Attachment Security and Cortical Activity Measured by Functional Near-Infrared Spectroscopy During a Working Memory Task in Healthy Young Adults

Abstract

Introduction:

Attachment style has been associated with socio-emotional outcomes, however little evidence suggests a possible association with executive functioning. Few studies have demonstrated that attachment style mediates working memory and learning relationships. We hypothesized that attachment style affects performance and cortical activity patterns of working memory.

Methods:

We compared working memory performance and cortical activity in securely and insecurely attached first-year college students (N=49) using three n-back task conditions. Cortical activity was recorded by functional near-infrared spectroscopy during these three conditions of the n-back task. Attachment style was assessed using the Relationship Scale Questionnaire, categorized into four groups.

Results:

Both study groups showed similar working memory performance. The cortical representation of working memory was different between the two groups. The securely attached group demonstrated higher activity in the right superior frontal and superior-medial frontal areas across all n-back conditions as well as in the right superior frontal cortex during the two-back and three-back conditions. The insecurely attached group displayed higher activity in the bilateral supplementary motor area and the left premotor area only during the three-back condition.

Conclusion:

These findings emphasize the potential influence of attachment style on the cortical representation of working memory. Different activity maps between the two groups may reflect varying cognitive strategies employed to achieve a comparable working memory performance. Moreover, these results suggest that each style may have a distinct strategy to achieve attachment-relevant and irrelevant neurocognitive tasks.

INTRODUCTION

Early experiences of infant-caregiver relationships constitute mental representations called internal working models (IWMs) of attachment. These working models comprise several elements, including episodic memories, beliefs, goals, and plans. However, executive functioning may also be related to attachment style. Executive functions encompass the flexible control of attention, the ability to hold information through working memory and the maintenance of inhibitory control. These three executive function skills are the fundamental abilities for children and adults to achieve all kinds of daily life goals (1). Zimmermann et al. (2015) proposed that problem-solving, information processing, decision-making, social evaluation, perception, pattern recognition, attention, memory, and cognitive control comprises subprocesses associated with attachment (2). Del Villano et. al. (2014) demonstrated that learning processes are closely related to working memory (WM) and are mediated by attachment style (3). Mares et al. (2020) stated that early relationships shape neuro-biological development, influencing social and psychological functioning, executive functions, information processing, and the development of language and cognitive abilities (4). Blair et al. (2018) used WM, planning, organization of materials, task completion, behavioral inhibition, set-shifting, emotional regulation, and monitoring skills to investigate attachment-developmental-cognitive model and its association with psychotic-like experiences (PLE). Their obtained results showed that high attachment insecurity and executive function deficits were significantly related to PLE. However, they underlined that they did not examine each executive function specifically concerning PLE (5). Menon et al. (2020) stated that executive functions and attachment are interconnected systems that thrive early in life. Therefore, they investigated executive functions (i.e., WM, cognitive flexibility, and inhibitory control) in three groups: secure, avoidant, and disorganized groups. Results showed resemblances between secure and avoidant children’s executive functions. Differences in the performance of executive functions amongst insecure subtypes were minimal (6).

Highlights

• Differently attached individuals displayed similar working memory (WM) performance.

• Attachment style may affect the cognitive mechanisms used in the WM task.

• Attachment may be associated with WM representations in terms of cortical activity.

• Attachment style may impact cognitive neural circuits beyond socio-emotional domain.

Published empirical research regarding attachment and executive functioning in middle childhood and adulthood is comparatively scarce. As mentioned above, an important issue, is that studies generally use several executive functions associated with attachment. This method may yield results that are hard to interpret. Working memory studies allow the analysis of atypical development patterns in children with neurodevelopmental conditions or those who demonstrate WM difficulties without specific diagnoses (7). If we consider that IWMs represent relations with one’s self and significant others, or emotional experiences, it is WM that handles the processing of emotional information, encoding, storing, retrieving and manipulating past information when needed in the face of novel conditions. Given existing literature and knowledge, it is reasonable to investigate the association between one of the specific executive functions –WM– and attachment.

