THE DEVELOPMENT AND TESTING OF A PAIRED - COMPARISONS FIGURAL SCALE TO MEASURE PREFERENCE FOR COMPLEXITY by SHELLEY GABRIELE WICHERT B.A., University of British Columbia, 1970 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS in the Department of Educational Psychology We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA April, 1973 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the Head of my Department or by his representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of The University of British Columbia Vancouver 8, Canada ii ABSTRACT The purpose of this study was to develop and to test a paired-comparisons figural scale to measure preference for complexity. A Random Shapes Scale (RSS) consisting of 18 sets of 3 random shapes was constructed. In each set of 3j one shape was of high complexity, one of medium complexity and one of low complexity. The random shapes were chosen from the eleven hundred generated by Vanderplas. Two existing measures of preference for complexity, the Barron-Welsh Art Scale (BW) and the Revised Art Scale (RA) were also used. Students in architecture, art, education, law and engineering (N=292) were tested using the RSS. Three weeks later the same groups of students (N=1S^) were retested on the RSS and completed the BW and KA as well. The BW and RA were significantly correlated with the RSS in three of the five groups tested. The internal consistency of the RSS calculated over all groups combined was .66; the stability coefficient was .71. The analysis of variance showed significant differences among the five groups tested. Therefore the RSS does differentiate among groups on the dimension of preference for complexity. The majority of the items were highly correlated with total test scores. This indicates that the items are homogenous. I The results of the conclusion that the RSS dimension of preference iii statistical analyses lead to the is a useful measure of a unitary for complexity. iv Table of Contents Page List of Tables v i List of Figures v i l List of Appendices Chapter 1 The Problem 1 The Need for a new Scale to • • 2 Measure Preference for Complexity 2 Related Research . • • 3 The Development of the Barron- • 4 Welsh Art Scale Selected Studies in which the . 3 Barron-Welsh Art Scale was Used Selected Studies in which the . 11 Revised Art Scale was Used The Development of the Rep Test . 12 The Development of the Bieri- • • ^ Blacker Test of Preference for Complexity Selected Studies in which the . . 14 Rep Test and Rorschach were Used Summary . 15 I Procedures Development of the Random . . , Shapes Scale Association Value of the . . Random Shapes Complexity of the Random . • . Shapes Materials . « • • • • • « Data Collection and Manipulation Results « • • • • • • • • • Analyses of the Reliabilities . of the Random Shapes Scale Item Analysis . Correlational Analyses • • • Analysis of Variance • , • • Summary, Conclusions and . . . Summaries and Conclusion . , References Appendices 18 19 20 2? 24 31 40 40 42 45 51 56 56 60 65 vi List of Tables Table Title Page 1 Fourteen Measures Calculated 23 for each of the Geometric Shapes 2 Scheme for Counterbalancing 27 Complexity 3 Summary of Item Scoring Procedure 32 for the Random Shapes Scale 4 Number of Subjects Tested by 35 Occupational Group 5 Item Numbers for each of Six 37 Scales of the Barron-Welsh Art Scale Booklet 6 Group Means, Standard Deviations, 41 and Reliabilities for the first Session of the RSS 7 Means, Standard Deviations, and 43 Test Retest Stability Coefficients for each Group g Total Groups Item Means, Standard 44 Deviations, and Correlations with Total Scores 9 Total Group Choice Distributions 46 Calculated as Percentages 10 Items in the RSS that Discriminated 47 Among Groups 1 1 Total Group Correlations Among 49 Measures of Preference for Complexity 12 Summary of Group Contrasts 52 1 3 Analysis of Variance 56 List of Figures Figure Title 1 Mean Score Over Items for Each Group on Test and Retest I LIST OF APPENDICES APPENDIX TITLE A Vanderplas Factor Loadings of Sixteeri Variables B Figure Preference Scale C Item Means, Standard Deviations, and Correlations with Total Score D Choice Distributions Calculated as Percentages E. Correlations among Measures of Preference for Complexity viii PAGE 66 67 74 80 86 1 Chapter 1 The Problem The purpose of this study is to develop and to test a paired-comparisons figural scale to measure preference for complexity. In %9k7 Welsh (Preliminary Manual: Welsh Figure Preference Scale, 1959) designed a figural-content scale wherein a rater indicated whether he liked or disliked each of four hundred drawings. A factor analysis of the ratings obtained yielded two major factors: one was response style; the other was a dimension which indicated whether individuals tended to prefer simple and symmetric forms or complex and assymmetric ones (Barron, 1952). From the original four hundred i-cems, Barron and Welsh derived two scales to differentiate among persons' preferences for complexity: First the Barron-Welsh Art Scale (BW) and later the Revised Art Scale (RA). Results from studies in which the BW was used led Barron to suggest that the basic distinction between the two types of perceptual preference (simplicity-complexity) is to be made in terms of "... a choice of what to attend to in the complex phenomena which make up the world we experience (pp. 398-399? 1952)." I 2 The Need for a New Scale to Measure Preference for Complexity Criticisms which will be described in detail in chapter 2, have been levelled at both the BW and the RA. Briefly, these.criticisms are that both scales measure several types of complexity which are not necessarily related, yet only one complexity score is arrived at. Secondly, the association value, or meaning of the drawings has apparently not been controlled for. Subjects may be responding to the meanings rather than to the complexity of the drawings. The purpose of this study is to construct a scale that will measure preference for complexity; and to avoid giving grounds for those criticisms levelled at the BW and the RA. The scale is to be evaluated by providing answers to the following questions; a. Can a figural-comparisons scale be devised to reliably measure preference for complexity? b. How stable is the scale over time? c. How internally consistent is the scale? d. Both the BW and the HA distinguish among different occupational groups. Does the scale also distinguish among different occupational groups? e. Are the test items homogeneous? f. Since the BW, the RA, and the scale are all designed to measure preference for complexity, how are these measures correlated? 3 Chapter 2 Related Research In this chapter, the development of existing measures of preference for complexity will be discussed. Selected studies in which these measures are used will be reviewed. Chapter 2 is organized into the following sections: a. a discussion of the development of the Barron-Welsh Art Scale (BW), b. selected studies in which the Barron-Welsh Art Scale was used, c. selected studies in which the Revised Art Scale (RA) was used, d. the development of the Kelly Repertory Test (Rep Test), e. the development of the Bieri-Blacker scoring modification of the Rorschach, f. selected studies in which the Rep Test and Rorschach were used, and g. summary. 4 The Development of the Barron-Welsh Art Scale The Welsh Figure Preference Test (WFPT) consists of four hundred black and white line drawings which range from very simple geometric figures to very complex designs. They were drawn with many variations to include differences in line quality, shape, content and other aspects of the figure (Preliminary Manual; Welsh Figure Preference Test, 1959). In the original form of the test the drawings were on cards which were presented individually. The test is now in the form of a fifty page booklet with eight drawings per page. The subject (S) is asked to decide for each of the drawings whether he likes it or does not like it. The responses "Like" (L) and "Don't Like" (DL) are marked on the answer sheet that accompanies the test. Welsh stated in the 1959 Preliminary Manual, that he found the absolute objectivity in the scoring system of the MMPI appealing. In 1947 he began to develop the WFPT using non verbal stimulus materials suitable for a wide range of subjects who could not be readily tested by conventional personality inventories and projective methods. As with the MMPI, the scoring system was objective. Twenty-seven different scores can be obtained from the WPFT; one of these is the BW. This is an empirically 5 derived scale, obtained by comparing the response frequen-cies of thirty-seven artists and art students with those of one hundred fifty 'people in general'. From the four hundred items in the WFPT, a scale of sixty-five items, all showing differences between the groups, was developed. Forty items were significantly more often disliked (p£.0l), and twenty-five more often liked (p£.05) by the artists than by the 'people in general' (Barron, 1952). The mean score of the artists was 40.25, that of the 'people in general' was 16.9. Barron and Welsh (1952) cross-validated the scale. The mean score of the thirty artists was 39.07, that of the thirty 'people in general' was 13.37. The original sixty-five item BW was later decreased in size to sixty-two