Skip to main content
Log in

Memory as Skill

  • Published:
Review of Philosophy and Psychology Aims and scope Submit manuscript

Abstract

The temporal structure for motivating, monitoring, and making sense of agency depends on encoding, maintaining, and accessing the right contents at the right times. These functions are facilitated by memory. Moreover, in informing action, memory is itself often active. That remembering is essential to and an expression of agency and is often active suggests that it is a type of action. Despite this, Galen Strawson (Proceedings of the Aristotelian Society, 103, 227–257, 2003) and Alfred Mele (2009) deny that remembering is an action. They claim that memory fails to admit of control. Remembering is automatic—once remembering starts, the process can neither be stopped nor intervened on. Moreover, the agent does not initiate remembering. An agent has control over an event or process if and only if she has the capacity and opportunity to initiate and intervene on that event or process. Actions are events over which an agent has control. Since it is automatic, we fail to have control over remembering. Thus, remembering is not an action. In this paper, I draw out an assumption of Strawson’s and Mele’s accounts: an event-type whose tokens exhibit automaticity cannot, for that reason, be an action (§2). Against this assumption, I draw parallels between skilled bodily action and memory. I show that memory exhibits two defining features of skill: it can be learned with practice and it admits of attributions of excellence (§3). These features reveal how intelligent control is exerted in the exercise of skill despite apparent automaticity—control is gained over time (§4). Since exercises of skill are by definition actions and since memory exemplifies the defining features of skill, memory is a skill and instances of remembering are actions too.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others

Availability of Data and Material

Not applicable.

Code Availability

Not applicable.

Notes

  1. I hereafter use ‘event’ to refer both to occurrences that are synchronous and to those which occur over time, viz. processes, except where context requires speaking of the two separately.

  2. One might worry that showing that memory meets the conditions essential to being a skill is insufficient to show that remembering is an action. More specifically, it may be enough to show that remembering can be controlled. But being controllable is only one necessary condition on action. If I cannot show that memory exemplifies whichever other conditions are individually necessary and jointly sufficient for counting as an action then I have not shown that remembering is an action.

    In response, assuming a causal theory of action, showing that memory can be controlled by the agent is sufficient to show that it can be caused in just the way other actions are caused, namely, by the agent’s intentions. Thus, for instance, if an expert rememberer desires to remember the cards in a deck and knows that she can do so by walking her memory palace then she can intentionally remember the cards. Because part of what is at issue is whether memory can be self-consciously initiated or self-consciously guided and because I think the more interesting cases of mnemonic action are those wherein the process is not self-consciously initiated or not self-consciously intervened on, my argument takes a circuitous route through showing that memory is a skill. The worry that remembering is never intentionally caused because it is sometimes neither self-consciously initiated nor self-consciously intervened on is thus structurally analogous to the worry that a pianist’s automatic glissando is never intentionally caused because it is sometimes neither self-consciously initiated nor self-consciously intervened on. If so, the former worry should dissipate in light of a recognition that the latter worry is ill-founded. I want to thank an anonymous reviewer for pushing me to clarify how my argument connects to action theory.

  3. Wayne Wu (2013) attacks this assumption as well. He asserts that so long as there is some feature of the relevant cognitive process that the agent attends to as a target of intervention, that process counts as controlled and, therefore, an action (253–4). Importantly, since the agent cannot possibly attend to all of the relevant features of any cognitive process (or bodily action) that they have control over, those features of the process which are automatic when not attended to are automatic in act-tokens where the agent exerts control by (in part) attending to some other feature. The limits of attention make it the case that automaticity is pervasive in action generally. This insight of Wu’s plays a substantial role in my account (see §4).

