Memory is a cognitive module in action organization in which information about objects, movements, events, environmental elements, and the action-related constellations between these entities are stored. Memory could be described as well as a process by which such information about the aforementioned elements are encoded, consolidated, stored, and recalled for use in attaining action goals. The structural organization of memory is based on units, categories, and expertise-dependent order formation and is therefore strongly related to learning processes.
The study of human memory has a long history, making it one of the oldest and most investigated topics in psychology. The initial scientific studies of memory are usually ascribed to Hermann Ebbinghaus’s work in 1885. He investigated the serial learning and forgetting of new information (nonsense syllables) and described the speed of forgetting by formulating the so-called forgetting curve. William James was the first scientist to address the topic of structure in memory with the distinction between primary and secondary memory, around the last decade of the 19th century. Over the years, the wealth of research assessing memory has greatly increased. The methods used have also evolved over time, spanning from experimental approaches in healthy subjects, to neurobiological and biophysical studies on neuronal cells, to neuropsychological studies on patients with brain damage, and in recent decades to neuroimaging research.
The first memory studies, and over the years the bulk of memory studies, have been conducted to learn about verbal memory using methods like serial learning, associative learning, free recall, and recognition tests. Through studying such learning processes, researchers have found that memory performance is constrained by both capacity and time-related storage (duration). Halfway through the 20th century, the idea of a distinctive short-term memory (STM) came forward, mostly supported by the investigations and theoretical work of Donald Broadbent in the 1950s and Richard Atkinson and Richard Shiffrin in the 1960s. From their points of view, STM temporarily allows a limited storage of information that can be conveniently retrieved without performing control processes like rehearsal. The capacity of information storage is limited as well. As a famous example for such capacity limits, in 1956 George A. Miller described the magical number 7 (±2) as the limited number of items that may be stored in STM. He was also able to show that this memory capacity could be increased through processes like chunking— that is, grouping items (e.g., digits) into reasonable frames or subgroups, such as three-digit-chunks. Based on various experiments in the 1960s using a partial report paradigm, George Sperling defined sensory memory, which refers to holding an iconic mental picture of the relevant items for approximately 100 to 500 ms after an item is presented. Based on such research lines, Atkinson and Shiffrin were the first to describe in 1968 a multistore model of memory featuring both (a) structure and (b) control processes. Concerning structure, the further developed Atkinson–Shiffrin memory model distinguished between three separate stores: sensory memory, STM, and long-term memory (LTM). Control processes are used for encoding, consolidating, and rehearsing information in memory; improving memory capacity; supporting the transfer from STM into LTM; and vice versa. This model became an influential framework in cognitive psychology and sport psychology (SP) because it provided an integrative theoretical perspective and a testable background for experimental studies.
In the years since the appearance of the original Atkinson–Shiffrin model, many alternative models have been developed, some of which hypothesize only one memory structure. In such models, sensory registers or short-term stores are hypothesized as temporary states of activation in a unique memory network. These perspectives are similar to the idea of a unified cognitive memory model like Adaptive Character of Thought–Rational (ACT– R), which was developed over two decades (from the 1970s to the 1990s) by John Robert Anderson. This model does not define memory structures by their storage duration. Rather, it distinguishes two types of information that may be stored in memory: declarative information, which consists of facts about our world, and procedural information, which is the knowledge about how to perform actions, for instance writing a letter. ACT–R theory also includes a computational implementation to study principles of human memory through simulation.
Much effort has been made in the field of memory research to define and describe working memory. While there are commonalities between working memory and STM, STM researchers emphasize time limitations, whereas working memory researchers address storage limitations from a functional point of view. The term working memory was originally used by George A. Miller, Eugene Galanter, and Karl H. Pribram in 1960 to describe the function of memory in the planning, implementation, and cognitive control of behavior. Because various working memory studies demonstrated that stimuli are differentially processed with respect to the sensory nature of information, Alan D. Baddeley proposed in 1986 a working memory model with three active components. The model distinguishes between two active slave systems: a phonological loop, responsible for the storage of auditory information, and a visuo-spatial scratch pad, which stores visual and spatial information. Furthermore, it has been assumed that these different types of representations are controlled by an attention-based central executive. With the help of this model, it is possible to explain why it is easier to perform a dual task with two differentsensory channels, with for instance auditory and visual information, compared to a dual task with the same kind of information, such as two types of visual information. This model of working memory remains influential even today.
