Sleep and environmental context: interactive effects for memory
Sleep after learning is often beneficial for memory. Reinstating an environmental context that was present at learning during subsequent retrieval also leads to superior declarative memory performance. This study examined how post-learning sleep, relative to wakefulness, impacts upon context-dependent memory effects. Thirty-two participants encoded word lists in each of two rooms (contexts), which were different in terms of size, odour and background music. Immediately after learning and following a night of sleep or a day of wakefulness, memory for all previously studied words was tested using a category-cued recall task in room one or two alone. Accordingly, a comparison could be made between words retrieved in an environmental context which was the same as, or different to, that of the learning phase. Memory performance was assessed by the difference between the number of words remembered at immediate and delayed retrieval. A 2 × 2 × 2 mixed ANOVA revealed an interaction between retrieval context (same/different to learning) and retention interval (sleep/wakefulness), which was driven by superior memory after sleep than after wake when learning and retrieval took place in different environmental contexts. Our findings suggest a sleep-related reduction in the extent to which context impacts upon retrieval. As such, these data provide initial support for the possibility that sleep dependent processes may promote a decontextualisation of recently formed declarative representations.
KeywordsDeclarative memory Sleep Environmental context Hippocampus Neocortex
During judicial proceedings, witnesses are often taken back to the scene of the crime as this is thought to enhance their memory of previous events. This course of action can be attributed to the environmental context reinstatement effect (Smith 1979), a common psychological observation whereby memory performance is typically better in a context which is the same as that of learning. This effect has been repeatedly demonstrated in studies of declarative memory. These include manipulations of context in terms of the physical environment (Godden and Baddeley 1975), olfactory and auditory modalities (Smith 1985; Schab 1990; Parker et al. 2001) and a combination of these including room size, odour and background music (Parker et al. 2007).
The context reinstatement effect is thought to be supported by the hippocampus, which binds together spatially and temporally disparate features of a declarative experience to create an integrated associative memory representation (O’Reilly and Rudy 2001; Davachi et al. 2003). Research has suggested that the reactivation of any component of the bound associative representation, for example, the emotional context within which information was initially encoded, potentiates the whole representation, bringing all of its elements nearer to the threshold for retrieval (Lewis et al. 2005).
Like environmental context reinstatement, sleep has been shown to have an important impact upon declarative memory performance at retrieval (Born et al. 2006; Diekelmann et al. 2009). Considerable evidence has demonstrated that post-learning sleep, as compared to wakefulness, leads to a reduction in the decay of declarative memories (Takashima et al. 2006) and in some cases provides an active strengthening of such representations (Plihal and Born 1997). Other work has suggested that sleep may even mediate a form of mental restructuring, whereby explicit knowledge is extracted from previously implicit representations (Wagner et al. 2004; Fischer et al. 2006).
Despite a wealth of literature focusing on the mnemonic influences of contextual reinstatement and post-learning sleep, research investigating how these factors interact has been limited. As a consequence, little is known about how contextual elements of declarative representations are affected by sleep dependent consolidation processes.
Research into sleep and emotional memory may be able to shed some light on this issue. During the retrieval of emotional representations, a mnemonic trade off often occurs, whereby central aspects of emotional experiences are remembered at the expense of contextual details (Reisberg and Heuer 2004). By manipulating central (object) and contextual (background) aspects of pictures to create negative and neutral scenes, Payne et al. (2008) were able to investigate how such emotional trade offs were affected by sleep. Post-learning wakefulness resulted in the forgetting of entire negative scenes, with memory for objects and backgrounds becoming equally impaired. However, an equivalent period of sleep led to a decay of only the background elements, with negative object elements remaining intact. These findings indicate that central and contextual elements of declarative memory may be processed differently across sleep and wakefulness, with sleep providing an active protection for the most salient or focal aspects of a memory. Nevertheless, it is important to note that the sleep-related benefits for central aspects of memory reported by Payne and colleagues were related to negative items alone as sleep provided no benefit for the memory of either central or background components of neutral scenes. This suggests that focal components of declarative memory may be processed differently by sleep depending on their emotional valence.
