It is interesting how humans and primates have unique abilities to recognize faces within a fraction of a second. Generally, it is observed that primates have a specific area known as the macaque middle face patch for recognizing faces. However, it is not yet known the principles used by cells to analyze faces. In humans, neurons found in the medial temporal lobe (MTL) are known to transmit the presence of a face or an object and amazingly, these neurons have been shown to focus on the firing of letter strings with names. In macaque monkeys, for instance, it has been determined that cells found in the macaque middle face patch are adjusted to the geometry of facial features. This essay focuses on neurons and faces recognition within a fraction of a second.
It is observed that primate societies depend on face recognition (Yovel and Freiwald 10). For several years, scientists have researched face recognition principles in macaque monkeys and humans, and in both species, they have identified several brains dedicated to face processing, with advanced details of functional properties (Yovel and Freiwald 10). According to Yovel and Freiwald (10), both species have well-organized systems with numerous face-selective cortical locations in spatial arrays and with specific functional activities. This implies that they have hierarchical and parallel means of processing information.
While studies have acknowledged a robust, advanced representation of facial recognition, they still note that how neurons achieve such outcomes in the human brain remains a mystery (Quiroga, Reddy and Kreiman 1102). In monkeys, Quiroga et al. (1102) pointed out that neurons located in the upper section of the ventral visual path react to intricate images, including objects and faces, and depict show some levels of consistency t metric features such as position, stimulus size and angle of observation. In addition, the researchers have previously demonstrated that neurons found in the human medial temporal lobe (MTL) fire discerningly at images of faces or objects (Quiroga et al. 1102). When Quiroga et al. (1102) focused on a certain category of MTL neurons that were discerningly activated by firing upon various images of a specific object, individual or landmarks or even on a string of letters bearing names, they noted that the outcomes were displayed invariant, thin and explicit code. These codes could be critical for the changes of complex visual images into long-term and more abstract memories.
Studies based on pictures such as the Sydney Opera and Halle Berry dressed as Catwoman alongside letter strings had different firing rates. These experiments depicted extreme cases of an abstract representation of images of a given object, as well as corresponding letter strings with names of the objects presented. In some instances, selective responses to an object and its letter string with the name were noted.
It was determined that neuronal responses could not be attached to any specific movement of an object because the noted selective responses were triggered at nearly 300 ms after the picture onset while major movements were noted at one second or after. It was noted that neuronal responses were extremely selective. At the same time, some of the responses were observed between 300 ms and 600 ms. The noted interval was associated with a latency of event-associated responses that relate with the recognition of stimuli found in the scalp electroencephalogram, specifically the P300. In some cases, studies have supported the creation of P300 in the hippocampal formation and amygdala, which also reflected outcomes noted by Quiroga et al. (1102). Researchers have therefore focused on common characteristics that are responsible for neuron activation. On this note, it is observed that the extent of variation on specific pictures of a given individual, including letter strings, photographs, or scenes among others, depends on the selectivity of cells. Thus, it is most improbable that the extent of invariance may be attributed or described through a simple set of determining features common among images. Quiroga et al.’s (1102) data support an abstract representation of the identity of the individual or object used in the study. The availability of high-level visual responses in (MTL) structures is normally associated with a functionally related system specialized for declarative, long-term memory creation and consolidation (Suzuki 657). This assumption is supported based on the following observations. First, there are recognized anatomical links between the advanced stages of the visual ladder in the ventral pathway and the MTL. Second, the well-described reactivity of the cortical stages has been noted to lead into the MTL on the objects, faces, and other spatial scenes. Finally, it is noted that any visual element that may be critically recalled later must be represented within the hippocampal system (Quiroga et al. 1106). The observation was noted as true even among patients who had lost MTL but displayed deficiency in image perceptions.
It has been observed that neurons found in the MTL could play a critical part in learning relations between abstract representations. Therefore, it is concluded the notable invariant responses perhaps emanate from observing extremely different images, letter strings with names, or any other visual stimuli related to a given image. However, all neuroscience studies have noted that how neurons process diverse percepts is intriguing and, therefore, is yet to be understood. On this note, two critical hypotheses based on representations of highly selective components such as grandmother, Gnostic and cardinal neurons have been put forward to explain neuron processing of images. They are considered as processes that depend on a larger representation of an extremely wide distribution of neurons. Further, it is observed that face recognition would need activation of several cells simultaneously, and, therefore, every cell is expected to react to several images with the same basic features. This situation is not similar to the observed sparse firing because many MTL cells do not react to the large majority of pictures observed. In addition, cells fire upon a given object or individual explicitly to ensure that the object present, in general, can be significantly decoded on an extremely low number of neutrons. This observation, however, is not meant to depict that available single neurons code entirely for certain percepts for many reasons. First, neurons respond differently to images of more than a single object or individual. Second, the recording period is generally restricted and, therefore, only a small fraction of stimulus space may be explored. Finally, every cell may represent more than a single class of images because neurons can discover other pictures within a short period. Nevertheless, it is noted that the division of MTL cells remains selective during activation by various images of individuals, objects, scenes, or animals. This is different from a fully distributed group code and depicts a thin, precise and invariant encoding of visual images in the MTL. Consequently, the noted abstract representation, as opposed to the metric representation in the initial stages of the visual pathway, could be imperative for storage that aids long-term memories. It also noted that other factors such as emotional reactions toward some pictures or objects could significantly affect the neuronal during image processing. The actions of hippocampal place cells affect the reaction of neurons, for instance, in rodents, which only fire if the rodent passes through a given spatial place in which the specific location field is identified autonomously of sensory signals.
