Mental representations are the fundamental component of cognitive processes as it is believed that much of cognitive processing can only take place with the aid of metal representations. Consequently this area of study is extremely important to psychologists and many cognitive theories have been modelled around the assumption that mental representations form the basis of cognitive processing. It has been identified through various forms of study, that incoming information which is received by the cognitive system must be represented in the form of a symbol or copy which signifies the real thing.
An object can often be represented in the form of pictures or by a verbal label, whilst spatial location can be represented by a map or by the numerics denoting a date or time. Mental representations can also be viewed from a number of perspectives and a broad division is often distinguished between prepositional language-like abstractions and analogical image-like copies of the real thing. Mental representations are essential in dealing with new experiences as they allow us to conceive of objects that are not immediately available to our senses.
The most commonly used construct to account for complex knowledge organisation is the schema. Schema theory was originally introduced by Bartlett (1932) and he proposed that what we remember is influenced by what we already know. The schemas work in a top-down direction to enable the interpretation of bottom-up flow information from the world around us. Schemas actively engage with new inputs to decide what will be stored and whether the memory representation of a new experience will be modified. Therefore new schemas can be stored as new schemas or modifications of old schemas which increase the store of general knowledge .
Whilst schema theory provides possibly the most comprehensive proposals of mental representation and their need in construction of Long term memory, critics suggest that because memories tend to be transformed toward what is familiar, mistakes and distortions can occur. Mental representations can be confused with real events, and certain details about a certain event can be lost in memory to a single representation. It is also difficult to ascertain how schemas are acquired in the first place, and one wonders how children remember or interpret new experiences without pre-existing schema in place.
In order to asses these criticisms and evaluate the value of schematic theory,it is probably best evaluated through studies of everyday memory. It is important to examine how well people remember everyday experiences and to what extent pre-existing schemas influence what is recalled. Brewer and Treyens (1981) devised an experiment to see if people’s memory for a scene was influenced by the schema which would normally relate to that particular scene. In their office scenario it seemed that most subjects recalled items with a high schema expectancy, many of which were not actually present in the room.
Likewise items which were not typical to the room were not named so frequently. Thus it seems that pre-existing mental representations can have a detrimental effect on what is actually observed or experienced. This can have significant implications with situations such as eye-witness testimony. Loftus et al (1979) carried out further experiments in this area, which tests the theory that new information is integrated with pre-stored memory representations. The main concern with this is whether people’s memory of an event can be unwittingly falsified at a later date.
It was found that although specific obvious information could not easily be distorted, people can be misled by false information if there is sufficient time delay, or the details are insignificant and peripheral to the main event (Cohen, 1993, p. 39). It therefore seems that highly specific information is not schema-based whilst other non-specific memories are sometime replaced by the most probable assertions. Knowledge which is derived from all areas of cognitive functioning, requires complex mental representations of facts and events.
The processing of information relies heavily on long term memory, and the use of schema theory to explain the construct of concepts as innate structures undoubtedly helps us to perceive the world around us. However, whilst this is invaluable, it must also be appreciated that complex and detailed information needs a flexible form of representation with a hierarchal organization. This is possibly because one of the main functions is to allow us to perceive analogies between experiences and events which appear on the surface to be quite different, but at some level of representation are similar.
This ability is fundamental to human cognition, as it enables us to learn from our experiences and apply what we have previously learned to novel situations. In order to achieve this, a hierarchal representation with various levels of specification and generality is somewhat essential. Whilst schemas can be described as declarative knowledge, our knowledge of what to do and how to perform actions is known as procedural knowledge. It encompasses sets of rules or procedures for skills as diverse as driving a car and calculating mathematical sums.
These rule-based representations are unconscious and it is only the output of the representational system which can actually be consciously accessed. A great advantage to unconscious processing is that learning new rules or changing old rules can be accomplished without disrupting other knowledge systems. In addition procedural representations can be accessed automatically and implemented with speed. Theses ideas have been developed as connectionist networks or neural networks in the form of computational models of the mind.
