DISPARATE EFFECTS OF REPEATED TESTING: Reconciling Ballard's (1913) and Bartlett's (1932) Results

Bartlett (1932) gave subjects a prose passage and showed how recall dropped when they were tested repeatedly. Ballard (1913), using poetry, and Erdelyi and Becker (1974), using pictures, reported improvements in performance (or hypermnesia) over repeated testing. We investigated two likely factors leading to the discrepant results: the type of material and the interval between tests. The primary cause of the differing outcomes is the interval between tests. In general, when the intervals between successive tests are short, improvement occurs between tests. When these intervals are long, forgetting occurs. The type of material used plays little role: Hypermnesia in recall of prose (even "The War of the Ghosts") occurred with short intervals between tests. We also report a striking confirmation of the power of tests to enhance memory: Repeated tests shortly after study greatly improved recall a week later. In the course of daily affairs, we are frequently asked or required to recollect the same events from memory. This repeated recollection may occur in educational contexts, as when people are tested on material during the course of a semester and later receive a cumulative final exam. Similarly, eyewitnesses to crimes may be queried repeatedly about what they observed by police, by friends, by lawyers, and then eventually in court. More commonly, we all repeatedly retell stories of favorite or notable events in our lives. Despite the ubiquity of such experiences, the strategy of repeatedly testing memory has not been the dominant method of studying how memories change over time. Rather, researchers have generally employed between-subjects comparisons, in which independent groups learn some material and then are tested at various times after the original learning to measure the course of forgetting. This design has the advantage of delayed tests being uncontaminated by recollections on previous tests, as would occur if a within-subjects design were used. However, the "contaminating" influences of previous tests can be of central interest in their own right, with the eyewitness testimony situation representing a prime example (Loftus, 1979). Therefore, the repeated testing of memory for the same events (without intervening study opportunities) also represents a valid, if less used, approach to changes in retention over time. Several experimental paradigms and research traditions have been used to examine effects of repeated testing on memory. For present purposes, we confine interest to those cases in which the tests were given without overt cues (free recall). The most famous example of this technique is that of Bartlett (1932), who gave college students the Indian folktale "The War of the Ghosts" and then tested them repeatedly. He presented no aggregate data to support his conclusions, but rather provided sample recall protocols from his subjects. He reported that their performance became increasingly poor over time; they forgot the story but reconstructed plausible accounts that were skewed to the more typical schema of a fairy tale, a style that was presumably more familiar to his students. Despite the fact that Bartlett (1932, pp. 3-4) criticized the pioneering work of Ebbinghaus (1885/1964), his conclusions about the course of forgetting were broadly consistent with Ebbinghaus's prior work. Of course, his theoretical emphasis was markedly different. Curiously, a second tradition of repeated testing research, begun by Ballard (1913), leads to the opposite conclusion from Bartlett's better publicized work. Ballard gave schoolchildren poetry and tested their memories both soon after learning and then again at various periods up to 1 week later. He reported that children frequently recalled lines of poetry on later tests that were not recalled on earlier ones, a phenomenon he termed reminiscence. Sometimes the gain in recall between tests outweighed forgetting, so overall performance improved. The basic phenomena were replicated (Brown, 1923; Williams, 1926) and became known in textbooks as the BallardWilliams reminiscence phenomenon (e.g., Osgood, 1953, pp. 564-566). ! Bartlett (1932) never cited these articles in Remembering, and never noted the discrepancy between these findings and his own. The topic of reminiscence fell into disrepute in the 1940s and 1950s, when researchers failed to obtain absolute gains in performance across repeated testing under some conditions (Buxton, 1943).2 The topic has enjoyed a resurgence of interest in the past 20 years, owing largely to the work of Matthew Erdelyi, who developed standard experimental settings that reliably produce the phenomenon. Erdelyi and Becker (1974) had subjects study either a large set of pictures or a set of words (the names of the pictures). Subjects received three tests without recall cues of any sort. Each test lasted 7 min, with short breaks Address correspondence to the authors at the Department of Psychology, Rice University, Houston, TX 77251-1892. 