Adapting Event-Related Potential Research Paradigms for Children: Considerations from Research on the Development of Recognition Memory

Abstract

Most developmental event-related potential (ERP) research uses experimental paradigms modified from research with adults (Brooker et al., 2020). One major challenge is identifying how to adapt these paradigms effectively for use with younger individuals. The current paper provides guidance for developmental adaptations by considering research on the development of recognition memory. We provide a brief overview of recognition memory tasks and ERP components associated with recognition memory in children and adults. Then, we provide some general recommendations, discuss common differences between ERP studies of recognition memory in adults and children (e.g., the type of stimuli presented, response modalities), and provide suggestions for assessing the effect of task modifications on ERP components of interest. Specifically, we recommend (a) testing both children and adults on the modified paradigm to allow for a continuity of findings across development, (b) comparing children of different ages on the modified paradigm based on expectations regarding when developmental change occurs for the cognitive process of interest, and (c) empirically assessing the effect of methodological differences between paradigms. To illustrate the latter, we analyzed data from our lab comparing memory-related ERP components when children experienced a 1-day, 2-day, or 1-week delay between encoding and retrieval.

Most developmental event-related potential (ERP) research uses experimental paradigms that were modified from tasks originally designed for adults (Brooker et al., 2020). One challenge for developmental researchers is identifying how to effectively adapt ERP research paradigms for younger individuals. Many design considerations overlap with best practices for conducting developmental research generally (e.g., using age-appropriate stimuli and instructions; Horst & Flack, 2020) and conducting electrophysiological research in adults (e.g., maximizing trial numbers; see Luck, 2014; Picton et al., 2000). However, there are some nuances of developmental cognitive electrophysiology that are worthy of discussion. Identifying best practices for developmental ERP research is important given the increased use of this methodology in infants and children and the specific challenges presented by this type of research (e.g., the balance of maximizing trial numbers while minimizing fatigue, data loss, and attrition; Boudewyn et al., 2018).

The aim of the current manuscript is to provide guidance on how to effectively adapt ERP research paradigms designed for adults for use with children by using research on the development of recognition memory as an example. We will begin by providing a brief review of tasks used to assess recognition memory and ERP components that have been associated with recognition memory in adults and children. Then, we will consider the task design recommendations provided by Brooker et al. (2020) within the context of ERP studies on the development of recognition memory. We will also identify some common methodological differences between ERP studies of recognition memory in children versus adults and then provide specific suggestions for assessing the effect of task modifications on ERP components of interest. To preview our main points, we suggest children and adults both be tested on the modified paradigm when it is appropriate for both age groups to support the continuity of findings across development (e.g., Rollins & Riggins, 2013; 2018). However, if tasks are not appropriate for adults when modified for children (e.g., Riggins et al., 2013), development can be assessed by comparing children of different ages on the modified paradigm based on expectations regarding when developmental change occurs for the cognitive process of interest (Riggins & Rollins, 2015). Further, we suggest empirically assessing the effect of methodological differences between paradigms when feasible. We demonstrate the latter by reporting new analyses assessing the effect of retention delay on ERP correlates of recognition memory.

Brief Overview of Recognition Memory Tasks and ERP Components Associated with Recognition Memory in Adults and Children

A variety of tasks are used for behavioral and electrophysiological research on recognition memory and its development. Recognition memory tasks can either include discrete encoding and retrieval phases (i.e., the study-test format) or intertwine encoding and retrieval by repeating stimuli within a single list (i.e., the continuous recognition procedure, Hintzman, 1969; Rugg & Nagy, 1989). Encoding and retrieval may be intentional (i.e., individuals may be instructed to remember items or to remember whether an item was previously encountered) or incidental (i.e., individuals may not be trying to remember items or consciously retrieve them; e.g., Curran, 1999; Haese & Czernochowski, 2015). At test, individuals either make a recognition memory judgment to individual items presented one at a time (e.g., Wilding & Rugg, 1996) or a forced-choice judgment about which item out of a set of items was previously encountered (e.g., Voss et al., 2008). The type of recognition memory judgment provided varies across paradigms, including old/new judgments, old/new judgments along with confidence ratings (e.g., Addante et al., 2012), and judgments regarding the quality of the recognition memory decision. For example, some recognition memory tasks assess the contributions of recollection and familiarity to recognition memory. Recollection refers to memory for contextual details whereas familiarity refers to the global assessment of memory strength (Yonelinas, 2002). Recollection and familiarity are can be assessed by the remember/know procedure (Tulving, 1985; e.g., Trott et al., 1999), process dissociation procedure (Jacoby, 1991; Herron & Rugg, 2003), or objective assessments of memory for details learned during the experimental paradigm (e.g., memory for whether the word was originally presented as singular or plural; e.g., Curran, 2000).

