Critically Evaluate Theoretical Accounts Of How Knowledge Is Represented At A Cognitive Level.

At birth we are known as a ‘tabula rasa’ meaning a blank slate; in which nurture influences our mental content (J. Locke, 1895). The famous empiricist Locke also theorised simple ideas gained through our senses were developed into complex mechanisms. Thinking alone, cannot supply us with the ability to interact with the environment therefore we perceive and make predictions about the world through internal cognitive representations regardless of it being a scientific fact or a self believed fact. Consequently we built up knowledge from prior events, memories, perception, culture and socialisation. These cognitions convey knowledge to be represented as a mind state. Knowledge is the familiarity one has with worldly information.

The theoretical accounts of knowledge processes must be carefully analysed and critiqued. The fundamental base of this arguement relies on cognitive understanding, in which the mind plays a key role in knowledge acquisition, contemplation and retention. The arguement will be to explore the most valid line of reasoning in how knowledge of the representing world is conceptualised into abstract cognitive ideas. References made to key research with in-depth analysis will create understanding into how the cognitive paradigm views knowledge representation.

Analogical and propositional representations of knowledge have been derived from many theoretical viewpoints. Analogical representations of knowledge are physical and geometric interpretations of objects. Internally stored information is analogous to the representing world. Each element of the analogous object is represented as one. These analogies are generic generalisations and are flexible enough to attach relationships between objects i.e. loose combination rules. We gain information direct through specific modalities for example sight so representations are modality dependent.

Analogical representations provide skills for visual scanning. The Cathode Ray Tube model of imaging and is similar to how television mimics the spatial content of images analogous to the world. This metaphor therefore accounts for the use of our mind’s eye (inner eye to gain knowledge) also has a limited area that larger images can overflow (Kosslyn et al 1983). We can define and construct images from a spatial display of surface representation and semantic interpretation. Surface representation of images acts as a spatial medium; depth orientations and spatial properties (Kosslyn. S.M. 1980). Remarkably parts of the image correspond to the represented image in such a way we can make inferences about spatial information e.g. distances between objects. Deep representation occurs from constant refreshing of images and transfer into long term memory. This highlights analogical representations are usually specialized for some specific application domain. Additionally Kosslyn denoted perceptual information is dependant on occipital retinotopic maps and linked this to the imagery buffer i.e. surface representation. However conflicting evidence suggests blind patients have normal imagery (Chatterjee, A. 1995). Those with half the cerebral cortex damage only gain insight to half the images and do have trouble with imagery suggesting perceptual mechanisms have been damaged (Farah, M.J. 1992). Recent research has determined this one to one restriction is violated by the use of using 3D maps, and result in multiple representations of another analog (Krawczyk, D. C. 2005).

Research into imagery used simple yet effective methodology by showing similar objects but mirrored that differ in orientation. The object can be rotated to achieve similarity, picture plane (shape layout) and depth orientations (spatial layout); reaction time to achieve resemblance depended on the angular differences (Shepard, R.N. 1982). The transformation process occurs on each retina to compose a mental analogy of a physical rotation (Metzler, J. 1974). The surface representation holds a matrix of points which has a limited size to how big a picture could be. This linear relationship outlines that we can make images that coincide with reality. However this matrix representational format does not provide evidence for our ability to use our imagination for problem solving. One could technically compute the desired end point rather than using complex matrix multiplication. Grain size problem is a fundamental limitation, if we use the matrix notion of transformation; it is not possible for the same angular rotation to cover the same matrix cells in fear of being skipped or repeated, especially true for the periphery of the figure.

Interpreting ambiguous figures depends on the subjective interpretation of the individual as these images do not have typicality amongst general knowledge. Images in the spatial medium are represented at different levels of resolution. An ambiguous image was shown (5secs) and gave a first interpretation, subjects were then asked to form a mental image of this and devise a second interpretation. It was found that no participants correctly found the alternative interpretation; however drawing the image helps generate the correct interpretations of the ambiguous image (Chambers, D. 1983). In conclusion the difference between visual perception and imagery is visual can be reinterpreted.