The present study examined the differential impact of post-learning sleep or wakefulness upon the contextual reinstatement effect across three experimental sessions occurring in the morning (AM) or evening (PM). To investigate this, we used two environmental contexts (1 & 2), which were dissociable in terms of room size, odour and background music. Participants learned one list of category exemplar words in context 1 and another in context 2. These words were retrieved in either the same context as encoding or the other (different) context immediately after training and following a 12-h retention interval containing a day of wakefulness or a night of sleep. The retrieval sessions consisted of a category-cued recall task, which was selected due to its sensitivity to environmental context effects in declarative memory (Parker et al. 2007). In this task, participants are presented with a number of word categories, e.g. ‘fruits’ or ‘vehicles’, and asked to recall words that were presented during the learning phase and also correspond to the category in question.
This study therefore followed a 2 × 2 × 2 mixed ANOVA design with within subject factors of ‘retrieval context’ (same or different to learning) and ‘retention interval’ (sleep or wakefulness), and a between subjects factor of ‘session order’ (AM/PM/AM or PM/AM/PM). We hypothesised that consolidation processes occurring during post-learning sleep would be associated with a decay of contextual information within declarative memory thus diminishing the impact of environmental cues at delayed retrieval.
Materials and methods
Thirty-two healthy participants (five male, mean age 20.19, S.D. ± 1.89) took part in three experimental sessions after informed consent was obtained in accordance with the School of Psychological Sciences Research Ethics Committee, University of Manchester. Participants were informed that they were taking part in an experiment assessing the effects of music and odours upon memory performance. Psychology undergraduates took part in exchange for course credit, whereas other participants were paid £10. Participants had no prior history of psychiatric, neurological or sleep disorders and were required to be free of psychological drugs, alcohol and caffeine for 24 h preceding and throughout the study period. Each participant reported a normal sleep-wake cycle across the month preceding the study and agreed to refrain from daytime napping between experimental sessions.
Materials and apparatus
All stimuli were taken from a standardised set of word category norms (Van Overschelde et al. 2004), which provides a number of category names (i.e. vehicles) and exemplar words which correspond to such categories (i.e. car, train, boat). These category norms were compiled through the assistance of hundreds of American University students who were presented with 70 word categories sequentially and, for each, asked to generate as many exemplars as possible within 30 s.
Examples of the category exemplar words used in this study
A sport (7)
A four-footed animal (10)
Water polo (0.05)
An article of furniture (20)
An occupation or profession (24)
A further eight categories were selected with four exemplars taken from each to create thirty-two filler words. These categories were metals, weather, reading materials, clergy members, carpenters tools, human dwellings, natural earth formations and diseases. Four filler words were placed at the beginning and another four at the end of each target word list for respective use as primacy and recency buffers.
To control for order effects, the four lists of target words were each placed in two pseudorandom orders, with the only constraint being that no two words from the same category could appear in sequence. The primacy and recency buffers were also re-ordered in this way. Each of the four word lists, complete with target and filler words, therefore had two different orders (i.e. A1 and A2, B1 and B2, C1 and C2, D1 and D2).
Within the normative data collected by Van Overschelde et al., a ‘response proportion’ is provided for each category exemplar. This represents the proportion (min = 0.00, max = 1.00) of American students who, in the data collected by Van Overschelde et al., generated that particular word for the category in question. Additionally, word categories are ranked by their ‘category potency’. This measure ranks the word categories in terms of the number of exemplars that were generated for them; thus, a high rank indicates that a larger number of exemplars were generated for the category in question.
There was no significant difference between the mean response proportions of each word list (list A: 0.076 [S.D. ± 0.022], list B: 0.070 [S.D. ± 0.023], list C: 0.073 [S.D. ± 0.020], list D: 0.073 [S.D. ± 0.020], P = 0.83) and no significant difference between mean category potency rankings of each word list (list A: 21.67 [S.D. ± 18.80], list B: 24.17 [S.D. ± 15.69], list C: 25.67 [S.D. ± 13.40], list D: 23.83 [S.D. ± 13.48], P = 0.98).
