Work of consciousness: what are Gabor's spots?
At one of the first lessons in my first neuroimaging course, a confusing dialogue occurred.
Professor : in this experiment, people look at the cross in the center of the screen, while the Gabor spot is shown to the left or right of the cross ...
Student : I'm sorry, but what is a Gabor spot?
Professor : Ah, well, this is a convolution of a sinusoid with a Gaussian curve.
He smiled at us, not paying attention to the fact that behind him there was an image of Gabor's spot on the screen. He raised his eyebrows expectantly. His whole posture said: "Well, now I understand?"
Student : Uh ...
Professor : No? Here, let me show you.
Still ignoring the presentation screen, he turned to the blackboard. On it, he drew a sine wave, and below it - a Gaussian curve.
“And now you carry out a convolution operation on them!” [ According to some experts in the comments, this function is a product of Gaussians and sinusoids, not a convolution / approx. trans. ]
The student gave up. Perhaps he had some ideas about the mathematical operation of convolution, but there was no necessary intuition. He needed someone to simply point his finger at the right place on the screen: here, this is Gabor's stain.
It could be a story about how I sometimes felt at the place of this student. Or a story about teaching. Or, perhaps, about how absolutely accurate information may seem to us meaningless. But I would like to focus now on the fact that the Gabor spot is more than just a convolution of a sinusoid with a Gaussian curve.
Suppose you have a bunch of kittens that were raised in an environment where there were no landmarks other than vertical ones. Kittens would only see vertical stripes. What happens if in a few months these kittens are released and allowed to interact with the ordinary world around them?
What will happen is that after an uncertain start, they will learn to interact with the outside world. They will begin to study it, play, behave like kittens. But if they come across a long, thin, horizontal object - for example, a black cable, stretched across a white carpet, they will behave as if it does not exist. They will not be frightened if he suddenly goes to their side and will not miss him if he jerks. They will show selective blindness to him, although their eyes will work fine. The root of their problems will be in the brain.
The primary visual cortex — the first of a series of cortical zones processing visual information — is sensitive to the orientation of the lines. There are cells called “simple” (or strip detectors, or face detectors) that are activated in response to different degrees of inclination, depending on the orientation of the faces of the objects observed. If we had walked along the surface of the primary visual cortex, we would have walked slowly from the areas responsible for vertical orientation to areas that are tuned to greater and greater angles of inclination. This means that one group of cells becomes very active when it sees a vertical line, another - if it is tilted slightly, and another one reacts to a horizontal line. Cells “tuned” to a vertical orientation will still respond in response to not exactly vertical lines, just smaller and smaller, depending on the difference between the “preferred” vertical orientation and the visible one.
Sensitivity of a neuron tuned to vertical stripes as a function of strip orientation
In the case of kittens, the cages that should have been responsible for horizontal orientation began to respond to other orientations due to sensory deprivation at a critical period of development. They do not have cells that react to horizontal stimuli, so at this stage of neuroprocessing the visual signal goes out.
Upon learning of this orientation setting, we can begin to understand one of the phenomena related to perception. Namely, if we look at the same oriented strip for a long time, then our ability to judge the inclination of subsequent stripes, which are approximately the same oriented, will decrease for a while, but this effect will be weakened more than by the inclined stripes. Now we know that this is due to the fatigability of the neurons - the more the neuron is activated, the more tired it will be later, and the less it will be able to accurately convey information about the orientation.
How do we know this? Thanks to thousands of experiments with Gabor's spots. Gabor spots are incentives that control early visual activity in a controlled way. They look like a sequence of black and white stripes, and they can be oriented in any way, they can be made well or poorly distinguishable, large or small, rotating or stationary. They must be in any imaging laboratory.
That day in class I encountered an error, a misunderstanding. My professor did not want to circumvent consciousness, or argue that perception ends with the physical description of stimulation. He simply assumed that the student was aware of the orientation setting, and tried to give out additional information.
But the spots of Gabor do not only control the primary visual cortex. The properties of the primary processing of visual information are extremely influential on how we envision the brain as a whole. They support our belief that somewhere there will be a neurocode for the sensation of time, space, our position in space, the meaning of words, the beauty of melodies, complex emotions like the pain of social rejection, the ability to judge other people's thoughts, self-awareness, political bias, character traits. These neural circuits may be difficult for us to discern as observers, but we believe that the code is hidden inside, and is ready for us to extract and analyze it, compare it with one of the cognitive concepts, and that the work of consciousness can in principle be ideally compared with the rules governing the work of neurons. If you dream even further, then after we crack this code, it should be possible to build machines capable of processing information as well as us, and indistinguishable from us in this regard.