There are two codes in the brain, one verbal and one visual based (Paivio, A. 1991). Paivio dual coding theory highlights verbal recall is generally higher for pictorial images than for items presented linguistically. The elaborative verbal imagery holds superiority for concrete words, in sense imagery as a memory code (Richardson, 1999). This concluded that dual code was created for concrete representations. However these results are based on correlations and not causal relationships. Visual information differs from the verbal code. Visual information determines spatial information and verbal interferes with spoken words, each can be manipulated. Words must be distinct in nature, familiar and able to connect with other information (Cornoldi, C. 1996). The theory is outdated yet still valid to human cognition. Bottom up research has stated it is likely that using the visuo-spatial sketchpad to create spatial mediums of images is demanding on the central executive and has links wih the phonological loop. (Baddeley et al., 1986). Therefore the uses for imagery are less complex and practiced (Baddeley and Leiberman 1984). Confounding variables of the research depend on whether participants close their eyes to use imagery if not they can still process images on this surface medium and may be influenced by propositional cues. Analogical is discrete if we map objects in to number systems then finite symbols.

Supporting research concluded that information can be verbal or requires visual or spatial processing. Some subjects were asked to memorize a block diagram of a letter. They were required to mentally travel the letter and indicate if the corner was on the extreme top or bottom. Other participants memorized a sentence (Brooks 1968). Each noun of the sentence was indicated by yes or no (verbally) or pointing to the words. The Reaction time to classify the words was slowest for those who gave a verbal response than those who pointed. This highlights the method for processing (verbal/spatial) effects reaction time in answering verbally or nonverbally.

Cognitive concepts allow us to make inferences from our prior knowledge, it limits the amount of info we need to relearn, perceive, remember and recognize. Conceptualization creates hierarchies of information, the preffered (basic and predictable) level categorizes common objects, the superordinate level categorizes the specific niche, and the subordinate level focuses on one aspect. Knowledgeable experts in a particular area would therefore make subordinate as quickly and accurately as basic (preferred) categorizations. It suggests the most typical information is presented in a structure. These generalisations are culture specific meaning general knowledge is constructed by societal expectancies e.g. language. This supports the notion of typicality gradients which will be later discussed.

Generic examples and multi-dimensional spaces are accounted for by the prototype theory. A prototypical item a new piece of similar information is collected and compared with its distinctiveness to the prototype (a new exemplar). A previous idea of exemplar based theory explains that the typicality was represented by an average of all the exemplars in a category consequently a cumulative (collective) average (Galton, A. 1880). However it is argued such theories are counterintuitive as it assumes information is consciously ready for classification.

Propositional knowledge aids us to make assertions of the world from a collection of symbols and interpretations and not from spatial representations. These are modality independent and visually perceived, with strong combination rules. Propositional interpretations enable us to be different and interpret world phenomena into a sentence-like description.

Concepts of objects and items which can relate are called semantic primitives. These propositional attributes ‘features’ are basic and descriptive in order to organise knowledge. Concepts were represented in tasks using case grammar to denote these relationships (Fillmore, C. J 1968). A restriction of the idea to use case grammar in investigating occurs as the difference between for example linguistic (extensional) and mental (intentional) representation is ignored (Laird, J. E 1987). Additionally the Conceptual Dependency theory outlines semantic primitives are in built representations of words. These semantic acts are described and then these were then characterised into semantic verbs linked to the real world (Shank, R. C. 1972). Concepts are context dependent and this knowledge becomes active in a particular situation. According to the 241PY lecture material, primitives are not defining attributes but are characteristics (Coleman. J. 1985).

Piaget suggested we assimilate new information into the existing knowledge base in order to achieve equilibrium with our environment (Piaget, J. 1971). Patterns in the environment arise in similar situations indicate expectancies; these expectancies are represented by a schema (Bartlett 1932). Conceptual dependency theory further supports the notion a collection of concepts and ideas and relation concepts are then represented as schemas.

