The influence of the self on attentional processes has long been recognized by psychologists. Since the “cocktail party effect” was first described more than 50 years ago (see Cherry, 1953; Moray, 1959), many studies have established that humans are equipped with a mechanism that enables self-relevant information to be attended to rapidly and reliably (e.g., Bargh, 1982; Brédart, Delchambre, & Laureys, 2006; Gray, Ambady, Lowenthal, & Deldin, 2004; Moray, 1959; Shapiro, Caldwell, & Sorensen, 1997; Sui, Zhu, & Han, 2006; Tong & Nakayama, 1999; Turk, van Bussel, Brebner, et al., 2011). Indeed, Turk, van Bussel, Brebner, et al. (2011) showed that when an item is cued as being relevant to self, ERP signals indicate a rapid increase in both visuospatial and executive attention to the item. The tendency for self-cues to capture attention is clearly advantageous, as information that is coupled with the self is likely to be of greater personal importance than material linked with other people. Reflecting this potential importance, information associated with oneself also elicits a robust memory advantage (the “self-reference effect” [SRE] on memory; for a review, see Symons & Johnson, 1997). The question of interest to the present inquiry is whether a link exists between these two features of self-referential material. Specifically, is the memory advantage associated with self-referential encoding dependent on the attentional resources recruited by self-cues?

Self-referential memory effects have been explored through a variety of experimental manipulations. The most widely used paradigm requires participants to explicitly evaluate target trait words in relation to self or to others (e.g., Conway & Dewhurst, 1995; Klein & Kihlstrom, 1986; Klein & Loftus, 1988; Rogers, Kuiper, & Kirker, 1977; Symons & Johnson, 1997). However, paradigms that do not require the direct evaluation of the self- or other-concept can also reveal a self-referential advantage. For example, Turk, Cunningham, and Macrae (2008) showed that co-presenting self-images with stimulus words produced higher subsequent word recognition scores than did co-presenting stimulus words with images of another person. Creating a self-relevant encoding context can also elicit this pattern of memory performance (Cunningham, Turk, MacDonald & Macrae, 2008; van den Bos, Cunningham, Conway, & Turk, 2010). Cunningham et al. showed that items encoded under conditions of imagined self-ownership were more likely to be recognized than items encoded as being owned by another person. Furthermore, Cloutier and Macrae (2008) showed that mere self-involvement at encoding (i.e., picking an outcome by blind selection) enhanced recollection of the outcome. These studies suggest that explicit self-evaluation at encoding is not essential to eliciting an SRE; rather, a simple association between the self and a stimulus at encoding is sufficient to enhance memory.

This finding is somewhat discordant with the standard cognitive account of the SRE, which relies on the application of self-knowledge at encoding to create an elaborate representation organized within the self-concept, and is therefore more easily retrieved (see Klein & Kihlstrom, 1986; Klein & Loftus, 1988; Symons & Johnson, 1997). A great deal of empirical evidence has shown that elaboration and organization both contribute to the SREs elicited in the trait-evaluation paradigm. For instance, Klein and Loftus showed that self-referencing elicits a pattern of memory performance indistinguishable from the organization effects of category sorting and the elaborative effects of a definition task (with unrelated and related lists, respectively). The importance of both organization and elaboration was also highlighted in Symons and Johnson’s SRE meta-analysis.

Although this account of SREs is influential, it is somewhat difficult to apply to the nonevaluative self-referential memory effects described above, as participants were not required to relate the incoming information to the rich, elaborative self-concept. This theoretical gap could be bridged by consideration of the importance of attention at encoding. Given the attention capture known to follow the perception of self-cues (as illustrated by, e.g., the cocktail party effect), this may facilitate resource-intensive processing in self-referential encoding contexts, such as owning objects and making outcome choices. This should elicit elaborate memory representations (i.e., enriched through the formation of semantic, pictorial, or affective associations) and increase subsequent recognition and recollection.