The prefrontal cortex (PFC) is the essential cortical region for executive functioning. Similar to the development of executive functions, the PFC displays a long postnatal maturation that extends beyond the adolescence period. This slow differentiation of the PFC provides an extended period of plasticity, which may be sensitive to environmental effects (8).

Functional near-infrared spectroscopy (fNIRS) is a novel neuroimaging practice that measures human cerebral cortical activity by detecting temporal changes in oxygenated hemoglobin (Oxy-Hb) levels. fNIRS sensitivity to motion artifacts is comparatively low, and brain activity is measured in a natural sitting position without physical restriction, allowing for a higher representative design, particularly in psychological experiments. Besides, fNIRS can reliably gauge cortical activity throughout the WM task, which is the n-back paradigm (9).

Considering the proposed relationship between attachment security and executive functions as well as the parallel development of executive functions and the PFC, we aimed to examine whether or not the previously reported association between executive functions, especially in terms of WM and attachment security is biologically represented in terms of PFC activity during an executive function task in healthy young adults. Furthermore, we hypothesized that WM performance and PFC activity would differ between securely and insecurely attached healthy young adults in a WM task.

METHOD

Subjects

Participants consisted of first- and second-grade university students. All grade-one students were informed about the study procedure in May 2016. Out of 494 grade-one students, 71 agreed to join the study. Thirty-three (67.3%) participants were female and the mean age was 19±1.6 years. Students with a current psychiatric diagnosis, a history of mental disorder, use of psychotropic medication, neurological disease, another chronic general medical condition, head trauma that resulting in unconsciousness for more than 30 minutes, and those with alcohol or substance abuse were excluded (N=22). Since the acquisition of neuroimaging data was postponed until April 2017, nine subjects had already progressed to term two during enrollment. All participants signed the written informed consent. Students were not promised extra credit for research participation, and refusal had no consequences. The Ethics Committee approved all study procedures on the 24th of March 2016 (Ankara University Ethics Committee Approval no: 03-24-2016/08/108).

Materials

Assessment of Attachment Styles

There are substantial differences among measurement techniques for adult attachment styles, yet a standard way has not yet been established. Nonetheless, the Relationship Scale Questionnaire (RSQ) is assumed to be capable of revealing essential aspects of individuals’ mental processes and behaviors in close relationships. The RSQ comprises thirty items, and each item is graded on a seven-point scale (from “not at all like me” to “very much like me”) to evaluate the attachment patterns of participants. The questionnaire is competent in distinguishing secure, avoidant, preoccupied, and fearful attachment styles (10). The reliability and validity of the Turkish version of the Relationship Scale Questionnaire were tested in a series of studies by comparing it with Turkish and United States samples. These studies demonstrated that Turkish students could sufficiently identify the four attachment styles. Furthermore, results pointed out that the four attachment styles were organized into two underlying dimensions of attachment: The model of self and others. Therefore, for the purpose of assessing attachment styles in young Turkish adults, RSQ is a valid and reliable tool (11). Consequently, we employed the Turkish version of the RSQ to measure participants’ attachment styles. We used Griffin and Bartholomew’s original RSQ scale’s Turkish version with the same items as the original scale to determine each attachment category.

The n-back Paradigm

The n-back paradigm is a WM task with varying task loads, commonly employed in neuroimaging studies. Research has revealed that the rostral PFC, dorsolateral PFC, middle and ventrolateral PFC, posterior parietal, premotor and dorsal cingulate cortices exhibit consistent activation during the n-back paradigm. However, the functional role of these PFC’s sub-regions in WM is controversial. Additionally, the anterior insula is crucial in buttressing subjective feeling states, cognitive functions, and processing salient information. Furthermore, prefrontal activity is sensitive to the assigned task difficulty during the n-back tasks (12).

