Action, meaning, and stimulus selection in paired-associate learning

Action, meaning, and stimulus selection in paired-associate learning

JOURNAL OF VERBAL LEARNING AND Action, VERBAL BEHAVIOR 7, 137-141 (1968) Meaning, and Stimulus Selection Paired-Associate Learning1 WILLIAM ...

392KB Sizes 0 Downloads 21 Views

JOURNAL

OF VERBAL

LEARNING

AND

Action,

VERBAL

BEHAVIOR

7, 137-141 (1968)

Meaning, and Stimulus Selection Paired-Associate Learning1

WILLIAM

in

D. ROHWER, JR., AND JOEL R. LEVIN

University

of California,

Berkeley

94720

The task of learning 12 pairs of high-frequency nouns by a study-test method was given to 112 fifth-grade children. In the experimental conditions, the study trial consisted of the presentation of each pair in the context of a three-word sentence wherein the two nouns were connected by a verb. In a three-way design, the variables manipulated were: the amount of overt activity implied by theverbs(actionvs. still); themeaningfulness of thesentence(norma1 vs. anomalous); and the character of the test-trial stimuli (subject nouns vs. verbs vs. subject nouns and verbs). Performance did not vary as a function of the amount of activity implied by the verbs, but normal sentences produced significantly more learning than anomalous ones. Verbs proved inferior to subject nouns as test stimuli, leading to theconclusion that the selection of verbs as functional stimuli during the study trials does not account for the sentential facilitation of noun-pair learning.

noun pairs in the context of sentences facilitates acquisition. Typically, the phenomenon has been detected with sentences in which one of the nouns occupies the subject position in the contextual string and the other occupies the position of direct object. In such sentences, the verb may be referred to as a connective which joins the two nouns to be learned. Rohwer (1966) found that conjunctions, as connectives, are less effective than verbs in promoting efficient learning, thereby posing the question: What are the properties of verbs that are responsible for the observed facilitation? Specifically, it may be asked if facilitation has been found because verbs often imply overt action involving the objects named by the subject and object nouns in a sentence. One of the ways verbs and conjunctions differ is that the former frequently refer to episodes of overt action while the latter do not, and it has been shown 1 This work was supported, in part, by a contract that paired objects presented pictorially are with the U.S. Office of Education (OE6-10-273) learned more rapidly when depicted in an through Project Literacy. The report was prepared at the Institute of Human Learning, which is supported by action episode than when depicted still grants from the National Science Foundation and the (Rohwer, Lynch, Suzuki, and Levin, 1967). If National Institutes of Health. .-- implied overt activity is the property of verbs

The fact that children learn paired nouns more efficiently when the nouns are presented in the context of a sentence than when presented alone (Rohwer, 1966) permits of multiple interpretations. After translation into empirical terms a number of such interpretations have been eliminated as inadequate to account for the phenomenon. Among the hypotheses that have already been disproven are those concerning intralist similarity (Rohwer and Lynch, 1967), semantic constraint (Rohwer and Lynch, 1966) and context availability (Rohwer, Shuell, and Levin, 1967). To date, a satisfactory account has neither been given nor validated, and one of the obstacles may be that the phenomenon has been described in insufficient detail to permit explanation. Thus, the purpose of the present experiment was to specify more precisely the conditions under which the presentation of

137

138

ROHWER AND LEVlN

responsible for facilitation, then verbs that do not refer to overt activity should not produce facilitation of noun-pair learning. A second factor of possible relevance is suggested by the demonstrated relationship between semantic structure and the learning of entire strings of words. Marks and Miller (1964) reported that normal sentences are better learned than anomalous sentences in which English word order is preserved but in which string meaning or semantic structure is violated by the particular combination of words concatenated. If the learning of constituent noun pairs is governed by the same law as the learning of entire strings, then less facilitation of PA learning should be produced by anomalous sentence contexts than by normal sentence contexts. Finally, a third issue concerns the nature of the functional stimulus when a noun pair has been presented in a sentence context. In a previous study (Rohwer, Shuell, and Levin, 1967) it was shown that as a stimulus for the object noun, the subject noun from such a context is not as effective as the entire portion of the sentence preceding the object noun. It might be inferred from this result that the sentence, as a configuration, is the functional stimulus (cf. Underwood, 1963). It is also possible, however, that the verb from such a context serves by itself as the functional stimulus. If so, the presentation of the verb from a contextual string as the test stimulus should produce performance equivalent to that produced by the presentation of the subject noun with the verb. METHOD Design and Materials. A 2 x 2 x 3 factorial design with two independent control groups was used to assess the relevance of three factors to the learning of paired nouns. The first factor, Activity, contrasted verbs implying overt action (Action) with verbs implying little or no overt action (Still). The second factor, Meaningfulness, compared the effects of normal sentence contexts with those of anomalous sentence contexts. A test for the identity of the functional stimulus was provided by the third factor, Test Stimuli, in which the nature of

