The Role Of Neurons In Humans’ Responding To The Faces, Scenes And Objects
It has been wondered that how we recognize stimulus around us. Brain imagining techniques have been used to determine neurons in which areas of the human brain respond to the faces, scenes and objects. Nancy Kanwisher and coworkers (1997) found that neurons in Fusiform Face Area (FFA) especially activated for faces, houses, household objects and hands via using fMRI. Furthermore, additional evidence comes from prosopagnosia which is difficulty in recognizing of familiar faces caused by damage to the temporal lobe. All in all, it can be said that neurons in Medial Temporal Lobe (MTL) is responsible for being activated to the image of faces, animals, objects or scenes.
One theory suggests that a single neuron that would only respond to specific stimuli. This idea first proposed by Jerry Letvin around 1969. This theory is known as grandmother cell theory or localist approach in psychology. The most important study about this theory has done by Quiroga and his colleges in 2005. Eight patients with pharmacologically intractable epilepsy participated to this study. They were shown different pictures of persons, animal, objects and landmark buildings. As a result they found that the a single unit in the left posterior hippocampus fired to all pictures of the Jennifer Aniston, but not other pictures at all or weakly fired to them. Another single unit located in the right anterior hippocampus was activated by pictures of Halle Berry as well as by a drawing of her, to Catwomen from Batman and the word “Halle Berry”. “This highlights the fact that this cell does not code high-level perceptual information, but rather, a memory (or conceptual information) about the person.”
According to Quiroga, the MTL and hippocampus is not responsible for recognizing objects. For instance, in the case of H.M. who had hippocampus removal could recognize objects so, what matters is remembering. “Despite these findings, Quiroga et al. (2005) explicitly rejected the grandmother cell hypothesis and concluded that each cell must represent more than one class of image (otherwise the cells would not be found in the first place).” As an advantage, this specificity coding would be energy saving. From evolutionary perspective, the situation that requires minimum amount of energy would be favored all the times. Also, we would be able to know what person is thinking easily due to one neuron is related to specific person or object.
Hubel and Wiesel (1962) first proposed the hierarchical organization of visual cortex. They differentiated the simple and complex cells. “Both these cells respond to oriented edges and gratings, however, the complex cells show a degree of spatial invariance.” They argued that inputs of complex cells come from a numbers of simple cells. This hierarchical model supports the idea of grandmother cell theory. Moreover, in the Quiroga’s experiment (2005) some selective neurons respond more than one person. For example the neuron that fired to the picture of Jenifer Aniston also fired to the picture of Lisa Kudrow. This is because they both took role in TV-series ‘Friends’. “It is important to note that this overlap only occurs in entities that are related to each other.” So, you should have specific interest to certain things. For example if you are interested in Harry Potter movie series, the neuron that is activated to the picture of Daniel Jacob Radcliffe would also be activated to the picture of Emma Watson. “The response of these MTL neurons to visual stimuli appears to depend, therefore, on a particular person’s past experiences.”
Empirical challenge to the grandmother cell theory comes from Waydo et al. (2006). They analyze responses of 1,425 MTL neurons in humans via using the data from Quian Quiroga and Kreiman (2010). The activation of neurons shows differences to the images. “Despite the selectivity of some neurons, they concluded that each image activated approximately 5 million neurons within MTL (a measure of population sparseness), and each neuron is involved in representing between 50 and 150 objects (a measure of lifetime sparseness).” They rejected grandmother cell theory with the accordance of their findings. The question comes up to the mind that if we have one neuron for each person, then what would happen if that neuron die or stop functioning? We would lose our ability to recognize this specific person or object suddenly. It seems very unlikely and not such findings exist up to now. Another problem arises from the structure of neurons activations. One neuron cannot activate itself. There has to be neuronal connections once the input received. This means it is not likely to happen as suggested in grandmother cell theory. Localist model in psychology are often rejected because it may not be possible to have singe neuron for each person or objects. There are too many things to recognize in the world. An alternative idea is specificity coding which involves more than one neuron to represent a stimulus. Distributed coding involves large number of neurons for representation of stimuli. It has an advantage because large number of neurons can create so many different patterns.
According to Quian Quiroga and Kreiman (2010), it does not make sense to refuse distributed coding when it is found that single neuron respond to multiple things. Additionally, sparce coding is representation of an object by small group of neurons. In one study that has applied to the participants with epilepsy, found neuron that only respond to picture of Steve Carell and not others. However, it revealed that they only record these neurons for 30 minutes then, more time would be helpful to find another faces that cause this neuron to fire. So, Quiroga and coworkers (2008) suggested that their neuron might be the example of sparce coding. Another study found that coding objects in visual system, tones in auditory system and odors in olfactory system might consist of pattern of small number of neurons, as suggested by sparce coding. A comparison of distributed coding and grandmother cell theory comes from Rolls et al.’s experiment in 1997. The way of to compare these two approaches is to determine if a single neuron is capable of encoding information for high-level vision (exp. face recognition) or multiple neuron would require to do so. “For instance, Rolls et al. (1997) identified a set of 14 face selective neurons in the temporal cortex of the macaque on the basis of the criterion that they responded at least twice as much to an optimal face stimulus, compared to an optimal nonface stimulus (out of a set of 68 images).”
All in all, the neurons that have been found in MTL would probably be grandmother-like cells but not actually it. The reason behind it the fact that MTL is related to memory not the presentation of visual input. Also some of that neuron responds more than one stimulus. That neuron might involve in representation of concepts of objects or person instead of specific stimuli. According to Meijer, the highly selective neurons that have been found in MTL are not grandmother cell due to responding more than one stimulus. The fact that the neuron that fired specifically all the pictures of Jennifer Aniston also fired to Lisa Kudrow means we associate them in our mind. So it is all about retrieval these concepts from our memory.
As a conclusion, grandmother cell theory suggests that we have one neuron for each person or object. There are few arguments that support the idea of grandmother cell theory. But generally neuropsychologists have been tried to falsify this theory. Lots of studies have done via using brain imagining techniques. Opposing views are distributed and sparce coding. According to these views, it is not one neuron but pattern different neurons activated to the specific stimuli. Distributed coding involves mass number of neurons but sparce coding involves small number of neurons for coding. There are some findings that support these models too. So it can be said that grandmother cell theory is generally not acceptable among psychologists.
