Attachment patterns and autobiographical episodic memory functioning: A systemic review of adult studies to advance clinical psychological science

Patterns of insecure attachment are associated with psychopathology but the mechanisms involved remain poorly understood. Cognitive science proposes that attachment patterns are influenced by the autobiographical memory system and in turn influence its ongoing functioning. Disturbances in autobiographical memory represent cognitive risks for later emotional difficulties. We systemically reviewed 33 studies (in 28 articles) examining the association between attachment patterns and autobiographical episodic memory (AEM) in individuals from the age of 16 (i.e., from young to older adulthood). Attachment patterns were associated with key areas of AEM phenomenology, including intensity and arousal; detail, specificity, and vividness; coherence and fragmentation; and accuracy and latency. These associations appeared to be moderated by contextual and individual factors; mediated by emotional regulation and schema-based processing; linked to mental health outcomes. Attachment patterns may also influence the impact of certain AEM-based manipulations. We conclude by providing a critical discussion and a research agenda for bringing together attachment, memory, and emotion, with a view to promote mechanism-driven treatment innovation in clinical psychology.

Patterns of insecure attachment are associated with psychopathology but the mechanisms involved remain poorly understood.Cognitive science proposes that attachment patterns are influenced by the autobiographical memory system and in turn influence its ongoing functioning.Disturbances in autobiographical memory represent cognitive risks for later emotional difficulties.We systemically reviewed 33 studies (in 28 articles) examining the association between attachment patterns and autobiographical episodic memory (AEM) in individuals from the age of 16 (i.e., from young to older adulthood).Attachment patterns were associated with key areas of AEM phenomenology, including intensity and arousal; detail, specificity, and vividness; coherence and fragmentation; and accuracy and latency.These associations appeared to be moderated by contextual and individual factors; mediated by emotional regulation and schema-based processing; linked to mental health outcomes.Attachment patterns may also influence the impact of certain AEM-based manipulations.We conclude by providing a critical discussion and a research agenda for bringing together attachment, memory, and emotion, with a view to promote mechanism-driven treatment innovation in clinical psychology.
A key tenet in developmental science is that experiences with caregiving figures contribute to different attachment patterns from early on in life (Ainsworth & Bowlby, 1991;Bowlby, 1969).These attachment patterns are thought to initially maximise the infant's survival (Bowlby, 1969) but also exert major influences later on adolescent and adult relationships (E.Waters, Weinfield, & Hamilton, 2000).Attachment ideas have attracted great interest from researchers and practitioners, providing a framework for understanding emotional disturbances across the lifespan.Although attachment insecurity in itself is not equivalent to psychopathology (Zilberstein, 2014), it is an established risk factor for mental health difficulties including depression and anxiety disorders (Mullen, 2019) and reduces the effectiveness of psychological treatments (Levy, Kivity, Johnson, & Gooch, 2018).However, the mechanisms through which attachment insecurity influences psychopathology remain poorly understood, greatly limiting our ability to develop more effective interventions to prevent and treat lifelong emotional difficulties (Barlow, Bullis, Comer, & Ametaj, 2013;Holmes et al., 2018) associated with attachment insecurity.Understanding how attachment influences mental healthat multiple levels of analysesrepresents a research priority (Insel et al., 2010).Bowlby (1969) argued that early attachment interactions result in affect-laden mental representations, also called "internal working models" (IWMs).Although their exact nature remains an issue of ongoing controversy (e.g., Rutter, 2014), many theorists posit that IWMs are intimately linked to the autobiographical memory system (Collins & Read, 1994;Conway, 2005;Crittenden, 2006;H. Waters & Waters, 2006), comprising script-like, semantic, verbal-based sequences of events of prototype distress episodes encompassing how other attachment figures recognise and respond to one's distress (H.Waters & Waters, 2006), which are in turn are derived from generalisation of unique affect-laden, sensory-based memories depicting specific interactions (Collins & Read, 1994).Not only IWMs have a cognitive basis in memory, but they can also serve as templates for subsequent information processing, including of autobiographical memories more broadly (Dykas & Cassidy, 2011).In this article, we will consider the role of autobiographical memories as one potential cognitive (modifiable) mechanism through which attachment patterns influence mental health in adulthood.

Conceptualising attachment patterns
Contemporary models in adults characterise attachment patterns using a dimensional approach (Fraley et al., 2015;Fraley, Waller, & Brennan, 2000).A dominant model by Brennan, Clark, and Shaver (1998) distinguishes two independent dimensions of attachment avoidance and attachment anxiety (Fig. 1).Individuals high in attachment anxiety over-activate the attachment system through hypervigilance to threat, preoccupation with loss and excessive care-seeking.Individuals high in attachment avoidance under-activate the attachment system by emphasising self-reliance, emotional suppression, and distancing from others.Individuals high in attachment security display low attachment anxiety and avoidance, thus having flexibility between relying on others and oneself according to the contexts and goals (Mikulincer & Shaver, 2007).

Autobiographical memory and psychopathology
Autobiographical memories refer to personally-meaningful memories formed in an everyday context (Conway, 2005;Conway & Pleydell-Pearce, 2000;Rubin, 2005).The Self-Memory Model (Conway & Pleydell-Pearce, 2000) argues that the autobiographical memory system is organised hierarchically, with the top starting with lifetime periods (e. g., "my first relationship"), followed by general events ("our holiday trips together"), which further contain event-specific knowledge ("the time we got lost kayaking around a Swedish Island while freezing").The latter involves highly-detailed and vivid recollections of individual episodes typically in the form of sensory-based mental images (Conway, 2001).We will focus on these autobiographical episodic memories (AEMs) as two of their key featuresmental imagery (Rubin, Schrauf, & Greenberg, 2003) and involuntary/intrusive retrieval (Berntsen & Hall, 2004) are key to emotional disorders (Brewin, Gregory, Lipton, & Burgess, 2010).Such an episodic memory system is also where affectladen and sensory-based memories of specific attachment-relevant interactions are formed (Collins & Read, 1994), which are the foundation for the presumably semantic, script-like aspects of IWMs (H.Waters & Waters, 2006).
A. Lau-Zhu et al. and recurrent trauma memories in post-traumatic stress disorder or PTSD (Ehlers, Hackmann, & Michael, 2004).Autobiographical memory disruptions are also apparent early in development signalling latent vulnerability in those susceptible to later psychopathology (McCrory et al., 2017).