Supporting this reasoning, research has suggested that self-referential memories tend to be rich, elaborate representations, as is shown by participants’ ability to remember features of the encoding event (Conway & Dewhurst, 1995; Conway, Dewhurst, Pearson, & Sapute, 2001). Van den Bos, Cunningham, Conway, and Turk (2010) found that memory for items encoded in a context of self-ownership are particularly associated with episodic recollection, as opposed to familiarity. In this study, self-relevance was ascribed to stimuli through imagined, hypothetical ownership. The participants were required to sort items into “self-owned” and “other-owned” baskets on the basis of a color cue, before being given a surprise memory test in the form of a two-step remember–know (R–K) task. Participants were instructed to respond “remember” if they recognized an item from the encoding phase and had a specific recollection of having seen the item (typical of elaborative encoding; Gardiner, 2008). A “know” response was to be given if participants recognized the item from the encoding phase, but only on the basis of a strong feeling of familiarity. The results showed a self-reference effect (i.e., better memory for self-owned than for other-owned items) in “remember” responses only. This finding echoes previous demonstrations in trait-evaluation research that a self-reference effect emerges only in “remember” responses, which has led to a rebranding of the SRE as the “self-reference recollection effect” (SRRE; Conway & Dewhurst, 1995; Conway et al., 2001).

The link between elaboration at encoding and “remember” responses is intrinsic. Elaboration involves the binding of contextual and associative details of the encoding event (see Jacoby, 1991; Yonelinas, 2002). If participants retrieve such details at test, they are instructed to give a “remember” response, indicating episodic recollection. Van den Bos et al.’s (2010) demonstration of a self-reference effect in “remember” responses therefore suggests that self-ownership gives rise to elaborative memory representations.

The use of the R–K task to provide recollection and familiarity estimates has been the focus of some disagreement, with some researchers arguing that the two types of estimates reflect a difference in levels of confidence rather than qualitatively different memory processes (e.g., Dunn, 2004; Hockley, 2008; Wixted & Stretch, 2004). Nonetheless, in an extensive review, Yonelinas (2002) showed that the results from the R–K task are comparable to those from other methods used to dissociate recollection and familiarity, namely the process-dissociation procedure (Jacoby, 1991). As Yonelinas (2002) documented, “remember” and “know” responses typically produce different processing speeds, receiver operating characteristic patterns, and event-related potential responses, and can be dissociated after brain injury. Moreover, recent consistent findings from studies using the R–K task have supported the approach (Bruno & Rutherford, 2010; Markopoulos, Rutherford, Cairns & Green, 2010). In the present experiment, participants were extensively trained and asked to respond “know” only to recognized items that they were highly confident of having seen. In this way, we ensured that they treated the “remember” and “know” options as being qualitatively different, and not merely as representing high and low confidence, respectively. Therefore, we acknowledge that while the R–K task is not a perfect way of separating recollection from familiarity (i.e., like all tasks, it is not process pure), but it remains one of the best methods available (Gardiner, 2008). Importantly, it should be noted that in the present experiment, while we relate our work to that of dual-process studies, the primary function of the R–K test was to provide measures of participants’ phenomenological experience, in a way similar to the procedure of Conway and Dewhurst (1995).

The elaboration of incoming material tends to be an effortful process requiring attentional resources (Gardiner, Gregg, Mashru, & Thaman, 2001; Gardiner & Parkin, 1990; Yonelinas, 2001). R–K studies (e.g., Gardiner & Richardson-Klavehn, 2000) have found that dividing attention at study drastically reduces recognition accompanied by recollective experience (i.e., “remember” responses) but not recognition accompanied by strong feelings of familiarity (i.e., “know” responses). Elaborative self-referential memory representations (e.g., of self-owned objects) are therefore likely to depend on the application of attentional resources at encoding to a greater extent than do similar representations linked with other people. Van den Bos et al.’s (2010) ownership study employed full-attention conditions, so it was not possible to determine whether the ownership effect depended on the availability of attentional resources. The present inquiry will redress this limitation by manipulating attentional resource availability in order to determine whether limited resources have a selectively deleterious effect on self-referential memory biases.

