|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001469419
007 cr mnu|||uuuuu
008 040524s2004 flua sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000321
Taylor, Kelli M.
The effects of overlearning on long-term retention
h [electronic resource] /
by Kelli M. Taylor.
[Tampa, Fla.] :
University of South Florida,
Thesis (M.A.)--University of South Florida, 2004.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 41 pages.
ABSTRACT: Overlearning is defined as the continued study of an item immediately after it has been correctly recalled once. Based on past studies, the effectiveness of overlearning is uncertain. In the present study, the effectiveness of overlearning was tested in 3 experiments. In Experiment 1, participants studied 10 city-country pairs (e.g. doba, chad) to either a low or high degree of learning and were tested 1, 3, or 9 weeks later. In Experiments 2 and 3, participants studied varying numbers of word-definition pairs (e.g. vizard, mask) for a constant total study time. They were tested between 1 and 4 weeks later. While overlearning improved test performance, this difference in recall diminished at longer retention intervals. Furthermore, overlearning was found to be inefficient because the increase in recall was not proportional to the increase in study time. Finally, for a given amount of study time, the underlearning of more words led to higher absolute recall totals than did the overlearning of fewer words.
Adviser: Doug Rohrer
total study time.
study list length.
degree of learning.
t USF Electronic Theses and Dissertations.
The Effects of Overlearning on Long-Term Retention by Kelli M. Taylor A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts Department of Psychology College of Arts and Sciences University of South Florida Major Professor: Doug Rohrer, Ph.D. Tammy Allen, Ph.D. Kevin Thompson, Ph. D Date of Approval: April 8, 2004 Keywords: repetition, learning, total study time, degree of learning, study list length Copyright 2004, Kelli M. Taylor
i Table of Contents List of Figures iii Abstract iv Introduction 1 Benefits of Overlearning 2 Limitations of Overlearning 3 Retention Interval 3 Efficiency 5 The Total Time Method 7 Overview of the Present Studies 9 Experiment 1 11 Method 11 Participants 11 Design 11 Procedure 11 Results and Discussion 12 Experiment 2 15 Method 15 Participants 15 Design 16 Procedure 16 Results and Discussion 17 Experiment 3 20 Method 20 Participants 20 Design 20 Procedure 20 Results and Discussion 21 General Discussion 24 Summary 24 Implications 27 Future Studies 28 References 31
ii Appendices 33 Appendix A: List of city-c ountry pairs for Experiment 1 34 Appendix B: List of vocabulary words for Experiment 2 35 Appendix C: List of additional vocabulary words for Experiment 3 36
iii List of Figures Figure 1. Hypothetical Results of an Overlearning Experiment 4 Figure 2. Hypothetical Recall Performan ce as a Function of List Length 8 Figure 3. Results of Experiment 1 13 Figure 4. Results of Experiment 2 18 Figure 5. Results of Experiment 3 23
iv The Effects of Overlearning on Long-Term Retention Kelli M. Taylor ABSTRACT Overlearning is defined as the continued st udy of an item immediately after it has been correctly recalled once. Based on past stud ies, the effectiveness of overlearning is uncertain. In the present study, the effec tiveness of overlearning was tested in 3 experiments. In Experiment 1, participants studied 10 city-country pairs (e.g. doba, chad) to either a low or high degree of learning and were tested 1, 3, or 9 weeks later. In Experiments 2 and 3, participants studied vary ing numbers of word-definition pairs (e.g. vizard, mask) for a constant total study time. They were tested between 1 and 4 weeks later. While overlearning improved test perfor mance, this difference in recall diminished at longer retention in tervals. Furthermore, overlearn ing was found to be inefficient because the increase in recall was not proportional to the increase in study time. Finally, for a given amount of study time, the underlearn ing of more words le d to higher absolute recall totals than did the overlearning of fewer words.