  4. Some philosophers have argued that full-blown automaticity is insufficient to imply lack of control (Fridland 2015b, 2016; Arango-Muñoz and Bermúdez 2018; Douskos 2019). These philosophers claim that many (skilled) actions are both automatic and controlled. They are automatic inasmuch as they often do not require our attention or effortful intervention. But they are also controlled inasmuch as it is open to us to correct mistakes in performance or inasmuch as those performances are flexible and sensitive to our (high-level) intentions or goals. I agree: some instances of (skilled) action are neither initiated nor intervened on, though they could be. Hence why I say that actions are not necessarily fully automatic. Regardless, if the reader feels uneasy about my use of ‘automaticity’ as a contrast to control, feel free to substitute whatever terminology implies lack of control.

  5. I want to thank Felipe De Brigard  for pushing me to clarify the notion of fluid remembering.

  6. Though I speak of fluid and intrusive memory as contraries, they actually fall on a wide spectrum of mnemonic activity. On one side of that spectrum are instances of intrusive memories like those experienced by, e.g., PTSD or hyperthymesia patients. On the other side, there are instances of intentional remembering. One of the loftier goals of this paper is to provide some groundwork for an explanation of the pathological nature of intrusive memories and episodic amnesia rather than just their being non-agentive (cf. Berntsen 2007, 2009). One way to account for the pathologies of intrusive memories and amnesia within the framework proposed by the paper is as follows. In the intrusive cases, if remembering is an action and action is to be understood at least partly in terms of control then a systemic lack of control in memory due to prior injury or insult constitutes a disruption of one’s agency. Call instances of intrusive memory mnemonic spasms. Mnemonic spasms, like their bodily counterparts, obstruct opportunities for exerting control in a way that is symptomatic and, so, are cause for concern regarding the patient’s health. In the case of amnesia, if control is to be understood at least partly in terms of the having of a capacity to initiate and intervene on the relevant process then amnesiacs lack that capacity due to some malfunction (Levy 2013: 714–5). I want to thank Colin Allen for pushing me to consider episodic amnesia and hyperthymesia.

  7. The contrast I draw between fluid and intrusive memory does not map neatly onto what some might call voluntary or involuntary memory. More specifically, Dorthe Berntsen (2007, 2009), a pioneering scholar and researcher on involuntary memory, defines voluntary autobiographical memories as those episodic memories that are initiated by the agent’s consciously deciding to remember and that consist in a negative feedback loop of specifying or revising a verbal cue followed by searching for mnemonic content that matches the original or revised cue, until a satisfactory match is found (2009: 21, 39, 86, 113–114). By contrast, involuntary autobiographical memories are episodic memories that are initiated without the agent’s conscious decision. Often, they are brought about by a cue’s being salient to the agent such that a memory is discriminated by that cue (2007: 20; 2009, passim). Whether and when a cue triggers an involuntary autobiographical memory depends on a host of factors, including whether and how much the agent is attending to other tasks, how congruent the memory is with the agent’s current mood, whether the cue is of something relevant to the agent’s current life-concept or goals, etc.

    My notions of fluid and intrusive memory cut across Berntsen’s voluntary and involuntary autobiographical memory on at least two dimensions. First, I am not limiting myself to episodic memory. Second, I do not restrict fluid remembering to remembering that is initiated by a conscious decision to remember, lest I beg the question. As such, many instances of so-called involuntary autobiographical memory may well be exercises of mnemonic skill on my account despite not being initiated by a conscious decision to remember. At the same time, I grant that some instances of non-intrusive memory may nevertheless be nonactions. On my account, whether a bit of mnemonic activity which was not so initiated constitutes an act-token depends on whether the agent can assume control over that activity after it has been initiated, e.g., by interrupting it (§4.2). That remembering is an act-type is implied by the agent’s enjoying being in a position to initiate and intervene on token-rememberings (§3). Thus, I find the term “involuntary” inapt and possibly question begging. Moreover, by my lights, Berntsen does not give a plausible account of the pathological nature of intrusive memories in cases of, e.g., PTSD or hyperthymesia beyond their distinct phenomenological profile (2009: 162–181). By contrast, my account provides the groundwork for a unified explanation of the pathological nature of PTSD, hyperthymesia, and episodic amnesia (fn.6). In any case, my account is consistent with Berntsen’s ecological approach to involuntary autobiographical memory (2007: 40–44). I want to thank Felipe De Brigard for informing me of Berntsen’s work and for pushing me to clarify my account relative to hers.

