Cognitive Neuroscience Society 2021 Data Blitz talk:
Hannah Marlatte, Derek Beaton, Sarah Adler-Luzon: Scene construction and PTSD:
Society for Neuroscience 2021 poster talks:
Zorry Belchev, Hannah Marlatte, Madison Fraser: The effect of hippocampal damage on integrating information across time:
PGSA poster day 2022 talk:
Pratyush Manon, Kyle Nealy, Asaf Gilboa: How does prior knowledge affect memory transformation in neural network models:
Cognitive Neuroscience Society 2020 talks:
Erik Wing: The varied influence of prior knowledge on perception, retention and new learning
Prateek Dhamija: Distinct electrophysiological responses to aversive first-order and higher-order conditioned stimuli
Ariana Giuliano: Differential influence of ventromedial prefrontal cortex lesions to schema and category knowledge
Rotman Research Institute 2020 conference talk:
Zorry Belchev and Adina Levi: Age differences in the relationships between two hippocampal-dependent domains: Memory and spatial cognition
Prefrontal cortex and strategic retrieval
In 2002, based on clinical observations of patients with confabulation, Moscovitch and colleagues hypothesized the existence of at least two memory monitoring processes (Gilboa and Moscovitch, 2002; Moscovitch and Winocur, 2002): a rapid, intuitive, automatic "felt rightness" mechanism that people have little control over, and the other is a deliberate problem-solving "editor" which detects memory contradictions, inconsistencies and discrepancies.
In 2004 I further developed the idea and brought to bear much of the existing neuroimaging literature in support of such a duality of memory monitoring (Gilboa, 2004; Neuropsychologia). In that paper I hypothesized that the former, critically supported by the ventromedial prefrontal cortex (VMPFC), is likely related to more primitive emotional and reinforcement processing systems of the brain and body, and the latter is executive problem solving routines applied to the memory domain and more related to (right) dorsolateral prefrontal cortex (DLPFC) functions.
Gilboa et al. (2006, 2009) were the first two empirical studies that provided direct evidence for the existence of the two systems. Working with a group of confabulating and non-confabulating patients with VMPFC lesions, Gilboa et al. (2006; Brain) showed that even when strategic search requirements of a task are minimized, confabulating patients still fail to detect implausible information, and endorse it as veridical memory with high levels of confidence.
In a later study (Gilboa et al., 2009; Journal of Neuroscience) we measured electrophysiological responses in patients with VMPFC lesions and healthy controls while they made "personal familiarity" judgment for famous, non-famous and personally familiar faces. This allowed us the temporal resolution to test the contention that "felt rightness" might be a process occurring rapidly before conscious processing of the information occurs. We demonstrated that patients were missing a very early posterior electrophysiological familiarity signature, which distinguished familiar from non-familiar faces in controls. They were also missing a very early anterior component, which was highly correlated with making rapid, accurate responses, whereas later differences correlated with accuracy but not RT, in keeping with the dual monitoring hypothesis.
I recently suggested (and presented evidence) that in addition to monitoring, confabulating patients also have deficits in memory control processes, and that regaining control abilities might account for the fact that many patients stop confabulating (Gilboa, 2010; Cognitive Neuropsychiatry). I also showed that the strategic retrieval framework could be efficiently used to distinguish confabulation from seemingly related disorders such as delusion. In an interesting twist on this theme, we have recently shown that severe prospective memory deficits due to frontal polar lesions can exist in the context of intact (implicit) intention superiority effect (Uretzky and Gilboa, 2010; Journal of Cognitive Neuroscience). We have begun collecting data from patients with VMPFC lesions, and they appear to show the reverse pattern: relatively intact detection of prospective cues coupled with a missing intention superiority effect, which we suspect operates based on similar principles as feeling of rightness (unpublished data).
