Effect of Prior Knowledge on Long-Term Memory
Introduction
Human memory has evolved with long-term storage capabilities to aid in immediate survival needs and future planning. The need to recall and make use of past experiences quickly and efficiently to inform current circumstances is evidenced in our ability to recognize friend or foe from a quick survey of identifying features. Long-term memory stores, filled with prior knowledge of the world gleaned from personal experience, support man in making judgements and decisions in uncertain future situations (Klein, Robertson, & Delton, 2010). The question of how best to represent the way in which this long-term memory information is collected, catalogued, recalled, and employed is one that has intrigued scientists and psychologists for hundreds of years. This paper will touch on broad concepts related to the representation, organization, and retrieval of long-term memory mapped out by these thinkers. It will delve into three elemental characteristics of long-term memory structures, that is, that they are highly organized, intricately interconnected, and constantly evolving. A design review will also include explication of scripts and tasks, subthemes within this exploration of the application of prior knowledge.
Representations of Long-Term Memory
The process of long-term memory-making and use is a complicated system which can be broadly defined in three stages. First, information is encoded into long-term memory through learning and training. This information is then maintainedover time in the mind’s storage system. Lastly, the information is retrieved through recognition, recall, or implicit demonstration of a relevant response based on prior experience (Baddeley, 1976, p. 9). Long-term memory can be separated into two distinct knowledge types: procedural (or implicit) memory - our ability to know how to complete a task, often employing our motor skills with little conscious thought, and declarative (or explicit) memory- our knowledge of facts (Baddeley, 1976, p. 7). Declarative memory can be further separated into two categories. The first, semantic memory, is our “mental thesaurus,” the knowledge a person possesses about words and verbal symbols, necessary for our use of language (Tulving, 1972, p.386). The second, episodic memory, is our capacity to retain and recall specific personal occurrences and mentally return to the scene to put past learnings to use in the present (Tulving, 1972, p.385-388). Memories morph and shift as new experiences build on memory stores, and there is a danger of losing or misidentifying information as edits are made to this knowledge. Research of the neurobiological structures of learning and memory find evidence that the implicit, semantic, and episodic memory systems are independently organized, filtering information to separate regions of the brain and allowing information to be processed in parallel (Poldrack & Packard, 2003, p.245). Activation of multiple memory stores through multiple modes of learning, a redundancy which supports recollection of memories from several knowledge systems, strengthens knowledge retention.
A Highly Organized System for Long-Term Memory Storage
With the complexity of retaining, storing, and retrieving information from distinct but connected memory stores, it comes as no surprise that long-term memory structures must be highly organized to support this process. Models and frameworks constructed by scientists and psychologists to represent long-term memory vary, often based on their preference for privileging either verbal or image-based representations of memory. Schemas and mental models are the most commonly presented cognitive representations of structures that link concepts and represent the relationships between and among those concepts.
Schema theories, representations of complex internal knowledge structures, abound through the 20th century in wide-ranging philosophical areas, from education to artificial intelligence. Sir Frederic Bartlett first defined schema as an “active organization” of past experiences which facilitates an action based on a set of responses to similar events which occurred at some time in the past (Bartlett, 1932. p.201). According to Bartlett’s Schema Theory, our response to an event is informed not by a sudden reaction to a stimulus, rather it is brought on by accessing and combining knowledge from one’s schemata - units of knowledge which contain an individual’s understanding of the world based on prior knowledge and experience - to produce the appropriate response (Bartlett, 1932).
Decades later, a similar schematic framework for representing knowledge storage was introduced by Marvin Minsky, a computer scientist attempting to produce machines which could “think” as humans do by reacting to a situation based on prior experience. Minsky’s frames, his structures for representing expectations in a generic situation, included a network of nodes and relationships holding hierarchies of knowledge, with undeniably true facts held above those “terminals” whose “slots” could be filled with alternate data to replace unnecessary knowledge and accommodate a new happening (Minsky, 1974). As new situations are encountered, one simply has to check out a frame - or perhaps combine several frame-systems- and plug in with new information to produce the expected behavior in that situation (Minsky, 1974). Minsky’s Frame Theory informed several related ideas, including Propositional Theory, a symbols-based representation of conceptual relationships, and the Script Construct, both discussed later in this review.
