Memory binding and theta EEG during middle childhood

Abstract

The ability to bind together the contextual details associated with an event undergoes dramatic improvement during childhood. However, few studies have examined the neural correlates of memory binding encoding and retrieval during middle childhood. We examined age-related encoding and retrieval differences using continuous electroencephalogram (EEG) measures in a sample of 6- and 8-year-olds. For the memory binding task, children were tested on memory for individual items (i.e., objects and backgrounds only) and combined object–backgrounds pairings (combination condition). Memory for individual item information was comparable across both age groups. However, younger children experienced greater difficulty (i.e., higher false alarm rate) in the combination condition. Theta (4–7 Hz) neuronal oscillations were analyzed to compare memory encoding and retrieval processes. Widespread retrieval-related increases in theta band EEG power (compared with baseline and encoding-related activation) were evident in both 6- and 8-year-olds. Regression analyses revealed that parietal theta EEG power during retrieval accounted for variability in memory binding performance. These findings suggest that theta rhythms are intricately linked to memory binding processes during middle childhood.

1 |. INTRODUCTION

Our ability to form memories of our personal past that are rich in detail and vary in complexity (e.g., memory for all the contextual details of one’s wedding day) is critical in constructing our autobiography and considered a hallmark of episodic memory (EM). EM depends on the ability to bind the individual elements of an event together into a cohesive representation (e.g., remembering who attended, what was said in your vows, where and when the ceremony and reception took place). This process of encoding the relations among stimuli and linking the components of an event into a unified and memorable whole is referred to as binding (Chalfonte & Johnson, 1996; Sluzenski et al., 2006). In general, children’s ability to bind items in memory undergoes significant developmental change during the transition from early to middle childhood (Drummey & Newcombe, 2002; Riggins, 2014; Sluzenski et al., 2006). However, few studies have examined the neural correlates of memory binding encoding and retrieval during this time period. The present study examined relations between psychophysiological measures and age-related differences in memory binding development. In the following sections, we review the developmental literature on memory binding and discuss its associated neural correlates.

1.1 |. Memory binding development in early and middle childhood

Developmental improvement in associative binding capacities has been studied in multiple ways using relational memory and source memory paradigms. Typically, these studies have examined item-item binding (e.g., pairing two separate pictures together or pairing an item with a color feature; Cycowicz et al., 2003; Sluzenski et al., 2006), item-source binding (e.g., pairing a fact with the individual who taught it; Drummey & Newcombe, 2002), or binding an item to a particular spatial (Bauer et al., 2012) or temporal context (Riggins et al., 2009).

Newcombe and colleagues examined the development of item-item binding in early childhood by testing 4- and 6-year-old children on their memory for individual objects, individual backgrounds, and combined object–background pairings. Younger children had greater difficulty remembering bound items, but the memory for individual features was comparable across age groups (Lloyd et al., 2009; Sluzenski et al., 2006). In particular, younger children had difficulty discriminating between original and rearranged combination pairs, and exhibited higher rates of false alarms (i.e., incorrectly attributing a new item as old) in the bound condition. Other developmental investigations using source memory paradigms found that memory for individual items alone (e.g., facts learned in an experimental setting) steadily improves from 4 to 10 years (Drummy & Newcombe, 2002; Rajan & Bell, 2015; Rajan et al., 2014; Riggins, 2014), whereas the ability to link an item to its corresponding source (e.g., puppet, experimenter) shows the most pronounced improvement between 4 and 6 years (Drummey & Newcombe, 2002; Riggins, 2014).

Given that memory binding improves from 4 to 6 years, some have concluded that the associative component of EM is relatively intact by early childhood (Shing et al., 2008). However, there is evidence to suggest that developmental improvement in binding item and source-specifying (Rajan & Bell, 2015) and feature information (S. Ghetti et al., 2010; Lorsbach & Reimber, 2005) may extend into middle childhood and beyond. Thus, it could be possible that memory binding shows adult-like qualities by age 6 or continues to show improvement. In order to answer this question, we investigated whether developmental differences in item-item memory binding would be evident between 6 and 8 years of age.

1.2 |. Neural correlates of memory binding

It is important to consider the underlying neurological systems that support age-related changes in memory binding. Developmental neuroimaging investigations reveal that age-related changes in the medial temporal lobe (MTL) and prefrontal cortex (PFC) structures (S. Ghetti et al., 2010; S. Ghetti & Bunge, 2012; Ofen et al., 2007) and functional connectivity between MTL and PFC structures (Menon et al., 2005) play an important role in item-context binding. Electrophysiological indices (using electroencephalogram [EEG] power or event-related potentials [ERPs]) have also been informative to our understanding of age-related changes in memory binding.

Investigations using ERPs, which measure brain electrical responses time-locked to a specific event (e.g., during item and source memory judgments) reveal differential responses in the ERP waveform known as “EM” effects. In adults, greater ERP amplitudes at left parietal electrode sites are observed for words recalled with accurate item-context judgments, referred to as a left parietal EM effect (Wilding & Rugg, 1996). ERP recollection-based processing indices have also been documented in children. For example, the parietal EM effect has been observed in older school-aged children (Czernochowski et al., 2005; de Chastelaine et al., 2007), whereas in younger preschool-aged children a late-slow wave component is associated with both the encoding (Geng et al., 2018) and retrieval of contextual details (Canada et al., 2019; Riggins et al., 2013).

