A central observation in the recognition memory literature is that neural processes occurring during encoding of stimuli are predictive of their later recognition and recall. Compared to items that are later forgotten, encoding of correctly recognised items has been associated with greater amplitude between 400 ms and 800 ms post stimulus onset across centro-parietal sites (the difference-due-to-memory ERP effect), increased EEG power in the theta and gamma frequency bands and decreased EEG power in the alpha and beta bands, and increased theta-gamma phase-amplitude coupling. Importantly, theories of encoding based on these findings imply that these effects should be domain-general. In this pre-registered study, we tested this assumption by exploring neural correlates of successful encoding in learning of novel names for novel concepts. Following the previous studies, we used three different measures of neural activity, ERPs, time-frequency representations of power, and phase-amplitude coupling; however, for either of these measures, we could not reject the null hypothesis of no difference between the novel names that were later recalled and those that were not. We provide three possible interpretations of our findings, and our main conclusion is that the existing theories of encoding may be underspecified and that properly-powered pre-registered studies are needed to further constrain these theories.

In novel word learning, it is hypothesised that a memory of a novel word starts as an episodic memory. At this stage, known as encoding (e.g., Craik et al., 2007; Davachi & Dobbins, 2008), detailed information about the novel word (e.g., its meaning or pronunciation) is intertwined with the spatiotemporal context it was acquired in (e.g., Dickerson & Eichenbaum, 2009). With time, the word-specific information is abstracted away from the episode-specific information, allowing the novel word to become integrated into the existing knowledge networks (e.g., Davis & Gaskell, 2009; McMurray et al., 2016). These networks constitute semantic memory, which is viewed as a store of our knowledge about the world and the meanings of verbal symbols (i.e., words) that we employ to make use of this knowledge (see Kumar, 2020, for a recent review).

While integration of novel words into semantic memory has been studied extensively in the past two decades (see, e.g., James et al., 2017, 2019; Walker et al., 2020; Weighall et al., 2017, for work with children and, e.g., Korochkina et al., 2021; Liu & Hell, 2020; Walker et al., 2019, for work with adults, as well as McMurray et al., 2016, for a review of studies published before 2016), to date, encoding has received little attention in the novel word learning literature. In contrast, in the recognition memory literature, encoding has been studied extensively, with the consensus that cognitive and neural processes that occur at this stage of learning of new information are pivotal to its later memorability (e.g., Paller & Wagner, 2002). Critically, though, most of these studies have used familiar words to study encoding. Yet, because familiar words are thought to be subserved primarily by semantic memory, the neurocognitive processes required to retain these words in episodic memory for a short amount of time may be quite different from those required to learn a list of previously unseen words, the referents of which may or may not be known (i.e., learning novel names for familiar or novel concepts, respectively). Therefore, notwithstanding the important contribution of the recognition memory literature to the study of memory formation, the generalisability of some of the findings in this literature remains unclear.

With the present study, we sought to address this issue and contribute to research on both word learning and encoding in general. In line with this, the objectives of the present study were to examine whether event-related potentials and neural oscillatory activity during encoding of novel names for novel concepts are predictive of later recall of these names.

ERP correlates of successful encoding

In addressing the first aim of this study, we built upon the previous research on recognition memory, where encoding is usually studied with the subsequent memory paradigm. This paradigm consists of two phases: in the study phase, participants’ electrophysiological activity is recorded as they study a list of familiar words. In the test phase, participants are shown the words from the study list together with new (not shown in the study phase) familiar words and asked to classify each word as ‘new’ (i.e., shown for the first time) or ‘old’ (i.e., was in the study list). Sometimes, the recognition task in the test phase is complemented with or replaced by a free recall task, in which participants are asked to write down or recite all the words from the study list that they can remember. A common finding from this paradigm is that, in the study phase, subsequently remembered words (i.e., correctly recognised as old and/or recalled) exhibit greater positivity between 400 ms and 800 ms post onset in the centro-parietal region than words that are subsequently forgotten (i.e., incorrectly recognised as new and/or not recalled); this phenomenon is known as the difference-due-to-memory (Dm) ERP effect or the subsequent memory effect (e.g., Münte et al., 2000; Neville et al., 1986; Paller, 1990; Paller et al., 1987, 1988; Röder et al., 2001; Rugg, 1995; Schott et al., 2002, and see Friedman & Johnson, 2000; Paller & Wagner, 2002; Wagner et al., 1999, for reviews). The difference-due-to-memory effect has sometimes been found to be greater for free recall as compared to recognition and cued recall (e.g., Paller, 1990; Paller et al., 1988) as well as for incidental as opposed to intentional learning tasks (see Wagner et al., 1999, for a review). In incidental learning, participants are not explicitly told to remember the words (in contrast to intentional learning) but perform an unrelated task (e.g., counting the number of letters) and are later asked to recall the words. In this paradigm, greater difference-due-to-memory effects have been reported for tasks that involved more extensive semantic processing (e.g., “Is the referent of the word a living thing?”) as opposed to those that required the participants to focus on structural or physical features of the target items (e.g., “Does this word contain exactly two vowels?”) (Wagner et al., 1999).

The difference-due-to-memory ERP effect has been proposed to reflect the difference in the effectiveness of, or strength of, encoding between remembered and forgotten words (e.g., Paller et al., 1988). Due to the timing and topography of this ERP effect, it is often regarded as a modulation of the Late Positive Component (LPC), although the relationship between the difference-due-to-memory effect and other event-related potential components is still debated (e.g., Wagner et al., 1999). The LPC is typically observed in word recognition tasks (both with and without explicit memory judgements) and constitutes a positive-going event-related potential (ERP) that is largest over parietal sites and peaks between 500 ms and 800 ms post stimulus onset (e.g., Curran, 1999; Rugg, 1995; Rugg & Curran, 2007). This ERP component has been associated with the retrieval of episodic information about both the target and the spatiotemporal context associated with it (i.e., recollection, see, e.g., Addante et al., 2012; Paller et al., 1995, and Diana et al., 2007; Eichenbaum et al., 2007; Yonelinas et al., 2010, for reviews).

While the difference-due-to-memory effect is well established in the memory literature, where familiar words serve as stimuli, to our knowledge, to date, only one study has examined whether this effect can be observed with novel words. Batterink & Neville (2011) presented their participants with short stories, in which novel words replaced real English words. The learning task was followed by two test tasks: in the semantic-relatedness judgement task, the participants saw pairs of words and judged whether the primes (trained novel words) were semantically related to the targets (familiar English words), and, in the translation task, participants were required to provide English translation equivalents for the trained novel words. Batterink & Neville (2011) then analysed whether, in the story task, the LPC response was greater if the novel words were responded to correctly versus incorrectly in the test tasks (i.e., whether there was a difference-due-to-memory effect). This pattern was observed for the translation but not for the semantic-relatedness judgement task, leading the authors to conclude that the amplitude of the LPC during encoding of novel words was predictive of subsequent recall but not recognition performance. Yet, it is unclear why, in the Batterink & Neville (2011) study, the translation task was considered a measure of (free) recall of the novel words as the participants had to produce English words. Thus, the first objective of the current study was to test Batterink & Neville’s claim by using a more standard measure of recall (retrieval of trained novel words cued by their definitions).

Neural oscillatory correlates of successful encoding

Our second aim was to study whether cued recall of novel words can be predicted by neural oscillatory activity during encoding. Macroscopic oscillations observed in an electroencephalogram are thought to reflect fluctuations in postsynaptic local field potentials of a large group of neighbouring neurons, often referred to as a neuronal ensemble (e.g., Buzsáki, 2006; Buzsáki & Draguhn, 2004). These neural oscillations are usually characterised by three properties: frequency (number of occurrences per unit of time; in Hz), amplitude (maximum value of current, or voltage; in μV) and phase (fraction of the cycle covered up to a specific time point; in radians (rad) or degrees [0°, 360°]). Because neural oscillations often exhibit characteristic rhythms, they are typically divided into several frequency bands: slow oscillations (<1 Hz), delta (1-3 Hz), theta (4-7 Hz), alpha (8-12 Hz), beta (13-29 Hz), and gamma (> 30 Hz). Thereby, low-frequency oscillations are thought to modulate neural activity over large spatial regions in long temporal windows, whereas high-frequency oscillations have been argued to modulate activity over small spatial regions and short temporal windows (e.g., von Stein & Sarnthein, 2000).

These distinct brain rhythms are known to exhibit two types of oscillatory behaviour, synchronisation and desynchronisation, that can occur either locally (~1 cm) or on a large scale (>1 cm). Local synchronisation takes place when neurons within a neuronal ensemble fire simultaneously in response to a stimulus and their postsynaptic activity synchronises. This gives rise to an increase in EEG power, where power is defined as amplitude squared (e.g., Fell & Axmacher, 2011; Fries, 2005). In contrast, the term large-scale, or phase, synchronisation is used when there is correlation of oscillatory phases of neural oscillations in spatially separated neuronal ensembles. In turn, desynchronisation refers to a local or a large-scale decrease in synchrony and is reflected in a reduction of power post stimulus onset compared to a pre-stimulus baseline period (e.g., Pfurtscheller & Aranibar, 1977). Numerous studies have examined neural oscillatory activity in encoding of familiar words, and there is a growing consensus that two distinct patterns are associated with memory encoding and retrieval: synchronisation of activity in theta and gamma bands and desynchronisation of activity in alpha and beta bands (see, e.g., Düzel et al., 2010; Fell & Axmacher, 2011; Hanslmayr et al., 2012; Nyhus & Curran, 2010, for reviews). We first provide a brief overview of the literature that has focused on each of these oscillatory patterns and then present a theory (Hanslmayr et al., 2016) that aims to integrate these seemingly opposing mechanisms.

At the cognitive level, the encoding process can be conceptualised as follows: for a memory trace of a stimulus to be created in episodic memory, numerous features attributable to this stimulus must be integrated into a perceptual whole (e.g., Nyhus & Curran, 2010). At the neurological level, the hippocampal memory indexing theory (Teyler & DiScenna, 1986; Teyler & Rudy, 2007) states that the formation of a unified episodic memory representation of a stimulus is achieved through binding together of individual cortical activity patterns by the hippocampus. Extending this theory, Nyhus & Curran (2010) proposed that this interaction between hippocampal and cortical structures is achieved by synchronised activity in the gamma and theta bands. Specifically, cortical rhythmic firing in the gamma band has been suggested to be a mechanism for binding of the cortical instantiations of the stimulus features into a unitary perceptual representation, while hippocampal theta oscillations have been argued to promote synaptic plasticity and the long-term potentiation required for memory formation (Nyhus & Curran, 2010). In line with Nyhus & Curran’s account, subsequently remembered items have been found to exhibit greater power in both the gamma and theta bands in the posterior cortex during encoding (e.g., Gruber et al., 2004; Hanslmayr et al., 2009; Osipova et al., 2006; Sederberg et al., 2003, 2007). Moreover, successful encoding has also been associated with gamma phase synchronisation between cortical and hippocampal neurons (e.g., Fell et al., 2001).

Besides synchronisation of activity within the theta and gamma frequency bands, interactions between these two brain rhythms have also been argued to play a functional role in memory formation. For instance, the theta-gamma code model states that theta-gamma cross-frequency coupling during encoding serves as a working memory buffer and thus supports encoding of the correct temporal order of episodic memories (Jensen & Lisman, 2005; Lisman, 1999; Lisman & Idiart, 1995; Lisman & Jensen, 2013). Building upon the premise of this model, Nyhus and Curran (2010) argued that binding and temporal ordering of the cortical instantiations of the individual stimulus features into a coherent episodic memory could occur by means of theta-gamma phase-amplitude coupling. Phase-amplitude coupling is a form of cross-frequency coupling that emerges when the amplitude of a high-frequency oscillation (here, gamma) couples to the phase of a low-frequency oscillation (here, theta) (see Canolty & Knight, 2010, for a review). This results in so-called nested oscillations, where the phase of slower oscillations (nesting) synchronises with the amplitude envelope of faster oscillations (nested) such that the fast rhythm’s power is always maximum at a certain phase of a slower oscillation. In accordance with these theoretical assumptions, several empirical studies have argued that the theta-gamma phase-amplitude coupling promotes integration of memories with their spatiotemporal context (e.g., Lega et al., 2014; Staudigl & Hanslmayr, 2013; Summerfield & Mangels, 2005; Tort et al., 2009).