In other words, we expect that self-relevant information may require resource-intensive processing and will therefore be affected by divided-attention (DA) manipulations. Therefore, we predict that only elaboration-based “remember” responses would be reduced under DA conditions, but not “know” responses.

The present inquiry

Participants were asked to sort items under conditions of imagined self- and other-ownership with full or divided attention, before completing an R–K recognition test. Under full-attention conditions, a standard ownership effect in memory (i.e., self-owned > other-owned item memory) was expected in “remember” responses. As memory for self-owned items is likely to be driven by attention-dependent elaborative encoding, DA should reduce or eliminate the ownership effect. Two levels of DA (easy and difficult) were employed to determine whether the self-memory bias is proportionately affected by resource availability.

Method

Participants and design

A group of 45 undergraduateFootnote 1 students (27 females, 18 males; mean age 21.9 years) from the University of Aberdeen took part in the experiment in return for course credit. All participants had normal or corrected-to-normal eyesight. The participants gave informed consent in accordance with the guidelines set by the University of Aberdeen’s Psychology Ethics Committee. A two-factor mixed design was employed, with one between-subjects factor (Encoding Condition: full attention, easy DA, difficult DA) and one repeated measures factor (Ownership: self-owned, other-owned).

Stimuli and apparatus

The stimulus set comprised 108 photographic images of grocery items (e.g., food and electrical items) adapted from online supermarket databases. The images (250 × 250 pixels/72 pixels per inch) were presented on a white background. The stimuli were divided into three equal sets (36 items) that were matched for item type, word length, and syllabic length. The use of these sets as self-owned targets, other-owned targets, and foils at recognition was counterbalanced across participants. The experiment was programmed using the E-Prime version 1.1 experimental software (Psychology Software Tools Inc., Sharpsburg, PA).

Procedure

Encoding phase

The participants were tested individually and were seated at a PC laptop and monitor. Participants were told that they were taking part in a shopping experiment and that they had to imagine that they and a fictitious other student (“John”) had each won their own basket of shopping items. They were then given instructions for the encoding phase (see Fig. 1 for a schematic representation of the tasks). A blank screen was presented with a shopping basket in each of the two bottom corners, one colored red and the other blue. Participants were informed that either the red or the blue basket was theirs (i.e., “self-owned”), and they were asked to imagine that everything that went into that basket belonged to them. The other basket, along with its contents, was designated as belonging to John (i.e., “other-owned”). The color of the self-owned basket and the onscreen locations of the red and blue baskets (bottom left or right) were counterbalanced across participants. In the encoding phase, a shopping item was presented in the center of the screen for 1,500 ms, after which a red- or blue-colored border appeared around the item and remained for a further 1,500 ms. Participants were instructed to use labeled buttons on the keyboard to assign the item to the red basket if the border was red, or to the blue basket if the border was blue. The next item was presented after an intertrial interval of 1,000 ms. The presentation order of the self-owned and other-owned items was randomized by the experimental program.

Fig. 1
figure 1

Schematic representation of the encoding and test phases

Full size image

Participants were also informed that they would be presented with a series of numbers onscreen during the ownership task. A number was presented beneath the shopping item for the duration of its 3,000-ms presentation. After every six trials, a number-related question was presented, with a response box in which participants could type their answer. All participants were presented with the same numbers but different questions, depending on the attention condition to which the participant had been assigned. In the difficult-DA condition, participants were prompted to report the preceding six digits in the order in which they had been presented. In the easy-DA condition, they were asked to report how many even numbers had been presented in the preceding six digits. In the control condition (full attention), they were asked to ignore the digits presented alongside the items, but instead to copy a three-digit number presented with the response box. All participants were told that it was very important that they perform well at both the sorting task and the digit task.