No one considers this task simple. Even in the case of “simple cells”, reality turns out to be much more complicated than simply assigning the neuron the role of orientation recognition on the one hand, and its connection with sensations on the other. First of all, sensations are not associated with individual neurons, but with the relative number of activities in neurons that prefer different orientations. The connection between sensations and this distribution of activity is not direct. If we consider a vertical strip and tilt our head or body sideways, so that the strip relative to our eyes is not vertical, neurons preferring an oblique position will have to work. But in fact, this strip continues to be vertical for us, judging by the neurons of the primary visual cortex (and we perceive it as vertical). All due to the fact that the vestibular information is associated with information about the orientation and corrects it. Moreover, orientation and space are also interconnected: a sequence of increasingly tilted orientations, dotting the surface of the primary visual cortex, is repeated there many times. This makes it possible to adequately recognize the inclination present in different areas of the visual field. Simple cells process not only this information - but, for example, the total employment of the visual field - and the Gabor spot with a large number of thin strips will be perceived differently than a spot with a small number of broad. Some simple cells will have wide adjustment lines, others will use narrow ones. Some will be embedded in information from the previous level of its processing, in the thalamus , almost additively, while others will use more complex calculations. And in addition to all this, simple cells selectively suppress each other, and in addition, their activity flexibly adapts depending on the data obtained from higher-order areas. Imagine what it would be like if they were not “simple”!
But, despite all the difficulties, the connection between the orientation of the lines and the nervous activity, and between the nervous activity and sensations is quite straightforward. So much so that disruptions in sensations can allow you to make reasonable assumptions about the work of the brain, as is the case with the illusion on the wall of a cafe, where the surrounding visual context affects our sense of orientation - perhaps due to local suppression.
The bricks seem trapezoidal, and the seam is oblique, although in reality the bricks are rectangular, and the seam is parallel to the ground.
Is it possible to expect such direct connections between cognitive processes and nervous activity to be the norm? You can often see how orientation adjustments are described as an introduction to more general ideas about the purpose of the brain — as a prototypical example of how the brain works. Also, research often begins with general cognitive questions (How do we define our car from hundreds of cars in the parking lot? How do we navigate through a busy street? Why are we surprised when a long-term noise source suddenly stops?) And ends at a very small scale of work neurons describing the results of the experiment. For the uninitiated, it may look as though connections with cognitive phenomena are considered so obvious that they do not need to be described further. In fact, this is most likely due to the belief that the space between nerve activity and cognitive processing can be filled in principle, and that gradually, with difficulty, we will finally fill it - as can be seen in the example of orientational adjustment.
However, with the increasing complexity of the phenomena studied by us, the complexity of establishing connections between neurons and reasonable activity also grows very rapidly. Most of the sense of orientation is very convenient for the activity of simple cells in the primary visual cortex, but it would be terribly inadequate to define the learning process simply through the plasticity of synapses. Even if we fully and accurately describe all the activity of neurons, we will need to find a fundamental way to connect them with the mind, and this method almost never appears because we are very closely looking at the nervous tissue.
If you ask yourself whether we rely on mental phenomena on nervous activity, the answer will definitely be “yes”. In this sense, all reasonable activity can be reduced to simple, tangible, consistent building blocks, combined on the basis of a finite number of clear guidelines. But out of this simplicity comes unexpected complexity. In this sense, learning in principle can be described at the nervous level, and we, in principle, can build machines that are aware of everything just as we are. The activity of simple cells that manifests itself when viewing the Gabor spot is a good example of how this can work with a mind function of any complexity.
On the other hand, some important phenomena occur both within and between people. Our sense of identity, for example, is a mixture of personal qualities and how they differ from the qualities of other people. One of my notable properties is that I am a foreigner. This is reflected in the way I process certain information related to the locals. For example, I cannot distinguish certain sounds uttered with a local accent, since I did not grow with them, and their signal fades in my auditory cortex just like the horizontal orientation of those unfortunate kittens. My belonging to foreigners can be discussed in terms of differences in nervous activity, and perhaps even with great accuracy, but is it reasonable to do so? My differences with the local ones will be different if I move from one country to another, and in any country other foreigners may differ from the local in a way that is not related to me in any way. It makes more sense to discuss any such differences at the nervous level as a function of cultural differences, and not as a function of a foreign brain generating a foreign mind. I do not have a foreign brain: I have a brain, and I am a foreigner. In the limit, any symbolization can be regarded as a result of culture, comparable to belonging to foreigners, and therefore it cannot be called a feature of the brain. And for intelligent activities need symbolization.
