Quilty-Dunn et al. convincingly show that language-of-thought hypothesis (LoTH) is alive and kicking in contemporary cognitive science. One domain they do not discuss, however, is episodic memory and imagination (i.e., episodic thought). This is not surprising: Traditionally, episodic thought has been widely viewed in terms of iconic forms of representation. Nonetheless, we believe that episodic thought is rife for being theorized in terms of LoTH. Most importantly, LoTH generates novel perspectives on how humans achieve such remarkable productivity and flexibility when thinking about other places and times.
Recent research on episodic memory and imagination suggests that different kinds of episodic thoughts (past memories, future imaginations, counterfactual imaginations, etc.) are cognitively not individuated through their contents (Addis, Reference Addis2020, Reference Addis2018; Mahr, Reference Mahr2020; Schacter et al., Reference Schacter, Addis, Hassabis, Martin, Spreng and Szpunar2012); that is, episodic contents are “taxonomically neutral” with respect to their cognitive role as imaginations or memories. For example, Mahr, Greene, and Schacter (Reference Mahr, Greene and Schacter2021; see also De Brigard, Gessell, Yang, Stewart, & Marsh, Reference De Brigard, Gessell, Yang, Stewart and Marsh2020) found that participants' ability to recall the contents of a previously imagined event only weakly predicts their ability to recall whether this event was about the future or the past. This finding suggests that whether a given episode is taken to represent the past or the future (say) is not determined by what is retrieved from memory (i.e., episodic content) but by processes that occur before or after such retrieval (Mahr, Reference Mahr2020).
With this in mind, we propose that there are two main ways in which LoTH can be cashed out in episodic thought. On the one hand, LoTH can help to conceptualize the structure of episodic contents: What is encoded into long-term memory and how these contents are later retrieved in the service of the construction of both episodic memories and imaginations. According to the “constructive episodic simulation” hypothesis (Schacter & Addis, Reference Schacter and Addis2007), episodic retrieval consists in the flexible recombination of the elements of previously encoded experiences. Although there is good evidence to support this idea (see, e.g., Schacter & Addis, Reference Schacter, Addis and Abraham2020, for a review), it remains unclear what mechanisms allow such flexible recombination of episodic elements in the service of episodic simulation. These processes are most commonly thought of in terms of associative inference (Addis, Reference Addis2020; Carpenter & Schacter, Reference Carpenter and Schacter2017; Horner & Burgess, Reference Horner and Burgess2013) even though – as Quilty-Dunn et al. point out – LoT-style representations like scene-grammars (Võ, Reference Võ2021), object files (Zimmer & Ecker, Reference Zimmer and Ecker2010), and event files (Hommel, Reference Hommel2004) play a role in structuring the information encoded into long-term memory. Similarly, these representations might play a role in structuring what content is retrieved and how it is composed. The fact that episodic contents could thus exhibit LoT properties – contributing to the flexibility and productivity of episodic simulation – has so far been underexplored. For example, evidence for the influence of “schemas” in episodic encoding and retrieval (Irish & Piguet, Reference Irish and Piguet2013; Renoult, Irish, Moscovitch, & Rugg, Reference Renoult, Irish, Moscovitch and Rugg2019), which also play a role in episodic simulation of future scenarios (Wynn, van Genugten, Sheldon, & Schacter, Reference Wynn, van Genugten, Sheldon and Schacter2022), might be understood in this light (e.g., Draschkow, Wolfe, & Vo, Reference Draschkow, Wolfe and Vo2014; Võ & Wolfe, Reference Võ and Wolfe2013).