1 Introduction Educators and trainers often use repetition as a teaching device. For instance, homework assignments often include many of the same type of math problem, repeated practice of spelling or vocabulary words, or repeated conjugations of a foreign language verb. Many of these tasks ar e an act of overlearning, which is defined as the continued study of material, in a single learning session, after it has been learne d to a criterion of one correct recall. For example, suppose a st udent uses flashcards to study vocabulary words and removes a card once that item has b een correctly recalled once. This is an example of adequate learning (AL), where the criterion is one correct recall. Any further study of the words, in the same study session, is known as overlearning (OL). Although overlearning helps students on a test the ne xt day and trainees on a skills assessment shortly after training, it is not clear whether it is beneficial ov er longer reten tion intervals. Moreover, it is uncerta in if the boost in recall due to underlearning is proportional to the amount of extra time required to overlearn. For example, twice the effort may not lead to twice the increase in recall. Information on the usefulness of overlearni ng over long periods of retention could be very helpful when applied to classroom s and training courses. For example, if overlearning is a good learning strategy, th en students should repeatedly study vocabulary words and solve many similar math problems in a single session. However, if overlearning is not particularly useful, then certain commonly used exercises could be
2 replaced by more effective ones. For example, teachers should not assign vocabulary exercises that include the same 20 words ev ery night for a week; instead, they should assign 20 different words each night for a total of 100 words that week. Although the assigning of more words will lead to the recall of a smaller proportion of the words, this underlearning of 100 words might lead to a gr eater absolute number of recalled words. Hence, studies on overlearning can evalua te the usefulness of such activities by examining the long-term retention of overlear ned material. Furthermore, the application of the results of overlearning studies can lead to more efficient cl assrooms and training programs. Benefits of Overlearning There is an overwhelming body of resear ch suggesting that overlearning is a valuable method of learning. That is, a nd perhaps not surprisingly, conditions that require overlearning produce greater recall than conditions that requ ire lesser degrees of mastery. For example, Krueger (1929) f ound overlearning trumped adequate learning when participants learned words and returned for a test up to 28 days later. Postman (1961) found that participants re quired less time to relearn lists of wo rds when the lists were originally overlearn ed. Schendel and Hagman (1982) found that overlearning helped to decrease the amount of time needed to retrain soldiers to assemble and disassemble a machine gun. Driskell, Willis, and Copper (1992) concluded from their meta-analysis that overlearning is an eff ective learning tool f or both physical and cognitive tasks (p. 618).
3 After findings such as these, it is no w onder that overlearning is often described as a great learning technique. In fact, Fitts claimed that the importance of continuing practice beyond the point in time wher e some criterion is reached cannot be overemphasized (1965, p.195) Likewise, Hall (1989) called overlearning practical because continued practice on material alrea dy learned to a point of mastery can take place with a minimum effort, and yet will preven t significant losses in retention (p. 328). Similarly, Foriska (1993) endorsed overlear ning by presenting it as the mechanism children use to move information from shor t-term memory to long-term memory. With conclusions such as these, it is not surpri sing that researchers and educators advocate overlearning. Limitations of Overlearning Although research on overlearning does s how that overlearni ng leads to better recall than lesser degrees of l earning, these results are not surprising, as greater effort almost always produces greater performance. However, the apparent benefits of overlearning are less impressive when one cons iders the factors of retention interval and efficiency. Retention Interval While overlearning has often b een shown to lead to better retention after short re tention intervals, stud ies that employ a longer retention interval have revealed a lesser benefit. In fact, mo st of the studies that reveal benefits of overlearning have very short re tention intervals. For instan ce, none of the studies in the meta-analysis by Driskell et al. (1992) used a retention interval that exceeded 28 days.
4 Similarly, Postman (1961) used a retention in terval of only seven days, and Ausebel, Stager, and Gaite (1968) used a reten tion interval of only two days. Likewise, studies that employ varying rete ntion intervals show that the boost due to overlearning diminishes with time. That is, as the time between the study session and the test session grows longer, th e benefits of overlearning de cline, as illustrated by the hypothetical data in Figure 1A. For example, Craig, Sternthal, and Olshan (1972) found that, at several levels of overlearning, the boost in recall due to ove rlearning decreased as retention interval increase d. Similarly, Reynolds and Gl aser (1964) found that the overlearners and adequate learners curv es converged, as shown in Figure 1A, as retention interval was lengthe ned. Specifically, they found that the differences between a 200% overlearning group and a 50% overlearnin g group existed at the 2-day interval on a recall task, yet these differences disappeared at the 21-day interval. Likewise, the metaanalysis done by Driskell et al. (1992) found th e benefits of overlearning to weaken as retention interval lengthene d. These findings indicate that while overlearning is beneficial over short reten tion intervals, this learning technique does not produce longterm retention. OLAL Retention IntervalRecall Figure 1 Hypothetical Results of an Overlear ning Experiment (A) Convergence of learning curves. (B) Divergence of Learning Curves. OLAL Retention IntervalRecall A B
5 One exception to this pattern of decl ining benefits was reported by Krueger (1929). His results showed a divergence of recall at longer retention intervals (Figure 1B). Specifically, he found that an increase in de gree of learning from adequate learning to overlearning led to an increase in rete ntion at both the 1and 28-day retention intervals. Furthermore, at the 28-day interval, the ratio of words recalled by the overlearners to the ratio of words recalled by the adequate learners was actually larger than the same ratio at the 1-day retention interval. These conflicting results in the literature leave an unanswered question about the usefulness of ove rlearning at longer retention intervals. Efficiency. Regardless of whether overlearning increases retention at certain retention intervals, it may not be the most e fficient study method. That is, the increase in study time needed for overlearning may produce diminishing returns in recall benefits. To be precise, if a list of paired-associates is presented 10 (low) times, 15 (moderate) times, or 20 (high) times, recall total w ill naturally increase as the number of presentations increases, but the increase between the low and m oderate groups will be greater than the increase betw een the moderate and high group s. One way of examining this inequity is by comparing varying degr ees of overlearning. Using this method, Krueger (1929) found that, as degree of ove rlearning increased, rete ntion also increased, but not proportionately to the increase in ove rlearning. In other words, when the number of trials was increased by 50%, retention increased by 48%. Yet, when the number of trials was increased by another 33 %, retention increased by only an additional 27%. This suggests that overlear ning becomes less efficient as degree of overlearning
6 increases. That is, the increase in study time needed to reach higher degrees of overlearning may produce a disproportionately small return in recall performance. Similarly, Bromage and Mayer (1986) found that as the number of trials increased, recall total also increased at a nega tively accelerated pace. That is, as degree of overlearning increased, the rate of increase in retention slowed. Similarly, Driskell et al ( 1992) found that participants did retained more as the degree of overlearning increased, but once again, this increase in study time led to diminishing returns. Likewise, Kratochwill, Demuth, and Conzemius ( 1977) found that increasing the number of study trials by 20 led to an increase in re tention of only one word, or only 25%. Once again, the proportional increase in retention was much less than the increase in study time. Findings such as these suggest that the benefits of overl earning do not increase at the same rate that degree of overlearning in creases, thus making it an inefficient learning strategy. That is, the proportio nal increase in retention due to overlearning is usually not as large as the proportion al increase in study time Therefore, although much of the overlearni ng literature describes overlearning as a beneficial learning t echnique, a closer examination of th e data reveals that the benefits of overlearning decline at longer retention intervals. In ad dition, there is much evidence suggesting that the benefits due to overlear ning are not commensurate with the increase in total study time.