Multiple ERP components have been associated with recognition memory. In adults, studies of encoding have identified a robust subsequent memory effect (i.e., Dm effect) across frontal, central, and parietal leads 400–900 ms poststimulus onset that is characterized by a larger amplitude response to subsequently recognized items than subsequently forgotten items (e.g., Friedman et al., 1996; Smith, 1993). The subsequent memory effect has also been associated with recollection, especially when recollection is indexed using Tulving’s (1985) remember/know paradigm (e.g., Friedman & Trott, 2000). Specifically, items given a “remember” judgment elicit a more positive-going response in comparison to items given a “know” judgment and missed items. Although relatively few developmental ERP studies have focused on encoding, these studies suggest that children also exhibit a subsequent memory effect (Geng et al., 2018; Rollins & Riggins, 2013; 2018). However, the latency, directionality, and topography of the subsequent memory effect changes across early childhood (Geng et al., 2018; Rollins & Riggins, 2013). By middle childhood, children show a subsequent memory effect that is influenced by subjective indices of recollection and comparable to the effect shown in adolescents and young adults (Rollins & Riggins, 2018).

The most common ERP components assessed during retrieval in adults are the FN400 and parietal old/new effect (for reviews see Friedman & Johnson, 2000; Rugg & Curran, 2007; Wilding, 2000). The FN400 (i.e., mid-frontal old/new effect) occurs 300–500 ms poststimulus onset and has been associated with familiarity (Curran, 2000; Rugg et al., 1998; Trott et al., 1999; Woodruff et al., 2006). For example, Curran (2000) showed that the FN400 was larger to new words than identical same-plurality words and similar different-plurality words. The left parietal old/new effect occurs 400–800 ms poststimulus onset and has been associated with recollection because the amplitude is higher for remembered and source-correct items than new, familiar, and source-incorrect items leads (for reviews see Rugg & Curran, 2007; Wilding, 2000). Studies of older children typically assess the FN400 and parietal old/new effect (e.g., Cycowicz et al., 2001; Czernochowski et al., 2009; Czernochowski et al., 2005; Mecklinger, Brunnemann, & Kipp, 2011; Rollins & Riggins, 2018; Sprondel et al., 2011, 2012). However, due to differences in the morphology of ERP waveforms observed in younger children, studies of early childhood typically assess the negative component (Nc) and late slow wave (LSW; Canada et al., 2020; Riggins et al., 2013; Riggins & Rollins, 2015). The Nc occurs approximately 350–800 ms and is typically maximal over frontal and central electrodes (for review see DeBoer et al., 2007). The Nc has been related to attention and is modulated by memory (Bauer et al., 2003; Marshall et al., 2002; Riggins et al., 2009; Rollins & Riggins, 2015). For example, novel items that were perceptually similar to previously encountered items elicited a larger (i.e., more negative) amplitude Nc response relative to those that were more distinct (Rollins & Riggins, 2015). The LSW occurs 900–1500 ms and has been associated with the retrieval of contextual details (e.g., Riggins & Rollins, 2015; Riggins et al., 2013). For example, Riggins et al. (2013) showed that the LSW was more positive in amplitude to source-correct items than source-incorrect and new items. Although the Nc and LSW show similar properties to the FN400 and parietal old/new effect, respectively, more evidence is needed to determine whether the Nc and LSW are precursors to the FN400 and parietal old/new effect. This would be an important step in connecting ERP research in children and adults, which often are disconnected in the literature.

General Recommendations for Adapting Laboratory Tasks for Children

General recommendations for adapting ERP studies from adults for children exist and include using practice trials, minimizing the number of trials, blocking trials to allow children breaks, equating task difficulty, and providing trial-by-trial feedback (Brooker et al., 2020). We discuss each of these below in detail and also provide additional guidance for how paradigms can be modified to be appropriate for children.

The inclusion of practice trials is common in developmental research to ensure that children understand task instructions. Practice trials are even more important when collecting electrophysiological data because they provide the opportunity to train children to minimize artifacts by remaining still, minimizing blinking, and relaxing their facial muscles (Brooker et al., 2020). For example, in our labs, we instruct and remind children throughout the task to sit “as still as a statue.” The use of practice trials in ERP studies of recognition memory differs based on the experimental paradigm. Practice trials are not typically used in passive viewing paradigms since children do not provide an overt behavioral response (e.g., Riggins et al., 2013; Riggins & Rollins, 2015). However, they are common in active retrieval tasks that require children to make an explicit memory judgment (e.g., Geng et al., 2018; Mecklinger et al., 2011; Rollins & Riggins, 2018). For tasks that involve a behavioral response, children can also practice how and when the response should be given (e.g., verbally after the stimulus has been removed from the screen so as to not contaminate the epoch used for averaging).