  8. More specifically, when I say that a condition on skill acquisition is that tasks be integrated with one another, I mean that the relevant activities are performed with a view to acting as the relevant principles prescribe (§4.2). The agent need not be aware of those principles, so long as someone training her is sufficiently aware of them. For instance, a novice piano teacher, having just found explicit instructions developed in the Faber and Faber method, may help her student engage in practicing extending her fingers outward as she rests her hands on the keyboard and pressing her fingers into ‘O’ shapes against her thumb with a view to getting the student to automatically assume what the Faber and Faber method prescribes as the correct hand position for play. Alternatively, a novice player may find these techniques and practice them herself. What makes these activities “integrated” is that they are organized in a particular way, namely, the way prescribed by the guiding principle(s). I’d like to thank an anonymous reviewer for pushing me to clarify integration of tasks under guiding principles.

  9. One of the necessary conditions for deliberate practice in Ericsson (2008) is the subject’s being motivated to improve. I leave this condition off because it is arguably the one that distinguishes everyday skills from expertise (Ericsson 2008: 991). But the distinction between everyday skill and expertise is controversial (Christensen et al. 2016, 2019; cf. Montero 2016). And since remembering is in most contexts an everyday activity, if it is a skill then it is among those whose acquisition does not require the agent’s being motivated to improve.

  10. There is evidence that episodic and semantic memory are systematically interdependent at least with respect to encoding and retrieval (Greenberg and Verfaellie 2010). If so, and if the exercise of episodic memory is indeed skillful, then, to the extent that the exercise of semantic memory is informed by the skillful aspects of the exercise of episodic memory (or vice versa), the exercise of semantic memory is likely also skillful. Thus, it is plausible that at least all of the declarative division of the classical Tulving taxonomy of memory is captured by the account of memory as skill (Tulving 1972).

  11. Though there is some disagreement about what exactly the faculty comprises (Miller, Galanter, and Pribram 1960; Baddeley and Hitch 1974; Cowan 1999; Miyake and Shah 1999; Oberauer et al. 2003; Postle 2006; Carruthers 2015), the consensus is that its main function is to maintain and process in consciousness information that is drawn from both current experience and long-term memory.

  12. Another goal of such studies is to test for what is called ‘far transfer.’ Transfer is far when a subject who has improved on a specific task does significantly better than controls on tasks that are unlike the trained task but are thought to rely on the same cognitive process(es). Acquiring a skill often leads to improvement in tasks that depend on the same motor or cognitive processes. Hence, skill acquisition tends to induce far transfer. As the novice pianist improves, she may well find herself better able to, say, discern changes in pitch in spoken Mandarin given prior familiarity with the language (Nan et al. 2018).

  13. Tasks are adaptive if they increase in difficulty when subjects answer correctly and decrease in difficulty when subjects answer incorrectly.

  14. Improvement can be measured on a number of behavioral dimensions and is most often related to increases in the efficacy, reliability, and style with which one acts in accordance with guiding principles as a result of practice (§3.2). Improvement may also be measured in terms of the efficacy or reliability of isolable cognitive processes relative to some baseline (see discussion of the Smith et al. 2009 study below). This means that improvement can be measured in terms of greater accuracy, vividness, chunking or parsing capacity or concatenation (see the case of SF below and fn.16, fn.22; cf. fn.18), core narrative structure, valence, etc. depending on the context. Empirical studies often focus on improvements in the exercise of dissociable cognitive capacities, e.g., auditory recall, relative to some baseline and use stimuli simple enough that experimenters can control for the relevant dimension(s) of improvement, e.g., number of items recalled. By contrast, mnemonists (§3.2) infer improvement on a number of dimensions relative to performance. More specifically, they focus on any and all of the dimensions listed in this footnote with the possible exception of valence and with the plausible inclusion of the development and mastery of novel techniques—some techniques allow one to remember (only) 999,999 individual items while others might allow one to remember 999,999,999 items (Foer 2011: 163–168). Achieving mastery of the latter system (or developing and mastering an even more impressive one) would count as improvement by the mnemonist’s lights, however one achieved it. I want to thank Felipe De Brigard for pushing me to clarify how improvement is measured.