On the other hand, LoTH can help to understand how episodic contents come to play their respective cognitive roles. In the minds of adult humans, episodic contents can fill a variety of different roles – for example, as imaginations of past counterfactuals (De Brigard, Addis, Ford, Schacter, & Giovanello, Reference De Brigard, Addis, Ford, Schacter and Giovanello2013) or representations of event types (Addis, Pan, Vu, Laiser, & Schacter, Reference Addis, Pan, Vu, Laiser and Schacter2009). A complete theory of episodic simulation requires an account of how “taxonomically neutral” episodic contents come to fill these roles. Mahr (Reference Mahr2020, Reference Mahr, Sant'Anna, McCarroll and Michaelian2022; see also Mahr & Csibra, Reference Mahr and Csibra2018) argued that (in addition to their contents) episodic thoughts have (at least) five discrete constituents: Temporal orientation (is this event occurring in the past, present, or future?; Mahr & Schacter, Reference Mahr and Schacter2022; Mahr et al., Reference Mahr, Greene and Schacter2021), specificity (is this a unique occurrence or a type of occurrence?; Addis et al., Reference Addis, Pan, Vu, Laiser and Schacter2009), subjectivity (who is the subject of this experience?; Pillemer, Steiner, Kuwabara, Thomsen, & Svob, Reference Pillemer, Steiner, Kuwabara, Thomsen and Svob2015), factuality (did this/will this really happen?; Johnson & Raye, Reference Johnson and Raye1981; Simons, Garrison, & Johnson, Reference Simons, Garrison and Johnson2017), and mnemicity (do I know about this through my own experience?; Mahr, Reference Mahrin press; Mahr et al., Reference Mahr, Van Bergen, Sutton, Schacter and Heyes2023). As a result, episodic thoughts fulfill many of the diagnostic features of LoT-style representations proposed by Quilty-Dunn et al.
First, episodic thoughts exhibit predicate–argument structure. Complete episodic thoughts are the result of the predication of episodic contents by compounds of these constituents according to syntactic rules. For example, an episodic memory can be analyzed as remember[past, specific, factual, self (EPISODE)], where “EPISODE” refers to episodic content. Evidence about the independence of episodic contents and assignments of temporality (Mahr & Schacter, Reference Mahr and Schacter2022; Mahr et al., Reference Mahr, Greene and Schacter2021) support this idea. Further, one can remember imagining suggesting that some of these predicates can be recursively embedded.
Second, these predicates are discrete: Episodic thoughts are the result of the composition of distinct conceptual constituents into an LoT-like “sentence.” Although there is not complete independence between these predicates, the space of possible episodic thoughts includes a large number of such sentences (see Mahr, Reference Mahr2020; Michaelian, Reference Michaelian2016). For example, the above might be easily amended to imagine[past, specific, factual, self (EPISODE)].
Third, several phenomena attest to the fact that this architecture exhibits a large degree of roll-filler independence – the same episodic content might fill different cognitive roles. For example, people are able to “recast” memories of past events into the future (e.g., one might imagine a basketball game in the past but also imagine the same game as occurring in the future; Thakral, Yang, Addis, & Schacter, Reference Thakral, Yang, Addis and Schacter2021), can be convinced to “disbelieve” their memories (Otgaar, Scoboria, & Mazzoni, Reference Otgaar, Scoboria and Mazzoni2014), and regularly change their assessment of whether they are remembering or imagining an episode (Loftus & Pickrell, Reference Loftus and Pickrell1995).
Finally, the predicates of episodic thought are abstract: Although (say) the pastness or futurity of an episode might commonly go along with different contents, temporal orientation itself cannot be depicted (Mahr, Reference Mahr2020; Matthen, Reference Matthen2010).
The hypothesis that episodic thoughts indeed exhibit these features crucially generates a package of unique predictions (for a first pass at testing role-filler independence, see Mahr et al., Reference Mahr, Greene and Schacter2021; and for discreteness, see Mahr & Schacter, Reference Mahr and Schacter2022). Even though research on the role of LoT in episodic thought is still in its inception, there are potentially large theoretical payoffs for taking LoTH seriously in this domain.
