User:Amull333/sandbox

Categorical perception is a phenomenon of perception of distinct categories when there is a gradual change in a variable along a continuum. It was originally observed for auditory stimuli but now found to be applicable to other perceptual modalities. Categorical perception paradigms have been used to study perception in several areas, including speech processing, sign language, and facial expression identification.

Contents

 * Identification and discrimination tasks
 * Identification
 * Discrimination
 * The motor theory of speech perception
 * Acquired distinctiveness
 * Categorical Perception and Sign Language
 * The Whorf hypothesis
 * Support
 * Learned CP
 * Learning Cues: Labels and Themes
 * Computational and neural models
 * Brain basis
 * Language-induced
 * Emotion
 * See also
 * References
 * Bibliography

Identification and discrimination tasks
The study of categorical perception often uses experiments involving discrimination and identification tasks in order to categorize participants' perceptions of sounds. Voice onset time (VOT) is measured along a continuum rather than as a binary. English bilabial stops /b/ and /p/ are counterparts of the same type of articulation, yet native speakers are able to distinguish the sounds primarily by where they fall on the VOT continuum. Two sounds with different VOT will be perceived as the same phoneme if they fall on the same side of the boundary. Participants take longer to discriminate between two sounds falling in the same category of VOT than between two on opposite sides of the phoneme boundary, even if the difference in VOT is greater between the two in the same category. This phenomenon defines categorical perception.

Identification
In a categorical perception identification task, participants often must identify stimuli, such as speech sounds, based on the binary ends of the continuum. An experimenter testing the perception of the VOT boundary between /p/ and /b/ may play several sounds falling on various parts of the VOT continuum and ask volunteers whether they hear each sound as /p/ or /b/. This is often called a forced-choice task because participants have to choose between one of the two options based on the presented stimulus. In such experiments, sounds on one side of the boundary are heard almost universally as /p/ and on the other as /b/. Stimuli on or near the boundary take longer to identify and are reported differently by different volunteers, but the sounds are still perceived as either /b/ or /p/, rather than as a third option.

Discrimination
A simple AB discrimination task presents participants with two options and participants must decide if they are identical. Predictions for a discrimination task in an experiment are often based on the preceding identification task. An ideal discrimination experiment would result in volunteers correctly discriminating stimuli that fall on opposite sides of the boundary more often, while discriminating at chance level on the same side of the boundary.

In an ABX discrimination task, volunteers are presented with three stimuli. A and B must be distinct stimuli and volunteers decide which of the two the third stimulus X matches. This discrimination task is much more common than a simple AB task.

The motor theory of speech perception
Main article: Motor theory of speech perception

Early work on categorical perception focused on how people identify and categorize speech sounds by researching English speech sound categories. Based on sound spectrograms, the speech sounds /ba/ and /pa/ are found to lie along an acoustic continuum called "voice-onset-time." Because the two sounds are on the same acoustic continuum, a /ba/ can be artificially morphed into a /pa/ by gradually increasing the voicing parameter.

Alvin Liberman et al. reported that when people listen to sounds that vary along the voicing continuum, they report only hearing /ba/s and /pa/s, even in the ambiguous sound morphs. This effect—in which a perceived quality jumps abruptly from one category to another at a certain point along a continuum, instead of changing gradually—he dubbed "categorical perception" (CP). He suggested that CP was unique to speech, that CP made speech special, and, in what came to be called "the motor theory of speech perception," he suggested that CP's explanation lay in the anatomy of speech production.

According to the motor theory of speech perception, the reason people perceive an abrupt change between /ba/ and /pa/ is that sound perception is influenced by sound production. In this example, voice onset time varies, creating the difference in sound: the "b" in /ba/ is voiced and the "p" in /pa/ is not. Synthetic morphing creates an auditory continuum between the two sound. However, unlike the synthetic morph, vocal cords cannot produce anything in between /ba/ and /pa/. Synthetic stimuli along the continuum will therefore be interpreted as either /ba/ or /pa/, depending on where the sound falls across a participants categorical boundary. A similar CP effect is found with ba/da, which also lie along a continuum acoustically. Vocally, /ba/ is formed with the two lips, /da/ with the tip of the tongue and the alveolar ridge. Our anatomy does not allow any intermediates, meaning that sound perception in this case will be categorical. Contrarily, where production is continuous, perception will be continuous. Further research found vowel categories like a/u were found to be much less binary than ba/pa or ba/da. Despite this, the motor theory of speech production has since been abandoned by most psycholinguists.

