Commentary

In their target articles, Jabès and Nelson (2015) propose a novel model outlining how distinct hippocampal regions may selectively support the gradual emergence of discrete explicit memory functions. The primary strength of this model lies in its recognition that the hippocampus is comprised of multiple functionally heterogeneous regions (or circuits). These regions, despite being highly interconnected, have long been believed to support different types of explicit memory functions (Deuker, Doeller, Fell, & Axmacher, 2014), and rather critically for Jabès and Nelson’s model, they demonstrate a differential rate of postnatal maturation (for review see Lavenex & Banta Lavenex, 2013).

More specifically, three distinct hippocampal circuits have been identified that show a differential rate of postnatal maturation. The first, which appears to be relatively mature at birth, contains the highly interconnected subcortical structures (subiculum, presubiculum, parasubiculum and CA2; Jabès, Lavenex, Amaral, & Lavenex, 2011). The second circuit demonstrates a more protracted postnatal development (˜0–2 years) and is comprised of the distinct layers in several hippocampal regions that receive direct projections from the entorhinal cortex (such as CA1, CA2, and subiculum; Jabès et al., 2011). Most striking, however, is the third circuit (containing the dentate gyrus (DG) and specific layers located downstream of the DG, particularly in the CA3 region), which appears to continue to develop throughout the first decade of human life (Lavenex & Banta Lavenex, 2013).

Hence, key to this model is the identification of a credible temporal sequence of emergent memory functions that map onto the above neuroanatomical data: a model that is both plausible (relative to the aforementioned time scales), and in keeping with each region’s previously hypothesized mnemonic function. In order to achieve this, Jabès and Nelson (2015) proposed a four-stage model in which the earliest developing hippocampal region (the subiculum) is proposed to support recognition memory (evident from birth or shortly thereafter). Next, the slower developing CA1 region (0–24 months; Fortman, Hewett, & Bennett, 2001) is hypothesized to support the later emergence (between 9 and 25 months) of a basic relational memory system in which the relations between items are considered to be encoded in a unitized and inflexible manner. With increasing age this system gives way to a more complex “flexible” relational memory system that no longer needs to reduce associated elements into singular, unitized representations. It is proposed that this complexity is afforded by the subsequent maturation of DG and CA3 regions. However, Jabès and Nelson argue that it is not until the DG has reached structural maturity (˜4 years), and the rate of neurogenesis in this region has stabilized, that the DG is optimally tuned to support enduring episodic memory abilities (i.e. the recollection of personal life events).

Importantly, this model is broadly in keeping with theoretical models of the function of the hippocampal subfields (e.g. Rolls, 2010) and the available animal data. However, it is noteworthy that our current understanding of the functional role played by individual hippocampal subfields has almost exclusively been attained through theoretical and computational models and from electrophysiological recordings in rodent hippocampal subfields but not from data acquired in human participants (Deuker et al., 2014). This is primarily due to the fact that examining the hippocampal subfields in vivo in human participants has proved to be extremely challenging. Although concerted efforts are being made to overcome these challenges (e.g. Bonnici et al., 2012; Carr, Rissman, & Wagner, 2010), the findings from the few studies that are currently available often show inconsistencies with the aforementioned computational and electrophysiological data (Deuker et al., 2014). It is therefore important to consider that our understanding of the functional role of the human hippocampal subfields may not be sufficiently advanced to enable us to draw confident links between their ontogeny and the emergent mnemonic capabilities of young infants. Similarly, there is a dearth of studies investigating the structural development of the individual hippocampal subfields in healthy human infants and children (see Lee, Ekstrom, & Ghetti, 2014, for a recent exception). Such data is critical if one wishes to confidently infer within-species links between structure and function in this manner.

In addition, although it is well established that human memory appears to undergo significant changes across infancy and childhood, the exact nature of these changes is far from clear (Mullally & Maguire, 2014). For instance, traditional neurocognitive accounts of memory development have assumed that relational memory (which is often viewed as the quintessential hippocampal-dependent memory function; Konkel & Cohen, 2009) emerges late in the first year of life. This sentiment is echoed in the above model, with a basic relational memory system incrementally emerging between 9 and 25 months of age and complimented by a later emerging “flexible” relational memory system. However, there is convincing evidence available to suggest that infants as young as 3 months of age are capable of forming enduring relational memories (e.g. Campanella & Rovee-Collier, 2005), and that 6-month-old infants are able to form associations between items that have never previously been encountered together (Cuevas, Rovée-Collier, & Learmonth, 2006). Whilst the findings of the former study suggest that basic relational memory processes may in fact be evident from as early as 3 months of age, the latter finding suggests that the type of flexibility assumed by the above model to emerge much later in life (>2 years of age) may in fact be in situ by 6 months of age. Such findings potentially skew the mnemonic times scales presented in the proposed model and introduce a possible discordance between the cognitive and neuroanatomical developmental trajectories described above.

Hence, for any such model to establish credibility, direct evidence linking structure and function in human infants/children is essential. In a recent review article, Mullally and Maguire (2014) challenged the field to actively intensify efforts to acquire such evidence. This challenge was two-fold. Firstly, it emphasized the need to acquire high-resolution structural magnetic resonance (MR) images to enable the examination of hippocampal subfield maturation throughout infancy and early childhood. Significantly, Lee et al. (2014) have since validated a hippocampal subfield segmentation method, previously used in adult participants (e.g. Ekstrom et al., 2009), in children aged between 8 and 14 years. This successful adaptation of high-resolution MR imaging methods to a paediatric population represents an important first step in attaining this goal. Secondly, despite the challenges that it imposes, Mullally and Maguire (2014) argued that for convincing links between function and structure to be made, functional neuroimaging studies (including functional magnetic resonance imaging (fMRI) studies) need to be more commonplace in the infant memory literature. Jabès and Nelson’s above model seems to intensify the need to acquire such evidence, which would enable the tentative links between function and structure forged here to be empirically tested. In addition, this more direct approach could fundamentally change the dynamics of the relationship between the developmental and adult memory literatures. Rather than the infant memory literature being the passive recipient of discoveries made in the adult and animal memory literatures, discoveries made in human developmental neuroimaging studies could instead take centre stage in the episodic memory debate – potentially driving forward our conceptual and neurobiological understanding of the developed episodic memory system.