While the idea of creating models to represent a concept had many precursors, psychologist Kenneth Craik was the first to assert that thinking and reasoning are accomplished through internal manipulations of a “small-scale model of external reality,” our mental models (Craik, 1943). Mental models are highly personalized thought processes, representing an individual’s understanding of the world based on his or her own experiences. They provide internal imagistic representations of objects or ideas, which can be explored and manipulated in the mind to explore a problem, and provide the mechanism through which new information is filtered and encoded (Jones, Ross, Lynam, Perez, & Leitch, 2011, p.45). Mental models can be formed naturally through experience or structured more precisely through training or study. As experience in a particular subject area increases, the model for that concept becomes more complex and one’s flexibility in switching between models increases. Considering mental models for the understanding of Human-Computer Interaction, Don Norman urges researchers not to limit the understanding of these models to a simple “quasi-pictorial” representation, but to consider how the “mental simulation” of representation, facilitated by mental modelling, allows an individual to test out a solution - or many solutions - to a problem before deciding a course of action (Rumelhart & Norman, 1983, p.70-71). Norman points to conceptual models as the bridge to connect this “knowledge in the head” to “knowledge in the world”, providing a simple explanation or illustration in the real world of how something works, in order to unlock knowledge in the mind of how to use it (Norman, 2013, p.25 & 75). Conceptual models do not need to be precise, they simply serve to rouse the memory.
Intricately Interconnected Systems
Regardless of the form of the components for knowledge retention, all representations and models for memory storage emphasize the intricate interconnectivity of these structures. In 1969, Collins and Quillian first proposed the basic network model of semantic memory. The system, based on a model for storing information in computer memory, is described as a hierarchical series of connected “nodes” formed from words or word concepts (Collins & Quillian, 1969, p.240). Each word in this cognitive mapping has stored with it information about other word nodes - properties of the word - with “pointers” to these nodes embedded in the original word concept (Collins & Quillian, 1969, p.240). To retrieve a memory requires one to traverse a path from initial concept through the web of properties affiliated with it to properly categorize and understand it. In humans, then, this linking of information across nodes corresponds to the “distance” traveled in the mind to retrieve it, and time to verify the truth between a word-property relationship depends on how far apart the representations of concept and its related nodes are, how many connected paths one has to take to verify its accuracy (Collins & Quillian, 1969, p. 247). The more particular a property is to a word concept, the closer it will be stored to the representation in memory (Kosslyn, 1980, p. 347-348). In 1979, Zenon Pylyshyn presented a similar view of interconnected memory knowledge, presenting a theory of propositional networks: abstract data structures accessed with little conscious attention which are conceptual or propositional in nature (Pylyshyn, 1973). What a person knows could be represented by a list of propositions; by applying deductive reasoning, all valid propositions to solve a question could be deduced from this list and put forth to answer it (Pylyshyn, 1973, p. 12) His research provided a direct critique of memory as sensory, perceptive, or pictorial structures, popular with his peers, insisting that data was the building block for knowledge stores.
Interconnected memory systems require stimulation to maintain strength and speed in accessing and retrieving knowledge in long-term storage. Relationships between concepts in a given network make the retrieval process more efficient, as evidenced by the testing of memory involving participant recollection of related word pair concepts by J.R. Anderson throughout the 1970s. Repetition of a word pair, including a particular person in a particular place, primed participants to connect person-place concepts and yielded faster recall (Anderson, 1974). When presented with new people-place concepts, recall errors and response time both increased. Anderson accounted for this phenomenon with his spreading activation theory of memory, in which he postulates that, as knowledge increases, the number of nodes built into semantic memory networks also increase; a fanning out of these branches of knowledge occurs and slows the pace of information identification and retrieval (Anderson, 1983).