items. Two items were removed because they were duplicates, and one item was deleted because it could not be included for projection on screen. The BW in its present form is published in an eighty-six item booklet with thirty-eight items keyed »DL' and twenty-four items keyed 'L'. Wrightsman (1964) found that the test-retest reliabil-ity coefficient (using a one hundred and sixty-six day interval) for the total BW to be .799 for a group of seventy-five college students. Aiken (1967) reported that after Welsh's 1947 sort of several hundred line drawings (N=143), a factor analysis of 6 the preferences revealed two major orthogonal factors. Factori was a tendency to 'Like' or 'Dislike' (response mode). Factor II was a tendency to prefer simple, symmet-rical figures as opposed to complex asymmetrical figures. Barron (1953a) found that artists liked highly complex, asymmetrical freehand figures that were "restless and moving" in their general effect as opposed to simple, ruled figures. Since the BW was scored so that preferences such as those of the artists, that is, preferences for the complex, received a high score, Barron called the BW a measure of the construct complexity. The tacit assumption was that the construct is unidimensional in nature, Evsenck (1970) questioned this assumed unidimension-ality. He stated that the literature "•••does not suggest that any attempt has been made to investigate the unidimen-sional nature of the test; it seems possible that here, as in the case of the Maitland Graves Design Judgement Test (Eysenck, 1963), the assumed unidimensionality may not in fact be found" (p. 523). Eysenck presented the eighty-six drawings of the BW on slides to small groups of students (N=lll). Product-moment correlations between items were calculated and the resulting matrix was analysed by the method of principal components. Twenty factors were extracted and rotated by Promax into oblique simple structure. I 7 Four independent higher order factors were extracted from the intercorrelations of the primary factors. • Eysenck gave the four factors the following interpret-ations: items that had loadings on Factor I were simple geometrical drawings such as circles, squares, zig-zag lines and the like. He therefore labelled Factor I as one of simplicity. Eysenck interpreted factors II, III , and IV as three independent complexity factors" (?. 3Zb) . Items that had loadings on Factor II were combinations of two simple figures, or simple figures that have been embellished, for example, a circle with a cursive line in it. Items that ' h a d loadings on Factor H I were complex, non-representational free hand drawings. Factor IV was characterized by items that were complex, representational free hand drawings. Eysenck found the interpretation for this factor least satisfactory since two items fit the interpretation doubt-fully, and one did not fit at all. On the basis of these findings, Eysenck concluded that there is more than one type of complexity, and that the scoring of the BW is faulty. Instead of having one single score, there should be four such scores, corresponding to the four factors extracted from the item correlation matrix. 8 Selected Studies in which the BW was Used Barron interpreted the BW as being a measure of complex-ity related to artistic taste. Many of the studies using the test have been concerned with the construct of creativ-ity. Aiken (1967) stated that in an unpublished report Mackinnon noted that the BW differentiated between 'creative1 and 'non-creative' artists, writers, architects, and research scientists. Baird (Buros 1972) concluded that as intriguing as these studies are, they are generally based on small samples and have not been cross-validated. The results of studies relating the BW to criteria of creativity are mixed. Rosen (1955) found a correlation' of .40 between the BW and ratings of art students' originality, and a correlation of .34 with course grades; Getzels and Csikszentmihali (1964) found that there was no correlation with ratings of art students' creativity and the BW, nor a correlation with art grades in art school and the BW. Helson (1966) found that the BW did not identify students nominated as creative by faculty, but that it did discriminate in her similar 1968 study. Schaefer (1969) found that it discrim-inated 'creative' from noncreative' boys in art, but did not discriminate in other creative areas, nor did it discrim-inate in any creative area for girls' groups. The BW predicted neither academic nor artistic achievement in a 9 school of design, in a study conducted by Skager, Klein, and Schultz (1967). In each of the last five studies cited, other measures of creativity were better than the BW. There does appear to be a correlation between the BW and aesthetic preferences. Barron (1952) found that low scorers on the BW tended to dislike modern, primitive, and sensual paintings; whereas for high scorers it was the reverse. Child (1962) found a correlation of .45 between BW scores and students' preferences for paintings scored according to a scale of aesthetic values established from a number of experts' judgements. He noted that the results from several studies have suggested that the BW scores are higher for people of higher social class. Child surmised that preference for asymmetrical designs may be mediated by education and social class, and possibly by opportunities to develop aesthetic sensitivity. Anderson (1965) pointed out that the BW yields a score on aesthetic judgement, which has been accepted by many as one component of artistic ability, but that there are better tests of aesthetic ability already available. Several researchers have reviewed studies in which the BW has been used. The following are their conclusions; Baird (1972) a^eed with Aiken (1967) that the correlations have not been large, the samples have generally been small, 10 and the results have seldom been validated. While the results are interesting and provocative, it is still not clear what aspect or correlate of creativity is being measured. Baird concluded that more research about the meaning and uses of the BW seems to be needed. Helmstadter's (1972) comments concerning the BW were less kind. He stated that one of the difficulties faced by the person who chooses to buy the BW will be that of dis-covering exactly what the scale measures, He concluded that lack of more extensive validity studies, of evidence of score consistency over long periods of time, and of a finished manual indicate that the BW should be considered promising set of stimuli which can be used in a variety of research situations involving art students and artists than as a refined instrument. As was mentioned earlier, Welsh, in his 1949 study of several hundred figures, found two major factors that were orthogonal. Factor 1 was an index of response set. Welsh (1959) pointed out that if a subject indicated 'don't like' for the majority of the items, he would get a high score on the BW, similar to the mean score of the artists. To avoid this difficulty the Revised Art Scale (RA) was developed. This scale consists of sixty items, thirty of which were scored 'like' and thirty of 11 which were scored 'don't like' by high scorers on the BW. In constructing this scale, high and low groups of scorers who also had average 'don't like' scores were used. Twelve subjects were selected for each of these two groups. The entire pool of subjects was two hundred and fifty. Helmstadter (1972) reported that in a study of one hundred psychiatric patients, the correlation between the RA and the BW was .85. Aiken (1967) stated that there is no correlation between total scores on the RA and 'Don't Like' scores on the BW. "Selected Studies in which the RA was Used Aiken (1967) reported that Raychaudhuri (1966a) found the RA distinguished among sixty painting artists, sixty musical artists and fifty non-artists. The respective means for these groups were 40.83, 29.00, and 19.00. Other studies (Raychaudhuri, 1961, 1962, 1963, 1966c) corroborated these results. Colman (1966) compared the RA with two measures of creativity; Hednick's Remote Associates Test and a Similarities Test. The correlation for the former with the RA was .52 (N=13 high scorers on the RA). The correlation for the latter with the RA was .53 (N=ll low scorers on the RA). 