Relating Hippocampal Circuitry to Function: Recall of Memory Sequences by Reciprocal Dentate–CA3 Interactions

摘要: Summary and Conclusionsbasedonpastevents(Sokolov,1963).Thismodel,which Theviewthatemergesofthehippocampusisarelativelycan be thought of as stored sequences, allows predic- simple one that can related to the functional require-tion expected events (if this happens, then I expect ments episodic memory more specifically thethat happen). The brain continuously compares these storage retrieval sequences in context.expectations with reality. If comparison shows a ThedentateandCA3storeautoassociationsandhetero-“mismatch” expectations or something altogether associations, respectively, synapses their re-novel, encoding attentional processes are current connections. During retrieval, networkstriggered. Evidence human hippocampus is in- worktogethertoproduceanaccuratetime-compressedvolved such comes from imaging recall sequence. role perforant(Dolan Fletcher, 1997) electrophysiological stud- path input CA3 relates general problem howies (Halgren et al., 1980; Knight, 1996). Related experi- linked con-ments rat rabbit show hippocampal neurons text. It proposed contextual information arrivingbecome habituated repetitive stimulus but respond through perforant enables subset cellsvigorously when standard replaced by (and perhaps also mossy cells). A subsetan“oddball”stimulus(Buzsa´kietal.,1979;Vinogradova, fired salient represented1984). Other experiments directly activity fiber activity. In way, specific memoriesrelatedtomatch/mismatchconditions(OttoandEichen- encoded context without having explicitlybaum, 1992). clearest demonstration exis- represent highly redundant detailed in-tence an internal model formation. From CA3, travels CA1, re-which was suddenly omitted. Cells gion has two functions. First, it converts hippo-in mammillary body, recipients hippo- campal representation back cortical one,campal output, fire exact registration ex- thereby making possible for cortex interpret thepected onset duration absent (Vino- output. Second, CA1 cells compare thegradova, 1984). Because region site predictions made based on se-convergence (via Schaffer quences sensory reality arrive per-collaterals) raw per- forant CA1. mismatch detected,forant cortex) (Vinogradova, 1984), learning initiated. To-CA1 well positioned perform match/mismatch gether, ideas provide coherent explanation ofcomputation. function each pathways shown Figure 1.