Template Theory
Template theory combines the concept of chunking with a retrieval structure to create a more detailed model for expertise. The strength of the theory lies in its ability to connect low-level and high-level information and is evident in its empirical success. Most of the research surrounding this theory has taken place in the domain of chess, which is useful for a number of reasons, namely that—as chess is the quintessential domain for expertise research—it is easy to compare and contrast.
Memory Systems
The authors of template theory assume a limited capacity of short-term memory (STM), typically around 3 chunks. This means that integration with long-term memory (LTM) is very important and must be addressed. According to template theory, chunks of chess pieces are contained in LTM, and the role of STM is to contain pointers to this information. Chess players access the LTM chunks by filtering information through a discrimination net.
When a chess expert sees an array of pieces and positions, the activity in his memory systems will depend on the novelty of the arrangement. In novel or atypical cases, the resulting chunk will only contain the pieces and their positions. In typical and familiar cases, however, the chunk will take the form of a discriminated node with semantic relevance, including information about what conditions might have created the particular arrangement as well as subsequent moves to execute.
From Chunks to Templates
As a player gains more experience with specific moves or arrangements, these chunks become templates. Templates contain not only chunks but also slots to accommodate variable features. For instance, there may be a slot for the piece (variable) contained at board position B5. The default value for this slot might be "none", but there could be 12 other possible values (black knight, white knight, black king, white king, and so on). From a connectionist perspective, it could be said that there are 13 tests below the node in the discrimination net.
A template is a useful and practical way to represent expertise in the domain of chess, as the slots of a node allow for access to and from several different paths. This theory assumes a much more dynamic role for the short-term store, incorporating new information into the template. The evolution of chunks into the more complex template data structures leads to a highly redundant and interconnected discrimination net, presumably accounting for the vast and complex range of phenomena exhibited by chess experts. This complexity also accounts for the length of time necessary to develop expertise in knowledge-rich domains.
More on template theory:
© Copyright Victor Long, Chandra Singh, and David Snitkof 2005All Rights Reserved