the test trial stimulus was varied (subject noun vs. verb vs. subject noun and verb). Four lists of 12 three-word sentences were constructed, one for each of the conditions produced by the factorial combination of the Activity and Meaningfulness factors. The sentences for the Normal-Action (NA) condition were produced first by simply inserting a semantically appropriate verb implying overt action between the two high-frequency nouns in each of 12 pairs (e.g., ROSES drink RAIN.). The sentences for the Normal-Still (NS) condition were then derived by substituting semantically appropriate verbs which did not imply overt activity for those used in NA materials (e.g., ROSES like RAIN.). Sentences for the Anomalous-Action (AA) condition were formed by substituting semantically inappropriate nouns in the directobject positions of the NA sentences (e.g., ROSES drink HATS.). Finally, Anomalous-Still (AS) sentences were constructed from those used in NS by replacing the direct object nouns with those used in AA (e.g., ROSES like HATS.). It is important to note that the list of noun pairs used in the Anomalous conditions was different from that used in the Normal conditions, even though the stimulus nouns were identical in both cases. This method of manipulating the meaningfulness of the two sets of strings was chosen over the alternative of varying the verbs used in the two conditions in order that the definition of the distinction between Still and Action verbs would be identical for both the Normal and the Anomalous strings. In contrast to the difficulties involved in an attempt to control for verb differences in the latter kind of design, it was a straightforward matter to control for possible differences in list difficulty in the present design. One group of Ss was asked to learn the N list of pairs and another was asked to learn the A list in the absence of any additional context. The studytrial materials for the two control conditions, N and A, consisted of the noun pairs used in the NA and NS sentences and the noun pairs used in the AA and AS sentences, respectively. Three different kinds of test-trial materials were used within each of the four Activity x Meaningfulness conditions. In the first (S) only the subject nouns were presented as stimuli to which the direct-object nouns were tobegivenasresponses.Thesecondsetofmaterials (V) consisted only of the verbs from each of the sentences as stimuli. And the third set of test-trial materials (S and V) presented both the subject nouns and the verbs from the study-trial sentencesas stimuli. The testtrial stimuli in both control conditions were simply the first nouns from each of the pairs. All materials were typed on white, 4 x 6-in. cards in &in. type, one pair per card. Whenever they appeared, all nouns were capitalized and underlined to duplicate the format used in previous research. All cards for each

FACILITATIONOF

139

PA LEARNING

well as those from the experimental conditions were included. Contrary to what has been obtained with pictorial materials (Rohwer, Lynch, Suzuki, and Levin, 1967) the Activity factor produced no significant differences in performance, F < 1. Either (a) sentences do not evoke mediate pictorial responses of the activities implied or, (b) the two categories of verbs, A and S, were not sufficiently different with respect to the amount of overt activity implied to produce effective differences in the nature of the pictorial responses evoked. Since different lists of paired nouns were used in the Normal and Anomalous conditions, it is critical to determine whether or not the two lists differed in difficulty before evaluating the effect of the Meaningfulness factor. There was no significant difference between the two control conditions, F < 1. As the two lists do not appear to differ in difficulty, the significant superiority of the Normal sentences over the Anomalous ones, F(l, 98) = 29.52, p < .Ol, can be attributed to the difference in semantic structure. Dunnett’s test indicated RESULTS that, relative to the appropriate control condiLearning was measured in terms of the tions, Normal sentences produced facilitation numbers of correct responses made on the test of PA learning, t(2, 98) = 2.38, p < .Ol, but trial. The results are presented in Table 1. An Anomalous sentences did not, t(2, 98) = .02, analysis of variance was performed in which p > .05. Clearly the effect of semantic structure the variance associated with conditions was on the learning of entire sentences holds for the partitioned in a manner similar to that sug- learning of constituent noun pairs as well. gested by Winer (1962, p. 263) such that all The effect of the third factor, Test stimuli, observations from the control conditions, as was significant, F(2, 98) = 24.13, p < .Ol, and condition were placed in small loose-leaf binders for presentation to Ss. Procedure. The task was administered to Ssindividually by two Es, according to a study-test method for one complete trial. Additional trials were deemed unnecessary in view of the fact that the phenomenon of interest has been shown to appear on the first learning trial and to remain constant across subsequent trials (Rohwer, 1966). One E presented instructions and the study- and test-trial materials, while the other recorded Ss’ responses. In the instructions, Ss were told that they would be shown a series of sentences (Controls: a series of word pairs) and that their task was to learn these in a way such that when shown part of each sentence (Controls: the first word from each pair) they could tell Ewhat the last word in the given sentence (Controls: second word in each pair) had been. The study-trial items were presented successively at a 3-set rate and were read aloud by E as they appeared. The test trial followed immediately in which the stimulus terms were presented at a 3-set rate and read aloud by E but in an order different from that of the study trial. Subjects. Of a total of 112 fifth-grade children drawn from two public elementary schools in a middle-class neighborhood, eight were randomly assigned to each of the 12 experimental and the two control conditions.

