The Effect of Recall on Recognition an Empirical Investigation of Consecutive Learning Measures

Surendra N. Singh, University of Kansas
Michael L. Rothschild, University of Wisconsin
ABSTRACT - As uncertainty grows as to the appropriateness of recall versus recognition in advertising copy tests, practitioners are considering the simultaneous use of both tests. Such a practice is economically viable if the same subject can do both tasks. This paper examines the impact of a preceding recall test on a following recognition test. An empirical test shows that in most cases recall does not effect recognition. An explanation is proposed as to why recall does, in some cases, have an impact.
[ to cite ]:
Surendra N. Singh and Michael L. Rothschild (1983) ,"The Effect of Recall on Recognition an Empirical Investigation of Consecutive Learning Measures", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 271-276.

Advances in Consumer Research Volume 10, 1983      Pages 271-276

THE EFFECT OF RECALL ON RECOGNITION AN EMPIRICAL INVESTIGATION OF CONSECUTIVE LEARNING MEASURES

Surendra N. Singh, University of Kansas

Michael L. Rothschild, University of Wisconsin

ABSTRACT -

As uncertainty grows as to the appropriateness of recall versus recognition in advertising copy tests, practitioners are considering the simultaneous use of both tests. Such a practice is economically viable if the same subject can do both tasks. This paper examines the impact of a preceding recall test on a following recognition test. An empirical test shows that in most cases recall does not effect recognition. An explanation is proposed as to why recall does, in some cases, have an impact.

INTRODUCTION

Recall is currently the most widely used measure of learning from print and broadcast messages, although many have suggested that it is not the most appropriate one (Haskins 1964; Krugman 1972). In recent years, the idea that recognition may be an alternative or better measure of learning in certain low involvement situations has been gaining ground (Krugman 1972; Bettman 1979; Singh 1982). Advertising practitioners have also shown a positive inclination to explore recognition as an advertising response variable. For example, Bruzzone Research Company is successfully using recognition as a measure of learning from television commercials.

The difference between recall and recognition is that recall is a procedure in which a subject is first given a set of information and is then given some cue to retrieve and report that information. Recognition differs from recall in that during the test, the subject is presented with the material that is supposed to be remembered. Hence, in recall the individual must reconstruct the stimulus which is not present; in recognition information allowing one to differentiate or discriminate is given (Bettman 1979).

At this point it is not clear which of the two measures, recall or recognition, is the most appropriate; researchers and practitioner; have tried, therefore, in many situations to measure both recall as well as recognition. There are two possible approaches to secure both recall and recognition scores:

a. Measure recall and recognition separately from two groups of subjects exposed to the same stimuli.

b. Measure both recall and recognition on the same group of subjects.

The latter approach is preferable for two reasons:

(i) it reduces the sample size by one-half and consequently reduces cost,

(ii) it greatly increases precision since the design is now within subjects rather than across subjects.

When measuring both recall and recognition on the same group, subjects are given the recall tests prior to the recognition tests. (If recognition tests are given first, they will artificially inflate the recall scores.) The objective assessment of the recognition scores will only be possible, though, if prior recall does not effect subsequent recognition. If it does, then two different Samples must be used to secure recall and recognition scores. This paper examines whether prior recall effects later recognition scores.

LITERATURE REVIEW

This section is organized into three parts which describe (1) the various types of recall and recognition tests, (2) models comparing recognition and recall, and (3) literature which specifically relates to the issue of effect of prior recall on later recognition.

Types of Recall and Recognition Tests

In recall tests, the subject is first exposed to a stimulus and is then given some cue to retrieve it. If the only cue given is a temporal cue (e.g., what commercials did you see while viewing the Brady Bunch?), then the test is called an unaided recall test. However, if information other than the temporal cue is provided then the test is called an aided recall test. There are different degrees of aid depending upon the extent of the retrieval cues provided. In a recognition test, on the other hand, the retrieval cue provided is the stimulus itself. There are basicaLly three types of recognition tests:

'Yes'/'No' Recognition Test. Subjects are shown 2 series of items one at a time. As each item appears, the subject is to respond "yes" if it was on the original test and "no" if it was not. Usually, half of the items on the test are from the original list, and the other half are distractors.

Forced-Choice Test. In this procedure, subjects see two or more items at a time during the test. The task is to pick out the original stimulus. If the subject sees two items at a time, then the test is called a two alternative forced-choice, if three, a three-alternative forced-choice, and so on. The non-stimulus items presented during the forced choice test are called distractor items or distractors.