Bringing together attachment, autobiographical memory, and emotions
IWMs underlying different attachment patterns can act as information-processing "filters" across a range of cognitive-affective domains (Dykas & Cassidy, 2011;Mikulincer & Shaver, 2007), including encoding, organisation and retrieval of autobiographical memories (Thompson, 2008).Dykas and Cassidy's dual-process model (2011) proposes that insecurely-attached (vs.securely-attached) individuals will process attachment-relevant information by i) excluding such information all together if it is likely to cause psychological pain (i.e., as an emotional regulation strategy) or ii) filtering it in a way that is consistent with their negative (vs. positive) attachment-related experiences (i.e., via schema-based processing).Consistently, avoidantlyattached individuals show tendencies to distance themselves from others during distress while anxiously-attached individuals appear to access attachment-relevant information more quickly such as their attachment figures' names (for a comprehensive review see Dykas & Cassidy, 2011).
Individuals are thought to be motivated to process attachmentrelevant information in autobiographical memory to maintain a coherent self-image, thus information that fits with attachment-related experiences is more likely to be later accessible and available (Conway, 2005;Conway & Pleydell-Pearce, 2000).Avoidant attachment has been linked to slower retrieval of childhood events (Dykas, Woodhouse, Jones, and Cassidy (2014)) and reduced benefit from security-based (memory-based) priming techniques (Bryant & Chan, 2017), while anxious attachment has been linked to increased memory errors for interpersonal events (Hudson & Fraley, 2018) and enhanced benefit from security-based priming techniques instead (Gillath & Karantzas, 2019).The possibility that IWMs bias information-processing in ways that reinforce their pre-existing structures may also partly explain their relative stability across the lifespan.

Review questions
The postulated attachment-related biases on autobiographical memories, alongside the data linking such memories and emotional psychopathology, suggest that attachment security/insecurity relates to later emotional disorders through its impact on AEM functioning.We aim to systemically review adult studies examining attachment patterns and AEM functioning, by building on earlier theoretical and empirical foundations connecting attachment and emotional memory in childhood (Chae, Goodman, & Edelstein, 2011;Farrar, Fasig, & Welch-Ross, 1997;Valentino, 2011).
Our primary question was: 1) what aspects of AEM functioning are associated with patterns of attachment security and/or insecurity in adults (from age 16), as identified in clinical and non-clinical studies?We chose to focus on age 16 onwards as it coincides with the beginning of young adulthood (Arnett, 2000), characterised by the development of self-concept (Sebastian, Burnett, & Blakemore, 2008) and the laying down of long-lasting self-defining memories (Conway, 2005).Most health problems also emerge for the first time in that period (Patel, Flisher, Hetrick, & McGorry, 2007), representing a pivotal time for treatment innovation.
Our secondary questions were: 2) which factors influence the associations between attachment patterns and AEM functioning; and 3) what are the clinical implications for psychological interventions harnessing AEM to improve outcomes in emotional disorders?

Search strategy
A review protocol was registered in PROSPERO (registration number: CRD42020216345) and the study selection process followed PRISMA guidelines (Moher, Liberati, Tetzlaff, & Altman, 2009).Searching databases were PsycINFO, MEDLINE and Embase.The search terms included the following concepts: "attachment" (search terms: "attachment" OR "internal working model") and "autobiographical memory" (search terms: "memory" OR "representation").No restrictions or limitations were imposed during the search (including no restrictions for publication dates).Both clinical and non-clinical studies were included.

Inclusion and exclusion criteria
Inclusion criteria were original empirical studies published in peerreviewed journals; using quantitative analyses; with participants aged 16 or above (to include young adults and onwards); and written in English.Studies were also included if attachment patterns and AEM functioning were assessed separately in different measures and within the same adult participants; attachment patterns were assessed directly in the participants (e.g., not in their parents); and attachment patterns were about relationships with people (e.g., not with objects/pets).Exclusion criteria were studies where participants' mean age was below 16, articles were not in English; and lacking measures of attachment and/or AEM.
Searches were performed on 17/01/2023.Titles and abstracts were used for initial screening followed by full-text screening by the lead author, with a second rater screening a subset of them (10%).Inter-rater reliabilities for this subset achieved good to excellent agreement: initial screening: Cohen's kappa =0.82; full-text screening: Cohen's kappa =1.00 (Altman, 1991).

Data synthesis
Study information was extracted for: sample size; basic demographics such as age, sex, ethnicity, and the country where the study was conducted; type of attachment measure; type of attachment pattern captured; type of AEM measure; study design (e.g., correlational vs longitudinal; sample type); and any mental health outcomes.A narrative synthesis was conducted to showcase the range of AEM features considered; candidate mediators and moderators of any attachment-AEM links; and possible impact on mental health outcomes.
A meta-analysis was not performed at this point in time for the following reasons: i) methodologies across studies (e.g., design of measures of attachment and AEM functioning) were highly heterogenousincluding a vast diversity of AEM paradigms employed (with over 20 different memory tasks as confirmed later in Results) -making meaningful combinations of studies challenging (Borenstein, Hedges, Higgins, & Rothstein, 2009); ii) most designs were correlational (rather than longitudinal, as also confirmed later in Results), an important risk of bias which would overall render effect estimates less meaningful and possibly misleading for prediction or causal inference; and importantly, iii) our key motivation was to scope the literature to uncover novel ideas for theory and clinical developments, as well as encourage improved study designs which in turn could inform robust and meaningful metaanalytic approaches in future.