Test phase

At the start of the test phase, participants received instructions for responding to a two-step (old–new followed by remember–know–guess) recognition memory test (Gardiner & Richardson-Klavehn, 2000). Presentation of these instructions took 5 min on average, depending on the amount of explanation required by the participant. A total of 72 previously seen items and 36 unseen distractors were presented individually in the center of the screen in a random order. The items were presented for 2,000 ms, during which time a response had to be made. Participants were told to use labeled buttons on the keyboard to respond “yes” if they recognized the item from the encoding phase, and “no” if they did not. If a “yes” response was selected, they were asked to specify the basis for their response. If they could consciously recollect having seen the item and could retrieve any information about this event (e.g., they could remember what they had thought at the time), they were instructed to press “remember.” If their recognition was based purely on a feeling of knowing that the item had been presented, in the absence of being able to recollect any further details, participants were instructed to press the “know” button. Finally, if their “yes” response had been a complete guess, they were instructed to press “guess.” The experimenter checked whether the instructions were understood by asking participants to explain the differences between the three response options in their own words. She made sure that participants did not regard the “remember” and “know” response options as meaning “sure” and “unsure,” respectively. When the recognition test was completed, participants were debriefed and thanked for taking part.

Results

Manipulation check and scoring

Performance on the DA tasks was scored as follows: In the easy-DA task, participants had to indicate the number of even digits in the preceding sequence of six digits. At study, this occurred 12 times, yielding a score from 0 to 12. The participants’ performance accuracy was then converted into a proportion score. In the difficult-DA task, participants had to indicate the correct sequence of six preceding digits. This also occurred 12 times. For each trial, participants received a score from 0 to 6 (depending on how many digits they had put in the correct order), yielding a score from 0 to 72. These scores were then converted into proportion scores, as was done for easy DA.

A single-factor (Condition) between-subjects ANOVA applied to the DA performance data at study showed no significant difference in distractor task performance between the easy- and difficult-DA groups, F(1, 28) = 2.465, MSE = 0.44, p = .128, indicating no difference between performance on the easy-DA (M = .91, SD = .09) and difficult-DA (M = .83, SD = .17) tasks. While the difficult-DA task involved more complex digit monitoring than did the easy-DA task, the results demonstrated that equivalent performance was maintained across both DA tasks.

Memory data

Analysis of recognition accuracy data

Each participant’s recognition score was corrected for the baseline false-alarm rate by subtracting the proportion of “old” responses to foils (false alarms) from the proportion of “old” responses to previously presented items (see Table 1). These global recognition scores were submitted to a 2 (owner: self or other) × 3 (attention condition: control, easy DA, or difficult DA) mixed-design ANOVA, with the between-subjects factor Attention Condition. This analysis revealed no significant main effect of owner [F(1, 42) = 1.305, MSE = .008, p = .260]. However we did observe a significant main effect of attention condition [F(2, 42) = 5.650, MSE = .408, p = .007]. Post hoc pairwise comparisons revealed better memory for items encoded in the control condition (mean: .643) than in either the easy-DA (mean: .502, p = .049, two-tailed) or the difficult-DA (mean: .411, p = .002, two-tailed) condition. However, no significant difference in memory performance emerged between the two DA conditions (p = .198, two-tailed). We also observed a reliable Owner × Attention Condition interaction [F(2, 42) = 5.239, MSE = .031, p = .009]. Simple main effects analysis revealed a significant effect of owner in the control group [t(14) = 3.833, p = .002, two-tailed], with better memory for items owned by the self than for items encoded in relation to another. However, we found no significant effect of owner for either the easy-DA or the difficult-DA group (ps > .254).

Table 1 Mean uncorrected hit and false-alarm rates
Full size table

We also explored the impact of attention condition on ownership. For self-owned items, there was a significant decrease in memory performance in the easy-DA group as compared with the control group [t(28) = 2.530, p = .015, two-tailed] and between the difficult-DA group and the control group [t(28) = 4.063, p < .001, two-tailed]; however, there was no difference in memory performance for information encoded in relation to self across easy- and difficult-DA groups [t(28) = 1.533, p = .133, two-tailed]. For other-owned items, we observed no significant difference between the control and easy-DA groups [t(28) = 1.365, p = .179, two-tailed] or between the easy- and difficult-DA groups [t(28) = .979, p = .333, two-tailed]. However, a significant difference in memory did emerge between the control and difficult-DA groups [t(28) = 2.344, p = .024, two-tailed]. This pattern shows that self–other differences in global recognition accuracy disappear under DA conditions, and DA tasks appear to have a greater impact on self-referential encoding.