Batch-Testing Procedure. In this method, all the original stimuli and all the distractors are presented at once. The subject then tries to indicate which items were on the original list.

Along with making a "yes"/"no" response, the subject is often required to indicate the degree of confidence with which she/he recognizes the item. Typically the confidence scale has three or four points ranging from absolute certainty to random guess. The confidence ratings are then used to transform the binary yes/no identification responses into a more sensitive 6 or 8 point scale. Apart from yielding a more sensitive recognition test, confidence ratings add a significant amount of information over and above what is obtained by the binary identification response (Clarke 1964). Since the confidence rating procedure yields the same criterion relationships as obtained by the binary yes/no decision procedure (Pollack and Decker 1964), it is possible to carry out statistical analyses on the transformed data without distortions. Finally, confidence ratings help to correct for random guessing which is one source of bias in recognition tests.

Models Comparing Recognition and Recall

The most prominent models comparing recognition and recall are the "two-stage theories" and the "episodic ecphory view." Among the earlier two stage theories, the most popular one is the "Dual Process Hypothesis" (Anderson & Bower 1972; Kintsch 1970). According to this view, recall consists of search and recognition. The subject, when faced with a recall problem generates a number of prospective candidates for recall during the search process and then recognizes one of them. The decision stage of recall is assumed to involve the same processes as are involved in recognition. Since this model assumes that recall consists of two stages (retrieval and search) and recognition only one (search), it also assumes that the two processes are different.

The dual process hypothesis assumes that recognition is a subprocess of recall. This implies that if an item can be recalled, it should also be recognizable and that the total number of recognized items should always be at least as much as the total number recalled.

Tulving and Thompson (1973) and Wiseman and Tulving (1976) have provided data that violate both these implications. The phenomenon has been called "recognition-failure of recallable words." Basically they have demonstrated that there are certain situations in which an item can be recalled but not recognized. Later two stage theories, such as those of Anderson and Bower (1976) and Kintsch (1974) (called "modified dual process theories"), are able to account for recognition failure of recallable words. According to the modified dual process view, the subject encodes propositions about the list words at the time of the encoding of items. These propositions are of the form, " is on the list." Recall occurs when a subject searches for and retrieves a proposition, and decides (recognizes) that the retrieved item was on the list.

Recognition testing may superficially seem the same as this decision process, but, in theory, it has its own search component. In recognition, the subject is given a potential list word, which sends the subject to a location in memory. At that point, propositions connected to that location are searched. The subject may or may not find the critical proposition created during the initial encoding. The modified dual-process model suggests that recognition testing, where a to-be-remembered item is actually presented, does not guarantee that the subject will find the appropriate location in memory for that item. Instead, there may still have to be a search for a key proposition about the item (Klatzky 1980). Thus the modified dual process model assumes the importance of context in which encoding takes place and it also assumes existence of retrieval processes in recognition.

Tulving (1976) has presented an alternative view (episodic ecphory) which is, in some respects, similar to two stage theories of recognition and recall. "The basic assumption in [the episodic ecphory view] is that remembering of an event both in recall and recognition comes about as a consequence of interactions between trace information, after effects of the initial encoding of the event, and appropriate retrieval information from other sources. Recognition and recall differ only with respect to the exact nature of the retrieval information available to the rememberer. In recognition, retrieval information is presented by a literal copy of the event or item to be remembered; in recall, the retrieval information is contained in cues other than copy cues. In other respects the process of utilization of trace information in the act of retrieval is thought to be essentially the same for recall and recognition" (Tulving 1976, p.37). Notice that there are two important sources of information here - memory trace and ecphoric information. "The probability of recollection depends on the amount of overlapping information between the [episodic] trace and ecphoric cues" (Tulving 1976, p.63). Tulving defines episodic trace as the perception and encoding of a unique event, such as the occurrence of a word in a list. This results in the creation of a unique trace in the episodic memory system. The properties of the episodic trace includes not only the perceptible properties of the unique event and its semantic meaning, but also the context in which it occurs and the specific encoding operations performed on the input into the system.