The Neuroimaging Task

The n-back task consisted of zero-back, two-back, and three-back conditions implemented as the activation paradigm. For each single task block, there were pre (32.5 seconds) and post (12 seconds) task baseline (control) blocks (Figure 1a). During those control blocks, subjects were presented with a “+” sign in the middle of the screen. Cue words (zero/two/three back) were presented at the top of the screen and remained there for two seconds before each condition to inform the participant about the presented n-back level. Participants performed four practice trials before the experiment: two trials for zero-back (A block), one for two-back (B block) and one for the three-back task (C block). The order of A, B, and C blocks was randomized (Figure 1b). Each n-back block consisted of six target and 14 non-target letters. Thus, the ratios of targets/all letters and non-targets/all letters were 0.3 and 0.7, respectively. The letters (C, D, F, G, H, J, K, and B) appeared on the monitor for 500 ms, with an inter-stimulus interval of 2000 ms (Figure 1c). Subjects were expected to press the space button whenever a target was displayed in the middle of a 19-inch screen by using a keyboard with the dominant hand, and to withhold the response otherwise. Subjects were positioned at a 55 cm distance from the monitor. The letter style and font size used for all letters were Palatino Linotype and 30, respectively. In the zero-back task, the letter “B” was the target; while, in the two-back and the three-back tasks, the target letter could be any letter following the n-back rule. The temporal order of target and non-target letters was randomly changed for each participant. We focused on the two behavioral outcomes of the n-back task, namely, i) the hit rate and ii) d-prime scores. D-prime scores were calculated with the following formula in Microsoft Excel (Microsoft Office, 2016): Hit Rate = True Positive / (True Positive + False Negative); False Alarm Rate = False Positive / (False Positive + True Negative); z (Hit Rate) = NORM.S.INV (Hit Rate); z (False Alarm Rate) = NORM.S.INV (False Alarm Rate); d’ = z (Hit Rate) – z (False Alarm Rate).

Figure 1.

The n-back task as the activation paradigm.

Therefore, the formulas given above indicate that the calculated d-prime score is the product of true-positive (hit), false-negative (miss), true-negative (correct rejection), and false-positive (false alarm) scores. Thus d-prime scores can be considered as the sensitivity index that represents how well a signal is distinguishable from the noise (13).

Functional Near-infrared Spectroscopy

We utilized the 52-channel fNIRS device (ETG-4000; Hitachi Medical Co., Tokyo, Japan) for all measures of cortical activity throughout the n-back paradigm. fNIRS does not precisely measure cortical activity levels; rather, it detects relative alterations in oxygenated (Δoxy-Hb) and deoxygenated hemoglobin (Δdeoxy-Hb) via source/detector pairs that emit and receive two wavelengths of infra-red light (695 and 830 nm). The weakening of light data is converted into hemoglobin concentrations based on the Beer-Lambert law. Therefore, fNIRS enables noninvasive monitoring of the alterations in cerebral hemoglobin concentration (in mMmm), the main near-infrared absorbing type in the brain. Source/detector pairs were positioned at a distance of 3.0 cm; defining the channels as the area between these pairs. A 3 cm distance between the optodes allows to quantification of Δoxy-Hb and Δdeoxy-Hb at a depth of 2-3 cm from the scalp corresponding to the cortical surface. The optodes were securely attached to participants’ scalps by a 3×11 probe holder. Specifically, channel 11 was placed over F7 on the left of the scalp, and T4 interspersed the midpoint between Channels 8 and 9 on the right side. Figure 2 displays the positions of fNIRS channels placed on the scalp, aligned with the international 10-20 system used for electroencephalogram (EEG) placement..

Figure 2.

Position of the measurement channels on the scalp according to the international 10–20 electroencephalography system.

To visualize channel projections on the cortex, the 52 optode channels derived from 33 optodes in the 3×11 shell were projected on the rendered cortex (Figure 3). Initially, we spatially registered the 3D digitizer file with the standard Montreal Neurological Institute (MNI) template cortex using “fNIRS functions” in the NIRS SPM (Statistical Parametric Mapping) toolbox. While NIRS offers several advantages, it also has disadvantages such as low spatial resolution and reduced signal-to-noise ratio (SNR) (9).

