"WWW-W‘- THE mm QF INSTRUCTIONS. TRANfiFER. AND CONTENT SN REASONING TIME Theta ‘ov flu Davao of D5, D. MKCBIGAN 3'? TE UNI‘JEBSETY Gwrge Lyie Pamfi i969 ‘fHESl‘ This is to certify that the thesis entitled THE EFFECTS OF INSTRUCTIONS, TRANSFER, AND CONTENT ON REASONING TIME presented by George Lyle Parrott has been accepted towards fulfillment of the requirements for Ph D degree in Psychology )[j 1 Y“ t/1j‘/ u1vé/ {J //1 - («LL//""]¥/ ‘L/ Major professor fl. Date [41’ («\v //7/ 1/ (7/14 y / '7’, r 0-169 .-.—mm—v—cn—» V . ABSTRACT THE EFFECTS OF INSTRUCTIONS, TRANSFER, AND CONTENT.ON REASONING TIME By George Lyle Parrott Using syllogistic reasoning as the focus, this paper has examined the effects on reasoning processes of: 1. speed and standard instructions; 2. type of warmup or transfer conditions; and 3. true and false content. After first obtaining simple reading time on all content types, then checking for the truth-falsity of the multiple-choice alternatives in the T and F scales, a 3 x 2 x 2 eXperiment was conducted using as the dependent variables: 1. premise reading time; 2. total reasoning time; 3. number of switchbacks; and 4. number correct. No regular subjects had prior training in logical reasoning. As expected, no differences obtained for number correct as a function of any of the main effects or interactions, and this suggested that the basic reasoning process leading to a correct solution was relatively matched across all groups. Any differences in solution time could not be associated with greater or fewer errors, but rather with the semantic processing of the given material. As predicted, speed instructions reduced reasoning time; no content-by-instructions interaction was indicated. Also in line with predictions, true items George Lyle Parrott took significantly less time to reason with than false, even though they were closely matched for reading time. In another comparison, symbolic items did not differ from true items on reasoning time as expected, but a measure of net processing time that corrected for the much shorter reading time required by the S items did indicate a significant difference between S and T items. Symbolic reasoning and processing times were much less than those for false items. The transfer comparison yielded a significant over- all E, but the source of the major facilitation was in the symbolic warmup, indicating a structural insight, rather than in the opposite content, indicating an arousal effect. A transfer—by-instruction interaction was present as the opposite content warmup showed negative transfer under speed demands while the symbolic warmup appeared even more effective at facilitating the false content reasoning. A significant transfer-by- content interaction was due to the transfer effect speeding up only the false content items. Several alternative analyses of the main effects as they related to encoding or premise exposure time and the number of switchbacks, another measure of encoding difficulty, were also reported; all generally supported the basic con- tent and transfer effects as measured by total reasoning time. George Lyle Parrott The discussion of this main study emphasized the importance of the basic information on the parameters of reasoning speeds as well as how both S—R and information processing theories might be used in integrating these results. An information processing model was discussed, and the nature of structural effects in reasoning processes examined as they might relate to the central processing of encoded information. A final chapter reported a correlation study of dogmatism, rigidity and reasoning times. As predicted, the Rokeach Dogmatism Scale correlated positively with overall reasoning time for false content and negatively with true content; the difference between the two correlations was statistically significant. No relation- ships were expected and none obtained between the Gough- Sanford Rigidity Scale and either content reasoning time. These results were fit quite neatly into Rokeach's (1960) belief-disbelief system, but another finding in the examination of the post-eXperimental measurement of D and R indicated that subjects who had been in the speed instructions group of the main experiment had significantly lower D scores than those in the standard instructions group. No effects on R scores were seen. These findings were quite unexpected, but it was suggested that the speed demand forced the subjects to open up their cognitive systems and this open state was reflected in the later D— George Lyle Parrott score measurement. THE EFFECTS OF INSTRUCTIONS, TRANSFER, AND CONTENT ON REASONING TIME By George Lyle Parrott A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1969 To Dr. Donald M. Johnson ii ACKNOWLEDGMENTS The completion of any course of personal study depends greatly on others. Dr. James G. Martin helped to form this student's attitude toward research which was further developed during four years of close association with Dr. Donald M. Johnson and R. Paul Stratton. The influence of Dr. James Phillips has been most helpful, and his personal attack on psychology a beneficial model. Dr. Charles Wrigley has given freely and most positively of his professional insights and scientific experience, and any graduate student must be better trained by contact with this teacher. Dr. Albert Rabin's assistance has been of additional benefit, as in the midst of the traditional experimental paradigm, an influence could be felt for consideration of individual difference data. Dr. Milton Rokeach's research and teaching has further broadened this student's interest. Last in order, but first in true influence, my wife, Bonnie's, consistent support, encouragement, and assistance made many, many problems less formidable and every task more productive. If any wife ever earned a share of her husbandfis Ph. D., its her. iii Chapter I. II. III. IV. VI. VII. TABLE OF CONTENTS INTRODUCTION..... ...... . ................... Syllogistic Reasoning: A Review......... A Theoretical Integration ................ Current Research Outline and Hypothesis.. EXPERIMENT I: READING TIME STUDY .......... Method.. ................................. Results .................................. Discussion ............................... EXPERIMENT II: CONCLUSION SCALING ......... Method.. ................................. Results.. ................................ Discussion. ............................ EXPERIMENT III: INSTRUCTIONS, CONTENT, AND TRANSFER EFFECTS ON THE REASONING PROCESS .................... . ............... Method ................................... Results .................................. Discussion .............. . ................ Passive and Active Reading. .............. Reasoning Theory ....... . ................ Structure- Process Model of Reasoning ..... Transfer Effects and Structural Implications ............................ Skilled and Naive Subjects ............... Future Research ......... . ................ EXPERIMENT IV: DOGMATISM, RIGIDITY AND REASONING ........ . .............. . .......... Method... ................................ Results .................................. Discussion ............................... SUMMARY .................................... REFERENCES.. ............................... 20 20 22 22 24 24 25 26 Chapter Page VIII. APPENDICES ............................... 75 Statement Scaling Form D...... ....... ... 75 Reasoning Scale Items.. ...... ...... ..... 78 Post-Reasoning Experiment Questionnaire. 83 Table 10. ll. 12. l3. l4. 15. 16. 17. LIST OF TABLES Page An Outline of the Experiment .................... 17 Total Reading Times in Seconds for Three Item Types........ ............. ...... ...... .... ...... 22 Summary of T-F Ratings for Conclusions .......... 25 Assignment Schedule................. ........ .... 28 Means and SDs (in parentheses) for Total Number of Items Correct as a Function of Treatment ..... 30 Analysis of Variance for Number Correct in 3 x 2 x 2 Study................................. 31 Means and SDs for Total Reasoning Times for All Experimental Groups (in seconds) ......... . ...... 32 Analysis of Variance Summary for S, T, and F Treatments in both Speed and Standard Instructions............ ..... ......... .......... 33 Reading and Reasoning Times on S, T, and F Content ........ . ..................... ..... ...... 36 ANOV for Total Reasoning Times (in seconds) ..... 36 Means and SDs in Seconds for Encoding Time ...... 41 Analysis of Variance on Encoding Times .......... 41 Total Number of Switchbacks in Each Experimental Group........... ....... ..... ....... 42 Analysis of Variance for Number of Switchbacks.. 42 A Comparison of Skilled and Naive Subjects on Premise Encoding Times and Number Correct....... 44 Correlations of Dogmatism and Rigidity Scores with Reasoning Time... .................. . ....... 64 Reasoning Instruction Effects on D and R Scores.. ........... . ........ . ..... .. ............ 