Psychon.

Sci.,

1970,Vol. 20 (2) 119

Comparison of associative learningstrategies

GORDON H. BOWER* and DAVIDWINZENZStanford University,

Stanford,

Calif. 94305

Ss learned paired-associate lists of nouns using one of four learning strategies:repetition of the pair, reading the pair as subject and object nouns in a meaningfulsentence, generation of a meaningfulsentence linking the twonouns, or visualization of amental image in which the word referents were in vivid interaction. Half of thepairs weretested for recall of the response member given the stimulus and half were tested formultiple-choice recognition of the response. Significant differences in recall andrecognition occurred with conditions ranking in the order: imagery (best), sentencegeneration, sentence reading, and rote repetition (worst). The results are interpreted interms of associative consequences of relational organization: S's memory benefits fromhis actively searching out, discovering, and generating (or depicting) predicative or"actor-action-object"relations between the words or referents of a pair.

This experiment is concerned with the remembered a noun pair much better ifeffect of various learning strategies upon they generated their own sentence linkingpaired-associate learning by college Ss. Ss the word pair rather than juststudying thespontaneously use many different pair embedded in a sentence constructedtechniques to learn paired associates; with by E. Finally, Bower (in press) citesnoun pairs, the techniquesmay range from evidence that having S form a mentalsimple repetition to the construction of picture of the words' referents in vividcatchy phrases or elaborate narratives interaction produces somewhat betterwoven around the word pair. Rohwer learning than does having S generate a(1966) reported that reading young sentence linking the word pair. Comparingchildren a declarative sentence linking the across these several studies, there seems toS-R pair as subject and object nouns be a progressive improvement in recallproduced better recall than having the going from rote repetition to studying thechildren simply study the pair without a word pair in a sentence, to generating asentence context. Further, Bobrow & sentence linking the pair, to forming anBower (1969) found that college students inteiactive mental image. The following

experiment compares these several learningstrategies within one experiment and one S"This research was supported by

Grant

. . , , .. ,MH-13950 to the fust authorfrom the National population, using both recall andInstitute

of

Mental Health. recognition tests of associative learning.

Conceivably, the differences amongconditions are attributable solely todifferences in response availability. If so,then differences among the conditions inrecall performance would disappearin testsof recognition memory.

METHODThree paired-associate lists of 30 items

each were constructed using unrelatedconcrete nouns. Each S learned andrecalled all lists. Study times and test timeswere 5 sec per item. After one study trialon each list, an "immediate" test followedfor the items of that list. The test stimuliwere presented in the same order as thatinwhkh the pairs had been studied, thusassuring a constantlag of30 itemsbetweenstudying and testing each pair. After thethree lists had been studiedand testedonceeach, a final test was given to all 90 pairs.On both immediateand final tests,half theitems were tested for cued recall and halfby multiple-choice recognition. Therecognition test presented the cue wordfollowed by five response words from theappropriate list; S was to select and sayaloud the correct response that had beenpaired with the cue word.No feedback wasgiven on test trials.

Each S was assigned randomly toone offour groups, received generalpaired-association instructions, and thenwas instructed specifically on how tomemorize the S-R pairs. In the repetitioncondition, S was asked to repeat orrehearse each pair silently to himselfduringthe study time. In the sentence-readingcondition, S saw each pair as capitalizedsubject and object nouns in a simpledeclarative sentence, was told to read thesentence aloud and use it to associate thetwo critical nouns. In thesentence-generation condition, S wasshown each pair and asked tomake up andsay aloud a meaningful sentence or phraseusing and relating the two words in asensible way. In the imagery condition, Swas asked to visualize a mental picture orimage in which the two referents (denotedby the words) were in some kind of vividinteraction.He was encouragedto make hisimage as elaborate, vivid, or bizarre as hewished. The Ss were 40 undergraduatesfulfilling a service requirement for theirintroductory psychology course, with 10Ss assigned to each of the four learningconditions.

RESULTSThe average number of correct responses

per list (15 recall and 15 recognition items)is given in Table 1 for the four types oflearning conditions. "Immediate" inTable 1 denotes tests given just after eachset of 30 items had been studied. "Final"denotes the test of all items at the end ofthe session after the three lists had been

Psychon.

Sci.,

1970,Vol. 20 (2)120

Table 1Average Correct Responses Per List for Four Study Conditions (Out of 15)

Sentence- Sentence-Repetition Reading Generation ImageryTest Type

RecallImmediate D „, ...Recognition5.57 8.17 11.50 13.10

12.03 13.10 14.80 14.67

.

