Radicalizing simulationism: Remembering as imagining the (nonpersonal) past

ABSTRACT

On the simulation theory of memory, to remember is to imagine an event from the personal past. McCarroll has recently argued that, because it implies not only that a genuine memory need not be caused by the rememberer’s experience of the remembered event but also that the rememberer need not even have experienced that event, simulationism is unable, first, to explain infantile amnesia (the inability to remember events that occurred in one’s early childhood) and, second, to rule out certain “impossible” memories (namely, memories of events that occurred before one was born). Responding to McCarroll, this paper argues that simulationism is in fact able to explain infantile amnesia but concedes that it is unable to rule out pre-birth memories. It goes on to argue, however, that, rather than leading us to reject the theory, this should lead us to endorse a radicalized simulationism on which to remember is simply to imagine an event from the past, regardless of whether that event belongs to the personal past.

KEYWORDS:

• Memory
• imagination
• simulation theory of memory
• infantile amnesia
• memory traces

Acknowledgments

Thanks to audiences at Memory and Mind: A Sofia-Grenoble Workshop, the Language, Epistemology, Metaphysics, and Mind Research Interest Group seminar at the University of Toronto, the Thumos seminar at the University of Geneva, and the Centre for Philosophy of Memory’s internal seminar for discussion and to Juan Álvarez, Anja Berninger, Chris McCarroll, André Sant’Anna, Íngrid Vendrell Ferran, and two anonymous reviewers for written comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Section 1 of this paper overlaps significantly with section 1 of Michaelian (Forthcoming a), with which it originally formed a single paper.

2. On the continuism-discontinuism debate, see, Michaelian et al. (2020), Langland-Hassan (Forthcoming), and Schirmer Dos Santos et al. (Forthcoming). This paper will assume, as is standard, that causalism aligns with discontinuism and simulationism with continuism, but see, Langland-Hassan (2021) and Sant’Anna (2021) for alternative views. Variants of causalism have proliferated in recent years (Michaelian & Robins, 2018). This paper will take a generic causalism along the lines of that developed by Martin and Deutscher (1966) for granted. Variants of simulationism that may be compatible with causalism have been proposed (De Brigard, 2014a; Hopkins, 2018; Shanton & Goldman, 2010). This paper will take the postcausal simulationism of Michaelian (2016c) for granted. Note that Michaelian (2021) defends a simulation theory that differs significantly from that originally proposed in Michaelian (2016c). This paper will, since the response to McCarroll developed here is available to partisans of both versions of the theory, focus on the simpler version proposed in Michaelian (2016c).

3. Relationalist alternatives to representationalism have been receiving increasing attention. While these are worth taking seriously (Aranyosi, 2020; Moran, Forthcoming; Sant’Anna, 2020), they will not be considered here.

4. The contentful character of memory traces has recently been contested (see, Hutto, Forthcoming; Hutto & Peeters, 2018; Michaelian & Sant’Anna, 2021; Werning, 2020). McCarroll’s critique of simulationism assumes that traces are contentful, as does Michaelian’s (2016c) formulation of simulationism, and contentless approaches will thus not be taken into account here. There is a general lack of clarity in the literature concerning the nature of memory traces (see, De Brigard, 2014b; Robins, 2017a, 2017b); given that, as argued below, both causalism and simulationism will invoke traces in explaining infantile amnesia, this lack of clarity may pose problems for both accounts.

5. See, Bernecker (2017) and Robins (2016, 2019, 2020b) for causalist treatments of confabulation.

6. References to McCarroll in what follows are to McCarroll (2020), Andonovski (2019), Perrin (2021), and Werning (2020) for critiques of other aspects of simulationism.

7. McCarroll argues that its commitment to NO-C renders simulationism unable to account not only for infantile amnesia but also for forgetting. See, Michaelian (Forthcoming a) for a full response to this aspect of McCarroll’s argument.

