4/29: Episodic memory.pdf

4/29: Episodic memory: Retrieval II 27 Pattern completion and recapitulation Replaying back to cortex perceptual features of what happened during an event Occurs during sleep Occurs whenever we're consciously recollecting something Remembering what you've seen or heard: recapitulation of encoding activation Reinstating perceptual information Paired associates: word and something else they had to remember Trial 1: picture and visual referent (picture of dog and word dog) Trial 2: picture and audio referent (picture of dog and sound of dog barking) Retrieval Only saw visually presented word Like when you bring back to mind a visual object vs. an auditory sound Results Collected data during study and retrieval phases Words --> pictures Perceive: cortical visual areas, ventral temporal lobe Retrieval (only seeing word): greater activation in high level visual areas Brain replays/reinstates the visual patterns Words --> sounds Perceive: auditory cortex Retrieval (only seeing word): reactivation of auditory cortex Remembering what you've imagined: recapitulation of encoding activation Same cortical activation when you're reactivating some prior thought Regenerating a thought through engagement of mental imagery Subjects imagined individual words Some of the words, subjects were instructed to imagine a scene that might be described by that adjective Sometimes asked to describe a famous person Co-occurrence of and external word, thinking about the meaning of the word, and different classes of mental images Test Ss presented words All they have to do is try to remember the kind of mental image they generated Tell if they imagined a scene or a face Results Cortical pattern at retrieval is similar to what you engaged in at encoding Replayed during retrieval "Pair-coding" response in non-human primates Memory depends upon reactivation Necessary role of MTL for that reactivation Listening in on neurons in the primate brain that represent high-level visual object Single-unit data Listen-in, find a neurons Present monkey a bunch of objects Hope that the neuron prefers one of the 24 randomly selected object Having figured out preferred/non-preferred objects for the neurons, they pair up the objects Learns association between pairs of objects Presents objects, either preferred or un-preferred object In contrast to response of neurons prior to learning, after learning the neurons suddenly prefers the one it was paired with Explanation Experiences changes response of neurons When things that initially weren't preferred, they end up preferring the ones they were associated with This is a "pair-coding" response Ruled out that this was a perceptual learning effect in cortex If you lesion MTL, the response to the associate goes away, but the response to the cue remains the same Loss of MTL representations = loss of stored association Role of MTL in cortical recapitulation 11 Listen in on high-level visual neurons Listen in on perirhinal cortex If the pair-coding response is actually a retrieval response, then it should be lagged in the cortex in response to preferred stimulus If the stimulus being presented is the preferred stimulus, it should be driving neurons --> perirhinal --> replays back in cortex If acquired response is actually a representation of the preferred stimulus, then when you present the non-preferred stimulus, should bring online neural representations of preferred stimulus Paired associates during study Present either preferred or non-preferred as retrieval cue Preferred: drive cortical neurons --> perirhinal --> cortex Drives the active retrieval, then no longer relevant, so they shut down Feed forward response Early visual areas that are coding for low-level visual properties Gets to IT/TE, high level representation It's early if it's the preferred object Feeds forward to MTL, then back out to cortex Non-preferred Significantly lag response Takes a few hundred milliseconds Takes time for neurons that code for the non-preferred object to feed to MTL, for pattern completion in MTL, for pattern activation in neocortex --> IT/TE These aren't representing non-preferred object, but they're reinstating it through active retrieval Conjunctive memory depends upon MTL Reinstates other features out in cortex that were around during learning Next midterm Review session: Monday May 10, 5:30-7:30 Memento screenings Recognizing the past: recollection and familiarity Unitary view = All memory judgements are based on how strong a stimulus is in memory Single kind of memory is build/encoded during experience The stronger the encoding of that episode is, and the greater the similarity of retrieval cue and episode, the more familiar the test item will feel The older a memory get, the stronger it gets Active signal detection = do I have enough activation coming out of memory that tells me that it's strong enough to decide that I've seen it before? Single kind of knowledge coming out of memory systems, and we make a decision Decide that if they're above decision bound, they're old d' = distances between distributions Unless d' is incredibly far apart, we'll have some overlap False alarm = false familiarity Dual process (two memory signals) = recollection and familiarity Recollection = Hippocampus system supports recall Associative/conjunctive information Depends upon conjunctive traces Cued recall, free recall Attention at time of encoding is really important Requires cognitive control Familiarity: Recognition memory is something like signal detection Can use recollective information to decide if the object is old or new Undifferentiated Can't point to particular features that are being reinstated Competing influence of R and F: "Becoming famous overnight" We have two processes Familiarity Recollection Usually, these work together to support a memory judgement Novel names of made up people and duped subjects by making names familiar and decreasing probability that people could identify source of familiarity Full and divided attention If recollection is dependent upon attention at encoding, then items that happened under divided attention might be familiar, but won't be recollected False alarms full attention: more false alarms in nonfamous + novel the ones that were already seen didn't feel familiar, so they made the right decision divided attention You can have a unitary process theory with multiple decision bounds and still handle these data Just by looking at behavioral data, it's hard to discern between the two theories Hippocampus Looks like it's building conjunctions that will later support assocaitions Isn't encoding indiv. items MTL Encoding indiv. items, but not conjunctions Lizz Fay Untitled