12 The BW and RA are the better known measures of prefer-ence for complexity- Two lesser known measures are the Kelly's Repertory Test (Rep Test) and the Bieri-Blacker scoring modification of the Rorschach. The Development of the Rep Test The Rep Test was designed to be used in a clinical or preclinical setting. The test is designed to measure role constructs. Kelly (1955) stated that the test is concerned with the subject's relationships to people who affected his life. Subjects are given a series of role titles, either verbally or in written form to which they must respond by naming a dif-ferent person of their acquaintance who fits each of the role titles. There are twenty-four such titles. Subjects are then asked to look at the names that they supplied in groups of three names, and to state how two of the people are similar and how they are different from the third. These responses are role constructs. Bieri and Blacker (1956) used a shortened version of the Rep Test (six role titles) to measure the complexity of an individual's perception of people. Responses were elicited in the same fashion as in the original Rep Test. Subjects responded to all possible combinations of three names such that no three names appeared together more than once. There -were twenty such combinations. The test was 13 scored by giving one point for each different role construct. The highest possible complexity rating was tw.enty, the lowest possible was one. The Development of the Bieri-Blacker As a second measure of complexity Bieri and Blacker (1936) developed a scoring modification of . the Rorschach. . Tb.e test consisted of ten large details (D) one D selected from each of ten cards. All other portions of the inkblot were blocked out. The inkblot details were chosen because of their judged ability to evoke a variety of responses, both in terms of content and in terms of determinants. Standard Rorschach instruc-' tions were given, except that subjects were asked to look at the whole inkblot; the blot was removed after subjects gave one response; and if subjects began to rotate the inkblot they were asked to keep it in the position in which it had been handed to them. Each subject went through the series of ten three times, each time being asked what else each inkblot could be. A total of thirty responses was obtained from each subject. Bieri and Blacker stated that they used the "Clopfter Scheme" for scoring the Rorschach records. Seven complexity measures were derived from the test. Four were determinant complexity scores measured empirically by counting the total number of different determinants used by the subject in making his responses; the total number of cards on which he repeated the same determinant at least two times; the total number of cards on which he repeated the same determinant three times; and the total number of cards on which the subject gave three different deter-minants. Two complexity measures were based on content complexity which was measured by counting the total number of responses which were repeated by a subject on the same card, and, the total number of cards on which a response was repeated. The seventh measure of complexity was the movement score, which was.calculated using the method traditionally employed in computing movement score on Rorschach records. Selected Studies in which the Rep Test and Rorschach were used Bieri and Blacker (1956) reasoned"... that the generality of cognitive complexity, operationally defined by response variability (p. 117)", could be measured using inkblots and people as stimuli. They administered both the Rorschach and the Rep Test to forty male undergraduates. Of the seven Rorschach measures of complexity, only the Movement score correlated significantly with the Rep Test (r=.31). The authors concluded that the results of this study suggest that some degree of generality in the 15 complexity of subjects* behavior can be demonstrated using two perceptual tasks involving personal and non-personal stimuli. Bieri and Blackerfs conclusion is somewhat optimistic since their data suggest very little generality in the complexity of subjects' behavior using the two perceptual tasks. Caracena and King (1962) ran a study similar to that of Bieri and Blacker. Three measures of complexity were used: the Rep Test, the Rorschach and the RA. The subjects were sixty undergraduates, thirty males, thirty females. The results showed no significant relations among the three measures of complexity. These results indicate that there are various types of complexity and that preference for one type is not related to preference for another type. Summary Portions of the Rep Test and of the Rorschach have been used as measures of preference for complexity. In the former, persons names'were used as stimuli; in the latter, inkblots were used as stimuli. Both tests share the same definition of cognitive complexity: response variability. There are few studies in which these measures of complexity have been used. 16 The RA and the BW, both portions of the WFPT, are the most .often used measures of complexity. In both, black and white free hand drawings are used as stimuli. Cognitive complexity is operationally defined for these two tests as preference for complex, asymmetrical drawings. The bulk of the research in which these two tests have been used is centered about the area of creativity. Because the RA and BW are widely used measures of preference for complexity and have been used in a great number of studies, they have been the focal point of this discussion. However, the RA and BW have several weaknesses. It is apparent that the assumption underlying the two tests ' is that preference for complexity is unidimensional. Studies by uysenck (1970) and Caracena and King (1962) cast doubt on the assumed unidimensionality. The complex drawings in the RA and BW are of three different types: complex geometrical shapes, non-representational freehand drawings and representational freehand drawings. Preference for one type of complexity may not necessarily indicate preference for another type. It would be better if the complex drawings were all of one type, or if three scores were derived from the test, one representing each of the types of complex drawings. Secondly, apparently no attempt has been made to 17 control for association value of the drawings. Some of the drawings have very obvious associations which may make the subject respond positively or negatively, depending on his experience. This could influence his score on the test. A final comment concerns the method of responding to the BW. The S is asked to indicate which drawings he likes and which drawings he does not like. Response mode defi-nitely influences scores. If a S were to indicate 'Don't Like' for all of the drawings, he would receive a score close to the mean score of the artists. This objection cannot be raised concerning the RA. The RA is the result of an attempt to remove the effect of response mode. In 'this scale, thirty items are keyed 'Like' and thirty are keyed 'Don't Like.' An alternative for measuring preference for complexity is described in Chapter 3. This alternative is designed to provide: a. Homogeneity in type or complexity measured. The measure of complexity is restricted to geometric shapes. b. stimuli which do not vary substantially in association value. 18 Chapter 3 Procedures In this study, five groups of students were tested using three measures of preference for complexity. Samples of students were drawn from Architecture, Art, Education, Law, and Engineering. Two of the existing measures of preference for complexity, the Barron-Welsh Art Scale (BW) and the Revised Art Scale (RA) were used. A third measure of preference'for complexity, the Random Shapes Scale (RSS) was developed. The design of this study will be described in two sections of this chapter: a. Development of the Random Shapes Scale, and b. Data collection and manipulation. 19 Development of the Random Shapes Scale The Random Shapes Scale is based on the preference-for-complexity constructs discussed by Barron and Welsh (1952). The scale is a measure of preference for geometric shapes that are relatively complex vs.. geometric shapes that are relativdy simple. A forced choice paradigm is employed.. Vanderplas (1963) constructed 1,100 random shapes using Method I as described by Attneave and Arnoult (1956). The random shapes used in the present study were selected from the Vanderplas set of 1,100. x Method I for constructing random shapes is as follows: Successive pairs of numbers between one and one hundred were selected from a table of random numbers. Using a sheet of one hundred by one hundred graph paper, each pair of the ran-domly chosen numbers was plotted on the grid.-When a number of points had been plotted (the number to be chosen arbitrarily), a ruler was used to connect the most peripheral points to form a polygon having only convex angles. After the peripheral points had been .ioined there were usually some points left within the ploygon. If an inside point was very close to the perimeter, it was included as 1 20 part of the perimeter since otherwise the point could lat6r cause an indentation that would divide the figure into two parts. A table of random numbers was used to decide which of the remaining points were to be connected to which other points. Since the finished figure was not to have lines that crossed, the successive connecting lines were decided upon by a process of elimination of those points that would produce intersecting lines when more than one point was left to choose from. Association Value of the Random Shapes Vanderplas (1963) noted that "a major difficulty in research on perception involves the well-known, but little understood, fact that perception is influenced by past exper-iences (p. 501)." This is an important point to keep in mind when using forms as stimuli, since forms do have meanings in the sense that they might be associated with objects, events, or concepts. These associations may influence the perception of forms and the relative preferences among them. Vanderplas and Garvin (1959) constructed one hundred eighty random shapes using Method I described by Attneave (1957). Six groups of shapes, each group consisting of thirty polygons having the same number of points, were