7 The Total Study Time Method Another source of literature related to overlearning that is not often cited in overlearning studies are studies employing what is known as the total time paradigm. Unlike the overlearning paradigm, where total study time is varied a nd study list length is held constant, in the total time paradigm, tota l study time is held constant and study list length is varied. For example, in a totaltime study, participants might be given five minutes to study a list of 5, 10, or 20 items. Of course studying 5 items will lead to the recall of a higher proportion of items, but it is unclear whic h condition will lead to a higher absolute recall total (i.e. the total number of words recalled). The implications are described in greater detail in the general discussion. Murdock (1960) used this design and hypot hesized that there is no effect of varying list length on absolute recall total. Hence, this hypothesis suggests that, when given a total time of 10 minutes to study 10 wo rd pairs, the same number of words will be recalled as when given 10 minutes to study 20 word pairs. Yet Roberts (1972) reported results contrary to th is hypothesis. He instead found the absolute recall total increased as list length increased when total time was held constant. For example, for a given 80-s study period, participants who studied 40 words recalled more words absolutely than participants who studied 20 words. Here studying more words in an allotted time led to an increas e in absolute recall even though the proportion of words recalled declined. That is, th e overlearning of 20 items produced a smaller absolute recall total than the underlearning (UL) of 40 items. From these results, it seems that studying more words in an allotted time can lead to a higher number of words recalled, and that
8 overlearning is not the best le arning strategy when absolute recall rather than proportion of recall is important. For example, when pr eparing for the verbal section of the GRE, there are thousands of vocabulary words that ma y appear on the test. Thus, it is best to know as many words as possible to increase the chance that you will know some of the words that do appear. In th is instance, then, absolute recall is more important than proportional recall. However, Roberts (1972) found that this increase in absolu te recall due to increased study list length, when total study time was held constant, asymptotically approached a ceiling. Specifically, he found there was no significant difference between the absolute number of words recalled af ter studying 20 or 40 words in a 40-s study period. Thus, for a given to tal study time, extreme underlearning (EUL) may provide either no further boost in absolu te recall total or even a decl ine. That is, as shown in Figure 2, does the level of ab solute recall continue to in crease as study list length increases, or will absolute recall begin to d ecline as study list length increases to extreme levels? Short (OL) Medium (UL) Long (EUL) Study List LengthAbsolute Recall Figure 2. Hypothetical recall performa nce as a function of list length (constant total study time). As the degree of underl earning increases to extreme levels, the increase in absolute recall total could eith er reach asymptote, or begin to decline.
9 Although Roberts (1972) found that unde rlearning many words can produce greater absolute recall than overlearning fewer wo rds, there are two characteristics of this study that hinder generalizability and ecological validity. The first is his use of a free recall paradigm, where participan ts are asked to recall a list of studied words. Situations in which free recall are needed do not normally present themselves in the real world. On the other hand, tasks that involve cued recall occur more often in real world settings. For example, if you are visiting a Spanish speaki ng country, you might need to know that the Spanish word for food is comida. The second limitation of Roberts study is his use of an immediate retention interval. Because it is usually important to retain information for more than a minute, of course it is important to test this effect at longer, more meaningful retention intervals. Overview of the Present Studies In the present studies, we pitted ove rlearning against underlearning in three different experiments to explore the bene fits and limitations of overlearning. In Experiment 1, the utility of overlearning was assessed with a design normally used in the overlearning literature as desc ribed further above. Specifica lly, every participant studied 10 items, and total study time was varied. In Experiment 2, we assessed the utility of overlearning by varying list le ngth rather than total study time, as in the total time studies. That is, participants studied lists of either 10 or 20 words for a period of 12 minutes. In Experiment 3, we once again he ld total study time constant and varied list length to assess whether any increase in ab solute recall achieved by an increase in
10 number of study words ultimately declines at long er list lengths. In all three experiments, participants returned for a test be tween one and four weeks later.