In this way the idea that representations enable the functioning of memory has further been modelled through connectionist models such as the ACT system, which was developed by Anderson (1983). The architecture uses not only procedural knowledge but also declarative memory which uses the schematic network and working memory that contains active and current information. These sets of productions are linked together and can be easily assembled and re-assembled to create different multi-component and action sequences.
They can represent complex plans because productions can be organized into hierarchal goal structures with higher level goals and lower level sub-goals (Kiss, 1993, p. 104). Computer based models are somewhat different from conventional experimentation, since the aim is to provide a general purpose cognitive system which can be used to develop ideas in all areas of human cognition. In this way although the use of representations can be simulated, the model cannot be tested in an empirical way since the computer system can still not exactly predict human performance.
In addition, the adequacy of any computer stimuli is related directly to the adequacy of our knowledge of the cognitive processes that the system is intended to simulate. The represented world is the physical world of patterns, objects, animals, plants and human faces. Perception involves a connection between three types of representation, namely those stored in long term memory, temporary representations which are constructed as intermediate stages in the processing of sensory input, and new representations which are the end result of perceptual processing.
One of the most important characteristics of cognition is that objects and events, however different, are defined in the mind as members of conceptual categories. These conceptual categories are mental representations which are stored in memory. The ability to group individually different items into conceptual categories according to their shared haracteristics is central to all cognitive activities including memory, communication and perception.
The categories then work to represent objects, events and entities using the maximum information and at the same time using the minimum cognitive effort. The defining feature approach assumes concepts can be categorized according to certain necessary properties. When these properties are joined together an object, entity or geometric shape can then be recognized precisely, with clear-cut boundaries being defined between members and non-members of that category.
The defining feature approach then assumes that concepts are mentally represented in this way. In addition to this approach concepts can also be organized in a hierarchal way, whereby concepts at a certain level adopt the defining properties of the original concept and acquire defining characteristics which make them distinguishable from all the other same level concepts (Eysenck & Keane, 2000, p. 285. Collins and Quillian (1969) proposed a model for hierarchal representation of biological concepts which have been based upon these assumptions.
The individual concepts within the hierarchy are said to be represented by definitions and that the defining properties of general concepts are inherited by their more specific subordinates. In order to test their model, Collins and Quillian used a task known as sentence verification. Interestingly, their findings supported the general theory of conceptual hierarchies but it was found that people’s knowledge of categories, and the difference of typicality for category members has relevance which must also be taken into account (Roth, 1995, p. 29).
These findings have prompted psychologists to review the defining feature approach of representation, and much of the theoretical and empirical work which has sought to uncover the flaws has been undertaken by Rosch (1973). Through study of everyday concepts, many categories of objects appear to have internal structure, in that the exemplars of a category such as furniture are not thought to have equal status. An explanation of this is that mental representations of categories are internally structured according to the typicality of their members.
People are able to categorize typical members more quickly than atypical members, and therefore it would seem that typicality is fundamental to the way categories are represented (Roth, 1995, p. 37). By outlining the difficulties in defining everyday categories, Rosch has theorised on the importance of ‘fuzzy’ representations, proposing that initially everyday categories are represented by single prototypes which have been based on typical instances. There are two derivatives from this idea. Firstly it is assumed that categories are represented by a summary list of typical features.
Secondly, it is assumed that category representations consist of specific and typical exemplars. Rosch also stresses the importance of hierarchal relationships in representing conceptual information, arguing that one level in such hierarchies would have special properties as a representation (Eysenck and Keane, 2000, p. 289). These theories stress what is know about the structure and formation of knowledge about concepts and about perceptual processes which allow us to decipher whether a particular object or entity is a member of a particular category.
In this way metal representations are used to assign conceptual categories on their visible features alone. This leads the question as to what stage the visible characteristics that access a concept are combined with other relevant attributes. It seems that the specific area of face recognition may hold the key, since our ability to recognize faces employs discrimination within a more basic level of object recognition.