In addition to a great deal of research examining the role of theta and gamma oscillations in memory formation, a smaller literature has focused on the oscillatory behaviour in the alpha and beta frequency bands, with the main finding that a power decrease in these two bands during encoding is positively related to later retrieval (e.g., Fellner et al., 2013; Hanslmayr et al., 2009, 2011; Klimesch, Schimke, et al., 1996; Sederberg et al., 2003, 2006, 2007). To account for this finding, Hanslmayr and colleagues have developed the information via desynchronisation hypothesis which states that there is an inverse relationship between the synchrony of the firing patterns of neural assemblies and the richness of information encoded in these firing patterns (Hanslmayr et al., 2012, and see, e.g., Goard & Dan, 2009; Schneidman et al., 2011, for non-human animal studies with a similar argument). According to Hanslmayr et al. (2012), the advantage of desynchronisation is that, unlike synchronous firing, it allows for creation of an infinite number of unique firing patterns (see Hanslmayr et al., 2012; Schneidman et al., 2011, for an illustration of this principle). Critically, this research group has shown that inducing synchronised activity in the left inferior prefrontal cortex at beta frequency with rTMS impaired encoding, suggesting that the relationship between memory formation and beta power decreases in this area could be causal (Hanslmayr & Staudigl, 2014, and see Hanslmayr et al., 2019, for a review of similar studies).

As summarised above, some studies in the memory literature have focused on the role of synchronised activity in the theta and gamma bands, while others have highlighted the importance of desynchronisation of activity in the alpha and beta bands for encoding of new memories. Importantly, Hanslmayr et al. (2016) have proposed a framework that coalesces these two, seemingly incompatible, mechanisms and the corresponding theoretical accounts. According to this framework, the disparity between oscillatory dynamics in the alpha-beta (desynchronisation) and in the theta-gamma bands (synchronisation) reflects the complementary nature of the memory systems involved in learning. In brief, Hanslmayr et al.’s model suggests that the alpha-beta desynchronisation is a mechanism by which the neocortex encodes patterns of activity generated by sensory stimulation (e.g., seeing or hearing a new word). Under this account, the decrease in synchronised firing in the alpha and beta bands in stimulus-specific neocortical structures could cause phase precession of oscillations in the theta band in the hippocampal structures. Phase precession occurs when, with each successive cycle of a neural oscillation, neural firing happens progressively earlier in relation to the phase of the oscillation (e.g., Jaramillo & Kempter, 2017; O’Keefe & Recce, 1993; Qasim et al., 2021). Importantly, phase precession has been suggested to be a mechanism by which sequential neural representations within one single theta cycle are temporarily linked to each other (see also Kovács, 2020). Moreover, Hanslmayr and collaborators have hypothesised that it is this regulatory influence of the phase of the theta rhythm that drives the nesting of the gamma oscillations within specific phases of the theta rhythm and, hence, phase-amplitude coupling in these two frequency bands. In turn, the theta-gamma phase-amplitude coupling is seen as a means of separation of stimulus-specific groups of neural assemblies from those that are not sensitive to the current sensory simulation, thus enabling long-term potentiation and binding of discontiguous patterns of cortical activity into one memory trace. This framework is, therefore, consistent with the Complementary Learning Systems account that postulates a division of labour between the hippocampal and the neocortical memory systems (e.g., Kumaran et al., 2016; McClelland, 2013; McClelland et al., 1995, 2020; O’Reilly et al., 2014).

To conclude, while there is a rich literature studying neural oscillatory activity during encoding using the subsequent memory paradigm, like the studies examining the difference-due-to-memory effect using ERPs, this literature has focused on encoding of familiar words. Consequently, it remains unclear whether brain oscillations during encoding of novel words would exhibit activity similar to that previously observed in the literature when participants memorised familiar words. Therefore, the second objective of the study reported here was to examine whether neural oscillatory activity (in addition to event-related potentials) during encoding of novel names for novel concepts can predict cued recall of these novel names.

Thus, the overarching research question of this exploratory study was whether there are differences between the neural correlates of successful versus not successful encoding in learning of novel names for novel concepts. To address this research question, we used three different measures of neural activity, ERPs, time-frequency power representations, and phase-amplitude coupling, because these measures are thought to reflect neural activity of at least partly different origin and subserve different (albeit possibly partly overlapping) neural and cognitive processes (e.g., Fries, 2005; Pfurtscheller & Lopes da Silva, 1999). This study was pre-registered, and the pre-registration is publicly available on the Open Science Framework (https://osf.io/jegv7).

Because the present study used data collected as part of our earlier study (see https://psyarxiv.com/vup25/ for a pre-print and https://osf.io/su7d3 for the pre-registration), the methodology for the study presented here is identical to Korochkina et al. (under review). Critically, however, these two studies concern different research questions and, consequently, have used data from different tasks and required different analyses. In the remainder of this section, we provide a brief overview of the methods and refer the reader to Korochkina et al. (under review) for further detail. The supplementary material for this paper includes stimulus lists for all tasks and model output from stimuli matching and is publicly available on OSF (https://osf.io/4kj3y/).

2.1. Participants

Seventy two1 native speakers of Australian English (28 male), aged 18–35 years (M = 20.94, SD = 3.86) and with no known hearing, neurological or psychiatric disorders, were recruited. None of the participants had started learning any other languages prior to the age of six and all reported Australian English as their first and dominant language. All participants were right-handed and had normal or corrected-to-normal vision. Participants received either monetary compensation (70 AUD) or course credit for their participation.

The experiment was approved by the Macquarie University Human Research Ethics Committee. Each participant was tested individually in a shielded room and gave written consent prior to the start of the experiment.

2.2. General procedure

The experiment was run using the Presentation software (Version 20.2, Build 07.25.18; Neurobehavioral Systems, Inc., Berkeley, CA, www.neurobs.com) at Macquarie University, Australia. Each participant attended two sessions on two consecutive days. In each session, in the learning phase, the participants learned novel names for a set of 20 novel concepts (e.g., a novel animal, piece of furniture, etc.) presented together with their descriptions. In both sessions, the learning phase was followed by a cued recall task for the set of novel names learned in that session. Following this, in the first session, the participants performed a working memory battery, whereas, in the second session, they completed a continuous primed lexical decision task and another cued recall task (for both sets). The second session lasted around 2 hours, and EEG was recorded throughout. In this article, we only report the EEG data from the novel word learning task and the accuracy data from first of the two cued recall tasks administered in Session 2.

2.3. Learning phase

2.3.1. Stimuli

We selected 40 existing concepts, each from a different semantic category, and used them to derive 40 novel concepts, with each novel concept representing a highly similar category coordinate to the existing concept. For example, one of the novel concepts was a plant that was very similar to cactus but had partly different semantic features (e.g., ability to survive extreme drought by curling into a tight ball and uncurling when exposed to water). Subsequently, descriptions for these 40 novel concepts were generated, with each description composed of 4 short sentences that did not include the category names of the concepts (sentence length: M = 9.98 words, SD = 2.58). Next, 23 Australian native speakers (5 male, age: M = 36.17, SD = 13.47) were asked to judge whether they were familiar with the described concepts and, if not, which existing objects these novel concepts reminded them of. Those items, which did not result in reporting of the existing concepts that the novel concepts were derived from, were replaced or modified (see Korochkina et al., under review, for more detail). The revised novel concepts and their descriptions were then given to another 10 speakers of Australian English (2 male, age: M = 28.20, SD = 5.75). For the revised items, there was high (90%) agreement in terms of the most similar existing concepts (which always belonged to the same semantic category as the novel concepts; see Table 1 for examples).

Table 1.
Sentences that formed the concept descriptions of (A) a novel fish similar to a shark, and (B) a novel plant similar to a cactus.
AB
It lives in saltwater and has a prominent dorsal fin. It has tiny round leaves and very strong roots. 
It is able to detect electromagnetic fields that living things produce. It is only found in deserts. 
It has a small body and an even smaller head and moves very slowly. It can survive extreme drought by curling into a tight ball. 
It always swims near the surface. It uncurls when exposed to water. 
AB
It lives in saltwater and has a prominent dorsal fin. It has tiny round leaves and very strong roots. 
It is able to detect electromagnetic fields that living things produce. It is only found in deserts. 
It has a small body and an even smaller head and moves very slowly. It can survive extreme drought by curling into a tight ball. 
It always swims near the surface. It uncurls when exposed to water. 

Following this, 40 novel words (20 monosyllabic and 20 disyllabic) were created such that they were phonotactically and orthographically legal in Australian English. The same 10 native speakers of Australian English judged that these novel words neither existed in English nor elicited strong associations with existing English words, and a recording of each word was made by a male native speaker of Australian English.

Finally, to ensure that each concept description and each novel name appeared in both sessions, 4 lists were constructed. First, we split the concept descriptions and the novel names into two groups of 20, which resulted in two sets of concept descriptions (1 and 2) and two sets of novel names (A and B). Each set of novel names contained 10 monosyllabic and 10 disyllabic words, and a Bayesian alternative to a t-test (as implemented in the R packages BEST and BayesFactor; see Supplementary material A on https://osf.io/4kj3y/ for model output) confirmed that the two sets were matched for orthographic Levenshtein distance 20 (OLD20; Yarkoni et al., 2008), number of letters, and bigram frequency (as given in the MCWord database; Medler & Binder, 2008). Next, four sets of concept description-novel name pairs were generated, with 20 pairs per set: 1A — Description Set 1 with Novel name Set A, 1B — Description Set 1 with Novel name Set B, 2A — Description Set 2 with Novel name Set A, 2B — Description Set 2 with Novel name Set B. We then created 4 lists such that there were all possible combinations of concept description-novel name sets across the lists: List 1 — set 1A in Session 1 and 2B in Session 2, List 2 — 1B and 2A, List 3 — 2A and 1B and List 4 — 2B and 1A. The participants were randomly assigned to the lists, and there was an equal number of participants (N = 9) per list. The lists are provided in the Supplementary material B (https://osf.io/4kj3y/).

2.3.2. Procedure

Each trial started with a fixation cross shown for 500 ms in the centre of the screen (Figure 1). Next, the fixation cross was replaced by a novel name (in black Arial font, font size 60), and the participants viewed it silently for 2000 ms, during which a recording of this name was played. Subsequently, the description of the novel concept was presented, with the four sentences appearing one by one (in black Arial font, size 25) with 3000 ms between each sentence. After the last sentence, all four sentences remained on the screen for another 9000 ms, together with the novel name. This was followed by a blank screen for 1000 ms, and then the written form of the novel name (Arial font, size 60) was presented again, this time in red and with an image of a microphone above it, to indicate that a verbal response was required. The participants were given 3000 ms to read the novel name aloud, after which the screen was blank for 1000 ms before the next trial started.

Figure 1.
Example of one trial in the learning phase.
Figure 1.
Example of one trial in the learning phase.
Close modal

In total, each of the 20 novel names and concept descriptions was presented 4 times, each time in a different pseudo-randomised order. The novel name shown last in one round was never first in the next round, and none of the novel names appeared at the same position more than twice across the four rounds. The learning phase took approximately 45 minutes to complete (including breaks).

2.4. Cued recall tasks

Immediately following the learning phase in each session, a cued recall task was administered. In this task, the participants were asked to provide the names of the novel concepts they had just been trained on. Each trial started with a 500 ms fixation screen. Then, the description of the novel concept (four sentences shown during training) appeared on the screen (in black Arial font, size 25) with an image of a microphone above it. The participants were given 5000 ms to provide the name of the concept corresponding to the description orally (recorded with a Rode NTG1 shotgun microphone) and, after the image of the microphone disappeared, 10000 ms to type the novel name in a response box displayed under the description. The task lasted approximately 5 minutes (including one break).

We asked the participants to respond both orally and by typing their responses to have a better understanding of how well the participants had learned the novel word forms: by asking the participants to provide the spoken form, we wished to see whether they have remembered the correct pronunciation of the novel name, while, by asking them to type the words, we sought to test their knowledge of the novel words’ spelling. However, to facilitate comparison with the earlier studies, we pre-registered that only written responses would be considered in the analysis. This is because, in the literature where a recall task was used, the participants’ memory was typically tested by asking them to write down all words from the study list that they could remember, and no time limit was applied.

