top of page
Dan

Population coding is not physically possible

Let's take a very simple example of a stimulus, the number 01. To get that complete stimulus from the environment to the brain, the two parts of the stimulus must be linked in your cognitive system. 0 must be associated with 1.


This means that if there are different neural populations that code for parts of a stimulus, those populations must meet at some point. In fact, if any more than one single neuron codes for a stimulus, the neurons must meet. Unassociated signals are no use at all - you don't want only part of the stimulus.


If anything demonstrates the recondite bankruptcy of neural coding theory, it's this: The stark physical reality is that no information survives a confluence of multiple neurons. When two neurons meet to cause an efferent neuron to fire, it's exactly the same as one neuron causing that neuron to fire. Fifty neurons - a billion - could feed into that neuron, but the result will be the same: the only signal that can come out of that neuron is 0 or 1.






A series of neurons is useless as a transmission system for anything except 0 or 1 / on or off. They can be networked, interconnected, braided, multiplied by a billion, but the output of any neural connection will be 0 or 1.


Of course, the obvious explanation is that the efferent neuron fires in response to a certain combination of neural signals. In other words, you have a grandmother neuron. However, there are two reasons why this doesn't help.


First, it doesn't chime with how cognition actually works. For example, If I ask you to picture a Greek-speaking platypus in a bikini on a paddleboard, you can do it. If you already had a neural pathway set up and ready to fire for that image, then you need help. Neuronal structures are plastic, but not that plastic.


Second, a neuron may code for a specific thing like Grandma or Jennifer Aniston, but you have to know that. All you're getting is a neural spike of 0 or 1 (or indeed, many if it's a population). That spike cannot be labelled. The neuron doesn't "know" that it's firing, any more than a button knows that it's being pressed, and even if it did (I note here that neurons "knowing" anything is magical thinking), it can't tell anyone. All it can "say" is 0 or 1.


In order to know that you're getting a spike indicating Jennifer, you'd need another neuron (or population) telling you that the population that's firing indicates Ms Aniston. But how would your brain know that that signal is indicating Jenny? All you're getting from that population is 0 or 1...


What we're skirting around here is the truism that ultimately, in order to be cognitively useful at all, neural signals must turn into something else. Something needs to convert signals from a specific set of neurons into a form that says "Jennifer Aniston", and not just 0 or 1. That thing cannot be a neuron, and therefore population coding does not make any kind of sense in the real world.







36 views0 comments

Recent Posts

See All

Introduction

This blog is for people who believe in Science. That is, those who are looking for the objective truth according to observation and...

Comments


bottom of page