In addition to ERP correlates of EM, neuronal oscillatory activity at different frequency ranges is also associated with various memory processes (see Klimesch, 1999; Nyhus & Curran, 2010 for reviews) but has received less attention in the developmental literature. Evidence from both human and animal studies suggest that theta oscillations in the cortical-hippocampal network are important for the encoding and retrieval of EM (Nyhus & Curran, 2010). In adults, task-related increases in theta power (4–7 Hz) are associated with several different memory processes, including rehearsal, short-term memory, episodic encoding, and episodic retrieval (Klimesch, 1999; Klimesch et al., 1997). In general, the research suggests positive associations between theta oscillations and memory binding; higher theta power is correlated with successful item-context encoding (Staudigl & Hanslmayr, 2013) and correct source memory retrieval (Gruber et al., 2008), and theta synchronization differentiates high and low EM performance (Dopplemayr et al., 1998). In particular, frontal-theta activity was associated with successful item-context binding during encoding (Summerfield & Mangels, 2005), and both frontal and parietal theta oscillations were enhanced during successful retrieval of source details (Addante et al., 2011; see Hsieh & Ranganath, 2014 for a review).

Less is known about the functional role of theta synchronization during childhood. In 2-year-olds, theta EEG differentiates encoding and retrieval processes. More specifically, encoding- and retrieval-related increases in theta band power were observed during an explicit memory task and distinguished high and low memory performance (Cuevas et al., 2012). However, this study examined recollection of item information alone and did not examine the relation of theta EEG and item-context binding. To our knowledge, only two prior studies have examined theta EEG correlates associated with memory binding processes. In one study examining developmental changes in source memory, Rajan and Bell (2015) found that both 6- and 8-year-olds exhibit memory-related increases in theta EEG power at frontal, temporal, and parietal electrode sites. However, this investigation only focused on retrieval processes. Given that age-related changes in encoding are evident in middle childhood and beyond (Geng et al., 2018; Rollins & Riggins, 2013), it is important to examine differences in theta EEG power between memory encoding and retrieval phases. Using measures of theta band EEG coherence, Blankenship and Bell (2015) found that frontotemporal functional connectivity during encoding and retrieval predicted item-item memory binding performance across middle childhood (9–12 years of age), but this study did not explicitly examine age-related changes in theta EEG activation during memory processing. Clearly, additional research is needed to examine whether age-related differences in theta band EEG power differentiate both the encoding and retrieval of individual items and their bound representation, which was the focus of the present investigation.

1.3 |. Overview of current study

The purpose of our investigation was to examine age-related encoding and retrieval differences in theta EEG activation during a memory binding task in 6- and 8-year-olds. Continuous EEG measures were collected during encoding and retrieval of individual item features (objects only, backgrounds only) and their bound representation (combined object–background pairings). Given that improvement in the binding of item information may extend into middle childhood and beyond, we hypothesized that 6-year-olds would have higher rates of false alarm errors in the bound condition, whereas memory for individual item information would be comparable for both age groups.

Unfortunately, prior theta EEG studies do not provide much information on the developmental trajectories of encoding and retrieval processes across middle childhood (Blankenship & Bell, 2015; Rajan & Bell, 2015). Therefore, our first research question was exploratory and examined whether encoding- and retrieval-related changes in theta EEG activation would vary as a function of age. With respect to memory processing stages, based on prior developmental research (Cuevas et al., 2012; Rajan & Bell, 2015), we hypothesized that task-related increases in theta power would be evident during encoding and retrieval of individual item information and during item-context binding. Our second research question focused on individual differences in task performance. Based on previous evidence of frontal and parietal theta oscillations being linked with successful episodic recollection in adults (Addante et al., 2011; Hsieh & Ranganath, 2014), we examined whether these same frontal and parietal sites accounted for variance in memory binding performance during middle childhood.

2 |. METHOD

2.1 |. Participants

Twenty-nine 6-year-olds (M = 6.10, SD = 0.26; 19 girls; 28 Caucasian, 1 African American) and 36 8-year-olds (M = 8.07; SD = 0.36; 17 girls; 28 Caucasian, four Hispanic, three African American, one American Indian/Alaska Native) participated in this study. Children were eligible for participation if they were born within 4 weeks of their expected due date, experienced no prenatal or birth complications and had no developmental or neurological diagnoses. An additional seven children were recruited but excluded from analyses, either for failing to meet inclusion criteria (developmental diagnosis: n = 2; premature birth: n = 1) or for failing to demonstrate understanding of the recognition task (n = 4). Specifically, following the procedure of Sluzenski et al. (2006), four 6-year-olds were excluded because they provided either all “yes” or all “no” responses during the test. Therefore, behavioral analyses focused on a sample of 65 children. With respect to parental education, all mothers and 99% of fathers graduated from high school (4.6% and 6.2% technical degree, 41.5% and 38.5% bachelor’s degree, 50.8% and 44.6% graduate degree, respectively). The average maternal and paternal age was 38 and 40 years, respectively. As compensation for participation, all children received a $10 gift card, and parents were entered into a lottery drawing for one $50 gift certificate to a local store of their choice.