Second, tasks should minimize recording time to the extent possible (Brooker et al., 2020). ERP studies with adults often last multiple hours because higher trial numbers improve the signal-to-noise ratio of the ERP waveforms (Luck, 2014). However, recording for long durations with children reduces task-directed attention, increases movement-related artifacts, and increases the probability of children refusing to complete the study (Brooker et al., 2020). For this reason, ERP tasks for toddlers and preschool-aged children typically last approximately 10–15 minutes (Brooker et al., 2020; see Howard et al., 2020, for empirical example). One approach to minimize recording durations for recognition memory research has been to selectively record EEG only during encoding or retrieval (Canada et al., 2020; Geng et al., 2018; Riggins & Rollins, 2015; Riggins et al., 2013; Robey & Riggins, 2016; Rollins & Riggins, 2013; 2018) or divide encoding and retrieval across multiple testing sessions (e.g., Leventon et al., 2014). However, studies of children 7 years of age and older have been successful collecting EEG data for both encoding and retrieval during a testing session lasting approximately 2 hours (Czernochowski et al., 2005; 2009; Haese & Czernochowski, 2016; Mecklinger et al., 2011; Sprondel et al., 2012).

Trial numbers for tasks used for developmental ERP research must be carefully selected based on anticipated data loss, which is common in ERP studies with young children, and the nature of the experimental paradigm. ERP studies require multiple test trials per condition to obtain an adequate signal-to-noise ratio. DeBoer et al. (2005) recommended a minimum of 10 trials per condition for developmental research; however, the number of trials needed likely varies by the component of interest and age group tested and has yet to be empirically established for many ERP components during development (cf. Pontifex et al., 2010). Most ERP studies of recognition memory meet or exceed the recommendation of a minimum of 10 trials per condition. When designing an ERP task for use with children, a conservative estimate is that one will lose approximately 50% of the test trials due to artifacts. The experimental paradigm also determines the number of trials that children are able to complete. Researchers can achieve higher trial numbers by having very young children complete passive viewing paradigms rather than memory tasks that require an active response because progression during a passive viewing paradigm is not dependent on the child’s behavioral response. Another important consideration when selecting trial numbers is whether stimuli will be selected for inclusion based on behavioral performance. Some research has trial numbers defined within the task itself. For example, studies that use an oddball task compare ERPs associated with standards and targets (Johnstone et al., 1996) and those of emotional expression may compare ERPs elicited to happy, angry, sad, and fearful faces (Moulson et al., 2009). However, trials for memory research are typically sorted into conditions based on behavioral performance. For example, studies of encoding typically separate stimuli into conditions based on whether they were subsequently recognized (i.e., hits) or forgotten (i.e., misses; e.g., Rollins & Riggins, 2013). This problem is exacerbated in studies of recollection (i.e., memory for details) because hits are further sorted into conditions based on accuracy for contextual details (e.g., location in which an item was previously encountered; Riggins & Rollins, 2015; Riggins et al., 2013).

If tasks require longer recording durations or greater trial numbers, performance and the number of trials available for analysis can be improved by dividing trials into multiple blocks or allowing children frequent breaks. The studies of recognition memory in older children that recorded for approximately 2 hours typically included multiple blocks (Czernochowski et al., 2005, 2009; Haese & Czernochowski et al., 2016; Mecklinger et al., 2011; Sprondel et al., 2012). Multiple blocks are particularly beneficial if researchers are concerned about participants exhibiting floor-level performance on a task. Even studies on face recognition in adults often include multiple study-test blocks for this reason (e.g., Rollins et al., 2020; Wiese et al., 2008). However, multiple blocks can be problematic if experience with the first block could influence the cognitive processes engaged in subsequent blocks. For example, an advantage associated with one block is that it allows for incidental encoding, which minimizes the possibility that developmental improvements in strategy use could account for age-related differences in recognition memory performance. If participants complete multiple blocks of a memory task, it is not possible to use incidental encoding because participants will know they are completing a memory task and what information they will be asked to recollect when they perform the second block. Another disadvantage of multiple blocks is that attrition may be higher; we have previously experienced that multiple blocks reduced children’s willingness to complete the entire experimental paradigm. If using a single block, trial numbers can be increased by allowing frequent breaks. For example, Mecklinger et al. (2011) gave participants a break every 15 trials. Similarly, in one study, children saw a familiar character every 35 pictures and received a sticker for a sticker sheet that depicted their progress on the task (Riggins & Rollins, 2015). The inclusion of short breaks is intended to improve children’s sustained attention during EEG recording by temporarily reducing mental fatigue and increasing their motivation to complete the task. However, breaks inherently increase the duration of the EEG recording, so the pros and cons of each option should be carefully considered.