Primacy and recency buffers were also balanced for response proportion (primacy buffers mean: 0.0806 [S.D. ± 0.027], recency buffers mean: 0.0813 [S.D. ± 0.021], P = 0.94) and category potency ranking (primacy buffers mean: 47 [S.D. ± 10.10], recency buffers mean: 50 [S.D. ± 19.41], P = 0.79).
Examples of the word categories and associated exemplars used in this study, along with their respective category potency and response proportion ratings, are shown in Table 1.
Declarative memory task
The encoding task was created using cogent 2000 developed by the cogent 2000 team at the functional imaging laboratory (FIL), University College London. The tasks were written and implemented using MATLAB© version 6.5 on a desktop PC with a dual core Xeon processor. Word lists were presented on a 20′′ computer screen in a large white font and black background. A fixation crosshair preceded the first word for 1,500 ms. Each word remained on the screen for a period of 2,000 ms, with a fixation crosshair placed in between for 1,500 ms. Using the corresponding numbers on the computer keyboard, participants were instructed to provide a pleasantness rating on a scale of 1 (very unpleasant) to 5 (very pleasant) for each word as it was presented. Participants were not informed that their memory for the words would be tested. In any experimental session, participants were presented with two lists, either A and B, or C and D.
A category-cued recall task was used to assess memory performance for the target words. Participants were presented with a response booklet that contained twelve word categories and space beneath each for four words. The categories corresponded to the two target word lists seen by participants during the previous encoding phase. Accordingly, participants were required to write down words that were examples of the category heading, and had also been presented in the previous learning phase, within a 4 min time limit. Participants were instructed to avoid guessing and only write down a word if they were quite certain it had appeared previously.
Odours were diffused using an electronic aerosol dispenser (Xanadu© Fragrance System, model: WF-0322). Two distinct fragrances were used (Xanadu fragrances ‘rose garden’ and ‘citrus lemon’). Fragrances were dispensed every 2 min with digitally timed actuations, thus ensuring that the odours were intense enough to be noticed by participants. A portable CD player was used to play the background music. The music was played quietly so that the sound would not impair participants’ concentration during encoding and retrieval. Subjective ratings of odour and background music pleasantness were not assessed in this study.
Room descriptions: environmental contexts
Context 1 was a long, spacious room and cream in colour. It contained computer equipment and a number of desks, chairs and cabinets. The odour was rose, and the background music was Beethoven’s 5th Symphony.
Context 2 was a small room on a different floor of the building to context 1. It contained only one desk and chair and was also cream in colour. The odour was lemon, while the background music was a modern dance song entitled ‘Connected’ by Paul Van Dyk.
An experimental 2 × 2 × 2 mixed ANOVA design was used. Within subject factors were ‘retrieval context’ (same or different to learning) and ‘retention interval’ (sleep or wakefulness), while the between subjects factor was ‘session order’ (AM/PM/AM or PM/AM/PM). The dependent variable was the difference between the number of words remembered at immediate and delayed retrieval sessions (delayed retrieval–immediate retrieval) (Payne et al. 2008).
Participants began an encoding phase with one of the word lists (A or B, C or D) in one of the two study rooms (context 1 or 2). In order to reduce word rehearsal, the encoding phase was followed by a 3-min distraction task of mathematical addition problems. Participants were then led to the other study room for a second encoding phase which followed the same procedures as the first, but with a different word list. Following a second distraction task, again to reduce word rehearsal, participants either remained in the same room or were taken back to the room where they began the session for category-cued recall of both word lists (immediate retrieval). Participants were then instructed to leave and carry on with their usual daily activities (AM/PM/AM group) or return home and sleep normally (PM/AM/PM group).
Twelve hours later, participants were taken back to the room in which they completed category-cued recall in session one and asked to complete the same task again (delayed retrieval). After completing the 3-min distraction task, this time to reduce interference between recalled words and those learned in the subsequent encoding phase, participants were then moved to the other study room. Here, they began the second encoding phase, which followed the same procedures as session one, but with a different pair of word lists. Accordingly, if participants had seen word lists A and B in the first session, they were presented with lists C and D in the second session and vice versa. After another distraction task, participants began a second category-cued recall task (immediate retrieval) in either room (context) 1 or 2.
