Human lesion studies offer a powerful means of validating and refining neurocognitive models of memory (Irish & van Kesteren Reference Irish and van Kesteren2018). In their integrative memory model, Bastin et al. provide a compelling overview of the processes of recollection and familiarity, invoking evidence from Alzheimer's disease (AD) in support of its central tenets. Although this new framework can accommodate many of the memory and pathological changes in AD, we note significant gaps that warrant consideration.

We propose to extend the integrative memory model by considering complementary human lesion findings from the syndrome of semantic dementia; a neurodegenerative disorder characterized by progressive atrophy to core nodes of the entity representation (i.e., anterior temporal lobes and perirhinal cortex) and relational representation (i.e., hippocampus) systems (Brambati et al. Reference Brambati, Rankin, Narvid, Seeley, Dean, Rosen, Miller, Ashburner and Gorno-Tempini2009). The cognitive profile of semantic dementia is commonly taken to represent the “other side of the coin” to AD, with profound semantic processing impairments emerging in the context of relatively spared episodic memory (Hodges & Patterson Reference Hodges and Patterson2007). Importantly, the neurocognitive sequelae of semantic dementia pose several challenges to the integrative memory model, which we address here.

First, the integrative memory model fails to consider how the inherent features of the to-be-remembered stimulus potentially influence processes underlying familiarity. The key role of the entity representation system in familiarity is primarily supported by evidence from nonverbal measures, such as object recognition tasks. Studies of recognition memory in semantic dementia, however, reveal the importance of stimulus modality in familiarity judgments. For example, when pictures of objects are used as stimuli, semantic dementia patients demonstrate intact recognition (Simons et al. Reference Simons, Graham and Hodges2002a), despite severely impaired recognition for words (Graham et al. Reference Graham, Patterson, Powis, Drake and Hodges2002). As such, the anterior temporal lobe degeneration in semantic dementia does not manifest in an all-encompassing recognition deficit, but rather produces an impairment specific to verbal stimuli. As suggested by the integrative memory model, perceptual fluency signals, likely mediated by intact occipitotemporal regions, are co-opted to support recognition of pictorial stimuli in semantic dementia, even when the conceptual representation of that stimulus is degraded (Simons et al. Reference Simons, Verfaellie, Galton, Miller, Hodges and Graham2002b). For verbal stimuli, however, no such compensatory strategy can be deployed, as the poor perceptual discriminability of written words precludes the use of perceptual fluency signals (Graham et al. Reference Graham, Patterson, Powis, Drake and Hodges2002). Collectively, these findings from semantic dementia suggest that the relative weightings of perceptual versus conceptual fluency signals during familiarity judgments vary, contingent upon the nature of the to-be-remembered stimulus.

Building on this argument, the role of the anterior temporal lobe in supporting familiarity judgments may also scale, depending on the “meaningfulness” or conceptual loading of pictorial stimuli. In line with the integrative memory model, damage to the perirhinal cortex in semantic dementia has been shown to impair discrimination of objects with highly ambiguous perceptual features (Barense et al. Reference Barense, Rogers, Bussey, Saksida and Graham2010). Importantly, however, in semantic dementia these deficits are amplified for items embedded within a distinct semantic framework (e.g., animals) as compared with non-meaningful items (e.g., blobs). This disproportionate impairment of semantically loaded items likely manifests because of the co-occurrence of both perirhinal and temporopolar atrophy in semantic dementia. By contrast, amnesic patients with exclusive medial temporal (including perirhinal) damage, and spared temporal poles, are more accurate at discriminating between perceptually similar items that are meaningful, compared with novel ones, potentially by harnessing intact semantic constructs to boost recognition (see Barense et al. Reference Barense, Rogers, Bussey, Saksida and Graham2010). Accordingly, we propose an extension to the integrative memory model, in which the role of the anterior temporal lobes in familiarity judgments is particularly pertinent for stimuli that are both perceptually ambiguous and conceptually meaningful (see also Chiou & Lambon Ralph Reference Chiou and Lambon Ralph2016).

Contemporary models of memory recognize the importance of a distributed core recollection network including medial temporal, frontal, and parietal regions in mediating successful episodic retrieval (Rugg & Vilberg Reference Rugg and Vilberg2013). While the hippocampus is an important node of this network, findings from semantic dementia suggest that models of episodic memory must look beyond the medial temporal lobes (Irish et al. Reference Irish, Bunk, Tu, Kamminga, Hodges, Hornberger and Piguet2016). Despite pronounced hippocampal atrophy from early in the disease course (Chapleau et al. Reference Chapleau, Aldebert, Montembeault and Brambati2016), semantic dementia patients demonstrate remarkably intact nonverbal memory retrieval. This profile of sparing reflects the preservation of frontal and parietal brain structures (Irish et al. Reference Irish, Bunk, Tu, Kamminga, Hodges, Hornberger and Piguet2016), including the angular gyrus, posterior cingulate cortex, and lateral prefrontal cortex. Moreover, when recollection is affected in semantic dementia, it relates primarily to prefrontal, rather than hippocampal, degeneration (Simons et al. Reference Simons, Verfaellie, Galton, Miller, Hodges and Graham2002b). These findings emphasize the multifaceted nature of recollection, and its dependency upon lateral prefrontal and parietal brain regions. Further iterations of the integrative memory model should consider how distinct aspects of recollection, such as strategic retrieval (Rugg et al. Reference Rugg, Fletcher, Chua and Dolan1999) and contextual binding (Ramanan et al. Reference Ramanan, Piguet and Irish2018), are differentially underwritten by subdivisions of lateral prefrontal and parietal cortices.

Finally, we believe it is important to validate the integrative memory model with respect to ecologically valid expressions of memory, which are essential for self-continuity and identity (Strikwerda-Brown et al. Reference Strikwerda-Brown, Grilli, Andrews-Hanna and Irish2019). The integrative memory model is predicated largely on evidence from highly decontextualized experimental measures (e.g., object recognition, source memory). Although useful, such laboratory tasks fail to capture the idiosyncrasies of self-referential episodic recall as experienced in the real world. Autobiographical memory represents the prototypical expression of contextually detailed episodic retrieval, imbued with vivid sensory-perceptual and semantic elements, and strong self-referential and emotional connotations. Studies of autobiographical memory in semantic dementia reveal intact retrieval of recently experienced events (Irish et al. Reference Irish, Addis, Hodges and Piguet2012; Piolino et al. Reference Piolino, Desgranges, Belliard, Matuszewski, Lalevée, De La Sayette and Eustache2003), attributable to preservation of sensory-perceptual representations stored in posterior parietal brain regions (Irish et al. Reference Irish, Landin-Romero, Mothakunnel, Ramanan, Hsieh, Hodges and Piguet2018). In contrast, recall of remote autobiographical events is impoverished in semantic dementia (Irish et al. Reference Irish, Hornberger, Lah, Miller, Pengas, Nestor, Hodges and Piguet2011), given the increased semanticization of episodic experiences with the passage of time (Moscovitch et al. Reference Moscovitch, Nadel, Winocur, Gilboa and Rosenbaum2006). Considering how profiles of autobiographical memory corroborate or challenge the integrative memory model framework will be crucial to validate and update the model with respect to self-defining expressions of the episodic memory system.