Quilty-Dunn et al. convincingly show that language-of-thought hypothesis (LoTH) is alive and kicking in contemporary cognitive science. One domain they do not discuss, however, is episodic memory and imagination (i.e., episodic thought). This is not surprising: Traditionally, episodic thought has been widely viewed in terms of iconic forms of representation. Nonetheless, we believe that episodic thought is rife for being theorized in terms of LoTH. Most importantly, LoTH generates novel perspectives on how humans achieve such remarkable productivity and flexibility when thinking about other places and times.
Recent research on episodic memory and imagination suggests that different kinds of episodic thoughts (past memories, future imaginations, counterfactual imaginations, etc.) are cognitively not individuated through their contents (Addis, Reference Addis2020, Reference Addis2018; Mahr, Reference Mahr2020; Schacter et al., Reference Schacter, Addis, Hassabis, Martin, Spreng and Szpunar2012); that is, episodic contents are “taxonomically neutral” with respect to their cognitive role as imaginations or memories. For example, Mahr, Greene, and Schacter (Reference Mahr, Greene and Schacter2021; see also De Brigard, Gessell, Yang, Stewart, & Marsh, Reference De Brigard, Gessell, Yang, Stewart and Marsh2020) found that participants' ability to recall the contents of a previously imagined event only weakly predicts their ability to recall whether this event was about the future or the past. This finding suggests that whether a given episode is taken to represent the past or the future (say) is not determined by what is retrieved from memory (i.e., episodic content) but by processes that occur before or after such retrieval (Mahr, Reference Mahr2020).
With this in mind, we propose that there are two main ways in which LoTH can be cashed out in episodic thought. On the one hand, LoTH can help to conceptualize the structure of episodic contents: What is encoded into long-term memory and how these contents are later retrieved in the service of the construction of both episodic memories and imaginations. According to the “constructive episodic simulation” hypothesis (Schacter & Addis, Reference Schacter and Addis2007), episodic retrieval consists in the flexible recombination of the elements of previously encoded experiences. Although there is good evidence to support this idea (see, e.g., Schacter & Addis, Reference Schacter, Addis and Abraham2020, for a review), it remains unclear what mechanisms allow such flexible recombination of episodic elements in the service of episodic simulation. These processes are most commonly thought of in terms of associative inference (Addis, Reference Addis2020; Carpenter & Schacter, Reference Carpenter and Schacter2017; Horner & Burgess, Reference Horner and Burgess2013) even though – as Quilty-Dunn et al. point out – LoT-style representations like scene-grammars (Võ, Reference Võ2021), object files (Zimmer & Ecker, Reference Zimmer and Ecker2010), and event files (Hommel, Reference Hommel2004) play a role in structuring the information encoded into long-term memory. Similarly, these representations might play a role in structuring what content is retrieved and how it is composed. The fact that episodic contents could thus exhibit LoT properties – contributing to the flexibility and productivity of episodic simulation – has so far been underexplored. For example, evidence for the influence of “schemas” in episodic encoding and retrieval (Irish & Piguet, Reference Irish and Piguet2013; Renoult, Irish, Moscovitch, & Rugg, Reference Renoult, Irish, Moscovitch and Rugg2019), which also play a role in episodic simulation of future scenarios (Wynn, van Genugten, Sheldon, & Schacter, Reference Wynn, van Genugten, Sheldon and Schacter2022), might be understood in this light (e.g., Draschkow, Wolfe, & Vo, Reference Draschkow, Wolfe and Vo2014; Võ & Wolfe, Reference Võ and Wolfe2013).