Quiroga et al. (1107) recently noted that place cells have also been discovered in the human hippocampus. These cells have also been shown to be extremely selective in response because they depict low baseline activities.
In short, Quiroga et al. have stressed that the neurons in MTL fire upon the presentation of letter strings with names, objects, or individuals selectively because they are “selectively activated by strikingly different pictures” (1107). This observation has been supported by other studies (Roy 300). On this note, it is expected that other primates could also possess the same characteristics.
To enhance understanding of how primates recognize faces, other researchers have focused on studying various specialized face areas (Freiwald, Tsao and Livingstone 1187). One critical area that many studies have focused on is the macaque middle face patch. It is generally composed of cells that are dedicated to selective recognition of faces. According to Freiwald et al. (1187), the middle face patch neurons are responsible for identifying and classifying faces through a method that is both holistic and partial. These cells have abilities to detect a specific collection of face parts. In addition, the cells are normally adjusted toward the facial features, and in most cases, cell tuning is ramped alongside one-to-one mapping of element level to the rate of firing. It was also observed that the tuning amplitude could be influenced by the availability of a full, upright face and features, which are perceived based on their outlooks consisting of a full, upright face. Therefore, they concluded that cells in the middle face patch were responsible for “encoding axes of a face space specialized for whole, upright faces” (Freiwald et al. 1187).
Experiments have revealed new coding rules that could assist in understanding the extraction of intricate forms in the inferotemporal cortex. In this regard, the cells found in the middle face patch have been found to detect a diverse range of faces based on their robust responses to real and cartoon faces relative to objects (Freiwald et al. 1187). While the mechanisms of cell operations are not known, it is believed that various cells use different methods for face recognition. In fact, it is noted that the macaque brain has a subset of certain regions that depict stronger Functional magnetic resonance imaging (fMRI) activation to faces relative to other classes of objects (Tsao 109). The fMRI was used to understand stimulus space based on the cells’ stimulus preferences. In the case of macaque monkeys, through the fMRI, it was established that the cortical part located in the temporal lobe was prone to activation much more by face relative to other objects without faces. Subsequent studies demonstrated that the middle face patch was majorly made up of face-selective cells (Freiwald et al. 1187). Thus, by focusing on this area, one can understand the most advanced mechanisms used in form coding for a similar constellation of cells, which are selective for a specific type of image. The researchers also established that the detection process was not utterly holistic because the cell responded to all faces. Just like in other studies, Freiwald et al. (1194) established that various cells were selective for various face parts and associations between parts while a similar cell could respond optimally to diverse constellations of face parts. These findings show that no single approach is effective for detecting the type of a face in the middle face patch.
Cells are specialized in the identification of faces. That is, some cells are more tuned to various categories of facial features. The middle face patch cells were found across the face with three prominent characteristics. First, axes of the space were more inclined to basic face elements and not the whole. Second, the size of the space is narrowed relative to the physical face space and the feature population of focus based on the face and eye layout structure. Finally, the point in the space was coded mainly by the firing rates of cells.
The essay has focused on neurons and face recognition within a fraction of a second in humans and primates. The principles of face recognition remain unknown. It was also established that the MTL firing upon presentation of letter strings with names, objects, or individuals was selective due to selective activation. The cells located in the middle face patch were specifically specialized for selective face recognition than other cells.
References
Freiwald, Winrich A, Doris Y Tsao and Margaret S Livingstone. “A face feature space in the macaque temporal lobe.” Nature Neuroscience 12.9 (2009): 1187-1197. Print.
Quiroga, R. Quian, L. Reddy, Kreiman G, C. Koch and I. Fried. “Invariant visual representation by single neurons in the human brain.” Nature 435 (2005): 1102- 1107. Print.
Roy, Asim. “An extension of the localist representation theory: grandmother cells are also widely used in the brain.” Frontiers in Psychology 4 (2013): 300. Print.
Suzuki, Wendy A. “Perception and the Medial Temporal Lobe: Evaluating the Current Evidence.” Neuron 61.5 (2009): 657–666. Print.
Tsao, Doris. “The Macaque Face Patch System: A Window into Object Representation.” Cold Spring Harbor Symposia on Quantitative Biology 79 (2015): 109-114. Print.
Yovel, Galit and Winrich A. Freiwald. “Face recognition systems in monkey and human: are they the same thing?” F1000Prime Report 5 (2013): 10. Print.