The concept of representation. We've actually been talking about representations all along, but today we're going to make it more explicit. And I'm going to consider the concept of representations and draw upon a variety of examples. A representation is something that stands for something else. We use representations all the time. So, for example, this arbitrary symbol here consisting of a horizontal line, a vertical line, and then a circular line conveys to us a meaning, the number of items. The arbitrary nature of this relationship is captured by the fact that we actually have many different possible ways to convey that same meaning. For example, we can write the word five. F-I-V-E, or we can pronounce those sounds. All three of these methods are arbitrary symbols conveying the same meaning. Representations are used not just in human language, but also in electronics and other technological systems. And I'm going to develop this analogy in a bit of specific detail, because I think it provides insight into the challenges faced by the brain in representing and transforming information. Many electronic devices use a type of code called digital coding. In a digital code, these symbols that we're talking about involve a series of zeros and ones that are used to convey the presence of components of that number in the form of powers of two. Which bits in this code are on, indicate which powers of two should be combined. So for example, if you have a zero, a zero, three ones and then three more zeros. What that means is we can ignore these zeros and just consider the ones. Based on it's position in the number. That corresponds to two to the fifth. This corresponds to two to the fourth. And this corresponds to two to the third. These numbers can be evaluated and added together to produce the total of 56. So that this symbol corresponds to this meaning. In neural representations, the symbol consists of a neural activity pattern. And the meaning can involve either a sensory stimulus, which has caused that activity pattern, or a desired movement, which is controlled by that activity pattern. So we can ask, in what ways does neural activity stand for information, and in particular, spatial information? Well, you've already seen one example kind of representation. And that is the brain map, which we see in the visual system and in the somatosensory system. And in a brain map, you can think of this as being like the brain's version of a digital code. If you plot neuron activity as a function of neuron identity or position on the retina, the zeros correspond to neurons that might not have anything in their receptive field right at the moment. The ones correspond to neurons that do have something in their receptive field and are therefore active. Of course, neurons are different from logic gates in that they don't have simply on and off patterns, but rather exhibit grated firing rates, so really it's more something like this. If you plot the number of spikes. First is the location of a stimulus. An individual neuron might have a response pattern that looks something like this, graded activity that reaches a peak for a particular location at the center of the receptive field, and lower activity for locations on either side. Different neurons have different receptive field locations, so that taken together, these neurons can form a map. Well, why do representations matter? Representations matter, because these codes have to be read and interpreted to produce behavioral responses. So, when we think about these neural activity patterns. Remember that the brain doesn't know about the sensory stimulus. It only knows about the neural activity pattern. It has to figure out how to infer what that sensory stimulus was based on that activity pattern. When neurons form a map, it's critical to use the activity of more than one neuron to try to figure out where the stimulus is located. And the reason is that the activity of any individual neuron is ambiguous. If this neuron is firing at this firing rate, you could be pretty sure the stimulus is here. But if it is firing at one of these firing rates, you can't tell the difference between whether a stimulus is located here, or here. So these maps have to be read. And the activity of groups of neurons has to be evaluated in combination. That's what we were talking about when we talked about how the map for visual stimulus velocity in area mt is converted into a command to generate a smooth pursuit eye movement. So if a particular neuron is firing at that firing rate. And another neuron is firing at that firing rate. Together, you can see that the most likely explanation is for a stimulus at that location, and adding a third neuron, might just provide the icing on the cake. Overall, it is the identity of the active neurons that is signaling the location of the stimulus. The central elements of a map are really very simple. Individual neurons show peaked tuning functions, and there are many of these neurons. If you have those two factors, then what you have is a situation in which all locations can be represented in a point-to-point fashion. But it turns out that that's not the only way that the brain can represent information. There's an alternative, called the brain meter, that I'll tell you about in the next video. [MUSIC] [MUSIC] [MUSIC] [BLANK_AUDIO]