Figure 3.

Optode channel projections on the standard MNI cortex.

Because the data is recorded from the scalp surface, mapping the activity of underlying brain space to the fNIRS channels space measurements depends on the positions of the sensors, the anatomy of the head/brain and the tissues between the brain and fNIRS probes. Although neuroimaging devices provide software and techniques to overcome displacement/misplacement issues, the data analysis process may introduce further complexities for researchers. Using the region of interest (ROI) approach can help mitigate these challenges. While no definitive method currently exists, employing quantitative methods like meta-analysis is recommended for selecting ROIs (14).

Yaple et al. (2019) conducted a quantitative meta-analysis to assess congruity and age-related alterations throughout the healthy adult life span within brain regions involved in performing the n-back task. The authors identified 15 clusters of concordant brain region activations linked to the task in young adults. That study critically showed that the prefrontal cortex involvement is congruent for young adults, to a lesser degree for middle-aged adults, and absent in older adults (15).

In the present study, we selected eleven regions of interest out of the 52 measurement channels with higher than 50% overlap with the Brodmann areas reported in young adults in the Yaple et al. study (15). These eleven regions of interest and their overlap rates with corresponding Brodmann areas (BA) were as follows; Left BA6 [Channels 46 (97%) and 47 (100%)], Left BA 10 [Channel 4 (63%)], Left BA9 [Channels 5 (67%) and 24 (72%)], Right BA6 [Channels 48 (100%) and 50 (90%)], Right BA9 [Channels 6 (68%) and 29 (56%)], and right BA8 [Channels 27 (81%) and 28 (85%)].

The fNIRS tool measures relative alterations of oxyhemoglobin concentrations based on a baseline. The activity during the ‘+’ sign served as the baseline activity. The pre-task baseline was defined as the mean over a 9 s baseline period preceding the task period, and the post-task baseline was defined as the mean over the last 7 s of the post-task baseline period; linear fitting was applied to the data between these two baselines. The tasks consisted of two runs of one-back, two-back, and three-back conditions, randomly presented to each participant. The sampling frequency was set at 10 Hz. Physiological activities such as systemic arterial pulse oscillations (0.1 Hz) and respiration (0.2–0.3 Hz) are often associated with fNIRS signal fluctuations (9). Therefore, we applied a moving average filter with 5 s windows to eliminate short-term motion artifacts and correct such fluctuations in the analyzed data.

An incisive alteration in signal exceeding 0.4 mMmm in more than twenty consecutive samples was labeled as a body movement artifact (BMA) by the ETG-4000 device. A researcher (AK) who was blind to the study group members re-evaluated these artifacts to identify the responsible individual channels. As a result, channels with BMAs were excluded from further analyses. Since oxy-Hb alteration is presumed to more directly reflect cognitive activation and provide a higher signal/noise ratio than deoxy-Hb, and the oxy-Hb is the most sensitive indicator of changes in regional cerebral blood flow (rCBF) (16), we utilized the Δoxy-Hb data to measure cortical activity.

Statistics

We compared socio-demographic and clinical data between the groups with the independent samples t-test and the Mann-Whitney U test, depending on the data distribution. We also compared the secure (SG) and insecure groups (ISG) with the Mann-Whitney U test according to the number of correct responses during the two-back and three-back conditions as well as d-prime scores. ΔOxy-Hb measurements obtained from 11 channels across all three n-back conditions were analyzed using a 2 (Group: Secure vs Insecure) × 3 (Condition: zero-back vs two-back vs three-back) × 11 (Channels) mixed analysis of variance (ANOVA) design. The between-subject independent variable was the group, and the within-subject independent variables were Condition and Channels, respectively. We applied Greenhouse-Geisser corrections to address violations of the sphericity assumtion, and employed Bonferroni corrections to mitigate the risk of Type-1 errors stemming from multiple testing. Finally, we calculated Spearman correlation coefficients to assess the correlations between RSQ subscale scores (secure, dismissing, fearful and preoccupied) and cortical activity throughout the n-back task. We utilized the False Discovery Rate (FDR) method to control Type-1 errors, resulting from multiple comparisons in these correlation analyses (FDR α=0.05) (17).