11 Experiment 1 This experiment tested the benefits of overlearning over long retention intervals using a paper/pencil task. Co llege students either overlearned or underlearned 10 citycountry pairs for either 5 or 20 minutes and were tested one, three, or nine weeks later. Method Participants. A sample of 130 undergraduate students participated in return for extra credit. Eight other students began the study but fa iled to complete it. Design. The two variables, lear ning level (Lo or Hi) and retention interval (1, 3, or 9 weeks) were manipulated between subjects. Participan ts were randomly assigned to conditions. Procedure. The experiment consisted of both a study session and a test session. Booklets were used during the study session. The first page of the booklets included a list of 10 city-country pairs, as listed in Appendix A. At the beginning of the study session, participants were given 60 s to study thes e pairs. Following th is initial study trial were 5 (Lo-Learners) or 20 (Hi-Learners) fee dback trials, each of a 1-min duration. Each trial corresponded to one page in the booklet. On these pages, th e cities were listed on the left, and participants were asked to reca ll the corresponding country in the blank space provided to the right. Handw riting time was reduced by including countries with only five or fewer letters. After 50 s, particip ants were asked to unfold the page, which revealed the correct answers immediately to the right of the part icipants answers.
12 Participants were then instru cted to study the city-country pa irs for the remaining 10 s of the 1-min trial. After the participants had completed five of these trials, the Lo-Learners left the room. After a brief rest, the Hi-Learne rs completed their remaining 15 trials. The words were randomly ordered each time they appeared so that pa rticipants would not merely memorize each definitions position within the list. The participants were tested one, three, or nine weeks late r. During the test, participants were given the name of the city and were asked to r ecall the country. They had three minutes to complete this task. Results and Discussion The results of the study phase are shown in Figure 3A. As shown, virtually all the Hi-Learners (overlearners) correc tly recalled all 10 countries on each of the last 10 trials. Because they completed twice as many trials as that needed to reach the criterion of one correct recall, these Hi-Lear ners clearly relied on overlea rning, as defined in the introduction. In contrast, on average, the Lo-Learners (underlearners) had not reached a criterion of even one correct recall of all 10 countries by the end of their study session. Thus, many of the Lo-Learners failed to reach even adequate learning. The test results are shown in Figure 3B. As expected, the Hi-Learners recalled more than the Lo-Learners, as indicated by a si gnificant main effect of degree of learning on recall, F (1,124) = 33.29, p < .01. However, the difference between the Hi-Learners and the Lo-Learners decreased dramatically wi th retention interval, as evidenced by the significant interaction between reten tion interval and degree of learning, F (2,124) = 8.65, p < .01. More specifically, the results of a Tuke y HSD test revealed that the difference in
13 recall between the Hi-Learners and Lo-Learners was significant at the 1and 3week intervals but failed to reach significance at the 9-week interval. A Hi-Learners (OL) Lo-Learners (UL) 0 0.5 1 5101520 Trial NumberProportion Recalled B Hi-Learners (OL) Lo-Learners (UL)0 0.5 1 1369 Retention Interval (weeks)Proportion Recalled Figure 3. Results of Experiment 1. (A) Study Phas e. The Hi-Learners likely overlearned the material, while the Lo-Learners failed to reach even adequate learning. (B) Test Results. Overlearning boosted recall, but th e difference between the Hiand Lo-Learners declined at longer retenti on intervals. Error bars indicate 1 standard error.
14 These results suggest that while overlearni ng does boost recall, the advantages are limited. One of these limitations is the redu ction of the benefits of overlearning as retention interval increases. For example, although ove rlearning led to a doubling of recall at the 1-week retention interval, this increase in recall decreased dramatically as retention interval increased. Furthermor e, this doubling of recall becomes less impressive when you consider that total study time quadrupled. Thus, even at the point of the largest increase in recall, the benef its of overlearning did not come close to equaling the increase in study time. Conseque ntly, overlearning is not an efficient study method. When interpreting these results, however, one caveat should be considered. In this experiment, participants did not learn to crite rion. That is, the de gree of overlearning was not strictly controlled for. Although we did not specifically require participants to reach a criterion of adequate learning a nd overlearning, the difference between the Loand Hi-Learners was arguably much greater than this. Consequently, any difference found between the Hi and Lo-Lear ners exaggerates the benef its of overlearning because the Lo-Learners failed to reach adequate learning.