On the other hand, LoTH can help to understand how episodic contents come to play their respective cognitive roles. In the minds of adult humans, episodic contents can fill a variety of different roles – for example, as imaginations of past counterfactuals (De Brigard, Addis, Ford, Schacter, & Giovanello, Reference De Brigard, Addis, Ford, Schacter and Giovanello2013) or representations of event types (Addis, Pan, Vu, Laiser, & Schacter, Reference Addis, Pan, Vu, Laiser and Schacter2009). A complete theory of episodic simulation requires an account of how “taxonomically neutral” episodic contents come to fill these roles. Mahr (Reference Mahr2020, Reference Mahr, Sant'Anna, McCarroll and Michaelian2022; see also Mahr & Csibra, Reference Mahr and Csibra2018) argued that (in addition to their contents) episodic thoughts have (at least) five discrete constituents: Temporal orientation (is this event occurring in the past, present, or future?; Mahr & Schacter, Reference Mahr and Schacter2022; Mahr et al., Reference Mahr, Greene and Schacter2021), specificity (is this a unique occurrence or a type of occurrence?; Addis et al., Reference Addis, Pan, Vu, Laiser and Schacter2009), subjectivity (who is the subject of this experience?; Pillemer, Steiner, Kuwabara, Thomsen, & Svob, Reference Pillemer, Steiner, Kuwabara, Thomsen and Svob2015), factuality (did this/will this really happen?; Johnson & Raye, Reference Johnson and Raye1981; Simons, Garrison, & Johnson, Reference Simons, Garrison and Johnson2017), and mnemicity (do I know about this through my own experience?; Mahr, Reference Mahrin press; Mahr et al., Reference Mahr, Van Bergen, Sutton, Schacter and Heyes2023). As a result, episodic thoughts fulfill many of the diagnostic features of LoT-style representations proposed by Quilty-Dunn et al.
First, episodic thoughts exhibit predicate–argument structure. Complete episodic thoughts are the result of the predication of episodic contents by compounds of these constituents according to syntactic rules. For example, an episodic memory can be analyzed as remember[past, specific, factual, self (EPISODE)], where “EPISODE” refers to episodic content. Evidence about the independence of episodic contents and assignments of temporality (Mahr & Schacter, Reference Mahr and Schacter2022; Mahr et al., Reference Mahr, Greene and Schacter2021) support this idea. Further, one can remember imagining suggesting that some of these predicates can be recursively embedded.
Second, these predicates are discrete: Episodic thoughts are the result of the composition of distinct conceptual constituents into an LoT-like “sentence.” Although there is not complete independence between these predicates, the space of possible episodic thoughts includes a large number of such sentences (see Mahr, Reference Mahr2020; Michaelian, Reference Michaelian2016). For example, the above might be easily amended to imagine[past, specific, factual, self (EPISODE)].
Third, several phenomena attest to the fact that this architecture exhibits a large degree of roll-filler independence – the same episodic content might fill different cognitive roles. For example, people are able to “recast” memories of past events into the future (e.g., one might imagine a basketball game in the past but also imagine the same game as occurring in the future; Thakral, Yang, Addis, & Schacter, Reference Thakral, Yang, Addis and Schacter2021), can be convinced to “disbelieve” their memories (Otgaar, Scoboria, & Mazzoni, Reference Otgaar, Scoboria and Mazzoni2014), and regularly change their assessment of whether they are remembering or imagining an episode (Loftus & Pickrell, Reference Loftus and Pickrell1995).
Finally, the predicates of episodic thought are abstract: Although (say) the pastness or futurity of an episode might commonly go along with different contents, temporal orientation itself cannot be depicted (Mahr, Reference Mahr2020; Matthen, Reference Matthen2010).
The hypothesis that episodic thoughts indeed exhibit these features crucially generates a package of unique predictions (for a first pass at testing role-filler independence, see Mahr et al., Reference Mahr, Greene and Schacter2021; and for discreteness, see Mahr & Schacter, Reference Mahr and Schacter2022). Even though research on the role of LoT in episodic thought is still in its inception, there are potentially large theoretical payoffs for taking LoTH seriously in this domain.
Financial support
This work was supported by a Walter-Benjamin Fellowship (MA 9499/1-1) by the German Research Foundation (DFG) to J. B. M.
Competing interest
None.