The mnemonic influence of self-cues on narrative recall

A major issue in educational design is the provision of techniques which enhance learners' memory processes. A mnemonic advantage has been observed in cognitive research for self-cues. Self-cues are one or a few words provided by the learner, rather than a third party, about a paragraph. This study explored the effect of self-cues on the recall of narrative texts in learners of English as a second language. Under two general conditions, participants either read and recalled two narrative texts or read them and provided self-cues before recall. The results indicated that: a) self-cues facilitated the quantity of the recall (i.e., number of paragraphs recalled), regardless of whether or not participants were allowed to inspect their cues during recall, b) self-cues had no significant effect on the completeness with which each paragraph was recalled, c) paragraphs recalled at an earlier stage of output were more complete than those recalled at a later stage.

Educational design is concerned with designing effective input to the human cognitive system. An understanding of how the cognitive system works would thus contribute to an effective design and implementation of instructional materials (Brien & Eastmond, 1994). Among aspects of the cognitive system which are highly relevant to the major components of curricula are memory, thinking and problem solving (Brunning, Schraw, & Ronning, 1999).

As far as memory is concerned, the relevant issue for educational design has been how to enhance learners' memory skills. This is in fact the area where educational theory can draw on the findings of cognitive research. In other words, research on memory enhancement may serve as an interface between cognitive research and educational theory.

Within the framework of cognitive research on memory, the present study was designed as an extension of research on the effect of self- versus other-generated cues on memory. A new dimension in the present study was the recruitment of English as second language (ESL) learners, instead of native speakers of English, as informants. The rationale behind this selection was to see if language status (i.e., first or second) would have a bearing on the effectiveness of memory processes.

Background

Cognitive research has shown that self-generated stimuli are recalled and recognized better than other-generated stimuli (Kinjo & Snodgrass, 2000; Moshfeghi & Sharifian, 1998a, 1998b; Gorrell, Tricou, & Graham, 1991; Hill, Shwob, & Ottman, 1993; Hill, Allen, & Gregory, 1990; Kim & Van Dusen, 1998). Self-generated items are those which are either partially or totally provided by the learner, usually with the help of a cue, whereas other-generated items are provided externally. A classic demonstration of this phenomenon comes from Bobrow and Bobrow's study (1969). These investigators asked participants to commit to memory simple subject-verb-object (SVO) sentences. There were two conditions of interest. In condition 1, participants were provided with sentences written by the experimenters. In condition 2, participants had to generate sentences to connect the subject nouns and object nouns. After studying the sentences, the participants were prompted with the first noun (i.e., subject) and were required to generate the second one (i.e., object). In condition 1, participants recalled 29% of the nouns; in condition 2 (i.e., generation of sentences by participants), they recalled 58% of the nouns. Presumably, in generating their own sentences, participants had to think more carefully about the meaning of the two nouns and their possible interrelationships.

Van Dam and his colleagues (van Dam & Brinkerink-Carlier, 1988, 1989, 1990; van Dam, Brinkerink-Carlier, & Kok, 1987) found a facilitating effect on text recall for self-generated cues. In one of their earlier experiments (van Dam, Brinkerink-Carlier, & Kok, 1987), they found that presentation of a self-generated list of short indications of the paragraphs of a text facilitated the retrieval of the text's contents during recall.

In another experiment (van Dam, Brinkerink-Carlier, 1988), participants studied an expository text that comprised a series of paragraphs referring to a common superordinate topic and were asked to formulate retrieval cues based on the distinct paragraphs of the text. Exposure to these retrieval cues at the onset of the recall facilitated the recall of these paragraphs. It may be argued that formulating these paragraph cues prior to recall might be functionally equal to an act of free recall and, therefore, two consecutive acts of free recall may result in the recall of an equal number of paragraphs.

Anderson and Armbruster (1984) report that such strategies can be effective only when they are encoded appropriately during learning. Under ordinary conditions of learning, assigning labels to the paragraphs of a text is not necessarily performed systematically. Some chunks of information may not attract attention and will, therefore, be stored with relatively low accessibility in memory. Asking participants to generate a list of retrieval cues prior to recall may evoke systematic labelling of the paragraphs, which may facilitate their accessibility.

In order to study the effects of self-generated cues, either during learning or at the onset of oral recall, on the recall of different text types, van Dam and Brinkerink-Carlier (1989) presented participants with a mixed text, containing a coherent story as well as a number of isolated plot-irrelevant paragraphs. Although more story paragraphs than isolated ones were retrieved, generation of cues during learning or at the onset of oral recall enhanced recall of the isolated paragraphs.