Participants completed another cued recall task at the end of Session 2 (following the primed continuous lexical decision task, see below). The procedure was identical to that described above, however, this time, descriptions of all 40 novel concepts (from both sets) were presented. The descriptions appeared in a random order (i.e., not blocked by set). This task took approximately 10 minutes and was administered to examine explicit memory and the amount of forgetting for the novel names trained in Session 1 (i.e., 24h ago); therefore, we do not discuss it here.

2.5. Acquisition and pre-processing of EEG data

EEG was recorded at a sampling rate of 2048 Hz from 64 Ag-AgCl-tipped electrodes attached to an electrode cap using the 10/20 system. The recordings were made with the BioSemi ActiveTwo EEG system, in which the function of the ground electrode of more conventional systems is accomplished by a feedback loop between two additional electrodes, the common mode sense (CMS; active electrode) and the driven right leg (DRL; passive electrode). In addition, external electrodes were placed on the left and right mastoids, and at the outer canthus of the left eye.

The raw signal was pre-processed in MATLAB (Version R2019a) using the EEGLAB toolbox (Delorme & Makeig, 2004). In the course of pre-processing, two datasets per participant were generated, one for the analysis of the event-related potentials (ERP dataset) and one for the time-frequency and phase-amplitude coupling analyses (TF dataset). For both datasets, we first re-sampled the continuous signal to 500 Hz and then filtered it with a 1 Hz high pass filter (Kaiser windowed sinc FIR filter, order = 802, beta = 4.9898, transition bandwidth = 2 Hz). This was followed by line noise removal using the PREP pre-processing pipeline (Bigdely-Shamlo et al., 2015). The subsequent pre-processing steps were performed separately for the two datasets. For the TF dataset, line noise removal was followed by identification of bad channels (using the functions from the PREP pre-processing pipeline). Next, following the literature on the time-frequency power correlates of the difference-due-to-memory effect, we re-referenced the signal to the average of all channels which had not been marked bad. In contrast, the ERP dataset, after line noise removal, was filtered with a 40 Hz low pass filter (Kaiser windowed sinc FIR filter, order = 162, beta = 4.9898, transition bandwidth = 10 Hz), re-referenced to the algebraic average of the left and right mastoids (as is common in the literature on the ERP correlates of the difference-due-to-memory effect) and subjected to the PREP pre-processing pipeline for bad channel identification. Subsequently, copies of both datasets were created (ERP dataset 2 and TF dataset 2, respectively).

Next, the original datasets (excluding bad channels) were put through an Independent Component Analysis (ICA; Chaumon et al., 2015), while copies of the original datasets (ERP dataset 2 and the TF dataset 2) were high pass filtered at 0.1 Hz to minimise slow drifts (Kaiser windowed sinc FIR filter, order = 8008, beta = 4.9898, transition bandwidth = 0.2 Hz). Following this, the demixing matrices obtained from the ICA (run on the original datasets) were applied to the ERP dataset 2 and the TF dataset 2 (filtered at 0.1 Hz). Then, we excluded individual components corresponding to activity resulting from the movements of the eyes and muscle motor units as well as electrocardiographic signals (using ICLabel, an automated EEG independent component classifier; Pion-Tonachini et al., 2019) and spherically interpolated the bad channels identified earlier.

Subsequently, the signal was segmented into epochs of 1.2 s (−200 ms to 1000 ms relative to stimulus onset) for the ERP dataset and 4.5 s (−1500 ms to 3000 ms relative to stimulus onset) for the TF dataset, where the 0 ms point was defined as the onset of the written word form on the screen. Note that this epoch length in the TF dataset (4.5 s) was due to the fact that we added buffer zones at either end of the time period of interest (0 ms to 1500 ms relative to stimulus onset) to avoid contamination from edge artifacts after Hilbert transform (see Cohen, 2014; Kramer et al., 2008, for a relevant discussion). Finally, in both datasets, all epochs were baseline corrected using the mean signal within 200 ms before stimulus onset, and epochs with amplitudes below −100 μV or above 100 μV within the time period of interest (−200-1200 ms for the ERP dataset and −200-1500 ms for the TF dataset) were discarded.

In the learning phase, each of the 20 novel names and concept descriptions was shown 4 times, which resulted in 80 trials per participant (see above). Because we did not expect any differences between the two conditions (novel names later recalled correctly vs. incorrectly; henceforth, recalled and not recalled novel names) at the first exposure, we pre-registered that only data from subsequent exposures (2, 3, and 4) would be considered for the analysis, resulting in a maximum of 60 trials per participant.2 We also pre-registered that only data from those participants for whom at least 20 trials per condition are retained after pre-processing would be included in the analysis. At the end of the pre-processing, there were 31 participants in the ERP dataset (12 male, age: M = 21.77, SD = 3.93) and 25 participants in the TF dataset (9 male, age: M = 22.32, SD = 4.10) for whom this criterion was met (note that the smaller number of participants in the TF dataset was due to the longer epochs (ERP: −200 ms to 1000 ms; TF: −200 ms to 1500 ms) resulting in more ‘bad’ trials being identified and excluded). For these participants, in the ERP dataset, there were a mean of 29.29 epochs (SD = 6.01) per participant in the recalled condition and a mean of 29.94 epochs (SD = 6.02) per participant in the not recalled condition, while, in the TF dataset, there were a mean of 28.12 epochs (SD = 5.73) and a mean of 28.52 epochs (SD = 6.35) per participant in the recalled and not recalled conditions, respectively. Raw and pre-processed data (https://osf.io/vn7hc/) as well as code used for data pre-processing (https://osf.io/agnf6/) are publicly available on OSF.

Three analyses were pre-registered, with all analyses contrasting neural activity during encoding of novel names for novel concepts that were either later recalled or not recalled. Each of the analyses focused on a different measure of neural activity: post stimulus event-related potentials in Analysis 1, post-stimulus time-frequency representations in five frequency bands (theta, 4-7 Hz; alpha, 8-12 Hz; beta, 13-29 Hz; low gamma, 30-60 Hz, and high gamma, 61-100 Hz) in Analysis 2, and post-stimulus phase-amplitude coupling between the theta and low gamma frequency oscillations in Analysis 3. Given the great deal of variability in the literature with respect to the choice of analysis of the EEG data (see, e.g., Jurkiewicz et al., 2020, for an overview of methods used to quantify phase-amplitude coupling), we pre-registered those analyses which, in our opinion, provide relatively comprehensive information about the data. After these three analyses were completed, we performed additional exploratory analyses. All analyses performed were mass univariate analyses, and correction for multiple comparisons used cluster-based statistics. In Analyses 1 and 2, we applied the Threshold Free Cluster Enhancement technique (TFCE; Smith & Nichols, 2009) in conjunction with bootstrapping, and, in Analysis 3, cluster-based statistics were computed using permutation. While these methods have been found to control the family-wise error rate to a similar degree (see Pernet et al., 2015, for a relevant comparison and discussion), we used the permutation technique in Analysis 3 to follow the recommendations in the literature (e.g., Seymour et al., 2017). In the remainder of this section, we detail all conducted analyses and their results. The code for all analyses is available on OSF (https://osf.io/agnf6/).

3.1. Cued recall task

In the cued recall task, mean accuracy, for the spoken responses, was 44% correct (SD = 12%) and, for the typed responses, 51% (SD = 13%). The seemingly higher accuracy for the typed responses is most likely due to the participants having more time for retrieval (5 seconds for the spoken responses followed by additional 10 seconds for the typed responses).

3.2. Post stimulus event-related potentials

In this pre-registered analysis, the dependent variable was amplitude measured every two milliseconds between −200 ms and 1000 ms relative to stimulus onset, for every electrode and trial. We contrasted the amplitude for recalled versus not recalled novel names by means of a mass univariate analysis. This analysis allows testing for significant differences between the condition(s) of interest across all electrodes and time points (Groppe et al., 2011a, 2011b). We performed the analysis in LIMO (LInear MOdelling of MEEG data), an open-source MATLAB toolbox (Pernet et al., 2011). In LIMO, mass univariate analysis is based on a hierarchical general linear model and modelling occurs at two levels. At the first level (single-subject level), beta coefficients for each level of the predictor(s) of interest are estimated separately for each participant, time point and electrode, while, at the second level (group level), these parameter estimates are used to test the statistical significance of the predictor(s) across all participants.

In the present study, model parameters at the single-subject level were estimated using the trial-based Ordinary Least Squares approach and the group-level analysis was performed using robust paired t-tests. Correction for multiple comparisons used the Threshold Free Cluster Enhancement method (TFCE; Pernet et al., 2015; Smith & Nichols, 2009), which, in LIMO, relies on bootstrapping. Note that this is true for all analyses, pre-registered and exploratory, that were conducted in LIMO, and, therefore, this information is not repeated in the following sections.

Topographic maps depicted in Figure 2 suggest that the ERP responses for the novel names that were later recalled were similar to those for the novel names that were not recalled. To facilitate comparison with results of the previous studies, where the difference-due-to-memory effect is typically reported at central and parietal sites within the time window between 400 ms and 800 ms post stimulus onset (e.g., Batterink & Neville, 2011; Paller & Wagner, 2002), Figure 3 shows mean amplitudes for the recalled and not recalled words computed by averaging the data across 15 centro-parietal and parietal electrodes (P5, P3, P1, Pz, P2, P4, P6, CP5, CP3, CP1, CPz, CP2, CP4, CP6), with the time window of interest shaded grey. The uncorrected t-values from the mass univariate analysis are shown in Figure 4, and none remained significant after correction for multiple comparisons.

Figure 2.
Topographic maps of amplitudes in the time window between 400 ms and 800 ms after stimulus onset during encoding of novel names that were later recalled versus not recalled.
Figure 2.
Topographic maps of amplitudes in the time window between 400 ms and 800 ms after stimulus onset during encoding of novel names that were later recalled versus not recalled.
Close modal
Figure 3.
ERP responses across a set of centro-parietal electrodes during encoding of novel names that were later recalled versus not recalled. The time window between 400 ms and 800 ms post stimulus onset is shaded grey.
Figure 3.
ERP responses across a set of centro-parietal electrodes during encoding of novel names that were later recalled versus not recalled. The time window between 400 ms and 800 ms post stimulus onset is shaded grey.
Close modal
Figure 4.
Results of the mass univariate analysis for the comparison of the ERP responses between the recalled and not recalled novel names, with significant t-values before TFCE correction.
Figure 4.
Results of the mass univariate analysis for the comparison of the ERP responses between the recalled and not recalled novel names, with significant t-values before TFCE correction.
Close modal

3.3. Post stimulus time-frequency representations of EEG power

3.3.1. Pre-registered analysis

The dependent variable was squared amplitude, or EEG power, for every two milliseconds between −200 ms and 1500 ms post stimulus onset in five frequency bands, theta (4-7 Hz), alpha (8-12 Hz), beta (13-29 Hz), low gamma (30-60 Hz), and high gamma (61-100 Hz), for every electrode and trial. To obtain power in these frequency bands, we first computed time-frequency representations for frequencies ranging from 4 Hz to 100 Hz, with 97 frequencies in between, using the continuous wavelet transform method with complex Morlet wavelets (e.g., Sinkkonen et al., 1995). To increase temporal precision at lower frequencies and frequency precision at higher frequencies, we used a variable number of wavelet cycles (see, e.g., Cohen, 2014, for a relevant discussion): this number varied from 3 to 10 as a function of the frequency of the wavelet with the increase occurring in 97 steps (as this was the number of steps used to increase the frequency of the wavelets from 4 Hz to 100 Hz). As is common in the literature, the result of the convolution was baseline normalised by converting it to the decibel scale using the −400 ms to −100 ms baseline relative to stimulus onset (see Cohen, 2014, for a discussion on advantages of the baseline normalisation).