con-structed. Each polygon had four, six, eight, twelve, sixteen, or twenty-four points. The shapes were shown to fifty subjects 21 who verbally responded 'Yes' if a shape reminded them of some-thing, named whatever it reminded them of if they could, and they responded 'No' if it did not remind them of anything. The association value of each shape was the percentage of subjects making 'yes' or 'content' responses; that is, subjects who could name what the shape reminded them of. The range of percentages of association values was between twenty and sixty-two with the mean of thirty-eight. Vepderplas (1963) replicated this study making a few modi-fications. He generated eleven hundred shapes which included one hundred each of four, six, eight, ten, twelve, fourteen, sixteen, eighteen, twenty, twenty-two, and twenty-four sided shapes. The ninety-seven subjects were tested in groups of three to five people. The subjects were asked to write a word or short phrase describing the shape, or an object or an event of which the shape reminded'them. The association value was determined by adding the number of responses made to each shape. The mean association was 32.6 with a standard deviation of II*.0. Vanderplas' interpre-tation of the distribution was that since most of the shapes appear to have a low to medium association value, they were relatively devoid of meaning, therefore analogous to nonsense syllables. 22 Complexity or the Random Shapes Vanderplas (1963), after generating eleven hundred shapes, obtained fourteen geometric measures for each of them. They are given in Table 1. The fourteen geometric measures were included, with the association value and the information content of each shape, in a factor analysis. A principal components analysis with Varimax rotation was used to discover whether a relationship existed between association value and information content and the fourteen measures. From the results of this factor analysis vanderplas inferred that the associative character-istics of the sample of shapes are relatively independent of ;their physical structure as measured by the fourteen values. In order to identify the domains of variation in the sixteen measures, the correlation matrix was transformed to an unrotated factor matrix. This analysis yielded both the eigenvalues and the cumulative sum of eigenvalues after each extraction of a latent root. After inspecting the table of cumulative eigenvalues, Vanderplas found that the major portion of the variance was accounted for by six factors. He labelled the principal component complexity (eigenvalue = 6.69) This factor was most closely associated with dumber of points, Perimeter, .perimeter squared, Mean interior angle, and the Ration of perimeter Squared to area. 23 Table 11 Fourteen Measures Calculated for each of the geometric Shapes Measures 1. Number of points 2. Area 3. Perimeter 4. Perimeter squared 5. Mean side length 6. Variance of side length distribution 7. Third moment of side length distribution Fourth moment of side length distribution 9. Mean interior angle 10. Variance of distribution of interior angles 11. Third moment of distribution of interior angles 12. Fourth Moment of distribution of interior angles 13. Ratio of perimeter squared to area 14. Maximum extent Zk Vanderplas used a computer program to order the eleven hundred random shapes on complexity factor scores. Attneave (1957), in a study using geometric measures of one hundred eighty random shapes similar to those used by Vanderplas, found that complexity was related to the number of points which determine inflections on the perimeter of the shape. The larger the number of points, the greater the complexity of the figure. Attneave discovered that the objective ratings of the complexity of random shapes agreed with subjective ratings of their complexity. He obtained subjective ratings by giving a group of subjects the same one hundred eighty shapes as had been rated objectively. Kach subject had rank ordered the shapes from least complex to most complex. Materials The RSS was constructed in the following manner: Highly complex shapes were defined as those shapes having z-scores on Vanderplas' complexity factor between 2.75 and 1.25; shapes of medium complexity were defined as those shapes having z-scores between 0.25 and -0.25; simple shapes were defined as those shapes having z-scores between -1.75 and -2.75-25 Fifty-four random shapes were chosen from the eleven hundred used by Vanderplas. Of the fifty-four shapes, eighteen are highly complex, eighteen are of medium complexity and eighteen are simple. The association or meaning of a form can influence perception. If the shapes had high association values, that is, if they were very meaningful, subjects' choices might be influenced more by what a shape meant to them than by its complexity. In an effort to control for association, association value was kept low: only those shapes having association values below 31.00 were chosen. The simple random shapes tended to have higher association values than did those of medium and high complexity. In order to hold the association value equivalent for the three levels of complexity, 31.00 was arbitrarily chosen as an upper limit. This value assured that all selected shapes were below the mean of 32.6, which Vanderplas found for his sample of 1,100 shapes. It was desirable that subjects consider all of the shapes in the group when making their choices. To avoid the possibil-ity of response mode and to ensure attention to characteristics of shapes, the number of sides was restricted to ten, tv/elve, fourteen, and sixteen, thereby eliminating obvious visual differences among the shapes. Complexity z-scores for ten-sided figures ranged from -4-85 t o +0.5b, twelve-sided figures from to +3.45, fourteen-sided figures from -2.56 to +3.68, and sixteen-sided figures from +0.62 to +5*47, 26 The three levels of complexity were narrov/ly specified, therefore the number of shapes to be chosen from was small. Using only ten to sixteen-sided figures and keeping the association values below 31.00 further limited the item population to 75 shapes. Of the 75, 20 snapes were simple, approximately 30 shapes were of medium complexity, and 25 shapes were of high complexity. From each of these 3 groups, 18 shapes were randomly sampled. The fifty-four shapes were graphed on a 100-by-l00 grid, cut out and pasted onto 20 x 7ir inch blabk strips of paper. Three shapes (one representing each of the three levels of complexity) were pasted onto each black strip. Three black strips were pasted on each of six 2.1& x 31i inch sheets of white paper. These were photographically reduced and reproduced in black and white and stapled together to form a six page booklet. The positions of the three shapes representing each level of complexity were varied so as to form a Latin Square on each of the six pages of the booklet. Table 2 shows the scheme for counterbalancing complexity. Table 2 Scheme for Counterbalancing Complexity-Page Number Complexity Association Item of Sides Order Value Label (A) IB) (c) (A) (B) (C) (A) (B) (0) 1 16 10 14 H L M 21.65 23.71 17.53 1 12 14 12 M H L 16.49 6.19 22.63 2 10 14 14 L M H 30.93 13.40 14.43 3 2 14 12 14 M L H 21.65 20.62 28.87 4 16 14 10 H M L 20.62 13.40 14.43 5 12 16 14 L H M 15.46 18.56 16.49 6 3 10 14 14 L M H 16.49 24.74 15.46 7 12 14 10 M H L 20.62 23.71 29.90 8 16 12 12 H L M 20.62 22.68 13.40 9 4 14 14 10 H M L 15.46 22.63 24.74 10 10 10 14 M L H 15.46 23.71 20.62 11 10 14 10 L H M 15.46 2k.65 23.37 12 5 10 16 12 L H M 19.59 19.59 26.30 13 12 10 14 M L H 3.25 27.34 20.62 14 16 14 10 H M L 16.43 22.63 22.63 15 6 14 16 10 M H L 22.63 21.65 20.62 16 16 10 10 H L M 10.31 15.46 25.77 17 10 14 14 L M H 23.37 17.53 22.63 13 Within Item: H - high complexity M - medium complexity L - low complexity Letters appeared on shapes as indicated by the instructions for the scale. The instructions for the RSS were: RANDOM SHAPES SCALE This task has been designed to measure prefer-ences for various geometric shapes. In it you will identify which of several shapes you like best, and which you like least. On the following pages there are eighteen groups of three shapes. Each group of three occupies one row on the page and is numbered. An example isi Each of the three shapes is lettered a, b, or c, from left to right. You should record which of the three shapes you like most and which you like least. Thus, if in the above example you liked shape (c) best and shape (b) least, your answer would look like this: LIKE MOST LIKE LEAST 30 Please provide an answer for each of the eighteen groups. We have found that initial impressions are much the same as those following prolonged debate. 