External Memory Systems: Cognitive Offloading, Transactive Memory and What Records Fail to Encode

Abstract

Humans have delegated memory to external supports since the invention of writing, and the concern that such delegation weakens internal memory is at least as old as Plato’s Phaedrus. The present review asks what contemporary evidence says about the cognitive consequences of externalising memory and about the limits of what records can preserve. We examine the extended-mind hypothesis and its critics, Wegner’s transactive memory framework, the “Google effect” and its replication status, cognitive offloading, the photo-taking-impairment effect and its boundary conditions, saving-enhanced memory, and lifelogging research with wearable cameras in amnesic patients. The evidence indicates that external storage reliably shifts encoding strategy, sometimes at a cost to content memory and sometimes to its benefit, with effects moderated by intention, control over what is captured, and the reliability of the store. Critically, no recording medium encodes the affective, interoceptive and self-referential components of experience; records act as retrieval cues for such states rather than as copies of them. For a programme of emotion preservation, this implies that the target cannot be the record alone but the coupling between record, rememberer and social partners.

Keywords: cognitive offloading; transactive memory; extended mind; autobiographical memory; photo-taking impairment; lifelogging

1. Introduction

The question of whether external memory aids strengthen or weaken the mind predates experimental psychology by more than two millennia. In Plato’s Phaedrus, Socrates relates the myth of Theuth, who presents writing to King Thamus as a remedy (pharmakon) for memory. Thamus replies that writing will produce forgetfulness, because those who trust external marks will cease to exercise their own memory; what has been invented is an aid to reminding, not to remembering (Plato, trans. 1995). The passage is notable for the distinction it draws between internal memory and externally cued recollection, and for its claim that the effect operates through trust in the external store. Both ideas recur in the empirical work reviewed below.

The contemporary literature is fragmented across philosophy of mind (Clark & Chalmers, 1998; Adams & Aizawa, 2001; Rupert, 2004), social psychology (Wegner, 1987; Sparrow, Liu, & Wegner, 2011), cognitive psychology (Henkel, 2014; Storm & Stone, 2015; Risko & Gilbert, 2016; Barasch, Diehl, Silverman, & Zauberman, 2017) and human–computer interaction (Hodges et al., 2006; Sellen & Whittaker, 2010; Loveday & Conway, 2011). The present review integrates these strands around two questions: what does externalising memory do to the internal memory of the person who externalises, and what dimensions of experience do external records systematically fail to encode? We argue that the first question has a defensible but heavily moderated answer, whereas the second has been underexamined and bears directly on any attempt to preserve emotional memory.

2. Where does memory end? The extended mind and its critics

Clark and Chalmers (1998) proposed that cognitive processes are not confined to the skull. Their central illustration contrasts Inga, who recalls a museum’s address from biological memory, with Otto, a patient with Alzheimer’s disease who retrieves the same address from a notebook he always carries. Because the notebook is reliably available, automatically consulted, readily endorsed, and was endorsed at the time of entry, the authors argued that it plays the same functional role for Otto as biological memory plays for Inga and should count as part of his memory. The claim is a parity argument: a process that would count as cognitive in the head should not be excluded merely because it occurs outside it.

The hypothesis has been contested on two main grounds. Adams and Aizawa (2001) argued that cognition is marked by non-derived content and a distinctive causal profile, which notebooks lack; coupling a cognitive system to an external resource does not make the resource cognitive, a move they call the coupling-constitution fallacy. Rupert (2004) objected on empirical grounds: internal memory exhibits regularities such as primacy, recency, interference and generation effects that external stores do not, so a science that treated Otto’s notebook as memory would lose explanatory generalisations. He proposed a weaker “embedded” view, on which cognition depends heavily on external resources without being constituted by them.

The outcome of this debate matters less here than the conditions it makes explicit. Clark and Chalmers’ criteria of availability, automatic reliance, endorsement and prior authorship specify when an external store becomes functionally integrated. Rupert’s objection specifies what is lost when it does: the store lacks the consolidation, reconstruction and affective modulation that characterise biological memory. Both points bear on cases where the externalised material is not an address but an emotionally significant episode.

3. Distributed memory: from transactive systems to search engines

3.1 Transactive memory

Wegner (1987) proposed that close partners and small groups develop transactive memory systems: differentiated stores in which each member holds some information and, crucially, metamemory about what the others know. The system can retain more than its members individually, provided directory information (“who knows what”) is accurate and retrieval from partners is reliable. Wegner also noted its vulnerability: losing a partner through separation or bereavement is a loss of memory access as well as of relationship. The framework is descriptive rather than mechanistic, but it introduced the idea that encoding is sensitive to expectations about external availability.