Subsequently, power in each of the five frequency bands was computed by averaging power across the individual frequencies in these bands (4-7 Hz for theta, 8-12 Hz for alpha, 13-29 Hz for beta, 30-60 Hz for low gamma, and 61-100 Hz for high gamma) separately for each participant, electrode, trial, and time point. Differences in power between the recalled and not recalled conditions were analysed with the mass univariate analysis using the LIMO toolbox in MATLAB. The topographic maps for each frequency band (Figure 5) suggest that EEG power was similar for the later recalled versus not recalled words in all frequency bands except for the alpha band, where power was lower for the recalled words across a set of frontal and central electrodes between 600 ms and 800 ms post stimulus onset. However, after correction, there were no significant t-values in either frequency band. Uncorrected t-values for each band can be found in Figure 6.

Figure 5.
Topographic maps of time-frequency power (in dB) in theta, alpha, beta, low gamma, and high gamma frequency bands during encoding of later recalled and not recalled novel names between 200ms and 1400 ms after stimulus onset.
Figure 5.
Topographic maps of time-frequency power (in dB) in theta, alpha, beta, low gamma, and high gamma frequency bands during encoding of later recalled and not recalled novel names between 200ms and 1400 ms after stimulus onset.
Close modal
Figure 6.
Results of the pre-registered mass univariate analysis for the comparison of EEG power in five frequency bands during encoding of the later recalled and not recalled novel names, with significant t-values before TFCE correction.
Figure 6.
Results of the pre-registered mass univariate analysis for the comparison of EEG power in five frequency bands during encoding of the later recalled and not recalled novel names, with significant t-values before TFCE correction.
Close modal

3.3.2. Exploratory analysis: power differences for recalled vs. not recalled words for each individual frequency and time point

Because single-subject time-frequency data is four-dimensional (electrode × frequency × time point × trial) and its analysis requires significant computational resources, in the literature, the data is usually converted to a three-dimensional matrix. Thereby, this conversion typically occurs at the expense of either the electrode or frequency dimensions. In the pre-registered analysis, we averaged across individual frequencies in five frequency bands and contrasted the recalled and the not recalled novel names in each of the five frequency bands separately for each electrode and time point. While this allowed us to explore the effect of interest at each electrode, it inevitably resulted in a partial loss of the frequency information. Consequently, in the exploratory analysis, we decreased the number of dimensions by averaging out the electrode dimension. In this analysis, recalled and not recalled words were compared for each individual frequency (4 frequencies in the theta range, 4-7 Hz; 5 frequencies in the alpha range, 8-12 Hz; 17 frequencies in the beta range, 13-29 Hz; 31 frequencies in the low gamma range, 30-60 Hz; 40 frequencies in the high gamma range, 61-100 Hz) and time point but averaging across selected electrodes. For frequencies in the theta, alpha and beta bands, 15 medial pre-frontal and frontal electrodes (Fp1, Fpz, Fp2, AF7, AF3, AFz, AF4, AF8, F5, F3, F1, Fz, F2, F4, F6) were selected and, for frequencies in the low and high gamma bands, 14 medial centro-parietal and parietal electrodes (P5, P3, P1, Pz, P2, P4, P6, CP5, CP3, CP1, CPz, CP2, CP4, CP6). These electrodes were chosen because, in the recognition memory literature, power increases/decreases during encoding of later remembered versus forgotten words are most often reported in the frontal regions for the lower frequency (below 30 Hz) bands (e.g., Caplan & Glaholt, 2007; Summerfield & Mangels, 2005) and in the parietal regions for frequencies in the gamma band (e.g., Hanslmayr et al., 2009; Osipova et al., 2006). In this analysis, no t-values remained significant after TFCE correction in either frequency band, but note that the uncorrected t-values seem to suggest that, for all alpha frequencies, power was lower for the recalled than for the not recalled words between 600 ms and 800 ms post stimulus onset (see Figure 7).

Figure 7.
Results of the exploratory mass univariate analysis for the comparison of EEG power for each frequency across a set of frontal electrodes for theta, alpha, and beta frequency bands and across a set of centro-parietal electrodes for low and high gamma frequency bands during encoding of novel names that were later recalled vs. not recalled, with significant t-values before TFCE correction.
Figure 7.
Results of the exploratory mass univariate analysis for the comparison of EEG power for each frequency across a set of frontal electrodes for theta, alpha, and beta frequency bands and across a set of centro-parietal electrodes for low and high gamma frequency bands during encoding of novel names that were later recalled vs. not recalled, with significant t-values before TFCE correction.
Close modal

3.4. Theta-gamma phase-amplitude coupling

3.4.1. Pre-registered analysis

In the literature, in tasks and designs similar to ours, theta-gamma phase-amplitude coupling is typically reported for theta oscillations in the medial frontal and pre-frontal areas and gamma oscillations in the medial parietal and centro-parietal areas (e.g., Köster et al., 2014). Consequently, for each participant, we restricted the data to 15 medial pre-frontal and frontal electrodes (Fp1, Fpz, Fp2, AF7, AF3, AFz, AF4, AF8, F5, F3, F1, Fz, F2, F4, F6; theta dataset) and 14 medial centro-parietal and parietal electrodes (P5, P3, P1, Pz, P2, P4, P6, CP5, CP3, CP1, CPz, CP2, CP4, CP6; gamma dataset). To extract information about the phase of the theta oscillations (4-7 Hz) and the amplitude of the low gamma oscillations (30-60 Hz), we relied on the code and recommendations given in Seymour et al. (2017). First, for each of the two datasets, a fourth-order bandpass Butterworth filter was applied (4-7 Hz for the theta dataset and 30-60 Hz for the gamma dataset), with the bandwidth defined as ±0.4 times the centre frequency as this has been shown to increase the ability to detect phase-amplitude coupling (e.g., Berman et al., 2012; Voloh et al., 2015). Next, the amplitude and phase information was extracted from the gamma (in steps of 2 Hz, i.e., for every alternate frequency in the 30-60 Hz range) and theta (in steps of 1 Hz, i e, for every frequency in the 4-7 Hz range) datasets, using the Hilbert transform (Le Van Quyen et al., 2001). To diminish contamination by edge artifacts, the first 1500 ms and the last 1500 ms of each trial were discarded, leaving the time window between 0 ms and 1500 ms post stimulus onset for the analysis.

The theta-gamma phase-amplitude coupling was quantified using the Modulation Index method (Tort et al., 2008, 2010), which is the most common approach in the EEG/MEG literature (but see, e.g., Jurkiewicz et al., 2020; Martinez-Cancino et al., 2020; Munia & Aviyente, 2019, for recent reviews and other methods). We first binned the theta frequency phases into 18 intervals of 20° (ranging from 0° to 360°) and then calculated the mean amplitude of the gamma frequency oscillations in each of these 18 phase bins, effectively transforming the mean amplitude per phase into a probability distribution-like function. Because it is assumed that, if there is no phase-amplitude coupling between the pair of frequencies, the amplitude distribution over the phase bins would be uniform, we quantified the deviation of the observed distribution from the uniform distribution by means of the Kullback-Leibler (KL) distance measure (Kullback & Leibler, 1951; Santner & Duffy, 1989) and then divided the KL distance by the logarithm of the number of phase bins (N = 18) to compute the Modulation Index. This was done separately for each participant and condition (recalled vs. not recalled) but averaging across trials. The result of this procedure was, for each participant and condition, a matrix (commonly referred to as a comodulogram) of Modulation Indices, with one value per amplitude and phase frequency pair. To normalise the Modulation Indices, the procedure described above was repeated with surrogate data (as is common in the literature). The surrogates (N = 200) were obtained by re-shuffling the trial and phase information and thus disrupting the coordination between the amplitudes and the phases in time which is assumed to exist if phase-amplitude coupling is present (e.g., Hurtado et al., 2004).

Then, to compare the differences in Modulation Indices across the recalled and the not recalled words, the single-subject comodulograms were subjected to a cluster-based nonparametric test (Maris & Oostenveld, 2007) as it is implemented in the FieldTrip toolbox (Oostenveld et al., 2011) in MATLAB. In brief, for every phase-amplitude pair, the signals from the two experimental conditions were compared by means of a paired two-tailed t-test, and the pairs with t-values exceeding a .05 significance threshold were grouped into clusters based on their phase-amplitude adjacency. The cluster-level statistics were then computed by taking the sum of the t-values within each cluster, and the largest of these statistics was carried forward. Subsequently, we corrected for multiple comparisons using the Monte Carlo method: we re-shuffled the condition label 1000 times (referred to as random partition) and, for each permutation, computed the largest cluster-level statistic. This resulted in a reference distribution, against which the maximum cluster-levels statistics from the original, not permuted, clusters were compared. We then computed the proportion of random partitions that resulted in a larger test statistic than that from the not permuted clusters (this proportion is called the Monte Carlo significance probability). We considered the differences between the two conditions statistically significant if the Monte Carlo significance probability was smaller than .05. Finally, to ensure the validity of the results, cluster-based nonparametric tests were also carried out on Modulation Indices normalised with the surrogate data. Figure 8 illustrates that none of the clusters passed a .05 significance threshold after correction and that similar results were obtained with surrogate-normalised Modulation Indices.

Figure 8.
T-values for the contrast between real (panel A) and surrogate-normalised (panel B) theta-gamma phase-amplitude coupling during encoding of novel names that were later recalled versus not recalled. In this analysis, gamma frequencies in the 30−60 Hz range were analysed. No t-values remained significant after correction for multiple comparisons.
Figure 8.
T-values for the contrast between real (panel A) and surrogate-normalised (panel B) theta-gamma phase-amplitude coupling during encoding of novel names that were later recalled versus not recalled. In this analysis, gamma frequencies in the 30−60 Hz range were analysed. No t-values remained significant after correction for multiple comparisons.
Close modal

3.4.2. Exploratory analysis: differences in theta-gamma phase-amplitude coupling for recalled vs. not recalled words for gamma frequencies from both low and high gamma bands

The exploratory analysis differed from the analysis described above in one aspect only: while, in the pre-registered analysis, we only considered the low gamma band (30-60 Hz), in the exploratory analysis, we used frequencies (in steps of 2 Hz) in the range between 30 Hz and 100 Hz to extract the amplitude information. This analysis was performed because, in the literature, the gamma band is defined inconsistently across the studies examining the theta-gamma phase-amplitude coupling, and, whereas most studies focused on frequencies below 60-70 Hz (e.g., 30-50 Hz in Sauseng et al., 2008), others used a broader range (e.g., 30-80 Hz in Demiralp et al., 2007; 30-90 Hz in Staudigl & Hanslmayr, 2013; but see, e.g., Friese et al., 2013, and Köster et al., 2014, for use of ranges such 50-70 Hz and 50-80 Hz, respectively). The results of this analysis were similar to the findings of the pre-registered analysis of theta-gamma phase-amplitude coupling: no cluster survived correction for multiple comparisons either for real or surrogate-normalised Modulation Indices (Figure 9).

Figure 9.
T-values for the contrast between real (panel A) and surrogate-normalised (panel B) theta-gamma phase-amplitude coupling during encoding of novel names that were later recalled versus not recalled. In this analysis, gamma frequencies in the 30−100 Hz range were analysed. No t-values remained significant after correction for multiple comparisons.
Figure 9.
T-values for the contrast between real (panel A) and surrogate-normalised (panel B) theta-gamma phase-amplitude coupling during encoding of novel names that were later recalled versus not recalled. In this analysis, gamma frequencies in the 30−100 Hz range were analysed. No t-values remained significant after correction for multiple comparisons.
Close modal

The aim of this study was to explore the neural correlates of encoding in learning of novel words. In previous studies on encoding in the recognition memory literature, using familiar words as stimuli, encoding of words that were later remembered as opposed to forgotten has been associated with greater amplitude across parietal electrodes between 400 ms and 800 ms post stimulus onset (i.e., difference-due-to-memory ERP effect; see, e.g., Friedman & Johnson, 2000; Paller & Wagner, 2002; Wagner et al., 1999, for reviews), enhanced desynchronisation in the alpha and beta bands (e.g., Hanslmayr et al., 2009; Klimesch, Schimke, et al., 1996; Sederberg et al., 2007; Weiss & Rappelsberger, 2000), enhanced synchronisation in the theta and gamma bands (e.g., Caplan & Glaholt, 2007; Gruber et al., 2004; Klimesch, Doppelmayr, et al., 1996; Summerfield & Mangels, 2005), and greater phase-amplitude coupling between the theta and gamma oscillations (e.g., Köster et al., 2014; Lega et al., 2014; Staudigl & Hanslmayr, 2013). In this literature, the difference-due-to-memory effect has been taken to reflect differences in the strength of encoding between words that are later remembered versus those that are not (e.g., Paller et al., 1988). With respect to changes in EEG power, desynchronisation in the alpha and beta bands has been associated with encoding of activity patterns resulting from sensory stimulation (Hanslmayr et al., 2012, 2016), while synchronisation in the theta and gamma bands and theta-gamma phase-amplitude coupling have been linked with temporal ordering of and binding of the cortical representations of the stimulus features into a uniform memory trace and long-term potentiation (e.g., Hanslmayr et al., 2016; Lisman & Jensen, 2013; Nyhus & Curran, 2010).

