Imagine you are asked to close your eyes and remember a childhood memory. The first thing that comes to mind is you, as a kid, playing in a park with a red ball; the colours are vivid, you can smell the freshly cut grass, and you can feel the shiny plastic from the red ball between your palms. You are then asked to remember a second childhood scenario. This time, you remember how, years later, you were doing schoolwork in your parents' lounge while your little sister was playing with the red ball. We know that we can remember things, and when we do, we create mental representations of what we are thinking – some of which can be so vivid we are confident we almost see them. But how does our brain create these scenarios?
Well, the Localization theory is a theory in neuroscience about how our brain works. It states that different brain functions are localized in other brain components. For example, a region in the frontal gyrus known as Broca's Broca's area is linked with language production. Patients with lesions in this area cannot produce language but rarely experience issues with comprehension. Thus, the function of language production and not language comprehension may be localized in this region.
The theory of localization has been applied to memory in different ways. In the late 20th century, the idea of concept-specific neurons (or memory-specific neurons) started to gain popularity. This occurred parallel to the term "grandmother cells" created by MIT professor Jerome Lettvin in 1969. It was a term used to describe the hypothesis that cells could selectively encode different memories – like one's grandmother.
The idea of grandmother cells was heavily criticized, and its use eventually declined in academia. It was too ambiguous of a term and lacked sufficient examples of empirical evidence to be integrated into scientific discourse. Still, the term opened the stage for debates surrounding memory localization. Since there was never any clarification on whether neurons would respond to the concept of "grandmother" or to just one's own grandmother, it led to discussions asking whether everyone's unique representation of concepts was innate to a particular neuron or a result of its unique wiring with other neurons. Additionally, it opened the stage to questions like, do concept-specific neurons respond to different stimuli that remind you of the concept they represent? Would grandmother's cells also fire when I smell the perfume she used to wear or the food she cooked? When reading the criticisms against the concept of "grandmother cells", you see that authors have different individual interpretations and questions about the idea. Still, from what I gathered ", grandmother cells" were meant to represent single-cell units capable of harbouring all information related to a specific concept or memory.
The idea of "grandmother cells" came back into popularity recently, following the discovery of what ended up being called the "Jennifer Anniston cells" and the "Halle Berry neurons" in 2005. Both terms arose after studies found subsets of medial temporal lobe neurons that would fire every time participants were presented with pictures of celebrities. These neurons even fired when the pictures were nothing alike, and the same effect occurred for pictures of landmarks like the Empire State Building and the Sydney Opera House. But contrary to grandmother cells, these neurons also fire for other related concepts. For example, Jennifer Anniston's neurons also fired for other cast members of friends, suggesting that the same group of neurons creates associations to represent similar concepts rather than one-to-one item-specific neurons. A more academically accepted idea is that memories may be physically represented by unique neuronal connections.
Theories relating memory representations to specific neuronal populations are not exclusive to long-term memory formation. Similar stimulus-specific neuronal responses have been proposed to underlie other types of memory, including Working Memory (WM). WM refers to our ability to retain and manipulate short-term information within a mental space, and we use it to keep track of and engage with conversations, make decisions, and plan (to name some examples). For instance, if I mention the park, the red ball, and your parent's lounge, your WM would be the one that enabled you to recall the context in which those concepts were mentioned. One of the critical aspects of WM is that it allows us to keep information in mind for a short period, a delay between when we first encounter information and when we must recall it. The exciting part is that early primate studies found some subsets of neurons that would persistently fire after the primate stopped looking at the stimulus. These neurons turned out to be "memory traces", literally representing an "item" (in this case, visual stimuli) being held in memory during a delay period.
Of course, WM is not limited to holding single items. Humans can generally retain around 4 "items" or "stimuli" in this mental space, which makes things a little bit more interesting. Well, turns out that maybe different items are held in WM by different "clusters of neighbouring cells" that maintain heightened activation during delay periods after the presentation of stimuli. This brings us to the final accepted theory- memories are not solely encoded in single units. Still, they are instead made up of sparse populations of neurons that come together to form unique recollections of reality.
Papers of Interest:
1. "Memory engrams: Recalling the past and imagining the future" by Sheena A Josselyn, Susumu Tonegawa
2. "Working models of working memory" by Omri Brak and Misha Tsodyks
3. "On the Biological Plausibility of Grandmother Cells: Implications for Neural Network Theories in Psychology and Neuroscience" by Jeffrey S. Bowers
4. "The Value of Failure in Science: The Story of Grandmother Cells in Neuroscience" by Ann-Sophie Barwich