65 vi LIST OF FIGURES Figures Page 1. ‘The "plan" for human information processing, specifically reasoning... ..... .. 12 2. Total reasoning and processing times under standard and speed instructions............. 34 3. Transfer effects on total reasoning time by standard and speed instructions ............. 38 4. A theoretical distinction between reading and reasoning ......................... . ..... 47 5. Berlyne's Analysis of the S-R attack on thinking .................................... 49 vii INTRODUCTION Interest in thinking has had a long history, from the early speculations by the Greek philosophers to the recent increase in both empirical and theoretical work in psy- chology the domain of cognition has been of interest. Within this broad domain, two major subdivisions become apparent: convergent and divergent cognitive processes. The divergent processes are those commonly associated with creativity and have been recently examined in an extensive experimental study (Johnson, Parrott, and Stratton, 1968). The convergent processes are those more closely allied to the traditional tOpic of reasoning or logic, and it is to a classic form of reasoning, the syllogism, that this paper is addressed. SYLLOGISTIC REASONING: A REVIEW The syllogism is an organization of three propositions; it is designed so that, given the validity of two of the statements, the third mg§§_follow. Across the three prOpositions, only three separate terms are normally given, two terms per propositions as in a subject- predicate relationship. Since the conclusion statement has only two terms, one of the premise or background terms has been excluded--it is labeled the middle term 1 2 and actually serves as the linking feature between the two separate premises. The content of the statements of a syllogism may be in three forms: words, symbols, or figures. Earlier research has examined the syllogism from the perspective of the logician; this orientation has led to studies of logical errors. The work of Simpson and Johnson (1966) typifies this concern. These authors examined the tendency of subjects to accept conclusion alternatives as logically true on the basis of the superficial similiarity of a common word in all assertions, or atmosphere error; and they also studied the logical fallacy of reversing the subject-predicate order, or con- version error. An example of an atmosphere error syllogism would be: "Some A is B; Some B is C; there- fore some A is C." A conversion error might involve changing an assertion of "all A is B" into "all B is A;" this conversion is not permitted by the rules of formal logic. Considerable previous research has also been reported studying the effects of syllogistic content on arriving at correct logical conclusions (Wilkins, 1928; Janis and Frick, 1943; Lefford, 1946; Morgan and Morton, 1944; Frase, 1966; and Wilson, 1965). Yet aspects of these studies leave some critical questions unanswered. Parrott and Johnson (1968) brought together several 3 of these early studies in an examination of the para- meters of the content effect. Virtually all the early studies which varied content across conditions made methodological mistakes which leave their conclusions very tenuous. The first, but probably less crucial problem is that of content specification or the appropriate classification of content. In the Wilkins (1928) study, the author used four types of content: 1. familiar ("all cats are mammals"); 2. symbolic ("all A is B"); 3. unfamiliar ("all zygatropes are metathanes"); and 4. suggestive ("all metal is wet"). Type 3 presents little classification difficulty, as the terms are all fictitious, and type 2 is certainly clear. But types 1 and 4 are quite difficult to reliably classify, as what is familiar to a graduate student in psychology may be highly suggestive to a freshman in education or foreign language. Wilkins personally classified the content of her items, as have others working with content effects (Janis and Frick, 1943; Lefford, 1946; Wilson, 1965). It should be evident that an a priori classification by a sample of one may be idiosyncratic. Any content manipulation must be based on a classification of content by a representative sample of judges, and Parrott and Johnson (1968) have presented data on the effects of true and false content on reasoning where all premises have been rated for truth- falsity by a representative sample of subjects. These 4 standardized statements will also be used in the study to be presented here. The second, most crucial, methodological problem in the early studies of content and reasoning (Wilkins, 1928; and Wilson, 1965) is that of confounding logical structure and content effects. Logical structure refers here to the four basic types of premises: 1. universal affirmative; 2. universal negative; 3. particular affirmative; and 4. particular negative. These early studies did not equate logical structure across the various content types, and Frase (1967) has shown logical structure may have a significant effect on the reasoning process. Parrott's studies controlled for logical structure and varied content, and the results clearly indicated a content effect on both number correct and reasoning time. The present series of studies will examine in more detail the time dimension in reasoning as an indicator of process. Very little research has been reported which attempts to examine the deductive reasoning process. The measure- ment of reasoning times may provide an insight into internal processing sequences. The time a subject takes to read the premises of a syllogism may represent an encoding phase, for the reading may involve something more than just following the superficial wording of the given statements, but also involve the internal holding of the given statements in storage in anticipation of their being later fed into a central processor for testing logical relationships Our knowledge of the time related processes in syllogistic reasoning is quite limited. Wilson (1965) examined the effects of group competition on the speed and accuracy of syllogistic reasoning, but his report leaves the reader aware of critical problems. First, Wilson used syllogisms which have content confounded with logical structure. This confounding of the content variable, which was of central importance with the uncontrolled variation in logical structure, makes all of Wilson's conclusions somewhat suspect. In addition, he reports that his data were collected in group testing conditions and does not describe how he measured accurately the reasoning time per item which he reports. The accuracy of his measurements for reasoning time must, therefore be held somewhat questionable; but, even more than this, it is possible that thetotal differences in reasoning times he reports could be due to original dif- ferences in reading times. That is, where he wishes to distinguish the effects of content on reasoning, all he may be reporting is a difference due to some regular, extraneous, variation in item length across the various content forms. With these reservations in mind, his data showed: 1. group competition increased reasoning time per item and improved accuracy, and 2. the order of solution time from the shortest to longest was familiar, biased, symbolic, and unfamiliar. Wilson offered 22 theoretical interpretation for this order of solution time and hence this report must stand as an initial empirical study of the syllogistic reasoning process. Frase (1966a) provided both data and theory in his study of the effects of affective and quantifying words on syllogistic reasoning. In some ways, this work is difficult to integrate into the trend of the earlier research in this area, for rather than focus on the problem of content as Wilson (1965) did, Frase manipulated the semantic incongruity of his premises. An example of a highly incongruous premise would be: "No natural things are creative." He found no significant effect of his incongruity manipulation on solution time, but to highlight the problems of interpreting results with items not balanced for logical structure, he found a significant effect on solution time of the "all" versus "some" quantifiers. "Some" premise statements required longer reasoning time. Frase noted Osgood's (1963) model of the language process for his theoretical integration. Osgood also emphasized the importance of quantifiers in sentence understanding as well as the lesser importance of affective characteristics. In studies of meaning, quanti— fiers obey a multiplicative model, while affective terms fit an additive model. Berlyne's (1960) notion of conceptual conflict, which would predict that more time would be spent on the more incongruous items was of central theoretical interest to Frase, but was not supported by his data. In a later paper, Frase (1967) likened the syllogism to a mediation paradigm of a simple chain type, and this notion also provided some predictions for empirical study in terms of associative difficulty