Theoretical and Applied Implications of Precisely Measuring Learning Rates

Abstract

Nist and Joseph (2008) have confirmed earlier research showing that adding and interspersing a large number of time-consuming learning trials targeting known items (e.g., incremental rehearsal [IR] or intersperal) retards student learning rates (Cates et al., 2003 Joseph & Nist, 2006; Schmidgall & Joseph, 2007). In addition, their current study has confirmed earlier research that adding and interspersing known items can improve the quality of learning trials, or the amount learned per unknown-word trials (Burns, Dean, & Foley, 2004; MacQuarrie-Klender, Tucker, Burns, & Hartman, 2002). Learning Levels Versus Learning Rates Learning is often described as a relatively permanent change in behavior or behavioral potential brought about by experience, with some experiences (e.g., an accident causing severe brain damage) being excluded. Across many applied and theoretical experiments, the experience is the instructional activity (e.g., intervention, remedial procedure). In the Nist and Joseph study, the learning experience was held constant across all conditions is the flashcard drill procedure. Across all conditions, each session included six trials, across six unknown words, repeated nine times per session. However, these flashcard trials were not the independent variable or the experience under investigation. Rather, the independent variable was the presence of additional flashcard trials targeting known words. These known-word trials were placed around the unknown-word trials using either IR (nine known-word trials placed among six unknown-word trials using a folding-in strategy) or interspersal (three known-word trials evenly distributed among six unknown-word trials). In the control condition, no known-word trials were added. In many intervention studies, the behavior change (learning) often serves as the primary dependent variable. Nist and Joseph used unknown words read accurately in isolation (acquisition and maintenance) and in sentences (generalization) as their dependent variables and repeated assessments were used to assess behavior change (learning) as intervention procedures were applied over time and across sets of unknown words. Their study has strong applied value because they compared three strategies and assessed changes in learning levels by including acquisition, maintenance, and generalization measures. Changes in levels of behavior (e.g., number correct, number maintained, number of words read correctly in sentences) do not take into consideration instructional time required to bring about that change. Nist's and Joseph's study is superior to most because they also measured learning rates. Assessing learning rates requires measuring the change in behavior (level change) while also taking into account the time required for the learning experience. Learning rates = learning level/instructional time (Cates et al., 2003; Skinner, Belfiore, Mace, Williams, & Johns, 1997; Skinner, Belfiore, & Watson, 1995/2002; Skinner, Fletcher, & Henington, 1996). Implications of Learning Levels Results Over 100 years ago, using himself as a subject, Ebbinghaus (1885) showed that increased practice enhanced learning. Thus, in some instances researchers comparing learning strategies may want to hold practice, often referred to as opportunities to respond (Greenwood, Delquadri, & Hall, 1984) or learning trials (Albers & Greer, 1991), constant. Holding the type (flashcard) and number of unknown-word learning trials constant across the three different learning experiences allowed Nist and Joseph to evaluate the relative quality of the learning experiences when the additional known-word flashcard trials were added (Skinner et al., 1997). Confirming previous research, Nist and Joseph found greater increases in learning levels under the IR procedure (Burns et al., 2004; MacQuarrie-Klender et al., 2002). Nist's and Joseph's findings have theoretical significance as they suggest that something involved with adding these known-word trials and folding them in enhances the quality of the unknown-word trials (amount learned per unknown-word trial). …

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 86 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.