If you have ever made a grocery list, programmed an appointment into your online calendar, or used your calculator to figure out the appropriate tip on a restaurant bill, you have used cognitive offloading as a strategy in your daily life. Cognitive offloading is defined as “the use of physical action to alter the information processing requirements of a task so as to reduce cognitive demand” (Risko & Gilbert, 2016, p. 677). Offloading can help overcome the well-established capacity limits of cognitive processes such as working memory or visual perception and has been shown to reliably improve performance in these domains compared to conditions in which offloading is prohibited (see Risko & Gilbert, 2016, for a review). This type of behavior is something that many of us engage in throughout the course of our daily lives.
Despite the prevalence of cognitive offloading in modern life, this type of behavior has been studied very little in comparison to the wealth of literature dedicated to investigating facets of internal information storage (i.e., storing mentally). Here, we seek to better characterize the individuals who tend to offload versus those that more often rely on their internal memory stores. We specifically examine whether one’s propensity for offloading information is tied to their working memory capacity (WMC), defined broadly as the number of items one can simultaneously hold in a highly accessible state (Cowan, 2017). We test the prediction that those who can hold fewer items in working memory will be more likely to offload.
Previous work on cognitive offloading
In recent years, the nascent field of cognitive offloading has examined two main questions: 1) what factors influence whether one chooses to offload; and 2) what are the downstream cognitive consequences of choosing to engage in offloading (Risko & Gilbert, 2016)?
To address the first question, researchers have assessed both traits and tendencies within individuals that might predict the likelihood of offloading. Recent work has also experimentally manipulated conditions under which participants may be more inclined to engage in offloading. Perhaps unsurprisingly, lower objective memory ability and lower subjective memory ratings corresponded to higher levels of offloading behavior (Gilbert, 2015b). More interestingly, these patterns held true even when subjective confidence was not correlated with objective memory performance (Gilbert, 2015b). Recent work by Risko and Dunn (2015) replicated these findings in a short-term memory task involving letter memory. They demonstrated that short-term memory capacity, measured by performance in a “no-choice condition” of their task where participants were not permitted to offload, was inversely related to the frequency of offloading in the choice condition during which participants were able to choose whether or not to offload information. This finding regarding short-term memory capacity and offloading will be further investigated herein.
Moving on to experimental manipulations that influence offloading behaviors, researchers have identified two main factors that make offloading more likely. Gilbert (2015a) administered a task that involved dragging numbered circles to the bottom of the screen in order from 1 to 10, except for the instruction to move either one or three circles to different locations on the screen. The tendency to choose to engage in offloading was influenced by increased memory load (three-circle condition compared to one-circle condition) or the addition of a secondary processing demand to the ongoing task (in this case completing a distractor task during the trial). In short, increasing the memory demands of the task and/or reducing available attentional resources that can be devoted to the task at hand made participants more likely to engage in offloading behavior.
With respect to the question of the downstream effects of offloading, as one might suspect, offloaded intentions are more likely to be fulfilled later compared to intentions that are not offloaded (Gilbert, 2015a, 2015b), suggesting an overall benefit for offloaded versus internally stored representations. However, in some cases individuals choose to offload even when it does not improve performance (i.e., when performance is already at ceiling; Gilbert, 2015a, 2015b; Risko & Dunn, 2015). Why might individuals choose to exert the effort to offload without an obvious performance benefit? Recent work suggests that individuals who engage in offloading even when there is no objective performance benefit may be driven to do so by the incorrect belief that offloading will lead to better performance (Risko & Dunn, 2015).
There are also some downsides to choosing to offload. In an experiment where participants studied trivia statements and believed that the studied information would or would not be stored for later access (here, believing that the statements would be stored and accessible later is thought of as akin to offloading), participants who thought they would have access to the stored information later showed poorer memory for the studied items (Sparrow, Liu, & Wegner, 2011). Similar effects have been found with visual stimuli. In a study where participants were asked to visit a museum and take photos of some artifacts and simply visit and observe other artifacts, those objects that were photographed were remembered more poorly later. This has been termed the photo-taking-impairment effect (Henkel, 2014). More recent work suggests that this effect may not be solely due to the belief that the camera will do the ‘remembering’ for them; recently, the photo-taking-impairment effect was observed even when participants did not believe they would have access to photos later (Soares & Storm, 2017). The authors of this recent work suggest that engaging in photo-taking might disrupt the typical processing or encoding of object features (Soares & Storm, 2017), but more work is necessary to disentangle the consequences of engaging in offloading for processes related to initial encoding and for subsequent recall of information.
These prior findings suggest that offloading can lead to impairments in performance when offloaded information is not available at the time the information is needed, but that offloaded information, when available for use later, typically benefits performance. In addition, lower subjective ratings of memory ability and greater cognitive demand during the task have been found to relate to higher levels of offloading behavior. However, less is known about the specific cognitive processes that may be involved in or disrupted by offloading.
