Brain oscillations track the formation of episodic memories in the real world
Graphical abstract
Introduction
Episodic memory refers to rich memories of personally experienced events. The details of these memories not only encompass the event itself but also the surrounding environmental setting, such as where and when the event occurred. Environmental context change can have a profound effect on episodic memory (Godden and Baddeley, 1975, Smith and Vela, 2001). Yet despite such context change being typical in day-to-day life, these changes are rarely incorporated in neuroscientific experiments examining episodic memory (often due to the need to conduct these experiments in magnetic resonance imaging [MRI] or magnetoencephalogram [MEG] suites). In these experiments, it is possible that mechanisms relating to the encoding of environmental context are supressed, as context remains largely consistent and therefore irrelevant to the task. This means that the neural correlates of episodic memory observed in the lab may provide an incomplete picture of episodic memory in the real world. While it is impossible to implement MEG or MRI in daily-life settings, progress has been made in the use of portable EEG outdoors (De Vos et al., 2014, Debener et al., 2012). Embracing these advances, we aimed to investigate the influence of vibrant real world environments on the electrophysiological correlates of episodic memory formation.
One of the most common approaches to studying episodic memory formation is the subsequent memory effect (SME). SMEs are the neural signature of successful memory formation, calculated by contrasting the neural activity at encoding which predicts later remembering with the activity that predicts later forgetting, hence isolating the activity unique to memory formation. Oscillatory SMEs are in part characterised by alpha and beta (8–12 Hz; 13–30 Hz) power decreases (Burke et al., 2015a, Fellner et al., 2013, Greenberg et al., 2015, Guderian et al., 2009, Hanslmayr et al., 2009, Meeuwissen et al., 2011, Noh et al., 2014, Weiss and Rappelsberger, 2000). Additionally, theta (3-7 Hz) has often been implicated in memory formation, although discrepancies exist in the literature with both theta power increases and decreases purported to underlie successful memory formation (Burke et al., 2015a, Burke et al., 2013, Fell et al., 2011, Guderian et al., 2009, Merkow et al., 2014, Noh et al., 2014, Nyhus and Curran, 2010, Staudigl and Hanslmayr, 2013). Nevertheless, beta power (13–20 Hz) decreases have been shown to reliably arise over task-relevant sensory regions during successful memory formation, a result attributed to information processing (Hanslmayr et al., 2012). Critically, a recent EEG-repetitive transcranial magnetic stimulation (rTMS) study has demonstrated that beta power decreases are causally relevant to this process (Hanslmayr et al., 2014). The predictability of these beta power decreases provide a reliable benchmark to contrast with real world recordings in order to identify whether the typical lab-based SME is observable in a real world environment.
Beyond the validation of previous lab-based findings, portable EEG technology allows the investigation of aspects of episodic memory that only occur in their entirety in the real world, such as item-to-context binding. Item-to-context binding can be assessed via contextual clustering - a behavioural phenomenon in which several events are recalled together based on contextual similarities they share. Contextual clustering has often been demonstrated for events which share a similar temporal context (i.e. events that occurred at similar points in time; Howard and Kahana, 2002). However, contextual clustering is not solely restricted to the time domain (e.g. Long et al., 2015; Polyn et al., 2009). Of particular relevance here, studies have also demonstrated spatial contextual clustering where events that occurred in similar locations are recalled together (Copara et al., 2014; Miller et al., 2013a). To date, this phenomenon is predominantly studied in virtual reality where participants navigate low-resolution environments with limited visuospatial information. Vestibular and locomotion cues are distinctly lacking in many virtual reality experiments, yet lesion studies in rats have shown that these cues have been shown to have a profound impact on spatial navigation (Stackman and Herbert, 2002, Wallace et al., 2002). The absence of such cues may impede the development of a comprehensive spatial contextual representation.