Van Dam and Brinkerink-Carlier (1990) further found that free recall of a text containing only isolated paragraphs was facilitated when half of the paragraphs were adapted so that they referred to a succeeding paragraph. They also found that insertion of single referring cues within isolated paragraphs and self-generation of retrieval cues during learning enhanced recall.

The present study extends the research on the effect of self-cues to the context of ESL education and also examines this effect in the recall of narrative texts. As mentioned earlier, ESL learners were recruited in this study as participants to see if this mnemonic effect would still be obtained in processing a second language. The specific question addressed in this study was whether generation of retrieval cues prior to recall has any influence on the written recall of the paragraphs of a narrative text in participants' second language. This study made use of a within-group design. The variables were recall condition (i.e., generate/non-generate) and accessibility of self-generated cues at recall (inspection/non-inspection).

Method

Participants

A standardised English proficiency test (TOEFL) was administered to a group of 50 adult ESL learners attending a language school. Thirty who scored between one standard deviation above and below the mean were identified as pre-intermediate and were selected to participate in this study. Participants were all male and aged between 23 to 28. They had studied English formally for at least 500 hours.

Materials

The materials were two texts (hereafter referred to as texts A and B) with narrative accounts, each composed of 14 self-contained paragraphs. These texts were selected from the book A Basic Course in Reading English for University Students (Jafarpur, Sedighi, & Tahririan, 1981), which is written for pre-intermediate students of English as a second language. The title of text A was "Slaves and Steam Engines" and text B "Pigeon Post".

Procedure

Participants were all tested under two conditions: Advance listing condition (AL) and posterior listing condition (PL). The 2 trial sessions were carried out with an interval of 24 hours. Under condition AL, half of the participants (N = 15) studied text A, while the other half studied text B, for future free recall. They were informed that they would later be required to reproduce the contents of the texts as completely as possible in their own words. While reading, they were allowed to make notes or underline portions of the texts, but not to use their notes during recall. The time allotted to reading the texts was 30 minutes.

Next, participants were given 15 minutes to perform an interpolated task, during which they answered questions regarding their field of study. Such a retention interval was included to prevent recall from short-term memory. The participants then received a form with one column of 14 numbered lines. To carry out the task of self-generation of cues, they were asked to generate from memory a list of short cues (one or a few words) for the14 paragraphs they had studied at the study phase. Participants were informed that the serial order of the generation of short indications did not have to coincide with the original serial order of the paragraphs in each text. After this listing, the papers were collected and participants were given enough time to recall and write down the contents of the texts they had studied at the study phase. During written recall, participants did not have access to their generated lists or the notes they had made while reading the texts.

Under condition PL, participants who had studied text A in condition AL received text B and those who had studied text B in AL received text A. The study procedure was the same for both conditions except that there was no cue generation after reading under PL condition. Upon reading the texts, participants performed a 15-minute interpolated task followed by a recall task during which participants wrote down the contents of the texts they had studied. Again participants were not allowed to use notes of any kind during recall. As soon as the written recall task was completed, participants' recall protocols were collected and participants were provided with a form including a column of 14 numbered lines to generate short cues for the paragraphs of the text, according to the same instructions provided in condition AL. As soon as participants completed the generation of cues, they were given a chance to recall and write down the contents of any additional paragraphs they could recall, to supplement their previous recalls.

Unlike in condition AL, participants were allowed at this stage of condition PL to inspect their self-generated cues. This was done to test the effect of access to self-generated cues at recall. The supplementary protocols and the sheets including participants' self-generated cues were then collected and made ready for analysis. Since there were two recall tasks involved in condition PL (i.e., one before self-generation and one after that), the first one will be referred to as PL1 and the second PL2 in the analysis of the data.

Data analysis and results

Scoring

In order for the scores to be reliable, the researcher as well as two other independent raters (university lecturers) scored the protocols. Then the average of the 3 scores was calculated for each condition.

The number of paragraphs recalled.
All written retrievals were scored for gist recall of the paragraphs. Scores ranged from 0 to 2 for each paragraph. A score of 2 was given for gist recall of the paragraphs; a score of 1 was given where the proportion of original words of the paragraphs was retained in the recall protocols; and a score of 0 was given where the words recalled did not apparently refer to the paragraphs. Scores of 1 and 2 were calculated as indices of recall.