15 Experiment 2 This experiment examined whether it is better to overlearn a small number of items or to underlearn a larger number of ite ms. That is, when absolute recall is more important than proportion of recall, is it bene ficial to spend your time studying more or less information? For example, as describe d in the introduction, s hould a teacher assign activities geared toward lear ning vocabulary words for the same 20 words every night for a week, or should she assign activities for 20 di fferent words each night, resulting in the studying of 100 words a week? To answer this question, college students devoted 12 minutes to learning a list of either 10 or 20 word-definition pairs. Thus, those participants who studied only 10 words in the 12-min study period were likely to ove rlearn the words. By contrast, those participants who studied 20 words in the 12-min study period were likely to underlearn the words. One or four weeks later, all participants returned for a test. Method Participants. The sample contained 88 University of South Florida undergraduates who participated in retu rn for extra credit. One hundred seven participants began the study, but 13 failed to return for the test, and four did not complete the task correctly. None part icipated in Experiment 1.
16 Design. Study list length (10 or 20) and reten tion interval (one or four weeks) were manipulated between subjects. Particip ants were randomly assigned to one of the four conditions. Procedure. The procedure used was similar to that used in Experiment 1, except that we varied the study lis t length rather than the total study time. Once again, participants attended both a st udy session and a test session In the study session, each participant was given a booklet and asked to st udy a list of 10 or 20 wo rd-definition pairs, as listed in Appendix B. Each definition wa s a single word with four letters or fewer (e.g., vizard-mask) in order to minimize writi ng time. Participants were given 120 s to study the word pairs. Following this 2-mi n study period, participants completed twenty 30-s feedback trials in which they tried to recall the definitions of the words. During each feedback trial, only five words were presented, with all words appearing equally often. Thus, the 10-word Learners saw each word 10 times, whereas the 20-word Learners saw each word only five times. Af ter 20 s of the recall phase of the feedback trials, participants unfolded the page, and the correct definitions appeared immediately to the right of their responses. Participants then studied the pairs for the remaining 10 s before turning the page to begin the next feedback trial. The words were randomly ordered each time so that participants w ould not merely memorize each definitions position within the list. The participants were tested either one or four weeks later. They were given the words and asked to recall the corresponding one-word definitions They were given four minutes to complete the test.
17 Results and Discussion The study data are shown in Figure 4A. As shown, most of the 10-word Learners (overlearners) were correctly r ecalling all ten definitions on each of the last five study trials. Because they underwent twice as many tr ials as that needed to reach a criterion of one correct recall, it is assumed they underwent 100% overlearning. In contrast, by the last study trial, the 20-word Learners ( underlearners) were recalling only 70% of the definitions correctly. Thus, these 20-word Learners failed to reach even adequate learning. Consequently, any difference f ound between the 10and 20-word Learners exaggerates the benefits of overlearning becau se the 20-word learners failed to reach adequate learning. The proportional test results are illustrated in Figure 4B. Not surprisingly, the 10word Learners, who studied each word twice as long as the 20-word Learners, recalled a greater proportion of words than the 20-word Learners, F (1,84) = 12.21, p < .01. However, as in Experiment 1, the benefits of overlearning decreased dramatically with retention interval, as indicated by the signifi cant interaction between retention interval and list length, F (1,84) = 6.15, p < .05. Interestingly, as shown in Figure 4C, when the absolute recall data were analyzed, we found that the 20-word Learners (underlearne rs) actually recalled more words than the 10-word Learners (overlearners), F (1,84) = 4.14, p < .05. Furthermore, as retention interval increased, the benefits of underlearning on absolute recall total stay ed constant, as demonstrated by the statistical nonsignifi cance of the list length x retention interval interaction (F<1).
18 A 10-word Learners (OL) 20-word Learners (UL)0.00 0.50 1.00 12345678910 Trial NumberProportion Recalled B 10-word Learners (OL) 20-word Learners (UL)0 0.5 1 14 Retention Interval (weeks)Proportion Recalled C 10-word Learers (OL) 20-word Learners (UL)0 5 10 14 Retention Interval (weeks)Absolute Recall Figure 4 Results of Experiment 2. (A) Study Ph ase. (B) Proportional Test Results. (C) Absolute Test Results. Error ba rs indicate 1 standard error.
19 As in Experiment 1, overlearning led to a higher proportion of words recalled than underlearning, but this difference decreased as retention interval increased. Once again, it seems that overlearning is not useful when material must be maintained over long retention intervals. Furtherm ore, overlearning appears to be inefficient. As shown in Figure 4C, the overlearners spent twice as much time studying each word, and still recalled fewer words absolutely than the underl earners at both one and four weeks. This finding suggests that, when given a set amount of time to study, instead of studying a small amount of material, you should study a la rger amount of material. In return, you will be able to recall more information absolutely.