1. Although Williams (1926) did replicate Ballard's (1913) reminiscence phenomenon under highly similar conditions, he actually failed to obtain the effect in most conditions. 2. Payne (1987) noted that reminiscence was redefined by researchers after Ballard as net improvements in recall over tests. They often failed to find such effects and concluded that the phenomenon was unreliable. Yet when defined as Ballard (1913, pp. 17-18) did originally ' 'remembering one or more [items] that were not remembered in a [prior] test" the phenomenon is regularly observed in most studies. Curiously, Bartlett (1932, pp. 67, 72) provided only two brief mentions of subjects correctly recalling on a later test a fact that was not recalled previously. 240 Copyright © 1992 American Psychological Society VOL. 3, NO. 4, JULY 1992 This content downloaded from 128.252.3.179 on Mon, 08 Jun 2015 15:15:26 UTC All use subject to JSTOR Terms and Conditions PSYCHOLOGICAL SCIENCE Mark A. Wheeler and Henry L. Roediger, HI occurring between tests. Subjects recalled an increasing number of pictures across the three tests, but overall recall of words was stable. Erdelyi and Becker labeled the improvement in recall across tests hypermnesia to distinguish the phenomenon from amnesia, or forgetting, over time. Others have reported hypermnesia even for verbal materials (Klein, Loftus, Kihlstrom, & Aseron, 1989; Payne & Roediger, 1987). Roediger and Thorpe (1978) obtained hypermnesia for verbal stimuli and distinguished between reminiscence and hypermnesia, in accord with the definitions of Ballard (1913) and Erdelyi and Becker (1974). Reminiscence refers to recall on a later test of items that could not be recalled on an earlier test. Virtually all studies of repeated free recall have shown this phenomenon. However, reminiscence is often neutralized or outweighed by intertest forgetting so that there is no overall net improvement in recall across tests, especially with verbal materials and short amounts of recall time (Tulving, 1967). Nonetheless, net improvements in recall across tests, or hypermnesia, can be reliably obtained in many situations, particularly those in which overall levels of performance are fairly high (Payne, 1987). Various theoretical approaches for explaining hypermnesia have been explored, involving, for example, an imaginal encoding format (Erdelyi, 1984) and the retrieval dynamics of cumulative recall curves (Roediger & Challis, 1989; Payne, 1986, 1987). Raaijmakers and Shiffrin (1980, 1981) proposed a variety of mechanisms for exploring hypermnesia. The impetus for the present experiments, however, was not to explain reminiscence or hypermnesia. Rather, we sought to understand how two research techniques that appear so similar on the surface could lead to such different results and interpretations about how memory changes over time. In the tradition of work begun by Bartlett (1932), subjects learned prose and were typically tested once about 15 min later, and then again either days or weeks later. In the research tradition initiated by Ballard (1913) and rejuvenated by Erdelyi and Becker (1974), subjects were more typically required to study a list of pictures or words and were tested repeatedly over short intervals of time during the experimental session. Two obvious differences between the traditions that might explain the discrepancy in outcomes are the type of material (prose or lists) and the amount of time between tests (short or long). We investigated both sets of factors in the current experiments. It is not obvious from prior work which should be more important. Ballard reported reminiscence with passages of poetry that may be considered connected discourse, like prose. He also obtained the effect with a 24-hr interval between successive tests. Similarly, Erdelyi and Kleinbard (1978) obtained hypermnesia for lists of pictures over a week after presentation by testing subjects three times a day. Finally, Roediger, Payne, Gillespie, and Lean (1982, Experiment 3) found hypermnesia with well-organized categorical materials retrieved from permanent memory (U.S. presidents, sports, and birds). In Experiment 1, subjects studied a series of pictures that were embedded in a story read orally. They then took a series of up to six tests (depending on the condition) shortly after the study phase and after a week. We attempted to replicate both the findings of Bartlett (1932) sharp forgetting between successive tests and those of Erdelyi and Becker (1974) and Ballard (1913) improvements over tests. In Experiment 2, we attempted to obtain hypermnesia in recall of a prose passage when tests were repeatedly given over short delays.