A third recommendation is to equate task difficulty across participants (Brooker et al., 2020). If individual differences in performance are present, task difficulty may account for variability in neural activity and introduce a confound into the experimental design. Most studies of recognition memory have not equated task difficulty across age groups. Rather, most studies expect behavioral differences and then identify parallels between the behavioral findings and ERP results. A concern with this approach is that differences in performance may causally contribute to age-related differences observed in ERP waveforms. Two studies suggest that both age and performance influence memory-related ERPs recorded at encoding (Geng et al., 2018) and retrieval (Canada et al., 2020). Continuing to tease apart the effects of age and performance will be important for ERP studies of recognition memory development in the future. One way to account for performance-related effects is to control for them statistically in the analyses. Alternatively, tasks can be designed to calibrate difficulty across participants. For example, some studies have varied the frequency of deviant tones in an oddball task (Rusiniaket al., 2013) and the stimulus presentation durations during a go/no-go task (Stieben et al., 2007) to equate task difficulty across participants. For recognition memory research, variability in performance on a standard recognition memory task could be reduced by having younger individuals encode stimuli for longer durations or complete more study-test blocks. However, varying encoding durations would limit ERP analyses to retrieval because stimulus presentation durations can influence the amplitude of ERP responses (e.g., Brisson & Jolicœur, 2007; Gontier et al., 2008). This approach still would not address within-group variability in performance and would require extensive piloting to determining the correct conditions to equate performance. For that reason, continuous recognition tasks may be better suited to addressing variability in performance. Difficulty on a continuous recognition task could be equated by varying the lag between first and subsequent presentations or the mnemonic similarity of novel items relative to those previously encountered (Rollins & Cloude, 2018). The creation and validation of such a task would be a valuable endeavor for future research on recognition memory.

One final recommendation is to provide children with trial-by-trial feedback (Brooker et al., 2020). Trial-by-trial feedback is often provided during practice phases of recognition memory studies to ensure adequate task understanding, and some studies have provided feedback regarding reaction time if a speeded response was required (e.g., Mecklinger et al., 2011). However, trial-by-trial feedback is not typically provided during recognition memory tasks. Although providing trial-by-trial feedback could have positive effects (e.g., re-engaging attention or promoting accurate task performance), doing so could also have negative consequences. For example, feedback may increase the length of the paradigm, lead to frustration (especially in underperforming individuals), and/or influence the individuals’ response criterion (i.e., their willingness to respond that an item was previously encountered or novel). In addition, some research suggests that the outcomes of previous trials can impact subsequent ERP trials (e.g., Arjona et al., 2014). Therefore, researchers should carefully weigh the advantages and disadvantages associated with providing trial-by-trial feedback in their paradigm.

Common Methodological Differences between ERP Studies of Recognition Memory in Children versus Adults

Due to the considerations noted above and limited cognitive abilities of children, there are a number of methodological differences between paradigms utilized to study recognition memory in children and adults. Specifically, differences are present in: 1) the type of stimuli presented, 2) response modalities, 3) retrieval format, and 4) delay between encoding and retrieval. Studies of recognition memory in adults use a variety of stimuli, including pictures (e.g., Curran & Cleary, 2003), spoken words (e.g., Wilding & Rugg, 1996), and written words (e.g., Curran, 2000). Due to limitations and variability in children’s reading ability, few developmental ERP studies have used words as stimuli. Sprondel et al. (2012) used written words with 13- to 14-year-old adolescents and adults, and Czernochowski et al. (2005) used spoken words and pictures with 6- to 8-year-old children, 10- to 12-year-old children, and adults. Developmental studies of recognition memory more commonly use pictures as stimuli, including colored drawings (e.g., Czernochowski et al., 2009), abstract sculpture-like objects (Boucher et al., 2016), and photographs of scenes (e.g., Leventon et al., 2014), and photographs of toys played with during encoding (e.g., Riggins et al., 2013). The approach of presenting children with toys during learning and recording ERPs elicited to photographs of the objects was extended from studies of recognition memory conducted with infants (e.g., Bauer et al., 2003). In one previous study children had prior experience with the 3-D objects and ERPs elicited to 3-D objects and 2-D photographs were compared in 18-month-old. Memory-related ERP effects were elicted to both 3-D and 2-D objects (albeit in different components, see Carver et al., 2006). This finding suggested that both 2-D and 3-D objects may be capable of eliciting memory-based processing.

Response modalities also vary between studies based on participant age. Many studies of infants and young children have used passive viewing paradigms, which do not require an overt behavioral response (e.g., Bauer et al., 2003; Riggins & Rollins, 2015; Riggins et al., 2013; Wiebe et al., 2006). However, there are limitations to this approach. Passive viewing paradigms do not allow for the examination of reaction time as a dependent variable, and exposure to stimuli during the passive viewing paradigm could influence performance on subsequent behavioral assessments of memory. Studies that do use active retrieval tasks often have children make verbal responses (e.g., Marshall et al., 2002; Robey & Riggins, 2016) due to children’s propensity to generate movement-related artifact while making a button press (DeBoer et al., 2005). In contrast, responses are typically collected via button press in studies of older children and adults (e.g., Boucher et al,. 2016; Czernochowski et al., 2005; Mecklinger et al., 2011).