Script theory is slightly more updated and much more detailed has provided more insight (Shank, R.C. 1977). Shank et al found conceptual clustering of information occurs; as classes of conceptual descriptions which can represent real world situations and develop the script construct to deal with generic knowledge of sequences of actions. These semantic networks are organized into bunches of conceptual hierarchies.

The dynamic memory theory outlines schemas which organize our knowledge about common situations: if you know the script applicable to the event, you can better remember the elements of the event (Holtz, V.F. 1992). Thus eye witness testimony shows memories of an event can be changed by misinformation even likely during interrogation (postevent misinformation) (Loftus, 1979). It Is therefore likely we can manipulate of schemas to account for additional data.

Furthermore Marvin Minsky proposed there are levels in the frame the top levels are knowledge that we know is true about a situation and the lower levels have gaps in which we fill with the ambiguous details (Minksy, M. L. 1977). A weakness of schemata is that it is defined as the form of mental representation for generic knowledge, but in most theories are then used as the term for the representation of all knowledge.

Perceptual symbol theory implements knowledge as a perceptual approach to recording information rather than as a construction process. During the perceptual process activation of the sensory motor areas in the brain occur (Pecher, D. 2003). Selective attention aids us to construct schemas; deep representations stored in long term memory. Mind brain identity theory suggest modalities connected to the brain help us interpretations of internal and external information.

Hierarchal organisation of schemas represents an interpreted structure of all exemplars from least to most typical. So unsupervised learning categorises and relates many objects in the most efficient manner. This can vary according to the population, the individual or the context. We drive to identify the more typical example of a category these create unclear or fuzzy boundaries between information which fits into this typicality gradient (Barsalou, L.W. 1987). Researchers found sentence verification tasks examined the difficultly of certain sentences deduced the hierarchal structure of information storage. Typicality effects are seen when participants are asked to verify ‘is a canary a bird’ and ‘is an ostrich a bird’. Decisions to answer ‘yes’ for question one may seem faster due our knowledge base being more sensitive towards recognising a canary as a better example of a bird rather than an ostrich. The second effect occurs when the participant is hesitant to choose which come from the categorisation similarity effect; indicating our knowledge base is organised to pick up even distant relationships between categories (Collins A, M. 1969).

In addition, procedural representations suggest our knowledge derives from interpreting the world logically. Logic exists in two forms semantics (meaning) and syntax (how it is constructed). We apply our knowledge based on informed logic; the use of modus ponens if P then Q, all birds have feathers, a canary has feathers - a canary is a bird. This rule has ecological validity and is used every day to analyse the relationships between concepts. Rule based theories claimed categorisation exist from the maintenance of reliable exemplars (Rouder, J.N. 2004).
A propositional representation has a structure that has no bearing to the semantics of the problem domain. Therefore, propositional representations can be made easily, by their own nature, universal. However, it makes them much less effective when navigating through search spaces.

Overall supporting evidence involves participants using imagery task which lack validity to tasks involved in the real world (ecological). Conclusively, analogical representation is purely intrinsic (imposed form the individual); when the object has similar constraints to the representing object. Hence it is typical for propositional representations that characteristics of the represented object differ from the representing object, so these characteristics are seen to be added extrinsic constraints. Whether or not we use analogical techniques depends on how well we have captured the critical features of world phenomena. In conclusion mental representations of knowledge are subjective in propositional sense or are analogous relating to spatial content of the real world. Procedural representations accounts for our ability to logically represent information internally.

To live is to know, without human motivation to interact with ones environment we may not acquire the skills needed in order to perceive and make internal generalisation and give structure to the wide information we receive daily. Knowledge can be stored as units intertwined over large regions of memory. It is believed that the differences between analogical and propositional representations of knowledge are overstated and over emphasised, each clearly has its own virtues and deficits for representation. As we age our familiarity of the worldly concepts increase and we achieve ‘wisdom’, suggesting a developmental repercussion of knowledge representation. Positive implications of research into knowledge systems and processing capabilities has the benefited education system suggesting that structure and categories aid children’s learning.