Categorisation Flashcards
what is categorisation?
Categorisation is when observers respond in an equivalent manner to discriminably different stimuli. In other words, we know that things are different yet we treat them as if they are the same. The early literature thought that categorisation occurred relatively late in development. However, recent literature suggests that categorisation has an early onset with even newborns displaying primitive categorisation abilities
why is categorisation important for development?
Categorisation is an important skill for infants as it facilitates processing, learning and the organisation of knowledge. The way in which we group items together or categorize them determines how we learn about the relationship between objects and how we generalise these relationships to novel items. Categories are especially valuable in infancy and early childhood when many new objects, events and people are encountered because without the ability to categorize children would have to learn and respond anew to each novel entity they experience. Not only do categories allow for us to process information more efficiently but they also allow for knowledge of new exemplars for example different breeds or dogs that all come under the dog category
what is habituation?
1.Habituation/novelty preference: Technique used to assess whether infants can discriminate stimuli. This involves showing the infant one type of stimulus repeatedly whilst infants’ looking is monitored. As the infant looks at the stimulus they encode it, but the more it is shown, the less infants look. This decline in looking following repeated presentation of the stimulus is called ‘habituation’. A novel stimulus is then shown, and infant looking to this stimulus is monitored. Infants may start looking again when the novel stimulus is shown, or they may continue to not look. If the novel stimulus reinstates infants’ looking, this is called ‘dis-habituation’. If there is dis-habituation it means that the infant must be able to tell the difference between the original and the novel stimuli. There are variants of this method, including the ‘novelty preference’ method, that are based on the same principles.
Franklin and Davies:
Franklin and Davies used the method of novelty preference which follows habituation and is the process in which a baby looks more at a new stimulus than a familiar one. This tells the researcher that the baby not only recognises the old picture but can tell the difference between the old and new one. Using this technique Franklin and Davies found that four month olds seem to perceive colour categorically. They started with a familiarisation phase whereby infants were repeatedly shown the same colour in both windows. They then followed with a test phase whereby the infants were shown a new colour in one of the windows. This colour was either from the same or a different category as the previous colour. The infants looking time was measured. Their results found that if the novel colour comes from the same category as in the familiarisation phase then the baby will show signs of habituation. Whereas if the colour is from a new category colour then there is a noticeable change where the attention of the infant is caught and looking time is increased. Thus suggesting that four month olds can categorise colour.
Quinn:
This technique has also been employed by Quinn et al., to observe whether four month olds can form category representations of cats that excludes dogs. Quinn used realistic photographic exemplars of cats, representing a variety of breeds, stances, colours and hair lengths. They familiarised the infant to the cat category and then in the test phase they displayed two new animals – one is a cat but one they have not previously seen and the other is a dog. They found that the novel category (dog) was attended to 65% of the time, whereas the same category (cat) was attended to 35% of the time.
To further examine this, Quinn wanted to make sure that they were actually discriminating between cats and treating them as if they were the same. Infants were familiarised to one cat and were then shown a novel cat and found the same results as the previous experiment. This suggests that they can indeed tell the difference between cats but treat them equally when they are shown the two animal categories.
High Amplitude sucking:
Another variation of habituation is the high amplitude sucking technique; from this we have learnt that infants can categorise speech sounds. The method involves a baby sucking on a dummy, and when they suck it, it will produce the sound, therefore if they suck more – more sound will be produced. There is first a baseline procedure where the researcher measures how much the baby will suck the dummy without the sound, then in another condition they create habituation where the baby will slowly decrease their sucking as they become bored of hearing the same sound. In the test phase a new sound will be produced either from the same category as the previous sound or from a new category. Results from Eimas have shown that when they suck again and a new sound is produced, they will suck more, but only if it is from a different category. This has shown that four month olds can discriminate between /ba/ and /pa/. It has also discovered that infants can make more phonetic discriminations than adults.