Another common difference between recognition memory research with children and adults is retrieval format. The use of incidental retrieval paradigms, such as passive viewing paradigms, is more common in child development research (e.g., Riggins & Rollins, 2015; Riggins et al., 2013) than studies with adults. Studies of recognition memory with older children and adults typically utilize intentional retrieval paradigms that require overt indications of recognition of the presented items or contextual details associated with them (e.g., Boucher et al., 2016; Czernochowski et al., 2005; 2009; Haese & Czernochowski, 2016; Leventon et al., 2014; Mecklinger et al., 2011; Sprondel et al., 2012). To explore the effect of this methodological difference on ERP correlates of memory, Robey and Riggins (2016) had young children complete a behavioral memory paradigm that required them to retrieve either item or source memory while they completed an incidental or intentional retrieval task. They observed that the overall amplitude of the Nc was larger across frontal and central leads when children engaged in intentional versus incidental retrieval. This observation is consistent with previous research in adults that compared electrophysiological responses associated with incidental and intentional retrieval (Nelson et al., 1998). The LSW was also impacted by incidental versus intentional retrieval. Although the pattern of the results was similar across conditions with new items eliciting a more positive amplitude ERP response than previously encountered items, the topographical distribution of the effect was distinct when children intentionally retrieved source information relative to the other three conditions. Whereas the effect was distributed over fronto-central electrodes/regions during incidental retrieval and the intentional retrieval of item memory, the effect was maximal over more posterior and left lateralized leads during the intentional retrieval of source memory. These findings exemplify how task adaptations for developmental samples can influence the neurocognitive processes recruited and the ERPs generated during performance.

A final difference between developmental studies of recognition memory in children relative to adults is the delay between encoding and retrieval. Most studies of recognition memory in adults and older children occur within a single testing session with a brief delay (e.g., 5 minutes) between encoding and retrieval (e.g., Czernochowski et al., 2005; Haese & Czernochowki, 2016; Mecklinger et al., 2011; Sprondel et al., 2012). However, consistent with ERP studies of recognition memory in infants (e.g., Wiebe et al., 2006), many studies with young children occur across multiple days with delays ranging from 24 hours to a week (e.g., Leventon et al., 2014; Riggins & Rollins, 2015; Riggins et al., 2013). This may be driven by infant and young children’s limited attention spans, which require visits on different days. However, it may also be driven by interest in long-term retention during development, which has been shown to improve dramatically (Lukowski & Bauer, 2014).

Recommendations for Implementing Developmental ERP Research with Adapted Paradigms

We have provided recommendations for how ERP research paradigms can be modified to be suitable for research with children and identified common differences between recognition memory tasks used with children relative to adults. Because task adaptations may impact the neurocognitive processes engaged, we propose three recommendations for carrying out developmental ERP research with paradigms adapted from research with adults.

Children and adults should both be tested on the modified paradigm when it is appropriate for both age groups.

This is common practice for ERP studies of recognition memory (e.g., Czernochowski et al., 2005, 2009; Marshall et al., 2002; Mecklinger et al., 2011; Rollins & Riggins, 2013; 2018; Sprondel et al., 2012). If the modified paradigm elicits the same neurocognitive processes as the paradigm on which the modification was based, the results obtained with adults should be consistent with previous research. For example, Rollins and Riggins (2018) had children, adolescents, and adults complete a modified version of Tulving’s (1985) remember/know paradigm. Due to a substantial body of research showing that items given a “remember” judgment elicit a larger amplitude parietal old/new effect than items given a “know” judgment and novel items (see Friedman & Johnson, 2000; Rugg & Curran, 2007), Rollins and Riggins (2018) anticipated observing the same pattern of results from the adults in their study. The adults did, in fact, show this pattern of results. If this pattern of results had not been obtained, it would have suggested that the paradigm elicited different neurocognitive processes than the standard remember/know paradigm.

If adapted tasks are not appropriate for adults, comparing children of different ages on the modified paradigm may shed light on developmental effects of interest.

The selection of age groups should be based on expectations for when developmental change occurs for the cognitive processes of interest. For one set of our studies, children engaged with toys in two locations, each of which was associated with a plush character (Riggins & Rollins, 2015; Riggins et al., 2013; Robey & Riggins, 2016; Rollins & Riggins, 2015). In one of our early studies using this paradigm with 5- and 6-year-old children (Riggins et al., 2013), we attempted to also test adults on the task; however, we observed very poor performance in our young adults, likely because the conditions (i.e., playing on the floor with children’s toys) were not age-appropriate for college students. Thus, to understand age-related differences, our subsequent research included 3-, 4-, 5-, and 6-year-old children due to behavioral research suggesting that early childhood represents a period of rapid change in children’s memory for contextual details (e.g., Riggins, 2014). Including multiple age groups of children has been a common approach for ERP studies on the development of recognition memory (Canada et al., 2020; Geng et al., 2018; Haese & Czernochowski, 2016; Leventon et al., 2014; Riggins & Rollins, 2015; Rollins & Riggins, 2018).

Examine the effect of task modifications on behavior and ERP components whenever feasible.

