MEMORY 

PRE-READING: predictive coding 

Baseyian inference: a statistical method to calculate the probability. TED talk guy says that this is how we calculate movement since we have many noise sources that mean we do not have robotic movements 

1. Source 1: sensory input 

2. Source 2: prior knowledge 

   1. = BELIEF  → where you think your movement will take you 

      1. BUT there is an extra edge of possibility of where it MIGHT END that you do not really anticipate 

      2. You have to predict the probability of different senenory inputs based on your belief (where it will land) → i.e. make a prediction of the future 

THIS IS THE BASEYIAN INFERENCE - but could also explain how we do movement 

The tickling experiment 

1. When one person generates the movement command the predict the sensory consequence and thus can not feel if they tickle themselves 

2. However, if someone else tickles you since you are not generating the movement command then you cannot subtract the sensory consequence and thus will fell the tickle and laugh 

Understand different memory modalities. 

Link each aspect of memory to specific neurological phenomena.

EPISODIC MEMORY - HIPPOCAMPUS 

H.M - 1957 - Removal of medial temporal lobes  - no new memories - hippocampus is in charge of encoding and consolidating memories 

1. The old memories seemed fine some maybe the neocortex stored those memories 

Connectivity to the rest of the corext 

1. HC is in the neocortex, does the HC lose all function after the memory is old?

1. Red bit is the hippocapusm, 

2. the blue arrows are connections to other parts fo neo-cortex 

3. and the green arrows are some sort of network for episodic memory which must exist or HM would have lost all his old memories  

1. THERE ARE THEN 2 THEORIES 

   1. A. says that there is a network and then th ehippocampus stops being involved and the memory is within that green neocortex 

   2. B. says that the hippocampus is always involved 

MRI STUDY : MULTI VOXEL MRI PATTERN ANALYSIS

1. 3 dimensional pixel

2. Recent and remote memories they had made over the last 2 weeks and then older ones - trained to think this way 

3. They found out which nteworks were active when the remote memory was thought about 

4. results : HC, tissues connected to hippocampus  vmPFC 

5. They trained teh machine to tell them whether they were thinking about a remote or older memory - for the hippocampus the machine could not base of of this alone whether the memory was remote or old (you can see that the lines cross over  

This supports explanation B - in that the hippocampus is always kind of active regardless of what type of memory - but older memories create a pattern of activation in the neocortex 

Hippocampal replay 

Place cells - 2014 O’keefe 1967 +1978 - a single cell recording from rats, there was a pattern in that the same neurons where activated when the rat walked in the same place

• They implanted electrodes in the hippocampus of rats to measure neural activity. He discovered place cells, which fire when a rat is in a specific location in its environment. This demonstrated that the hippocampus plays a key role in spatial navigation and memory by forming an internal "cognitive map" of the surroundings.

• These findings showed that single-cell recording is an effective method for identifying how individual neurons encode specific types of information, such as location.

Wilson and McNagughton 1994 - Hippocampus replay 

1. Reactivation of place cells in rats when they are asleep 

2. Yes it is in humans (Peigneux 2004) - found that areas active during a spatial navigation task were also active during Slow Wave Sleep using PET. A link between the amount of reactivation in the hippocampus during sleep and memory performance the next day

3. Not only related to spatial things - Rasch et al - conscious memory - slow wave sleep

What is hippocampal replay doing? 

1. Hebbian learning - brain is constantly changing its connections but also reinforcing those connections again and again - the more you do things the more likely it is to shape the brain 

   1. Even though episodic memories are unique, recalling them repeatedly could reinforce the neural connections involved, making them stronger and easier to retrieve in the future.

2. System level consolidation - how long term memories become stablisied and re organised in the brain over time 

   1. H.M did not lose his long term memory which means that memories cant be stored in the hippocampus forever (probably move to necortext).  There must be a transfer of the memory, eventually the memories move from the HC to frontal or temporal lobe 

3. The replay during sleep also happens in neocortex (JI and Wilson) suggesting that a process of transferring happens outside the hippocampus 

Two possible theories:

1. Hippocampus-driven consolidation → The hippocampus actively “teaches” the neocortex by repeatedly replaying memories.

2. Cortex-driven consolidation → The neocortex organizes memories independently, integrating information over time.

Sparse coding of the hippocampus - a subset of neurons are active in response to a given memory whereas others remain inactive 

1. It allows for more memories to be encoded 

2. Allows for pattern separation - making similar representations more distinct 

3. Auto-association = activation of an entire memory from activation of part of the neurons involved = role of spatial coding 

   1. Chadwick, Bonnici and Maquire 2014 - can we keep different memories apart and is the HC involved in this - films with overlapping info from 4 different events - Some where good in telling these apart - the firing patterns found that people who found this task harder had similar representation in their CA3 

      1. People with bigger CA3s were more likely to tell these events apart 

CA-3 - only has 46 mossy fibers going in but within this, there are 12,000 connections within the HC - this tells us that there is a lot of organisation within the HC 

Procedural memory-  error based learning - cerebellum and motor cortex  

Milner 1962 - something blocks you, you can only see your actions through a mirror in front of you and you will be asked to draw something 

What we learn from this?

Have an expectation → you don't meet it → try something else = error-based learning 

1. Errors produced information that are essential for our model to learn 

   1. Prism adaptation - Gaveau 2014 - Prism adaptation is a sensorimotor learning process where individuals adjust their movements in response to altered visual input.

   2. Connection to the Cerebellum – Provides insight into how the cerebellum and motor cortex work together in procedural learning.

What is evidence for error signal 

1. You can't tickle yourself - your brain predicts exactly what it will feel like so you brain has predicted it perfectly - this means that there is no error-based learning 

2. Tickling in schizophrenia - they assume that the tickling comes from someone else - individuals with schizophrenia often struggle to correctly determine whether they caused an action or whether an external source caused it - people with schizophrenia predictive model is weaker than people without it - so the brain will fail to predict an action that comes from itself- individuals may experience self-produced sensations as unexpected or externally caused

Semantic memory - distributed plus hub in the temporal lobe 

HM had spared semantic memory as well - Vargha-Khadem et al. (1997)

Paterson et al., 2007 → Semantic knowledge is stored in multiple sensory, motor, and association areas of the

1. Motor cortex, auditory cortex, and visual cortex 

2. Anterior temporal lobes (ATL) act as a central hub, integrating and generalising information from different brain regions. This hub helps us understand concepts across different modalities.

3. When thinking about a “dog,” the ATL links together visual (appearance), auditory (barking), and functional (pet, guard animal) information into a unified concept

4. Voxel Based Morphometry (VBM) reveals semantic memory impairments are correlated with anterior temporal lobe atrophy.