Results
To organise our findings, we will begin with an overview of key study characteristics (Table 1).We will then describe the broad patterns of associations found between attachment insecurity and AEM phenomenology; candidate factors influencing (i.e., moderators) and explaining A. Lau-Zhu et al. such associations (i.e., mediators or mechanisms); potential relevance of these associations to mental health outcomes and to the effectiveness of AEM-based manipulations.

Measures of attachment patterns
Most studies (20 [61%]) included the Experience in Close Relationship self-report questionnaire (ECR; Brennan et al., 1998) and its variants (Fraley et al., 2000;Fraley, Heffernan, Vicary, & Brumbaugh, 2011;Wei, Russell, Mallinckrodt, & Vogel, 2007), which yielded dimensions on attachment avoidance and attachment anxiety.Other dimensional measures used included the Relationship Questionnaire (Griffin & Bartholomew, 1994) and the Older Adults Marital Attachment Scale (Wang, Wang, Feeney, & Li, 2016;Wang, Wang, Wang, & Feeney, 2018) the latter developed specifically in a Chinese sample.A minority of studies (7 [21%]) used self-report measures that yielded categories of secure and insecure attachment styles.These were the Relationship Questionnaire (Bartholomew & Horowitz, 1991) and the Attachment Scale (Hazan & Shaver, 1987).Two remaining studies used the Adult Attachment Interview (AAI; George, Kaplan, & Main, 1996), which is the goldstandard interview that yields categories of secure and insecure attachment patterns.While the self-report measures mostly focus on romantic/marital relationships, the AAI focuses on parents.The relative merits and drawbacks of different attachment measures have been comprehensively reviewed elsewhere (Ravitz, Maunder, Hunter, Sthankiya, & Lancee, 2010).

Measures of AEM functioning
A variety of memory activation methods were used.Well-established cuing methods include the retrieval of specific memories in relation to word cues in variants of the Autobiographical Memory Test (Williams, Nurs, Tyers, Rose, & MacLeod, 1996) or in response to four emotion words (happiness, sadness, anxiety and anger) in the Memory for Childhood Experiences Task (Mikulincer & Orbach, 1995).Pictorial cues were used in an adapted Autobiographical Memory Interview (Cao, Madore, Wang, & Schacter, 2018).
Two studies capitalised on ways to record the initial encoded event which then served as stimuli for a delayed recall test, using a diary method (Gentzler & Kerns, 2006) and by checking historical records (Edelstein et al., 2005).

What aspects of AEM functioning are associated with individual differences in attachment patterns?
Most studies examined AEM phenomenology (Table 1) and were synthesised below by i) memory intensity and arousal; ii) memory details, specificity, and vividness; iii) memory coherence and fragmentation; and iv) memory latency and accuracy.A minority of studies described memory content.Both attachment categories and dimensions were used to describe the key findings (Fig. 4).

Memory intensity and arousal
Attachment anxiety appeared to be associated with increased intensity while attachment avoidance with decreased intensity.In a seminal study by Mikulincer and Orbach (1995), anxiously-attached adults reported increased intensity of early childhood memories for both dominant emotions (e.g., sadness in a sad memory) and nondominant emotions (e.g., anger in a sad memory), whereas avoidantlyattached adults reported decreased intensity (specifically for sad and anxious memories) relative to securely-attached adults.
Consistent with Mikulincer and Orbach's early findings, subsequent studies have found that attachment anxiety was associated with increased intensity of negative memories (Sutin & Gillath, 2009; Study 1) and trauma memories (Ogle, Rubin, & Siegler, 2015), as well as increased reports of sense of "reliving" for early memories ( Öner & Gülgöz, 2022).Cortes and Wilson (2016;Studies 1-3) showed that individuals with high attachment anxiety perceived negative memories (of transgression) subjectively closer in time relative to positive memories (of kindness), independent of actual calendar time, which may partially explain their higher perceived intensity of negative memories at retrieval.Also consistent with Mikulincer and Orbach's work, attachment avoidance was associated with reduced intensity of caregiver memories (Haggerty, Siefert, & Weinberger, 2010) and non-valence specific memories (Sutin & Gillath, 2009; Study 2).However, a recent study by Dykas et al., (2014) using the Mikulincer and Orbach (1995)'s memory paradigm replicated the findings on memory intensity in anxiously-attached adults but not in avoidantly-attached adults.Note some key methodological differences in the study by Dykas and others include a younger sample and a different attachment measure.
Others have shown similar effects of attachment anxiety and avoidance on memory intensity.Gentzler and Kerns (2006) showed that both anxiously-and avoidantly-attached individuals underestimated the intensity of positive affect previously experienced for positive events.Both of these insecure patterns were also associated with comparable neurophysiological markers of increased emotional arousal (i.e., enhanced right-sided parietal activity as revealed by electroencephalography or EEG) during memory retrieval relative to a resting period (Kungl, Leyh, & Spangler, 2016).Unlike securely-attached adults, the "fading affect bias" was absent in insecurely-attached adults (Zengel et al., 2019), which refers to a tendency for affective intensity to fade less over time for positive compared to negative memories of current relationships.This suggests that insecure attachment overall is related to a reduction in benefits from positive affect generated by relationshipbased memories.However, Crawford and others (2021) failed to find an association between such bias and attachment patterns.Discrepancies between both studies may be due to the use of categorical (Zengel et al., 2019) versus dimensional analyses (Crawford et al., 2021) of attachment patterns.
Overall, memory intensity of specific memories has been the most researched index in this section.Of the eight studies described with such an outcome, four findings indicated that anxious attachment was Note.Three adults with different attachment patterns may differentially recall the same autobiographical event.The securely-attached adult on the left-hand side recalls a negative memory of an interpersonal conflict; the anxiouslyattached adult on the right-hand side recalls the same event but experiences it as more emotionally intense; the avoidantly-attached adult in the middle recalls the same event and experiences it as less emotionally intense, less detailed, less coherent, and less quickly accessible.associated with increased intensity and four indicated that avoidant attachment was associated with decreased intensity instead, with mostly small-to-medium effects across studies (Table 1).