Ecphoric information refers to specific and temporal cues manipulated in the experiment. Recall and recognition are different inasmuch as the retrieval information that is present at the time of recall is different from that at the time of recognition. They are the same inasmuch as in both cases ecphory is a consequence of_ appropriate combining of trace information with retrieval information. [Tulving 1976. p.681]

Figure 1 presents a description of the episodic ecphoric view. The subject perceives the stimulus event (E1) and it is encoded into a memory trace (TE). This memory trace may be recoded (into trace TR) as a result of additional input (E2) into the memory system. When the recoded trace (TR) is contacted, activated, matched or complemented by the information provided by the retrieval query (together with one or more specific cues (Q) ), stored information is retrieved; this results in conscious memory. At this point, if conditions warrant it, the conscious contents of the memory are overtly reported (R).

FIGURE 1

EPISODIC-ECPHORY VIEW

The only difference between the two stage theories (especially of Kintsch, 1974) and Tulving's Episodic/Ecphory View is that while two stage theories insist that recall involves two stages of retrieval and search, the ecphoric view does not make any such distinction. It seems that as long as modified dual process theories have accepted the involvement of retrieval processes in recognition, the notion of two stages in recall is no longer useful. Hence this paper uses the episodic ecphory view.

Literature Review Relating to Effect of Prior Recall on Recognition

In examining prior work in this area, one finds a mix of findings. In a study by Dallett, et al. (1968), subjects were shown a series of pictures and were tested with another series in which one half of the original pictures appeared unchanged while the other half were replaced with similar but different pictures. The subjects were tested for their recognition of the pictures. One group of subjects was given the title of the picture and requested to recall it before recognition, while the other group was tested for recognition alone. Both immediate and one week delayed retention intervals were used. Results showed that giving the title of the picture and requesting recall before recognition testing made no difference in recognition accuracy. Dallett, et a . Is explanation of these results was that "prior recall either does not affect recognition, or helps and hinders it to an equal degree." ( p . 315)

Darley & Murdock (1971) used word lists and showed that prior recall had no effect on subsequent recognition. Subjects were shown 10 lists of words. Each list was either followed by an immediate free recall test or was not tested. Half the subjects in each condition received a final free recall test on the words from all 10 lists; the other half received a three-alternative forced choice recognition test. Initial testing facilitated retrieval on the final recall test but had no overall effect on recognition performance.

The results were explained by Darley and Murdock in terms of information availability vs. accessibility in memory. The probability that an item is retrieved from memory can be considered to be a function of both the availability of the item's trace in memory storage and the accessibility of that trace. Since in the recognition test the item is physically present (the ultimate retrieval cue), its memory trace is felt to be more accessible. If the prior recall test affects recognition it would have an effect on availability. Since recognition was not affected, it was concluded that availability (or trace storage) of an item's memory trace is not affected by prior recall test. However, since prior recall does facilitate subsequent recall, it would imply that prior testing aids later retrieval by increasing accessibility of items stored in memory. Lockhart (1975) has shown a facilitation effect of prior recall on the recognition of the last few items of a word list. However, he concludes that effect of recall on subsequent recognition frequently will be concealed because recallable items will usually be recognized even without the aid of the recall test. Thus, he made a distinction between effect of recall on recognition vs. the effect of recallability on recognition. Rabinowitz, et al. (1977) also provide evidence to support the idea that recalled words are more accurately and more quickly discriminated from new items in a recognition test than are nonrecalled words.

Postman, et al. (1948) gave two groups of subjects six trials on a list of 48 nonsense syllables. Subjects in group l were tested for recognition followed by a 10 minute recall test. For group 2 the order of tests was reversed, i.e., recall was followed by recognition. Prior recognition facilitated recall but prior recall depressed recognition. The explanation offered for lower recognition scores after 10 minutes of recall testing was that weak traces became weaker during the 10 minute delay.

Belbin (1950) exposed a group of 64 subjects to a poster on a blank wall . One half of the subjects received a recall test ("prompted by standardized questions"). The other half engaged in an unrelated activity following the viewing and took the recognition test after the same lapse of time. Prior recall had a depressing effect on recognition. Belbin explained this depression in recognition scores on the basis of errors in the recall test. On the recognition test the absence of some erroneously recalled detail or the presence of some non-recalled detail seemed to determine the recall groups' (group 1) rejection of the pictures as being the same. Kay & Skemp (1956), repeated Belbin's experiment with some changes and came up with similar but much less dramatic results.