2.2 |. Procedure

EEG recording:

EEG recordings were acquired during baseline and the memory binding task. Based on publication guidelines for studies using EEG methodology (Keil et al, 2014), we report the following details. Recordings were made from 24 left and right scalp sites: frontal pole (Fp1, Fp2), medial frontal (F3, F4), lateral frontal (F7, F8), frontal-central (FC1, FC2, FC5, FC6), central (C3, C4), temporal, (T7/T8), central-parietal (CP1, CP2, CP5, CP6), medial parietal (P3/P4), lateral parietal (P7/P8), and occipital (O1, O2). All electrode sites were referenced to Cz during recording. EEG was recorded using a stretch cap (Electro Cap, Inc., Eaton, OH; E1-series cap) with tin electrodes in a modified 10/10 system pattern. After the cap was placed on the head, a small amount of abrasive gel was placed into each recording site and the scalp gently rubbed. Conductive gel was then added to the recording sites. Electrode impedances were measured and accepted if below 10 kΩ. The electrical activity from each lead was amplified using separate Bioamps (James Long Company, Caroga Lake, NY). During data collection, the high-pass filter was a single-pole resistor-capacitor (RC) filter with a 0.1 Hz cut-off (3 dB or half-power point) and 6 dB per octave roll-off. The low-pass filter was a two-pole Butterworth-type with a 100 Hz cut-off (3 dB or half-power point) and 12 dB octave roll-off.

Activity for each lead was displayed on the monitor of an acquisition computer. The EEG signal was digitized online at 512 samples per second for each channel so that the data were not affected by aliasing. The acquisition software was Snapshot-Snapstream (HEM Data Corp., Southfield, MI), and the raw data were stored for later analyses. Prior to the recording of each subject, a 10 Hz, 50 μV peak-to-peak sine wave was input through each amplifier. This calibration signal was digitized for 30 s and stored for subsequent analysis.

EEG analysis:

Spectral analysis of the calibration signal and computation of power at the 9–11 Hz frequency band was accomplished. The power figures were used to calibrate the power derived from the subsequent spectral analysis of the EEG. Then, EEG data were examined and analyzed using EEG Analysis System software developed by James Long Company. Data were re-referenced via software to an average reference configuration and then the artifact scored for eye movements using a peak-to-peak criterion of 100 μV or greater. Gross motor movements over 200 μV peak-to-peak were also scored. These artifact-scored epochs were eliminated from all subsequent analyses. The data were then analyzed with a discrete Fourier transform (DFT) using a Hanning window of 1 s width and 50% overlap. Power was computed for the theta frequency band (4–7 Hz). Power was expressed as mean square microvolts and the data were transformed using the natural log (ln) to normalize the distribution. Based on prior biobehavioral investigations of explicit (Cuevas et al., 2012) and source memory (Rajan & Bell, 2015) development, we focused on task-related changes in theta EEG power at frontal, temporal, and parietal electrode sites.

Baseline EEG:

Baseline EEG was recorded while children were shown a neutral cartoon video (M = 174.2 s, SD = 11.0, artifact-free DFT windows, M = 184, SD = 77). This obtained a period of physiology that contained eye movements and gross motor activity comparable to what was exhibited during the memory binding task. Parents were instructed not to talk to their children during the EEG recording.

Memory binding task:

This task was originally developed by Sluzenski et al. (2006) as a measure of memory binding in early childhood. Children were assessed on their memory for individual features (object alone, background alone) as well as the combination of these two features (object in the background). The encoding, training, and test phases were administered on a computer monitor. Stimulus presentation and participant responses were controlled and recorded using SuperLab Pro 4.0 (San Pedro, CA). For the encoding portion of the task, children were shown 48 pictures comprising 16 subsets, which were presented in a random order to all participants. For each subset, children were shown the following in a fixed order (5-s duration per picture): a picture of an animal (e.g., bear), followed by a location (e.g., basketball court), followed by the animal in that respective location (e.g., bear in basketball court). In another subset, the child was shown a picture of a kangaroo, a picture of a snowfield, and a picture of the kangaroo in the snowfield. Similar to the procedure used in Sluzenski et al. (2006), unique animal–location pairings were chosen in order to reduce the possibility of guessing based on knowledge of their natural habitats (see Figure 1).

FIGURE 1.

Example of encoding and test stimuli for the object, background, and combination conditions

After a 20-min delay, we tested children’s memory for individual items (objects and backgrounds) and their paired combination. Children were tested in a blocked format (16 pictures each) in the following fixed order: objects only (i.e., animals), backgrounds only, and object–background combinations. For each of the three conditions, half of the pictures were old and half new. In the combination condition, “old” pictures consisted of identical animal–background pairs shown during encoding, whereas “new” pictures consisted of re-combined pairs of familiar stimuli. Thus, the old/new discrimination judgment included distinguishing between studied and novel animals, studied and novel backgrounds, and studied (e.g., kangaroo in the snowfield) or rearranged pairings (e.g., chimpanzee in basketball court; see Figure 1). Children were instructed to respond “yes” to pictures they had seen before and “no” if the picture was new.