Memory details, specificity and vividness
Attachment avoidance may reduce the level of details retrieved in memory, as prominent theory has argued that such a pattern reflects habitual attempts to keep aversive memories at bay (Williams et al., 2007).Consistent with this account, Sutin and Gillath, (2009; Studies 1-2) found that attachment avoidance was associated with less detailed memories for positive (Study 1), negative (Study 1), and non-valence specific memories (Study 2; indexed by lower self-reported agreement with statements such as "I can picture this memory in great detail in my mind").Similarly, avoidant attachment was associated with fewer unique details in a study using a more objective method of quantifying details (with research-led coding of memory elaboration; McCabe & Peterson, 2012).However, Wang et al. (2018) showed that avoidant attachment was associated instead with more memory details.It is possible that this avoidant strategy can backfire at times (Williams et al., 2007) resulting in more vivid memories akin to processes involved in the experience of intrusive memories in PTSD (Ehlers et al., 2004).
The above "avoidant strategy" account does not seem to explain the links between attachment anxiety and memory details.Sutin and Gillath (2009) found that attachment anxiety was also associated with less detailed positive (Study 1) and non-valence specific memories (Study 2).Similarly, Wang et al. (2018) found that anxious attachment was associated with fewer details of marriage-related memories.Such reduction in memory details may be better explained by excessive arousal associated with anxiety which impedes optimal cognitive performance (Yerkes & Dodson, 1908).Despite links with reduced details, anxiety has been found at times to be correlated with higher vividness of negative AEMs ( Öner & Gülgöz, 2016), although not with vividness of positive memories ( Öner & Gülgöz, 2016) or compassion-focused imagery (Naismith et al., 2018).High vividness may be of the small number of details retrieved.While details and vividness are both related to perceptual impressions of the memory content, they may represent distinct and dissociable aspects of memory that are related in a complex manner (Richter, Cooper, Bays, & Simons, 2016).
Patterns on past memories could extend to future imagined events.A study by Cao and others (2018) demonstrated that securely-attached adults generated more episodic details for both remembered and imagined attachment-relevant relative to attachment-irrelevant events.This attachment-relevance enhancement on recall was absent in both insecurely-attached individuals.
Finally, another memory feature related to detail is specificitythe extent to which a retrieved memory refers to a unique event rather than a broader event category (i.e., overgeneral memory).Using the wellestablished test of memory specificity, Beyderman and Young (2016) found no association between anxious/avoidant attachment and memory specificity in African-American adults, contrary to predictions from developmental psychopathology models linking insecure attachment with overgeneral memory (Valentino, 2011).Their clinical sample exhibited previous substance abuse which could have greatly impaired memory retrieval all together.
Overall, memory details has been the most researched index in this section.In the five studies described with such an outcome, four findings indicated that avoidant attachment was associated with decreased details and three indicated that anxious attachment was associated with decreased details too, with small-to-medium effects across studies (Table 1).

Memory coherence and fragmentation
Memory coherence refers to one's ability to re-construct past experiences through a connected and logical narrative (Vanderveren, Bijttebier, & Hermans, 2020).Attachment avoidance is thought to recruit a "defensive" strategy to deactivate the attachment system, contributing to less coherent memory descriptions (Mikulincer, Shaver, Cassidy, & Berant, 2009).Accordingly, attachment avoidance was associated with less coherent AEMs of both positive and negative memories of marriage in older Chinese adults (Wang et al., 2018) and of both first memories and adolescence memories in US-based young adults (McCabe & Peterson, 2012).Quinn, Spiby, and Slade (2015) failed to find links between attachment patterns and more fragmented (i.e., less coherent) memories for childbirth trauma in mothers.Differences in methodologies for assessing memory coherence (objective coding) and fragmentation (self-report) may partly explain such discrepancies.In sum, two out of the three studies described has shown that attachment avoidance was associated with less memory coherence with small effects (Table 1).

Memory latency and accuracy
Attachment insecurity may also impact memory performance.Slower retrieval of childhood memories has been found for avoidantlyattached relative to securely-attached adults (Dykas et al., 2014;Mikulincer & Orbach, 1995).Slower retrieval of negative memories has also been associated with greater levels of attachment avoidance measured dimensionally (Kohn et al., 2012).Conversely, faster retrieval of (anxious/sad) memories has been found for anxiously-attached adults (Mikulincer & Orbach, 1995).A more recent study found slower retrieval of AEMs for both avoidantly-and anxiously-attached adults (Luo et al., 2020; Study 1), but only avoidantly-attached adults showed a concurrent neurophysiological marker of reduced emotional processing, which presumably reflects attempts at reducing memory accessibility (i.e., a diminished late positive potentialas captured using EEGwhen comparing negative relative to neutral memories).The absence of such biological correlates in the anxious attachment group suggests that a different mechanism is at play.
Memory accuracy is often challenging to study for autobiographical memories as researchers tend to have little control over encoding of the original event.However, a study by Edelstein and others (2005) took advantage of a community-based cohort study with access to confirmed records from 14 years earlier during childhood alongside a follow-up during adulthood.Attachment avoidance (but not attachment anxiety) was associated with lower accuracy for facts about childhood sexual abuse (e.g., frequency and extent of the abuse) in those with high levels of abuse severity.
Overall, retrieval latency has been the most researched index in this section.In the three studies described with such an outcome, all of them indicated that avoidant attachment was associated with slower retrieval, with small-to-medium effects across studies (Table 1).

Memory content
A minority of studies examined the types of events retrieved.For instance, associations were found between dismissing/preoccupied attachment and increased number of memories about relationships and family (Elnick et al., 1999); fearful/profound-trust attachment and selfdefining, life-threatening memories (Goldner & Scharf, 2017); and avoidant attachment and more negative memories involving caretakers (Haggerty et al., 2010).