Currently, together with my graduate student Vanessa Ghosh, we began to explore the role of schema in the evolution of confabulation and the role of the vmPFC in both schema representation and confabulation. In all of my previous articles (Gilboa, 2004, 2010; Gilboa et al., 2006, 2009) I have hypothesized that among other factors, vmPFC damage leads to confabulation because of aberrant interactions between schemata and current memory cues (see Ghosh and Gilboa, 2014; Neuropsychologia for review). Interestingly, recent human (van-Kesteren et al., 2010; PNAS) and rat (Tse et al., 2011; Science) studies have revealed that vmPFC is important during schema-dependent memory encoding and updating.
In the first human lesion study on the topic, Vanessa found that patients with confabulation have nebulous schema structure and that this impairment is associated with damage to sub-callosal cortex and posterior orbitofrontal cortex (Ghosh et al., 2014; Journal of Neuroscience). This finding is particularly intriguing because the task itself did not require episodic memory functioning as all information was available to the patients on-line. Using EEG with vmPFC patients and healthy controls, Vanessa is now exploring the physiological mechanism that may underlie the manner in which vmPFC might mediate schema functions. We have also developed a task that would allow us for the first time to observe not only how new information is integrated into an existing schema, but also how schemata change when new information is studied (accommodation).
Medial Temporal Lobe and remote memory representation
A second major area of interest I have been pursuing is the role of MTL structure in remote memory. Prevailing views of MTL function posit three basic principles: (i) the hippocampus and neighbouring MTL cortex (MTLC; perirhinal, entorhinal and Parahippocampal cortices) constitute a fast memory acquisition system of declarative memories, either by serving as a link system between memory traces or through a parallel set of memory representations. (ii) Posterior neocortical regions (primarily the anterior and lateral temporal neocortex) support the actual memory traces; however due to its limited plasticity it is a slow-learning system that can only form enduring memory representations via the prolonged support of the MTL system. (iii) All declarative memories go through a process of consolidation through which the neocortical representations become independent of their MTL contribution.
Much of my work on MTL and remote memory representations in recent years has challenged principles (i) and (iii), bringing together apparently separate debates in the literature. On the one hand, contrary to principle (i) it has been suggested that distinct subregions within the MTL serve distinct memory processes within the declarative memory system (Aggleton and Brown, 1999; Eichenbum et al., 1994). Thus, the EHS is specifically implicated in recollective processes of declarative memory retrieval, whereas MTLC support familiarity-based memory judgments. Gilboa et al. (2006; Hippocampus) presented direct evidence that this distinction might also apply to remote memory. Using a unique autobiographical recognition paradigm couple with ROC curve analysis, they showed specific recollection deficits following hippocampal damage, while more extensive MTL lesions led to deficits on both processes. More recently, my PhD student Shani Waidergoren demonstrated that using process dissociation procedures (PDP) in semantic memory can also expose the contribution of recollection-like processes to semantics (Waidergoren et al., 2012; Neuropsychologia). This is one of only a handful of studies that demonstrates the idea that the hippocampus is critical for aspects of semantic memory, and that this contribution reflects an inherent property of semantic memory and its processes, rather than a reflection of interaction between memory systems. However, to expose that dependence semantic memory needs to be probed in particular ways.
Similarly, principle (iii) has also been challenged by the Multiple Trace Theory (Nadel and Moscovitch, 1997), suggesting that distinct memory types show different patterns of consolidation. Semantic memories go through relatively brief periods of consolidation, and older semantic memories become independent of their MTL component, as suggested by the prevailing views of declarative memory. By contrast, episodic memories depend on their MTL component for as long as they exist. Every time an episodic memory is recalled, new traces that link its components are created within the MTL system, strengthening the representation of more remote memories. Using multivariate analysis and combining whole-brain volumetric data and performance on semantic vs. episodic autobiographical memory, Gilboa et al. (2005; Hippocampus) showed that whereas episodic memory performance was related to a pattern of tissue loss in both MTL and temporal neocortex (regardless of age), semantic autobiographical memory was related only to tissue loss in temporal and frontal neocortex, in keeping with the predictions of MTT.