The generated lists (i.e., cue words).
In scoring the self-generations, where the cues generated referred to their respective paragraphs, they were scored positively. Scores for the generated cue words also ranged from 0 to 2. A score of two was given where the original word/words of the paragraphs were generated; a score of 1 was given where the generated cue words were synonymous to the original words of the paragraphs; and a score of 0 was given where the generated word was irrelevant to the paragraphs.

The completeness of recalled paragraphs.
In order to obtain a measure of the completeness of the paragraphs recalled, each paragraph of the original text was analyzed in terms of its idea units. Then, the number of recalled idea units relating to each individual paragraph was divided by the total number of the idea units of the same paragraph and the result was considered as the proportion score for the recall completeness of that paragraph. The proportion scores for paragraphs were added together and divided by the total number of recalled paragraphs to obtain the mean proportion of recalled idea units, which was considered as completeness score for each participant in either condition.

The serial position of output.
In order to investigate whether the completeness of recall of paragraphs within each condition was the function of their serial position of output, the mean proportion of recalled idea units was calculated for the paragraphs that were recalled in each serial position of output. This was done by adding together the proportion scores of each paragraph recalled in each serial position of output and divided by the number of paragraphs recalled at that serial position (e.g., when 18 paragraphs were recalled at the 7th serial position of output, the 18 proportion scores of recalled idea units were added up and divided by 18 to get the mean proportion score for the 7th serial position of output). The scores obtained in each condition from the two different texts were added together and a single mean was calculated for each condition.

Results

The analysis of the data, using ANOVA, revealed a significant difference between the three mean scores obtained for recalls under AL, PL1, and PL2 conditions, F (2,87) = 5.19, p = 0.007. Post hoc analyses of the mean pairs, adjusting p levels using the Bonferroni inequality, revealed that: a) AL mean score was significantly higher than PL1 mean score, t(29) = 2.84, p = 0.0081, b) PL2 mean score was significantly higher than PL1 mean score, t(29) = 5.54, p = 0000.0, c) there was no significant difference between PL2 and AL, t(29) = 0.38, p = 0.71.

Therefore, it can be concluded that: a) self-generation of the retrieval cues prior to recall had a significantly facilitating effect on the recall of the paragraphs of the narrative texts, when inspection of the generated lists during written recall was not allowed, b) the self-generation of the retrieval cues subsequent to recall had a significantly facilitating effect on the recall of the paragraphs of the narrative texts, when the inspection of the generated list during written recall was allowed, c) there was no significant recall advantage for the inspection condition over the non-inspection condition. It should of course be borne in mind that these two conditions differed from each other in more respects than just the inspection.

The completeness of recalled paragraphs.
The results of a t test indicated that there was no significant difference between the mean scores for completeness proportion under conditions AL and PL1, (p = 0.43). This means that the preliminary generation of lists of retrieval cues for the paragraphs of a narrative text does not have any effect on the completeness of recall of the paragraphs.

The effects of the serial position of output.
The means of the proportion of the recalled idea units for the paragraphs that had been recalled in each serial position of output in conditions AL and PL1 were calculated and are presented in Figures 1 and 2. Since only a few participants recalled more than 10 paragraphs in each passage, only the first 10 paragraphs of each passage were taken into account and the calculations were based on those 10. This was done to prevent the effect of these few participants' recall on the overall results.

Figure 1: The serial position of output of paragraphs 1 to 10 and the
completeness with which the paragraphs were recalled in AL

Discussion

The results of the present experiment revealed that self-generation of cues by participants prior to recall of the contents of the paragraphs significantly facilitated access to the paragraphs of the selected narrative texts, whether or not inspection of the self-generated cues was allowed during recall. The findings of this study support the finding of van Dam, Brinkerink-Carlier, and Kok, (1987) with expository text, van Dam and Brinkerink-Carlier (1989) with story paragraphs, and van Dam and Brinkerink-Carlier (1990) with isolated paragraphs. Yet, van Dam and Brinkerink-Carlier (1988) illustrated, in their experiment, that inspection of short indications of paragraphs of a narrative text did not show any facilitative effect on oral recall of the paragraphs, t(29) = 0.64, which is in consonance with the findings of this study and the two experiments mentioned above.

Figure 2: The serial position of output of paragraphs 1 to 10 and the completeness
with which the paragraphs were recalled in PL (initial reproductions)

It was assumed, prior to this study, that the facilitating effect of a self-generated list might be increased if inspection of the list was allowed during recall. The results, however, did not support this hypothesis.