Which factors are likely to influence the associations between attachment patterns and AEM functioning?
The distinct profiles of AEM phenomenology by attachment patterns (Fig. 4) seem to also depend on additional moderating factors, as explored in seven studies.One moderator considered relates to the characteristics of the event.Attachment patterns have been linked with recall of affect experienced only for events involving positive, interpersonal, daily events experienced as adults (Gentzler & Kerns, 2006) or with recall of memory intensity involving only negative (anxious/sad) events in childhood (Mikulincer & Orbach, 1995).Furthermore, attachment avoidance was related to decreased memory intensity when recalling negative memories only involving caretakers but not memories involving non-attachment figures (Haggerty et al., 2010).It is possible that attachment-related biases in memory are primarily heightened when recalling personally-relevant events.
Personal characteristics could also play an important role.A study in older Chinese adults (Wang et al., 2016) suggested a stronger effect of attachment avoidance on AEM functioning (e.g., detail) in older than younger adults relative to the effect of attachment anxiety, because the latter decreases with age (Cusimano & Riggs, 2013), although another study found that older age does not always influence attachment effects (Cao et al., 2018).Gender may shape the content of the memories retrieved: men (but not women) with lower attachment avoidance retrieved fewer memories of relationship-maintaining events (e.g., anniversaries and reunions) whereas women (but not men) with higher attachment anxiety retrieved more between-couple events (Wang et al., 2016).Such gender effects were derived from a study in Chinese participants thus may also reflect cultural differences in AEM functioning.Individuals also vary in how they use mental imagerya key component of AEM functioning (Conway & Pleydell-Pearce, 2000).Preliminary evidence indicates that those with generally weaker imagery ability (of imagining everyday scenarios) may benefit less from generating a compassion-focused image (Naismith et al., 2018).
Overall, these findings tentatively suggest that processes related to demographics (e.g., age and gender), psychological traits (e.g., imagery abilities) and nature of the event (e.g., content and emotionality) could amplify and/or weaken attachment-related biases in memory processing.

What potential mechanisms underpin the associations between attachment patterns and AEM functioning?
Three studies interrogated potential processes mediating attachment and AEM.Using structural equation modelling to test hypothesised links between different memory stages, Öner and Gülgöz (2016) found that attachment avoidance may involve dampening down of the personal relevance of intimacy-related events in romantic relationships at the time of the initial experience (i.e., encoding), which can have knock-on effects on all subsequent memory stages, resulting in reduced rehearsal of such events (i.e., consolidation) and subsequently less intense and vivid memories (i.e., retrieval).Evidence on the role of memory rehearsal/consolidation comes from two additional studies.Attachment avoidance appears to be associated with reduced experiences of event elaboration from parents during childhood which could result in subsequent reduced recollection of early memories during adulthood ( Öner & Gülgöz, 2022).Similarly, Edelstein and others (2005) suggested that attachment avoidance involves talking to others to a lesser extent about previous incidents of childhood sexual abuse (possibly affecting rehearsal) which in turn can contribute to reduced memory accuracy.These ideas are consistent with the hypothesis put forward by Mikulincer and Orbach (1995), whereby attachment avoidance reduces relational memory accessibility by influencing emotional regulation processes, specifically recruiting "repressive defensiveness' -the ability to selectively suppress painful attachment-related information (Dykas & Cassidy, 2011).
Öner and Gülgöz (2016) also found that attachment anxiety may increase rehearsal (but not increase encoding) of negative memories, further enhancing retrieval of those memories.This finding is in line with the notion of schema-consistent processing whereby informationprocessing is skewed by IWMs in a manner that is self-reinforcing (Dykas & Cassidy, 2011), such that anxiously-attached individuals may see others in a negative-biased manner (e.g., unreliable or unpredictable) and retrieved negative memories consistent with such views.

Are attachment patterns linked to mental health outcomes through their effects on AEM functioning?
Only five studies considered mental health outcomes.Sutin and Gillath (2009) directly tested the hypothesis that attachment patterns influence anxiety/depression through AEM functioning.They found that reduced memory details (Studies 1 & 2) and memory intensity (Study 2) partially mediated the association between attachment avoidance and more depressive symptoms (Study 1).Increased frequency of negative memories (Study 2) mediated the association between attachment anxiety and more depressive symptoms.Unlike depressive symptoms, attachment patterns did not appear to be associated with anxiety symptoms via AEMs.Beyderman and Young (2016) found an association between attachment avoidance and depressive symptoms; but overgeneral memory neither mediated such association nor was associated with attachment patterns.
Regarding PTSD, Ogle et al. (2015) found that increased physical reactions, voluntary rehearsal and involuntary recall (but not emotional intensity) mediated the link between attachment anxiety and more PTSD symptoms.Importantly, involuntary recall appeared to play a unique role in the development of psychopathology as it is the only feature of trauma memory that mediated the relationship between attachment anxiety and more PTSD symptoms in individuals with experiences of childhood traumas.Another study set out to explore the links between attachment patterns and PTSD (and the possible mediating role of memory fragmentation) but found no associations between attachment and fragmentation (Quinn et al., 2015).
Two studies included measures of processes that are thought to play a role in psychopathology (although mental health was not directly assessed).Attachment anxiety was linked to increased perception of a transgression memory (involving a romantic partner) as having happened closer in time which fuelled "kitchen thinking" -the tendency to bring up past relational memories even in a new and unrelated context (Cortes & Wilson, 2016).This thinking pattern has conceptual overlaps with rumination thinking in depression (Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008).Anxiously-attached adults, unlike their securely-attached or avoidantly-attached counterparts, failed to report positive mood change after positive (vs.neutral) AEM retrieval and were worse at problem solving (Mikulincer & Sheffi, 2000), another common challenge in depression (Williams et al., 2007).
Overall, an emerging picture indicates that attachment-related biases in AEM functioning exist, and these biases could also have a knockon effect on mental health and associated risk processes, but the evidence base is too small to draw any firm conclusions.