After another 12 h, participants were brought back to the room where the second category-cued recall task had taken place in session two and were asked to complete it again (delayed retrieval).
If participants had slept between session one and two, they remained awake between session two and three, and vice versa. This therefore created a within subjects comparison of the interaction between retention intervals of post-learning sleep or wakefulness and environmental context reinstatement or replacement at retrieval, allowing examination of how this affects declarative memory performance.
Following a retention interval of sleep, participants were asked to fill out a sleep questionnaire before undertaking the delayed retrieval task. This included a question on approximate times of sleep the night before and wakefulness in the morning, which enabled us to deduce the number of hours sleep they had attained across the preceding night. Other questions included average duration of nocturnal sleep (hours), whether their sleep patterns had remained relatively constant across the preceding month and finally if they had, at any point in their life, suffered from any form of sleep disturbance or disorder.
No participant reported an inconsistent sleep pattern over the month preceding the study, or any history of sleep disturbance or disorder at any point in their life. There was no significant difference between the groups in terms of mean hours slept during the sleep retention interval (AM/PM/AM group: 7.42 [S.D. ± 1.06], PM/AM/PM group: 6.58 [S.D. ± 1.68], P > 0.1) or the average duration of nocturnal sleep (hours) (AM/PM/AM group: 7.44 [S.D. ± 1.03], PM/AM/PM group: 6.94 [S.D. ± 1.29], P > 0.2).
Immediate and delayed retrieval
Mean category-cued recall scores for the number of words retrieved (24 maximum) in the same or different context to learning at immediate (before sleep or wake) and delayed (after sleep or wake) retrieval
Environmental context at retrieval
9.84 (S.D. ± 4.54)
9.13 (S.D. ± 3.92)
9.94 (S.D. ± 3.88)
10.31 (S.D. ± 3.13)
9.75 (S.D. ± 4.36)
9.28 (S.D. ± 3.90)
10.19 (S.D. ± 3.69)
9.78 (S.D. ± 3.30)
Context reinstatement effects after sleep or wakefulness
Difference in the mean number of words remembered (delayed retrieval–immediate retrieval) for category-cued recall in each study condition
Environmental context at retrieval
−0.09 (S.D. ± 1.15)
0.16 (S.D. ± 1.27)
0.25 (S.D. ± 1.37)
−0.53 (S.D. ± 1.37)
Our data were analysed using a 2 × 2 × 2 mixed ANOVA with ‘retention interval’ (sleep or wakefulness) and ‘retrieval context’ (same or different to learning) as within subject factors and ‘session order’ (AM/PM/AM or PM/AM/PM) as the between subjects factor. This showed no main effect of retrieval context F(1, 30) = 1.01, P > 0.3 or retention interval F(1, 30) = 0.78, P > 0.3. However, a significant interaction was found between these factors F(1, 30) = 6.95, P = 0.013.
No main effect was found for session order F(1,30) = 0.003, P > 0.9, and no significant interaction was found between this factor and either retention interval F(1, 30) = 1.88, P > 0.1 or retrieval context F(1, 30) = 0.64, P > 0.4. There was also no significant interaction between all three factors F(1, 30) = 0.006, P > 0.9.
The conditions of retrieval (retention interval and retrieval context) were further examined using paired samples t tests. These showed that significantly more words were remembered following a period of sleep, as compared to a period of wakefulness, when the context at retrieval differed from that at encoding t(1,31) = 2.73, P = 0.01. However, when the retrieval context was the same as that of encoding, no significant difference was found in memory score between conditions of post-learning sleep or wakefulness t(1,31) = −1.16, P > 0.2 (see Fig. 2).
Overall, our findings suggest that the extent to which a change in environmental context impacted upon retrieval success was reduced after consolidation across sleep, but not after consolidation across an equivalent period of wakefulness.
Environmental context effects
We examined classical context-dependent memory effects within category-cued recall performance. A paired samples t test showed that, following a retention interval of wakefulness, significantly more words were remembered in the same retrieval context (as compared to a different retrieval context) t(1,31) = −2.10, P = 0.044 (see Fig. 2). This pattern is indicative of a classical context-dependent memory effect, but was only observed after retention across wakefulness, and not after retention across sleep t(1,31) = −1.00, P > 0.3.