People who are interested in what happens at these two ends of the spectrum — the reduction of cognitive phenomena to nervous activity, and the refraction of these phenomena through cultural and interpersonal lenses — consider the opposite position to be true, but trivial, without explanatory power. It seems to me that this may be due to the disagreement between the nature of the causal relationships and whether it can develop in one direction or in many. In any case, the relationship between the nervous and cognitive sides of the same coin is full of subtleties.
Studying these subtleties, we can ask whether it is possible to associate any human mental phenomena with exact nerve states that do not vary in time, and therefore, we can always use the nervous state to describe a cognitive result. The answer to this question will be negative. Many nerve states can lead to the same cognitive result (you can solve a math problem based on the sense of numbers, on visualization, on the possibility of verbalization), and various cognitive results can flow from one state (for example, your joyous excitement can flow into euphoria or into anxiety).
But perhaps in this changeability some kind of inherent nervous activity hides, or does this activity go into one or another state depending on what the brain is doing? If we could fully describe this background nervous activity, could we know what kind of mental activity will manifest? Maybe. But, most likely, the properties of consciousness work according to their own rules, which do not exist on the lower level. For example, in the words about what follows from the other, there can be a meaning, and in the words that the nervous picture that led to the first thought gave rise to the nervous picture that led to the second thought, there can be no meaning. Without a description of the thoughts, the connections between the two nerve paintings are not at all obvious. This means that the way the mind is organized may not be the best guide to how the brain is organized - it may be that mind has its own opinion.
Conversely, our assumptions that neural effects simply describe cognitive phenomena are not a given, and we cannot treat the assumptions about the connections between neurons and thoughts superficially. Personally, reaching the end of work on cognitive neuroscience, I try to ask myself if I can now say something new about cognitive phenomena, the studies of which were stated in this work without touching the work of the brain. If I cannot do this, then probably the mind was not the protagonist of this story - he was only the hero of the second plan. This principle helps me to remember that nervous activity is the work of the mind in the same sense that the spot of Gabor is a convolution of a sinusoid with a Gaussian curve: nervous activity explains the work of the mind by unconditionally true concepts, which at the same time are unconditionally limited.
Professor : in this experiment, people look at the cross in the center of the screen, while the Gabor spot is shown to the left or right of the cross ...
Student : I'm sorry, but what is a Gabor spot?
Professor : Ah, well, this is a convolution of a sinusoid with a Gaussian curve.
He smiled at us, not paying attention to the fact that behind him there was an image of Gabor's spot on the screen. He raised his eyebrows expectantly. His whole posture said: "Well, now I understand?"
Student : Uh ...
Professor : No? Here, let me show you.
Still ignoring the presentation screen, he turned to the blackboard. On it, he drew a sine wave, and below it - a Gaussian curve.
“And now you carry out a convolution operation on them!” [ According to some experts in the comments, this function is a product of Gaussians and sinusoids, not a convolution / approx. trans. ]
The student gave up. Perhaps he had some ideas about the mathematical operation of convolution, but there was no necessary intuition. He needed someone to simply point his finger at the right place on the screen: here, this is Gabor's stain.
It could be a story about how I sometimes felt at the place of this student. Or a story about teaching. Or, perhaps, about how absolutely accurate information may seem to us meaningless. But I would like to focus now on the fact that the Gabor spot is more than just a convolution of a sinusoid with a Gaussian curve.
Suppose you have a bunch of kittens that were raised in an environment where there were no landmarks other than vertical ones. Kittens would only see vertical stripes. What happens if in a few months these kittens are released and allowed to interact with the ordinary world around them?
What will happen is that after an uncertain start, they will learn to interact with the outside world. They will begin to study it, play, behave like kittens. But if they come across a long, thin, horizontal object - for example, a black cable, stretched across a white carpet, they will behave as if it does not exist. They will not be frightened if he suddenly goes to their side and will not miss him if he jerks. They will show selective blindness to him, although their eyes will work fine. The root of their problems will be in the brain.
The primary visual cortex — the first of a series of cortical zones processing visual information — is sensitive to the orientation of the lines. There are cells called “simple” (or strip detectors, or face detectors) that are activated in response to different degrees of inclination, depending on the orientation of the faces of the objects observed. If we had walked along the surface of the primary visual cortex, we would have walked slowly from the areas responsible for vertical orientation to areas that are tuned to greater and greater angles of inclination. This means that one group of cells becomes very active when it sees a vertical line, another - if it is tilted slightly, and another one reacts to a horizontal line. Cells “tuned” to a vertical orientation will still respond in response to not exactly vertical lines, just smaller and smaller, depending on the difference between the “preferred” vertical orientation and the visible one.