Do attachment patterns influence the effects of AEM-based manipulations?
Several studies have suggested that the effects of AEM-based interventions could depend on attachment patterns.Anxiously-attached (relative to securely-and avoidantly-attached) adults failed to experience changes in positive affect following positive memory retrieval (Mikulincer & Sheffi, 2000; Study 1).Likewise, the use of nostalgic memory was associated with more recovery from sadness and more increases in happiness in securely-attached relative to insecurelyattached adults (Cavanagh et al., 2015).Possibly, insecurely-attached adults spontaneously interpret those memories in (negative) ways that impede their beneficial effects.
Access to memories of negative experiences with attachment figures was reduced for individuals with higher attachment avoidance, yet this effect was "corrected" by asking the same individuals to perform a concurrent demanding task (i.e., a controlled writing task) intended to disrupt self-regulatory processes (Kohn et al., 2012).This represents a possible strategy for avoidantly-attached adults to access aversive memories for further processing.
Some interventions have the potential to cause harm.A distinction is typically made between imagery through a first-person perspective (i.e., seeing the memory through one's own eyes) versus through a thirdperson perspective (i.e., seeing onself as outsider in the memory), with the latter thought to facilitate adaptive emotional distancing (Kross & Ayduk, 2017).However, when visualising a relationship-based memory (of a transgression involving the partner) from a third-person (vs.firstperson) perspective, individuals with high attachment anxiety made less positive evaluations of their relationship, whereas individuals with low attachment anxiety made more positive evaluations and also reported less distress (Marigold et al., 2014; Studies 1-2).Third-person imagery may more readily activate and amplify pre-existing conceptual beliefs (e.g., of the self and others) which are likely to be negative in insecurely-attached adults (Libby & Eibach, 2011).A study on spouses (of military members who were deployed) also found that retrieving relational memories resulted in lower negative emotions in individuals with low attachment avoidance but higher negative emotions in those with high attachment avoidance (Borelli et al., 2014).Overall, some ways of recalling attachment memories may not always bring positive consequences in insecurely-attached individuals.
Attachment patterns did not influence the benefit of interventions recruiting positive attachment memories, including interventions using compassion-focused imagery (Naismith et al., 2018; Studies 1-2) or security priming where an attachment figure is visualised to offer comfort and increase "felt security" (Sutin & Gillath, 2009;Study 2).
In sum, an individual's pre-existing attachment pattern may be relevant to revealing the effects of some manipulations that capitalise on AEMs.The application of AEM-based techniques likely requires a thoughtful approach based on a sound understanding of the underlying mechanisms, ideally involving a nuanced match between attachment patterns and memory procedure to maximise the clinical effects while avoiding unintended harms.

Discussion
We reviewed the literature on attachment patterns and AEM in adults (from aged 16) to explore its relevance for clinical psychological science and practice.Our eligible studies (Table 1) spanned across psychology subdisciplines (e.g., social, clinical, developmental, educational, cognitive, psychoanalytic, etc.), underscoring the topic's broad relevance.Work directly examining mental health outcomes in this area remains scarce, despite the well-established links between AEM and psychopathology transdiagnostically (Dalgleish & Brewin, 2007;Hitchcock, Werner-Seidler, Blackwell, & Dalgleish, 2017;Williams et al., 2007).Below we present an overview of key findings, and consider methodological, theoretical, and clinical issues to bridge the sciences of attachment, memory, and psychopathology for treatment innovation.

Overview of findings
Compared to securely-attached adults, avoidantly-attached adults appeared to demonstrate a relatively stable profile of AEM biases with decreased accessibility to attachment-relevant information across a range AEM properties.Recalling (negative) attachment memories in avoidantly-attached adults can be less intense, detailed, coherent and slower (Fig. 4).For anxiously-attached individuals, evidence indicates that recalling comparable memories can be more intense (Fig. 4) although possible also less detailed.
A complex set of mediating and moderating factors seems to characterise the link between attachment and AEM.At least some of the attachment-related memory biases can be explained (mediated) by emotional regulation and self-relevant processing, as well as be influenced (moderated) by processes related to both event features (e.g., valence and interpersonal nature) and personal characteristics (e.g., age and trait imagery).Tentatively, attachment patterns and mental health outcomes/processes are linked through key AEM properties (e.g., intensity), with emerging evidence for depression and PTSD.Finally, the benefits of (at least some) AEM-based therapeutic techniques may depend on attachment patterns.

Methodological considerations
The vast number of studies using correlational designs (see Table A.1 in the Appendix for details on quality appraisal) highlights the need for further longitudinal and appropriately-powered studies (as most effects described were in the small-to-medium range) with repeated assessment timepoints to provide more robust tests of causality from attachment/ IWMs to AEM functioning, and in turn to mental health.A truly developmentally-sensitive approach would track younger to older adulthood, include genetically-informative design features to tease apart the relative contributions of genes and environment (Fearon, Shmueli-Goetz, Viding, Fonagy, & Plomin, 2014), explore a global and diverse perspective (Bauer, 2019), and embrace robustness and transparency (Munafò et al., 2017).
Different measurements of attachment may reflect different aspects of attachment, which in turn may explain some of the inconsistent findings.The AAI is an interview that assesses one's "state of mind" with regard to attachment with early caregivers followed with subsequent standardised coding yielding discrete classifications (George et al., 1996), whereas self-reported measures such as the ECR (Brennan et al., 1998) typically assess attachment in the context of general adult Note.Internal working models (IWM) represent hypothetical cognitive structures underpinning individual differences in attachment patterns (e.g., secure, anxious and avoidant attachment) in adulthood; IWM are derived from semantic and episodic aspects of autobiographical memories (AMs); IWM, through two key informationprocessing routes (emotion regulation and schema-based processing), also influence AM functioning including its phenomenological qualities; AMs in turn shape cognitivebehavioural processes involved in the maintenance of mental health and emotional psychopathology.
relationships yielding continuous measures.It also remains to be explored whether attachment-related biases in memory are influenced by the attachment contexts and relationship foci.
AEM phenomenology comprises a range of features but most studies focused exclusively on single features, thus we lack an understanding of their interrelationships.Crucially, despite the centrality of involuntary retrieval in AEM functioning (Berntsen, 2009) and psychopathology (Ogle et al., 2015), only a few studies considered intrusive images/ memories.