Whether each task modification made influences behavior or neurocognitive processes engaged is an empirical question. For example, Robey and Riggins (2016) assessed the effect of retrieval paradigm (incidental vs. intentional) and memory type (item vs. source) on behavioral performance as well as ERP effects associated with memory. Some empirical questions about task modifications may need to be assessed in individuals old enough to successfully contribute to both conditions. For example, whether stimulus modality (words vs. pictures) influences memory-related ERP components could be explored in school-aged children or older.

Current Analyses

To illustrate the possible utility of the proposal to empirically examine the effect of task manipulations on ERP components of interest, a secondary goal of the current manuscript was to explore the effect of different delay latencies on memory-related ERP components. To meet this goal, we pooled data from existing studies (Riggins & Rollins, 2015; Riggins et al., 2013) and conducted new analyses exploring the effect of a 1 day, 2 day, and 1 week delay period between encoding and retrieval on the Nc and LSW ERP components.

Method

Participants

Ninety-one children (M = 5.104 years, SE = 0.062, range: 4.03–5.96 years, 46 females, 45 males) were selected from previously published studies on recognition memory in early childhood (Riggins et al., 2013; Riggins & Rollins, 2015). Of the sample, 73% of the children were White or Caucasian, 12% Black or African American, 9% multiracial, 3% Asian, and 3% did not wish to disclose their race. Additionally, 87% identified as not Hispanic or Latino, 10% identified as Hispanic or Latino, and 3% did not wish to disclose their ethnicity. The majority of the children were from middle to high SES families (68% of guardians reported an annual household income exceeding $75,000/year).

Children encoded items in a play-like setting and then completed behavioral retrieval following a delay of one day (n = 32, 12 females, 20 males), two days (n = 19, 10 females, 9 males), or one week (n = 40, 24 females, 16 males). There were no sex differences across delay conditions, F(2, 88) = 0.020, p = .980. However, there was a difference in age across delay conditions, F(2, 88) = 38.994, p < .001. Children in the one week delay condition (M = 5.563 years, SE = 0.044) were approximately 7 months older than children in the one (M = 4.740 years, SE = 0.095) and two day delay conditions (M = 4.753 years, SE = 0.120), ps < .001, who were similar to each other in age (p = 1.00).

An additional 44 children participated in this research. Eight children were excluded for not meeting inclusion criteria (i.e., were born more than 3 weeks premature or parents reported a diagnosis of a learning or developmental disorder), six children were excluded due to technical difficulties (i.e., either the video of the session or EEG data was not recorded), six were excluded due to attrition (i.e., not completing the second session), and five children refused to wear the EEG cap or had it removed briefly after recording began. Therefore, a total of 110 children provided complete behavioral and electrophysiological data. A total of 19 children (17% of the available sample; eight from the one day condition, nine from the two day condition, and two from the one week delay condition) were excluded due to not providing the required number of ERP trials (see EEG Data Processing and Analysis below for more detail).

Procedure

All procedures were approved by the Institutional Review Board prior to data collection. Participants were recruited from a database maintained by the University. Parents provided informed consent for their children. Children received a small gift and a certificate for participating.

During the encoding portion of the memory paradigm, children engaged with age appropriate items in a play-like setting (e.g., a fireman’s hat). Children in the one and two day delay conditions interacted with 54 items and children in the one week delay condition interacted with 60 items. Items were split into sets, and the order of the sets were counterbalanced across participants. Items within sets were presented in a random order. The experimenter presented each item individually, provided it with a verbal label, and instructed the child to perform an action with the item. Children were not informed that their memory for the items would be subsequently assessed.

Memory retrieval was assessed using electrophysiological and behavioral assessments. Children returned to the lab for a second visit after a delay period of one day, two days, or one week (M = 6.78 days; range = 5–9 days). All children completed the electrophysiological assessment of memory retrieval prior to the behavioral memory retrieval assessment. First, children were fitted with a stretchy Lycra cap appropriate for their head size. Then, EEG data was continuously recorded with a sampling rate of 512 Hz from a Biosemi Active 2 system from 64 active Ag-AgCl scalp electrodes and two vertical and two horizontal electrooculogram (EOG) channels. During EEG recording, children passively-viewed 4.5 × 8” digital color photographs of the previously viewed items and novel items. Children in the one and two day delay conditions saw 54 previously encountered items and 27 novel items and children in the one week delay condition saw 60 previously encountered items and 30 novel items. Stimuli were presented on a neutral black screen and in a random order using E-Prime software. Children viewed stimuli for 500 ms with an ISI that ranged between 1250 and 1700 ms. A fixation cross was presented during the ISI. Children saw each stimulus presented multiple times; however, current analyses selectively included data from the first presentation.

Following the EEG assessment, children were physically shown the previously viewed and novel items and asked whether each item had been previously encountered. Previously viewed items that were accurately identified as previously encountered are referred to as “hits” and novel items that were accurately identified as novel are referred to as “correct rejections.”

