2 January 2022 
11 mins In this book, Rodrigo Quiroga illuminates the inner workings of the brain, specifically the aspect of how memory works in humans.
###The Neuron and its models
The fundamental unit of the brain - the neuron - consists of a long body, called the axon, and thread-like extensions, called the dendrites. The neurons are inter-connected to for a web though gaps called synapses where electro-chemical signals (neurotransmitters) flow from the axon of one neuron to a dendrite of another. Glutamates are neurotransmitters that strengthen or weaken the synapses between neurons - forming the basis for building memories. Glutamates and dopamines stimulate activity (released by excitatory neurons), while gamma-aminobutyric acid (GABA) suppress activity (Inhibitory neurons). Computational neuroscience, generally deals with the study of these neural activities and firing patters that lead to different cerebral functions. It is one thing to understand primitive cerebral functions form the firing patterns and it is another to understand the emergent phenomenon behind behavioral processes.
Two fundamental ideas in computational neuroscience are Hopfield's model of neural network and Hebb's theory of neural activations. Hopfield's model explains how some random stimulus on a given neural network can find the nearest memory associated with that stimulus (in terms of neuron activations), while Hebb's theory explains how those neural networks form through reinforced connections between them. Interconnected neurons that always fire together form a module called Hebbian cell assembly - representing some memory. Over time the connections get weaker, if they are not activated often - explaining memory loss / shrouded memories. Hopfield's model - memory retrieval; Hebb's theory - memory formation.
###Processing, Filtering, Storage, & Abstraction in the Brain
The brain's capacity is limited - in terms of the number of neurons. And they are solely responsible for all of human cognitive behavior, perception, regulation of various bodily mechanisms, and so on. The brain deals with this by pattern matching and extrapolation. For instance, our eyes, do not capture everything we see in our field-of-view in full detail (owing to its structure) and do not transmit all of its signals to the brain for a complete perception. Experiments indicate that our eyes transmit information at the rate of around 1MB/s to the brain. The brain too processes the visual signals depending on whether the subject is in the focus. Subjects out of focus need only simple processing like contrast and edges for a quick representation, while subjects in focus are processed much more. This is called center-surround organization. Furthermore, the brain make our eyes move involuntarily (called saccades) to bring different things within our field of view into focus. This also provides a means of creating redundancy for a more robust perception. All these, happen in our cerebral cortex. These representations of visual signals in our brain has deep repercussions from how patients after eye surgeries are treated, to studying brain degradation due of illnesses or old age, to artists exploiting them to create visual experiences.
... memories are usually generated from perceptions, since we tend to create memories of things we see or hear. But the most important reason to discuss vision as we probe the subject of memory is that the brain uses very similar strategies to see and to remember. Both processes are based on the construction of meaning, an interpretation of the outside world that relies on selecting a minimum of information and making abstractions—while discarding a multitude of detail.
Forgetting is an important and an essential behavioral process. While the lament towards forgetting important memories is aflood in literature, there are few works exploring the consequences of never forgetting. Forgetting, or discarding information is crucial for building abstract representations that can help in extrapolation and pattern matching. Luis Borges, beautifully writes in "Funes the Memorious" about a guy who can't forget anything -
Let us not forget that he was almost incapable of having general, Platonic ideas. Not only was it difficult for him to understand that the generic term “dog” could embrace so many disparate individuals of diverse sizes and shapes; it bothered him that the dog seen in profile at 3:14 would be called the same as the dog at 3:15 seen from the front. His own face in the mirror, his own hands, surprised him every time … [Funes] was the lonely and perceptive spectator of a world at once multifarious, instantaneous, and almost intolerably precise.
Herman Ebbinghaus proposed a theory of enhancing memories - converting short-term memories to long-term ones through repetition. He called this memory consolidation. He constructed "forgetting curves" to depict how memories fade away in time (in accordance Hebb's theory), and how memories can be strengthened through space repetition. Repeated experiments have revealed that our memorizing rate is even slower than our processing rate of 1 MB/s - just 1 B/s to 2 B/s. Note that this is the information contained in the representation in our brains. These compressed abstract representations leads to unconscious inference where signals are interpreted, filtered and stored.
The construction of a schema based on such inferences—remembering our subjective interpretation of reality rather than reality itself—is precisely the source of false memories that lead us to be certain of events that never happened.
Mnemonics is the art of enhancing memory through various techniques - usually through associating a memory with some place - method of loci. According to this method, memory retrieval can be made easier though the following techniques -
- The memories are organized or have some sort of structure. This avoids interference, where some memories block others.
- Creating mental images. Mental images turn a set of disjoint memories into a connected one ans gives a spatial aspect to them. More memorable memories can be given more remarkable visuals.
- Creating associations between memories.