Using functional neuroimaging Gilboa et al., (2004; Cerebral Cortex) showed that vividness of the memory (i.e. the amount of experiential detail retrieved) rather than remoteness determined hippocampal activity in healthy controls. They also showed that remote memories were associated with a more distributed pattern of activation along the rostro-caudal axis of the hippocampus, in keeping with the predictions of MTT. This latter unexpected finding has recently been replicated using more advanced imaging techniques (Bonnici et al., 2012; J Neurosci) and figures prominently in recent literature exploring functional diversity along the long axis of the hippocampus.
Neocortical plasticity
Traditional neuropsychological and neurobiological theories of declarative memory suggest the existence of two complementary memory systems: a hippocampal-based system, which specializes in rapid acquisition of specific events, and a neocortical system, which slowly learns through statistical regularities (Principle ii above). Initial acquisition of declarative memories usually depends on the hippocampal memory system, which continues to support these representations until the slow-learning neocortical network can independently represent them through the passage of time and/or repetition. Direct neocortical acquisition of information may occur only through intense and relatively invariable repetition.
In our work we employed Fast Mapping (FM) (Sharon et al., 2011; PNAS) and obtained evidence that challenges this widely held view. FM supports the acquisition of novel associations between words and their referents in childhood and is believed to support the rapid acquisition of vocabulary in children as young as 16 months of age. Through FM children gain information about a words meaning by generating 'working hypotheses' based on the way and the context in which it is used in a sentence. In our study, participants were told the task was a perceptual task and on each trial they saw a novel and a familiar picture and had to answer a question which contained the target's label (e.g. "Is the Numbat's tail pointed upward?"). By comparing the novel and familiar items, participants infer that Numbat is the name of the novel item (the process of FM). We administered our task to four patients with dense MTL amnesia, two with Anterior Temporal Lobe (ATL) lesions and to healthy matched controls. On a surprise associative recognition memory test administered at delays of 10 minutes and 1 week, our MTL amnesic patients performed as well as controls. By contrast, on a matched standard explicit encoding (EE) task (e.g. "Remember the Mangosteen" with only Mangosteen on screen) patients were markedly impaired compared to controls and performed at chance. These data show that learning through FM is independent of episodic memory and the MTL, and indicate that FM is mediated by extra-hippocampal structures. Chance performance on FM by two ATL patients suggests these neocortical structures are crucial for FM.
We have since conducted two neuroimaging studies, one examining encoding using a subsequent memory effects paradigm (Atir-Sharon et al., submitted) and the other examining retrieval and the effects of overnight sleep on memories that are directly acquired by neocortical plasticity (Merhav et al. submitted). Neuroanatomically, both studies appear to converge with our patient studies in showing preferential involvement of the hipoocampus in encoding and retrieval of items in EE and preferential involvement of the ATL in encoding and retrieval of items acquired through FM. More interestingly, however, is that together with behavioural and lesion studies examining interference effect during direct neocortical learning, we have begun to uncover the mechanisms and principles that govern this unexpected form of learning.
Our studies have uncovered some paradoxical characteristics of this kind of learning that are radically different from normal declarative learning. First, we found evidence that memories acquired through FM go through some process of consolidation or cohesion that can be interfered with using behavioural amnesic agents (Merhav et al., 2014; Hippocampus; Merhav et al., in preparation). Moreover, we found that these representations are susceptible to Catastrophic Interference as predicted by models of neocortical learning (McClleland et al., 1995) but never demonstrated in a biological system until now (Merhav et al., 2014; Hippocampus). Finally we found using fMRI that associations acquired through EE show evidence for overnight systems consolidation mediated by vmPFC-hippocampal interaction. By contrast, associations acquired through FM show little change in network representations and instead might become more stable through changes in local circuit cohesion possibly mediated by synaptic consolidation processes (Merhav et al., under review).