Encoding based interference
Unlike the first session, the encoding phase of session two was preceded by a delayed retrieval test. Despite the use of a distraction task before this second encoding session, it is important to consider the possibility that recalling words so soon before encoding could have induced pro-active interference upon new word learning in session two. Such interference would be reflected by a poorer performance in the subsequent immediate recall test, regardless of whether the word’s retrieval context was the same or different to that of learning. To test this, we conducted a 2 × 2 repeated measures ANOVA on performance in the immediate retrieval session with ‘session’ (one or two) and ‘retrieval context’ (same or different to learning) as factors. A main effect of session F(1, 31) = 8.08, P < 0.01 showed superior memory performance in the immediate retrieval test of session two, relative to that of session one, thus indicating that the results of this study were not confounded by pro-active interference. There was no main effect of context F(1, 31) = 0.11, P > 0.7 and no significant interaction between factors F(1, 31) = 0.006, P > 0.9.
Retrieval expectancy and sleep
Participants were not explicitly informed that their memory for the category words would be tested. However, it is likely that their observations of our study methods during the first two experimental sessions allowed them to predict the occurrence of subsequent retrieval tests, resulting in a higher rate of learning and superior memory performance during these tests. To assess whether retrieval test expectancy may have modulated subsequent sleep dependent consolidation processes, we compared the post-sleep delayed retrieval performance of the PM/AM/PM group to that of the AM/PM/AM group, where post-learning sleep occurred following the first and second encoding phases, respectively. A 2 × 2 mixed ANOVA with between subjects factor ‘post-sleep retrieval session’ (PM/AM/PM session 2 or AM/PM/AM session 3) and within subjects factor ‘retrieval context’ (same or different to learning) showed that participants in the AM/PM/AM group (i.e. those who were likely to expect the retrieval tests) performed significantly better at post-sleep delayed retrieval F(1, 30) = 6.79, P < 0.01, suggesting that retrieval test expectancy influenced mnemonic processes occurring over the sleep retention interval. There was, however, no main effect of retrieval context F(1, 30) = 0.32, P > 0.5 and no interaction between factors F(1, 30) = 0.001, P > 0.9.
Circadian influences at encoding and retrieval
Since experimental sessions taking place before or after retention intervals of diurnal wakefulness or nocturnal sleep were conducted at different times of day (e.g. in the morning after sleep, and in the evening after wake), it is possible that our results were influenced by circadian factors at either encoding or retrieval. To test for this possibility, we conducted a 2 × 2 repeated measures ANOVA with factors ‘time of day’ (morning or evening) and ‘retrieval context’ (same or different to learning) for words remembered at immediate retrieval. This showed no main effect of either time of day F(1, 31) = 1.42, P > 0.2 or retrieval context F(1, 31) = 0.11, P > 0.7 and no significant interaction between these factors F(1, 31) = 0.79, P > 0.3.
The same ANOVA was conducted for words remembered at delayed retrieval. This also showed no main effect of time of day F(1, 31) = 1.02, P > 0.3 or retrieval context F(1, 31) = 0.70, P > 0.4 and no significant interaction between these factors F(1, 31) = 0.003, P > 0.9. These results indicate that the differences in performance which we observed after retention across intervals of wakefulness and sleep were not due to circadian variations at either encoding or retrieval.
In this study, we investigated the influences of post-learning sleep, relative to wakefulness, upon environmental context effects in declarative memory. Our results show a significant interaction, with superior memory performance after nocturnal sleep relative to daytime wakefulness, but only when the environmental context of retrieval was different to that of learning. In other words, a change in environmental context impaired memory performance when learning was followed by a retention interval of wakefulness but not when it was followed by a retention interval of sleep. In line with our hypothesis, these findings are indicative of a sleep-related reduction in the extent to which contextual cues impact upon retrieval. Our data therefore provide initial support for the suggestion that declarative representations may become actively decontextualised during post-learning sleep.