A Constantly Evolving System
Long-term memory is a constantly evolving system. Jean Piaget, a Swiss psychologist known for his theories of child development, laid the groundwork for the interpretation of memory as an evolving system. Piaget supported the idea of schema as the building blocks of knowledge, and built upon this foundation with processes of adaptation, namely assimilation and accommodation, which enabled a child to reframe his knowledge and move through one stage of understanding to the next, more advanced, level. Assimilation is the process by which a new situation or information is processed through an existing schema. Accommodation is enacted when existing schema are not sufficient to explain a new situation, when there is a mismatch of information presented and disequilibrium occurs, and a new schema is formed. Rumelhart and Norman also considered the evolving nature of schematic networks, separating the process of learning and knowledge retrieval into three parts: accretion, tuning, and restructuring. Accretion is the normal process of learning, a daily accumulation of information which requires no changes to the information processing system, simply adding new data structures to existing memory stores (Rumelhart & Norman, 1976, p.3,13). Tuning involves modifying existing schemata as new knowledge is acquired, gradually transforming old structures into new though practice, learning, and experience (Rumelhart & Norman, 1976, p.6, 14). Restructuring is the process by which new conceptual categories, new memory structures, are formed, either through replicating patterns of existing schemata or by building new schemata by combining recurring patterns of old (Rumelhart & Norman, 1976, p. 6, 14). This process requires far more time and cognitive effort than either accretion or tuning. It is likely then, that if a “close enough” match can be made through less intensive processes, it probably will be. The mind, after all, strives to conserve energy where it can.
Conclusion
Through a highly organized, intricately connected, and constantly evolving system, human memory stores enable us to go about daily life, taking in a barrage of new experiences with aplomb. One’s ability to innately comprehend expected behavior in a novel situation, by fitting new information within an existing network of knowledge, increases flexibility with new concepts and eases anxiety resulting from a mismatch of understanding. It is a designer’s responsibility to understand these processes of long-term memory representation, organization, and retrieval in order to access and make use of users’ prior knowledge stores, increasing familiarity and comfort with new concepts to yield the most successful outcome.
References
Anderson, John R. (1983). A spreading activation theory of memory. Journal of Verbal Learning and Verbal Behavior, 22(3), 261–295.
Anderson, John Robert. (1974). Retrieval of propositional information from long-term memory. Cognitive Psychology, 6(4), 451–474.
Baddeley, A. D. (1976). The psychology of memory. New York.
Bartlett, F. C. (1932). Remembering: A study in experimental and social psychology (Vol. 14). Cambridge University Press.
Collins, A. M., & Quillian, M. R. (1969). Retrieval time from semantic memory. Journal of Verbal Learning and Verbal Behavior, 8(2), 240–247.
Craik, K. J. W. (1943). The nature of explanation. Cambridge: Cambridge University Press.
Jones, N., Ross, H., Lynam, T., Perez, P., & Leitch, A. (2011). Mental models: An interdisciplinary synthesis of theory and methods. Ecology and Society, 16 (1), 46-46.
Klein, S. B., Robertson, T. E., & Delton, A. W. (2010). Facing the future: Memory as an evolved system for planning future acts. Memory & Cognition, 38(1), 13–22.
Kosslyn, Stephen Michael. (1980). Image and mind. Harvard University Press.
Minsky, Marvin. (1974). A Framework for Representing Knowledge. The Psychology of Computer Vision. McGraw Hill.
Norman, D. (2013). The design of everyday things: Revised and expanded edition. Basic books.
Poldrack, R. A., & Packard, M. G. (2003). Competition among multiple memory systems: Converging evidence from animal and human brain studies. Neuropsychologia, 41, 245–251.
Pylyshyn, Z. W. (1973). What the mind’s eye tells the mind’s brain: A critique of mental imagery. Psychological Bulletin, 80(1), 1.
Rumelhart, D. E., & Norman, D. A. (1976). Accretion, tuning and restructuring: Three modes of learning. La Jolla Center for Human Information Processing.
Rumelhart, D. E., & Norman, D. A. (1983). Representation in memory. La Jolla Center for Human Information Processing.
Schank, R. C., & Abelson, R. P. (1975). Scripts, plans, and knowledge. IJCAI, 151–157.
Tulving, E. (1972). Episodic and semantic memory. Organization of Memory, 1, 381–403.