Fragmentary Understanding of Memory

My thesis argues that memory resembles navigation by fragments, as proposed in [1]. To relate to it, imagine you are in Paris and you already know quite well the neighbourhood of where you are staying. Now, it is easy to get to the Eiffel Tower: you may need some help with an overall direction, but once you follow it, it does not really matter how exactly, very soon you will see the Tower and you will nd your way to it. When you get closer you will remember the immediate surroundings of the Tower and how to nd the nearest coffee place. Similarly from there you can get to Notre Dame: now you may just follow the river, it is one-way and you may only remember just a couple of spots along the way, but near the Cathedral you recognize every pigeon. Here the remembered fragments are the two landmarks, their neighbourhood and the river, to some extent. Keep this in mind. I prepared a guided tour for you, where the fragments I learned to navigate through are memory phenomena that can be studied from a point of view of navigation by fragments. Pick a sustainable vehicle, I suggest a sailing boat as it must have the right speed, or you may imagine this trip as cycling uphill (as I certainly felt it), and follow my thoughts. We start with the hippocampus and its relation to memory and navigation. We discuss how the discovery of spatially selective cells there led to study memory systems in our brain as attractor neural networks, what fragmentary knowledge about the representation of space could be learnt from the hippocampus and the open questions that remain. Just across a bridge, in Chapter 2, we will attempt to answer some of those questions, studying a mathematical model of an attractor neural network in CA3. We will expand our knowledge about the quasi-continuous maps our model forms for multiple sample environments, their storage and their usage. We will argue that CA3 network storage can in fact be thought of as a fragment assembly. We must take a series of one-way turns to reach some understanding of how human recall relates to virtual rat navigation, but I will be your guide. In chapters 3 and 4 we will discuss, using free recall as a model example, how human memory, too, can be thought of as navigation by fragments. We will present a series of experiments and simulations of a Potts network that together point at the semi-random nature of human recall. We suggest that when given a plain environment to learn - a hexagonal grid on a screen, as an empty box for a rat in a lab, human participants tend to memorize locations on the screen by `seeing' there various familiar fragments. And the more restrictive the memorization task is, the least they can reach these attractive patterns. We will discuss how common biases and unanimity across participants in fragment activation can be predictive of human recall capacity. Finally, we will briefly visit Milan of Mind Wandering and Rome of Remembering Poetry. We will argue that, despite being seemingly (and luckily) far from each other, both of these places-processes have in common their functional reliance on fragmentary schemata. First, we will propose an experiment that aims at quantifying the effect of recently acquired episodic schemata on mind wandering in participants with a lesion to vmPFC and in their healthy controls. Separately, we will suggest a mechanism of selective involvement of poetic meter variables as schemata helping remember non-words in non-poems. In the end we will gather to review the pictures from the journey and discuss the takeaways. Let's go!