Theoretical implications
Notwithstanding the key limitations of the literature as described above, we have drawn up a schematic model bringing together attachment, memory and psychopathology (Fig. 5).Focusing on the first link from attachment to memory, the reviewed findings broadly accord with Dykas and Cassidy (2011)'s dual process model.In the majority of studies, it appeared that avoidantly-attached individuals display consistent signs of underactivating their attachment memories, as indexed across multiple memory outcomes.These include a reduction in memory intensity (Haggerty et al., 2010;Sutin & Gillath, 2009), details (e.g., McCabe & Peterson, 2012;Sutin & Gillath, 2009), coherence (e.g., McCabe & Peterson, 2012;Wang et al., 2018) and retrieval speed (e.g., Dykas et al., 2014;Kohn et al., 2012).This memory profile may be the consequence of downregulation to avoid emotional pain as initially proposed by Mikulincer and Orbach (1995), although this emotional "blocking" may also limit the benefits from such memories (e.g., from AEM-based interventions).One study used an experimental manipulation that purportedly disrupted self-regulatory processes (through a "controlled" writing task; Kohn et al., 2012) and found that this "normalised" retrieval speed to negative memories in avoidantly-attached individuals, supporting the role of emotional regulation in initially reducing memory accessibility.Nevertheless, despite long-standing proposals that emotion regulation plays a key role in mediating, no studies reviewed have included an independent measure assessing this construct.
If attachment-related memories are accessed and processed, then these are likely to be consistent with pre-existing self-images (Conway, 2005;Conway & Pleydell-Pearce, 2000).Accordingly, anxiouslyattached individuals showed more sensitivity to negative attachment memories (e.g., with negative self-views), as reflected in experiencing more intensity or sense of "reliving" while recalling such memories (Cortes & Wilson, 2016;Ogle et al., 2015;Öner & Gülgöz, 2022;Sutin & Gillath, 2009).Again, despite long-standing proposals that schemarelevant processing also plays a key mediating role between attachment and cognition (Dykas & Cassidy, 2011), none of the studies reviewed had directly studied it either through experimental manipulation or including independent measurement.
Turning to the second link linking memory and mental health, attachment-related memory biases likely impact on established aspects of cognition (e.g., thinking styles) and behaviour (e.g., problem solving) highlighted in cognitive-behavioural models of emotional disorder maintenance (Harvey, Watkins, Mansell, & Shafran, 2004), while other levels of analyses (e.g., biology and sociality) remain underinvestigated.Only a few studies from the reviewed literature have focused on this second link as it pertains to attachment patterns (Ogle et al., 2015;Quinn et al., 2015;Sutin & Gillath, 2009) and therefore it remains inconclusive.Given the paucity of research in this area, there is also a lack of mechanistic investigations underlying the link from attachment-based memory biases to psychopathology.Future research could consider recent frameworks that have proposed key mechanisms impacted by attachmentnegative expectancies, interpretation biases and defensive strategies (Kobak & Bosmans, 2019) which are likely to dynamically influence intra-and interpersonal processes relevant to mental health outcomes.
While some memory properties (e.g., specificity and fragmentation) have well-known theoretical and/or empirical links to psychopathology, causality between other properties (e.g., intensity and latency/accuracy) and mental health remains to be established.There are likely to be multiple mechanisms depending on the specific aspect of AEM in consideration.