Prior to testing, practice trials were administered to ensure children understood the rules of the task. For the practice trials, children were first shown two animal pictures, two backgrounds pictures, and two animal–background pictures not included during the test phase of the study. To test children’s memory on practice trials, one old and one new pictures were presented first for the animal condition and then one old and one new pictures were presented for the background condition. An old–new pairing was then presented for the animal–background combination, but the test stimuli differed slightly. Most notably, the new animal–background pairing consisted of an old animal and old background paired together in a new combination, and the experimenter explicitly made sure to point out that the animals had been switched in this case. After successful completion of the practice trials, testing began. The dependent variables of interest were the proportion of hits (i.e., old items correctly identified as old), the proportion of false alarms (i.e., new items incorrectly identified as old) and d’ sensitivity scores for the animal, background, and combination conditions. A standard correction was used in the calculation of sensitivity scores (d’) for hit rates of 1 and false alarm rates of 0. The adjusted value of 1 – 1/(2N), where N is the maximum number of hits, was used when the proportion of hits was equal to 1. The adjusted value of 1/(2N), where N is the maximum number of false alarms, was used when the proportion of false alarms was equal to 0.

An event mark was placed on the electrophysiological record so that EEG recordings could be synchronized with the encoding and retrieval phases of the memory task. Encoding-related EEG was synchronized to the presentation of each 5-s stimulus trial. For encoding, the mean artifact-free DFT windows for object, background, and combination conditions were 50 (SD = 25), 59 (SD = 31), and 58 (SD = 29), respectively. Retrieval-related EEG started immediately after stimulus presentation and continued until the child responded. Only correct trials were used in the analyses; the mean artifact-free DFT windows for object, background, and combination conditions were 36 (SD = 18), 35 (SD = 17), and 37 (SD = 20), respectively. The average number of retrieval trials from which electrophysiological data were collected was 14.80 (SD = 1.44), 15.70 (SD = 0.61), and 13.93 (SD = 2.02) for the object, background, and combination conditions, respectively.

2.3 |. Verbal intelligence proxy

The Expressive Vocabulary Test (EVT; Williams, 1997) was administered to examine expressive vocabulary and word retrieval. The EVT is a nationally standardized instrument that has been normed for ages 2½ through 90⁺ years and is co-normed with the Peabody Picture Vocabulary Test–III (Dunn & Dunn, 1997). Expressive language is highly correlated with performance on measures of general intel ligence (Marchman & Fernald, 2008) and can be viewed as a proxy for verbal intelligence. Our dependent measure of interest was the participants’ percentile rank based on age.

3 |. RESULTS

3.1 |. Behavioral performance

A 3 (test item type: Object, background, and combination) × 2 (age: 6 or 8 years) mixed analysis of variance (ANOVA) was conducted for each of the three dependent measures of interest (hits, false alarms, and d’ sensitivity scores) with test item type as the within-subjects factor and age as the between-subjects factor. Figure 2a shows the proportion of mean “yes” responses as a function of age and test item type. For the pattern of hits, there was a main effect of test trial type, F(2, 62) = 15.95, p < .001, η_p² = .34. Posthoc analyses using Bonferroni correction revealed that the hit rate for the background condition was higher than both the object (p < .001) and combination (p = .001) conditions. The main effect for age and the test item type × age interaction were not significant (all ps > .121).

FIGURE 2.

(a) Mean proportion “yes” responses as a function of age and test item type for hits and false alarms. (b) d’ sensitivity scores as a function of age and test item type. Higher d’ values indicate a better ability to discriminate between old and new items. **p < .01

For the pattern of false alarms, there was a main effect for age, F(1, 63) = 8.85, p = .004, η_p² = .12, with 6-year-olds committing more false alarm errors than 8-year-olds overall. There was also a main effect of test trial type, F(2, 62) = 27.06, p < .001, η_p² = .47. Posthoc analyses revealed that false alarms errors were higher in the combination condition, compared to both object (p < .001) and background (p < .001) conditions. In addition, the false alarm rate was higher for object compared to background trials (p = .003). These main effects were superseded by an age × test trial type interaction F(2, 62) = 4.17, p = .02, η_p² = .12. Follow-up comparisons using Bonferonni correction revealed an age difference in the combination condition, t(63) = 2.79, p = .007, d = .68, with 6-year-olds committing a higher proportion of false alarms than 8-year-olds, whereas the false alarm rate in the object and background conditions was equivalent between age groups (all ps > .16; see Figure 2a).

For d’ scores, there was a main effect for age, F(1, 63) = 8.94, p = .004, η_p² = .12, with recognition sensitivity higher in 8-year-olds. There was also a main effect of test trial type, F(2, 62) = 39.93, p < .001, η_p² = .56. Follow-up analyses revealed that sensitivity scores for the combination condition were lower than both the object and background conditions, and that sensitivity in the background condition was higher than the object condition (all ps < .001). The age × test trial type interaction was also significant, F(2, 62) = 4.35, p = .017, η_p² = .12. Follow-up comparisons using Bonferonni correction revealed an age difference only in the combination condition, t(63) = −2.96, p = .004, d = .73, with 8-year-olds exhibiting higher d’ scores, indicating greater recognition sensitivity (see Figure 2b).