Hanawalt & Tarr (1961), using 23 statements (e.g., "Hotel Steaks Are Big") as stimuli, tested the effect of prior recall on a 5 alternative recognition test at immediate (8 minute) and delayed (48 and 52 hours) retention intervals. Results suggested that a depressing effect was present when recognition followed immediately after recall but the effect upon delayed recognition was that of facilitation. Hanawalt & Tarr explain the differences between their results and that of Belbin's (1950) on the basis of differences in learning situation, the recall activity, and the type of recognition test.

The studies reviewed above have given- different reasons for the results obtained. Some have offered very superficial explanations (Dallett, et al. 1968) and others have given more profound explanations (Darley & Murdock 1971). However, none of these studies has offered any satisfactory explanation for the effects obtained on the basis of any theoretical model of recall and recognition. In the following section a hypothesis is developed concerning effect of prior recall on recognition based on Tulving's Episodic Ecphory View.

Hypothesis Formulation

Tulving's Episodic Ecphory View suggests that recognition and recall differ only with respect to the exact nature of the retrieval information available to the rememberer. The probability of recollection depends on the amount of overlapping information between episodic trace and ecphoric cues.

At the time of the recall, the subject has a memory trace, TR, of the unique event to be recalled. The only ecphoric information provided is the temporal cue, QT. Immediately after the recall test, the subject is again given the temporal cue, QT, but also a literal copy of the event or item to be remembered. Thus, the total ecphoric information provided is (QT + Q) where Q is all other ecphoric information other than QT.

The temporal cue, QT, is being provided in both the recall and recognition tests but TR remains the same before and after recall. Hence, there should be no effect of prior recall on recognition because the only way prior recall can have an effect on recognition is if prior recall changes trace TR. We assume that this does not happen. The following hypothesis is tested:

H1: Prior free recall will have no effect on immediately following forced-choice recognition

where:

(a) Recall is free (not cued) recall, i.e., the only ecphoric cue provided is a temporal cue.

(b) Recognition is a force-choice test.

While the literature presented above is extensive, it is felt that an introduction to the concept is necessary for an appreciation of the issue being tested. There are several problems mentioned in this review which are not addressed by the study which follows, but are to be dealt with in future work. The current work is reported because it will help to set methodological groundwork for following work.

METHOD

The experimental design was a 2 (presence or absence of recall test) x 2 (types of recognition tests: 5 alternative and 9 alternative tests) x 3 (levels of repetitions: 1, 2 and 4) split plot design. The first two factors, presence or absence of recall and type of recognition test, were between subject factors whereas level of repetition was a within subject factor.

Procedure

Subjects were recruited and asked to watch a 30-minutes news show on a videotape. Viewing took place in small groups of about 15 subjects. The cover story was that the study concerned a comparison of three late night news shows. To aid the subjects in evaluating television news, they would be shown a videotape consisting of ten minutes from each of three television stations from other parts of the country. After viewing was over, subjects were asked to fill out an evaluation form comparing the news from the three stations. Subjects were asked to come back after two weeks to evaluate cable affiliate news shows. At this second session, instead of showing cable affiliate news shows, subjects were tested for recall and recognition.

Subjects were divided into two groups of approximately equal size. Subjects in group 1 were given an unaided recall test of commercials seen on the videotape. Following the recall test, half of the subjects in group 1 received a 9 alternative verbal recognition test while the other half received a five alternative verbal recognition test. Subjects in group 2 received no recall test but were given the verbal recognition test; again half were given a 5 alternative test and half a 9 alternative test. Finally, each subject filled out a questionnaire providing demographic information. At that point, subjects were debriefed and the purpose of the experiment was revealed.

Stimulus Materials: Commercials

Three thirty-second predominantly rational commercials representing three low-involvement product categories were selected. Rational commercials were used because they would give subjects something to write about for recall and recognition of "claims." The three commercials were selected out of an original pool of 200 through a two stage process using both expert and nonexpert judges. None of the selected commercials ever been aired in the test city and none of the brands represented in the commercials were marketed in the test CitY

Stimulus Materials: Overall Videotape

Commercials were embedded in a 30-minute tape of program material. 1* or three commercials were inserted between the news and weather and between the weather and sports of each of the three newscasts. The first two and last two commercials on tape were fillers to avoid primacy and recency effects among experimental commercials. No commercial was shown twice in a row. The tape contained one experimental commercial at one repetition level, one at two and one at four repetition levels.