  15. In digit span, subjects are given a string of digits and then asked to repeat that string back to the experimenter in the order received (forward span) or starting from the last digit (backward span). For an overview of the history and use of digit span and related tasks, see Wambach et al. 2011.

  16. It is worth noting that SF’s training did not transfer far (fn.12)—his verbal working memory stayed at around 7 (± 2) elements. It is likely that a lack of variety in SF’s training was its undoing with respect to far transfer. It is also worth noting that SF’s working memory capacity may well have remained at the normal limit throughout training. That is, at the height of his practice, SF could have been encoding around 7–9 digits into a single chunk and bringing about 7–9 chunks into working memory at recall (≈ 49–81 digits) (fn.22). What allowed for the increase in the number of digits encoded into single chunks and for the possibility of reliable decoding of chunks was likely the development of knowledge structures or templates for understanding the digits in terms of, e.g., running times (Guida et al. 2012). I want to thank Colin Allen for pushing me to clarify the distinction between number of chunks and number of digits in SF’s performance.

  17. Note: this list is not meant to be exhaustive of the possible non-epistemic dimensions along which assessments of mnemonic activity can be made.

  18. One might worry that mnemonists fail to exhibit far transfer (fn.12). We tend to think of skills as exhibiting some degree of far transfer, that is, improvement in tasks that engage cognitive processes beyond the specific cognitive processes that were trained on. If mnemonists fail to exhibit any degree of far transfer, memory as they practice it may not be a skill at all.

    In response, mnemonists exhibit a degree of far transfer comparable to that of several other skills. On the one hand, it is true that some mnemonists cannot easily transfer an ability to recall digits to recalling faces or names or vice versa (Foer 2011: 168). But a comparable claim applies to other skills as well: many pianists may well fail to transfer an ability to play some pieces in a particular style to playing some other piece or to playing in another highly specific style. There are often intrinsic limits to how much an individual can master (fn.19). On the other hand, mnemonists are tested on a variety of distinct tasks that almost certainly involve some degree of transfer between cognitive processes, e.g., memorizing decks of cards or sets of digits and lines of poems or names and faces. That the same mnemonists can be competitive across these tasks suggests that there is some degree of far transfer. I want to thank Felipe De Brigard for pushing me to clarify how mnemonists likely exhibit far transfer.

  19. One could object that some attributions of excellence do not track the exercise of a capacity that one could shape. For instance, it seems felicitous to say, “Jones is an excellent digester.” Digestion is not a capacity over which we have any control and, so, is not one we can shape. It seems, then, that admitting of attributions of excellence is not even a necessary condition for skill, let alone a defining feature.

    In response, the felicity of “Jones is an excellent digester” depends on the possibility of shaping other capacities which have downstream effects on digestion, e.g., mental and physical tolerance for, say, spicy foods. After all, one learns to control what, when, how, where, and why one eats. And one can come to control a number of other capacities, e.g., for exercise, which have long term impacts on digestive health. Being an excellent digester, then, means having mastery over capacities the exercise of which redounds well on digestion. Or, at the limit, it means having traits that makes one well suited to such mastery and that to a lesser degree result in better or more tolerant digestion. As Amy Kind (2021) points out, all skills have as part of their enabling conditions biological grounds in, e.g., genes (341–2). Kind’s point applies to memory as well (including hyperthymesia, fn.6). What distinguishes memory from digestion is that the control gained through deliberate practice is gained over the mnemonic events themselves rather than just the exercise of other, mnemonic-adjacent capacities. I want to thank Kate Stanton for this objection.

