Context effects and the recall of comparative sentences

JOHN BENJAFIELD 0 LINDA GIESBRECHT 0 0 Brock University , St. Catharines, Ontario. Canada Clark and Card (1969) have proposed that semantic components underlie memory for comparative (C) sentences. To test this hypothesis. six groups of 15 Sseach were given different sets of C sentences. In line with the theory. Ss tended to remember unmarked adjectives better than marked ones and positive constructions better than negatives. However, contrary to the theory, they also tended to bias their responding either toward the negative or the equative form. A two-stage theory of recall. based on the memory schema-memory trace distinction. is proposed to account for these data. - Four pairs of unmarked-marked adjective pairs were selected from the set Clark and Card used: happy-sad; clean-dirty: pretty-ugly: and good-bad. Sixteen noun triples were used to generate 16 comparative sentences and 16 recall cues. Thus. some Ss heard this cue and this sentence: "The books: Robin Hood isn't better than Little Women." The 16 sentences had the folIowing properties: four positive comparatives containing unmarked adjectives: four negative comparatives containing unmarked adjectives: four positive comparatives containing marked adjectives: and four negative comparatives containing marked adjectives. Thirty-two additional sentences were generated bv changing the comparative-unmarked adjective-sentences' to the~ equative-unrnarked adjective and equative-marked adjective forms and by also changing the comparative-marked adjective "This work was supported by the National Research Council of Canada. The authors are grateful to Brian Doan for his help in the preparation of materials and the running of Ss. Comparative-Unmarked/ Comparative-Marked Equative-Unmarked/ Equa rive-Marked Com para tive-U nrnarked,' 6.1 Equative-Marked Comparative-Marked/ Equative-U nmarked Comparative-Unmarked! Equative-Unmarked Comparative-Marked! Equative-Marked Comparative-Unmarked Comparative-Marked Equa tive-Unrnarked/ Equa rive-Marked Com parative-U nmarked/ Eq ua tive- ~Iarked Cornparative-Marked/ Equative-U nmarked Cornparative-U nmarked! Equa tive-Unmarked Comparative-Marked/ Equa tive-Marked Response Bias Verba tim Recall se n ten ce s t o t heir e qua t ive-u nmarked adjective and equative-rnarked adjective forms. Each Sheard 16 of the sentences. played on a tape recorder. in blocks of 4. Each of the sentences in each block was separated bv a 3-sec interval. After each block. there was a 3-sec interval. f~llowed by the first noun cue. The'S then had 20 sec in which to write down what he could remember, followed by the word "stop." than the next noun cue. and so on. After S had tried to recall the fourth sentence. the next block was presented. until S had tried all 16 sentences. The S wrote each sentence on a separate page of a booklet. He was not allowed to go backward and change his previous attempts. The Ss were 90 underzraduate student volunteers at Brock University. Fifteen Ss were assigned randomly to each of six treatment conditions. The Ss were told to try to recall the sentences exactly. but. if they could not. to write down anything they could remember. The Ss were tested in groups of Sizes 3-15. Experimental Design The six experimental groups each received a different set of sentences, formed by combining the four basic sentence types two at a time to Yield six sets of 16 sentences each: IIi comparative-unmarked adjective and comparative-marked adjective: (2) comparative-unmarked adjective and equative-unmarked adjective: (3) comparative-unmarked adjective and equative-rnarked adjective: (4) comparative-marked adjective and equative-unmarked adjective: (5) comparative-marked adjective and equative-rnarked adjective: and (6) equative-unrnarked adjective and equative-marked adjective. The 16 sentences of each set were arranged in a 4 by 4 Latin-square design so that S encountered one of each adjective pair in each block of 4, of each construction in each block, and of each adjective pair in each construction. Thus. in each block. S heard two positive and two negative constructions. RESULTS The mean number of sentences in each category recalled. as well as the mean number correctly recalled. are given in Table 1. Items included under "response bias" were recognizable comparison constructions containing at least one correct noun and the correct underlying adjective or its opposite. Items included under "correct recall" were recalled correctly verbatim. In the four treatments in which Ss were presented with both marked and unmarked adjectives. they consistently preferred to use the unmarked forms [F( 1.56) = 57.11. p < .001) and. overall. remembered more sentences containing unmarked adjectives verbatim than they did those containing marked adjectives CONTEXT EFFECTS AND RECALL OF SENTENCES 135 [F(l,56) = 14.60, P < .001]. Three of the four adjectives yielded more errors toward the unmarked than the marked form. The percentage of errors toward the unmarked form for pretty-ugly ranged from 88% to 100% across the four conditions: for good-bad from 90% to 100%; for happy-sad from 83% to 100%. For clean-dirty, however, Ss tended to recalldirty in place of clean, the percentage of such errors ranging from 60% to 100%. This last finding was also present in the Clark and Card experiment. It may be that Ss regard "dirty" as unmarked, Le., descriptive of the most frequently occurring, normal state of affairs, and "clean" as exceptional, or marked. The Ss did not prefer positive over negative constructions [F(l ,84) < 1, n.s.] , but they did get more positive than negative constructions correct [F( 1,84) = 24.92. p < .001]. In the group given comparative-unmarked adjectives and comparative-marked adjectives, negative constructions were used as responses in recall more frequently than were positive constructions [t(l4) = 2.67, p < .02]. This finding replicates the aforementioned pilot study in which 18 Ss gave an average of 5.3 positive responses and 8.6 negative responses [t(l7) = 2.80, p < .02]. Although overall Ss preferred definite to indefinite constructions [F(l,84) = 4.15. P < .05]. there was an interaction of Type of Construction Preferred by Condition [F(5,84) = 2.94. P < .05]. such that indefinite constructions were preferred in the group given comparative-unmarked adjectives and com parative-marked adjectives. The Ss did not get significantly more definite than indefinite constructions correct [F(l ,84) = 1.03. n.s.]. Finally, Ss preferred equative to comparative constructions [F(1.56) = 27.03. P < .001] and also got more equative than comparative constructions correct [F( 1.56) = 20.50, p < .001] . The number of recognizable comparative sentences produced varied widely between conditions [F(5.84) = 4.41. P < .01] . as did the number of sentences correctly recalled [F(5,84) = 10.82. P < .001]. These effects are neither surprising nor very interesting. given the fact that there is no r (...truncated)