is shown 50 pictures of cats in a day, he or she may confuse a small dog for a cat a while later. Eventually, we realize the differences between these categories. Similarities over repeated exposures strengthen concepts and categories while dissimilarities mitigate the overlearning of a single instance. This gradual exposure which takes an average of a large number of recent examples is termed interleaved learning. Here, for example, learning a new set of information AC will not interfere with our previously learned AB, in a way that we can actually distinguish between AB and AC, without letting one replace another. Thus, consolidation avoids this catastrophic interference (loss of previously learned information due to acquiring new information) and ensures the transformation of newly encoded memory trace from temporary to a more stable permanent form. This is accompanied by the existence of 2 independent yet complementary systems: Pattern separation and Pattern completion at a hippocampal level paired with gradual interleaved learning at a neocortical level and that spontaneous thought variability support this episodic to semantic transformation.
Finally, in a paper titled “Mind-wandering and self-referential thought”, the author describes how most mind wandering episodes involve future oriented and goal-oriented contents which may help anticipate future events, and predict possible outcomes for different sets of actions that could inform decisions and behaviors, although the extent of such thoughts being beneficial to guide decisions and behavior needs to be investigated further. Mind wandering might contribute to our sense of self and personal identity as a result of the frequent explorations of the past and the future, providing a sense of continuity of self through time and contributing to the creation and maintenance of self-models.
References:
- Corballis Michael, Wandering tales: evolutionary origins of mental time travel and language, Frontiers in Psychology, Vol 4 Year 2013, https://www.frontiersin.org/article/10.3389/fpsyg.2013.00485; DOI=10.3389/fpsyg.2013.00485; ISSN=1664-1078
- William A. Roberts, Mental Time Travel: Animals Anticipate the Future, Current Biology, Volume 17, Issue 11,2007, Pages R418-R420, ISSN 0960-9822, https://doi.org/10.1016/j.cub.2007.04.010. (https://www.sciencedirect.com/science/article/pii/S096098220701250X)
- Ólafsdóttir, H. F., Bush, D., & Barry, C. (2018). The Role of Hippocampal Replay in Memory and Planning. Current biology: CB, 28(1), R37–R50. https://doi.org/10.1016/j.cub.2017.10.073
- Corballis Michael C., Language, Memory, and Mental Time Travel: An Evolutionary Perspective, Frontiers in Human Neuroscience, Vol 13 Year 2019, https://www.frontiersin.org/article/10.3389/fnhum.2019.00217; DOI -10.3389/fnhum.2019.00217; ISSN=1662-5161
- Christoff, K., Irving, Z., Fox, K. et al. Mind-wandering as spontaneous thought: a dynamic framework. Nat Rev Neurosci 17, 718–731 (2016). https://doi.org/10.1038/nrn.2016.113
- Joshua Shepherd, why does the mind wander? Neuroscience of Consciousness, Volume 2019, Issue 1, 2019, niz014, https://doi.org/10.1093/nc/niz014
- Mills, Caitlin, Arianne Herrera-Bennett, Myrthe Faber and Kalina Christoff. “1 Why the mind wanders: How spontaneous thought’ s default variability may support episodic efficiency and semantic optimization.” (2017). Online Publication date: April 2018, The Oxford Handbook of Spontaneous Thought: Mind-Wandering, Creativity, and Dreaming; Edited by Kalina Christoff and Kieran C.R. Fox; DOI:10.1093/oxfordhb/9780190464745.013.42
- D’Argembeau, Arnaud. (2018). Mind-wandering and self-referential thought. 10.1093/oxfordhb/9780190464745.013.14.
- Featured image: Photo by Hadija Saidi on Unsplash