1'herefore, please record your initial impressions. Turn to the answer sheet and wait for the signal to begin. Scoring the RSS was based on Spearman's rank order correlation coefficient: R = 1 - 6 d 2 N(N2-1) To obtain the score for a person on an item, his rank order of preferences for the three shapes was correlated with the a priori ranking of geometric complexities of the item's three shapes. Given any group of three shapes, only six preferential orders are possible; these six orders produce four different correlation coefficients: +1.0, +0.5, -0.5, and -1.0. Scoring and statistical manipulations were facilitated by eliminating fractional and negative values. The rank correlation scores on RSS items were linear transformations of these coefficients, obtained by first multiplying by two, then adding two. The item scores possible were then 3, 1, and 0. Since no justification could be found to argue that there was twice the difference in complexity 31 between raw scores of +0.5 and -0.5 as between raw scores of +1.0 and +0.5 or -0.5 and -1.0, an ordinal scale was postulated. Therefore, a factor or one was added to the two lower scores, thus making possible a score of Zf, 3, 2, and 1 for each item. The range of scores possible for the full eighteen item scale is therefore 18.to 72. The item scoring procedure is summarized in Table 3. The test booklet containing both the BW and the RA. was used in the original format: an eleven-page booklet having eight items on each of the first ten pages, and six items on the last page. Both the BW and the HA were scored according to the keys accompanying the Welsh Figure Preference Test Manual. Data Collection and Manipulation The collection, coding and transformation of the data in this study occurred between the months of January and April, 1973. Subjects Two groups of graduate students were tested: students in first year Law, and students in first, second and third year Architecture. Three groups of senior undergraduate students were tested: students majoring in Electrical 32 Table 3 Summary of Item Scoring Procedure for the Random Shapes Scale Complexity Possible Preference Rankings Ranking (a priori) 1 1 1 2 2 3 3 2 2 3 1 3 1 2 3 3 2 3 1 2 1 Rho 1.0 0.5 0.5 -0.5 -0.5 - 1.0 Preference Score 4 33 Engineering, Art, ana Education. All groups except the Art students were drawn from the University of British Columbia. The art students were drawn from the Vancouver Institute of Art. In the initial recruitment of subjects, appropriate faculty members were contacted. The purpose of the study was explained to them and their cooperation was solicited. The faculty members then asked their students whether they would take part in a study in 'perception.' Intact classes were tested. Students were told that participation was totally voluntary; however, very few — eight Architecture and seven Art students — refused to complete the scales. Class size varied from eight to fifty-two student®. Groups were chosen according to the following rationale: According to studies in which the BW and RA were used (see Chapter 2) artists always ranked highest in preference for complexity, followed by architects who did not rank as high as the artists, but usually higher than 'people in general.' Mathematicians generally ranked lower than 'people in general.1 A technically oriented group, electrical engineers, was included to see whether there was a difference among people in technical and other professions. It was of interest to know whether these groups could be differentiated, and ranked in a similar order on the RSS as they had been on the BW and 34 RA. Since architects, artists and mathematicians are known to differ from one another in their preference for complexity, the question of whether lawyers and teachers would differ from one another and from the other groups was raised. To answer this question, students in Law and Education were also tested. Senior and graduate students were selected because the probability was higher that they would exhibit the character-istics of their respective professions than would students who were just beginning their studies. The testing was accomplished in two sessions per group. .During the first session every subject completed the RSS, (N=292). The interval during the retest varied from three to four and one—half weeks. During the retest session, each subject first completed the RSS, then the BW and RA. The number of subjects per group per session is shown in Table 4» In the first session each subject was given the RSS and an answer sheet on which 'name'-and 'faculty' were requested as well as responses to the RSS. Data for the first session were collected with a period of five weeks. 35 Group Architecture Art Education Law Engineering Total Table 4 Number of Subjects Tested by Occupation Group Session 1 BSS 65 53 51 73 50 292 Session 2 BSS and BW and RA 30 31 45 65 47 218 RSS completed twice 28 24 38 50 44 184 In the second session, each subject was given a precollated packet containing the RSS and answer sheet, the BW and RA booklet and two Optical Mark Reader Exam cards on which to record responses, and a dark pencil. Data Coding and Transformation The data from the RSS were coded onto standard eighty column coding sheets and then keypunched. Information was coded at the item level. A Fortran computer program, Program DAVID (Kaufman, I973) was prepared to find totals un the RSS for each subject ana to store them on magnetic tape. These totals were then keypunched onto IBM cards. A second computer program, program OMR: MARKER (Coulthard and Herring, 1973) was prepared to read sets of scales and sets of responses, from Optical Mark Reader Exam cards, mark each response in a given set against all of the scales for that set, and print and punch the totals of correct responses for each respondent, under each scale. The responses to the eighty-six item Barron-Welsh Art Scale Booklet were totalled for each respondent under six scales: the RA, the BW and each of the four factors found by Eysenck (I967). (See discussion in Chapter 2.) The item numbers for each of the six scales are listed in Table 37 Table 5 Item Numbers for each of Six Scales of the Barron-Welsh Art Scale Booklet Scale RA BW Eysenck's Factors FI FII Fill FIV Like Dislike Like Dislike Like Like Like Like 4 2 6 1 3 1 12 21 5 3 9 2 7 3 13 24 6 7 12 3 16 35 31 43 9 11 19 7 13 37 36 60 12 16 25 3 22 43 50 66 13 IS 26 10 32 54 54 67 14 22 23 11 33 57 57 63 15 23 30 16 41 53 63 69 21 27 31 17 42 64 73 32 24 37 36 20 45 65 30 34 25 33 44 22 47 66 32 23 39 46 23 43 67 30 40 49 29 51 36 31 41 50 32 . 54 36 42 53 33 61 43 43 63 34 63 44 51 73 35 71 46 52 77 37 72 49 55 78 39 75 50 56 79 41 81 60 59 82 42 68 61 S3 45 69 62 84 47 70 63 85 51 76 64 52 77 65 54 80 66 55 S3 71 56 84 75 57 *5 81 58 62 65 66 67 71 72 74 81 Eysenck had examined the items that had high loadings each factor. In his paper he gave the item numbers. These items were used to develop the four keys corresponding to each of the Eysenck factors. I 40 Chapter 4 Results Results of data analyses are organized into four sections in this chapter: a. Analyses of the reliabilities of the Randon Shapes Scale, including internal consistency and stability over time. b. Item analysis. c. Correlational analyses including the correlation between the Random Shapes Scale, the Revised Art Scale, the Barron-Welsh Art Scale, and Eysenck's four factors. d. Analysis of variance, to discover whether there are significant differences among groups. Reliabilities of the Random Shapes Scale A fortran program at the University of British Columbia computing centre, program PIA, was used to calculate the means, standard deviations, and internal consistencies of the Random Shapes Scale (RSS), for each of the five groups: architecture, art, education, law, and engineering; as well as for the total of all groups. The results of these analyses are presented in Table 6. The internal consistency of the RSS varies over groups, with extremes of 0.61 for both education and law and 0.72 for architecture. The internal consistency for the scale \ 41 Table 6 Group Means. Standard Deviations and Reliabilities of the first Session of the RSS Group Mean Standard Deviation Cronbach 's N Alpha Architecture 43.20 Art Education Law Total (All Groups Com-bined) 44.53 46.60 44.55 Engineering 45.04 7.10 6.77 6.09 6.32 6.22 44.76 6.73 0.72 0.67 0.61 0.67 0.61 0.66 65 53 51 73 50 292 42 for combined groups is 0.66. Cronbach's alpha, a conservative estimate (Nunnally, 1967) was used to determine the internal consistency. A University of British Columbia computer program, program STRIP, was used to calculate the stability (test-retest correlation) of the RSS. Stability, like internal consistency, was calculated for each of the five groups as well as for all groups combined. The results of the stability analyses are presented in Table The stability coefficient varies over groups, as does the internal consistency. However, the stability coefficients are in every case higher. The extremes are 0.65 for both education and architecture, 0.79 for art. The correlation coefficient for combined groups is 0.71. Item Analysis A program from the University of British Columbia computing centre, program PIA, was used to calculate item correlations with total scores on the RSS. These correlations were calculated for each group; architecture, art, education, law, and engineering. The correlations for the groups are presented in Appendix C. The results of the analysis for all groups combined are presented in Table 8. 43 Table 7 Means, Standard Deviations, and Test Retest Stability Coefficients for each Group Group Means Test Retest Architecture 42.43 39.70 Art 45.42 43-92 Education 46.35 46.70 Law 44.06 43.10 Engineer 45.32 44.16 Total Groups 44.