RESULTS

Self-Report Questionnaire Results

Based on the two-factor solution of the RSQ, 15 students were securely attached, while 34 were insecurely attached. The RSQ subscale scores indicated that 11 students were fearful, two were preoccupied, and 21 were dismissively attached among the ISG.

Behavioural Results

We initially compared n-back performance among secure and insecure groups. The comparison of the two groups for the two-back and three-back, true-positive, and d-prime scores is displayed in Table 1. Although the secure group’s performance was numerically higher than that of the insecure group, we did not detect a statistically significant difference between the two study groups regarding two-back and three-back performance.

Table 1.

Comparison of the two groups in terms of n-back performance

	SG	ISG	Mann-Whitney U test
2-back-TP median (minimum-maximum)	5 (1–6)	5 (0–6)	Z=-1.46, p=0.143
3-back-TP median (minimum-maximum)	4 (1–7)	3 (0–6)	Z=-0.91, p=0.362
2-back-d-prime median (minimum-maximum)	3.24 (0.33–3.29)	2.58 (-0.18–3.30)	Z=-1.29, p=0.197
3-back-d-prime median (minimum-maximum)	1.45 (0.33–3.33)	1.52 (-0.07–3.24)	Z=-0.65, p=0.94

ISG: Insecure group; SG: Secure group; TP: True positives.

Neuroimaging Results

We observed a significant main effect for Condition [F (2.88)=4.86, p=0.01, η²p=0.099]. Post hoc tests revealed that this was rassociated with higher activity in the two-back condition compared to the zero-back condition (Mean Difference: 0.021, Standard Error=0.007, p=0.02, 95%CI: 0.003–0.039). The main effect of Group was also [F (1,44)=4.05, p=0.05, η²p=0.084]. This significance arises from greater activity among the securely attached participants compared to insecurely attached participants (Mean Difference: 0.015, Standard Error=0.007, p=0.05, 95% CI: 0.002–0.029). While the Group × Condition interaction was not significant, Group × Channel [F (6.54, 287.55)=5.95, p=0.000, η²p=0.119] and Group × Condition × Channel interactions were significant [F (9.34, 410.94)=1.97, p=0.039, η²p=0.043]. We present the approach to this triple interaction from the ‘Group’ perspective in Table 2 and Figure 4.

Table 2.

Significant ‘Channel × Condition × Group’ interaction from the perspective of ‘Group’ comparisons at each level of ‘Channel’ and ‘Condition’