Unlike discrimination for most object or biological entities, perception of a face involves a more complex interpretation, taking into account not only the subordinate categories, but also facial expression and speech (Bruce, 1993, p. 141). Many studies have been conducted in order to investigate how people recognize and perceive faces, and whether face features are indeed processed independently. Bradshaw and Wallace (1971) constructed faces out of Identikit, and asked subjects to decide if two faces were different.
They found that the more features that differed, the sooner on average subjects would notice a difference if they based their decision on a sequential comparison of the individual components. In another study by Sargent (1984), an experiment was devised in order to prove that the more features that differed between faces, the faster a ‘different’ response could be made. This lead to the conclusion that face features are processed interactively, and that a configuration emerges from a set of features which is more that the sum of parts (Bruce, 1993, p. 46). However, these studies do not explain what configural processes are involved, and it must be considered whether the results are due to the fact that perception of one part interacts of that of another, or whether the parts are not actually made explicit at all in the visual perception of faces. Tanaka and Farah (1993) have attempted to address these issues by investigating whether parts of the face were made explicit in face representation.
They found that memory for face features was poorer when tested in isolation than in the normal context of the whole face, supporting the claim that upright and normally arranged faces are processed holistically without separate explicit representations of parts. For complex patterns such as scrambled or upside down faces, there appeared to be no advantage for testing their parts in the context of the whole pattern. This would suggest that these patterns are not processed holistically (Bruce, 1993, p. 150).
Because face recognition is viewed as being a unique process on its own it is interesting to asses how brain injured people fair without this capability, and without the necessary mental representations needed to judge facial expression or recognize someone who used to be familiar to them. Neuropsychological evidence is consistent with the theory that there is a sequence of operations establishing face familiarity, identity and name retrieval. Young et al (1985) conducted a study of everyday identification to establish exactly what sort of error in identification are normally experienced.
From the results of their study they have proposed a model for functional components which is involved in person identification. Through this model it is suggested that certain sequences have to be processed in order to identify a person, but should failure occur at any stage, identification would not be possible (Bruce, 1995, p. 174). Not only is what we see and how we perceive things important through mental representation, but also how we convey our thoughts. Language is defined as having two main functions, namely external communications with other people and internal representations of our own thoughts.
Cognitive psychologists are concerned with how the knowledge underlying language use is represented in the mind. At the level of discourse representations, it is sometimes difficult to separate out knowledge of a language from general knowledge of the world. Psychologists such as Schank (1972) believe that representations of real life knowledge are used to interpret inputs, whether they are actual real experiences, or are words or texts. In this way people continually use their knowledge of situations and events to make assumptions and construct interpretations of the world around them.
In some ways there is no difference between an actual experience and reading a story about the experience. People are believed to activate their general knowledge about situations and events in order to understand linguistic descriptions. Therefore, it can be assumed that understanding language is part of our experience of the world and not a separate activity (Greene, 1995, p. 24). It would therefore seem that mental representations are essential to effective cognitive processing. Underlying our ability to make sense of the world is the ability to recognize objects or faces or being able to interpret novel experiences or to solve problems.
Schema theory is vital in the perception of mental representations, since it means that not all aspects of a visual scene need to be analysed and that everyday scenes provide clear expectations about what we are likely to experience. In this way schemas actually reduce the amount of processing needed by the cognitive system. Conceptual categorization also makes essential use of mental representations, by applying abstract representations or concepts to objects or entities. This gives us the ability to recognize or make sense of perceptions and attach labels to them.
The significance of the need for mental representations is clearly shown through the study of brain injured patients who suffer dissociations and are deprived of important aspects in normal everyday situations. Deficits in the functioning of schematic hierarchy and conceptual categorization result in incorrect representations being made. Therefore, without the ability to make use of mental representations the cognitive system cannot function properly, and without full use of the cognitive system there can surely be no grasp on reality.