Acquired distinctiveness
Under the assumptions of the motor theory of speech, motor production mediates sensory perception, which implies that the CP effect could relate to how we learn to speak. Eimas et al. (1971), however, found that infants can already distinguish auditory categories before they can speak. Likewise, Kuhl (1987) found that chinchillas also have speech CP, even though they never learn to speak, which implies the speech production theory might be false. Lane (1965) went on to show that CP effects can be induced by learning, and that CP is also applicable in non-speech contexts. He demonstrated this using visual stimuli, which means that humans can distinguish categories along a continuum regardless of whether we can produce the stimulus ourselves. He concluded that categorical perception is not limited to sound perception, which has been corroborated by other researchers since then

Categorical Perception and Sign Language
Language-based categorical perception is not unique to spoken languages. Sign language speakers display categorical perception for signed words. Like spoken words, signed words differ from one another in visual “phonemes”, namely speed, handshape, location, and palm orientation. Some words and letters vary based on only one of these visual phonemes, like the letter U and the letter V or the words “sorry” and “please”. One study observed categorical perception in relation to the distance between the index and middle fingers, which distinguishes the signed letter U from the letter V. The study found that sign language speakers display a categorical boundary that determines whether they perceive the signed letter to be U or V.

The Whorf hypothesis
According to the Sapir–Whorf hypothesis, also known as linguistic relativity, language affects the way that people perceive the world. For example, colors are perceived categorically only because they happen to be named categorically: Our subdivisions of the spectrum are arbitrary, learned, and vary across cultures and languages. Other researchers have suggested instead that colors are perceived the same way regardless of the words we use. Berlin & Kay (1969) suggest that we all see colors within categories as more alike, with imprecise boundaries based on the color spectrum, rather than the color words we use. Regier and Kay (2009) challenged this idea with a study demonstrating that linguistic categories do affect categorical perception, but primarily in the right-eye visual field, and that this effect is eliminated with a concurrent verbal interference task. This could be due to interference from the language centers of the brain, typically localized in the left hemisphere.

Support
The Sapir-Whorf hypothesis is supported by examples of speakers of one language perceiving categories differently from speakers of another language.

One study reported evidence that linguistic categories affect categorical perception primarily in the right-eye visual field. The right-eye visual field is controlled by the left hemisphere of the brain, which also controls language faculties. A second study presented evidence that in color discrimination tasks, native English speakers discriminated easier between color stimuli across a determined blue-green boundary than within the same side, but did not have the same boundaries when given the same task with Berinmo colors "nol" and "wor". Berinmo speakers performed oppositely, doing better on the task using Berinmo color words. Similar patterns of language-based variation in color perception have been supported by other studies since.

Learned CP
Several experiments have demonstrated that categorical boundaries can be altered by learning. Nativist theorists have proposed that categorical boundaries are innate, however recent demonstrations show that boundaries can be modified, gained, or even lost as a result of learning.

Learning alters the perception of a given stimulus based on prior experience or knowledge. Learned categorical perception can be divided into different processes: categorical expansion and categorical compression. A categorical expansion occurs when the classifications and boundaries for a category become broader, encompassing a larger set of stimuli. This can be achieved by comparing stimuli from one end of the continuum to the other (e.g. comparing a bright red to a bright blue), which is called between-group comparison. A categorical compression effect narrows category boundaries to reduce the set of included stimuli, which can be done by comparing points close together on the continuum. This is called within-group comparisons and it creates more distinct and rigid categorical boundaries.

Likewise, it is still possible to distinguish between stimuli, even if they fall on the same side of the boundary. We can still perceive the differences, even though within-category differences sound/look much smaller than the between-category differences. This is true even when the difference between of the stimuli (i.e., changes in voicing or wavelength) is quantitatively the same as it would be across the boundary.

Another method of comparison is to look at both supervised and unsupervised group comparisons. Supervised groups are those for which categories have been provided, meaning that the category has been defined previously or given a label; unsupervised groups are groups for which categories are created, meaning that the categories will be defined as needed and are not labeled.

Learning Cues: Labels and Themes
Cues used in learned categorical perception can foster easier recall and access of prior knowledge in the process of learning and using categories. An item in a category can be easier to recall if the category has a cue for the memory. Learning categories and strengthening categorical perception can be improved by identifying themes within categories, adding verbal labels to the categories, making categories smaller, and by targeting similar defining features within categories.

Themes consist of unique traits that help identify how to categorize the stimulus. For example, shape perception uses angles to help categorize shapes. The number of angles and their size cue different categories, like square or triangle. Three angles would cue a triangle, whereas four would cue a square. Opposing themes help distinguish categories too, like circles which have no angles. It is easier to categorize circles from other shapes because circles oppose the theme of angles.