Channel	BA (percent overlap), (cortical projection)	MNI coordinates			Task	Group	MD	SE	P
		X	Y	Z
Ch 27	R-BA 8 (0.60) (sup. frontal), R-BA 6 (0.40) (sup-med. frontal)	41	9	62	0-back	SG >ISG	0.032	0.014	0.026
Ch 6	R-BA 8 (0.19) (sup-mid frontal) BA 9 (0.67) (sup. frontal)	50	29	41	2-back	SG >ISG	0.053	0.022	0.019
Ch 24	L-BA 8 (0.82) (inf. frontal) L-BA 9 (0.17) (mid. frontal)	-11	40	57	2-back	SG >ISG	0.063	0.030	0.040
Ch 27	R-BA 8 (0.60) (sup-frontal), R-BA 6 (0.40) (sup-med frontal)	41	9	62	2-back	SG >ISG	0.075	0.028	0.010
Ch 29	R-BA 6 (0.20) (inf. frontal pars opercularis & triangularis)	68	-26	42	2-back	SG >ISG	0.110	0.031	0.001
Ch 6	R-BA 8 (0.19) (sup-mid. frontal) BA 9 (0.67) (sup. frontal)	50	29	41	3-back	SG >ISG	0.064	0.021	0.003
Ch 27	R-BA 8 (0.60) (sup. frontal), R-BA 6 (0.40) (sup-med. frontal)	41	9	62	3-back	SG >ISG	0.89	0.029	0.003
Ch 29	R-BA 6 (0.20) (inf. frontal pars opercularis & triangularis)	68	-26	42	3-back	SG >ISG	0.094	0.031	0.004
Ch 46	L-BA 6 (1) (precentral, mid. frontal)	-6	7	73	3-back	ISG >SG	-0.055	0.021	0.013
Ch 47	R-BA 6 (1) (supplementary motor area, sup. frontal)	16	0	75	3-back	ISG >SG	-0.074	0.027	0.009
Ch 48	R-BA 6 (0.99) supplementary motor area, sup. frontal)	30	-11	73	3-back	ISG >SG	-0.048	0.019	0.013

BA: Brodmann area; Ch: Channel; Inf: Inferior; ISG: Insecure group; L: Left; MD: Mean Difference; Med: Medial; Mid: Middle; R: Right; SE: Standard Error;SG: Secure group; Sup: Superior.

Figure 4.

Cortical activity during the three n-back conditions in the two study groups.

The ISG’s rate of true negatives was positively correlated with activity in Channel 29 during the two-back condition (r=0.37, p=0.029), and the rate of false negatives was negatively correlated with activity in Channel 27 (r=-0.38, p=0.027). There were no correlations between behavioral performance in the 3-back condition and brain activity.

DISCUSSION

In a cross-sectional design, we compared cortical activity throughout a WM task between securely and insecurely attached healthy young adults. No differentiation was observed between the two study samples in terms of n-back task performance.

While to our knowledge no previous studies have directly investigated the relationship between a WM task and attachment security, based on prior research, we expected SG to outperform the ISG in a WM task. Some previous studies suggested a potential role of attachment in cognitive functions. For example, Cao et al. (2018) examined the attachment and episodic specificity association in memories that are attachment-relevant and irrelevant. Individuals with secure attachment produced more internal details and fewer external details in attachment-relevant tasks compared to attachment-irrelevant tasks (18). Similar studies focused on how emotional and attachment-related stimuli impact cognitive outcomes. A common theme in these studies is the comparison between clinical and healthy subjects or the use of attachment-relevant or irrelevant stimuli to assess executive functions.

Moreover, studies exploring the role of attachment on general cognitive ability are scarce. In a prospective study, Bernier et al. (2015) showed that kindergarteners, who were more securely attached to their mothers in toddlerhood, outperformed the insecurely attached ones on all executive functioning tasks. Teachers also reported that these securely attached children exhibit fewer executive functioning difficulties in everyday school situations (19). However, the present research did not identify a relationship between n-back performance and attachment security. Inconsistent results could potentially stem from participant selection. Our study enrolled first-grade university students from a prestigious medical faculty chosen through a highly competitive examination. Consequently, our sample consists of young healthy adults from the top percentile of academic success. Although this sample selection was specifically preferred to increase the internal validity of the study design, it might have masked a possible relationship between attachment and WM performance. Additionally, we did not use any emotional stimuli that could potentially interfere with WM performance. Our results may suggest that attachment system may not exert a substantial impact on observable outcomes in individuals from the perspective of an external observer.

Nonetheless, we showed that attachment security and insecurity are linked to diverse representations of WM as evidenced by cortical activity. Despite ongoing discussions about the psychometric properties of the n-back task and its utility in measuring individual differences in WM, it has been shown that n-back task is valuable for experimental studies and predicts inter-individual differences in WM, particularly under higher loads. The n-back tasks involve several processes, such as identification, selection, maintenance, decision-making, interference, suppression, and separation. As the ‘n’ value increases, the subject’s memory load increases and performance is expected to decline (20).