3.2 |. Theta EEG results

With the inclusion of EEG, there was a further reduction in sample size. Four children (two 6-year-olds and two 8-year-olds) did not contribute to the psychophysiological data due to excessive gross-motor artifact throughout the protocol. Therefore, 61 children contributed electrophysiological data to the following analyses.

3.3 |. Statistical analysis

The EEG data were analyzed using three separate repeated measures multivariate analysis of variance (MANOVAs) for each condition of interest (i.e., object, background, and combination conditions). For each analysis, region (i.e., medial frontal, lateral frontal, temporal, medial parietal, lateral parietal), hemisphere (i.e., left, right) and processing stage (i.e., baseline, encoding, retrieval) served as within-subjects factors and age (i.e., 6 or 8 years) was the between-subjects factor. For ease in examining any interaction effects among these variables, follow-up MANOVAs were performed. A multivariate approach for examining interaction effects has been suggested by Keselman (1998). In order to limit the familywise type 1 error rate, a Bonferroni procedure was adopted. We set the familywise type 1 error rate to α = .10 instead of α = .05 because this latter value would be too conservative for highly correlated variables, such as regional EEG power values (Yoder et al., 2004). We were especially interested in examining any main effects or interaction effects involving age and/or processing stage (comparing baseline, encoding, and retrieval processing).

3.4 |. Object condition

For the object condition, there was a main effect of age, F(1,53) = 4.25, p = .04, with 6-year-olds exhibiting higher theta power values than 8-year-olds, likely reflecting a developmental decrease in the contribution of theta-range power with age (Saby & Marshall, 2012). Interactions involving age and processing stage were not significant (all ps > .14). The three-way stage × region × hemisphere interaction was significant, F(8, 46) = 2.83, p = .012, η_p² = .33. To examine this interaction, we collapsed across the nonsignificant factor (i.e., age) and completed separate follow-up MANOVAs (within-subjects factors: stage, hemisphere) on the theta EEG power values for each region. The results of the follow-up regional MANOVAS are displayed in Table 1, and the means for theta EEG power during baseline, encoding, and retrieval for the object condition are displayed in Figure 3. Only main effects and interactions involving processing stage are highlighted.

TABLE 1.

Summary of regional multivariate F analyses for memory processing comparisons during object, background, and combination conditions

	F3/F4	F7/F8	T7/T8	P3/P4	P7/P8
Condition
Object
Stage	17.58*** (.37)	45.41*** (.61)	16.50*** (.36)	33.77*** (.54)	36.03*** (.56)
Hemisphere	–	20.06*** (.26)	–	–	–
S × H	–	3.49* (.11)	–	6.68** (.19)	6.83** (.19)
Background
Stage	10.53*** (.26)	27.53*** (.49)	24.06*** (.45)	16.03*** (.36)	15.24*** (.35)
Combination
Stage	44.00*** (.60)	35.81*** (.56)	42.60*** (.60)	20.44*** (.42)	26.12*** (.48)

^*p < .05;

^**p < .01;

^***p < .001.

Effect sizes (ηp²) are in parentheses.

FIGURE 3.

Theta EEG power values for baseline, encoding, and retrieval (collapsed across age) during the object condition. ***p < .001

There was a main effect of the processing stage across all electrode regions. We conducted pairwise comparisons for these electrode sites after adopting a Bonferonni procedure to control for the overall level of significance (five electrode regions with three comparisons for each site; p = .10/15 = .007). Posthoc analyses revealed that encoding-related theta EEG power values were higher than baseline power values across medial frontal, lateral frontal, and temporal regions (ps < .001) and that retrieval power values were higher than baseline power values across all electrode regions (ps < .001). For medial parietal and lateral parietal regions, retrieval theta EEG values were higher than encoding power values (ps < .001).

There was a stage × hemisphere interaction for lateral frontal, medial parietal, and lateral parietal regions. Pairwise comparisons (Bonferonni procedure: three comparisons across three regions; p = .10/9 = .01) revealed higher right hemisphere theta power values than the left hemisphere during the encoding and retrieval phases at lateral frontal sites (ps = .001). For lateral parietal sites, retrieval theta EEG was higher in the left hemisphere (p = .001) and the medial parietal region failed to reach significance.

3.5 |. Background condition

For the background condition, the main effect of age and interactions involving age and processing stage were not significant (all ps > .068). There was a significant stage × region interaction, F(8, 46) = 10.35, p < .001, η_p² = .64. To examine this interaction, we collapsed across all nonsignificant factors (i.e., age, hemisphere) and completed separate follow-up MANOVAs (within-subjects factors: stage) on the theta EEG power values for each region. The results of the follow-up regional MANOVAS are displayed in Table 1, and the means for theta power during baseline, encoding, and retrieval for the background condition are displayed in Figure 4. There was a main effect for processing stage across all regions. Pairwise comparisons for these electrode sites were conducted (Bonferonni procedure: five regions with three comparisons for each; p = .10/15 = .007). For lateral frontal electrode sites, theta EEG power values during encoding were higher than baseline power values (p < .001). EEG power values were higher during the retrieval phase than during baseline across all electrode sites (all ps ≤ .005). In general, retrieval power values were higher than encoding power values across all electrode sites except lateral frontal (all ps ≤ .003).