conditioned leg kicking
The conditioned leg kicking technique, is based on operant conditioning. Infants are placed face up in a cot with a mobile overhanging. During the reinforcement phase, the overhanging mobile is attached to the infant’s ankle via a ribbon, such that kicks activate the mobile. During non-reinforcement, the mobile remains in view but the ribbon is disconnected so that leg movements do not cause mobile movements. Each session begins with a two minute non-reinforcement phases during which a base-rate of the infants kicking is measured. During the reinforcement phases the infants learn to associate leg kicking with the movement of the overhanging mobile. In categorisation tasks, infants are familiarised with two or three different mobiles and then tested with a novel mobile from the familiar category or a novel mobile from a novel category. The degree to which the conditioned leg kicking is transferred to the novel mobile from the familiar category, but not the novel member of the novel category is used as an index of categorisation.
preferential looking
technique used to assess whether 0-12 month old infants can discriminate two stimuli, and have a bias to looking longer at one stimulus than the other. Stimuli are presented to the left and right of infants’ central point of fixation, and infant looking to each stimulus is monitored. There are various measures that can be coded: total looking time at each stimulus; which stimulus was looked at first; number of looks to each stimulus. Pairs of stimuli are usually shown on multiple trials. If the infant looks reliably longer at one stimulus than another, then they must be able to tell the difference between the stimuli. Note, that a preference to look longer at a stimulus does not necessarily mean that the infant actually ‘likes’ that stimulus more – they may simply find it more interesting or salient. Previous studies have shown that even newborns, as young as nine minutes old demonstrate preferential tracking of faces over non-face stimuli (Goren et al.,), exhibiting a preference for both real and schematic faces. These results suggest that face processing may have a prenatal component
sequential touching
is a technique that has been used with older infants at around 12-30 months. This procedure involves presenting infants with a number of exemplars from two categories simultaneously. The exemplars are small, three dimensional toy models that are places before the infant in a random arrangement. Categorisation is inferred if the infant touches exemplars from one category in a sequence before touching the members of other categories (Mandler).
generalised imitation
is used to test more sophisticated categories in 14-30 month old infants. For example, do they have a category of animals such that different animals do different things. In this procedure, an infant is presented with a small model of a real world object and an experimenter then models an action appropriate for that object. Categorisation is inferred if the infant generalises the action to other members of the same category but not to members of other categories (Mandler).
Clifford:
Clifford et al., used a visual oddball task whereby infants saw a display of coloured faces. For 60% of the trials the faces were standard (green), for 20% of the trials they were green but a different type of green (deviant-within category) and for another 20% they were blue (deviant-between category). Infants were shown the faces and their brain activity in response to each of the faces was recorded. They expected that there would be more brain activity in response to the deviant faces when they are from a different category than from the same category. Their results confirmed their hypothesis. They found a dip in brain activity when the colour is from a different colour category – thus the brain was reacting more strongly in response to a novel stimulus. This is a fairly automatic response, thus when babies are categorising they are not necessarily aware that they are doing so.
what kind of categories can infants represent?
Through the use of the different methods described above infants have been shown capable of representing a variety of categories at different levels of inclusiveness for example at the superordinate (animal), basic (cat) and subordinate levels (siamese, tabby). For example: Visual preference research has found that 3-4 month olds can categorise at a basic level of inclusiveness e.g. infants can form categories of cats that excludes dogs.
Object examining studies have found that infants can begin to categorise on the global level of category distinctions.
category formation vs category possession
There is an issue surrounding the research of categorisation representation in infants as to whether or not they are formed online e.g. during the experiment (category formation) or whether the experiment is simply tapping into category representations that were formed on the basis of real life experiences before the experiment began (category possession).
dual process account:
A dual process account for thinking about category representations of infants begins with the idea that seeing is not the same as thinking. This view holds that category representations formed on the basis of perceptual attributes are merely perceptual schemas that define what a group of things look like but that do not contain they content required to define the meaning of something. Thus, true category representations are based on image schemas that are categories based on meaning. The dual process theory thus suggests that infants possess both perceptual schemas that can be used for identifying entities and image schemas that can be used for conceptualising entities.
single process account:
Another view, argues that category representations of infants develop gradually, through a process of quantitative enrichment. In this single process view infants develop category representations for an animal by encountering various animals over time and joining together into a common representation of perceived attributes.
correlated features theory:
Rosch argued that categorisation is highly determined because objects in the world do not appear to observers as unstructured sets of equally likely occurring attributes. Rather the world is structured so that objects categories are marked by bundles of correlated attributes. For example, birds fall into one grouping because they have feathers, beaks and the ability to chirp. Whereas dogs fall into another grouping because they have fur, tails and the ability to bark.