The ‘Condition’ main effect observed in the ANOVA indicates that regardless of group membership and measurement channels, all three conditions of the n-back task (zero, two, and three-back) elicit distinct brain activity patterns. The two-back condition which imposes a higher load corresponds to a higher cortical activity than the zero-back task. Unexpectedly, brain activity between the three-back and zero-back conditions was not different. The relatively low d-prime scores in the three-back condition in both groups (Table 1) reflect the challenging nature of the three-back task, which might have caused a Type-2 error possibly related to a floor effect.

The significant ‘Group’ main effect indicates that the participants showed distinct cortical activity throughout the task paradigm depending on their attachment style. Post-hoc analysis showed cerebral cortical activity was higher in the SG than in the ISG. Vrtička and Vuilleumier (2012) put forward a functional neuroanatomical explanation for the impact of the adult attachment style on social processing, which suggests that attachment anxiety and avoidance may lead to lower activity in dorsolateral and orbitofrontal cortices during emotion regulation (21). In another study, attachment security was associated with brain activity during mental state judgments in the Eyes Task (22). However, based on our information, this study initially displays a relation among attachment security and cortical activity during a neuro-cognitive task. This outcome suggests that attachment security may impact neural circuits beyond the socio-emotional domain. This suggestion is consistent with previous studies, which point out that early environmental stress may have a role in the emergence of attachment patterns and may impact neurogenesis, synaptogenesis, and neuronal networks, mediated through epigenetic, neuroendocrine, and cellular adaptation mechanisms. In a review, McCrory et al. (2010) explain early adversity’s functional, structural, and genetic-environmental effects. They suggest that the hypothalamic-pituitary-adrenal (HPA) axis, hippocampus, amygdala, OFC, PFC, cerebellum, and corpus callosum shows structural and functional abnormalities in maltreated or traumatized adults (23). These areas are related to memory, emotion regulation, monitoring and stress response. Although they show abnormalities, these results are obtained by comparing healthy subjects to clinical subjects. Another study showed reciprocal developmental connection among WM and social experiences in elementary school students (24). Buonomano et al. (2009) used computational modeling work and showed the way that memory and processed information can be combined to be overwritten in the identical circuits. They showed cortical neurons do not simply encode current stimulus but generate representations of each incoming stimulus in the context of the previous stimulus (25). Hasson et al. (2015) support this idea and ask, “How does the very first information persistently shape the present moment?” and answer this question by stating that at the biological level, preliminary processed data constantly impacts information processing in the present time and propose that memory is innate to all neural processes. The processing-memory framework suggests that the information processing neurons and information storing neurons are the same neurons and same circuits (26). Although billions of pieces of information are processed in subcortical circuits daily, we can consciously process much less at the cortical level. In this research, we do not claim that WM has only a cortical representation or is encoded only in cortical circuits. We mean that the idea that memory is constantly combined and overwritten in cortico-cortical, cortical-subcortical, and subcortical-subcortical circuits, just as Buonomano and colleagues showed and Hasson and colleagues agreed, offers an important and valuable perspective. Attachment is an innate and inevitable system in our biological codes, and its purpose is that the organism survives and reaches the reproductive age. This system is necessary to develop successful reproductive strategies. In addition, throughout life, we try to find a solution to any challenge or obstacle we face by applying different solution strategies. Each attachment style develops its way of coping with the obstacles when encountered. These coping styles may or may not be effective.

We conclude that attachment styles are part of such a neuronal network. In other words, emotional, behavioral, and cognitive domains are affected differently for every individual. These findings suggest that attachment patterns need new explanations other than self-regulatory behaviours, which involve synchronizing emotional arousal and cognitive control. This suggestion is not only limited to the socio-emotional field but is also related to all processes we encounter throughout our lives such as decision-making, learning, monitoring, making judgements, school competence, spouse selection, and psychopathologic conditions. In summary, a new understanding regarding attachment system may open a new road to improve our understanding of how the brain functions.