  20. Arango-Muñoz and Bermúdez (2018) present Strawson’s argument in a way that appears to depend on Mele’s trying condition: if φ-ing is a mental action then one can control φ-ing by both intentionally trying to φ and intentionally trying not to φ. But the agent cannot intentionally try not to imagine (Strawson 2003: 240). Since episodic remembering is a reconstructive process that heavily overlaps with imaginative processes, it is a form of imagination. Thus, the agent cannot intentionally try not to episodically remember. Hence, remembering is not under the agent’s control. Hence, remembering is not an action.

    There are a few reasons to worry about this presentation of Strawson’s argument. First, the inference from episodic memory being a nonaction to all remembering being nonaction is too quick. Second, it is not clear that episodic memory is a form of imagination such that the same conditions for counting as, say, daydreaming or mind-wandering apply to it. If anything, when asked to not remember something, the agent can exploit the epistemic norms that govern memory to her advantage. That is, she can intentionally not remember by intentionally misremembering. Third, even granting that episodic memory is just a form of imagination, it is not clear that we really cannot intentionally try to not imagine or remember. Both Strawson’s (2003) argument and Arango-Muñoz’s and Bermúdez’s (2018) extension of it are intuition pumps. Neither appeal to what psychologists call the ‘white bear phenomenon’ or ‘ironic processing’, wherein attempting to suppress a thought makes its appearance more likely (cf. Strawson 2003: 240, fn.30). But even if they had, there is evidence that sufficient practice leads to successful suppression (Cunningham and Egeth 2016). This evidence is a further point in favor of the account of memory as skill. I want to thank Colin Allen for pushing me to consider Arango-Muñoz’s and Bermúdez’s (2018) presentation of Strawson’s argument.

  21. For evidence of the deployment of attention by elite athletes during performance see, e.g., Davids et al. (1999). For a detailed neurophysiological and behavioral account of the role of attending to internally generated information in both top-down (voluntary) and bottom-up (involuntary) conscious episodic recollection, see De Brigard (2012).

  22. In the interest of space, this story of the dart player’s progress does not discuss the role of what is sometimes called ‘concatenation’ or ‘chunking’ and ‘parsing’ in the empirical literature on the acquisition of motor skills (Verwey 2010; Verwey et al. 2011; Wymbs et al. 2012; Fridland 2019).

  23. Such experiences are grouped under ‘metacognitive feelings’ or ‘metacoginitive judgments’ in the empirical literature on metacognition and under ‘metamemory’ in the literature on memory (for overviews, see Dunlosky and Bjork 2008; Proust 2013; Dunlosky and Tauber 2016). I use ‘self-monitoring’ rather than ‘metacognition’ or ‘metacognitive feelings/judgments’ for two reasons. First, unpacking the entirety of the literature on monitoring (and control) in metacognition/metamemory would take us beyond the scope of this paper. So, I use ‘self-monitoring’ to capture metacognitive monitoring generally. Second, since self-monitoring in skillful bodily action is not usually identified with metacognition and part of the aim of this subsection is to draw a parallel between the use of self-monitoring in skillful bodily action and its use in skillful remembering, I use ‘self-monitoring’ throughout.

References

Download references

Acknowledgements

I want to thank Wayne Wu, Kate Stanton, John McDowell, and Colin Allen for their tireless support, helpful comments, and encouraging feedback. I also want to thank my partner, Morgan Misko, to whom I am infinitely indebted, for her unending care, immense compassion, and unwavering faith in me. I’d also like to thank Felipe De Brigard and an anonymous reviewer for comments on an earlier draft of this paper.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Seth Goldwasser.

Ethics declarations

Conflicts of Interest/Competing Interests

Not applicable.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This paper was awarded the 2021 Philosophy of Memory Essay Prize, hosted by the Centre for Philosophy of Memory at the Université Grenoble Alpes.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Goldwasser, S. Memory as Skill. Rev.Phil.Psych. 14, 833–856 (2023). https://doi.org/10.1007/s13164-021-00605-x

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s13164-021-00605-x

Navigation