Subjects

Seventy-eight undergraduate students were recruited for the experiment and randomly assigned to cells. Cell sizes varied from 15 to 22.

Measurements

As indicated above, one half of the subjects received an unaided recall test first, followed by either a five or nine alternative verbal recognition test. The other half did not receive any recall test but was tested on either five or nine alternative recognition tests.

Recall Test. Subjects were asked to recall the product category, brand name, and claim(s) for as many of the commercials shown to them two weeks earlier as possible. There were three test plus four filler commercials.

Recognition Test. Both the nine and five alternative recognition tests involved recognition of product category, brand name and claim in sequential tests. Subjects also indicated the degree of confidence they had in their answer on a three-point scale from "absolutely confident" to "not confident."

The forced choice test was used in this study to avoid the response biases of the simpler "yes-no" and "batch testing" procedures (Shepard and Cheng 1963). By increasing the number of alternatives, the chance probability level was felt to be reduced.

RESULTS

Tables 1 through 3 show the recognition scores. All scores represent a composite of the three experimental commercials so as to avoid commercial specific results and to increase generalizable results. The entries in these tables represent the mean score across subjects and commercials on the -3 to +3 confidence/accuracy scale described above. Analysis of variance results for the data are summarized in Table 4. There are two error terms in the ANOVA Table; the first error term relates to the between factor effects; the second error term relates to the within factor effects and its interactions with between factors.

The major concern of this paper was to examine the effect of recall on recognition. Only those results which relate to this issue are discussed.

Effect of Prior Recall on Product Category Recognition

There was no main effect of the testing condition (presence/absence of recall test) on product recognition (F = .06; P < .808), nor was there any interaction between the testing condition and the type of recognition test administered (F - 0.32; P < 0.571). Likewise, there was no interaction between testing condition and repetitions of commercials (F = 1.24; P < 0.292). There was no three way interaction (F - 0.04; P < 0.964).

Effect of Prior Recall on Brand Name Recognition

The testing condition did have a significant main effect on brand name recognition (F = 4.41; P < 0.039). The recognition-only group did generate a significantly higher score than did the recall and recognition group. There was no significant interaction between testing condition and type of test administered (F = 0.50; P < 0.482), or between testing condition and repetitions of commercials (F - 0.02; P < 0.982). There was no 3-way interaction (F - 1.61; P < 0.204).

Effect of Prior Recall on Claim Recognition

There was no significant main effect of testing condition on claim recognition (F - 0.11; P < 0.736) and no interactions of testing condition with the type of recognition test administered (F - 1.35; P < 0.249) or with repetitions of commercials (: = 0.15; P < 0.863).

Again, there was no three-way interaction (F - 2.05; P < 0.133).

DISCUSSION AND CONCLUSION

It is clear from the results that prior recall neither has a facilitating effect nor depressing effect on recognition scores for product recognition and claim recognition. Recognition scores from brand names were significantly lower when preceded by the recall test. Subjects in the recognition only condition correctly recognized significantly more brand names than did subjects in the recall plus recognition condition. The results offer only partial support to the hypothesis that prior recall will not effect later recognition.

It is possible that errors in recall are compounded in the recognition task, as suggested by Belbin (1950) but that would not account for a difference in effect across product, brand and claim. It is also possible that the time spent doing the recall task caused a weakening of memory traces to occur; this did not occur in the straight recognition task. This explanation, put forth by Postman, et al. (1948) also does not seem reasonable here because it is not consistent across product, brand and claim, and because this short additional delay seems negligible given the two week delay between viewing and responding.

What seems more plausible is that there is some difference between product, brand and claim which is independent of the hypothesis. Two differences seem likely; both suggest that the brand name task was more difficult than either the product or claim task.

Products (salad dressing, trash bags, pie) and claims (lower in calories, stronger, natural ingredients) were familiar to the subjects, but all the brand names were novel. In addition, the similarity of distractors was not tested; the brand name test may have been more difficult.

TABLE 1

PRODUCT CATEGORY RECOGNITION  (SCORES TRANSFORMED TO +3 TO -3)

These issues do not explain why recall would inhibit recognition but do suggest that the brand name test could differ from the product and claim test. These issues could, then, interact with either the Belbin or Postman hypotheses to show that in cases of difficult memory searches recall could suppress recognition but in cases of easier memory search recall would not suppress recognition.