Sensitivity of a neuron tuned to vertical stripes as a function of strip orientation
In the case of kittens, the cages that should have been responsible for horizontal orientation began to respond to other orientations due to sensory deprivation at a critical period of development. They do not have cells that react to horizontal stimuli, so at this stage of neuroprocessing the visual signal goes out.
Upon learning of this orientation setting, we can begin to understand one of the phenomena related to perception. Namely, if we look at the same oriented strip for a long time, then our ability to judge the inclination of subsequent stripes, which are approximately the same oriented, will decrease for a while, but this effect will be weakened more than by the inclined stripes. Now we know that this is due to the fatigability of the neurons - the more the neuron is activated, the more tired it will be later, and the less it will be able to accurately convey information about the orientation.
How do we know this? Thanks to thousands of experiments with Gabor's spots. Gabor spots are incentives that control early visual activity in a controlled way. They look like a sequence of black and white stripes, and they can be oriented in any way, they can be made well or poorly distinguishable, large or small, rotating or stationary. They must be in any imaging laboratory.
That day in class I encountered an error, a misunderstanding. My professor did not want to circumvent consciousness, or argue that perception ends with the physical description of stimulation. He simply assumed that the student was aware of the orientation setting, and tried to give out additional information.
But the spots of Gabor do not only control the primary visual cortex. The properties of the primary processing of visual information are extremely influential on how we envision the brain as a whole. They support our belief that somewhere there will be a neurocode for the sensation of time, space, our position in space, the meaning of words, the beauty of melodies, complex emotions like the pain of social rejection, the ability to judge other people's thoughts, self-awareness, political bias, character traits. These neural circuits may be difficult for us to discern as observers, but we believe that the code is hidden inside, and is ready for us to extract and analyze it, compare it with one of the cognitive concepts, and that the work of consciousness can in principle be ideally compared with the rules governing the work of neurons. If you dream even further, then after we crack this code, it should be possible to build machines capable of processing information as well as us, and indistinguishable from us in this regard.
No one considers this task simple. Even in the case of “simple cells”, reality turns out to be much more complicated than simply assigning the neuron the role of orientation recognition on the one hand, and its connection with sensations on the other. First of all, sensations are not associated with individual neurons, but with the relative number of activities in neurons that prefer different orientations. The connection between sensations and this distribution of activity is not direct. If we consider a vertical strip and tilt our head or body sideways, so that the strip relative to our eyes is not vertical, neurons preferring an oblique position will have to work. But in fact, this strip continues to be vertical for us, judging by the neurons of the primary visual cortex (and we perceive it as vertical). All due to the fact that the vestibular information is associated with information about the orientation and corrects it. Moreover, orientation and space are also interconnected: a sequence of increasingly tilted orientations, dotting the surface of the primary visual cortex, is repeated there many times. This makes it possible to adequately recognize the inclination present in different areas of the visual field. Simple cells process not only this information - but, for example, the total employment of the visual field - and the Gabor spot with a large number of thin strips will be perceived differently than a spot with a small number of broad. Some simple cells will have wide adjustment lines, others will use narrow ones. Some will be embedded in information from the previous level of its processing, in the thalamus , almost additively, while others will use more complex calculations. And in addition to all this, simple cells selectively suppress each other, and in addition, their activity flexibly adapts depending on the data obtained from higher-order areas. Imagine what it would be like if they were not “simple”!
But, despite all the difficulties, the connection between the orientation of the lines and the nervous activity, and between the nervous activity and sensations is quite straightforward. So much so that disruptions in sensations can allow you to make reasonable assumptions about the work of the brain, as is the case with the illusion on the wall of a cafe, where the surrounding visual context affects our sense of orientation - perhaps due to local suppression.
The bricks seem trapezoidal, and the seam is oblique, although in reality the bricks are rectangular, and the seam is parallel to the ground.
Is it possible to expect such direct connections between cognitive processes and nervous activity to be the norm? You can often see how orientation adjustments are described as an introduction to more general ideas about the purpose of the brain — as a prototypical example of how the brain works. Also, research often begins with general cognitive questions (How do we define our car from hundreds of cars in the parking lot? How do we navigate through a busy street? Why are we surprised when a long-term noise source suddenly stops?) And ends at a very small scale of work neurons describing the results of the experiment. For the uninitiated, it may look as though connections with cognitive phenomena are considered so obvious that they do not need to be described further. In fact, this is most likely due to the belief that the space between nerve activity and cognitive processing can be filled in principle, and that gradually, with difficulty, we will finally fill it - as can be seen in the example of orientational adjustment.