3.2 The Google effect and its replication

Sparrow et al. (2011) extended this framework to information technology in four experiments. Participants who typed trivia statements into a computer and were told the entries would be saved recalled fewer statements than those told they would be erased; participants primed with difficult questions showed slower colour-naming of computer-related words in a modified Stroop task, interpreted as activated thoughts of search; and when statements were saved into named folders, participants remembered folder locations better than the statements. The authors concluded that the internet functions as a transactive partner and that people preferentially encode where information can be found rather than the information itself.

The replication record is mixed. The Social Sciences Replication Project (Camerer et al., 2018), which ran high-powered replications of 21 experiments published in Nature and Science between 2010 and 2015, included one experiment from Sparrow et al. (2011) and did not obtain a significant effect in the original direction. Because only one of the four experiments was tested, the outcome does not invalidate the paper, but the claim selected for replication should be treated as unestablished. The broader pattern, that expected external availability reduces encoding effort, has convergent support from paradigms reviewed below.

4. Cognitive offloading as strategy

Risko and Gilbert (2016) proposed that the use of external action to reduce internal processing demand, which they term cognitive offloading, should be studied as a metacognitively controlled behaviour rather than an incidental consequence of technology. The decision to offload depends on the perceived cost of internal processing, the perceived reliability of the external resource, and confidence in one’s own memory. Paradigms include intention-offloading tasks in which participants may set reminders for delayed actions, and tasks in which a display can be rotated physically rather than mentally. People offload more when they judge their internal capacity to be low, and somewhat more than optimal performance requires, suggesting a bias toward externalisation when it is cheap.

Two consequences follow. First, offloading is not intrinsically detrimental: delegating a reminder frees capacity and improves prospective-memory performance. Second, the effect on subsequent internal memory depends on what offloading does to encoding. Storm and Stone (2015) found that saving a file of to-be-remembered items before studying a second file improved memory for the second file, a “saving-enhanced memory” effect that disappeared when saving was unreliable. Offloading, on this account, releases resources for subsequent encoding, but only when the store is trusted. The dependency on reliability echoes both Clark and Chalmers’ endorsement criterion and Plato’s diagnosis that the effect of writing operates through trust.

5. Recording experience: cameras, photographs and lifelogs

5.1 The photo-taking-impairment effect

Henkel (2014) had participants tour a museum, either observing designated objects or photographing them. The following day, photographed objects were recognised less accurately, and fewer visual details and locations were recalled, than objects that had only been observed. In a second experiment, photographing a specific detail by zooming in eliminated the impairment and preserved memory for the whole object, indicating that attentional engagement during capture, not the camera itself, drives the effect. Samples were modest.

Barasch et al. (2017) reported an important boundary condition. When participants chose freely what to photograph, memory for visual aspects of the experience was enhanced relative to no-photo controls, while memory for concurrent auditory information was impaired. Volitional photography thus appears to direct visual attention toward the scene at the expense of other modalities rather than uniformly reducing encoding. The effect of recording therefore depends on the agency and attentional stance of the recorder and on which sensory channel the record captures.

5.2 Lifelogging and SenseCam

Continuous capture pushes externalisation to its limit. Hodges et al. (2006) introduced SenseCam, a wearable camera that automatically photographs the wearer’s field of view at short intervals or when sensors detect changes in light, motion or heat. An initial case study with a severely amnesic patient showed that periodic review of SenseCam images supported retention of autobiographical events over months, whereas a written diary did not. Loveday and Conway (2011) examined a further amnesic patient and reported that image review gave access to otherwise inaccessible everyday memories, and that the images acted as cues triggering recollective experience, including sensory and emotional details not visible in the images themselves, rather than merely supplying facts about the day. They interpreted this within Conway’s model of autobiographical memory, in which event-specific knowledge remains stored but requires appropriate cues for access.

Sellen and Whittaker (2010) offered a constructive critique of the lifelogging programme. They argued that “total capture” conflates recording with remembering, that autobiographical memory serves functions (recollecting, reminiscing, retrieving, reflecting, remembering intentions) that an archive does not automatically serve, and that a lifelog is at best a set of cues whose value depends on the rememberer’s capacity to reconstruct from them. Indiscriminate capture also creates a retrieval problem: the more stored, the harder it is to find what matters.

6. What records fail to encode

Across these paradigms, the record is treated as a stand-in for the memory. The lifelogging literature shows this to be a category error. Autobiographical remembering is a reconstructive process that reinstates aspects of the original encoding state, including bodily and affective components. An image records the light that reached a lens; it does not record interoceptive state, the felt significance of the scene, the goals active at the time, or the self-relevant appraisal that determines whether an event is later judged central to identity. Loveday and Conway’s (2011) observation that SenseCam images elicited emotional and sensory details absent from the images is instructive: the affective content was retrieved from the rememberer, not from the record.

Three further limitations follow. First, recording alters the experience recorded: capture redistributes attention across modalities and can disengage the recorder from the scene. Second, records are selective in ways the recorder may not control. Clark and Chalmers’ endorsement criterion presupposes that the person authored the store; automatically curated archives violate that condition, so the resulting system is embedded rather than extended in Rupert’s (2004) sense. Third, transactive systems depend on partners who shared the experience and can be interrogated. A human partner remembers what was felt and can renegotiate the account; a store returns what was entered.

