Soraci, S. A.; Carlin, M. T.; Toglia, M. P.; Chechile, R. A.; Neuschatz, J. S. (2003). Generative processing and false memories: When there is no cost. Journal of Experimental Psychology: Learning, Memory and Cognition, 29, 511-523.
The present study emphasizes the effect generative processing has on false memories. Previous research has demonstrated the generation effect on non-words, word-fragment completion tasks, homographs and motor movement. The purpose of this study is to demonstrate how the generation effect affects false memories. The DRM paradigm has shown that participants will falsely recall words not presented in the study lists as well as report high confidence in their memory of critical foils. Another aspect of false memories is the Fuzzy Trace Theory. According to this theory (Reyna & Brainerd, 1995; Reyna & Kiernan, 1994), subjects will create verbatim and gist memories when presented with a list of words. Toglia et al. (1999) manipulated levels of processing within the DRM paradigm. Semantic processing was found to improve recall and elicit more false memories. Toglia et al. called this pairing a more is less pattern as enhanced recall consistently led to more false memories.
While several studies have further demonstrated the more is less pattern, there have been some that have been exhibited greater accuracy and fewer false memories. McDermott (1996, Experiment 2) demonstrated this using blocked versus randomly presented DRM lists. Benjamin (2001) and Goodwin, Meissner, and Ericsson (2001) also have demonstrated this pattern. Benjamin presented DRM lists three times, which led to more hits and fewer false memories. Goodwin et al. used acquisition lists composed of consecutive items that could form compound words.
Begg et al. (1989) proposed that generative encoding involves two processes: an initial “stimulus-recruitment” phase and a second, more discriminatory, phase. The present study was conducted to determine if generative encoding enhanced recall without increasing false memories using the DRM paradigm. Experiments 1 and 2 utilized word fragment completion tasks and recognition tests. Experiment 3, instead of a recognition test, employed free-recall. Experiment 4 moved away from the DRM paradigm, using congruous and incongruous response cuing.
Method:
In Experiments 1 and 2, using the DRM paradigm, participants were presented with 40 words or word fragments and were then asked to type the word (in the generation condition) or re-type it (in the read condition) before going on to the next word or word fragment. Following the study period, the participants were given 3 minutes to solve a series of math puzzles to provide as a distracter task. When the three minutes were up, the recognition task began. As words were presented on the screen, participants were asked to indicate whether or not they had seen the word during the study period by pressing Y or N.
Experiment 2 differed from Experiment 1 in item list length. Experiment 1 employed 5 item lists and Experiment 2 used 10 items per target.
Sixty items blocked by category were presented in Experiment 3. Each of the six blocks contained either whole words or word fragments with a single letter missing. Three of the six categories were presented in the generate format with the other three being presented in the read format. Each word or word fragment was presented for five seconds, allowing the participant to read the word and write it down. At the end of the list, participants gave the experimenter the sheets on which they wrote down the words and were then given a 4 minute free-recall test. Following the free-recall test, participants rated their confidence on a 3-point scale.
Experiment 4 used 32 pairs of words to create acquisition lists, differing from each other by a single letter (e.g. SNEAKER and SPEAKER). Four formats were used: congruous generate, congruous read, incongruous generate, incongruous read. Soraci et al. created eight 32-word lists to ensure that each word served in each of the four conditions. Participants then were asked to write or generate the word they saw. Once completed, they were given a 3-minute distracter task followed by 40oitem recognitions lists. Afterwards, participants rated their confidence on a 3-point scale.
Results:
Experiments 1 and 2 provided a strong argument that the generation effect enhanced recall and reduced false memories. In both experiments there was a significant main effect of item type but not for format. In Experiment 1, there was a significant difference in hit rates for both conditions, but there was no difference in false alarms. In Experiment 2, there was a significant generation effect for previously studied items, but no difference between the conditions for critical foils or their high associates.
Experiment 3 yielded a significant generation effect and item type effect and a significant interaction. It also replicated the generation at no cost pattern and is consistent with the results of Experiments 1 and 2. These results generalize the effect to free-recall tests.
Experiment 4 yielded significant main effects of format and item type. However, they did not produce a significant effect of congruity. The two-way interaction of format and item type and the three-way interaction of format, congruity, and item type were the only significant effects.
These studies provided further evidence that the generation effect improves recall without the cost of false memories and differs from Toglia’s more is less pattern. In Experiment 4, however, both more is less and generation at no cost were exhibited. The authors propose that the cuing processes that go along with congruous and incongruous generation provide a structure consistent with the results.