TABLE MEAN

NUMBERS

OF CORRECT

RESPONSES

AS A FUNCTION

Normal

1 OF ACTIVITY,

MEANINGFULNESS,

AND

TEST STIMULI

Anomalous

Test stimuli

Action

Still

Action

Still

Total

Nouns Verbs Nouns and verbs Total Controls

6.00 3.50 9.00

7.37 3.87 7.87

5.00 1.12 4.25

4.62 2.00 4.75

5.75 2.62 6.47

Note.--MSE(98)=

6.27 4.12 5.60.

3.64 3.62

140

ROHWER

accounted for a substantial percentage of the total variance, 26%. Contrary to the prediction that verbs are the major functional stimuli, the Scheffk method of comparisons revealed that V-test stimuli produced fewer correct responses than either N- or N- and V-test stimuli, p < .Ol, whereas the latter did not differ, p > .05. As assessed by Dunnett’s test, facilitation was produced by N stimuli, t(4,98) = 1.87,~ < .05, and by Nand V stimuli, t(4, 98) = 2.69, p < .Ol, but not by V stimuli; that is, the combined control conditions did not differ significantly from the V conditions, t(4,98) = I .26, p > .05.

None of the tests for interaction was significant. Only one, Meaningfulness x Test stimuli, P(2, 98) = 1.70, p > .05, produced an F-ratio in excess of unity. DISCUSSION

In terms of the intent of the present experiment, namely, to specify some of the conditions under which verb connectives do and do not facilitate the learning of noun pairs, the following conclusions are clearly indicated. Verbs that imply relatively little overt activity involving the two objects named by the noun pairs produce as much facilitation as verbs implying considerable overt activity. Semantically anomalous sentence contexts, however, do not facilitate the learning of constituent noun pairs whereas normal sentence contexts do. Perhaps the most intriguing aspect of the results concerns the failure of V-test stimuli to elicit as many correct responses as elicited by the N-test stimuli. It might be argued that the presentation format, in which nouns but not verbs were capitalized and underlined, is responsible for performance differences associated with the three types of test stimuli. The format does indeed de-emphasize the verbs, but this is not a sufficient explanation since, as study-trial connectives, verbs nevertheless facilitate performance. If the process whereby verb connectives facilitate the learning of noun pairs consists of the selection of the verbs as

AND

LEVIN

functional stimuli, then the verb in each string, when presented overtly, should reliably elicit the appropriate response noun. The present results demonstrate that this is not the case, at least not for the majority of the pairs in a list. It may be, however, that the verbs serve as functional stimuli for some but not all of the pairs; the verb test stimuli did elicit correct responses, even though the number of these was quite small. Three additional interpretations should also be mentioned. The first is that the subject noun and the verb serve as a single configurational stimulus for the object noun. The second is that independent associations are formed between the subject and object nouns and between the verb and object noun. The third is that an association is formed between the subject noun and the verb and between the verb and the object noun such that on test trials the verb mediates between the subject and object nouns. Although no one of these three interpretations is entirely discounted by the present results, each implies the prediction that the N and V condition should produce the best performance and this prediction was not confirmed. The interesting discrepancy is that the presentation of verb connectives during the study trial produces a major difference in performance whereas its presentation during the test trial does not. One implication of this, namely that the effect of verb connectives on the learning of noun pairs occurs principally at input, is consistent with results of an experiment reported previously (Rohwer, Shuell, and Levin, 1967) in which the presence of verb connectives at the time of output alone was shown to be irrelevant to PA performance. Given that the locus of the effect is at input, the puzzle is to describe the process occurring at that time which facilitates learning. REFERENCES MARKS, L. E., AND MILLER, G. A. The role of semantic and syntactic constraints in the memorization of English sentences. J. verb. Learn. verb. B&w., 1964,3,1-5.

FACILITATION

ROHWER, W. D., JR. Constraint, syntax and meaning in paired-associate learning. J. verb. Learn. verb. Behav., 1966,5,541-547. ROHWER, W. D., JR., AND LYNCH, S. Semantic constraint in paired-associate learning. J. educ. Psychol., 1966,57,271-278. ROHWER, W. D., JR., AND LYNCH, S. Form class and intralist similarity in paired-associate learning. J(verb. Learn. verb. Behav., 1967,6,551-554. ROHWER, W. D., JR., SHUELL, T. J., AND LEVIN, J. R. Context effects in the initial storage and retrieval of noun pairs. J. verb. Learn. verb. Behav., 1967,6, 796-801.

OF PA LEARNING

141

W. D., JR., LYNCH, S., SUZUKI, N., AND J. R. Verbal and pictorial facilitation of paired-associate learning. J. exp. child Psycho/ 1967,5,294-302. UNDERWOOD, B. J. Stimulus selection in verbal learning. In C. N. Cofer and Barbara S. Musgrave (Eds.), Verbal learning and behavior: problems and processes. New York: McGraw-Hill, 1963. WINER, B. J. Statistical principles in experimental design. New York: McGraw-Hill, 1962. ROHWER, LEVIN,

(Received August 9, 1966)