Failure to obtain a generation effect during naturalistic learning

Abstract

Many experiments have obtained a generation effect (GE) with various kinds of laboratory items. Six of the present seven experiments failed to find a GE when the responses were answers to general information questions that had been learned by college undergraduates who had either read or generated the answers during learning several days before the retention test. A GE also did not occur when those same answers were used as responses in paired-associate learning and were tested 20 min after learning. The GE appeared only when subjects learned lists of answers in the absence of the question context, followed by recognition testing. Implications ofthese findings are drawn both for the generality of the GE, especially to the kind of items and naturalistic situations in which learning occurs outside the laboratory, and for the theoretical mechanisms that may underlie the GE in traditional laboratory situations.Although some previous research has not found any difference in eventual recall after learning had occurred through reading as opposed to generatingthe to-be-acquired responses(e.g., when the responsesare nonsense syllables), the typical finding that has been of much current interest and that is called the "generation effect" (GE) is the following: Eventual recall is better after people learn items by generatingthe responses themselves than after they learn the itemsby reading the responsesas experimenter-provided words. However, as McDaniel, Riegler, and Waddill (1990) recently pointed out, "If the generation effect were restricted to one very particular paradigm, then its potential application (e.g., in educational situations) and illumination of learning and memory processes in general would be of questionable value" (p. 796).In accord with this point by McDaniel et al. (1990), we describe below our recurring failure to obtain a GE on items and procedures that we had hoped would produce one. On the assumption that the GE would generalize to everyday materials such as general information items (e.g., "What is the capital of Finland?" Answer = Helsinki) that students learn in naturalistic situations, we compared conditions in which people learned those items by generating as opposed to reading the answers. Our initial goal was to determine how the GE affected various aspects of metamemory such as the feeling of knowing and judgments of learning (note-because a precondition for such an investigation was that a GE should occurand this precondition was never fulfilled in any of our cued-recall experiments-we do not mention further any of the metamemory aspects). Because all but one of our attempts failed to produce a generation effect, these findings confirm the possibility of the limited boundary conditions mentioned by McDapiel et al. In our experiments, the GE did not occur for the ordinary factual items and postacquisition retention intervals that are a part of everyday learning in naturalistic situations. EXPERIMENT 1This experiment explored the possibility of a GE for a situation that involved learning unk...

A generation effect can be found during naturalistic learning

“…In fact, a large body of research exists on the generation effect with materials very much like those found in an educational setting (see, e.g., Doctorow, Wittrock, & Marks, 1978;Gardiner & Rowley, 1984;Graf, 1980;Pressley et al, 1987;Wittrock, 1990;Wittrock & Carter, 1975). I believe that the present three experiments illuminate the reason why Carroll and Nelson (1993) failed to obtain a generation effect with naturalistic materials when, as mentioned above, a vast body of research suggests that the generation effect can be extended to naturalistic materials. In conclusion, the generation effect is not limited to traditional laboratory situations; it has been repeatedly extended to materials and situations like those encountered in the real world.…”

“…The procedures and stimuli in Experiment 1 mimicked the significant characteristics of Carroll and Nelson's (1993) research (i.e., the general-information questions were from the same norms and difficulty range used by Carroll and Nelson in their Experiments 2 and 3, and the read and generate tasks were blocked on the study list), with one critical exception: For half of the subjects, the read task was not preceded by an initial attempt to answer the question.…”

“…

Recently, Carroll and Nelson (1993) presented research suggesting that general-information questions might represent a boundary condition for the generation effect. The present research focused on whether the generation effect did, in fact, generalize to such questions.

…”

“…Recently, however, the generality of the effect was challenged. Specifically, Carroll and Nelson (1993) cast doubt on one's ability to extend the laboratory findings to more naturalistic learning situations. Carroll and Nelson (1993) presented subjects with general-information questions (e.g., Which planet was the last to be discovered?).…”

Benefits of accumulating versus diminishing cues in recall

“…An alternative way of manipulating difficulty is to vary the amount of information provided by the retrieval cue (e.g., the number of letters of a to-be-remembered word; Benjamin, 2005; Carpenter & DeLosh, 2006; Carroll & Nelson, 1993). In spacing paradigms, difficulty increases with lag; in cue informativeness paradigms, difficulty increases with the poverty of the cue.…”