Automatizing alphabet arithmetic: I. Is extended practice necessary to produce automaticity?

Abstract

Four experiments tested the necessity of extended practice in producing automaticity in an alphabet-arithmetic task in which subjects verified equations of the form A + 2 = C, asking whether C was two letters down the alphabet from A. Experiment 1 trained subjects on 40 alphabet-arithmetic facts for 12 sessions, demonstrating that extended practice was sufficient to produce automaticity. Experiment 2 produced the same degree of automaticity in a single session by having subjects rote memorize 6 facts, suggesting that extended practice is not necessary. Experiments 3 and 4 explored procedural differences between Experiments 1 and 2 to determine what was responsible for the large difference in the time required to develop automaticity. Experiment 3 compared learning rates with different numbers of facts (6, 12. and 18), and found learning rate to depend on the number of presentations of individual items, not on the number of items to be learned. Experiment 4 compared learning by performing the task (as in Experiment 1) with learning by remembering the facts (as in Experiment 2) and found no important differences between them. The results of all 4 experiments cannot be predicted by approaches that define automaticity in terms of resources or by listing properties, although they are readily predictable from theories that assume memory retrieval is the process that underlies automaticity. Is extended practice necessary to produce automaticity? Can automaticity to be produced without it? Extended practice is certainly sufficient. Many of the properties of automaticity can be produced by extended practice (e.g., Logan, 1978; Shiffrin & Schneider, 1977) or by using materials such as familiar words that have a history of extended practice outside the laboratory (e.g., Neely, 1977). However, sufficiency does not imply necessity. Existing data provide no answer. Common theoretical approaches to automaticity provide no answer either. Approaches that define atuomaticity in terms of manifest properties, such as speed, effortlessness, and autonomy (property-list approaches), are stipulative or descriptive but not predictive (e.g., Hasher & Zacks, 1979; LaBerge & Samuels, 1974; Shiffrin & Schneider, 1977). They provide no underlying mechanism from which predictions about the properties of automaticity or necessity of extended practice can be deduced.

The loss of repetition priming and automaticity over time as a function of degree of initial learning

“…The hundreds or thousands of trials typical of many automaticity studies increase the likelihood that the solution will be retrieved from memory rather than computed from first principles, and they increase the number of items for which a solution is available in memory. Large numbers of trials are not required for automatic performance, nor do they guarantee it (Logan, 1988;Logan & Klapp, 1991).…”

Strategy transitions during cognitive skill learning in younger and older adults: Effects of interitem confusability

“…The progression of values is of interest. In the first epoch, all values are close to 400 msec/addend, a value typical of the alphabet enumeration process that underlies computational solutions (Logan & Klapp, 1991). As training progresses, the eight curves in Figure 2 diverge; three patterns can be discerned.…”

An exemplar-based random walk model of speeded classification.

“…Indeed, it accounts quantitatively for the power law decreases in the mean and standard deviation of response times that are often observed with training (Logan, 1988(Logan, , 1992Newell & Rosenbloom, 1981 ) and deals quantitatively with entire distributions of reaction time data (Logan, 1992). The theory has also shown considerable power in accounting for the development of automaticity in memory search (Strayer & Kramer, 1994), lexical decision (Logan, 1990), alphabet arithmetic (Logan & Klapp, 1991), and repetition priming (Logan, 1990).…”