as measured in syllogism solution time. This orientation might be termed neo—associationist and indicates the potential for taking S-R theory into the domain of cognitive processes. Unfortunately, predictions which Frase derived from mediation theory were 22E supported in terms of reasoning time differences across the four forms which he used. Parrott and Johnson (1968) report a study designed to clarify some of the effects of content on reasoning accuracy and time. Syllogisms were developed with three types of content, true, false, and mixed. The truth— falsity of the premise content in these items was scaled by having a large group of subjects rate each premise on a seven point scale. Sample items from the three scales are given below. Example from "true" premise scale. Varsity football is a very p0pular sport here at MSU. Popular sports have large crowds. Varsity football has large crowds here at MSU. Varsity football is pOpular at Notre Dame. MSU is a first-rate academic institution. John Hannah likes football. None of the above. ‘ m-RC’JNH Example from "false" premise scale. Physics is the most pOpular course of study here at MSU. POpular courses of study are very easy. Physics is very easy here. Elementary education is easier than physics. Physics is very difficult here. Physics students are very intelligent. None of the above. m-P-QJNH Example from "mixed" premise scale. Varsity football is a very pOpular sport here at MSU. All popular sports have very small crowds. Varsity football here has small crowds. Varsity football here has large crowds. Varsity football here is not very pOpular. Varsity football here provides much income for the university. None of the above. U'l #C’JNH These items were carefully matched for logical structure, and the average length of each item was equated across the three scales to control for reading time differ- ences. The data collected with this material provided the clearest insight yet into the effects of semantic ' content on accuracy and solution time in syllogistic reasoning. In terms of number correct, the mixed items were most difficult, then the true items, then the false. While these results are not of central interest here, the theoretical interpretation was based on Henle's (1962) discussion of the source of errors in reasoning. Henle noted that where errors are made in arriving at a desired logical conclusion, the problem may result from a lack of task comprehension, or at times, from a misperception of 9 the premises.) Subjects in Parrott and Johnson's (1968) study were given only the most elementary instructions, and thus there was great room for a misunderstanding of the task. The mixed premises would be further made difficult by a probable tendency to misperceive the single false premise. The tendency to misperceive premises would be much less for the false premise form, for the subject would very quickly note that the given background material in each item was of an unusual content, and additionally the tendency to accept a factually true, but logically invalid conclusion would be much less for the false items than for the true. In a second experiment in Parrott and Johnson's study, reasoning times for the three forms were collected; false items took longer than either the true or mixed items. An "opposite-content" or "opposite" transfer design was also used to study the effects of true-false and false-true presentation orders. True items following false took less time to reason with than true items in first position, but no facilitation was noted for false items after true. These data were interpreted in terms of category availability; false premises do not fit readily accessible categories and hence take longer to process than true premises or mixed premises, which would both be handled in often used claSsification or encoding categories. In terms of errors, the Henle (1962) suggestions fit these data; and for transfer effects, 10 it was noted that perhaps the false items produce what Berlyne (1957, 1965) has called "conceptual conflict" leading to arousal and greater speed of information processing. That is, the increase in arousal is not so great as to exceed the organism's "optimum level of cue function," as described by Hebb (1955). Here, we have inferred that because the false items take longer to reason with, they produce conceptual conflict leading to arousal and faster subsequent reasoning. It should be pointed out that arousal occurs as part of an extended reasoning process. If the above argument holds, any reasoning format that takes more time for processing, not just more reading time but actually thinking time, should facilitate later reasoning with other content. From this understanding, Parrott (1968, unpublished) has conducted a study of the reasoning time required for symbolic items and their possible transfer effects. The following is an example of a symbolic item. All A are B. All B are C. All A are C. All X is Y. All B is Y. All C is large. None of the above. Ln-thr-t Using symbolic and semantic items, grouping the true and false content under the general category of semantic, Parrott found that symbolic items took virtually 11 the same time to reason with as false items, and both false and symbolic took longer than true items. This effect is especially interesting for symbolic items, as their actual reading time was less than that for either true or false items by a significant degree (p <.001). Now, if arousal facilitates reasoning, then symbolic items should produce arousal and facilitate reasoning for subsequent true or false items. In addition the symbolic items might be facilitating in that they provide structural insight into the interrelationships in the semantic items. The semantic items are identical in structure to the symbolic items; even the alternatives are structurally identical across the true, false, and symbolic forms and are in the same position or order in all forms. Therefore, it was predicted that symbolic items would facilitate reasoning with later true and false items, but the pilot data did 222 bear out this prediction. Overall the symbolic pretraining had no significant effect on reasoning with semantic content, and therefore some theoretical reconsideration may be necessary. A THEORETICAL INTEGRATION The following figure represents an attempt at theoretical integration of the reasoning processes suggested by the time data collected so far. Figure 1 outlines an information processing model of reasoning. 12 .,___. Encode Dg; (Test #1) /\ (Test #2) this git Categorx\ Construct noapmawmoa Om mwwHHOQ man zmH "——* Code \ Reading / Reasoning *% Response Processe Figure 4. A theoretical distinction between reading and reasoning. 48 In the Osgoodian (1957) S-R model, oral reading or passive reading as the term is used here, involves stimulus-response connections at a middle level. The basic S-R pathways for passive reading are the same as for active or comprehension reading, but the level of internal complexity reached by the passive reading is lower. Because of the generality of S-R connections, this model might suggest that some representation coding was taking place even in passive reading, and thus the suggestion implicit in this model is that passive read- ing, because it does use some of the same S-R units of active reading, should provide positive transfer to later active reading. Admitedly then, the S-R paradigm can handle the active-passive reading comparison and produce some testable hypotheses. In the information processing model, the passive reading process is organized around a completely separate internal process unit from the active, comprehension reading. This information processing model of active-passive reading posits two separate encoding- decoding units; the passive reading activates the visual-oral recoding unit and does not involve any more complex units in the other systems. When the subject is required to actively encode information, that is for comprehension, the Metaplan calls up the active processing subroutine which stores input, and ultimately carries out the desired transformations on the input. Thus the 49 information processing model sketched here posits different conceptions of the reading process from the S-R model, for the active-passive reading task is seen as using two separate subroutines. Little transfer would be expected from passive to later active reading with this model, nor would passive reading be expected to produce any storage of input. These comparisons represent testable differences between the two approaches which both fit the data now reported. REASONING THEORY The S-R system appears to fit the active-passive reading data, and the neo—behavioristic model may be extended to apply to reasoning as well. Berlyne (1965) applies S-R conceptions to the analysis of higher-order responses, and Figure 5 diagrams his system. A. Berlyne's Transformation Chain So --—-(r _ s ) --- (r _ s ) --- (r s s )---R l l 2 2 "' n n