Future researchers may wish to control for difficulty of task (novelty of stimulus; similarity of distractors) so that the recall recognition relationship is more constant.

TABLE 2

BRAND NAME RECOGNITION  (SCORES TRANSFORMED TO +3 TO -3)

TABLE 3

CLAIMS RECOGNITION   (SCORES TRANSFORMED TO +3 TO -3)

TABLE 4

ANALYSIS OF VARIANCE RESULTS

REFERENCES

Anderson, J. R. and Bower, G. R.- (1972), "Recognition and Retrieval Processes in Free Recall," Psychological Review, 79, 97-123.

Anderson, J. R. (1974), "A Propositional Theory of Recognition Memory," Memory and Cognition, 2, 3, 406-12.

Belbin, E. (1950), "The Influence of Interpolated Recall Upon Recognition," Quarterly Journal of Experimental Psychology, 2, 163-69.

Bettman, J. R. (1979), "Memory Factors in Consumer Choice: A Review, Journal of Marketing, 43, 2, 37-53.

Clarke, F. R. (1964), "Confidence Ratings, Second Choice Responses, and Confusion Matrices in Intelligibility Tests," in J. A. Swets (ed.), Signal Detection and Recognition by Human Observers, New York: John Wiley & Sons, 620-48.

Dallett, K., Wilcox, S. G. and D'Andrea, L. (1968), "Picture Memory Experiments," Journal of Experimental Psychology, 76, 2, 312-20.

Darley, C. F. and Murdock, B. B., Jr. (1971), Effects of Prior Recall Testing on Final Recall and Recognition," Journal of Experimental Psychology, 91, 66-73.

Hanawalt, N. G. and Tarr, A. G. (1961), "The Effect of Recall on Recognition," Journal of Experimental Psychology, 62, 361-67.

Haskins, J. B. (1964), "Factual Recall as a Measure of Advertising Effectiveness," Journal of Advertising Research, March, 2-8.

Kay, M. and Skemp R. (1956), "Different Thresholds for Recognition: Further Experiments on Interpolated Recall and Recognition," Quarterly Journal of Experimental Psychology, 8, 153-62.

Kintsh, W. (1970), "Models for Free Recall and Recognition," D. A. Norman (ed.), Models of Human Memory, New York: Academic Press.

Kintsh, W. (1974) The Representation of Meaning in Memory , Potomac, Maryland: Lawrence Erlbaum Associates.

Klatzky, R. L. (1980), Human Memory: Structures and Processes. San Francisco: W. H. Freeman and Company.

Krugman, H. E. (1972), Why Three Exposures May be Enough," Journal of Advertising Research, December, 12, Nb. 6. 11-14.

Lockhart, R. S. (1975), "The Facilitation of Recognition by Recall," Journal of Verbal Learning and Verbal Behavior, 14, 253-58.

Pollack, I. and Decker L. R. (1964), "Confidence Ratings, Message Reception and Receiver Operating Characteristics," J. A. Swefs (ed.), Signal Detection and Recognition by Human Observers. New York: John Wiley & Sons, 592-608.

Postman, L., Jenkins, W. O. and Postman D. L. (1948), "An Experimental Comparison of Active Recall and Recognition." American Journal of Psychology, 61, 511-19.

Rabinowitz, J. C., Mandler, G. and Patterson K. E. (1977), "Determinants of Recognition and Recall: Accessibility and Generation," Journal of Experimental Psychology General, 106, 302-29.

Shepard, R. N. and Chang J. J. (1963), "Forced Choice Tests of Recognition Memory Under Steady State Conditions," Journal of Verbal Learning and Verbal Behavior, 2, 93-101.

Singh, S. N. (1982), Recognition As a Measure of Learning From Television Commercials, Unpublished Dissertation, University of Wisconsin-Madison.

Tulving, E. and Thomson D. M. (1973), "Encoding Specificity and Retrieval Processes in Episodic Memory," Psychological Review, 80, 352-73.

Tulving, E. (1976), "Ecphoric Processes in Recall and Recognition," in J. Brown (ed.), Recall and Recognition, New York: John Wiley & Sons, 36-73.

Wiseman, S. and Tulving E. E. (1976), "Encoding Specificity: Relation Between Recall Superiority and Recognition Failure," Journal of Experimental Psychology: Human Learning and Memory, 2, 349-61.

----------------------------------------