However, with the increasing complexity of the phenomena studied by us, the complexity of establishing connections between neurons and reasonable activity also grows very rapidly. Most of the sense of orientation is very convenient for the activity of simple cells in the primary visual cortex, but it would be terribly inadequate to define the learning process simply through the plasticity of synapses. Even if we fully and accurately describe all the activity of neurons, we will need to find a fundamental way to connect them with the mind, and this method almost never appears because we are very closely looking at the nervous tissue.
If you ask yourself whether we rely on mental phenomena on nervous activity, the answer will definitely be “yes”. In this sense, all reasonable activity can be reduced to simple, tangible, consistent building blocks, combined on the basis of a finite number of clear guidelines. But out of this simplicity comes unexpected complexity. In this sense, learning in principle can be described at the nervous level, and we, in principle, can build machines that are aware of everything just as we are. The activity of simple cells that manifests itself when viewing the Gabor spot is a good example of how this can work with a mind function of any complexity.
On the other hand, some important phenomena occur both within and between people. Our sense of identity, for example, is a mixture of personal qualities and how they differ from the qualities of other people. One of my notable properties is that I am a foreigner. This is reflected in the way I process certain information related to the locals. For example, I cannot distinguish certain sounds uttered with a local accent, since I did not grow with them, and their signal fades in my auditory cortex just like the horizontal orientation of those unfortunate kittens. My belonging to foreigners can be discussed in terms of differences in nervous activity, and perhaps even with great accuracy, but is it reasonable to do so? My differences with the local ones will be different if I move from one country to another, and in any country other foreigners may differ from the local in a way that is not related to me in any way. It makes more sense to discuss any such differences at the nervous level as a function of cultural differences, and not as a function of a foreign brain generating a foreign mind. I do not have a foreign brain: I have a brain, and I am a foreigner. In the limit, any symbolization can be regarded as a result of culture, comparable to belonging to foreigners, and therefore it cannot be called a feature of the brain. And for intelligent activities need symbolization.
People who are interested in what happens at these two ends of the spectrum — the reduction of cognitive phenomena to nervous activity, and the refraction of these phenomena through cultural and interpersonal lenses — consider the opposite position to be true, but trivial, without explanatory power. It seems to me that this may be due to the disagreement between the nature of the causal relationships and whether it can develop in one direction or in many. In any case, the relationship between the nervous and cognitive sides of the same coin is full of subtleties.
Studying these subtleties, we can ask whether it is possible to associate any human mental phenomena with exact nerve states that do not vary in time, and therefore, we can always use the nervous state to describe a cognitive result. The answer to this question will be negative. Many nerve states can lead to the same cognitive result (you can solve a math problem based on the sense of numbers, on visualization, on the possibility of verbalization), and various cognitive results can flow from one state (for example, your joyous excitement can flow into euphoria or into anxiety).
But perhaps in this changeability some kind of inherent nervous activity hides, or does this activity go into one or another state depending on what the brain is doing? If we could fully describe this background nervous activity, could we know what kind of mental activity will manifest? Maybe. But, most likely, the properties of consciousness work according to their own rules, which do not exist on the lower level. For example, in the words about what follows from the other, there can be a meaning, and in the words that the nervous picture that led to the first thought gave rise to the nervous picture that led to the second thought, there can be no meaning. Without a description of the thoughts, the connections between the two nerve paintings are not at all obvious. This means that the way the mind is organized may not be the best guide to how the brain is organized - it may be that mind has its own opinion.
Conversely, our assumptions that neural effects simply describe cognitive phenomena are not a given, and we cannot treat the assumptions about the connections between neurons and thoughts superficially. Personally, reaching the end of work on cognitive neuroscience, I try to ask myself if I can now say something new about cognitive phenomena, the studies of which were stated in this work without touching the work of the brain. If I cannot do this, then probably the mind was not the protagonist of this story - he was only the hero of the second plan. This principle helps me to remember that nervous activity is the work of the mind in the same sense that the spot of Gabor is a convolution of a sinusoid with a Gaussian curve: nervous activity explains the work of the mind by unconditionally true concepts, which at the same time are unconditionally limited.
Source: https://habr.com/ru/post/409639/