8. Thanks to two anonymous referees for encouraging me to think more carefully about this issue.

9. The simulationist might be tempted to argue, at this point, that, while I can indeed imagine my first birthday party, I cannot do so in the relevant sense of “imagine”. As Langland-Hassan (2021) has pointed out, the simulationist owes us a description of the kind of imagination of which he takes memory to be a form. Langland-Hassan himself suggests that the kind of imagination at issue is what Van Leeuwen refers to as “constructive imagination”, which the latter characterizes as “the capacity to form novel representations” (Van Leeuwen, 2013, p. 224). One might, in principle, suggest that the kind of imagination at issue is, instead, the judgment-involving imagistic imagination described elsewhere by Langland-Hassan (2020) and maintain that I am unable to imagine my first birthday party because, while I can form a mental image of my first birthday party, I cannot, given that I am aware that I lack sufficient knowledge of that event to enable me reliably to form accurate mental images of it, judge that that image accurately represents my birthday party. There are several problems with this strategy. First, given the possibility of nonbelieved memories (Mazzoni et al., 2010), it is not clear that simulationism takes or ought to take memory to be a form of judgment-involving imagistic imagination. Second, there is no apparent reason to rule out the possibility that I might not be aware that I lack sufficient knowledge of my first birthday party to enable me reliably to form accurate mental images of that event. Finally, and most seriously, if my mental images of the event are sufficiently schematic, I might correctly take myself to be able reliably to form accurate mental images of it and therefore judge that a given image accurately represents my first birthday party.

10. See 118–119 on “lost in the mall” memories (more on which below).

11. In order to enable us to assume with confidence that not all events undergone by the subject count as being experienced by him, more would have to be said about the relevant notion of experience. If what matters here is conscious experience, the assumption seems safe, but, as emphasized below, neither causalists (whose theory includes a previous experience condition) nor simulationists have said much about experience.

12. Analogous memories might and presumably do occur in non-laboratory settings.

13. It may be necessary to distinguish between lucky and nonlucky veridical LITM memory. In any case of veridical LITM memory, another agent (such as an experimenter) provides the subject with accurate information about an event from his personal past. In some cases, the agent intends to provide accurate information and does so. There is no luck at work in such cases, and it is reasonable for the simulationist to treat the representations that the subject comes, as a result, to entertain as successful memories. In other cases, however, the agent intends to provide inaccurate information but inadvertently provides accurate information. There is a form of luck at work in such cases, and the simulationist may therefore wish to treat the resulting representations as unsuccessful memories. STM’s proper function condition may be sufficient to enable him to do so, or it may be necessary to add a separate anti-luck condition to the theory (Michaelian, 2021, Forthcoming b). For the sake of simplicity, we may simply assume that the veridical LITM cases at issue here are cases in which whatever conditions end up being necessary are satisfied.

14. Indeed, it is difficult to see how one might coherently move from STM to STM_+PE in order to bring the theory into line with our intuitions about Dalí cases and yet resist incorporating an appropriate causation condition into the theory in order to bring it into line with our intuitions about the cases of absent and deviant causation that originally motivated the causal theory (Martin & Deutscher, 1966). But “STM_+PE+AC” would no longer be a simulation theory – it would be a version of the causal theory, roughly in the vein of that defended by Michaelian (2011).

15. McCarroll suggests a third possible move: the simulationist might appeal to the “internality” condition on remembering introduced in Michaelian (2016b). As McCarroll argues convincingly that the move cannot succeed, and as Michaelian et al. (2020), Michaelian & Sant’Anna (2021), forthcoming b) argues that the internality condition ultimately turns out to be a poor fit for simulationism, this move will not be considered here.

16. The notion of the personal past was introduced but not defined by Tulving (e.g., Tulving, 1983). Borrowing the notion from Tulving, Michaelian (2016c) considers difficulties involved in defining it in terms of the experienced past and suggests that the personal past be defined in terms not of the events that the subject has experienced but rather of the events in which he has been involved but does not say what it is for a subject to be “involved” in an event, rendering this definition uninformative.

17. This possibility will be considered in section 4 below.

18. The version of the simulation theory developed in Michaelian (2021) requires, for successful memory, reliability not just at the level of the retrieval process but also at the level of the metacognitive monitoring process that accompanies retrieval. If Dalí was sincere when he claimed to remember pre-birth events, this may have been due in part to a failure of metacognitive monitoring of a kind that, on the version of simulationism in question, would imply that the relevant memories are unsuccessful. There may thus be an additional reason to suppose that simulationism implies that most apparent memories of the relevant sort are merely apparent.