Analysis of remember/know data

In a paradigm with only two response options (“remember” and “know”), these responses are mutually exclusive: “Know” responses are instructed to be given only for items that have failed to trigger any recollective experience. It is not possible for participants to indicate a situation in which they experience remembering and knowing for a particular item. Therefore, “know” experiences are likely to be underestimated relative to “remember” experiences. When a “guess” category is also included, as in this experiment, the “remember” and “know” response options are technically no longer mutually exclusive. However, in the present experiments, participants did not often use the “guess” option (the overall guess rate was 4 %), suggesting that, in practice, the mutual exclusivity of “remember” and “know” responses may still be an issue. To solve this problem, we applied Yonelinas and Jacoby’s (1995) independence correction for “know” responses—namely,

$$ \mathrm{independent}\,\mathrm{know}={{\mathrm{know}} \left/ {{\left( {1-\mathrm{remember}} \right)}} \right.}. $$
(1)

These “independent-know” responses will be referred to as “I-know” responses below. False-alarm rates were then subtracted from the hit rates for each response type (“remember” and “I-know”) to correct for response bias. Note that there were no separate false-alarm rates per ownership condition.

“Remember” responses

The corrected remember hit rates were submitted to a two-factor (Attention Condition × Owner) mixed analysis of variance (ANOVA), which showed a main effect of owner, F(1, 42) = 4.327, MSE = .034, p = .044], with memory for self-owned items being characterized by more “remember” responses than items owned by others that were remembered. We also observed a main effect of attention condition [F(2, 42) = 4.100, MSE = .352, p = .024]: “Remember” responses were higher in the control condition than in the easy-DA condition, and higher for the easy-DA than for the difficult-DA task. However, these effects were complicated by a significant Owner × Attention Condition interaction [F(2, 42) = 6.206, MSE = .049, p = .004]. Single-factor (Attention Condition) ANOVAs confirmed that, as predicted, the proportion of “remember” responses for correctly identified self-owned items was significantly reduced by DA [F(2, 42) = 6.197, MSE = .322, p = .004]. The “remember” rates for difficult DA (mean .276) and easy DA (mean .391) were significantly lower than for the full-attention control condition (mean .567; ts = 3.495 and 2.115, ps = .001 and .040, respectively). The two DA tasks did not differ significantly from one another (p = .175). In contrast with the effect of attention on memory for self-owned items, divided attention did not affect the remembering of other-owned items [F(2, 42) = 1.888, MSE = .079, p = .164].

Furthermore, repeated measures (Ownership) ANOVAs revealed a significant ownership effect in the full-attention condition [F(1, 14) = 9.130, MSE = .093, p = .009] but no ownership effect in the easy-DA condition [F(1, 14) = 2.985, MSE = .220, p = .106], and a reversed but nonsignificant ownership effect in the difficult-DA condition [F(1, 14) = 2.805, MSE = .017, p = .167].

“I-know” responses

The corrected “I-know” hit rates were submitted to a two-factor (Attention × Ownership) mixed ANOVA, which showed no main effects of attention [F(2, 42) = 2.151, MSE = 0.143, p = .129] or ownership [F(1, 42) = 0.009, MSE < .001, p = .923], as well as no significant interaction between the two [F(2, 42) = 0.596, MSE = 0.010, p = .556].

Discussion

In the present inquiry, we explored the extent to which dividing attention at encoding impacts on the memory advantage usually associated with self-owned items (Cunningham et al., 2008; van den Bos et al., 2010). It was found that under full-attention conditions, an “ownership effect” (i.e., better memory for self-owned than for other-owned objects) emerged in “remember” responses, whereas no effect of ownership was observed in “know” responses. This pattern of responses in the full-attention condition replicated van den Bos et al.’s (2010) finding that ownership effects are observed in recognition accompanied by recollective experience, but not in recognition accompanied by feelings of “just knowing” (see also Conway & Dewhurst, 1995; Conway et al., 2001).