Based on the findings in the recognition memory literature, in the present study, we assumed that the neural correlates of successful encoding are domain-general and tested the prediction that similar effects should be observed for different types of stimuli. We achieved this by examining encoding of novel names for novel concepts (rather than existing, known words), and, to the best of our knowledge, our study is the first pre-registered study to attempt this. Building upon the earlier findings, we contrasted three measures of neural activity — event-related potentials, time-frequency power, and theta-gamma phase-amplitude coupling — for those stimuli that were later successfully recalled with those that were not. We were not able to reject the null hypothesis in either of the pre-registered or exploratory analyses. While null results are hard to interpret, we believe that our study nevertheless contributes to the literature on encoding, while also highlighting potential methodological issues in this research field. It is of note that our study differed from the earlier studies not only in terms of the novelty of the to-be-encoded information but also in terms of the overall exposure and testing protocols. For instance, in our study, the cued recall task implicated retrieval of both word-form and conceptual information, with the latter being acquired incrementally across the encoding trials. In contrast, in the earlier studies, there was greater alignment between information that was presumably being encoded during the EEG recording (i.e., word-form information) and information that was necessary for successful performance on a subsequent test (typically in form of a word recognition task). While this methodological difference could have contributed to the disparity between the outcomes of the previous studies and ours, in the remainder of this article, we focus on three other possible factors that could account for our results — (1) low signal-to-noise ratio, (2) genuine differences in encoding of familiar versus and novel words, (3) absence of differences in encoding of familiar and novel words, with the existing theories possibly being underspecified.

The first possibility that needs to be considered when interpreting the outcomes of the present study is that the null results could be explained by a low signal-to-noise ratio. In our study, each participant (N = 72) learned 20 novel words, and this number of stimuli was determined by practical concerns such as length and difficulty of the experiment. Moreover, to equalise the number of epochs per participant across the conditions of interest (recalled vs. not recalled novel words) while having the largest possible sample size, we excluded participants for whom less than 20 epochs per condition remained after pre-processing. This resulted in a sample of 31 and 25 participants for the ERP and the time-frequency analyses, respectively, with a mean of approximately 30 trials per participant and condition. In contrast, most of the previous studies had substantially more data points per participant (e.g., 96 words in Klimesch, Schimke, et al., 1996; 120 items in Staudigl & Hanslmayr, 2013; 200 words in Paller et al., 1988; 210 words in Paller, 1990; 150 items in Hanslmayr et al., 2009; 250 words in Gruber et al., 2004; 240 in Osipova et al., 2006). Admittedly, not all of these studies report whether there was an equal number of epochs per participant per condition and some even explicitly mention an imbalance (e.g., Fellner et al., 2013), which could have resulted in higher signal-to-noise ratio in one of the conditions, thus affecting the estimates of the effects of interest. In addition, with respect to the number of participants, our study is either similar to or favourably compares to the previous studies, where sample sizes of 10–20 participants are quite common (e.g., 21 participants in Batterink & Neville, 2011; 10 in Klimesch, Schimke, et al., 1996; 18 and 23 in Staudigl & Hanslmayr, 2013; 20 per group in Hanslmayr et al., 2009; 12 per group in Paller, 1990; 10 in Paller et al., 1988; 16 in Fellner et al., 2013; 19 in Gruber et al., 2004; 13 in Osipova et al., 2006). Nevertheless, because a decrease in statistical power due to a low number of data points per participant cannot always be compensated by increasing the number of participants (e.g., Stevens & Brysbaert, 2016), it is possible that, in our study, detection of the effect of interest was precluded by a low signal-to-noise ratio and, consequently, insufficient statistical power. Alternatively, it could be that the differences in neural activity reported in the recognition memory literature primarily manifest under very challenging learning conditions (i.e., a large number of stimuli; but note that the learning task in our study was quite challenging too, with the participants being exposed to each novel word and its definition only four times and with no intermittent testing). If this is the case, then the reported effects should be seen as neural correlates of specific but not overarching encoding processes (i.e., only apparent under challenging conditions). It would be interesting to test this preliminary hypothesis in future experiments.

Despite the possibility of limited signal-noise ratio in our study, it is of note that there are other studies in the literature where significant effects were observed with a number of participants and trials similar to ours (e.g., Friese et al., 2013, with 23 participants and an average of 26.1 trials (SD = 7.7) per condition; Köster et al., 2014, with 22 subjects and an average of 26.55 trials (SD = 7.75) per condition). Consequently, pre-registered studies with a large number of both participants and items are needed to test replicability of both our study’s findings and those of previous studies.

Another possibility that could account for our results differing from the patterns predicted is that encoding of novel words involves neurocognitive processes that differ from those necessary for encoding of familiar words. Learning a novel name for a novel concept involves encoding the information about the novel word form, the novel concept, and the association between them (e.g., Hawkins et al., 2015; McMurray et al., 2016). In contrast, in the case of already familiar words, both the word form and the corresponding concept are already stored in semantic memory and are therefore well-connected with other items in the lexical-semantic network (e.g., Davis & Gaskell, 2009). Consequently, for familiar words, encoding should entail merely a re-activation of their representations in episodic memory, whereas, for novel words, it should require creation of new episodic memory traces. Therefore, it is conceivable that theories developed based on the results with familiar words (e.g., Hanslmayr et al., 2012, 2016; Nyhus & Curran, 2010) may be underspecified in so far as how they apply to novel words. Yet, to date, with the exception of the present work, only one study has examined this issue and only by means of one measure of neural activity, the ERPs (Batterink & Neville, 2011). Before we turn to a discussion of the Batterink & Neville (2011) study, we note that neither their study nor ours directly compared familiar and novel words, as would be required to test the hypothesis that processes involved in encoding of familiar words differ from those engaged in encoding of novel words. This hypothesis could be addressed by future research.

In Batterink & Neville (2011), recognition of trained novel names was assessed by means of a task in which familiar words had to be judged for semantic relatedness to the trained novel words. The trained novel words were considered recognised if correct judgements were given for both related and unrelated trials. Recall was measured using a translation task, in which, for each trained novel word, the participants were required to write down its translation equivalent in their first language (English). No differences were found between the novel names that were later recognised versus not recognised, whereas greater amplitude in the LPC spatiotemporal window was reported for items that were later recalled as opposed to not recalled. However, as, in the translation task, participants were required to produce the English translations of the novel names rather than the novel names themselves, it is contentious whether this can be considered an adequate measure of recall. Indeed, in all other studies in the literature, recall was usually assessed by asking the participants to say or write down the recently trained words that they could remember, either in the presence of cues (e.g., word stem; cued recall) or without them (free recall). Therefore, while the Batterink & Neville (2011) study may speak to whether there are differences in encoding of novel names that were later recognised versus not recognised, it cannot inform differences in encoding between novel words that are later recalled versus not recalled, which was the focus of our study.

Whether success of encoding is measured with recognition or recall tasks is no trivial matter because these paradigms are thought to reflect different memory processes. While recognition is defined as the subjective experience of one’s memory for the stimulus, recall is believed to rely on the process of recollection, conceptualised as retrieval of the stimulus and the episode-specific details associated with it during encoding (but see, e.g., Yonelinas, 2002, for an overview and discussion of different models of memory). It has further been argued that recall requires more elaborative memory processes and is thus more sensitive to encoding strength than recognition (e.g., Paller, 1990; Paller et al., 1988). This argument builds upon studies in the recognition memory literature, where greater ERP difference-due-to-memory effects had been reported for free recall compared to recognition and cued recall (Paller, 1990; Paller et al., 1988). In our study, success of encoding was determined based solely on the participants’ performance on the cued recall task, and one of the future research directions could be to investigate, in high-N pre-registered experiments, whether Paller and colleagues’ finding can be extended to the recall of novel names of novel and/or familiar concepts. Because studies where encoding was examined by means of changes in EEG power have also focused primarily on recognition tasks (e.g., Hanslmayr et al., 2009; Staudigl & Hanslmayr, 2013; Summerfield & Mangels, 2005), a parallel line of research could test Paller et al.’s hypothesis using the time-frequency equivalents of the ERP difference-due-to-memory effect to further elucidate the neurocognitive processes that generate it.

Finally, another possible interpretation of our findings is that the neural correlates of successful encoding are indeed domain-general and, therefore, similar for familiar and novel words,3 but that the experimental effects (on amplitude, time-frequency power, or cross-frequency coupling) associated with the encoding process may surface only under certain conditions (note that this interpretation assumes that the null effects in the present study truly reflect an absence of effects, an assumption that would require further empirical validation). In fact, in the literature on familiar words, reports of null results are not uncommon: for instance, Johnson et al. (1985) and Paller et al. (1987) failed to observe differences in ERP amplitudes as a function of later memory. Hanslmayr et al. (2009) contrasted time-frequency power representations for later remembered as opposed to later forgotten familiar words across two incidental-learning tasks: one group of participants were required to make semantic judgements for the to-be-remembered words (living/nonliving object; deep encoding group), while the other group had to judge whether the first and the last letters of the to-be-remembered words were in alphabetical order (shallow encoding group). For the deep encoding group, no differences between the two conditions were found in the theta frequency band, whereas, for the shallow encoding group, null results were reported for beta and gamma frequency bands. These findings seem to suggest that different encoding tasks may be associated with different oscillatory subsequent-memory effects (Hanslmayr et al., 2009).

In addition, if the assumption is true that changes in neural oscillatory dynamics during encoding reflect domain-general memory processes (e.g., Hanslmayr et al., 2012; Nyhus & Curran, 2010), similar effects should be found in encoding of other types of stimuli. However, here too, there have been both positive and null results, sometimes even within one study. For example, Osipova et al. (2006) reported increased synchronisation in the theta and gamma bands for later recognised compared to not recognised images, however, no differences between the two conditions were observed in the alpha and beta bands. This is in contrast with a study in which the participants were exposed to words superimposed on movie clips and later tested on their memory for the word-movie pairs in a surprise recognition test (Staudigl & Hanslmayr, 2013). Contrary to expectation, differences between the successfully recognised versus unrecognised items were found in the theta frequency band (albeit only for frequencies in the range between 3.5 Hz and 4.5 Hz over a small cluster of MEG sensors) but not in the gamma frequency band (Staudigl & Hanslmayr, 2013). Of course, one could argue that different neurocognitive processes may underlie encoding of pictorial and linguistic stimuli, and this is a view that we share. However, results such as those in Osipova et al. (2006) and Staudigl & Hanslmayr (2013) could also indicate that there are no fundamental differences in encoding of novel and familiar words or other types of stimuli. Therefore, based on these and other findings in the literature, one tentative conclusion could be that the most prominent theoretical accounts of encoding were developed based only on a subset of available findings (i.e., those where the effects were observed).

Consequently, promising directions for future research would be, firstly, to test whether the previous findings (both those where the effects were present and those where they were absent) can be replicated and, secondly, to re-evaluate whether the existing theories of encoding (e.g., Hanslmayr et al., 2012; Nyhus & Curran, 2010) can indeed account for most of the findings in the literature. Depending on the outcome of these two research streams, these theories could then be constrained by examining under which conditions they apply. One possibility would be to directly contrast incidental and intentional learning paradigms. Most previous studies examined incidental learning with a very large number of stimuli, while, in our study, participants were explicitly told to try to remember the novel words. Consequently, one working hypothesis could be that the difference-due-to-memory effects in ERP and time-frequency data may characterise primarily unintentional learning and possibly under challenging learning conditions.