. RecallFinal DRecognition5.23 8.03 10.93 12.67

11.73 13.13 14.70 14.87

studied and tested once each. The pattern 0f the relationship between the criticalof results and the outcome of statistical noUns in the two cases. Ss who mustsearchanalysis were much the same for the out and construct a linking (predicative)immediateand the final test results. relationship among two nouns are more

There were significant differences in likely to activate and thereby associate theimmediaterecall amongthe four mediation two full semantic concepts than are Ss whoconditions, F(3,36) = 32.24, p<.001; all merely read a large number of similarordered pairwise t tests were significant at sentences.or beyond the .05 level. The recognition The recall difference between thedata also showed significant differences imaging and sentence-generation conditionsamong conditions,F(3,36) = 7.68, p< .01; is interesting, and its significance is openhowever, ordered t tests revealed that the for speculation. Paivio (1969) and Boweronly significant pairwise gap was that (in press) have conjectured that imaging Ssbetween the sentence-reading vs the benefit because they may establish a dualsentence-generation conditions, memory code, with both verba) andt(18) = 3.64, p< .01. imaginal memory traces. Bower proposed

DISCUSSION that the code in the verbal memory systemIn genera], the comparative results on corresponds to an "actor-action-object"

recall are much as conjectured from the relational bonding between the semanticprior pieces of evidence: Visualization of concepts, while the code in theinteractive scenes produced best recall, pictorial-imaginal memory system depictsfollowed by sentence generation, then the visual schema or scene correspondingsentence reading, and finally rote to the semantic relations found. Laterrepetition of the pair. We may suppose, presentation of the cue wordmay reinstatewith Asch (1969), that associativelearning either the sentence memory code (fromis largely determinedby theperson finding which the other main noun can be selectedand remembering some sort of relation for recall) or the code for visualizing theamong the two items. Rote repetition of constructed scene (from which S willtwo words insures the relation of temporal "recognize" the main objects and thuscontiguity between the two "recall" the name of the desired object).phonetographic units, but few relations Imaging Ss exceed sentence-generation

Ss,

other than contiguity are aroused, by hypothesis, because the former Ss haveactivated, or used for storing the S-R pair, both memory codes available whereas theDeclarative sentences, embedding the latter Ss predominantly have only thecritical wordsas subject and object nouns, verbal code available. A relevant bit ofassert predicative relations among the evidence is that Ss instructed tovisualize orwords, usually treating them as "actor" mentally image the scene described by aand "object of action." Bobrow& Bower declarative sentence give associative recall(1969) conjectured that the advantage in of the main nouns at a level significantlyrecall of sentence-generating Ss over higher than that of Ss who simply read thesentence-reading Ss is due to differencesin sentence,

probability or amount of "comprehension" The discussion above applies to the

recall results. The pattern of results is quitesimilar with the multiple-choicerecognition tests except that an obviousceiling is apparent in the better learningconditions. Recognition even on the finaltest was 98% and 99% for thesentence-generation and imagingconditions, respectively. The significantgap in recognition is between thesentence-reading vs sentence-generation Ss.which suggests an "active-idiosyncratic vspassive-prescribed" dimension to the Ss'associative strategies. It may be noted thatthis was also the largest gap betweenconditions in recall probability. Therecognition data will be interpreted asdisconfirming the response-availabilityaccount of mnemonic differences: Evenwith response-availability factors equatedby recognition tests, performancefollowing the various learning strategiesstill differs. Without strong arguments tothe contrary, the recognition data will beinterpreted as showing the same differencesamong conditions as were shown by therecall tests. Depending on one's theory,this may imply that the mediationconditions differ in the amount or qualityof information that is stored about eachpair rather than differing in terms ofretrieval efficiency ofa constant amount ofstored information.

REFERENCESASCH,

S. Reformulation of the problem ofassociation. American Psychologist,

1969,

24,

92-102.

BOBROW, S. A., & BOWER,

G. H.Comprehension and recall of sentences.Journal of Experimental Psychology,

1969,

80,455-461.BOWER,

G. H. Mental imagery and associativelearning. In L. Gregg (Ed.), Cognition inlearning and memory. New York: Wiley, inpress.

PAIVIO,

A. Mental imagery in associativelearning and

memory.

Psychological

Review,1969, 76,

241-263.

ROHWER,

W. D.. JR.

Constraint,

syntax andmeaning in paired-associate learning. Journalof Verbal Learning & Verbal

Behavior, 1966,5,541-547.