Younger found that infants were able to detect correlations among feature categories that assist them in forming natural object categories. Younger tested 10 month olds using habituation procedure. Infants were initially exposed to as many as 12 different animals and 2 features were correlated within a set of habituation stimuli. For example, animals with feathered tails had ears and animals with furry tails had antlers. The infants sensitivity to the pattern of correlation existing within the habituation set was demonstrated by their responses to new stimuli following habituation. They found that infants as young as 10 months generalised their habituation to a new animal that preserved the experienced pattern of correlation but showed an increase in looking to a stimulus that contained a novel combination of the same features that violated their expectations. This also demonstrates that it is implicit and they are not aware they are making these distinctions.
feature distributions theory
Feature distributions: infants also use feature distributions to assist with categorisation and there is little disagreement in the literature that perceptual factors influence both early and mature generalisation. French et al., found that 3-4 month infants are presented with images of dogs and cats, under some conditions infants form a category representation for cat that excludes dogs and a category representation for dog that includes cats. It has been shown that this pattern of categorization arises because the distribution of perceptual features in cats (e.g., length of head and ears, separation between eyes and ears, etc.).
They measured the nose length of all of the cats and dogs that they were familiar too. In the first condition infants were able to discriminate between cats and dogs solely on the basis of their nose length – as there was no nose length of a dog that was similar to that of a cat. What they found was that when the features of the cats and dogs did not overlap the infants could form categories and they did this with a range of different features e.g. tail length, leg length etc. The more separate the feature distributions were the better the infants were at categorizing.
However, in real life, there are dogs with very little noses the same as those of cats so they also tested them on stimulus with feature distributions that did overlap, thus the nose length is not a reliable way of discriminating between the two categories and the infants cannot categorize. This tells us that infants are actually doing something very powerful, they are looking at certain features and working out the probability of those features occurring across multiple features and then using that as a way of dividing up the stimuli. The results also suggest that the infants are generating the cat and dog categories online and in response to the particular exemplars encountered during familiarization. Changing the distribution of features in the familiarization exemplars changes the pattern of novel category preferences observed in the infants. Thus they are merely going by how the pictures look and not that they have concepts – domain general learning.
features theory
Spencer et al., established that infants focus on certain features when identifying category membership. Spencer et al., carried out experiments that were conducted with infant and adult participants in an effort to determine the perceptual cues that are used to categorically differentiate between two common animal species, cats and dogs. The stimuli were photographic exemplars of cats, dogs and cat–dog hybrids (i.e. cat head attached to dog body and dog head attached to cat body). Experiments 1 and 2 utilized the familiarization/novelty-preference procedure and showed that 4-month-old infants relied on head/face information to categorically differentiate between cats and dogs under conditions of short exposure duration. Thus, information from the head region provided infants with the necessary and sufficient basis to form a representation for cats that excludes dogs and vice versa. Therefore, infants were focusing on certain features of the animals in order to conclude their category membership.
prototype formation
Another way that infants can categorize is through prototype formation. A prototype is the average of previously encountered exemplars – thus the most representative of the stimuli. To test whether 6 month olds can form prototypes of faces, Rubenstein et al., showed infants eight different attractive female faces in order to familiarize them with the faces. They then tested whether a prototype created by averaging the eight female faces together seemed familiar to the infants. According to prototype theory, if babies have the ability to form prototypes by averaging the faces together, the prototype face should seem even more familiar than any of the faces the baby saw because it is similar to each of the faces and is this easily recognised by virtue of this similarity. Indeed, this is exactly what Rubenstein’s study found. Prototype faces were more familiar to the infants than either a novel female face or the individual faces the infants originally saw. The results show that by 6 months of age infants have the ability to create prototypes of female faces.