Another possibility could be that the presence and the strength of effects may vary with the analysis approach. For example, in their seminal study, Klimesch, Schimke, et al. (1996) split the frequencies in the alpha range into lower and upper alpha bands and determined the frequency windows for these two bands separately for each participant by using the mean peak frequency across all electrodes as an anchor point (and see, e.g., Demiralp et al., 2007, for a similar approach). The motivation behind this approach was that frequencies of the macroscopic oscillations have been found to vary as a function of factors such as age, memory performance, and attentional demands (e.g., Klimesch, Schimke, et al., 1996; Kwon et al., 2015; Strunk et al., 2017). This finding is in line with accounts in the literature that attribute modulations of the alpha rhythm to inhibition of task-irrelevant areas of the brain to achieve optimal task performance (e.g., Gips et al., 2016; Jensen & Mazaheri, 2010). This, together with the fact that patterns of oscillatory activity may vary greatly across individuals, suggests that clearer criteria are needed regarding when differently operationalised measures can be considered to reflect the same underlying process. This issue of the validity of the measures used in the literature is particularly prominent in studies examining encoding-related oscillatory activity in the gamma band as well as theta-gamma phase-amplitude coupling, with extremely variable gamma band definitions (e.g., 30-50 Hz in Sauseng et al., 2008, 44-64 Hz in Sederberg et al., 2007, 50-80 Hz in Friese et al., 2013, 30-80 Hz in Demiralp et al., 2007, 55-70 Hz in Hanslmayr et al., 2009, 60-90 Hz in Osipova et al., 2006, > 25 Hz in Gruber et al., 2004).

Having discussed three possible accounts of our results, we would like to acknowledge yet again that the outcome of our study could potentially be attributed to other factors and that many questions remain to be addressed in the future research. For instance, as we note in the Methods section, it is still unclear at what point in time brain activity patterns for words that are later remembered begin to differentiate from those for words that are later forgotten. In our study, we considered brain responses during the second, the third, and the fourth exposures to the novel words, however, it is possible that differences in encoding between the words that were later recalled and those that were not recalled were present at the first but not at the subsequent exposures. Our study was not designed to test this hypothesis, yet we think that it would be insightful to examine this issue in follow-up studies. This could be combined with an investigation of individual differences in the time course of the encoding process, particularly when the training protocol includes repeated exposure (as was the case in the present study).

To conclude, based on the theoretical claims of the previous studies, our logical prediction was that the effects reported in these studies should be observed also with novel words, and the purpose of our exploratory study was to test this prediction. Our study failed to confirm these predictions, and we have provided three possible interpretations of our results. Firstly, lack of significant effects in our study could be due to low signal-to-noise ratio, but, given that some other studies have reported significant results with a comparable number of data points, this suggestion needs to be confirmed with high-power pre-registered studies. Secondly, our results could be due to genuine differences in encoding of familiar and novel words, and we have discussed potential reasons for these differences, while also laying out some ideas for future research. Finally, we have discussed a possibility that the neural correlates of encoding are domain-general but that the experimental effects typically interpreted as reflecting these domain-general encoding processes may manifest only under certain conditions. We have further suggested that, due to a great deal of variability across the previous studies with respect to methods and results, our findings might still be in accordance with those of the previous studies, but that, in light of these inconsistencies, the existing theories may need to be further constrained. Critically, in our view, these issues highlight the complex and multifaceted nature of the memory literature, while they also foreground the need for pre-registered replications to better understand what drives some of the reported effects. To our knowledge, the present study is the first of such pre-registered attempts, and we hope that more such studies will appear in the future.

Maria Korochkina: Conceptualisation, methodology, software, validation, formal analysis, investigation, data curation, writing (original draft), writing (review and editing), visualisation.

Paul F. Sowman: Formal analysis, writing (review and editing).

Lyndsey Nickels: Conceptualisation, methodology, validation, writing (review and editing), supervision.

Audrey Bürki: Conceptualisation, methodology, validation, writing (review and editing), supervision.

The authors would like to thank Andrew Roberts for his help with recording of the stimuli.

MK was supported by a PhD stipend from Macquarie University [CTiMQRES IDEALAB; Allocation number 2019129]. Costs for participant reimbursement were covered by the Cognitive Science Postgraduate Grant (CPGR4 2020) to MK. LN received support from an Australian Research Council Discovery Project Grant [DP190101490]. Funding sources had no involvement in study design, collection, analysis, and interpretation of data, in the writing of the report or decision to submit the article for publication.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Pre-registration, stimuli, participant data, analysis scripts, and pre-print are publicly available on this paper’s project page on the OSF platform at https://osf.io/g7ftz/ (doi: 10.17605/OSF.IO/G7FTZ).

1.

We did not pre-register a sample size for this study because we were not able to conduct a power analysis due to lack of information necessary for data simulation (see Korochkina et al., under review, for more detail). Nevertheless, we pre-registered that all data would be collected within a 6-month time window.

2.

We note that it is unclear at what point in time differences emerge in encoding of items that are later remembered as opposed to forgotten and, consequently, differences between the recalled and not recalled words in our study could have been present at the first exposure (1) but absent at subsequent exposures (2–4). Our study was not designed to test this hypothesis: (a) with 20 novel names per participant, there were not enough data points to compare, for exposure 1, the EEG responses for those novel names that were later recalled with EEG responses for those novel names that were not recalled; (b) at the time of the first exposure, the participants had not yet learnt anything about the meaning of the novel words, and consequently, it would not be appropriate to use performance on the recall task where novel names were cued by their definitions, as a predictor in this analysis.

3.

Note, however, that, while domain-general learning theories suggest that similar mechanisms support learning of information of different types, it is unclear whether domain-generality necessarily implies that every domain would show identical effects. This issue is relevant not just for the present study but for the field of cognitive science in general, and we hope that it will be addressed in future research.