#4 43.26 Standard Deviation Test Retest 6.09 6.36 5.36 7.03 6.64 6.64 7.25 3.32 3.65 3.30 7.96 3.20 Stability " N Coefficient 0.65 0.79 0.65 0.63 0.75 0.71 30 31 45 65 47 218 Table S Total Groups Item Means, Standard Deviations and Correlations"'"with Total Scores (N=348) Item Mean Standard R(Total] Number Deviation 1 2.09 1.088 0.4542 2 2.80 0.966 0.4002 3 2.13 1.070 0.5570 4 2.37 0.973 0.2955 5 2.87 0.959 0.4672 6 1.91 0.950 0.2228 7 2.57 1.027 0.4335 8 2.23 0*968 0.3627 9 2.25 0.965 0.1989 10 2.90 0.916 0.3274 11 2.28 0.761 0.2128 12 3.00 1.048 0.4676 13 2.64 0.879 0.4813 14 2.89 1.012 0.3530 15 2.28 0.768 0 . 2996 16 2.77 1.029 0.4697 17 2.23 0.955 0.3804 18 2.57 1.060 0.4836 1A11 R(T) significant at p^.01 Items 1 through 18 are significantly correlated with total scores on the RSS in all groups except law and engineering. Items 6, 9, and 11 are not significantly correlated with total scores In the law group. Items if, 6, 9> 10> 11» am 15 are not significantly correlated with total scores in the engineering group. Since most items in the other groups are highly correlated with total scores (p^.-Olj, it is safe to conclude that the items in the RSS are homogeneous. Program PIA was used to calculate the percentage of subjects scoring 1, 2, 3» or 4 on each item for each group. The results of these analyses are presented in Appendix De Choice distributions for each item over all groups combined are presented in Table 9. Examination of the choice distributions shows that items 1, 2, 5, 6, 8, 11, and 17 did not discriminate as well among groups as did the remaining items. Table 10 shows how each group was split on those items that were the best discriminators» Correlational Analyses A University of British Columbia computer program, program STRIP, was used to calculate the correlations among the RSS, the RA, the BW, and estimates of Eysenck's factors, 46 Table 9 Total Group Choice Distributions Calculated as Percentages Item Number Item Score 1 2 3 4 1 39 28 17 16 2 12 23 39 26 3 36 31- 17 16 4 19 41 24 16 5 12 18 42 28 6 42 33 17 8 7 17 32 27 24 § 26 37 25 12 9 25 36 27 12 10 9 21 42 28 11 53 28 6 12 13 17 28 42 13 11 31 42 16 14 13 19 35 33 15 15 46 34 4 16 15 23 33 29 17 26 37 26 11 18 20 26 30 24 47 Table 10 Items in the RSS that Discriminated Among Groups Item Group Division Number 3 Architecture, law, and Engineering vs Art and Education 7 Architecture vs Art and Law vs Engineering vs Education 9 Architecture and Law vs Art vs Education vs Engineering 10 Architecture, Education, Law, and Engineering vs Art 12 Architecture, Education, and Law vs Art and Engineering 14 Architecture vs Art, Education, Law and Engineering 15 Architecture, Art, Law, and Engineering vs Education 16 Architecture vs Art vs Education vs Law vs Engineering 18 Architecture, Art, and Education vs Law vs Engineering FIf FII, Fill, and FIV. These correlations were calculated within each group —• architecture, art, education, law and engineering — as well as for the five groups combined. The results of the correlational analyses for individual groups are presented in Appendix E; the results of the correlational analysis over all groups combined are presented in Table 11. The RA and BW are highly correlated with one another in all groups. The extremes of the correlation coefficients are 0.80 for engineering and 0.96 for law. The correlation coefficient for all groups combined is 0.90. These coefficients are significantly different from zero (a6.01). In the architecture, art, and law groups, correlations between the RSS and the RA are significant. The correlation between the RSS and the BW is significant in the art and law groups. The three scales are not correlated in other groups, nor are they correlated when all groups are combined. Eysenck's Factor I, which is a factor of simplicity, is significantly correlated with the RSS for all groups combined and in the engineering group. Factor I is not correlated with the Rss in any of the other groups. FI is significantly negatively correlated with total scores 49 Table 11 Total Group: Correlations Among Measures of Preference for Complexity (N-213) Scale RSS RA BW . F1 F2 F3 F4 RSS 1.00 RA 0.05 1.00 BW 0.08 0.90* 1.00 F1 -0.32* -0.56* -0.57* 1.00* F2 -0.36* -0.26* 0.33* 0.72* 1.00 F3 -0.47* 0.19* 0.19* 0.52* 0.70* 1.00 F4 0.21* 0.34* 0.24* -0.29* -0.22* -0.37* 1.00 *P ^ .01 50 in the RA. and the BW in all groups combined, as well as all groups analysed separately, except one, engineering. In this group the correlation between FI and the BW is not significant. Eysenck's Factor II, complexity due to geometric shapes, has a significant negative correlation with the ESS in all groups combined, and in the architecture, education, and engineering groups. FEI is significantly negatively correlated with the RA and BW In all groups comoined, the architecture, and law groups. FII is also significantly negatively correlated with the BW in the art group. Eysenck1s Factor ill, Complexity due to non-representational freehand drawings, is significantly negatively correlated with the RSS in all groups combined, the engineering, and education groups. The correlation is significant and positive in the art group. Fill is also significantly posi-tively correlated with the RA and BW in all groups combined and the law and art groups. Eysenck's Factor IV, complexity due to representational drawings, is significantly positively correlated with the RSS in all groups except education. FIV is also significantly positively correlated with the RA and the BW in all groups except engineering and architecture, where it is significantly correlated with the RA but not with the BW. FI is significantly positively correlated with FII and FI.II in all groups except art and law in which it is correlated with FII hut not with Fill. FII is significantly positively correlated with Fill in all groups. It is significantly positively correlated with FIV in the architecture, art, and law groups, yet significantly negatively correlated with FIV in all groups combined, in education, and in engineering. Fill has the same pattern of significant correlations as FII. •Analysis of Variance A University of British Columbia adaptation of program BMD:o8V from UCLA BHD documentation was used to perform the analysis of variance. This program performs an analysis of variance using a mixed repeated measures design. The art group, (N=24) determined the number of subjects per cell for the other groups, from which subjects were randomly eliminated. The sources of variance in the mixed repeated measures design are differences due to group membership (G), testing seesion (.T), items (J), the interactions of group by time; group by item; time by item; and group by time by Item. The results of the analysis of variance are presented in Table 12. Table 12 Analysis of Variance Source df ' MS F Groups (G) k 13.01 6.87**-Times (T) 1 88.69 100.58** Items (I) 17 3.1.7 3.21** P/G H15 1.90 M? k 2.19 2.if8* MI 68 2.84 2.87** TI 17 3.26 3.79** TP/M 115 0.88 IP/M 1955 0.99 MTI 68 2.55 2.97** TIP/M 1955 0.86 *pi=.05 53 There is a significant difference (p<~,Ol) between the five groups, the items, the interaction between group and item, time and item, and group, time, and item, on the item scores in the RSS, The interaction between group and time is significant (p^-c05). Bonferonni contrasts were performed comparing law to architecture, engineering, art, and education; law and architecture to engineering, art, and education; law, architecture, and engineering to art and education; law, architecture, engineering, and art to education. The groups compared were ranked from lowest to highest according to means,. The results of the Bonferonni contrasts are shown in Table 13» All contrasts are significant (p^.05). There is a significant difference among students in architecture, art, education, law, and engineering. The interaction between group and time is significant. However, all of the group means were systematically lower in the retest than in the test. The group means for the test and the retest are shown in Figure 1, Table 13 i Summary of uroup Contrasts J.aw vs. Architecture, Engineering, Art, and Education = ±.447£ 0.093 Law and Architecture vs. Engineering, Art, and Education = 2.127 £ 0.453 .Law, Architecture, and Engineering vs. Art and Education ^ = 1.997< V}//, 0.323 Law, Architecture, Engineering, and Art vs. education (fy = 1.357 £ 0.003 ^All contrasts significant (p_^.05) Means Law =2.23 Architecture = 2.28 Engineering =2.41 Art = 2.48 Education = 2.52 MEAN Score over Items 2.90 2.85 2.80 2.75 2.70 2.65 2.60 2.55 2.50 2.45 2.40 2.35 2.30 2.25 2.20 2.15 2.10 2.05 2.00 Test _ Retest / V / V I « f i » Architecture Art Education Law Engineering Figure .U Meajis Score over Items for Each Group on Test and Retest 56 Chapter 5 Summary, Conclusions and Recommendations Summaries of the findings presented in chapter 4 are given here. Conclusions concerning the Random Shapes Scale are discussed, and recommendations for future research are made. Summaries and Conclusions The internal consistency of the RSS calculated over all groups combined is .66. The test - retest correlation coefficient calculated over all groups combined is .71® Eysenck's four factors are all significantly correlated with one another (p<.Ol). According to Eysenck, his factors are orthogonal. However, since the factors are highly intercorrelated it would appear that they are not orthogonal; that there is a great deal of overlap. Eysenck reported the item numbers that loaded on