FIGURE 4.

Theta EEG power values for baseline, encoding, and retrieval (collapsed across age and hemisphere) during the background condition. **p < .005; ***p < .001 [Corrections added on 17th May, 2021 after first online publication: **p < .05 was changed to **p < .005]

3.6 |. Combination condition

For the combination condition, the main effect of age and interactions involving age and processing stage were not significant (all ps > .064). There was a significant stage × region interaction, F(8, 46) = 5.17, p < .001, η_p² = .47. To examine this interaction, we collapsed across all nonsignificant factors (i.e., age, hemisphere) and completed separate follow-up MANOVAs (within-subjects factors: stage) on the theta EEG power values for each region. The results of the follow-up regional MANOVAS are displayed in Table 1, and the means for theta power during baseline, encoding, and retrieval for the combination condition are displayed in Figure 5. There was a main effect for processing stage across all regions. Pairwise comparisons for these electrode sites were conducted (Bonferonni procedure: five regions with three comparisons for each; p = .10/15 = .007). Posthoc analyses revealed that encoding-related theta EEG power values were higher than baseline power values across lateral frontal and temporal regions (ps ≤ .001) and that retrieval power values were higher than baseline power values across all electrode regions (ps < .001). Retrieval power values were higher than encoding power values across all regions except lateral frontal (ps < .001).

FIGURE 5.

Theta EEG power values for baseline, encoding, and retrieval (collapsed across age and hemisphere) during the combination condition. ***p ≤ .001

3.7 |. Brain-behavior associations

Prior literature indicates that frontal and parietal theta oscillations support episodic encoding and retrieval in adults (Addante et al., 2011; Hsieh & Ranganath, 2014), but less is known about brain–behavior associations of theta EEG and memory for contextual information during childhood development. Therefore, we completed multiple regression analyses to examine whether frontal and parietal theta EEG contributes to individual variation in memory binding performance during middle childhood. Our dependent measure of interest was combination condition performance, as this was the only condition that tapped into item-item memory binding (i.e., object–background pairings). Based on the results of our investigation into task-related changes in theta EEG activation during the combination condition, we focused our analyses on retrieval. Retrieval-related theta EEG was higher than both baseline and encoding phases. Therefore, our predictors included retrieval theta EEG power values for two frontal regions of interest (medial frontal and lateral frontal) and two parietal regions of interest (medial parietal and lateral parietal). Background variables (age, verbal intelligence proxy) were also included in the regression model. Regression analyses included 55 participants because six cases required removal of P3/P4, P7/P8, or F7/F8 electrodes due to electrical failure or high signal impedance (four 6-year-olds and two 8-year -olds).

The results of the multiple regression analyses investigating age, EVT percentile rank scores, and frontal theta EEG was significant, F(6, 48) = 2.77, p < .05 and accounted for 26% of the variance in memory binding performance (Table 2). An examination of the regression weights revealed that age and EVT scores accounted for unique variance in memory binding performance but frontal theta EEG power did not. The results of the multiple regression analyses investigating age, EVT percentile rank scores, and parietal theta EEG was significant, F(6, 48) = 3.61, p < .01, and accounted for 31% of the variance in memory binding performance (Table 3). An examination of the regression weights revealed that age, EVT scores, and left hemisphere medial parietal theta EEG at P3 accounted for unique variance in memory binding performance.

TABLE 2.

Regression analysis of frontal theta EEG predicting memory binding performance (N = 55) [Corrections added on 17th May, 2021 after first online publication: “thet” was changed to “theta” below.]

	B	SE (B)	β	t
Dependent variable: Combination condition performance
Retrieval EEG: Frontal theta
Age	.77	.23	.46	3.33**
EVT	.02	.01	.33	2.56*
F3 EEG	.16	.33	.08	0.49
F4 EEG	.04	.36	.02	0.11
F7 EEG	−.55	.43	−.25	−1.28
F8 EEG	.49	.50	.23	0.97
R 2	.26*

Note:

^*p < .05.

^**p < .01.

Retrieval EEG and d’ sensitivity scores in the combination condition.

TABLE 3.

Regression analysis of parietal theta EEG predicting memory binding performance (N = 55)

	B	SE (B)	β	t
Dependent variable: Combination condition performance
Retrieval EEG: Parietal theta
Age	.76	.23	.45	3.35**
EVT	.01	.01	.29	2.31*
P3 EEG	1.14	.50	.51	2.29*
P4 EEG	−.35	.42	−.17	−0.83
P7 EEG	−.61	.38	−.26	−1.62
P8 EEG	−.17	.48	−.07	−0.34
R 2	.31**

Note:

^*p < .05.

^**p < .01.

Retrieval EEG and d’ sensitivity scores in the combination condition.