We found no difference between memory performance in the same context following retention intervals containing sleep or wakefulness. A number of studies that, presumably did not include manipulations of environmental context between learning and retrieval, have demonstrated superior memory performance following sleep relative to wakefulness with verbal declarative memory tasks similar to the one used in this study (Tucker et al. 2006; Payne et al. 2009). Why is it that a comparable effect was not found in this experiment? Firstly, the category-cued recall task was chosen because previous research has demonstrated its sensitivity to environmental context effects in declarative memory (Parker et al. 2007). To our knowledge, this task has not been used in sleep and memory research before; thus, it is possible that sleep does not promote verbal declarative memory consolidation under these conditions. More specifically, however, a lack of sleep-dependent memory effects in the same context condition may actually pertain to the incidental nature of our encoding task. A recent study demonstrated that the beneficial impact of sleep for memory only occurred when participants were informed that their memory for the learned materials would be subsequently tested (Wilhelm et al. 2011). In this study, participants were not explicitly informed that their memory would be tested after learning, which may have weakened potential sleep-related memory improvements. Moreover, we reasoned that by the second encoding phase, participants would have had sufficient time to observe our study methods and potentially anticipate the occurrence of subsequent retrieval tests. This is supported by the observation that post-sleep retrieval performance was superior in the AM/PM/AM group (who had slept after the second encoding phase) when compared with the PM/AM/PM group (who had slept after the first encoding phase). Nevertheless, while this difference may offer one possible explanation of our findings, it is still not completely clear why we failed to reveal a sleep dependent memory effect in the same context condition.
Decontextualisation: an active process?
In terms of elucidating a potential mechanism for an active, sleep-dependent decontextualisation of declarative memory, we can turn to the standard model of consolidation (Marr 1971; McClelland et al. 1995). This model proposes that declarative representations are initially encoded within both the limited capacity hippocampus and the long-term neocortical store. However, in order to bring a coherent memory trace to consciousness, the hippocampus is thought to integrate or ‘bind’ highly distributed neocortical memory components, which represent various aspects of an experience.
Over time, a progressive strengthening of cortico-cortical connections and concurrent weakening of hippo-cortical connections are believed to enable declarative representations to become independent of the hippocampus and integrated within the neocortical store, which forms the locus of their retrieval. Accordingly, the hippocampus gradually regains a refreshed encoding capacity, which allows for additional learning to take place. Adaptations of the standard model suggest that sleep may be involved in this process (Diekelmann and Born 2010). Specifically, experience-dependent patterns of activity are believed to be reactivated in this hippocampal-neocortical memory network during the slow wave sleep (SWS) stage, and this is thought to facilitate a shift in retrieval dependency from the hippocampus to the neocortex.
Contextual information in both spatial and temporal modalities has been shown to be highly reliant on the hippocampus after initial learning (Davachi and Wagner 2002; Davachi et al. 2003). Two fundamental assumptions of the standard consolidation model relate to the limited capacity of the hippocampus and the temporary manner in which it encodes episodic experiences. Accordingly, if contextual components of declarative memory remain entirely hippocampal and this region forms only a short-term store, it is possible that this information may be lost as hippo-cortical connections are weakened and retrieval becomes entirely dependent upon the neocortex. It is noteworthy, however, that this proposal may be difficult to reconcile with the multiple trace hypothesis, which states that episodic memories never become totally independent of the hippocampus (Nadel and Moscovitch 1997; Moscovitch and Nadel 1998).
Under the standard consolidation model, our findings may therefore suggest that declarative memory components that relate to the environmental context of the original learning episode are lost during SWS. Future research should include overnight sleep recording with polysomnography (PSG) to explore the possibility that a potential sleep-dependent decontextualisation of declarative memory may take place exclusively during SWS.
Classical context dependent memory effects
As described earlier, a number of studies have demonstrated context-dependent memory effects across a range of modalities (Godden and Baddeley 1975; Parker et al. 2001). While this study focused primarily upon the interaction of sleep and environmental context effects for declarative memory, we also found a classical context-dependent memory effect. Following a retention interval of daytime wakefulness, fewer words were remembered in a different context than in the same context to learning. Accordingly, the results of this study demonstrate both a classical context-related memory impairment after wakefulness and a sleep-related reduction in the impact of contextual cues upon recall.