each factor. However, since he did not give the exact loadings for each item, the scoring of the factors was approximate. The RSS was significantly correlated with the BW and the RA among the architects and significantly correlated with the RA in,the art and law groups. No correlation existed between the RSS and the two art scales in the engineering and education groups. It is interesting to 57 note that scores on the RA for these two groups were significantly negatively correlated with Eysenck's Factor ill, complexity due to non-representational free-hand drawings. The two sets of correlation patterns cited may indicate that students in engineering and education responded to a different aspect of the RSS than did students in architecture, art, and law. Since the RSS and the RA are significantly correlated in three of the groups tested, it can be concluded that the RSS does measure a type of complexity that the RA also measures. The fact that the RSS, the RA, and the BW are all significantly negatively correlated with Eysenck's Factor II, complexity due to varying geometric shapes, lends further support to this conclusion. Results of the analysis of variance showed significant differences among the five groups tested. It can be concluded from this analysis that the KSS does differentiate among groups on the dimension of preference for complexity. The item analyses performed showed that the majority of items were significantly positively correlated with total scores on the RSS. This indicates that the items are homogeneous. 58 Of the eighteen items in the RSS, ten discriminated among the five groups. Inspection of these ten effective items showed that seven items each had two random shapes with the same number of sides, and one shape that was different. For example, the number of sides of the shapes in item 3 was 10, Ik, and 14. Of the eight items that did not discriminate effectively, only four were like the item just described. Possibly choosing among three shapes was made easier when two shapes had the same number of sides. The Random Shapes Scale was designed to measure preference for complexity. The results of the statistical analyses lead to the conclusion that the RSS is a useful measure of a unitary dimension of preference for complexity. The means were lower on the retest for all groups than the means had been for the test; the variances were higher on the retest than they had been on the test. It would be informative to conduct a time series study to discover whether group means continued to drop with each testing session. It would be of interest to design a random shapes scale using the ten items found to be effective in this study, and to add eight more items all having two shapes out of each set of three with an equal number of sides. A limitation of the present study is that, although association value was controlled, the letters placed in each shape in order 59 to identify it modified it by means of adding detail which contributed to meaningfulness, for example eyes. This would change the association value of the shape, making it more meaningful. In a future study, the Identifying letters should be placed under each of the shapes rather than in it. 60 References Aiken, L. R., Jr. A Review of Research on the Welsh Figure Preference Test. Greensboro, N.C., Creativity Research Institute of the Richardson Foundation, Inc., June 1967. Anderson, G. V. The Sixth Mental Measurements Yearbook. 0. Buros (ed), Highland Park, New Jersey, McGraw-Hill, 1967. Attneave, F. Physical Determinants of the Judged Complexity of Shapes» Journal of Experimental Psychology. 1957,(53), pp. 221-227. Attneave, F», and Arnoult, M. D. The Quantitative Study of Shape and Pattern Perception* Psychologic^,JulletjLn, 1956, (53)* PP. 452-471* Baird, L* L* The Seventh Mental Measurements Yearbook. 0. Buros (ed), Highland Park, New Jersey, McGraw-Hill, 1972. Barron, F. Personality Style and perceptual Choice. Journal of Personality, 1952, (20), pp. 385-401. 61 Barron, F. Complexity-simplicity as a personality dimension* Journal of Abnormal Social Psychology, 1953a, (48), pp. 163-172* Barron, F- and Welsh, G. S. Artistic perception as a possible factor in personality style, Its measurement by a figure preference test. Journal of Psychology, 1952,(33)* pp.. 199-203. Bieri, J. and Blacker, E. The generality of cognitive complexity in the perception of people and inkblots. Journal of Abnormal Social Psychology. 1956,(53)> pp* 112-117-Caracena, P. F. and King, G. F. Generality of individual differences in complexity, Journal of Clinical Psychology, 1962, (30), pp. 588-600. Child, I. L. Personal preferences as an expression of aesthetic sensitivity. Journal of Personality, 1962, (30), PP. 496-512. Coulthard, W. J. and Herring W. QMS:H&&KEB: A computer program prepared to read sets of scales and sets of responses, mark each response in a given set against all scales for that set, and print, and punch totals. University of British Columbia, 1973* 62 Eysenck, H. J. and Castle, M. A Factor-Analytic Study of the Barron-Welsh Art Scale. Psychological Records, 1970, 20, (4), PP. 523-526. Getzels, J.. W. and Csikszentmihalyi, M. Creative Thinking in Art Students: An exploratory Study. An unpublished report through the U.S. Office of Education, Cooperative Research Project No. E-008, University of Chicago, 1964, PP- vii-202. (Eric Ed 003 377) Halm, J. BMP:08V: Analysis of Variance, University of British Columbia, 1970. Helmstadter, G.C. The Seventh Mental Measurements Yearbook. 0. Buros (ed), Highland Park, New Jersey, McGraw-Hill. Helson, R- Effects of sibling characteristics and parental values on creative interest and achievement. Journal^of Personality, 1968, 36, (4), PP. 589-607. Helson, R. Personality of women with imaginative and artistic interests: The role of masculinity, originality, and other characteristics in their creativity. Journal of Personality, 1966, (34), PP. 1-25. 63 Kaufman, D. DAVID: A computer program to find totals on the RSS for each subject. Unpublished Manual, University of British Columbia, 1973-Kelly, G. A. Psychology of Personal Constructs. W. W. Norton, and Company Inc., 1955» PP- 219-267. Le„ C. STRIP: Small triangular regression package, University of British Columbia, 1971. Nunnally, J. C. Psychometric Theory, New York, McGraw-Hill, 1967* Raychaudhuri, M. Perceptual preference pattern and creativity. Indian Journal of Applied Psychology, 1966a (3)r PP. 67-70. Bosen, J. C. The Barron-Welsh Art Scale as a predictor of originality and level of ability among artists. Journal of Applied Psychology, 1955, (39), PP« 366-367. Schaefer, C. E. The prediction of creative achievement from a biographical inventory. Educational and Psychological Measurement,, 1969, (.29), (2), PP. 431-437. 64 Skager, R. W.; Klein, S. P.; and Schultz, C. B. The prediction of academic and artistic achievement at a school of design. Journal of Education Measurement. 1967, (4), pp. 105-117. Vanderplas, J. M. and Garvin, E. A. The association value of random shapes. Journal of Experimental Psychology, 1959, 57, (3), PP. 147-155* Vanderplas, J„ M. Associative processes and task relations in perceptual learning. Journal of Perceptual and Motor Skills, 1963, 16, pp. 501-509. Vanderplas, J* M. Final Report: Physical and Linguistic Factors in Form Perception. NIMH Grant No. M-4237, 1963, pp. 8-16* Vanderplas, J. M.; Sanderson, W. A.; and Vanderplas, J. N. Statistical and -associational characteristics of 1100 random shapes.. Journal of Perceptual and Motor Skills. 1965, (21), pp. 414. Welsh, G. S. Preliminary Manual Welsh Figure Preference Test (Research ed.). Palo Alto: Consulting Psychologists Press, 1959* Appendices Factor loadings of sixteen variables Figure preference scale Item means, standard deviations, and correlations with total score Choice distributions calculated as percentages Correlations a ong measures of preference for complexity Appendix A Factor Loadings of Sixteen Variables 1100 Random Shapes - varimax Rotation Variable 1 2 J l . 6 Number of points 819 -098 0 3 1 -072 - 3 5 9 3 7 3 Area 280 - 0 0 2 0 9 0 -147 - 0 5 4 908 Perimeter 879 0 4 9 1 5 5 -076 - 1 0 8 409 Perimeter squared 9 0 2 0 6 6 0 7 5 -060 -062 3 3 6 Side length mean - 2 9 7 3 6 1 111 0 6 2 816 -042 Side length variance 0 1 2 9 2 1 - 0 0 8 008 194 0 4 5 Side length skewness 1 3 5 618 0 4 5 0 3 6 - 6 5 5 0 6 2 £ide length kurtosis 0 2 6 9 5 4 - 0 3 1 0 2 7 003 0 3 4 Angle mean 660 -w 299 - 0 9 4 - 4 4 7 416 Angle variance 586 007 751 - 0 4 3 -140 0 7 5 Angle skewness -168 0 0 3 9 2 1 - 1 0 1 1 4 1 109 Angle kurtosis 457 008 868 -056 - 0 1 3 0 9 5 Perimeter squared/area 970 0 6 3 116 036 - 0 4 7 -088 Maximum extent 526 281 1 3 3 -027 -029 658 Association value 179 002 - 1 3 8 767 146 -234 Information content 049 -016 - 0 5 9 897 -072 041 67 Appendix B Figure Preference Scale 3 69 73 Appendix C Item Means, Standard Deviations and Correlations with Total Score 1. Architecture 2. Art 3. Education Law 5-. Engineering 75 Architecture: Item Means, Standard Deviations and Correlations with Total Scores (N=69) Item Number Mean Standard Deviation R( Total) 1 1.74 0.94-3 - 0.4084-** 2 2.90 0.887 0.4472** 3 2.06 0.991 0.6839** k 2.23 0.950 0.2271* 5 2.97 0.884 0.5286** 6 1.70 0.786 0.3356** 7 2.61 1.052 0.3887** 8 2.06 0.883 0.4881** 9 2.26 0.958 0.2734* 10 2.84 0.927 0.4211** 11 2.17 0.613 0.3648** 12 2.81 1.067 0.5003** 13 2.5'f 1.04-4 • 0.5953** 14 3.07 1.054 0.2788* 15 2.20 0.734- 0.3230** 16 2.38 1.131 0.4200** 17 2.20 O.S94 0.3769** 18 2. 