Conjunctive or context-invariant coding in the human hippocampus? by Rodrigo Quian Quiroga instant download

Conjunctive or context-invariant coding in the human hippocampus? by Rodrigo Quian Quiroga instant download

Concept cells are neurons in the human tive coding in the rodent and monkey hiplikely to observe conjunctive, rather than inhippocampal formation with selective and pocampus, no such evidence exists in variant neurons, which is clearly not the invariant responses to specific individuals humans [1], and that conjunctive coding case [1,6,7]. For the same reason, given or objects. They have been proposed to leads to combinatorial explosion and, therethe ubiquitous finding of context-invariant be the building blocks of episodic memofore, is not suitable to encode human memcells (and the lack of conjunctive coding), ries and show context-invariant coding, in ory [5,10]. The only experimental evidence it is unlikely that two different hippocampal contrast to the conjunctive, contextto support the existence of ‘index neurons’ pathways would support conjunctive repdependent coding described in other speand conjunctive coding in humans offered resentations and concept cells, respeccies, with neurons responding to conjuncby Kolibius et al. [3] is a recent paper [11], tively, as argued by Kolibius et al. [3]. tions of features related to specific which has several caveats. First, the neuroMoreover, the coding of associations with memories, significantly changing their firing nal activations during encoding and recall concept cells [2] is not evidence of an interwhen varying the context [1,2]. In a recent were compared in only one trial, and it is unmediate representation between concept article in TiCS [3], Kolibius and colleagues clear how reliable such coding is. Second, it cells and index neurons, because the neuproposed a memory model based on is possible that subjects could have been rons responding to associated concepts ‘index neurons’ and conjunctive coding, thinking about the same concept during respond to one concept OR the other, linked this model with engram theory, and encoding and recall. Furthermore, the disOR both together [2,7] (i.e., it is a logical argued that concept cells arise from index ambiguation of overlapping memories OR, not an AND), thus being an invariant neurons. (e.g., two memories involving the same perand not a conjunctive code. son) was not tested. Third, these results Kolibius et al. base their model on ‘indexing contrast with a recent paper [6], in which There are still very few experiments theory’, by which the hippocampus prosubjects were explicitly asked to disambigexplicitly addressing the question of vides an index to neocortical areas, actiuate overlapping memories/stories and no whether, and to what extent, conjunctive vated by experiential events [4]. In general, evidence of conjunctive coding was found or invariant representations are present I agree with this view, although I believe with a much larger number of trials (bein the human hippocampus. Invariant, that assemblies of concept cells, rather tween 10 and 15 per story; Figure 1). They context-independent representations than conjunctive representations of specific also contrast with a previous study [7], appear to be better suited for the explicit memories, are what index to neocortical showing no-conjunctive coding when subcoding of abstractions and generalizaareas, and that a second key function of jects learned new associations. Moreover, tions. However, there is no free lunch the hippocampus is to encode associabased on the arguments presented by and this is at the cost of being prone to tions between concepts, providing the Kolibius et al. [3], one would expect that inimore interferences and false memories, skeleton of episodic memories [2,5]. tially nonresponsive neurons should start fircompared with specific context-depening to conjunctions of persons with dent conjunctive representations [5], a Assemblies of concept cells can also be particular contexts/stories (in [6]) or to spetradeoff that deserves further study and considered engrams, since, as engrams, cific associations ( in [7]), which was not the might be key to understanding the neurothey show reactivations [6–8], but reprenal underpinnings of unique human cogcase.sent concepts rather than specific memonitive abilities and intelligence. ries. In line with results in rodents Kolibius et al. propose that a subset of described by Kolibius et al. [3], modeling index neurons may create concept cells Acknowledgments studies have shown how engrams of from overlapping memories (e.g., seeing a I thank the Biotechnology and Biological Sciences