Given the effortful nature of elaborative encoding (Gardiner et al., 2001; Gardiner & Parkin, 1990; Yonelinas, 2001), it was expected that the advantage for self-referenced (i.e., self-owned) items would be diminished while participants were completing a divided-attention task at encoding. This pattern was found, with no ownership effect emerging under conditions of easy or difficult divided attention. This suggests that ownership effects only occur when sufficient attentional resources are available.

It seems plausible to conclude from this pattern that attentional resources are required when producing elaborative memory representations for self-owned items, which is not possible under conditions of serious resource depletion. This pattern of memory performance augments the evidence that self-referential encoding, relative to the encoding of material about other people, triggers the formation of a rich, elaborative memory representation (Klein & Loftus, 1988; Symons & Johnson, 1997), which is attentionally demanding. To our knowledge, the present study provides the first demonstration that the self-reference effect in memory is underpinned by differences in the attentional processing engaged at encoding.

Corresponding with the idea that self-processing is attention-resource-intensive, Turk, van Bussel, Brebner, et al. (2011) reported differences in electrophysiological responses in self- and other-owned trials. They found that ownership cues triggered an increase in the central midline P300 component associated with attention and cognitive processing, as well as a narrowing of spatial attention (P1) to the object location in occipital electrode sites. Similarly, Turk, van Bussel, Waiter, and Macrae (2011) reported fMRI data on ownership in which activation for self-owned objects appeared to start in caudal anterior cingulate cortex (ACC), a region defined as functionally important in modulating attention to salient objects (Carretié, Hinojosa, Martin-Loeches, Mercado, & Tapia, 2004; Chiu, Holmes & Pizzagalli, 2008). The activity in caudal ACC was significantly correlated with subsequent memory bias, indicating an important role for attentional processing on ownership effects in memory. The present study offers further support for this notion that elaborative self-referential encoding through ownership is attentionally demanding.

Interestingly, memory for other-owned items was unaffected by dividing attention at encoding, suggesting that relatively little elaboration of other-owned items takes place even under full-attention conditions. It is possible that with more power, a small effect of divided attention would also have been observed on memory for other-owned items, as some elaboration would be expected to support memory even in this condition. However, the interaction between attention and ownership was expected, because divided attention should have a larger effect on self-referential encoding, reducing the elaboration with which it is distinguished.

The novel demonstration in the present inquiry that attentional input at encoding is crucial for the elicitation of self-reference effects casts new light on the potential links between the ways in which the self impacts on cognition. In particular, the link between the reliance on attention for self-reference effects and the well-known attention-capturing effect of self-cues (e.g., the cocktail party effect; Moray, 1959; see also Bargh, 1982) is an interesting theoretical angle. The two effects could be causally related, as the attention recruited by self-cues could be the mechanism by which resource-dependent elaborative encoding and successful retrieval is initiated. Cues relating to other people, which do not attract the same degree of attention, could as a result fail to benefit from these memory-enhancing processes; thus, dividing attention has little effect on memory for other-relevant information. Alternatively, the two effects could simply operate in parallel; people may engage in elaborative encoding in a relatively deliberate way, because this is critical for ensuring that personally important information is not lost, without a reliance on the incidental attention-capturing effects of self-cues. While more work will be required to distinguish between these accounts, the finding that attentional resources underlie self-reference effects in memory provides a significant step toward understanding the mechanisms that drive the impact of self on cognition.

In conclusion, the present inquiry has shown that the processes that underlie the formation of elaborative memory representations in response to self-cues are attentionally demanding. While an ownership effect is elicited under full-attention conditions, this memory advantage in lost when attention is divided at encoding. Consistent with research on the attentional requirements of elaborative encoding in general, and with the patterns of brain activation elicited in ownership studies, the present findings offer a clear demonstration that the memory advantages associated with self-referential encoding are dependent on the availability of attentional resources.