The above aspects have indeed been found in people with strong synesthesia - where they mix perceptions from different senses - "seeing" or "hearing" numbers so on. Interestingly, mnemonists like Solomon Shereshevsky and Kim Peek have indeed reported struggling against an overwhelming and uncontrollable tide of memories and associations, sparked by each and every word - similar to that of Funes. This lessened their ability to reason or keep up with some speech or a book. Borges had this to conclude about Funes' ability -
To think is to forget differences, to generalize, to abstract.
With the above knowledge of rote memorization and remembering through mundane associations, we can explore how we remember things rather than how much we remember. This is essential in our age of internet and information bombardment. We can now store data digitally and retrieve them much sooner than recalling them from our memories. So, what we need now is how we make associations and process the data into some context.
To be creative and intelligent, we must go beyond merely remembering and undertake completely different processes: we must assimilate concepts and derive meaning. Focusing on memorization techniques limits our ability to understand, classify, contextualize, and associate. Like memorization, these processes also help to secure memories, but in a more useful and elaborate way; these are precisely the processes that should be developed and encouraged by the educational system.
Learning must reinforce core concepts - not be repeated memorization but through repeated but different associations to the same concept. A concept must be learned through different nuances, different contexts, though different associations. This helps us retains the concept in our memory and reason with it, in different and newer contexts - helping us become more creative and supposedly more intelligent.
The brain has a limited capacity, and we should focus its resources on processes of comprehension and thought, not on memorization.
The Atkinson-Shiffrin model classifies memory into 3 types - sensory, short-term memory and long-term memory based on how long they last. Sensory memory lasts for a fraction of seconds, while short-term memory can last for few minutes. Sensory memory can be converted to short-term memory based on focus and attention, while short-term memory can be consolidated to a long-term memory based on repetition and retrieval.
Fascinatingly, humans can still learn motor skills - those that involve physical movement along with some sensory stimuli - even without any ability to form long-term memory. This has been tested and verified in patients whose hippocampus - place where the long-term memories are stored - was removed. Therefore, this learning of motor skills, or muscle memory if you will - is called non-declarative memory and is not stored in the hippocampus, although it is classified under long-term memory. Sensory memory can also become a type of non-declarative memory if it is strong enough to cause a strong emotional response. This is called emotional memory. For example - a like or dislike of a particular food can become ingrained in the memory to become an emotional memory. Alternatively, the declarative memory is the usual memory of facts, events, concepts, experiences and so on. But of course, memories can be more nuanced and finer and finer subclassifications can be made.
What we mean by "thinking" is usually the processing of concepts - basic units of thought. How these "concepts" are stored and processed in the brain and its relation to the representations of input signals is an interesting one. In case of visual system (knows as the ventral visual pathway), the brain contains specialized neurons that respond only to certain signals - like faces, hands, fruits etc. There are neurons in different areas that respond to different information and along the pathway - the complexity of the encoded information increases.
we move from a representation of local contrasts in the retina, to one of borders in V1 (primary visual cortex), to one of faces in the inferior temporal cortex.
These increasing complex representations are then sent to the hippocampus where the neurons start to encode concepts. A famous example is the "Jennifer Aniston neuron" which responded to the "concept" Jennifer Aniston and only Jennifer Aniston. This was because the patient tested was familiar with the actress. Further experiments have shown that we can localize neurons that activate when a certain concept is presented before them - albeit thought any sensory stimuli like pictures, voice, or text.
Furthermore, the same neurons that fire at a certain concept, also fire when a closely related concept is presented. Therefore, the associations between concepts are also stored in the hippocampus - though partial neuronal overlap between the two concepts (recall Hebb's principle). Note that these concepts constitute the Declarative memory and hence are stored in the hippocampus. The primitive representations in the ventral visual pathway is an example of sensory memory and thus, exists only for a small period of time and cannot be used for reasoning.
...it is no coincidence that these neurons encode associations, since associations constitute the basis of memory
Higher the neural plasticity, quicker and stronger are these associations made in the brain.
The scant information we choose to process depends on the task we have at hand—for example, we see the same book very differently if we are looking for something to read than if we need a way to raise the computer monitor. This flexibility in attributing meaning, in selecting which information to process and which to discard, is what defines our intelligence. ... Our capacity to manage and relate abstractions, coded by concept neurons in the hippocampus, is the basis of our memory—and, perhaps, the cornerstone of what makes us human.
The book was an overall "huh?! cool" for me, with some interesting neuroscientific results scattered here and there. The last chapter, titled "Can Androids Feel?" however, was a total mishmash of ideas ranging from materialism/functionalism, mind-body problem, to the connection between language and intelligence. The author failed to provide a coherent and insightful account on any of these topics.
All the above excerpts are attributed to their original author(s). Book info -
Rodrigo Quian Quiroga, The Forgetting Machine: Memory, Perception, and the "Jennifer Aniston Neuron", 2017. BenBella Books. ISBN-13: 9781944648541 ISBN-10: 1944648542