It is also worth noting that a number of studies investigating spatial context have relied on random travel patterns to dissociate spatial and temporal contextual effects. A large number of random trajectories would inevitably mean that spatial and temporal context incidentally coincide at various points during the experiment, introducing a confounding variable and potentially trivial explanation of spatial clustering. In our experiment, we aimed to overcome this issue by using novel navigational paths that allow the observation of the independent contributions of temporal and spatial context to episodic memory formation.
On an oscillatory level, Long and Kahana (2015) demonstrated that temporal clustering correlates with gamma power increases in the left inferior frontal gyrus and the hippocampus during encoding. However, to the best of our knowledge, no other experiment has further investigated the relationship between neural oscillations at encoding and contextual clustering. Therefore, it remains unknown whether these patterns of activation are unique to subsequent temporal clustering or part of a more general associative mechanism. A priori assumptions follow that subsequent temporal and spatial clustering would encompass the medial temporal lobe (MTL) – the home of place and time cells (Eichenbaum, 2014, MacDonald et al., 2011, O’Keefe, 1976). Given the intimate relationship between place cells and theta band activity, it may also be plausible to suggest that the spatial clustering effect would be observable within the theta frequency (Burgess and O’Keefe, 2011, O’Keefe and Recce, 1993). In support of this, a recent virtual reality study found that subsequent memory for spatial contextual location was predicted by 'slow' theta power decreases (2-5Hz; Crespo-García et al., 2016)
It is of course important to identify potential oscillatory confounds that may arise in ‘real world’ paradigms that are not present in lab-based experiments. Numerous mobile brain body imaging (MoBI; Makeig et al., 2009) studies have demonstrated that both event-related potentials (ERPs) and oscillatory activity can be observed in moving participants (De Sanctis et al., 2014, Gramann et al., 2010, Gwin et al., 2010, Malcolm et al., 2015, Wagner et al., 2014). However, in relation to oscillatory activity, movement-related changes in power changes across the frequency spectrum (~1 Hz to 90 Hz). More specifically, in comparison to standing, walking can produce alpha/beta band power suppression and gamma power increases in sensorimotor areas (Castermans et al., 2014, Seeber et al., 2015, Seeber et al., 2014, Wagner et al., 2016, Wagner et al., 2012), whilst a loss of balance has been linked to an increase in theta band activity (Sipp et al., 2013). Importantly, these latter findings share spectral similarities with the SME. Therefore, in order to avoid potential contamination of these effects, the EEG data obtained is this experiment was acquired solely while participants were stationary.
In this experiment, we asked two questions; (1) can oscillatory lab-based episodic memory studies be validated in real-life settings? and (2) what are the neural correlates of temporal and spatial contextual clustering? Following a predefined route and led by the experimenter (see Fig. 1a and b), participants were presented with words to encode and associate with their current location (see Fig. 1c), a situation similar to remembering several text messages on the way to the supermarket. Participants were shown 4 lists of 20 words, where each list was presented on a spiralling route (see Fig. 1a). These spiralling routes were used to help disentangle the relationship between temporal and spatial context (see methods for details). After being shown a list of words, participants were removed from the environment and completed a free recall test. Finally, participants guided the experimenter to where they thought each recalled word was shown and the location was marked by GPS. We aimed to replicate the well-documented low to mid frequency power decreases (<30 Hz) in lab-based subsequent memory studies (e.g. Burke et al., 2015b; Hanslmayr and Staudigl, 2014), in particular the beta power decreases over the left inferior frontal gyrus elicited by verbal SME paradigms (Hanslmayr et al., 2011). Furthermore, we aimed to identify and dissociate the neural correlates of spatial and temporal contextual encoding. To this end, we contrasted neural activity associated with subsequent temporal clustering with that of subsequent spatial clustering. In short, this is the first experiment directly observing the neural correlates of episodic memory encoding in the real world, allowing both the validation of a large body of the episodic memory literature and the identification of how real world context affects the neural correlates of encoding.