Clinical implications
The cognitive literature reveals that at least some of the AEM properties are amenable to modification (Hitchcock et al., 2017).To address memory intensity/arousal, well-established and effective CBT techniques are already available such as imaginal exposure (Foa, Hembree, & Rothbaum, 2007), cognitive restructuring within reliving (Grey, Young, & Holmes, 2002) and imagery rescripting (Arntz, 2012).The latter is a recognised approach for addressing salient memories of childhood traumaoften involving attachment figure(s) (Arntz & Weertman, 1999) and holds promise as brief, stand-alone approaches to address a range of emotional disorders (Arntz, 2012).Experimental psychology has also provided innovative early-stage techniques to modify memory features.One line of development harnesses repeated retrieval practices to enhance memory specificity/detail, including Memory Specificity Training (Barry, Sze, & Raes, 2019), Episodic Specificity Induction (Jing, Madore, & Schacter, 2016), Memory Flexibility Training (Moradi et al., 2014) and Positive Memory Training (Steel et al., 2020).Another emerging approach is the use of (visuospatial) competing task techniques to reduce memory vividness (Engelhard, van Uijen, & van den Hout, 2010;Rackham & Lau-Zhu, 2021) and intrusive recall (Lau-Zhu, Henson, & Holmes, 2019, 2021).Experimental approaches to change memory fragmentation or latency/ accuracy are currently underexplored.
AEM's malleability opens exciting therapeutic opportunities to improve mental health in insecurely-attached individuals.Various cognitive-behavioural models increasingly incorporate attachment ideas more explicitly (Gilbert, 2014;Kellogg & Young, 2006;Maccallum & Bryant, 2013), but have yet to fully harness the full range of possibilities for modifying AEMs.The effectiveness of current psychological treatments for them may be enhanced through an augmented focus on AEM.For example, current CBT protocols could be enriched with AEM-based techniques, as done for social anxiety (Wild, Hackmann, & Clark, 2008) or bipolar disorder (Steel et al., 2020), albeit here also matched to patients' attachment patterns.For instance, conventional CBT techniques (e.g., changing core beliefs) may be aided by increasing memory specificity to facilitate access to a broader range of evidence or by decreasing intensity of distressing memories to support emotional regulation.Novel intervention protocols could also consider targeting multiple AEM foci simultaneously as currently explored for both adults (Holmes et al., 2016;Steel, Wright, et al., 2020) and youth (Lau-Zhu, Farrington, & Bissessar, 2022;Pile et al., 2020).Critically, the development of new AEM-based techniques (e.g., varying imagery perspective in relational memories) shall consider whether their effectivenessincluding the possibility of harmdepends on pre-existing attachment patterns.
AEM-based approaches hold relevance to help prevent mental health problems in populations where attachment insecurity is prevalent, for example in the context of childhood maltreatment (Boroujerdi, Kimiaee, Yazdi, & Safa, 2019) and autism spectrum (Gallitto & Leth-Steensen, 2015), and thus where lifelong psychopathology risk is high.Prevention work can embrace digital mediums (as explored in many experimental psychology techniques) for wider and global reach (Holmes et al., 2018).Given that not every insecurely-attached individual would need an intervention, work on identifying those at "most" risk is warranted.For this we need to weave in data about possible moderators and mediators (Fig. 5), which remain a scientific gap.
Finally, attachment patterns can both persist and change across the lifespan (Fraley, 2019).A tantalizing yet underexplored possibility is whether modifying attachment-related memories can create long-lasting changes in attachment pattern.While attachment styles are thought to stabilise in adulthood, recent neuroscientific accounts posit that memories may be rendered labile again under certain boundary conditions (Visser, Lau-Zhu, Henson, & Holmes, 2018).Accordingly, adults' attachment appear to change with psychological treatments even within CBTwhere the theoretical focus is not explicitly on changing attachment (Taylor, Rietzschel, Danquah, & Berry, 2015).The success of changing attachment styles may depend on the varying degree to which the autobiographical memory system is (intentionally or incidentally) engaged across treatments.Attending to AEM-based images specifically underpinning one's attachment pattern (e.g., with a caregivers or partners) could help modify one's attachment more generally.This cognitive malleability is also likely to be potentiated during adolescence and young adulthood (Lau & Waters, 2017), which we can leverage as an optimal time window for changing attachment insecurity.
Within the framework of cognitive therapy, an AEM-based approach with attachment memories also has clinical potential to impact on core beliefs (about the self, others and the future) which are notoriously difficult to change (James & Barton, 2004).Attachment and core beliefs are proposed to overlap with autobiographical memory serving as the critical bridge (e.g., Platts, Tyson, & Mason, 2002).However, for core belief work, memory-and imagery-based techniques remain underutilised (Stopa, 2009) and their benefits on symptom relief underexamined (Çili & Stopa, 2015).Tuning into attachment-related memory biases could more readily evoke encapsulated core beliefs and render them more amenable to modification.

Conclusion
Unlike the more common approach of adapting successful therapeutic models in adults to developmental populations (Benjamin et al., 2011), developmental science could improve cognitive approaches in adults too, here with the rapidly expanding work connecting attachment, memories, and emotions.Attachment insecurity has been a longestablished transdiagnostic risk factor for later psychopathology but the underlying mechanisms have remained elusive.The focus on a key cognitive, causal, and modifiable factorautobiographical memoryopens the door for exciting future opportunities (Table 2).Mechanistically-informed interventions (Barlow et al., 2013;Holmes et al., 2018) harnessing psychological sciences hold promise for reducing mental health burden in the area of attachment insecurity across a range of clinical populations.

Role of funding sources
ALZ was supported by Oxford Health NHS Foundation Trust for doctoral training in clinical psychology.

Contributors
All authors designed the study and wrote the protocol.ALZ conducted literature searches, data extraction, and wrote the first draft of the manuscript.All authors contributed and have approved the final manuscript.

Declaration of Competing Interest
We declare no conflicts of interests.

Table 2
Outstanding Questions for Clinical Psychological Science Bridging Attachment, Autobiographical Memories and Mental Health.

Question Box
• Which autobiographical memory features lie in the causal pathway between attachment patterns and emotional disorders (and for which specifically)? • What are the overlapping and distinct autobiographical memory profiles between different forms of insecure attachment (e.g., avoidant vs. anxious)?• How could we best assess attachment-related biases in autobiographical memory in clinical settings to inform treatment selection?• Would treatment effectiveness for insecurely-attached individualsacross psychiatric diagnosesimprove with an added/renewed focus on autobiographical memory?• How do we best proceed with prevention work for insecurely-attached individuals that takes into account latent vulnerabilities in autobiographical memory?• Can we harness the malleability of autobiographical memories to create longlasting changes in internal working models (and thus attachment)?
A. Lau-Zhu et al.

Fig. 2 .
Fig. 2. Early Memories of Child-Parent Interactions Reported in the Adult Attachment Interview for Three Attachment Patterns.Note.Extracted from Main et al. (2008).

A
.Lau-Zhu et al.

Fig. 4 .
Fig. 4. Hypothetical Depiction of the Phenomenology of Autobiographical Episodic Memories by Patterns of Attachment.Note.Three adults with different attachment patterns may differentially recall the same autobiographical event.The securely-attached adult on the left-hand side recalls a negative memory of an interpersonal conflict; the anxiouslyattached adult on the right-hand side recalls the same event but experiences it as more emotionally intense; the avoidantly-attached adult in the middle recalls the same event and experiences it as less emotionally intense, less detailed, less coherent, and less quickly accessible.

Fig. 5 .
Fig. 5.A Schematic Model of MechanisticProcesses Connecting Attachment, Autobiographical Memory, and Mental Health.Note.Internal working models (IWM) represent hypothetical cognitive structures underpinning individual differences in attachment patterns (e.g., secure, anxious and avoidant attachment) in adulthood; IWM are derived from semantic and episodic aspects of autobiographical memories (AMs); IWM, through two key informationprocessing routes (emotion regulation and schema-based processing), also influence AM functioning including its phenomenological qualities; AMs in turn shape cognitivebehavioural processes involved in the maintenance of mental health and emotional psychopathology.

Table 1
Key Design Characteristics of Included Studies.