EEG Data Processing and Analysis

Data were re-referenced offline using Brain Electrical Source Analysis (BESA) software (MEGIS Software GmbH, Gräfelfing, Germany) to an average reference configuration. Missing data was interpolated for a maximum of 8 bad channels per participant. Blinks were corrected using the algorithm developed by Ille et al. (2002). Data were high-pass filtered at .1 Hz and low-pass filtered at 40 Hz and hand-edited to omit movement and system related artifact. We epoched EEG recordings from −100 ms to 1500 ms relative to stimulus onset. The BESA artifact tool was applied to remove trials containing non-blink related artifacts with an amplitude threshold of 250 uV and gradient criterion of 75 μV.

Consistent with recommendations by developmental ERP researchers (DeBoer et al., 2005; 2007), participants were required to contribute a minimum of 10 trials per condition to be included in the analyses. Mean trial numbers (standard deviation, range) contributing to an average ERP after artifact rejection for hits by delay condition were as follows: one day, 32 (9, 11–47), two day, 31 (9, 13–45), one week, 39, (22–54). Mean trial numbers (standard deviation, range) for correct rejections by delay condition were as follows: one day delay, 16 (4, 10–23), two day delay, 17 (4, 10–23), one week delay, 19, (11–40). The number of trials contributing to the hit F(2, 88) = 7.643, p < .001, and correct rejection, F(2, 88) = 5.223, p = .007, conditions varied across delay duration. Hits were lower for children in the one day and two day delay conditions than children in the one week delay condition. Similarly, correct rejections were lower for children in the one day than the one week delay condition; the number of trials contributed by children in the two day delay condition was comparable to the one day and one week delay conditions. These differences in trial numbers are consistent with the number of available trials. Critically, the current analyses examined mean amplitudes, which are relatively unaffected by differences in trial numbers across groups and conditions (Luck, 2014).

As noted in the introduction, we designed our paradigm conservatively estimating that we would lose approximately 50% of the available trials. A challenge for memory research is that the number of available trials is dependent both on behavioral performance and the quality of the EEG data. Considering the number of stimuli available to be classified as hits and correct rejections, we were able to retain 60–65% trials. However, that percent increased to 65–75% when we used the number of hits and correct rejections as the denominator. Thus, researchers designing tasks for which ERP trials are not contingent on behavioral performance will not need to estimate data loss as conservatively.

Mean amplitudes were used to examine the effect of delay on ERP correlates of memory during the 350–500 ms and 800–1100 ms time windows. Time windows were selected based on developmental research of memory effects and previous analyses of these data (Bauer et al., 2003; Marshall et al., 2002; Riggins & Rollins, 2015; Riggins et al., 2013; Robey & Riggins, 2016). Data were analyzed using an omnibus ANOVA with Delay (1 day, 2 day, 1 week) as the between-subjects factor and the following within-subjects factors: 2 Condition (hits, correct rejections) x 7 Sagittal Plane (left lateral, left central, left medial, midline, right medial, right central, right lateral) x 5 Coronal Plane (frontal, fronto-central, central, centro-parietal, parietal) at the following leads: F5, F3, F1, Fz, F2, F4, F6, FC5, FC3, FC1, FCz, FC2, FC4, FC6, C5, C3, C1, Cz, C2, C4, C6, CP5, CP3, CP1, CPz, CP2, CP4, CP6, P5, P3, P1, Pz, P2, P4, P6. When appropriate, the Greenhouse Geisser procedure was used to correct for violations of sphericity. Findings reported include a main effect of Delay or a main effect of or interaction with Condition.

Results

Behavioral Results

To examine the effect delay had on children’s ability to discriminate between old and new stimuli, a univariate ANOVA was conducted on d-prime (d’) scores (i.e., higher scores indicate greater discriminability). Delay significantly influenced children’s ability to discriminate between old and new items, F(2, 88) = 3.797, p = .026 (Figure 1). Children that experienced a delay of one week (M = 2.94, SE = .121) performed significantly worse than children that experienced a one day delay (M = 3.437, SE = .135), p = .022; children in the two day delay condition (M = 3.120, SE = .175) performed comparably to children in the one day and one week delay conditions (ps ≥ .466).

Figure 1.

Behavioral data

ERP Results

Nc (350–500 ms).

The amplitude of the Nc was not significantly influenced by Delay or Condition or their interaction, Fs ≤ 1.57, ps ≥ .198, 𝜂_p² = .035 (Figure 2).

Figure 2.

Grand Average Waveforms for Hits and Correct Rejections

Note. The grand average waveforms depict mean amplitude responses elicited to hits and correct rejections in groups of children that had a 1 day, 2 day, and 1 week delay between encoding and retrieval at Fz and Pz. The boxes reflect the time windows analyzed for the Nc (350–500 ms) and LSW (800–1100 ms) components.

LSW (800–1100 ms).