Section snippets
Participants
29 University of Birmingham students (18–39 years, 69% female) were recruited through a participant pool and rewarded with financial compensation for participation. Nine participants were excluded from the sample due to issues in recording leading to insufficient trials (n=4), poor weather conditions leading to insufficient trials (n=2) or extreme performance in the task (recalled <15 items, or forgot <15 items across all blocks; n=3). Recording complications meant that one block was lost for 3
Behavioural results
On average, participants recalled 50.45% of each 20 word list and when attempting to locate where each word was presented, were on average 14.74 m away from the presentation location. Eighty percent of participants showed less temporal contextual error (i.e. more temporal contextual clustering) than spatial contextual error (see Fig. 3). A one-sample t-test revealed significantly greater spatial clustering than expected by chance, t(19)=−5.728, p<0.001, 95% CI [−2.155, −1.001], matching previous
Discussion
Here, we identified the oscillatory subsequent memory effect (SME) in a real-world environment. Moreover, we examined the influence of real world contextual factors (i.e. space) on episodic memory relative to contextual factors available within the lab (i.e. time). Participants donned a portable EEG setup and were presented with verbal stimuli on a tablet across the university campus. Each list was presented on a spiral path that disentangled temporal and spatial context. Successful memory
Acknowledgements
This work was supported by grants awarded to S.H. by Deutsche Forschungsgemeinschaft [Emmy Noether Programme Grant HA 5622/1-1]; and the European Research Council [Consolidator Grant Agreement 647954].
References (77)
- et al.
Models of place and grid cell firing and theta rhythmicity
Curr. Opin. Neurobiol.
(2011) - et al.
Human intracranial high-frequency activity maps episodic memory formation in space and time
Neuroimage
(2014) - et al.
Human intracranial high-frequency activity during memory processing: Neural oscillations or stochastic volatility?
Curr. Opin. Neurobiol.
(2015) - et al.
About the cortical origin of the low-delta and high-gamma rhythms observed in EEG signals during treadmill walking
Neurosci. Lett.
(2014) Effects of time lag and frequency matching on phase-based connectivity
J. Neurosci. Methods
(2015)- et al.
Recalibration of inhibitory control systems during walking-related dual-task interference: A Mobile Brain-Body Imaging (MOBI) Study
Neuroimage
(2014) - et al.
Towards a truly mobile auditory brain-computer interface: exploring the P300 to take away
Int. J. Psychophysiol.
(2014) - et al.
Brain oscillatory subsequent memory effects differ in power and long-range synchronization between semantic and survival processing
Neuroimage
(2013) - et al.
Tracking the mind's image in the brain I: time-resolved fMRI during visuospatial mental imagery
Neuron
(2002) - et al.
Decreases in theta and increases in high frequency activity underlie associative memory encoding
Neuroimage
(2015)
Entrainment of prefrontal beta oscillations induces an endogenous echo and impairs memory formation
Curr. Biol.
How brain oscillations form memories--a processing based perspective on oscillatory subsequent memory effects
Neuroimage
A distributed representation of temporal context
J. Math. Psychol.
Measuring directionality between neuronal oscillations of different frequencies
Neuroimage
Successful memory formation is driven by contextual encoding in the core memory network
Neuroimage
Hippocampal “time cells” bridge the gap in memory for discontiguous events
Neuron
Linking brain, mind and behavior
Int. J. Psychophysiol.
The aging brain shows less flexible reallocation of cognitive resources during dual-task walking: a mobile brain/body imaging (MoBI) study
Neuroimage
Diverse phase relations among neuronal rhythms and their potential function
Trends Neurosci.
Nonparametric statistical testing of EEG- and MEG-data
J. Neurosci. Methods
Using single-trial EEG to predict and analyze subsequent memory
Neuroimage
Functional role of gamma and theta oscillations in episodic memory
Neurosci. Biobehav. Rev.
Place units in the hippocampus of the freely moving rat
Exp. Neurol.
Motor imagery activates primary sensorimotor area in humans
Neurosci. Lett.