Where (r - s ) = situational thought = u n n n B. Berlyne's Complete Thinking Process S o -- ¢l -- ul —- 02 —- ... un -—R Where 0 = transformational thought Figure 5. Berlyne's analysis of the S-R attack on thinking An So might be a syllogism premise which when read keys Off a specific chain of internal responses. Given 50 two premises or SO units and a transformational thought to bring them together, the series of situational thoughts would lead to a final reasoning conclusion. Using these systems of transformation chains and situational thoughts, it should be clear that the S-R approach can be adapted to fit much of what is called thinking. But it is quite alien to general S-R conceptions to posit situational thoughts and transformation units; these links appear to better fit a theoretical orientation prepared to accept complex internal units. In terms of syllogistic reasoning, a situational thought could be equated with a premise and a transformation unit with the combination rule to bring two premises together to form a concluSion. Hence the Berlyne S-R system can be applied to reasoning, but the S-R building blocks no longer seem to fit the classic behavioristic conceptions. The behavioristic analysis may be forced into the domain of cognition, but a transformational thought fits much more neatly into an honestly labeled cognitive theory. Cognitive and S-R theories cannot be sharply separated today, and even their theoretical units share much in common (Millenson, 1967). Therefore this paper will not attempt to further match their respective theoretical prowess; rather the data collected here will be applied back to rebuilding the structure-process model of the internal workings of the mind during reasoning. 51 STRUCTURE-PROCESS MODEL OF REASONING The data collected during the subject's reasoning trials indicated that symbolic items were handled differently than either the true or false content materials. In the transfer conditions, the correlations of the mean reasoning times between the first presented set of items and the second set were computed and showed that the symbolic items did not correlate as highly with the content items as the true and false items correlated with each other (p <.001). These data fit the original structural dimensions proposed in Figure l, for it was hypothesized that the true and false material would be handled by available category systems while the symbolic material would require a new category storage construction. This distinction between semantic and symbolic material is also consistent with Guilford's (1967) model of intellect. In addition, the data on switchbacks show that false items were more difficult to encode than either symbolic or true items. This might indicate that the subject is attempting to store the false premise in a true premise category and finding difficulty in adjusting the match between category and input as predicted in the category transformation unit in Figure 1. In contrast to the pilot data indicating that symbolic material might take much longer to reason with than true items, this study indicated that only the false 52 3 items take longer. This result also fits our initial process conception of a transformation requirement in the manipulation of false content. The examination of the processing transformation times also shows this difference for the false items, and it indicates the symbolic items do, in fact, take longer to process than the true items. This is also support for the initially proposed model. In terms of the processing of the various content types, a comment must be made about errors. It is most gratifying to note that none of the experimental treat- ments had any reliable effect on number correct. The trend of the differences Obtained suggested little theoretical difficulty, but further research could re- design the item format to produce even better control of the number correct variable. Since the number correct did not fluctuate as a function of any of the treat- ment effects or interactions, it may be assumed that the logical processes or internal transformations were made equally effectively in all groups. 3. One possible hypothesis is that these students were more prepared to deal with symbolic abstractions than those studies in the pilot project; the comparison of pilot subjects to those here on symbolic reasoning time with the 3 test indicated a significant difference between the two groups (p < .001). 53 The instructions for speed in processing, which may be seen as a "Metaplan" in the Miller, Galanter and Pribram (1960) system or as a 9set" in the more classic sense, were highly effective as predicted. Through the emphasis on speed-with-accuracy, a reduction in reasoning time of almost 25% as compared with standard instructions was obtained with no decrement in accuracy. This finding has not been shown by others working in the area of reasoning, and it implies that normal information processing performance is far from the optimal limit our subjects can achieve. This is somewhat in contrast to the work of Fitts (1966) and Suedfeld (1966), but the exact specification of speed and accuracy used here makes these results distinct from both the previous studies with simple speed manipulations. As Simon (1965) points out: "...the establishment of time constants for elementary processes will permit strong quantitative predictions of times required for specific compound problem solving and learning tasks." (p. 316) The results reported here for reading, standard reasoning and speed reasoning all add to our very meager knowledge of performance parameters in complex cognitive tasks. TRANSFER EFFECTS AND STRUCTURAL IMPLICATIONS The interpretation of the transfer effects shown in this study does not fit the suggested arousal-process formulation develOped in the introduction, as the Opposite content condition does not behave as expected 54 at all. It is very difficult to see why subjects should perform so differently on these items than on previous studies (Parrott, 1968; Parrott & Johnson, 1968), but one explanation is in the selection of only five items to represent each content set here as opposed to 10- item sets in earlier studies. It may be that certain items provide greater transfer than others and that these items were not represented in the current scales. The actual transfer effects are confounded, in addition, by an interaction with the instructional set. Generally, symbolic warmup facilitates later semantic reasoning, as it reduces reasoning times across both speed and standard instructions. This result would be support for a "structural" insight effect, where problem solution is facilitated by experience with the most basic organizational aspects of the problem. The positive transfer for the symbolic items is indicated most for the false content comparisons; this is rather consistent with the earlier suggestion that the processing aspects would be most likely to be improved for the most complex processing problems, specifically the false content items. It is clear from Figure 3 that under speed instructions symbolic items produce the greatest facilitation, or reduce the time the most for later false processing, while under standard instructions the true items show even better positive transfer than the symbolic. These results all seem to fit a strong 55 "structural insight" interpretation, where the critical problem-solving advantage is understanding the basic structure of the problem, rather than the arousal hypothesis first considered. The following attempts a pppp Egg integration of the transfer effects. Assumption #1: Problem structure is very important in playing the problem-solving game. a. Superficial content can mask basic structure and provide undertainty as to the correct solution process. Assumption #2: Structural learning, under normal conditions, is superior with familiar content as Opposed to symbolic terms. That is, where time for processing is relaxed, learning is more efficient when cast into normal conceptual elements. a. Under demands for speed, content tends to be complicating, and com- pletely symbolic processing provides basic structural insights that are missed in the speed processing of familiar content. Implicit in Assumption #1 is that false content might ‘trransfer negatively or slow down later reasoning with true liisems under standard instructions, and this is also shown 1]] our data. In addition, under speed instructions, the iialse content also increases later true reasoning times, “daile the true items do not show any appreciable positive tiransfer effect. These results are well contained in a l'lig'hly structure-oriented set of process postulates. These tflneoretical suggestions run counter to some data on Structural pretraining (Rundquist, et a1, 1965), but the CPiJSical difference between the concept attainment task uEMadin this earlier study and the reasoning task used here 56 may require different transfer processes due to the basic characteristics of the two problem types. Concept attainment is a process Of information storage-retrieval and hypothesis-construction—testing, while syllogistic reasoning is a task of pure structural processing, basically of finding the middle-term and using it to relate A and C. The series of test operations outlined in Figure 1 under the section "central processer" also fits our structural transfer results. If,as has been suggested, syllogistic reasoning is basically structural processing, then each TOTE unit operates most efficiently in the pure structural mode. The best test of this is to compare just the conclusion exposure times for the S, T, and F items; the prediction based on pure structural advantage would order the times: S, T, and then P; and this is how the data do rank. Yet the times for the S and T are very similar (13.00 seconds as compared to 13.85 seconds) and no statistically significant difference is indicated. S is less than F and T is less than F all at p <:.01.4 As there is a marked difference in the reading times for the S, T, and F forms, these conclusion exposure times could be roughly corrected for this confounding, 4. Individual 2 tests. 