The significant ‘Group’ × ‘Condition’ × ‘Channel’ triple interaction indicates that young adults with secure and insecure attachment display distinct cortical activity depending on task conditions and cortical areas of interest. When the post-hoc analyses are examined from the ‘Group’ perspective, results demonstrate that the SG displayed higher activity in superior frontal and superior-medial frontal areas (BA 8) during all n-back conditions, including the zero-back condition. Furthermore, this area was more active in the SG than in the ISG during two-back (channels 6 and 24) and the three-back conditions (Channel 6). This area, therefore, seems to be associated with a general requirement of the n-back task. A functional magnetic resonance imaging (fMRI) study set out that activity in BA 8 is sensitive to parametrically differing levels of uncertainty (27). High activity within this area may suggest a higher effort to dissolve uncertainty in all steps of the n-back task among securely compared to insecurely attached participants.

Similarly, channel 6, which projects to the right BA 9, 46 and a lesser extent 8, was more active in SG than in the ISG during two-back and three-back conditions. This area is generally engaged in WM and planning. The n-back task is primarily a WM task, and securely attached participants activated these areas directly associated with WM more than insecurely attached participants.

On the other hand, channels 46, 47 and 48 were more active during the three-back condition in the ISG than in the SG (Table 2). These channels project the precentral gyrus and supplementary motor areas (SMA). Given that the only motor requirement in the n-back task is pressing a button for success in both groups, this finding is unlikely to reflect a higher motor involvement among the ISG group. It has been reported that cognitive and motor systems share similar evolutionary roots, which influence each other bi-directionally during development. Moreover, certain brain areas integrate motor as well as cognitive functions. Supplementary motor areas and pre-SMA have distinct cytoarchitectural properties and involve several cognitive and motor processes. Some neuroimaging studies have revealed that to maintain verbal information, the precentral gyrus and the supplementary motor area participate in the sub-vocal rehearsal system of the phonological loop of WM (28). Higher activity in this area among the ISG may remind us that insecurely attached participants may have employed such sub-vocal rehearsal strategies more than securely attached ones during the most challenging condition of the n-back task. Aron et al. (2007) showed that the SMA also plays a crucial part in inhibiting undesired responses (29). Therefore, higher activity in the SMA may suggest a higher involvement of an inhibition strategy among the ISG than the SG. We detected (i) a positive correlation between activity in Channel 29 and the amount of true negative reactions in the n-back task and (ii) an inverse correlation between the rate of false-negative responses and activity in Channel 27 only among the ISG. Since negative responses in the n-back task are connected with inhibition, that is, ‘being able not to press the response button’ , these correlations might also suggest higher involvement of inhibitory mechanisms in the ISG.

In conclusion, these findings suggest that the SG and ISG may have employed different cognitive mechanisms as the task difficulty increased in the n-back task. However, the two study groups exhibited comparable performance throughout the two-back and three-back conditions; obtained results may indicate that employing one cognitive mechanism was not more advantageous than the other.

The present study has certain limitations. The first is related to the study sample. We registered first-grade medical university students, and the results may not be generalized to all healthy young adults. Additionally, the number of participants decreased during the study as some participants dropped out during the research process. Secondly, fNIRS allows measuring cortical activity, and subcortical regions cannot be evaluated. However, WM depends mostly on cortical activity. Therefore, the first hypothesis of this study was not supported; that is, the two groups displayed a similar WM performance.

Nevertheless, some authors called attention to considering individual differences, such as cognitive style and encoding strategies, to interpret the findings of neuroimaging studies. For example, Sanfratello et al. (2014) indicated that even in a healthy and homogeneous group, individuals may display different brain activities while performing the same task; therefore, it should be kept in mind that individual differences may exist (30). This study, to our knowledge, is the first one to demonstrate that individual differences in attachment style may be associated with representations of WM in terms of cortical activity.