20 Experiment 3 The results of Experiment 2 suggest that underlearning more information produces greater absolute reca ll totals than overlearning le ss information, but does this finding extend to extreme underlearning? The pur pose of Experiment 3 was to find such a limit, if one exists. As in Experiment 2, total study time was held constant. Study list length was manipulated, with participants st udying list lengths of 5 (EUL), 10 (UL), and 20 (OL). Because the extreme underlearning cond ition required very rapid presentation rates, we were required to use a computer. Using a computer based task also allowed us to precisely control for total time spent on each word. Method Participants. Participants were 32 undergraduates from the University of South Florida who participated in re turn for extra credit. Twelve additional students began the study but failed to complete the test phase. N one participated in Experiments 1 or 2. Design. List length (5,10, and 20 words) was a within-subjects variable. Procedure. The procedure was similar to that of Experiment 2. Participants attended both a study session and a test sess ion, and the study session was controlled by a computer. During the study session, each par ticipant completed a 5-word practice list before completing 5-, 10-, and 20-word experi mental lists, as listed in Appendix C. For each list, participants were gi ven a 1-s prompt to get ready to study and then shown each
21 word-definition pair (e.g. vizard-mask) for th ree seconds. After this initial study period, the participant completed eight (5-word Learne rs), four ( 10-word L earners), or two (20word Learners) feedback trials depending on the list length. Before each feedback trial began, a 1-s Get ready to remember prompt appeared. Then each word from the list appeared individually and the participant had five seconds to type in the correct definition. Then, if the partic ipant answered correctly, the word correct appeared on the screen. If the participant answered incorr ectly, the word incorrect appeared along with the correct answer. This feedback remained for three seconds in either instance. After every word on the list a ppearred, the participants rest ed for five seconds before beginning the next feedback tr ial. The words were randomly ordered each time so that participants could not merely memorize ea ch definitions position within the list. Furthermore, the order of the conditions wa s also random to prevent confounds such as fatigue and interference. One week later the participants were test ed by paper and pencil. They were given the words and asked to reca ll the corresponding one-word de finitions. They had five minutes to complete the task. Results and Discussion The results of the study phase are shown in Figure 5A. It appears that by the fifth trial, in the 5-word condition, most participan ts were recalling all five words correctly. Thus, it is assumed that by th e eighth trial, these partic ipants had experienced 100% overlearning. In contrast, in both the 10and 20word conditions, participants on average failed to reach adequate learning.
22 As shown in Figure 5C, we were succe ssful in finding an upper limit to the finding of Experiment 2. As in Experiment 2, as list length increased, absolute recall also increased, F (2,62) = 8.39, p <. 01. However, recall did not increase monotonically as list length increased. Post hoc tests attribute th e main effect of degree of learning to the differences between the overlearners and th e two underlearning conditions. There was however, no significant difference between th e 10and the 20-word conditions. Thus suggesting that the increase in study list leng th from 10 to 20 words did not result in an increase in recall. These results imply that while it is better to spend your tim e underlearning more information, there is a limit at which studyi ng more information in an allotted time will no longer increase recall. Once again, this experiment did not use a learning to criterion design, so the results should be interpreted cautiously.
23 A 5-word Learners (OL) 10-word Learners (UL)0 0.5 1 12345678 Trial NumberProportion Recalled B 0 0.5 1 5 (OL)10 (UL)20 (EUL) Study List LengthProportion Recalled C 0 5 5 (OL)10 (UL)20 (EUL) Study List LengthAbsolulte Recall Figure 5. Results of Experiment 3. (A) Study Phase. The 5-word Learners (overlearners) likely overlearned the words, whereas the 10-word Learners (underlearners) and the 20-word Learne rs (extreme underlearners) failed to reach even adequate learning. (B) Propor tional Test Results. The 5and 10-word Learners both recalled a higher pr oportion of words than the 20-word Learners. (C) Absolute Test Results. The 10-a nd 20-word Learners both recalled a higher absolute total of words than the 5-word l earners. Error bars indica te 1 standard error. 20-word Learners (EUL)
24 General Discussion Summary Although overlearning resulted in a boost in proportion of words recalled in all three experiments, closer insp ection of the results revealed that there were limitations to these benefits. More specifically, the be nefits of overlearning decreased at longer retention intervals, and overle arning also proved to be an in efficient study method. This observed decline in the benefits of overlearning at longer retention intervals is in agreement with past studies (Craig, Sternthal, & Olshan, 1972; Reynolds & Glaser, 1964). That is, while overlearners recalle d a much higher proportion of words than the underlearners one week after the study session, this difference declined with retention interval (i.e. the time elapse d between the study session and the test session), as shown in Figures 4B and 5B This result was consistent when tested using two different designs. In Experiment 1, overlearning was shown to be less effective at longe r retention intervals when study list length was held constant and total study time was varied. Likewise, Experiment 2 revealed a rapidly diminishi ng benefit of overlearning when study list length was varied, and total study time was held constant Thus, we found that overlearning proved less useful at long retention intervals wi th two different experimental designs. This convergence suggests that overlearning is not a good learning strategy when long-term retention is important.