57 and the difference between the S and T conclusion times would increase. This correction would provide some difficulty for the structural hypothesis, however. Another set of comparisons in reference to structural factors focuses on the individual items. The classic atmosphere error as first suggested by WOodworth and Sells (1935) can be extended to predict faster reasoning on items of the following form: Logical Form All A is B A All B is C A or Some A is B I Some B is C I The ordering of these two relatively easily pro— cessed forms would be AA and then II, if the "some" effect is assumed to produce somewhat less semantic halo than the "all." The data for the symbolic problems show just this pattern of results. .The most rapidly solved form is AA, then the premise combinations of both IA and II are almost equal in total time, while the "No A is B, All B is C," or EA form is most slowly solved.5 These data also fit into some of the literature on the utility of negative information (Donaldson, 1959; Hovland and Weiss, 1953); for many problem solving tasks negative 5. These differences approach acceptable significance levels ( .10 > p;> .05), but are to be taken here as only tentative. 58 information is not economically processed. The effects of structure on encoding time parallel the orderings for total reasoning time and may suggest that the very encoding of negative information is more difficult than positive information. This may imply that negative state- ments must go through some type of transformation process to be adapted for storage. These orderings for structural effects obtain only where content is not present; complete- ly different orderings can be achieved through the manipulation of the semantic nature of the premises. SKILLED AND NAIVE SUBJECTS The comparison of skilled and naive subjects was greatly restricted by the small number of skilled reasoners coming to the experiment, but some tentative conclusions may be advanced on their performance. First, though the evidence is very weak, it appears that even subjects with some experience in reasoning are still affected by the content of the material. The order of their encoding times closely follows that for the naive subjects. Where the skilled subjects can be discriminated from the naive may be on their performance under speed instructions. Assuming that they have some prior experience in this task to draw on, they can work even more rapidly and effectively than the naive subjects when the situation demands it. This superior performance would certainly be expected, but it is slightly surpris- ing not to see it demonstrated under standard instructions 59 as well; one possibility is that prior training may make them cautious, hence slower, and because these items are so basically simple, very little might be expected due to subject differences on number correct. FUTURE RESEARCH Because of the difficulty with 23222 matching of conclusion truth-falsity across the true and false premise scales and the suggested transfer effect on number correct, a new item format might be adopted for subsequent research. Instead of a multiple-choice design, a single alternative might be used which could be responded to on an "agree- disagree" basis. Future research might concentrate on some of the prOposed process steps (see Figure 1) not considered in this report. Outlined here is a comparison of an S-R and information processing distinction between active and passive reading. The storage phase in the active reading or reasoning model implies that the baCkground information is held in its originally encoded form while additional information is encoded and final pro- cessing is taking place. This means that we may interrupt our subject's processing activity and get an accurate readout of our original input. Another possible study might involve the examination of structural insights as a function of warmup activity. DO subjects really obtain conscious insight into the nature Of these prob- lems by working on symbolic material first? A sensitive "TJE '2. 6O post-experimental questionnaire might probe this insight question. Another line Of research suggested by this study might examine the effect of logical structure (the A, E, I, and 0 forms of information statement) on semantic categorization or encoding and information processing. The data from this study can only be taken as suggestive on the nature of these effects, but it would appear that the universal affirmative or A state- ment is even easier to encode than the particular affirmative or I, and most difficult is the universal negative or E. This order of difficulty fits implicit considerations in the information processing model discussed here (Figure 1), as it might be assumed that the negative statements need to go through a trans- formation process to fit an available category or storage unit. EXPERIMENT IV: DOGMATISM, RIGIDITY AND REASONING The relationship of personality factors to reasoning has not been entirely neglected in the psychological literature. The influence of attitudes on syllogistic reasoning has been studied by several investigators (Morgan & Morton, 1943, 1944; Henle & Michael, 1956; and Thistlewaite, 1950), and Frase (1966) has studied the relationship between belief in conclusions and the amount of incongruency of the sentence used in the conclusion. Of more relevance to a discussion of specific per- sonality variables as they relate to reasoning is some of the research with the Rokeach (1960) Dogmatism Scale. Beech (1964, unpublished) used both the Dogmatism Scale and the Cough-Sanford Rigidity Scale and compared sub- jects on analogy problems emphasizing induction or deduction. It was hypothesized that rigid subjects should differ from non-rigid subjects on the problems which emphasized deduction, but that on these problems no differences should exist between open as compared to closed subjects. On the problems emphasizing in- duction, also called synthesis, Open subjects should be superior to closed subjects, but no differences should exist between rigid and non-rigid subjects. Significant 61 62 differences were found between Open and closed subjects, but no differences were found between rigid and non- rigid subjects. Levy and Rokeach (1960) found significant differences between high and low dogmatic subjects on a modified Kohs Test (of perceptual synthesis), but no differences were found on the Witkin Embedded-Figures Test. Huberman (1961) also found differences between high and low dogmatic subjects on perceptual synthesis, and in addition found differences between these extremes on the Witkin Test. Rokeach, McGoveny and Denny (1955); Rokeach and Vidulich (1960); Rokeach, Oram, Laffey and Denny (1960); Rokeach, Swanson and Denny (1960); Fillenbaum and Jackman (1961); Vidulich (1961); Beech (1964, unpublished); Conway (1963); and Hoppe (1962) have all found significant differences in synthesis of belief time between high and low dogmatics on what is now the classic flJoe Doodlebug" problem. In the process of syllogistic reasoning, the subject is presented with two background statements, and his task is to synthesize these two statements into one final form. Thus the process of syllogistic reasoning appears quite analogous to that of many other synthesis problems. In this study, two types of item content were used: true item content and false item content. Based on these earlier studies with Open and closed subjects, and on Rokeach's (1960, p. 185) hypothesis that 62 differences were found between Open and closed subjects, but no differences were found between rigid and non- rigid subjects. Levy and Rokeach (1960) found significant differences between high and low dogmatic subjects on a modified Kohs Test (of perceptual synthesis), but no differences were found