The results have also shown that self-generation of the retrieval cues does not affect the completeness with which the paragraphs are recalled. These results also support the findings of van Dam, Brinkerink-Carlier, and Kok, (1987) that during oral recall of the paragraphs of an expository text, inspection of self-generated cues has no effect on the completeness of recalled paragraphs, t(34) = 0.97.

In order to obtain an impression of the effect of serial position of the output on the completeness with which the paragraphs were recalled, Figures 1 and 2 are presented. These figures clearly indicate a decrease in the completeness values of the paragraphs recalled at later serial positions of output, suggesting that paragraphs recalled at earlier serial positions were recalled more completely, whether the lists of cue words were generated prior to recall or not. These findings are consistent with the findings of Smith, d'Agostino, & Starling-Reid (1970) and van Dam, Brinkerink-Carlier, and Kok, (1987). Smith, d'Agostino, & Starling-Reid found a strong negative correlation between output order and completeness of recalled categories from a categorized list of items. van Dam, Brinkerink-Carlier, and Kok, also reported a similar relationship between the serial position of output and recall of the paragraphs of an expository text in an oral recall test.

Overall, research has so far shown a facilitating effect on recall for the self-generation of retrieval cues, both in first language and second language. It is suggested then that curriculum developers allow for maximal self-generation on the part of learners in devising teaching methodologies and designing course syllabi for learners. Teachers may also benefit from this robust finding in designing their classroom activities.

References

Anderson, T.H. & Armbruster, R.B. (1984). Handbook of reading research. New York: Longman.

Brian, R., & Eastmond, N. (1994). Cognitive science and instruction. Englewood Cliffs, NJ: Educational Technology Publications.

Brunning, H. R. Schraw, G. J. & Ronning, R. R. (1999). Cognitive psychology and instruction (3rd Ed.). NJ: Prentice-Hall.

Bobrow, S.A., & Bower, G.H. (1969). Comprehension and recall of sentences. Journal of Experimental Psychology, 80, 455-461.

Jafarpur, A., Sadighi, F., & Tahririan, M.H. (1983). A basic course in reading English for university students. Shiraz: Shiraz University Press.

Kim, S. & Van Dusen, L. M. (1998). The role of prior knowledge and elaboration in text comprehension and memory: A comparison of self-generated elaboration and text-provided elaboration. American Journal of Psychology, 111, 353-379.

Kinjo, H. & Snodgrass, J. G. (2000). Does the generation effect occur for pictures? American Journal of Psychology, 113, 95-117.

Gorrell, J., Tricou, C. & Graham, A (1991). Children's short- and long-term retention of science concepts via self-generated examples. Journal of Research in Childhood Education. 5, 100-108.

Hill, R. D., Shwob, S. L. & Ottman, S. (1993). Self-generated mnemonics for number recall in young and old adults. Perceptual & Motor Skills. 76, 467-470

Hill, R. D., Allen, C. & Gregory, K. (1990). Self-generated mnemonics for enhancing free recall performance in older learners. Experimental Aging Research, 16, 141-145

Moshfeghi, F. & Sharifian, F. (1998a). Generation effect and associative learning, Journal of Education and Psychology Research, 1, 1-20

Moshfeghi, F. & Sharifian, F. (1998b). The effect of two different degrees of generation on memory. Proc. of the Fourth Conference on Linguistics. Tehran: Allameh Tabatabaie University Press.

Smith, A.D., d'Agostino, P.R., & Starling Reid, L. (1970). Output interference in long-term memory. Canadian Journal of Psychology, 24, 85-89.

van Dam, G. & Brinkerink-Carlier, M.(1988). The effects of self-generation of retrieval cues during learning or recall of a text. Journal of General Psychology, 115, 151-159.

van Dam, G. & Brinkerink-Carlier, M. (1989). The influence of self-cueing on recall of a text containing both plot-relevant and plot-irrelevant materials. Journal of General Psychology, 116, 109-119.

van Dam, G. & Brinkerink-Carlier, M. (1990). Free recall of a text containing isolated paragraphs only. Perceptual & Motor Skills, 70, 523-530.

van Dam, G., Brinkerink-Carlier, M., & Kok, I. (1987). The effects of self-generated cues on recall of the paragraphs of a text. Journal of General Psychology, 114: 135-146.