Circadian influences at recall and retention
Research has demonstrated that a number of processes thought to underlie memory, such as synaptic plasticity, can be influenced by the time of day effects (Gerstner and Yin 2010). Since our experimental sessions were separated by 12 h and therefore took place in the morning or evening, it was important to assess the potential influence of circadian variances within our data.
While we did not assess participant vigilance levels or retrieval function at any stage of the experiment, we did compare participant retrieval performance during testing in the morning to performance during equivalent testing in the evening. We also tested for an interaction between the diurnal time at which subjects performed the task and the impact of retrieval context (same or different to learning) upon memory. These analyses indicated that the time of learning or testing had no influence upon encoding, retrieval or the impact of environmental context upon retrieval.
In addition to encoding and retrieval, circadian factors could also influence off-line consolidation processes. For example, it is possible that memory consolidation may occur more efficiently at night than during the day, irrespective of whether or not sleep occurs. Because our study design did not control for such circadian influences, we cannot rule these out as an alternative explanation for our data. Future work could control this better by including experimental conditions of nocturnal wakefulness and day-time sleep.
Inter-session performance differences at immediate retrieval
In the second experimental session, the encoding phase was preceded by a delayed retrieval test, which encompassed the words learned during session one. Despite the use of a mathematical distraction task, it is possible that word recall in this delayed test may have lead to a proactive interference of words encoded just a few minutes afterwards, thus negating immediate retrieval performance for these words. Interestingly, our results demonstrated quite the opposite effect, with participants performing significantly better on the immediate retrieval test of session two, relative to that of session one.
Previous research has shown that memory formation is enhanced when encoded information is deemed to be relevant for future plans (Szpunar et al. 2007). As mentioned earlier, participants were not explicitly informed that their memory would be tested immediately after encoding. It is therefore unlikely that participants would have anticipated an immediate memory test during the first learning session. However, since participants should have gained an understanding of the study methods by the beginning of the second encoding phase, it is likely that they would have predicted the occurrence of an immediate retrieval test in session two.
With the knowledge of these upcoming tests, participants may have, either implicitly or explicitly, directed a greater degree of cognitive processing resources towards category word learning during the second encoding phase, leading to superior immediate retrieval performance during the latter part of the study.
Although these results demonstrate a significant difference in immediate retrieval performance between sessions one and two, this effect was not modulated by the environmental context of retrieval and could not therefore have confounded the main results of this study.
We intended to assess how a retention interval of sleep, relative to wakefulness, would affect the change in context-dependent memory effects between immediate and delayed retrieval. However, since category words may have been re-encoded during the immediate retrieval phase, our findings are open to alternative interpretation. In the case of words retrieved in an incongruent context, such re-encoding may have lead to the formation of weak connections between the new (‘different’) context and the recalled word. Because previous research has shown that sleep preferentially consolidates weaker memory traces (Drosopoulos et al. 2007), this contextual association may have been selectively strengthened over sleep, resulting in a better delayed retrieval performance in the different context after sleep relative to wakefulness. Although plausible, this explanation requires the somewhat problematic assumption that contextual associations formed during re-encoding would be weaker (and therefore more likely to consolidate across sleep) than the initial trace formed during encoding, even though these re-encoded associations are formed after the initial traces. Furthermore, although this scenario could potentially explain why participants performed better in the ‘different’ context after sleep than after wake, it fails to explain why participants who re-encoded in the ‘different’ context subsequently forgot so many words when they were retested in this context after wakefulness. Despite these concerns, this alternative interpretation cannot be fully ruled out, and future work should attempt to resolve this issue by adopting a methodological framework that does not allow for re-encoding to take place.
This work was funded by the Engineering and Physical Sciences Research Council (EPSRC) and Unilever. PAL and SJD were supported by a Biotechnology and Biological Sciences Research Council (BBSRC) New Investigator Award to PAL.
Conflict of interest
The authors declare that they have no conflict of interest.
- Reisberg D, Heuer F (2004) Memory for emotional events. In: Reisberg D, Hertel P (eds) Emotion and memory. Oxford University Press, Oxford, pp 3–41Google Scholar