Addante, R. J., Ranganath, C., & Yonelinas, A. P. (2012). Examining ERP correlates of recognition memory: Evidence of accurate source recognition without recollection. NeuroImage, 62(1), 439–450. https://doi.org/10.1016/j.neuroimage.2012.04.031
Batterink, L., & Neville, H. (2011). Implicit and explicit mechanisms of word learning in a narrative context: An event-related potential study. Journal of Cognitive Neuroscience, 23, 3181--3196. https://doi.org/10.1162/jocn\_a\_00013
Berman, J. I., McDaniel, J., Liu, S., Cornew, L., Gaetz, W., Roberts, T. P. L., & Edgar, J. C. (2012). Variable bandwidth filtering for improved sensitivity of cross-frequency coupling metrics. Brain Connectivity, 2(3), 155–163. https://doi.org/10.1089/brain.2012.0085
Bigdely-Shamlo, N., Mullen, T., Kothe, C., Su, K.-M., & Robbins, K. A. (2015). The PREP pipeline: Standardized preprocessing for large-scale EEG analysis. Frontiers in Neuroinformatics, 9(16). https://doi.org/10.3389/fninf.2015.00016
Buzsáki, G. (2006). Rhythms of the Brain. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780195301069.001.0001
Buzsáki, G., & Draguhn, A. (2004). Neuronal oscillations in cortical networks. Science, 304(5679), 1926–1929. https://doi.org/10.1126/science.1099745
Canolty, R. T., & Knight, R. T. (2010). The functional role of cross-frequency coupling. Trends in Cognitive Sciences, 14(11), 506–515. https://doi.org/10.1016/j.tics.2010.09.001
Caplan, J. B., & Glaholt, M. G. (2007). The roles of EEG oscillations in learning relational information. NeuroImage, 38(3), 604–616. https://doi.org/10.1016/j.neuroimage.2007.07.054
Chaumon, M., Bishop, D. V. M., & Busch, N. A. (2015). A practical guide to the selection of independent components of the electroencephalogram for artifact correction. Journal of Neuroscience Methods, 250, 47–63. https://doi.org/10.1016/j.jneumeth.2015.02.025
Cohen, M. X. (2014). Analyzing Neural Time Series Data. The MIT Press. https://doi.org/10.7551/mitpress/9609.001.0001
Craik, F. I. M., Hasselmo, M. E., & Davachi, L. (2007). Encoding. In H. L. Roediger, Y. Dudai, & S. M. Fitzpatrick (Eds.), Science of Memory: Concepts (pp. 122--144). Oxford University Press.
Curran, T. (1999). The electrophysiology of incidental and intentionalretrieval: ERP old/new effects in lexical decision andrecognition memory. Neuropsychologia, 37(7), 771–785. https://doi.org/10.1016/s0028-3932(98)00133-x
Davachi, L., & Dobbins, I. G. (2008). Declarative memory. Current Directions in Psychological Science, 17(2), 112–118. https://doi.org/10.1111/j.1467-8721.2008.00559.x
Davis, M. H., & Gaskell, M. G. (2009). A complementary systems account of word learning: Neural and behavioural evidence. Philosophical Transactions of the Royal Society B: Biological Sciences, 364(1536), 3773–3800. https://doi.org/10.1098/rstb.2009.0111
Delorme, A., & Makeig, S. (2004). EEGLAB: An open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of Neuroscience Methods, 134(1), 9–21. https://doi.org/10.1016/j.jneumeth.2003.10.009
Demiralp, T., Bayraktaroglu, Z., Lenz, D., Junge, S., Busch, N. A., Maess, B., Ergen, M., & Herrmann, C. S. (2007). Gamma amplitudes are coupled to theta phase in human EEG during visual perception. International Journal of Psychophysiology, 64(1), 24–30. https://doi.org/10.1016/j.ijpsycho.2006.07.005
Diana, R. A., Yonelinas, A. P., & Ranganath, C. (2007). Imaging recollection and familiarity in the medial temporal lobe: A three-component model. Trends in Cognitive Sciences, 11(9), 379–386. https://doi.org/10.1016/j.tics.2007.08.001
Dickerson, B. C., & Eichenbaum, H. (2009). The Episodic Memory System: Neurocircuitry and Disorders. Neuropsychopharmacology, 35(1), 86–104. https://doi.org/10.1038/npp.2009.126
Düzel, E., Penny, W. D., & Burgess, N. (2010). Brain oscillations and memory. Current Opinion in Neurobiology, 20(2), 143–149. https://doi.org/10.1016/j.conb.2010.01.004
Eichenbaum, H., Yonelinas, A. P., & Ranganath, C. (2007). The medial temporal lobe and recognition memory. Annual Review of Neuroscience, 30(1), 123–152. https://doi.org/10.1146/annurev.neuro.30.051606.094328
Fell, J., & Axmacher, N. (2011). The role of phase synchronization in memory processes. Nature Reviews Neuroscience, 12(2), 105–118. https://doi.org/10.1038/nrn2979
Fell, J., Klaver, P., Lehnertz, K., Grunwald, T., Schaller, C., Elger, C. E., & Fernández, G. (2001). Human memory formation is accompanied by rhinal–hippocampal coupling and decoupling. Nature Neuroscience, 4(12), 1259–1264. https://doi.org/10.1038/nn759
Fellner, M.-C., Bäuml, K.-H. T., & Hanslmayr, S. (2013). Brain oscillatory subsequent memory effects differ in power and long-range synchronization between semantic and survival processing. NeuroImage, 79, 361–370. https://doi.org/10.1016/j.neuroimage.2013.04.121
Friedman, D., & Johnson, R. (2000). Event-related potential (ERP) studies of memory encoding and retrieval: A selective review. Microscopy Research and Technique, 51(1), 6–28. https://doi.org/10.1002/1097-0029(20001001)51:1
Fries, P. (2005). A mechanism for cognitive dynamics: Neuronal communication through neuronal coherence. Trends in Cognitive Sciences, 9(10), 474–480. https://doi.org/10.1016/j.tics.2005.08.011
Friese, U., Köster, M., Hassler, U., Martens, U., Trujillo-Barreto, N., & Gruber, T. (2013). Successful memory encoding is associated with increased cross-frequency coupling between frontal theta and posterior gamma oscillations in human scalp-recorded EEG. NeuroImage, 66, 642–647. https://doi.org/10.1016/j.neuroimage.2012.11.002
Gips, B., Eerden, J. P. J. M., & Jensen, O. (2016). A biologically plausible mechanism for neuronal coding organized by the phase of alpha oscillations. European Journal of Neuroscience, 44(4), 2147–2161. https://doi.org/10.1111/ejn.13318
Goard, M., & Dan, Y. (2009). Basal forebrain activation enhances cortical coding of natural scenes. Nature Neuroscience, 12(11), 1444–1449. https://doi.org/10.1038/nn.2402
Groppe, D. M., Urbach, T. P., & Kutas, M. (2011a). Mass univariate analysis of event-related brain potentials/fields I: A critical tutorial review. Psychophysiology, 48(12), 1711–1725. https://doi.org/10.1111/j.1469-8986.2011.01273.x
Groppe, D. M., Urbach, T. P., & Kutas, M. (2011b). Mass univariate analysis of event-related brain potentials/fields II: Simulation studies. Psychophysiology, 48(12), 1726–1737. https://doi.org/10.1111/j.1469-8986.2011.01272.x
Gruber, T., Tsivilis, D., Montaldi, D., & Müller, M. M. (2004). Induced gamma band responses: An early marker of memory encoding and retrieval. NeuroReport, 15(11), 1837–1841. https://doi.org/10.1097/01.wnr.0000137077.26010.12
Hanslmayr, S., Axmacher, N., & Inman, C. S. (2019). Modulating human memory via entrainment of brain oscillations. Trends in Neurosciences, 42(7), 485–499. https://doi.org/10.1016/j.tins.2019.04.004
Hanslmayr, S., Spitzer, B., & Bauml, K.-H. (2009). Brain Oscillations Dissociate between Semantic and Nonsemantic Encoding of Episodic Memories. Cerebral Cortex, 19(7), 1631–1640. https://doi.org/10.1093/cercor/bhn197
Hanslmayr, S., Staresina, B. P., & Bowman, H. (2016). Oscillations and Episodic Memory: Addressing the Synchronization/Desynchronization Conundrum. Trends in Neurosciences, 39(1), 16–25. https://doi.org/10.1016/j.tins.2015.11.004
Hanslmayr, S., & Staudigl, T. (2014). How brain oscillations form memories — A processing based perspective on oscillatory subsequent memory effects. NeuroImage, 85, 648–655. https://doi.org/10.1016/j.neuroimage.2013.05.121
Hanslmayr, S., Staudigl, T., & Fellner, M.-C. (2012). Oscillatory power decreases and long-term memory: The information via desynchronization hypothesis. Frontiers in Human Neuroscience, 6, 74. https://doi.org/10.3389/fnhum.2012.00074
Hanslmayr, S., Volberg, G., Wimber, M., Raabe, M., Greenlee, M. W., & Bauml, K.-H. T. (2011). The Relationship between Brain Oscillations and BOLD Signal during Memory Formation: A Combined EEG-fMRI Study. Journal of Neuroscience, 31(44), 15674–15680. https://doi.org/10.1523/jneurosci.3140-11.2011
Hawkins, E., Astle, D. E., & Rastle, K. (2015). Semantic advantage for learning new phonological form representations. Journal of Cognitive Neuroscience, 27, 775–786. https://doi.org/10.1162/jocn\_a\_00730
Hurtado, J. M., Rubchinsky, L. L., & Sigvardt, K. A. (2004). Statistical method for detection of phase-locking episodes in neural oscillations. Journal of Neurophysiology, 91(4), 1883–1898. https://doi.org/10.1152/jn.00853.2003
James, E., Gaskell, M. G., & Henderson, L. M. (2019). Offline consolidation supersedes prior knowledge benefits in children’s (but not adults’) word learning. Developmental Science, 22(3), e12776. https://doi.org/10.1111/desc.12776
James, E., Gaskell, M. G., Weighall, A., Henderson, L. (2017). Consolidation of vocabulary during sleep: The rich get richer? Neuroscience Biobehavioral Reviews, 77, 1–13. https://doi.org/10.1016/j.neubiorev.2017.01.054
Jaramillo, J., Kempter, R. (2017). Phase precession: A neural code underlying episodic memory? Current Opinion in Neurobiology, 43, 130–138. https://doi.org/10.1016/j.conb.2017.02.006
Jensen, O., Lisman, J. E. (2005). Hippocampal sequence-encoding driven by a cortical multi-item working memory buffer. Trends in Neurosciences, 28(2), 67–72. https://doi.org/10.1016/j.tins.2004.12.001
Jensen, O., Mazaheri, A. (2010). Shaping Functional Architecture by Oscillatory Alpha Activity: Gating by Inhibition. Frontiers in Human Neuroscience, 4, 186. https://doi.org/10.3389/fnhum.2010.00186
Johnson, R., Pfefferbaum, A., Kopell, B. S. (1985). P300 and Long-Term Memory: Latency Predicts Recognition Performance. Psychophysiology, 22(5), 497–507. https://doi.org/10.1111/j.1469-8986.1985.tb01639.x
Jurkiewicz, G. J., Hunt, M. J., Żygierewicz, J. (2020). Addressing Pitfalls in Phase-Amplitude Coupling Analysis with an Extended Modulation Index Toolbox. Neuroinformatics, 19(2), 319–345. https://doi.org/10.1007/s12021-020-09487-3
Klimesch, W., Doppelmayr, M., Russegger, H., Pachinger, T. (1996). Theta band power in the human scalp EEG and the encoding of new information. Neuroreport, 7(7), 1235–1240. https://doi.org/10.1097/00001756-199605170-00002
Klimesch, W., Schimke, H., Doppelmayr, M., Ripper, B., Schwaiger, J., Pfurtscheller, G. (1996). Event-related desynchronization (ERD) and the Dm effect: Does alpha desynchronization during encoding predict later recall performance? International Journal of Psychophysiology, 24(1–2), 47–60. https://doi.org/10.1016/s0167-8760(96)00054-2
Korochkina, M., Bürki, A., Nickels, L. (2021). Apples and oranges: How does learning context affect novel word learning? Journal of Memory and Language, 120, 104246. https://doi.org/10.1016/j.jml.2021.104246
Korochkina, M., Nickels, L., Bürki, A. (under review). An investigation of behavioural and electrophysiological markers of integration in learning of novel names for novel concepts. School of Psychological Sciences, Macquarie University. https://psyarxiv.com/vup25/
Köster, M., Friese, U., Schöne, B., Trujillo-Barreto, N., Gruber, T. (2014). Theta–gamma coupling during episodic retrieval in the human EEG. Brain Research, 1577, 57–68. https://doi.org/10.1016/j.brainres.2014.06.028
Kovács, K. A. (2020). Episodic Memories: How do the Hippocampus and the Entorhinal Ring Attractors Cooperate to Create Them? Frontiers in Systems Neuroscience, 14, 68. https://doi.org/10.3389/fnsys.2020.559186
Kramer, M. A., Tort, A. B. L., Kopell, N. J. (2008). Sharp edge artifacts and spurious coupling in EEG frequency comodulation measures. Journal of Neuroscience Methods, 170(2), 352–357. https://doi.org/10.1016/j.jneumeth.2008.01.020
Kullback, S., Leibler, R. A. (1951). On information and sufficiency. The Annals of Mathematical Statistics, 22(1), 79–86. https://doi.org/10.1214/aoms/1177729694
Kumar, A. A. (2020). Semantic memory: A review of methods, models, and current challenges. Psychonomic Bulletin Review, 28(1), 40–80. https://doi.org/10.3758/s13423-020-01792-x
Kumaran, D., Hassabis, D., McClelland, J. L. (2016). What Learning Systems do Intelligent Agents Need? Complementary Learning Systems Theory Updated. Trends in Cognitive Sciences, 20(7), 512–534. https://doi.org/10.1016/j.tics.2016.05.004
Kwon, G., Lim, S., Kim, M.-Y., Kwon, H., Lee, Y.-H., Kim, K., Lee, E.-J., Suh, M. (2015). Individual differences in oscillatory brain activity in response to varying attentional demands during a word recall and oculomotor dual task. Frontiers in Human Neuroscience, 9, 381. https://doi.org/10.3389/fnhum.2015.00381
Le Van Quyen, M., Foucher, J., Lachaux, J.-P., Rodriguez, E., Lutz, A., Martinerie, J., Varela, F. J. (2001). Comparison of Hilbert transform and wavelet methods for the analysis of neuronal synchrony. Journal of Neuroscience Methods, 111(2), 83–98. https://doi.org/10.1016/s0165-0270(01)00372-7
Lega, B., Burke, J., Jacobs, J., Kahana, M. J. (2014). Slow-Theta-to-Gamma Phase–Amplitude Coupling in Human Hippocampus Supports the Formation of New Episodic Memories. Cerebral Cortex, 26(1), 268–278. https://doi.org/10.1093/cercor/bhu232
Lisman, J. E. (1999). Relating hippocampal circuitry to function: Recall of memory sequences by reciprocal dentate-CA3 interactions. Neuron, 22(2), 233–242. https://doi.org/10.1016/s0896-6273(00)81085-5
Lisman, J. E., Idiart, M. A. P. (1995). Storage of 7 ± 2 Short-Term Memories in Oscillatory Subcycles. Science, 267(5203), 1512–1515. https://doi.org/10.1126/science.7878473
Lisman, J. E., Jensen, O. (2013). The theta-gamma neural code. Neuron, 77(6), 1002–1016. https://doi.org/10.1016/j.neuron.2013.03.007
Liu, Y., Hell, J. G. (2020). Learning Novel Word Meanings: An ERP Study on Lexical Consolidation in Monolingual, Inexperienced Foreign Language Learners. Language Learning, 70(S2), 45–74. https://doi.org/10.1111/lang.12403