25 In addition, overlearning also seems to be a highly inefficient study method, as suggested by past studies (Driskell et al ., 1992; Kratochwill, De muth, & Conzemius, 1977). In fact, in Experiments 1, 2 and 3, the percentage increase in recall gained from overlearning the material was less than the percentage increase in total study time. That is, the boost in recall due to overlearning was not proportional to the increase in total study time. In Experiment 1, for example, th e Hi-Learners studied four times longer than the Lo-Learners, and at best, outperformed them by only double (Figure 3B). Similarly, the overlearners (10-word Learners) in E xperiment 2 spent 100% more time studying each word than did the underlearners (20-word Learners), as shown in Figure 4B. In return for this effort, the overlearners recalle d only 70% more than the underlearners at the 1-week interval and only 26% more at the 4-week interval Likewise, in Experiment 3, the overlearners (5-word Learners) studied each word twice as long as underlearners (10-word Learners) and s till recalled only 8% more (Figure 5B ). This is evidence that the time committed to overlearning is nowhere near proportional to the benefits achieved. The inefficiency of overlearning becomes even more apparent when the absolute recall totals of Experiments 2 and 3 are examined. In Experiment 2, although twice as much time was spent learning each word in the 10-word list, participants recalled a higher number of words from the 20-word list (Fi gure 4C). Furthermore, this increase in absolute recall total (i.e. the total number of words recalled) remained virtually consistent across both retention intervals. Similarly, in Experiment 3, when less time was devoted to learning each word in both the underlearning (10-word) and the extreme underlearning (20word) conditions, participants neverthele ss recalled more than when the words were
26 overlearned (5-word condition) (Figure 5C). These results suggest that underlearning 10 or 20 words will probably result in greater absolute recall total than overlearning 5 words in the same amount of time. However, underlearning does not always produc e greater absolute recall totals than overlearning. That is, underlearning can be too extreme to result in any recall benefits. For example in Experiment 3, increasing th e list length from 10 to 20 while holding the study time constant did not produce an increase in absolute recall (Figure 5C). This suggests that the participants did not have enough time to study each word in the 20-word (extreme overlearning) condition. As a result, they did not benefit from the increase in list length. These resu lts are evidence, that in extreme measures, underlearning can be equally as inefficient as overlearning. In summary, while overlearning did boost re call on a test a week later, this boost in recall disappeared as retention interval le ngthened. Therefore, it may not be wise to spend time overlearning material if it needs to be recalled in the distant future. Furthermore, while overlearning led to increa sed recall when compared to underlearning, this increase in recall was never proportional to the increase in total study time. Thus, the time invested to overlearn was never proportiona l to the increase in recall. Finally, when more words were studied in a given amount of time, a higher number of words were recalled at test. Even more striking, this bene fit did not decrease with retention interval as did the benefits of overlearning. So, rega rdless of retention in terval, underlearning more words led to greater absolute recall.
27 Implications A lot of research has been done that could help improve certain facets of education. Unfortunately, this research is not often used to evaluate and update educational procedures. The results from the current studies have practical implications for education and learning in general. One problem often described by educat ors is that students do not retain information past test day. Cooper et al. (1996) expressed this c oncern of teachers by relaying that students forget a large amount of material during summer breaks. The finding that overlearning is only advantageous at short retenti on intervals suggests that, if the material being taught is something that should be remembered over long periods of time, as is most information taught in sc hools, perhaps overlearning is not the best method to use. For example, the purpose of teaching vocabulary is to give students a large, ever growing pool of terminology. Thus, rather than spending 30 minutes to overlearn a relatively short list of words, the present data suggest it might be better to underlearn a longer list. Thus, as in Experiments 2, this should lead to the retention of a larger number of words, regardless of retention interval. Another situation in which absolute to tal recall is important occurs when studying for the GRE. On the GRE, any of thousands of words could be presented to you. Thus, the total number of words recalled is importa nt because, as the total number of words you know increases, your chances to do well also increase. So, the findings of Experiments 2 and 3, suggest that underlearning may produce highe r absolute recall than overlearning.
