Schema learning with low redundancy patterns

Schema learning with low redundancy patterns EL VIS C. JONES Frostburg State College, Frostburg, Md. 21532 Ss tend to perform poorly on schema learning tasks with patterns below 50% redundancy. The present experiment attempted to determine if this was due to either (1) inability to abstract the schema with low-redundancy patterns or (2) a limited usefulness of the schema as an aid in the efficient encoding of low-redundancy patterns. Ss who were pretrained on the schema prototype performed better when transferred to 40% redundant patterns than Ss who received only 40% redundant patterns. This indicates that poor performance with 40% redundant patterns is largely due to a failure to abstract the schema. The use of schematic learning sets, with a gradual decrease in redundancy, was suggested as a possible means for improving S's ability to abstract schemata from low-redundancy patterns. On tasks that require Ss to abstract a schema from sampies of stimuli, performance is significantly influenced by the level of constraint redundancy (Rc) employed in generating the stimulus sampIes (Mueller, Edmonds, & Evans, 1966). In several pilot studies, the present E has found that in the absence of knowledge of re sults , Ss who reproduced 40% Rc patterns did not perform significantly better (p> .40) than Ss who reproduced nonschematic (0% Rc) patterns. Poor performance on low Rc patterns could be attributed to either or both (I) an inability to abstract the schema contained in the patterns or (2) to a limited usefulness of the schema as an aid in the efficient encoding of low Rc patterns. With low Rc, the central tendencies of the patterns may not occur with sufficient frequency to be detected by the Ss. Also, with low Rc, the schema may not be sufficiently descriptive of the individual patterns to be of help in encoding the individual patterns. Thus, the present experiment represents an attempt to determine the extent to which poor performance with low Rc patterns is due to (1) or (2), above. METHOO The Ss were 60 experimentally naive students enrolled in an introductory psychology course. They were assigned randomly to three groups of 20 Ss. All Ss were trained in a reproduction task with 40 histoform patterns assembled into booklets. These patterns were constructed according to the rules of the VARGUS 9 program (Evans & Mueller, 1966). The schema consisted of the following average column heights (4,9,7, 5,8,3,9,5, 2). Each pattern consisted of nine columns of reclining capital Os, produced by an IBM Model 11 electric typewriter. The base of each column was a reclining parenthesis sign ( (). The patterns were reproduced by 74 say 'turn the page' and you will find a new pattern. The booklet contains a total of forty complete patterns and forty patterns with missing columns. Throughout the booklet we will repeat the same steps as I have just described. Once we begin, never turn any page until you are told to do so. Never turn back to any of the preceding pages. Are there any questions?" RESULTS ANO DISCUSSION As indicated by Fig. 1, Ss in the 40-40 group demonstrated no significant improvement over the 40 patterns in their booklets. The performance of the 100-100 group rapidly improved and remained nearly perfect throughout the booklet. The performance of the ·100-40 group paralleled that of the 100-100 group over the first 20 patterns but deteriorated when the Ss reproduced the last 20 (40% Rc) patterns. The difference in performance on the last 20 trials was significant (Scheffe's test, F = 18.61, df= 1/57, P < .01). The 100-40 group's performance on the final 20 patterns was significantly superior to that of the 40-40 group (Scheffe's test, F = 15.74, df= 1/57, P < .01). Although a number of alternative interpretations are possible, it could be assumed that the difference between the 100-40 and the 40-40 groups was due to the fact that the 100-40 group had knowledge of the schema. Thus, the poor performance of the 40-40 group is not entirely attributable to the limited usefulness of the schema in efficiently encoding individual patterns. The data suggest that the 40-40 group probably was a mimeograph machine onto 8~ x ll in. paper and assembled into a booklet. The paper used in the booklet was sufficiently thick to preclude S's seeing through any page, and a blank sheet was used as a cover for the booklet. The patterns were reproduced on the odd-numbered pages of the booklet. The even-numbered pages contained the same patterns, except that three columns were omitted from each pattern. The columns to be omitted were determined randomly for each pattern. The booklets received by the 100-100 group contained 40 copies of the schema prototype. The booklets received by the 100-40 group contained 20 copies of the schema prototype, followed by 20 variants of the prototype, with an Rc level of 4WO. The 40-40 group received booklets containing 40 patterns that were 4WO redundant. The last 20 patterns in the booklets received by the 40-40 group were 40-40 identical to the last 20 patterns received by 100-100 the 100-40 group. ..------. 100-40 The following instructions were read to 30 all Ss: "00 not turn the cover page of the ~ booklet until I tell you to do so. When I ü W say 'turn the page,' you are to immediately 0:: 0:: 0 turn it and look at the first page. On the ü first page you will see a pattern composed CI) 20 z of zeros. You will have 15 sec in which to :;: memorize the pattern. After 15 sec I will :::> ..J again say 'turn the page.' When you do so, 0<.> you will see another pattern. This pattern Ci 1.0 is exactly like the one on the first page, z z except that there will be three columns <t w which have been omitted. You are to draw :;: in the three missing columns so that they 2 3 4 5 6 7 8 9 10 are the same height as they were in the original pattern. Oraw in the missing BLOCKS OF 4 PATTERNS columns by drawing a short horizontalline Fig. 1. Average number of columns at the height you remember the missing column to have been. You will have only eorreetly reprodueed by Ss reeeiving 40 15 sec in which to do this, so work as patterns, a1l 40% redundant (40-40 group), rapidly as possible. If you do not a1l 100% redundant (100-100 group), or 20 remember the exact height of a missing patterns at 100% redundaney followed by column, go ahead and guess. 00 not leave 20 patterns at 40% redundaney (l00-40 any column blank. After 15 sec I will again group). .- Psychon. Sei., 1970, Vol. 21 (2) unable to abstract the schema from Edmonds, Evans. & Mueller, 1966) with patterns with such low Rc. high Rc schema families, then transferring Conversely, the difference between the them to a new schema family composed of JOO-JOO and 100-40 groups on the last 20 low Re patterns. Having "learned how to patterns could be largely attributcd to the learn" schemata may enable Ss to abstract limited usefulness of the schema in schemata even from patterns with 4Q9c Re. encoding individual patterns a (...truncated)