Maris, E., Oostenveld, R. (2007). Nonparametric statistical testing of EEG- and MEG-data. Journal of Neuroscience Methods, 164(1), 177–190. https://doi.org/10.1016/j.jneumeth.2007.03.024
Martinez-Cancino, R., Delorme, A., Kreutz-Delgado, K., Makeig, S. (2020). Computing Phase Amplitude Coupling in EEGLAB: PACTools. 2020 IEEE 20th International Conference on Bioinformatics and Bioengineering (BIBE), 387–394. https://doi.org/10.1109/bibe50027.2020.00070
McClelland, J. L. (2013). Incorporating rapid neocortical learning of new schema-consistent information into complementary learning systems theory. Journal of Experimental Psychology: General, 142(4), 1190–1210. https://doi.org/10.1037/a0033812
McClelland, J. L., McNaughton, B. L., Lampinen, A. K. (2020). Integration of new information in memory: New insights from a complementary learning systems perspective. Philosophical Transactions of the Royal Society B: Biological Sciences, 375(1799), 20190637. https://doi.org/10.1098/rstb.2019.0637
McClelland, J. L., McNaughton, B. L., O’Reilly, R. C. (1995). Why there are complementary learning systems in the hippocampus and neocortex: Insights from the successes and failures of connectionist models of learning and memory. Psychological Review, 102(3), 419–457. https://doi.org/10.1037/0033-295x.102.3.419
McMurray, B., Kapnoula, E., Gaskell, M. G. (2016). Learning and integration of new word-forms: Consolidation, pruning, and the emergence of automaticity. In M. G. Gaskell J. Mirkovic (Eds.), Speech Perception and Spoken Word Recognition (pp. 116–142). Psychology Press.
Medler, D. A., Binder, J. R. (2008). MCWord: An On-Line Orthographic Database of the English Language. http://www.neuro.mcw.edu/mcword/
Munia, T. T. K., Aviyente, S. (2019). Time-frequency based phase-amplitude coupling measure for neuronal oscillations. Scientific Reports, 9(1), 12441. https://doi.org/10.1038/s41598-019-48870-2
Münte, T. F., Urbach, T. P., Düzel, E., Kutas, M. (2000). Event-related brain potentials in the study of human cognition and neuropsychology. In F. Boller, J. Grafman, G. Rizzolatti (Eds.), Handbook of Neuropsychology (Vol. 1, pp. 139–235). Elsevier Science Publishers B.V.
Neville, H. J., Kutas, M., Chesney, G., Schmidt, A. L. (1986). Event-related brain potentials during initial encoding and recognition memory of congruous and incongruous words. Journal of Memory and Language, 25(1), 75–92. https://doi.org/10.1016/0749-596x(86)90022-7
Nyhus, E., Curran, T. (2010). Functional role of gamma and theta oscillations in episodic memory. Neuroscience Biobehavioral Reviews, 34(7), 1023–1035. https://doi.org/10.1016/j.neubiorev.2009.12.014
O’Keefe, J., Recce, M. L. (1993). Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus, 3(3), 317–330. https://doi.org/10.1002/hipo.450030307
Oostenveld, R., Fries, P., Maris, E., Schoffelen, J.-M. (2011). FieldTrip: Open Source Software for Advanced Analysis of MEG, EEG, and Invasive Electrophysiological Data. Computational Intelligence and Neuroscience, 2011, 1–9. https://doi.org/10.1155/2011/156869
O’Reilly, R. C., Bhattacharyya, R., Howard, M. D., Ketz, N. (2014). Complementary learning systems. Cognitive Science, 38(6), 1229–1248. https://doi.org/10.1111/j.1551-6709.2011.01214.x
Osipova, D., Takashima, A., Oostenveld, R., Fernandez, G., Maris, E., Jensen, O. (2006). Theta and gamma oscillations predict encoding and retrieval of declarative memory. Journal of Neuroscience, 26(28), 7523–7531. https://doi.org/10.1523/jneurosci.1948-06.2006
Paller, K. A. (1990). Recall and stem-completion priming have different electrophysiological correlates and are modified differentially by directed forgetting. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16(6), 1021–1032. https://doi.org/10.1037/0278-7393.16.6.1021
Paller, K. A., Kutas, M., Mayes, A. R. (1987). Neural correlates of encoding in an incidental learning paradigm. Electroencephalography and Clinical Neurophysiology, 67(4), 360–371. https://doi.org/10.1016/0013-4694(87)90124-6
Paller, K. A., Kutas, M., McIsaac, H. K. (1995). Monitoring conscious recollection via the electrical activity of the brain. Psychological Science, 6(2), 107–111. https://doi.org/10.1111/j.1467-9280.1995.tb00315.x
Paller, K. A., McCarthy, G., Wood, C. C. (1988). ERPs predictive of subsequent recall and recognition performance. Biological Psychology, 26(1–3), 269–276. https://doi.org/10.1016/0301-0511(88)90023-3
Paller, K. A., Wagner, A. D. (2002). Observing the transformation of experience into memory. Trends in Cognitive Sciences, 6(2), 93–102. https://doi.org/10.1016/s1364-6613(00)01845-3
Pernet, C. R., Chauveau, N., Gaspar, C., Rousselet, G. A. (2011). LIMO EEG: A Toolbox for Hierarchical LInear MOdeling of ElectroEncephaloGraphic Data. Computational Intelligence and Neuroscience, 2011, 1–11. https://doi.org/10.1155/2011/831409
Pernet, C. R., Latinus, M., Nichols, T. E., Rousselet, G. A. (2015). Cluster-based computational methods for mass univariate analyses of event-related brain potentials/fields: A simulation study. Journal of Neuroscience Methods, 250(Supplement C), 85–93. https://doi.org/10.1016/j.jneumeth.2014.08.003
Pfurtscheller, G., Aranibar, A. (1977). Event-related cortical desynchronization detected by power measurements of scalp EEG. Electroencephalography and Clinical Neurophysiology, 42(6), 817–826. https://doi.org/10.1016/0013-4694(77)90235-8
Pfurtscheller, G., Lopes da Silva, F. H. (1999). Event-related EEG/MEG synchronization and desynchronization: Basic principles. Clinical Neurophysiology, 110(11), 1842–1857. https://doi.org/10.1016/s1388-2457(99)00141-8
Pion-Tonachini, L., Kreutz-Delgado, K., Makeig, S. (2019). ICLabel: An automated electroencephalographic independent component classifier, dataset, and website. NeuroImage, 198, 181–197. https://doi.org/10.1016/j.neuroimage.2019.05.026
Qasim, S. E., Fried, I., Jacobs, J. (2021). Phase precession in the human hippocampus and entorhinal cortex. Cell, 184(12), 3242-3255.e10. https://doi.org/10.1016/j.cell.2021.04.017
Röder, B., Rösler, F., Neville, H. J. (2001). Auditory memory in congenitally blind adults: A behavioral-electrophysiological investigation. Cognitive Brain Research, 11(2), 289–303. https://doi.org/10.1016/s0926-6410(01)00002-7
Rugg, M. D. (1995). ERP studies of memory. In M. D. Rugg M. G. H. Coles (Eds.), Electrophysiology of Mind: Event-related Brain Potentials and Cognition (pp. 132–170). Oxford University Press.
Rugg, M. D., Curran, T. (2007). Event-related potentials and recognition memory. Trends in Cognitive Sciences, 11(6), 251–257. https://doi.org/10.1016/j.tics.2007.04.004
Santner, T. J., Duffy, D. E. (1989). The Statistical Analysis of Discrete Data. Springer New York. https://doi.org/10.1007/978-1-4612-1017-7
Sauseng, P., Klimesch, W., Gruber, W. R., Birbaumer, N. (2008). Cross-frequency phase synchronization: A brain mechanism of memory matching and attention. NeuroImage, 40(1), 308–317. https://doi.org/10.1016/j.neuroimage.2007.11.032
Schneidman, E., Puchalla, J. L., Segev, R., Harris, R. A., Bialek, W., Berry, M. J. (2011). Synergy from silence in a combinatorial neural code. Journal of Neuroscience, 31(44), 15732–15741. https://doi.org/10.1523/jneurosci.0301-09.2011
Schott, B., Richardson-Klavehn, A., Heinze, H.-J., Düzel, E. (2002). Perceptual Priming Versus Explicit Memory: Dissociable Neural Correlates at Encoding. Journal of Cognitive Neuroscience, 14(4), 578–592. https://doi.org/10.1162/08989290260045828
Sederberg, P. B., Gauthier, L. V., Terushkin, V., Miller, J. F., Barnathan, J. A., Kahana, M. J. (2006). Oscillatory correlates of the primacy effect in episodic memory. NeuroImage, 32(3), 1422–1431. https://doi.org/10.1016/j.neuroimage.2006.04.223
Sederberg, P. B., Kahana, M. J., Howard, M. W., Donner, E. J., Madsen, J. R. (2003). Theta and gamma oscillations during encoding predict subsequent recall. The Journal of Neuroscience, 23(34), 10809–10814. https://doi.org/10.1523/jneurosci.23-34-10809.2003
Sederberg, P. B., Schulze-Bonhage, A., Madsen, J. R., Bromfield, E. B., McCarthy, D. C., Brandt, A., Tully, M. S., Kahana, M. J. (2007). Hippocampal and neocortical gamma oscillations predict memory formation in humans. Cerebral Cortex, 17(5), 1190–1196. https://doi.org/10.1093/cercor/bhl030
Seymour, R. A., Rippon, G., Kessler, K. (2017). The detection of phase amplitude coupling during sensory processing. Frontiers in Neuroscience, 11, 487. https://doi.org/10.3389/fnins.2017.00487
Sinkkonen, J., Tiitinen, H., Näätänen, R. (1995). Gabor filters: An informative way for analysing event-related brain activity. Journal of Neuroscience Methods, 56(1), 99–104. https://doi.org/10.1016/0165-0270(94)00111-s
Smith, S., Nichols, T. (2009). Threshold-free cluster enhancement: Addressing problems of smoothing, threshold dependence and localisation in cluster inference. NeuroImage, 44(1), 83–98. https://doi.org/10.1016/j.neuroimage.2008.03.061
Staudigl, T., Hanslmayr, S. (2013). Theta oscillations at encoding mediate the context-dependent nature of human episodic memory. Current Biology, 23(12), 1101–1106. https://doi.org/10.1016/j.cub.2013.04.074
Stevens, M., Brysbaert, M. (2016). When do we have enough power in language research? Evidence from priming studies. Ghent University. http://crr.ugent.be/papers/When\%20do\%20we\%20have\%20enough\%20power\%20in\%20language\%20research.pdf
Strunk, J., James, T., Arndt, J., Duarte, A. (2017). Age-related changes in neural oscillations supporting context memory retrieval. Cortex, 91, 40–55. https://doi.org/10.1016/j.cortex.2017.01.020
Summerfield, C., Mangels, J. A. (2005). Coherent theta-band EEG activity predicts item-context binding during encoding. NeuroImage, 24(3), 692–703. https://doi.org/10.1016/j.neuroimage.2004.09.012
Teyler, T. J., DiScenna, P. (1986). The hippocampal memory indexing theory. Behavioral Neuroscience, 100(2), 147–154. https://doi.org/10.1037/0735-7044.100.2.147
Teyler, T. J., Rudy, J. W. (2007). The hippocampal indexing theory and episodic memory: Updating the index. Hippocampus, 17(12), 1158–1169. https://doi.org/10.1002/hipo.20350
Tort, A. B. L., Komorowski, R., Eichenbaum, H., Kopell, N. (2010). Measuring phase-amplitude coupling between neuronal oscillations of different frequencies. Journal of Neurophysiology, 104(2), 1195–1210. https://doi.org/10.1152/jn.00106.2010
Tort, A. B. L., Komorowski, R. W., Manns, J. R., Kopell, N. J., Eichenbaum, H. (2009). Theta–gamma coupling increases during the learning of item–context associations. Proceedings of the National Academy of Sciences, 106(49), 20942–20947. https://doi.org/10.1073/pnas.0911331106
Tort, A. B. L., Kramer, M. A., Thorn, C., Gibson, D. J., Kubota, Y., Graybiel, A. M., Kopell, N. J. (2008). Dynamic cross-frequency couplings of local field potential oscillations in rat striatum and hippocampus during performance of a T-maze task. Proceedings of the National Academy of Sciences, 105(51), 20517–20522. https://doi.org/10.1073/pnas.0810524105
Voloh, B., Valiante, T. A., Everling, S., Womelsdorf, T. (2015). Theta–gamma coordination between anterior cingulate and prefrontal cortex indexes correct attention shifts. Proceedings of the National Academy of Sciences, 112(27), 8457–8462. https://doi.org/10.1073/pnas.1500438112
von Stein, A., Sarnthein, J. (2000). Different frequencies for different scales of cortical integration: From local gamma to long range alpha/theta synchronization. International Journal of Psychophysiology, 38(3), 301–313. https://doi.org/10.1016/s0167-8760(00)00172-0
Wagner, A. D., Koutstaal, W., Schacter, D. L. (1999). When encodong yields remembering: Insights from event-related neuroimaging. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 354(1387), 1307–1324. https://doi.org/10.1098/rstb.1999.0481
Walker, S., Gaskell, M. G., Knowland, V. C. P., Fletcher, F. E., Cairney, S. A., Henderson, L. M. (2020). Growing up with interfering neighbours: The influence of time of learning and vocabulary knowledge on written word learning in children. Royal Society Open Science, 7(3), 191597. https://doi.org/10.1098/rsos.191597
Walker, S., Henderson, L. M., Fletcher, F. E., Knowland, V. C. P., Cairney, S. A., Gaskell, M. G. (2019). Learning to live with interfering neighbours: The influence of time of learning and level of encoding on word learning. Royal Society Open Science, 6(4), 181842. https://doi.org/10.1098/rsos.181842
Weighall, A. R., Henderson, L. M., Barr, D. J., Cairney, S. A., Gaskell, M. G. (2017). Eye-tracking the time‐course of novel word learning and lexical competition in adults and children. Brain and Language, 167, 13–27. https://doi.org/10.1016/j.bandl.2016.07.010
Weiss, S., Rappelsberger, P. (2000). Long-range EEG synchronization during word encoding correlates with successful memory performance. Cognitive Brain Research, 9(3), 299–312. https://doi.org/10.1016/s0926-6410(00)00011-2
Yarkoni, T., Balota, D., Yap, M. (2008). Moving beyond Coltheart’s N: A new measure of orthographic similarity. Psychonomic Bulletin Review, 15(5), 971–979. https://doi.org/10.3758/pbr.15.5.971
Yonelinas, A. P. (2002). The Nature of Recollection and Familiarity: A Review of 30 Years of Research. Journal of Memory and Language, 46(3), 441–517. https://doi.org/10.1006/jmla.2002.2864
Yonelinas, A. P., Aly, M., Wang, W.-C., Koen, J. D. (2010). Recollection and familiarity: Examining controversial assumptions and new directions. Hippocampus, 20(11), 1178–1194. https://doi.org/10.1002/hipo.20864
This is an open access article distributed under the terms of the Creative Commons Attribution License (4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Supplementary Material