Like themes, labels are also important to learned categorical perception. Labels are category titles that focus on similarities, which helps categorical processing. The strength of a label can be determined by three factors: analysis of affective (emotional) strength, permeability of boundaries, and a judgment of discreteness. Sometimes labels are already designated, and some need to be created, like in the case of supervised/unsupervised categories. The source of the label doesn’t appear to matter, however there is a positive correlation between strength of the label, which is based on the three listed factors, and the degree to which the label affects perception. This means that stronger labels have more influence over perception.

Several brain structures are well-suited for learned categorical perception. The prefrontal cortex is involved in forming strong categorical representations. The inferotemporal cortex codes for different object categories and discriminates along diagnostic category dimensions, which distinguish categorical boundaries.

Learned categorical perception occurs not only in humans but has also been demonstrated in animal species as well. Studies have targeted categorical perception using humans, monkeys, rodents, birds, frogs. The studies focus primarily on learning the boundaries of categories, and where categorical inclusion begins and ends. These studies also support the hypothesis that categorical perception has a learned component.

Computational and neural models
Computational modeling (Tijsseling & Harnad 1997; Damper & Harnad 2000) has shown that many types of category-learning mechanisms (e.g. both back-propagation and competitive networks) display CP-like effects. In back-propagation networks, also known as nets, the hidden-unit activation patterns that represent an input build up within-category compression and between-category separation as they learn. Other kinds of nets display similar learning effects.

CP seems to facilitate efficiency by reducing the amount of information within a category. Different inputs are compressed onto similar internal representations if they all generate the same output, and if they generate different outputs, they become more separate. The network's "bias" is what filters inputs onto their correct output category. The nets accomplish this by selectively detecting the invariant features that are shared by the members of the same category and that distinguish them from members of different categories. This skill is guided by error-correcting feedback, and the nets learn to ignore other variations as irrelevant to the categorization

Brain basis
Neural data provide correlates of CP and of learning. Differences between event-related potentials were found to be correlated with differences in the perceived category of the stimulus the subject viewed. Neural imaging studies have shown that these effects are localized in specific brain regions in subjects who have successfully learned the category, and are absent in subjects who have not.

Tasks involving categorical perception activate the left prefrontal cortex. The left prefrontal cortex shows activation when perceiving speech units, while this is not the case in posterior areas. Earlier speech sound processing activates elsewhere, such as the left superior temporal gyrus.

Language-induced
Both innate and learned CP are sensorimotor effects: The compression/separation biases are sensorimotor biases, and likely originated from experience. The neural net I/O models are also compatible with this fact: I/O biases derive from I/O history. However, some categories are abstract, or impossible to experience.

There are some neural net simulation results suggesting that once a set of category names has been "grounded" through direct sensorimotor experience, the names can be combined into Boolean combinations (man = male & human) based on language alone, which can expand into still higher-order combinations (bachelor = unmarried & man) which generate even higher-order, more abstract categories. “Bachelor” inherits the compression/separation of “unmarried” and “man”, and also adds a layer of separation/compression of its own.

So far, only learned and innate sensorimotor CP have been demonstrated in humans. In order to support the previously described Whorf Hypothesis, further research will have to address whether human perception can be altered by changing concepts, rather than just vocabulary.

Emotion
Emotions and their relation to categorical perception are often studied using facial expressions. Facial expressions contain large amounts of valuable information, which is important because emotions fall into very different and distinct categories that are important to understand for successful social interaction.

Emotions are divided into discrete categories which entail separate and distinct sets of reactions, consequences, and expressions. There are six basic emotions that are generally considered universal regardless of age, gender, race, country, or culture. These six basic emotions are considered categorically distinct: happiness, disgust, sadness, surprise, anger, and fear. Of these, happiness is the most easily identified, even though categorical perception does not require lexical categories. According to the discrete emotions approach, people experience one emotion at a time, meaning though facial expressions are on a continuum, only one emotion can be expressed on the face at a time.

Emotion perception based on facial expressions reveals slight gender differences in the definition and boundaries of the six expression categories. Anger is perceived faster and more easily when it is displayed by males. However, the same effects are seen for happiness when presented by women. These effects may be because traits related to the two emotions (anger and happiness) are more closely associated with other features typical to masculine and feminine faces respectively.

Categorical perception for facial expressions is not language dependent. Before speech processing in infants, they can distinguish emotional responses. Additionally, people from language groups that may not have a label for a specific emotion still show categorical perception for distinct facial expression traits (e.g. in the case of disgust). Categorical perception of emotion has also been studied using eye tracking, which showed an implicit response to differences in facial expressions without requiring a language-based response.