3.8 |. Summary of EEG findings

Across all three conditions (i.e., object, background, and combination) theta EEG power values differentiated the various processing stages (baseline, encoding, and retrieval) of the memory binding task for both 6- and 8-year-olds. Baseline-to-retrieval-phase increases in theta power were evident across all electrode sites for the object, background, and combination conditions. In addition, retrieval power values were also higher than encoding power values across select parietal electrode regions for the object condition and most regions for the background and combination conditions. In contrast, baseline-to-encoding phase increases in theta power were only evident at select medial frontal, lateral frontal, and temporal electrode regions for the three conditions. An examination of brain–behavior associations revealed that left parietal theta EEG power during retrieval contributed to variability in memory binding performance

4 |. DISCUSSION

The present investigation provides an important contribution to our understanding of neural oscillations and memory processes during middle childhood development. To our knowledge, this is the first investigation examining age-related differences in encoding- and retrieval-related theta EEG power during memory binding. Regression analyses also provided valuable insight to brain–behavior associations and revealed that individual differences in memory binding performance are associated with parietal theta rhythms. These findings have important implications for our understanding of the neural systems that support recollection for contextual details during middle childhood.

With respect to behavioral performance, 6- and 8-year-olds were tested on memory for individual items (object and backgrounds) and their paired combination. Memory for individual item information was comparable across both age groups. However, younger children experienced greater difficulty discriminating between original and rearranged pairs in the combination condition, resulting in an increased false alarm rate. In contrast, the hit rate in the combination condition was equivalent across age, ruling out the possibility that younger children simply had difficulty remembering paired features. These findings extend the existing developmental literature on age-related changes in memory binding. While pronounced improvement in the ability to recall contextual information occurs between ages 4 and 6 during early childhood (Drummey & Newcombe, 2002; Sluzenski et al., 2006; Riggins, 2014), our findings suggest this does not mean that memory binding skills are therefore intact by age 6. We acknowledge as a limitation that the developmental time span between 6 and 8 years of age is rather narrow. However, our results lend support to the idea that there is room for improvement in memory binding processes during middle childhood and beyond. In fact, Lorsbach and Reimer (2005) tested 9-, 12-, and 21-year-old students on memory for item, location, and item-location binding. Feature and binding memory improved across childhood and into adulthood, and age differences were greatest for tasks that required memory for combined features.

There is also evidence to suggest that more complicated binding structures exhibit later developmental trajectories. In one investigation, Guillery-Girard et al. (2013) investigated item-feature (i.e., factual associative memory), item-location (i.e., spatial associative memory), and item-sequence (i.e., temporal associative memory) binding in a sample of children, adolescents, and adults (ages 6–23 years). Factual associative memory continuously increased until 10 years of age, and temporal associative memory increased from 9 to 10 years of age. In contrast, spatial associative memory showed a more protracted developmental trajectory, with continuous improvement occurring into adulthood. In addition, the ability to form increasingly complicated two- and three-way binding structures undergoes substantial development between age 7 to adulthood (Yim et al., 2013). Thus, future investigations should examine how the nature of the bound representation influences age-related differences in memory binding in order to obtain a more comprehensive picture of its development.

With respect to memory errors, the false alarms that younger children made in the combination condition were to previously presented object and background items that were rearranged to form a novel pairing. Younger children may have been relying on a global sense of familiarity to the rearranged items, which resulted in the endorsement of the item as previously studied. A child could avoid making such an error if they recollected that the object had been paired with a different background at study. Thus, the discrepancy in the false alarm rates for rearranged pairings could be attributed to the fact that familiarity develops earlier than recollection (A. Ghetti & Angelini, 2008). Another potential explanation, as noted by Richmond and Power (2014), is that the task may also have recruited the use of inhibitory control, as children needed to refrain from the tendency to say “yes” to trials that contained familiar items and may have experienced difficulty rejecting the familiarity signals from rearranged pairings. Executive control processes facilitate the binding of factual information with accompanying source-specifying details (Raj & Bell, 2010; Rajan & Bell, 2015; Rajan et al., 2014) and is associated with non-unitized memory binding representations (Blankenship & Riggins, 2015). Future investigations should examine the extent to which individual differences in memory binding can be explained by executive control processes.

4.1 |. Theta EEG and memory processes

In adults, theta synchronization is associated with episodic encoding and retrieval (Klimesch, 1999), but less is known about the functional role of theta activation during middle childhood. Our results revealed that both 6- and 8-year-olds exhibit task-related increases in theta EEG power when comparing baseline to memory processing, as well as differences in theta power between encoding and retrieval phases. We failed to find any significant interactions between memory processing stage and age, suggesting that this pattern does not vary as a function of age. When considering 6- and 8-year-olds’ behavioral performance and high accuracy (i.e., hit rate) for both age groups, this could potentially explain why children exhibited similar patterns of task-related changes in theta EEG power at 6 and 8 years. Relative to baseline, we found widespread retrieval-related increases in theta EEG power (all electrode sites) for the object, background, and combination conditions. Encoding-related increases, on the other hand, were only evident for select regions (medial frontal, lateral frontal, and temporal) across the three conditions. In addition, theta EEG power differentiated between memory encoding and retrieval processes; retrieval power was higher than encoding for the object (medial parietal and lateral parietal regions), location, and combination conditions (most electrode regions).