on the Witkin Embedded-Figures Test. Huberman (1961) also found differences between high and low dogmatic subjects on perceptual synthesis, and in addition found differences between these extremes on the Witkin Test. Rokeach, McGoveny and Denny (1955); Rokeach and Vidulich (1960); Rokeach, Oram, Laffey and Denny (1960); Rokeach, Swanson and Denny (1960); Fillenbaum and Jackman (1961); Vidulich (1961); Beech (1964, unpublished); Conway (1963); and Hoppe (1962) have all found significant differences in synthesis of belief time between high and low dogmatics on what is now the classic "Joe Doodlebug" problem. In the process of syllogistic reasoning, the subject is presented with two background statements, and his task ; is to synthesize these two statements into one final form. Thus the process of syllogistic reasoning appears quite analogous to that of many other synthesis problems. In this study, two types of item content were used: true item content and false item content. Based on these earlier studies with Open and closed subjects, and on Rokeach's (1960, p. 185) hypothesis that 62 differences were found between Open and closed subjects, but no differences were found between rigid and non- rigid subjects. Levy and Rokeach (1960) found significant differences between high and low dogmatic subjects on a modified Kohs Test (of perceptual synthesis), but no differences were found on the Witkin Embedded-Figures Test. Huberman (1961) also found differences between high and low dogmatic subjects on perceptual synthesis, and in addition found differences between these extremes on the Witkin Test. Rokeach, McGoveny and Denny (1955); Rokeach and Vidulich (1960); Rokeach, Oram, Laffey and Denny (1960); Rokeach, Swanson and Denny (1960); Fillenbaum and Jackman (1961); Vidulich (1961); Beech (1964, unpublished); Conway (1963); and HOppe (1962) have all found significant differences in synthesis Of belief time between high and low dogmatics on what is now the classic "Joe Doodlebug" problem. In the process of syllogistic reasoning, the subject is presented with two background statements, and his task is to synthesize these two statements into one final form. Thus the process of syllogistic reasoning appears quite analogous to that of many other synthesis problems. In this study, two types of item content were used: true item content and false item content. Based on these earlier studies with Open and closed subjects, and on Rokeach's (1960, p. 185) hypothesis that 63 open and closed subjects should differ on the synthesis type of problem solving, it was hypothesized that the Dogmatism score should correlate positively with the reasoning time for false items and negatively with true items. True items fall into the closed subjects tightly knit, well-organized belief system, therefore the actual processing time for these statements will be less for the closed subject than for the Open subject with the highly complex and interwoven belief system; he will be faced with greater uncertainty at each processing step than the simplex system. False items fall outside the closed subject's belief system, and he thus has to somehow either transform the input or his own system in order to handle the task--either alternative taking more time than for the Open subject who is prepared for very flexible information input and handling. The Gough- Sanford Rigidity Scale tends to measure a behavioral type of intention or performance, and thus no relationship between it and either form of reasoning was expected. METHOD Subjects: The data used here pertain to the subjects from Experiment III who were in the standard instructions T-F and F-T presentation orders; total N=84. Materials and Procedure: Subjects participated in the normal EXperiment III assignments; after the reasoning task, they completed the Rokeach Dogmatism Scale and the Cough-Sanford Rigidity Scale. Responses were on a two 64 choice, "agree-disagree," scale, and D and R scores were calculated as the number of items with "agree" responses. These scores were then correlated with the reasoning times for Treatment One data only, in order to avoid confounding with transfer effects. RESULTS Table 16 summarizes the correlations of D, R, and reasoning. Table 16 Correlations of Dogmatism and Rigidity Scores with Reasoning Time Content D R True -.18 .07 False +.23 .04 While neither D correlation is significantly different from Zero, the difference between the D cor- relations (to T and F reasoning times) has a p ,<'05 (one— tailed), and it is clear no significant effects or differences are shown between the R scores and the reasoning time measure. As part of the study of the relationship of Dogmatism, Rigidity, and reasoning, D and R scores were collected on 311 subjects in Experiment III. An examination of these scores as a function of one of the main treatment effects produced the following data. 65 Table 17 Reasoning Instruction Effects on D and R Scores Instructions Standard . Speed Dogmatism Score X' 15.92 13.69 SD 5.26 5.70 Rigidity Score X 9.87 9.26 SD 3.18 3.54 The difference between the D scores after reasoning under standard and speed instructions is statistically significant at p '<.Ol (F = 9.88; 1/228 df), and the comparisons between the R scores does not even approach statistical significance (F = 1.96; 1/228 df, p:> .05). In order to understand whether the change in D scores is toward greater or lesser dogmatism, these means may be compared to that Obtained in a large sample of students similarly enrolled in introductory psychology, In this sample of 271 students, the mean D score was 15.88 (SD = 5.37)and the mean R score was 9.38 (SD = 3.46). DISCUSSION As predicted, the direction of the Dogmatism Scale correlations indicated that high dogmatiCs are faster reasoners on familiar material and slower on material outside their normal belief system. The difference between these two correlations was statistically sig- nificant, and thus these data support Rokeach's (1960) 66 general theoretical system. It should be reasonable to suggest that the high dogmatic is faced with less un- certainty in the processing of familiar beliefs for he has fewer complex interconnections between belief elements than the low dogmatic, who has many inter— connections, hence decision points, in his diffuse belief system. On unfamiliar beliefs, or in this case, beliefs contrary to everyday experience, the high dogmatic is at a loss; for his processing system has little preparation for their processing. Rigidity as defined by the Gough-Sanford Scale should not be expected to relate to the solution times on these items, as it measures more of a behavioral or perseveration type of cognitive dimension. Hence, we see here a clear distinction between dogmatism and rigidity as Rokeach (1960) has also reported. The post hoc examination of the dogmatism and rigidity scores as a function of the experimental treat— ments in Experiment III is much more tentative and the results taken as suggestive only. It does not seem un- reasonable to assume that requiring subjects to reason with items under the pressure to be both fast and accurate might have some effect on the general organiza- tion of their belief or information processing systems. Across all types Of items an Open information process- ing system would be most effective and rapid; and hence the demand of speed and accuracy may serve to "program" 67 the subjects to be more Open than is their normal state. It is clear that the change is toward more Open systems when the treatment groups are compared with the large unselected sample of introductory psychology students. These results are "suggestive" in that they would pre- dict that environmental demands might alter an individual's normal cognitive system. Future studies should examine the validity of this effect, its temporal duration, the generality of the effect across different situations, and the possible use of the open state to facilitate learning. Through an item analysis, the exact nature of the restructuring might be examined. The fact that the post hoc comparison does not show any effect on the Rigidity Scale scores is another clear indication of the difference in the cognitive-behavioral nature of dogmatism and rigidity. Our task was a conceptual one, and it involved the need fOr conceptual restructuring, which dogmatism measures; the rigidity dimension is really not activated by our problem demands and thus no changes in rigidity scores need be expected for this conceptual activity. 