@Article{,
author = {John R. Anderson},
title = {"Acquisition of cognitive
skill"},
journal = {Psychological Review},
year = {1982},
OPTkey = {},
OPTvolume = {89},
OPTnumber = {},
OPTpages = {369--406},
OPTmonth = {},
OPTnote = {},
OPTannote = {}
}
There are three stages that coorespond to Fitt's (1964) stages. The first is the declarative stage.Second is knowledge compilation. Third is the procedural stage. This theory is based on the ACT production system.
In ACT there are two memories-- declarative and procedural. The procedural memory is made up of productions and the declarative memory is made of chunks. Production have preconditions of what is in the goal and declarative memory. Goals are organized on a goal stack. When two productions conflict, the one worth more specific requirements gets chosen.Each production has a strength which represents the frequency with which that production has been successfully applied. A prod actually fires if it applies before a more specific production is selected. Productions gave variable slots that get filled with chunks when they match.
Novices record how to do tasks declaratively and general productions interpret it much like a programming language interpreter. Forming new productions is potentially dangerous because productions have direct control over behavior.
Having the wrong production could mean death for a living creature.
Knowledge acquisition phase has a couple of methods. One of them is composition, whereby a sequence of productions is compiled into a single production with the same effect.(p369)
Proceduralization is when a new production is made that does not require memory retrieval. Evidence for this is in telephone number dialing (noted in Anderson 1976) where somwtimes you cannot retrieve a number but you can dial it. In effect you create knowledge specific productions, which frees up WM for other simultaneous tasks.
In the final stage, the productions are tuned. Generalization is when productions accept more general cases (see p373 for an example). It takes two similar productions and makes a new one that does what the two of them have in common and variablizes. Other systems, like Brown 1977 generalize by changing the range of a continuous variable.