The role of the anterior temporal lobes in the comprehension of concrete and abstract words: rTMS evidence
Cortex
Memory search and the neural representation of context
Trends Cogn. Sci.
Tracking the mind's image in the brain II: Transcranial magnetic stimulation reveals parietal asymmetry in visuospatial imagery
Neuron
Synchronous cortical oscillatory activity during motor action
Curr. Opin. Neurobiol.
High and low gamma EEG oscillations in central sensorimotor areas are conversely modulated during the human gait cycle
Neuroimage
Rats with lesions of the vestibular system require a visual landmark for spatial navigation
Behav. Brain Res.
Theta oscillations at encoding mediate the context-dependent nature of human episodic memory
Curr. Biol.
Level of participation in robotic-assisted treadmill walking modulates midline sensorimotor EEG rhythms in able-bodied subjects
Neuroimage
Long-range EEG synchronization during word encoding correlates with successful memory performance
Cogn. Brain Res.
Controlling the false discovery rate: a practical and powerful approach to multiple testing
J. R. Stat. Soc. Ser. B
Brain computer interface to enhance episodic memory in human participants
Front. Hum. Neurosci.
Synchronous and asynchronous theta and gamma activity during episodic memory formation
J. Neurosci.
Mental space travel: damage to posterior parietal cortex prevents egocentric navigation and reexperiencing of remote spatial memories
J. Exp. Psychol. Learn. Mem. Cogn.
Mechanisms and functions of theta rhythms
Annu. Rev. Neurosci.
Cited by (46)
Impacts of connections to the outside on underground space occupants’ psychophysiological health: A virtual reality-based experimental approach
2024, Tunnelling and Underground Space TechnologyGamma oscillations and episodic memory
2023, Trends in NeurosciencesSleep spindles track cortical learning patterns for memory consolidation
2022, Current BiologyThe phase of Theta oscillations modulates successful memory formation at encoding
2021, NeuropsychologiaCitation Excerpt :The aim of Experiment 2 was to address the role of ongoing low-frequency oscillations—before the to-be-encoded stimulus presentation—in subsequent recognition performance using EEG. Several neuroimaging studies have shown that frontotemporal regions are actively engaged during episodic encoding (Wagner 1999; Kirchhoff et al., 2000; Hanslmayr et al., 2011; Park and Rugg 2011; Griffiths et al., 2016). However, the results show contradictory evidence for Theta frequency fluctuations of the local field potential: some provide evidence in favor of encoding-related increases (Herweg et al., 2016; Addante et al., 2011; Summerfield and Mangels 2005; Osipova et al., 2006; Hanslmayr et al., 2011), other results report decreases (Fellner et al. 2016, 2019; Michelmann et al. 2018).
Alpha/beta power decreases during episodic memory formation predict the magnitude of alpha/beta power decreases during subsequent retrieval
2021, NeuropsychologiaCitation Excerpt :For remembered items, the largest cluster revealed a significant correlation between encoding- and retrieval-related alpha/beta power which included encoding time-points from 100 ms to 1500 ms and retrieval time-points ranging from 125 ms to 1950 ms [pclus = 0.006, summed t-statistic = 3484.52, cluster size = 1425, Cohen's dz = 0.59]. While temporally broad, inspection of the time generalisation matrix suggests that this cluster peaked at encoding time-points between 500 and 1000 ms and at retrieval time-points between 300 and 1800 ms. These results indicate that the correlated alpha/beta power decreases have different time-courses during encoding and retrieval, which seemingly align with the timing of previously-reported subsequent memory effects and retrieval success effects, respectively (e.g. Fell et al., 2008; Griffiths et al., 2016; Hanslmayr et al., 2009; Karlsson et al., 2020; Martín-Buro et al., 2020; Michelmann et al., 2016). Critically, these memory-related decreases in alpha/beta power also overlapped with the significant reduction in alpha/beta power that accompanies stimulus perception and retrieval (see supplementary results, see supplementary Fig. 2), rather than the alpha/beta rebound that follows stimulus-induced suppression.