68 SUMMARY Using syllogistic reasoning as the focus, this paper has examined the effects on reasoning processes of: 1. speed and standard instructions; 2. type of warmup or transfer conditions; and 3. true and false content. After first obtaining simple reading time on all content types, then checking for the truth-falsity of the multiple-choice alternatives in the T and F scales, a 3 x 2 x 2 experiment was conducted using as the dependent variables: 1. premise reading time; 2. total reasoning time; 3. number of switchbacks; and 4. number correct. No regular subjects had prior training in logical reasoning. As expected, no differences obtained for number correct as a function of any of the main effects or interactions, and this suggested that the basic reasoning process leading to a correct solution was relatively matched across all groups. Any differences in solution time could not be associated with greater or fewer errors, but rather with the semantic processing of the given material. As predicted, speed instructions reduced reasoning time; no content-by-instructions interaction was indicated. Also in line with predictions, true items took significantly less time to reason with than false, even though they were closely matched for reading time. In another comparison, symbolic items did not differ from true items on reasoning time as expected, but a measure 69 of net processing time that corrected for the much shorter reading time required by the S items did indicate a significant difference between S and T items. Symbolic reasoning and processing times were much less than those for false items. The transfer comparison yielded a significant over- all E, but the source of the major facilitation was in I? the symbolic warmup, indicating a structural insight, ”* rather than in the opposite content, indicating an arousal effect. A transfer—by-instructions interaction ’ 4 was present as the Opposite content warmup showed negative transfer under speed demands while the symbolic warmup appeared even more effective at facilitating the false content reasoning. A significant transfer-by- content interaction was due to the transfer effect speeding up only the false content items. Several alternative analyses of the main effects as they related to encoding or premise exposure time and the number of switchbacks, another measure of encoding difficulty, were also reported; all generally supported the basic content and transfer effects as measured by total reasoning time. The discussion Of this main study emphasized the importance of the basic information on the parameters of reasoning speeds as well as how both S-R and information processing theories might be used in integrating these results. An information processing model was discussed, 70 and the nature of structural effects in reasoning processes examined as they might relate to the central processing of encoded information. A final chapter reported a correlation study of dogmatism, rigidity and reasoning times. As predicted, the Rokeach Dogmatism Scale correlated positively with overall reasoning time for false content and negatively :1 with true content; the difference between the two fin correlatiOns was statistically significant. NO relation- ships were expected and none Obtained between the Gough— J " Sanford Rigidity Scale and either content reasoning time. , These results were fit quite neatly into Rokeach's (1960) belief-disbelief system, but another finding in the examination Of the post-experimental measurement of 2 and 3 indicated that subjects who had been in the speed instructions group of the main experiment had significantly lower Q scores than those in the standard instructions group. No effects on 3 scores were seen. These findings were quite unexpected, but it was suggested that the speed demand forced the subjects to open up their cognitive systems and this Open state was reflected in the later 2- score measurement. BIBLIOGRAPHY BIBLIOGRAPHY Beech, R. P. The influence of dogmatism and rigidity on reasoning. Master's thesis, Michigan State University, 1964. Berlyne, D. E. The influence Of complexity and novelty in visual figures on orienting responses. Journal pf Experimental Psychology, 1958, 55, 289-296. Berlyne, D. E. Structure and direction 13 thinking. New York: Wiley & Sons, 1965. Conway, J. A. Problem—solving in small groups as a function of "Open-minded" and "closed-minded" behavior: some implications for development of administrative theory. Ph. D. thesis, State University of New York at Albany, 1963. Donaldson, M. Positive and negative information in matching problems. British Journal pf Psychology, 1959, 50, 235-262- Fillenbaum, S. and Jackman, A. Dogmatism and anxiety in relation to problem solving--an extension of Rokeach's results. Journal of Abnormal and Social Fitts, Paul M. Cognitive aspects of information processing: III. set for speed versus accuracy. Journal pf Experimental Psychology, 1966, 71, 849-857. Frase, L. T. Validity judgments of syllogisms in re- lation to two sets of terms. Journal pf Educational Psychology, 1966, 57, 239-245. Frase, L. T. Mediation paradigms and syllogistic reasoning behavior. Proceedingg pf the 74th. Annual Convention pf the APA, 1966. Frase, L. T. Cybernetic control of memory while reading connected discourse. Journal pf Educational Psycholqu, 1969, 60, 49-55. 71 72 Guilford, J. P. The nature pf human intelliggnce. New York: McGraw-Hill, 1967. Hebb, D. 0. Drive and the C.N.S. (Conceptual nervous system). Psychological Review, 1955, 62, 243-254. Henle, M. On the relation between logic and thinking. Psychological Review, 1962, 69, 366-378. Henle, Mary and Michael, Mirian. The influence of attitudes on syllogistic reasoning. Journal pf Social Ps cholo , 1956, 44, 115-127. Hoppe, R. A. Individual and group memorization and problem-solving: a test of the Lorge-Solomon pooling-of-abilities model. Ph. D. thesis, Michigan State University, 1962. Hovland, C. I. and Weiss, W. Transmission of information concerning concepts through positive and negative instances. Journal pf Experimental Psycholpgy, 1953, 45, 175-182. Huberman, J. A re-examination of certain aspects of Rokeach's study on dogmatism. Master's thesis, University of British Columbia, 1961. Janis, T. L. and Frick. F. The relationship between attitudes toward conclusions and errors in judging logical validity of syllogisms. Journal pf Experimental Psychology, 1943, 33, 73-77. Johnson, D. M., Parrott, G. L. and Stratton, R. P. Production and judgment of solutions to five problems. Journal pf Educational Psychology, 1968, 59, 1-21. Lefford, A. The influence of emotional subject matter on logical reasoning. Journal pf General Psychology, 1946, 34, 127-151. Levy, J. M. and Rokeach, M. The formation of new perceptual systems. in M. Rokeach, The open and closed mind. New York: Basic Books, 1960. Martin, E. Stimulus meaningfulness and paired associate transfer. Psychological Review, 1968, 75, 421-441. Millenson, J. R. An isomorphism between s-r notation and information processing flow diagrams. Psychological Record , 1967, 17, 305-319. 73 Miller, G. A., Galanter, E., and Pribram, K. Plans and the structure of behavior. New York: Holt, Rinehart and Winston, 1960. Morgan, J. J. and Morton, J. T. Distorted reasoning as an index of public opinion. School and Society, 1943, 57, 333-335. Morgan, J. J. and Morton, J. T. The distortion of syllogistic reasoning produced by personal con- victions. Journal pf Social Psychology, 1944, 20, 39-59. Osgood, C. E. A behavioristic analysis of perception and language as cognitive phenomena. In Contemporary approaches 22 cognition. Cambridge, Massachusetts: Harvard University Press, 1957, 75-118. Osgood, C. E. On understanding and creating sentences. American Psychologist, 1963, 18, 735-751. Parrott, G. L. Effects of instructions and transfer on syllogistic reasoning. unpublished pilot study. Michigan State University. 1968. Parrott, G. L. and Johnson, D. M. Effect of premise content on accuracy and solution time in syllogistic reasoning. paper at Midwestern Psychological Association, Chicago, May 2, 1968. Rokeach, M. The open and closed mind. New York: Basic Books, 1960. Rokeach, M., McGovney, W. C., and Denny, M. R. A distinction between dogmatic and rigid thinking. Journal pf Abnormal and Social Psychology, 1955, 51, 87-93. Rokeach, M., Oram, A., Laffey, J. J. and Denny, M. R., On party-line thinking: an experimental analogy. in M. Rokeach, The Open and closed mind. New York: Basic Books, 1960. Rokeach, M., Swanson, T. S. and Denny, M. R. The role of past experience: a comparison between chess players. in M. Rokeach, The Open 22g closed mind. New York: Basic Books, 1960? Rokeach, M. and Vidulich, R. N. The formation of new belief systems: the roles of memory and the capacity to entertain. in M. Rokeach, The Open and closed mind. New York: Basic Books, 1960. 