19. There is no need to rehearse these arguments here; see, Michaelian and Robins (2018) for an overview.

20. An opponent might object that the simulationist approach to pre-birth memories has the consequence that the frequency with which we remember such events depends in an important sense on the environment in which we happen to find ourselves: in an environment in which subjects receive suitable testimony more often than they do in the actual environment, memories of pre-birth events might be considerably more frequent than they are in the actual environment. This consequence is not obviously problematic, but it does mean that the adequacy of the simulationist approach depends on the accuracy of the assumption that we receive testimony of the relevant sort only infrequently. Research on the role of interactions with caregivers in the development of memory (e.g., Fivush & Graci, 2017) may provide a starting-point for assessing the accuracy of this assumption.

21. Aranyosi (2020) offers an additional reason in favor of rejecting the personal past condition, arguing that it amounts to a means of smuggling factivity – which the simulationist rejects (Michaelian, 2016c) into the simulation theory. Referring to an event in his grandfather’s life, he writes:

Suppose I can imagine this [event] so well and in so much detail that it really feels as my own past. Is this the same as remembering that [event]? It is not clear, as far as simulationism is concerned. It does appear as though I am deploying the episodic construction system, which aims at representing my past, because it really feels as my past. Now, if Michaelian claimed this can’t be remembering because it is not really my past, then he would smuggle some form of factivity back into the analysis. (377)

Aranyosi’s thought here appears to be that, if the personal past is defined in terms of the events in which the subject was involved, then to say that a given event belongs to the personal past is to say that it is “something that happened to me”, which implies that it is “something that happened” (377), i.e., that the event actually occurred. Now, it is not clear, as noted above, whether the personal past should be defined in terms of the events in which the subject was involved, but let us suppose, for the sake of argument, that, if an event belongs to the personal past, then it is an event in which the subject was involved. What STM says is that a subject remembers only if his episodic construction system aims to produce a representation of an event belonging to his personal past. What the factivity condition says is that a subject remembers only if his episodic construction system produces a representation of an event that actually occurred. The point to note is that it does not follow from the fact that a subject’s episodic construction system aims to produce a representation of an event belonging to his personal past that his episodic construction system produces a representation of an event that actually occurred, even if we suppose not only that, if an event belongs to the personal past, then it is an event in which the subject was involved but also that, if a subject was involved in an event, then the event actually occurred; what follows is merely that a subject remembers only if his episodic construction system aims to produce a representation of an event that actually occurred. But it is, of course, one thing for an episodic construction system to aim to produce a representation of an event that actually occurred and another for it to succeed in producing a representation of an event that actually occurred. Aranyosi’s argument thus does not go through.

22. Note that an analogous objection can be directed against STM. There would seem to be, by simulationist lights, no important difference between an accurate representation of a personal future event produced by a properly-functioning episodic construction system and an accurate representation of a personal past event produced by a properly functioning episodic construction system. STM thus suggests that one can remember the personal future. This is, on the face of it, no less absurd than STM_−PP’s suggestion that one can remember the (nonpersonal) future. Simulationists who are persuaded by the objection against radical simulationism thus should not be tempted to retreat from STM_−PP to STM.

23. Thanks to an anonymous reviewer for suggesting this objection.

24. There may be other possibilities. It is worth noting that there are candidates, in addition to phenomenology and representational format, for the ingredient that distinguishes episodic from semantic memory. Tulving’s later definitions of episodic memory, e.g., characterize it in terms of its evolutionary history, its development across the lifespan of the individual, its proneness to dysfunction, and its probable human uniqueness, in addition to its phenomenology (Tulving, 2002b).

25. Cf., Michaelian (2016c) on process monitoring.

Additional information

Funding

This work is supported by the French National Research Agency in the framework of the “Investissements d’avenir” program (ANR-15- IDEX-02) and by CAPES-COFECUB (Grant Sh 967/20).