Of critical interest for the present report, the memory effect in the LSW was not significantly modulated by the duration of the Delay between encoding and retrieval, ps ≥ .069, 𝜂_p²s ≤ .059. However, consistent with previous reports of these data (Riggins & Rollins, 2015; Riggins et al., 2013), memory influenced the amplitude of the late slow wave (Figures 2). We observed a significant Condition x Coronal Plane interaction, F(4, 352) = 19.15, p < .001, 𝜂_p² = .179 (see Figure 3). At frontal and fronto-central leads, correctly rejected items (Frontal: M = 2.089, SE = .324, Fronto-central: M = 3.271, SE = .259) elicited a more positive amplitude response than hits (Frontal: M = .506, SE = .283, Fronto-central: M = 2.059, SE = .264), Fs(1, 90) = 16.902–21.758, ps < .001, 𝜂_p² = .158–195. The main effect of Condition was not significant over central, F(1, 90) = 2.267, p = .136, 𝜂_p² = .025, or centro-parietal leads, F(1, 90) = 1.667, p = .200, 𝜂_p² = .018. The effect was significant and reversed over parietal leads; hits (M = 1.45, SE = .255) elicited a more positive amplitude response than correctly rejected items (M = −.054, SE = .283), F(1, 90) = 24.543, p < .001, 𝜂_p² = .214.

Figure 3.

Topographical Maps Depicting Differences in the ERP Response to Hits and Correct Rejections

Note. These topographical maps depict differences in the ERP response to hits relative to correct rejections. Red values reflect hits eliciting a larger mean amplitude response than correct rejections and blue values reflect correct rejections eliciting a larger mean amplitude response than hits. The top row reflects the average topographical map for the Nc (350–500 ms) and the bottom row reflects the average topographical map for the LSW (800–1100 ms). No amplitude differences were observed in the Nc. In the PSW, correct rejections elicited a larger mean amplitude response than hits over frontal and fronto-central leads and the effect reversed over parietal leads. The memory effect did not significantly differ among children that experienced a 1 day, 2 day, or 1 week delay.

Discussion

Most developmental ERP research with children uses experimental paradigms that were adapted from tasks originally designed for adults (Brooker et al., 2020). This is certainly true in the domain of recognition memory. When adapting adult ERP memory paradigm for use in developmental samples, we recommend examining the effect of task modifications on ERP components of interest by (a) testing both children and adults on the modified paradigm when appropriate to support the continuity of findings across development, (b) assessing developmental effects by comparing children of different ages on the modified paradigm based on the developmental trajectory of cognitive process of interest, and (c) empirically assessing the effect of methodological differences between paradigms.

Common differences between ERP studies of recognition memory conducted with children relative to adults include 1) the type of stimuli presented, 2) response modality, 3) retrieval format, and 4) delay between encoding and retrieval. The goal of the present analyses was to explore the effect of the latter (i.e., different delay durations) on memory-related ERP components in children. Children from two previously published studies on recognition memory encoded toys in a play-like setting with an experimenter (Riggins & Rollins, 2015; Riggins et al., 2013). One day, two days, or a week later, ERP data were collected while children passively viewed pictures of the previously encoded and novel items. Following the electrophysiological recording, children made recognition memory judgments about the items. Consistent with forgetting curves (e.g., Murre & Dros, 2015) children more effectively discriminated between old and new stimuli when they experienced a shorter delay between encoding and retrieval. Specifically, children in the one day delay condition performed better than children in the one week delay condition.

Consistent with previous reports of these data, memory influenced the amplitude of the LSW (Riggins & Rollins, 2015; Riggins et al., 2013; Robey & Riggins, 2016). Correctly rejected novel items elicited a more positive amplitude response than hits over frontal and fronto-central leads. However, the pattern reversed over parietal leads with hits eliciting a more positive amplitude response than correctly rejected novel items, which resembles the parietal old/new effect commonly reported in ERP studies of recognition memory with adults (for review see Friedman & Johnson, 2000; Rugg & Curran, 2007).

Critically, this pattern of results was similar across delay durations, suggesting that the length of delay did not modulate the effect of memory on the LSW. This finding is important for recognition memory research in infants and young children since encoding and retrieval are typically separated across multiple sessions (Leventon et al., 2014; Riggins & Rollins, 2015; Riggins et al., 2013; Wiebe et al., 2006). However, because studies with older children and adults typically include a single session (e.g., Czernochowski et al., 2005; Haese & Czernochowki, 2016; Mecklinger et al., 2011; Sprondel et al., 2012), it will be informative for future research to establish whether a within-session delay yields a different pattern of results than longer days.

In conclusion, the present report provides multiple recommendations for how to effectively adapt ERP paradigms designed for adults for research with younger individuals. Specifically, we discussed general recommendations (i.e., providing practice trials, minimizing the number of trials, blocking trials to allow children breaks, equating task difficulty, and providing trial-by-trial feedback), examples of modifications used by ERP studies of recognition memory, and specific recommendations regarding 1) testing children and adults on modified paradigms, 2) assessing development by comparing children across developmental stages, and 3) empirically assessing the effects of task modifications. These recommendations can be used to support researchers as they develop novel paradigms and grow the field of developmental cognitive electrophysiology.