Individual differences in WMC and relation to other cognitive domains
Working memory was recently described as reflecting “an ability to maintain information in the maelstrom of divergent thought” (Engle, 2018, p. 192), and we predict that the capacity of this cognitive system may relate to whether one offloads information for future use as a memory aid. However, extant models of working memory and methods of testing WMC do not consider or test whether, when given the opportunity, people offload information from working memory to external sources. This limits the extent to which these models apply to real-world situations in which individuals use external memory aids.
The present work bridges the study of cognitive offloading with the literature examining individual differences in WMC. WMC has long been recognized as related to performance in other cognitive domains. For example, individuals with higher WMC show better performance on reasoning tasks (Kane et al., 2004), reading comprehension (Daneman & Carpenter, 1980) and controlled search of long-term memory (Unsworth, Brewer, & Spillers, 2013). In addition, people with higher WMC estimates exhibit smaller Stroop interference effects (Kane & Engle, 2003), reduced interference from background noise (Conway, Cowan, & Bunting, 2001), and the tendency to deploy proactive cognitive control (Redick, 2014; Richmond, Redick, & Braver, 2015; Wiemers & Redick, 2018). However, individuals with high WMC are not universally advantaged, and deficits have also been associated with higher WMC. Individuals with higher WMC have more difficulty recovering items in a directed forgetting task (Delaney & Sahakyan, 2007), exhibit poorer recognition memory on a surprise recall test for neutral words in a Stroop task (Shipstead & Broadway, 2013), and show longer response times on trials for which participants must override the prepotent response in a cognitive control task (Richmond et al., 2015).
While robust in many areas, the literature on individual differences in WMC says less about whether high and low WMC individuals differ in their tendency to store information internally (using working memory) versus externally (by offloading information to the external world). As WMC is operationalized as performance on measures of working memory (e.g., operation span; Unsworth, Heitz, Schrock, & Engle, 2005), high-span individuals excel by demonstrating that they can successfully retain multiple memory items over the short term. However, the tasks used to measure WMC do not typically afford the opportunity to offload, and just because individuals can store multiple items in working memory does not mean that they will when given an alternative (i.e., writing an item down). A primary aim of the present work is to investigate whether WMC is related to how often one offloads during a short-term memory task (when given the choice).
If there is indeed a relationship between offloading and capacity, one could predict that those who can remember more would offload less. Or, it could be that those with higher capacity also excel at strategic allocation of resources and therefore might be able to identify when offloading may be useful. Risko and Dunn (2015) showed evidence for the former; those who performed better on a verbal short-term memory task were less likely to offload when given the opportunity. The present study seeks to replicate and extend this work by using an expanded battery of tasks to measure WMC on a construct level (i.e., with the aggregate of performance on multiple tasks). This approach will allow for measurement of an underlying WMC construct, and will use complex span working memory measures that are common to the individual differences in WMC literature (Conway et al., 2005). This approach is in line with recent studies focused on individual differences in WMC and its relation to cognitive functioning in other domains (e.g., Redick, Calvo, Gay, & Engle, 2011; Richmond et al., 2015; Shipstead & Broadway, 2013).
Current research questions and hypotheses
Here, we aim to elucidate the relationship between memory ability and offloading behavior in a simple short-term memory task. In this study, we plan to assess four research questions. The first two questions are replications of work presented by Risko and Dunn (2015): 1) does having the choice to offload items from memory by writing them down lead to better performance than using memory alone; does this differ by load; and 2) within a given set size, does choosing to write in the choice condition benefit performance? For both questions 1 and 2, we expect to replicate the effects reported in Risko and Dunn (2015; experiments 1a and 1b), showing that offloading benefits performance and is more beneficial at larger set sizes.
Next, we expand upon previous work by asking: 3) does WMC predict the likelihood that one chooses to offload beyond what is already explained by performance in the no-choice block? We expect that WMC will account for additional variance in offloading frequency, and that it will do so in a negative fashion (i.e., individuals with higher WMC engage in lower levels of offloading). Finally, we examine the predicted performance advantage on trials where the participant can choose to offload (choice condition) versus trials where they must rely on internal memory (no-choice condition). We ask, 4) do WMC and frequency of offloading explain the difference between performance in the choice and no-choice conditions? This fourth question will focus on set sizes thought to be at or above the typical upper limit of WMC (set sizes 6, 8, 10), where offloading should be of benefit. We expect that WMC and the frequency of offloading will each explain unique variance in difference scores. We predict that WMC and choice minus no-choice accuracy will be inversely related to one another (i.e., higher WMC estimates will be related to smaller difference scores), whereas we expect that frequency of offloading will be positively related to choice minus no-choice performance (i.e., higher levels of offloading will predict a larger difference in choice – no-choice performance).