These findings parallel evidence of retrieval-related increases in theta EEG power being more widespread across the scalp and differentiating from memory encoding during explicit memory development in toddlers (Cuevas et al., 2012). In addition, these electrophysiological results are consistent with prior behavioral data highlighting the importance of retrieval in supporting memory binding (Lloyd et al., 2009). To elaborate, efficiency in memory binding could be attributed to improvement in encoding the link between item and context or could be attributed to successful retrieval of stimuli that co-occur. To answer this question, Lloyd et al. (2009) investigated 4- and 6-year-olds’ performance on a binding task that varied as a function of list length to tax working memory or long-term memory. Younger children had more difficulty remembering bound items, but only when there was a long list of study items, suggesting that age-related improvement in memory binding is primarily driven by enhanced retrieval from long-term memory, which may potentially be linked to widespread retrieval-related increases in theta power.

Another novel contribution was our examination of whether task-related changes in theta EEG power predict variability in memory binding performance. Regression analyses revealed that higher memory binding performance was associated with retrieval-related increases in parietal theta EEG. With respect to parietal cortex involvement, prior studies in adults suggest that parietal network structures are involved in the perceptual binding of feature information (Uncapher et al., 2006), allocating attention toward relevant contextual features to aid retrieval search, monitoring, and verification (Cabeza et al., 2008; Olson & Berryhill, 2009), and in gauging the subjective vividness and confidence of one’s retrieved memories (Ally et al., 2008). Understanding the contribution of the parietal cortex to contextual recollection during childhood development would be an important question for future developmental studies to address.

In adults, theta power increases at frontal and parietal scalp locations are associated with successful item-source retrieval (Addante et al., 2011), and phase synchronization between frontal and posterior (including parietal) regions is thought to reflect top-down control to the hippocampus and posterior cortex to modulate episodic retrieval (Nyhus & Curran, 2010). Although memory-related increases in theta power were observed at frontal scalp locations during retrieval, frontal theta power failed to predict memory binding performance. Similar findings were reported in an ERP investigation examining age-related changes in recollection. Whereas both children and adults demonstrated parietal old/new effects associated with item-source recollection, children failed to exhibit late frontal ERP components associated with monitoring and control of memory retrieval, which may reflect a reliance on parietal structures to support recollection due to protracted PFC development (Czernochowski et al., 2005).

While we did not have specific hypotheses about laterality effects, we found that only left parietal theta activation during retrieval predicted memory binding performance. In a study examining EEG correlates of item recognition memory in middle childhood, Diaz et al. (2018) report similar findings; only posterior parietal EEG in the left hemisphere was associated with individual differences in performance. The question of laterality effects needs to be further explored. It is possible that laterality effects depend on the type of material (i.e., verbal vs. nonverbal) used (Kim, 2011). For example, in a meta-analysis, Kim (2011) reported that studies using verbal materials exhibited left-lateralized effects, whereas studies using pictorial stimuli reported bilateral activation. Still, in another meta-analysis, Spaniol et al. (2009) report retrieval-related activation primarily in the left hemisphere for both verbal and nonverbal materials (Spaniol et al., 2009). However, these meta-analyses focused on adult populations. Clearly, further research is needed to examine whether laterality effects depend on the type of material (i.e., verbal vs. pictorial) used during middle childhood.

The present study has limitations. First, we focused on the time period between 6 and 8 years by design in order to examine whether memory binding improvement occurs beyond the early childhood age range (i.e., 4–6 years) documented in prior studies. However, we acknowledge this is a narrow age range and future studies would benefit from examining middle childhood beyond this restricted range. In addition, the cross-sectional design limits our understanding of the developmental trajectory of memory binding processes that could be better understood using longitudinal data. In addition, although the noninvasive nature of EEG is advantageous in developmental populations, there are limitations related to poor spatial resolution that make it difficult to infer localized changes in cortical versus sub-cortical neural activity. Future work incorporating neuroimaging techniques with better spatial resolution can help guide interpretations about functional specificity of prefrontal, hippocampal, and parietal structures in supporting memory binding.

In conclusion, we found that item-item associative binding continues to improve during middle childhood. Developmental differences were observed in the pattern of memory binding errors. Younger children had greater difficulty discriminating between original and rearranged pairs, which resulted in higher false alarm rates. In addition, our results suggest that theta synchronization is related to episodic encoding and retrieval and may be detected by scalp-recorded brain electrical activity. Widespread retrieval-related increases in theta band power (compared with baseline and encoding-related activation) were evident in middle childhood. We propose that theta synchronization potentially reflects the interaction between cortical structures and hippocampal-dependent binding mechanisms necessary for retrieving the contextual details of a memory episode (Nyhus & Curran, 2010). With respect to our examination of brain–behavior associations during middle childhood development, retrieval-related increases in parietal theta EEG power predicted variability in memory binding performance, potentially reflecting attentional allocation during memory retrieval search processes (Cabeza et al., 2008). Future electrophysiological investigations should employ a wider age span in order to understand the specific role that theta oscillations play in supporting memory binding processes across early and middle childhood development.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.