7. Implications for emotion preservation

The literature reviewed constrains any programme that aims to preserve emotional memory, and the constraints can be stated as testable propositions.

First, a record of an emotional episode is not the emotional memory; it is a cue. The preservation target should therefore be the reliability with which a record reinstates the affective state and self-relevant appraisal of the original episode in a rememberer, not the fidelity of the record to the scene. This predicts that records enriched with the rememberer’s own contemporaneous appraisal will cue emotion more reliably than higher-resolution records without it.

Second, capture must not degrade encoding of the components it cannot capture. The photo-taking findings predict that procedures requiring attentional selection by the person, as in Henkel’s zoom condition and Barasch et al.’s volitional condition, will produce less impairment than passive or directed capture, and that capture of one modality will impair memory for others unless the procedure deliberately samples them.

Third, authorship and endorsement matter. Following Clark and Chalmers and the saving-enhanced memory results, stores that the person fills and trusts should support subsequent encoding, whereas stores perceived as unreliable or externally curated should not. This is testable by manipulating perceived reliability and authorship of the archive.

Fourth, emotional memory has historically been distributed across social partners. A preservation programme should evaluate whether records function as shared cues within a transactive system and support reconstruction of an episode between people, rather than substituting for the partner; records may be most valuable when they preserve the directory information a lost partner held.

Finally, the replication history of the Google effect counsels modesty: claims about external stores and internal memory should be tested with preregistered, adequately powered designs before informing practice.

8. Conclusion

Thamus’s objection to writing, that trust in an external mark reduces the exercise of memory, has a partial empirical descendant: expected external availability changes what is encoded, and directed recording can thin memory for the recorded scene. But the evidence does not support a general degradation account. Offloading releases capacity, reliable saving improves subsequent learning, volitional photography enhances visual memory, and reviewed images restore access to episodes in amnesia. What unites the findings is that the effect of an external store depends on the coupling between store and rememberer: reliability, authorship, attentional stance at capture and the availability of partners who shared the episode. The dimension no record captures, the felt state of the person who was there, is also the dimension that gives an episode its value. An account of external memory that omits it describes archives, not remembering.

References

Adams, F., & Aizawa, K. (2001). The bounds of cognition. Philosophical Psychology, 14(1), 43–64.

Barasch, A., Diehl, K., Silverman, J., & Zauberman, G. (2017). Photographic memory: The effects of volitional photo taking on memory for visual and auditory aspects of an experience. Psychological Science, 28(8), 1056–1066.

Camerer, C. F., Dreber, A., Holzmeister, F., Ho, T.-H., Huber, J., Johannesson, M., Kirchler, M., Nave, G., Nosek, B. A., Pfeiffer, T., Altmejd, A., Buttrick, N., Chan, T., Chen, Y., Forsell, E., Gampa, A., Heikensten, E., Hummer, L., Imai, T., … Wu, H. (2018). Evaluating the replicability of social science experiments in Nature and Science between 2010 and 2015. Nature Human Behaviour, 2(9), 637–644.

Clark, A., & Chalmers, D. (1998). The extended mind. Analysis, 58(1), 7–19.

Henkel, L. A. (2014). Point-and-shoot memories: The influence of taking photos on memory for a museum tour. Psychological Science, 25(2), 396–402.

Hodges, S., Williams, L., Berry, E., Izadi, S., Srinivasan, J., Butler, A., Smyth, G., Kapur, N., & Wood, K. (2006). SenseCam: A retrospective memory aid. In P. Dourish & A. Friday (Eds.), UbiComp 2006: Ubiquitous computing (Lecture Notes in Computer Science, Vol. 4206, pp. 177–193). Springer.

Loveday, C., & Conway, M. A. (2011). Using SenseCam with an amnesic patient: Accessing inaccessible everyday memories. Memory, 19(7), 697–704.

Plato. (1995). Phaedrus (A. Nehamas & P. Woodruff, Trans.). Hackett.

Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676–688.

Rupert, R. D. (2004). Challenges to the hypothesis of extended cognition. Journal of Philosophy, 101(8), 389–428.

Sellen, A. J., & Whittaker, S. (2010). Beyond total capture: A constructive critique of lifelogging. Communications of the ACM, 53(5), 70–77.

Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google effects on memory: Cognitive consequences of having information at our fingertips. Science, 333(6043), 776–778.

Storm, B. C., & Stone, S. M. (2015). Saving-enhanced memory: The benefits of saving on the learning and remembering of new information. Psychological Science, 26(2), 182–188.

Wegner, D. M. (1987). Transactive memory: A contemporary analysis of the group mind. In B. Mullen & G. R. Goethals (Eds.), Theories of group behavior (pp. 185–208). Springer-Verlag.