4-6 1.030 0.4607** *P **p < .01 76 Art Item Means, Standard Deviations and Correlations v/ith Total Scores (N = 66) Item Number Mean Standard Deviation R(Total) 1 1.91 0.996 0.4273 2 2.33 0.946 0.3515 3 2.15 1.033 0.4594 4 2.52 0.925 0.4749 5 2.91 0.996 0.5491 6 2.09 1.033 0.3212 7 2.63 1.047 0.4613 3 2.21 1.003 0.3572 9 2.20 0.933 0.2457 10 2.39 0.935 0.3779 11 2.41 0.316 0.3032 12 2.35 1.104 0.2912 13 2.53 0.743 0.3217 14 2.36 0.919 0.2924 15 2.33 0.341 0.2962 16 2.65 1.007 0.5773 17 2.26 0.926 0.3990 13 2.74 1.034 0.4633 All R(T) significant (p^.01) 77 3. Education: Item, Means Standard Deviations and Correlations1 with Total Scores (N=67) Item Number Mean Standard Deviation R(Total) 1 2.39 1.196 0,6709 2 2.93 0.951 0.2061 3 2.33 1.138 0.6437 4 2.43 0.918 0.2048 5 2.96 0.999 0.4289 6 1.81 0.198 0.2744 7 2.60 1.008 0.3892 8 2.39 0.880 0.2074 9 2.15 1.082 0.3737 10 3.03 0.828 0.4375 11 2.07 0.606 0 . 2063 12 3.33 0.853 0.3932 13 2.75 0.720 0.4195 14 3.12 0.939 0.2591 15 2.30 0.773 0.4029 16 3.09 0.748 0.2210 17 2.21 0.955 0.3970 18 2.73 1.045 0.2738 All R(T) are significant (p^ .05) 78 4* Law: Item Means, Standard Deviations and Correlations with the Total Score (N=94) Item Mean Standard R(Total) Number Deviation 1 2.31 1.121 0.4120** 2 2.78 0.924 0.5833** 3 1.97 1.046 0.5199** 4 2.H 1.037 0.3419** 5 2,70 0.987 0.4273** 6 1.96 0.956 0.0526 7 2.43 1.026 0.4985** ) 8 2.27 1.023 O.3526** 9 2.20 0.858 0.0955 10 3.04 0.886 0.2760** 11 2.39 0.841 0.1439 12 3.00 1.101 0.5001** 13 2.59 0.880 0.4636** 14 2.73 1.043 0.4378** 15 2.21 0.742 O.2942** 16 2.82 1.051 0.4887** 17 2.18 0.967 0.3349** 18 2.5 A- 1.048 0.621** **p^.01 79 5. Engineer Item Means, Standard Deviations and Correlations with the Total score (N=52) Item Number Mean Standard Deviation R(Total) 1 2.00 0.981 0.2680* 2 2.42 1.080 0.3954** 3 2.21 1.115 0.4673** 4 2.19 0.962 0.1627 5 2.87 0.856 .Q.4491 ** 6 1.98 0.930 0.2262 7 2.62 0.964 0.4419** 8 2.23 0.993 0.3350** 9 2.54 0.909 0.0100 10 3.17 0.778 0.0732 11 2.29. 0.817 0.1758 1 12 3.02 0.971 0.6051** 13 2.85 0.928 0.5508** 14 2.65 0.998 0.5351** 15 2.38 0.738 0.1351 16 2.92 0.997 0.4904** 17 2.35 1.036 0.4666** 18 2.35 1.107 0.5257** * P ^ .05 **p< .01 80 Appendix D Choice Distributions Calculated as Percentages 1. Architecture 2. Art 3. Education Law 5. Engineering Architecture Choice Distributions Calculated a s Percentages Item Item Score Number 1 2 3 1 54 26 13 2 7 23 42 3 36 32 22 4 20 52 12 5 6 23 39 6 49 33 16 7 19 26 30 30 39 25 9 25 36 23 10 10 22 42 11 10 64 25 12 16 20 30 13 19 32 26 14 12 17 23 15 16 51 30 16 30 23 25 17 22 46 22 IS '22 29 30 82 Art Choice Distributions Calculated as Percentages Item Number Item Score 1 2 3 4 1 44 32 14 11 2 11 20 41 29 3 30 41 12 17 4 14 38 32 17 5 12 18 36 33 6 38 32 14 17 7 14 35 21 30 8 30 30 27 12 9 30 30 29 11 10 18 38 30 14 11 12 44 35 9 12 Id 15 30 36 13 39 45 8 14 11 18 45 26 15- 15 45 30 9 16 14 33 27 26 17 24 35 32 9 IS 15 24 32 29 I » Education Choice Distributions Calculated as Percentages Item Item Score ' t Number 1 2 3 4 1 33 22 13 27 2 9 22 36 33 3 30 31 15 24 4 16 37 33 13 5 15 7 45 33 6 30 16 6 7 16 30 31 22 a 15 43 30 12 9 36 30 13 16 10 6 15 49 30 11 15 63 22 0 12 3 16 25 55 13 4 23 55 12 14 10 7 42 40 15 IS 36 45 1 16 1 19 43 31 17 27 36 27 10 IS 13 13 37 27 Architecture Choice Distributions Calculated as Percentages Item Score 1 2 3 32 26 22 11 24 41 45 26 18 20 37 21 15 23 3$ 39 34 13 20 37 22 27 36 21 21 45 27 5 21 37 13 46 31 14 19 20 13 30 44 17 20 35 16 50 31 15 21 31 29 35 26 19 31 27 5. Engineer Choice Distributions Calculated as Percentages Item Item Score Number 1 2 3 4 1 37 3$ 13 12 2 27 23 31 19 3 35 . 29 17 19 4 27 38 23 12 5 10 15 54 21 6 37 37 19 8 7 13 33 33 21 8 27 37 23 13 9 13 35 37 15 10 6 6 54 35 11 15 48 29 3 12 12 12 40 37 13 10 23 40 27 14 13 33 29 25 15 10 48 37 6 16 12 19 35 35 17 25 33 25 17 18 29 29 21 21 Appendix E Correlations Among Measures of Preference for Complexity 1. Architecture 2. Art 3. Education Law 5, Engineering 87 1. Architecture Students: Correlations Among Measures of Preference for Complexity (N=30) *P ^ .05 .01 Scale SSS RA BW F1 F2 RSS 1.00 RA 0.38* 1.00 BW 0.28 0.93** 1.00 F1 0.29 -0.51** -0.55** 1.00 F2 0.34* -0.37* -0.49** 0.81** 1.00 F3 0.24 0.21 0.13 0.42** 0.52' F4 0.34* 0.31* 0.20 0.26 0.50' F3 F4 8J5 2. Art Students Correlations Among Measures of Preference for Complexity (N®31) Scale RSS RA BW F1 F2 F3 F4 RSS 1.00 RA 0.47** 1.00 BW 0.45** 0.95** 1.00 F1 0.24 -0.49** -0.51** 1.00 F2 0.24 -0.19 -0.32* 0.71** 1.00 F3 0.62** 0.54** 0.45** 0.20 0.39* 1.00 F4 0.30* 0.55** 0.44** 0.05 0.35* 0.62** 1.0< *P < .05 **p«i .01 89 3. Education Students: Correlations Among Measures of Preference for Complexity (11-45) Scale RSS RA BW F1 F2 F3 RSS 1.00 RA -0.01 1.00 BW 0.07 0.94** 1.00 F1 -0.19 -0.38** -0.48** 1.00 F2 -0.30* -0.03 -0.23 0.84** 1.00 F5 -0.31* 0.24 0.09 0.66** 0.88** 1.00 F4 0.19 0.35** 0.38* -0.42** -0.47** -0.49** *p < .05 **p ^ .01 I 90 Law Students Correlations Among Measures oiv Preference for Complexity (N=65) Scale RSS RA BW F1 F2 RSS 1.00 RA 0.36** 1.00 BW 0.40** 0.96** 1.00 F1 -0.14 -0.68** -0,70** 1.00 F2 -0.09 -0.40** -0.53** 0.78* 1.00 F3 0.12 0.45** 0.34** 0.11 0.30 F4 0.22* 0.59** 0.47** -0.01 0 © *P 5 .05 .01 F3 F4 0.77** 1.00 91 Engineering Students: Correlations Among Measures of Preference for Complexity (N=47) Scale " RSS RA BW F1 F2 F3 F4 RSS 1.00 RA 0.08 1.00 BW -0.01 0.80** 1.00 F1 -0.60** -0.29* -0.24 1.00 F2 -0.34** -0.08 0.06 0.62** 1.00 F3 -0.74** 0.23 0.40** 0.67** 0.69** 1.00 F4 0.56** 0.22 -0.06 -0.43** -0.35**-0.62** 1.00 *P <r .05 »*p<- .01
- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- Development and testing of a paired-comparisons figural...
Open Collections
UBC Theses and Dissertations
Featured Collection
UBC Theses and Dissertations
Development and testing of a paired-comparisons figural scale to measure preference for complexity Wichert, Shelley Gabriele 1973
pdf
Page Metadata
Item Metadata
Title | Development and testing of a paired-comparisons figural scale to measure preference for complexity |
Creator |
Wichert, Shelley Gabriele |
Publisher | University of British Columbia |
Date Issued | 1973 |
Description | The purpose of this study was to develop and to test a paired-comparisons figural scale to measure preference for complexity. A Random Shapes Scale (RSS) consisting of 18 sets of 3 random shapes was constructed. In each set of 3, one shape was of high complexity, one of medium complexity and one of low complexity. The random shapes were chosen from the eleven hundred generated by Vanderplas. Two existing measures of preference for complexity, the Barron-Welsh Art Scale (BW) and the Revised Art Scale (RA) were also used. Students in architecture, art, education, law and engineering (N=292) were tested using the RSS. Three weeks later the same groups of students (N=184) were retested on the RSS and completed the BW and RA as well. The BW and RA were significantly correlated with the RSS in three of the five groups tested. The internal consistency of the RSS calculated over all groups combined was .66; the stability coefficient was .71. The analysis of variance showed significant differences among the five groups tested. Therefore the RSS does differentiate among groups on the dimension of preference for complexity. The majority of the items were highly correlated with total test scores. This indicates that the items are homogenous. The results of the statistical analyses lead to the conclusion that the RSS is a useful measure of a unitary dimension of preference for complexity. |
Subject |
Paired-association learning Similarity (Psychology) |
Genre |
Thesis/Dissertation |
Type |
Text |
Language | eng |
Date Available | 2012-04-11 |
Provider | Vancouver : University of British Columbia Library |
Rights | For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use. |
IsShownAt | 10.14288/1.0107079 |
URI | http://hdl.handle.net/2429/41942 |
Degree |
Master of Arts - MA |
Program |
Psychology |
Affiliation |
Arts, Faculty of Psychology, Department of |
Degree Grantor | University of British Columbia |
Campus |
UBCV |
Scholarly Level | Graduate |
AggregatedSourceRepository | DSpace |
Download
- Media
- 831-UBC_1973_A8 W52.pdf [ 2.02MB ]
- Metadata
- JSON: 831-1.0107079.json
- JSON-LD: 831-1.0107079-ld.json
- RDF/XML (Pretty): 831-1.0107079-rdf.xml
- RDF/JSON: 831-1.0107079-rdf.json
- Turtle: 831-1.0107079-turtle.txt
- N-Triples: 831-1.0107079-rdf-ntriples.txt
- Original Record: 831-1.0107079-source.json
- Full Text
- 831-1.0107079-fulltext.txt
- Citation
- 831-1.0107079.ris
Full Text
Cite
Citation Scheme:
Usage Statistics
Share
Embed
Customize your widget with the following options, then copy and paste the code below into the HTML
of your page to embed this item in your website.
<div id="ubcOpenCollectionsWidgetDisplay">
<script id="ubcOpenCollectionsWidget"
src="{[{embed.src}]}"
data-item="{[{embed.item}]}"
data-collection="{[{embed.collection}]}"
data-metadata="{[{embed.showMetadata}]}"
data-width="{[{embed.width}]}"
async >
</script>
</div>
Our image viewer uses the IIIF 2.0 standard.
To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0107079/manifest