28 Overlearning techniques are also ofte n used in math courses. Teachers may assign students 20 or more of the same type of math problem for homework in one night. Once students figure out how to complete the first problem correctl y, every subsequent problem constitutes overlearning. As eviden ced by our results in Experiment 1, while this overlearning may boost their scores in the immediate test, it may not produce longterm learning. Furthermore, as evidenced by the results of Experiments 1 and 2, the extra time students spend to overlearn is not likely to produce a proportional gain in recall. There are, of course, instances in whic h overlearning would be the best strategy. For instance, if the propor tion of information recalled were important, overlearning strategies would be the most useful. For example, it is important when administering CPR that you remember and complete all the steps. In this case, absolute recall is less relevant, because recalling every detail is crucial. Thus, for instances such as these, overlearning would be helpful, but as demons trated in Experiments 1 and 2, the material would have to be retrained peri odically to reduce forgetting. In conclusion, before deciding on a study method, be sure to consider the desired result. If proportion of recall is desired, and the retention interval is short, overlearning could be useful. However, if your goal is to retain as much information as possible regardless of retention interv al, underlearning a lot of info rmation would be optimal. Future Studies When planning future studies of the benef its of overlearning, a few considerations should be made. Unlike the present st udies, many learning experiments require participants to learn words to a criterion. That is, in the st udy session, the feedback trials
29 continue until the participant has correctly recalled every word the required number of times. For example, in another study we recently completed, participants were required to learn word pairs to a criterion of either 1 or 9 correct recalls before moving on. This process can be likened to the procedure often used when stu dying with flash cards. Thus, in the condition in which participants are lear ning to a criterion of one, if they correctly recall the definition of a word, that word is removed from the list. They continue this process until no words are left in the list. The use of this design will control for degree of learning. It will also allow us to test the a dvantages and disadvantages of overlearning at precise degrees. In each of these experiments, the sizes of the study list lengths were limited due to time constraints. In fact, th e longest study list was 20 word s. Study books for tests such as the SAT or GRE contain vocabulary lists thousands of words long. Thus, in future studies the parameters of list length and total study time shoul d be expanded. While there should not be qualitative differences in the results as compared to the results of our experiments, the results would more clos ely approximate the study process for such exams. The gender composition of the sample s hould also be considered in future experiments. In the present studies, many mo re females than males were tested. Thus, although we found no gender differences, this highly unbalanced ratio of females to males resulted in very low pow er for these tests. Even t hough gender differences usually do not exist in memory tasks, future studi es should use more balanced samples.
30 Finally, although it is not lik ely that school-aged childre n differ qualitatively in learning styles, these studies should be replicated with a sample of young children before we generalize these results to that populat ion. Results from school-aged populations will help answer questions about how children learn and retain in formation. They will also be useful in increasing classroom effectiveness.
31 References Ausubel, D. P., Stager, M., & Gaite, J. H. (1968). Retroactive facilitation in meaningful verbal learning. Journal of Educational Psychology 59 250-255. Bromage, B. K., & Mayer, R. E. (1986). Quanti tative and qualitative e ffects of repetition on learning from technical texts. Journal of Education Psychology, 78 271-278. Cooper, H., Nye, B., Charlton, K., & Lindsay, J. (1996). The effects of summer vacation on achievement test scores: A narr ative and meta-analytic review. Review of Educational Research 66 227-268. Craig, C. S., Sternthal, B., & Olshan, K. (1972) The effect of overlearning on retention. Journal of General Psychology, 87 85-94. Driskell, J.E., Willis, R. P., & Copper, C. ( 1992). Effect of overlearning on retention. Journal of Applied Psychology, 77 615-622. Fitts, P. M. (1965). Factors in comple x skill training. In R. Glaser (Ed.), Training Research and Education (pp. 177-197). New York: Wiley. Foriska, T. J. (1993). What every educator should know about learning Schools in the Middle, 3 39-44. Hall, F. H. (1989). Retention and forgetting: The functional appro ach continued. In Learning and Memory (2nd ed., pp. 317-345). Boston, MA: Allyn & Bacon.
32 Kratochwill, T. R., Demuth, D. M., & Con zemius, W. C. (1977). The effects of overlearning on preschool childrens rete ntion of sight vocabulary words. Reading Improvement, 14 223-228. Krueger, W. C. F. (1929). The effect of overlearning on retention. Journal of Experimental Psychology, 12 71-78. Murdock, B. B. (1960). The immediate retention of unrelated words. Journal of Experimental Psychology, 60, 222-234. Postman, L. (1962). Retention as a function of degree of overlearning. Science, 135 666667. Reynolds, J. H., & Glaser, R. (1964). Effects of repetition and spaced review upon retention of a complex learning task. Journal of Educational Psychology, 55 297308. Roberts, W. A. (1972). Free recall of word lists varying in length and rate of presentation: A test of total-time hypotheses. Journal of Experimental Psychology, 92 365-372. Schendel, J. D. & Hagman, J. D. (1982). On sustaining procedural skills over a prolonged retention interval. Journal of Applied Psychology, 67 605-610.
34 Appendix A List of city-country pairs for Experiment 1. City Country Axim Ghana Bari Italy Chiba Japan Doba Chad Lugo Spain Mago Tonga Pune India Rabat Malta Talara Peru Yumen China
35 Appendix B List of vocabulary words for Experiment 2. Word Definition Word Definition anta pier acrogen fern cerate wax beldam hag cicatrix scar cess tax excrescence wart elver eel fosse moat emmet ant peruke wig mentum chin stannum tin mome fool tippet cape ruga fold vizard mask salver tray weir dam talar robe Note: The overlearners studied the 10 wo rds in the left column. The underlearners studied all 20 words.
36 Appendix C List of additional vocabular y words for Experiment 3. Note: These words were used in addition to the words used in Experiment 2. Word Definition Word Definition acarus mite pneuma soul brad nail portent omen bullock ox roughleghawk cabal plot swagmanhobo concordat pact victual food kismet fate vitellus yolk occident west wheal welt phyllome leaf