74 Rundquist, E. A. and Jones, E. I. Preparation for problem solving: structural vs strategy pretraining. USN PRA TB No. 66-1. Simon, H. A. The use of information processing languages in psychology. Colloqnes Internationaus Du Centre National De La Recherche Scientifique. Paris, France July 5—10, 1965. - Simpson, M. E. and Johnson, D. M. Atmosphere and conversion errors in syllogistic reasoning. Journal pf Experimental Psychology, 1966, 72, 197—200. Suedfeld, P. Information processing: the effects of differential pattern complexity and input rate. Psychonomic Science, 1966, 6, 249-250. Thistlewaite, D. L. Attitude and structure as factors in the distortion of reasoning. Journal pf Abnormal and Social Psychology, 1950, 45, 442-458. Vidulich, R. A re-examination of the distinction between rigid and dogmatic thinking. unpublished manuscript, 1961. Wilkins, Minna C. The effect of changed material on the ability to do formal syllogistic reasoning. Archives of Psychology, 1928, 16, No..102. Wilson, W. R. The effect of competition on the speed and accuracy of syllogistic reasoning. Journal pf Social Psyphology, 1965, 27-32. Woodworth, R. S. and Sells, S. B. An atmosphere effect in formal syllogistic reasoning. Journal of EXperimental Psycholpgy, 1935, 18, 451-4607- APPENDICES APP END IX A STATEMENT SCALING FORM--D Instructions: You are to read the following statements and rate or score them by marking on the accompanying IBM answer sheet a number from one to seven on the scale which you have been given below. These statements may vary from one end of the scale to the other; your own judgment is the best answer. A statement neither one way or the other or indeterminate should receive a score of four (4). If you have any questions, please ask the experimenter. DO NOT MAKE ANY MARKS IN THIS TEST "BOOKLET!" Rating Scale for all the following items: 1 - 2 — 3 - 4 - 5 - 6 _ 7 completely % & % completely TRUE FALSE Use this scale for each Of the following items: CX)\]O\U1-§bo \O 10. ll. 12. 13. Some elephants are cats. Some felines are elephants. Some elephants are not felines. NO cats are huge. Some airplanes have powerful jet engines. A rocket engine is more powerful than a jet engine. Some jet engines are used in racing boats. Some airplanes have prOpeller-type engines. Some automobiles leak during rainstorms. Some automobiles may not leak during rainstorms. Some automobiles do not leak during rainstorms. No automobiles leak during rainstorms. No politician is a dictator. 75 STATEMENT SCALING FORM-—D Continue using the same rating scale as before: 1-2-3-4-5-6-7 completely % & % completely TRUE FALSE 14. No politician is assassainated. 15. No dictators are politicians. 16. Both politicians and dictators exercise political ! control over their countrymen. 17. All mammals can see well at night. 18. A11 cats have teeth. 19. All mammals give birth to live young. 20. All cats have furry skins. 21. Some females are cows. 22. Some giraffes are females. 23. Some females are not giraffes. 24. No cow has a long neck. 25. Some men eat special rocket food. 26. There is no such thing as special rocket food. 27. The fastest runner in history did not even achieve 30 miles per hour. 28. High speed running requires much training. 29. Some trucks are birds. 30. Some birds can fly. 31. Some trucks are not birds. 32. No truck is a bird. 33. No students at MSU are idiots. 34. No students at MSU do very well on national achieve— ment tests. 35. There are no idiots at MSU. 76 STATEMENT SCALING FORM-~D Continue using the same rating scale as before: 1 - 2 - 3 - 4 - 5 - 6 - 7 completely % & % completely TRUE FALSE 36. Working hard is a sign of a desire for achievement-- a very good thing in these times. 37. All ivory hunters kill elephants. 38. All mammals hunt elephants. 39. All cats hunt mice. 40. A11 elephants are larger than cats. Take careful note that you have just recorded your last rating on item #40 on the IBM answer sheet. Thank you very much for your assistancell 77 APPENDIX B APPENDIX B The following is the example item. A. All football players are strong and healthy. All strong and healthy peOple are much above average in intelligence. 1. All professional football players are well paid. 2. All football players wear protective clothing while playing. 3. All strong and healthy football players would like to be professionals. 4. All football players are much above average in intelligence. 5. None of the above. The following 5 items represent the symbolic content materials. 1. All A is B. All B is C. 1. All X is Y. 2. All A is C. 3. All B is Y. 4. All C are large. 5. None of the above. 2. No A is B. Only C is B. 1. No A is C. 2. No A is X. 3. No C is A. 4. Only Y matters. 5. None of the above. 78 opha‘ 3. Some A is B. Some B is C. 1. Some A are C. 2. Some C are F. 3. Some A are not C. 4. No A are C. 5. None of the above. 4. Some A is B. All B is C. 1. Some A is C. 2. C does not exist. 3. A does not make sense. 4. Y does not relate to this. 5. None of the above. 5. No A is B. All B is C. 1. Some A is B. 2. Some C are A. 3. Some A are not C. 4. No B is X. 5. None of the above. The following 5 items represent the true content materials. 1. All cats are mammals. All mammals have teeth. 1. All mammals can see well at night. 2. All cats have teeth. 3. All mammals give birth to live young. 4. All cats have furry skins. 79 5. None of the above. No politicians expect to remain in office forever. All dictators expect to remain in office forever. 1. No politicians are dictators. 2. No politicians are assassinated. 3. No dictators are politicians. 4. Both dictators and politicians exercise political power over their countrymen. 5. None of the above. Some automobiles have convertible t0ps. Some convertible t0ps are prone to leaking during rainstorms. 1. Some automobiles leak during rainstorms. 2. Some automobiles may not leak during rainstorms. 3. Some automobiles do not leak during rainstorms. 4. No automobiles leak during rainstorms. 5. None of the above. Some airplanes fly faster than sound. Breaking the sound barrier requires a powerful jet engine for propulsion. 1. Some airplanes have powerful jet engines. 2. A rocket engine is more powerful than a jet engine. 3. Some jet engines are used in racing boats. 4. Some airplanes have prOpeller-type engines. 5. None of the above. No elephants are cats. All cats are felines. 1. Some elephants are cats. 2. Some felines are elephants. 8O 3. 4. 5. Some elephants are not felines. No cats are huge. None of the above. The following 5 items represent the false content material. 1. All mammals are cats. All cats hunt elephants. l. 2. 3. 4. 5. All ivory hunters kill elephants. All mammals hunt elephants. All cats hunt mice. All elephants are larger than cats. None of the above. No students at MSU study hard for their classes. Only idiots study hard for their classes. 1. 2. 5. No students at MSU are idiots. No students at MSU do very well on national achievement tests. There are no idiots at MSU. Working hard is a sign of a desire for achievement-- a very good thing in these times. None of the above. Some trucks travel under water. Some things that travel under water are birds. 1. 2. Some trucks are birds. Some birds can fly. Some trucks are not birds. No truck is a bird. None of the above. 81 Some men run Breaking the food. 1. 2. 3. 4. 5. Some men There is faster than sound. "sound barrier" requires special rocket eat special rocket food. no such thing as special rocket food. The fastest runner in history did not even achieve 30 miles per hour. High speed running requires much training. None of the above. No females are cows. All cows are giraffes. 1. kin->0.) Some females are cows. Some giraffes are females. Some females are not giraffes. No cow has a long neck. None of the above. 82 APPENDIX C APPENDIX C POST-REASONING EXPERIMENT QUESTIONNAIRE NAME: 1. Have you had any training in logical reasoning, i.e. in a philos0phy, speech, math, or other class in either high school or college? YES NO If "YES" please explain. Was-there anything that specifically helped you in your work on the reasoning test items? 2. Attempt to recall the process you went through in the solution of the items presented to you in the reasoning test; outline below your strategy or system for coming to a conclusion for the test items: 3. a. (for subjects who had items with "All A is B" format) What did you do when you read this symbolic form? Did you work with the material in terms of "A" and "B", or did you reword the statements into familiar terms, i.e. "All cats are mammals?" b. (for subjects who had items in unfamiliar terms, i.e. "All mammals are cats," or "All trucks fly under water.") What did you do when you read these "false" statements? Did you work with the material in this false wording, or did you reword the statements, and then seek a solution, finally rewording the solution back into false terms for your actual response on the reasoning panel? Describe in detail how you handled the "false—content" items: 83 84 Do you have any additional comments about this experiment you would like to make? Had you heard any details about this eXperiment before you actually participated? Did you have any information about the specific task this experiment involved? Please refrain from discussing the details of this experiment with your classmates, as they may yet participate, and each subject should come to the experiment with the same level of information. Specifically, please do not discuss the content of the problems or the nature and format of the problems!