Memory and Performance
Research over the past 40 years has revealed the close relationship between memory and performance. For example, the chess studies of Adriaan de Groot and later William G. Chase and Herbert A. Simon have highlighted differences in STM performance among expert and novice performers. These authors used different methods to learn about the cognitive mechanisms underlying expert performance. De Groot used think-aloud protocols, in which participants were instructed to verbalize their thoughts during the task, while Chase and Simon used a 5-second recall task to learn about how experts “chunk” meaningful game constellations. These chess studies revealed that experts are better than novices at storing task-relevant information in STM. However, their superiority is limited to meaningful game constellations. The experts’ advantage is no longer evident when players must reproduce meaningless constellations of chess pieces. Chase and Simon developed a chunking theory and proposed that experts are better able to memorize perceptual information because they adopt an organized pattern of information and a large number of chunks in LTM through practice. However, later research has shown that it is speculative to make direct statements as to how far chunked-representations in LTM mediate the formation of a chunk pattern in STM. As a consequence, William G. Chase and K. Anders Ericsson developed at the beginning of the 1980s the skilled memory theory, which argues that experts’ LTM structures, as well as their encoding and retrieval skills, spill over into working memory, supporting its capacity. This research perspective was later generalized to studies on the relationship between STM and performance in sport. New studies employing a variety of tasks such as menu orders, medical expertise, and text comprehension highlighted the limits of skilled memory theory. As a consequence, in1995 Ericsson and Walter Kintsch presented their concept of long-term working memory. From this point of view, experts do not only use encoding and retrieval skills but also develop a retrieval structure to address relevant information patterns in a particular task domain. Skilled performers are able to use long-term working memory to anticipate future retrieval demands and identify the task-relevant information in the environment and in memory. This work has made major contributions to the study of the functional links between retrieval processes in LTM and the coding and chunking processes in working memory, and it shows that the capacity for task-related information storage also increases as a function of performance in domains like high performance sport.
Further research has addressed the storage of knowledge components in LTM. As opposed to the focus on information storage capacity in working or STM, these studies have been more concerned with how knowledge is structured and networked in LTM. Hence, a major issue in this domain is whether we can confirm that improving performance is also accompanied by a higher degree of hierarchy in the knowledge structure. A wide range of methods and populations have been used to study expertise-dependent differences in the classification and memory representation of context-specific problem states, for instance, among springboard divers, judokas, triathletes, and weight lifters. Such research has revealed that the nodes of experts’ representation structures in memory possess far more features than those of novices.
Such nodes of representation in motor memory might involve formats such as propositions, relational structures of many kinds, and concepts. Researchers from various fields such as cognitive psychology, cognitive robotics, and SP have provided evidence in recent years of so-called basic action concepts (BACs) in the control of human movements. BACs are based on the cognitive chunking of body postures and movement events concerning common functions in the realization of action goals. Based on this definition of representation units in motor memory, Thomas Schack, Franz Mechsner, Bettina Bläsing, and other researchers studied the link between memory and motor skills in various kinds of sport and dance to investigate the nature and role of LTM in skilled athletic performance. In high-level experts, these representational frameworks were organized in a distinctive hierarchical treelike structure, were remarkably similar between individuals, and were well matched with the functional and biomechanical demands of the task. In comparison, actionrepresentations in low-level players and nonplayers were organized less hierarchically, were more variable between persons, and were less well matched with functional and biomechanical demands. The results from a number of different studies in domains such as golf, soccer, windsurfing, volleyball, gymnastics, and dancing have demonstrated that mental representation structures in memory are functionally related to motor performance.
- Atkinson R. C., & Shiffrin R. M. (1968). Human memory: A proposed system and its control processes. In K. W. Spence & J. T. Spence (Eds.), The psychology of learning and motivation: Advances in research and theory (Vol. 2, pp. 89–195). New York: Academic Press.
- Baddeley, A. D. (1986). Working memory. Oxford psychology. Oxford, UK: Clarendon Press.
- Cowan, N. (2008). What are the differences between long-term, short-term, and working memory? Progresses in Brain Research, 169, 323–338.
- Ericsson, K. A. (1985). Memory skill. Canadian Journal of Psychology, 39, 188–231.
- Krakauer, J. W., & Shadmehr, R. (2006). Consolidationof motor memory. Trends in Neuroscience, 29, 58–64.
- Schack, T., & Mechsner, F. (2006). Representation ofmotor skills in human long-term-memory.Neuroscience Letters, 391, 77–81.
- Starkes, J. L., Deakin, J., Lindley, J. M., & Crisp F. (1987). Motor versus verbal recall of ballet sequences by young expert dancers. Journal of Sport Psychology,9, 222–230.