ABSTRACT INCONGRUITY: THE EFFECT OF SYSTEMATICALLY ALTERED DRAWINGS ON ADULTS' RATINGS OF INCONGRUITY, UNFAMILIARITY, AND RELATED DIMENSIONS By Richard Robert Knight Drive-reduction conceptions of motivation have failed to explain a variety of behaviors. Among these are orientation, exploration, manipulation, and play. The explanation of these behaviors, particularly exploration, has been the aim of a considerable body of research. There have been some attempts to determine physiological con- comitants of exploration, but the majority of research has examined the conditions of stimulation in which exploration is elicited. In general, these conditions are novelty, complexity, and incongruity. The least is known about incongruity. Various definitions and theore- tical treatments of this term have hypothesized the existence of an expectation, whose disconfirmation results in changes in activation or arousal. Because of the importance of previous experience implied by ex- pectations, incongruity may be a useful variable in developmental studies of exploration, play, fear, and cognitive development, by explaining phenomena in these areas in terms of arousal processes. Lack of any systematic attempt to define incongruity has limited its usefulness, however. The aim of this research was to clarify the Richard Robert Knight meaning of incongruity; to define it in terms of specific alterations of specific figures. Accordingly, ten drawings were made, five of living figures and five of nonliving figures. Each was altered in four ways: Missing part; Misplaced part; Wrong part; Misplaced and Wrong part. Each manipulation was performed on both the head and leg of the living figures, resulting in three classes of figures: Living, head man- ipulated; Living, leg manipulated; Nonliving. The resulting drawings were seventy in number, ten unaltered; sixty altered. College students (N=240) rated the seventy drawings on six di- mensions: Incongruity; Complexity; Interestingness; Pleasingness; Unfamiliarity; Conflictfulness. Each subject rated each drawing once on only one dimension. The ratings were highly reliable and were not influenced by several potential confounding factors. Analyses of variance of the ratings showed that 'Manipulation' and 'Class of figure' influenced the ratings on all six scales. Their interaction influenced ratings on five scales. The manipulation rated most incongruous was 'Misplaced and Wrong part'; the least incongruous was 'Missing part'. 'Living figures with the head manipulated' were rated most incongruous, and 'NOnliving' the least incongruous. The most incongruous combination of 'Manipulation' and 'Class' was 'Living figures with head Misplaced and Wrong'; the least was 'Nonliving figures with a part Missing'. It was also noted that manipulations performed on the head resulted in particularly high ratings of incongruity. Ratings on all scales except pleasingness correlated positively Richard Robert Knight with each other. Pleasingness ratings correlated negatively with all others. This indicates that incongruity shares meaning with complexity, interestingness, unfamiliarity, conflictfulness, and displeasingness. Some evidence was presented that incongruous figures are more than simply novel; that in addition to novelty they have a component that results in judgments of incongruity. That component is some combination of specific alterations made on specific kinds of figures. A differential effect of incongruity and unfamiliarity upon looking time was hypothesized. If incongruity and novelty are different dimensions of stimulation, then familiarization would result in a greater decrease in looking time to less incongruous figures than to more incongruous figures. Previous studies have used figures of only moderate incongruity. Inclusion of the broader range of incongruity in the presently-used figures would likely enhance the effect of incongruity on visual ex- ploration and physiological processes associated with arousal, and thus help to explicate the process by which these effects are produced. Suggestions for further research included the effect of incongruity on visual exploration, play, and the deve10pment of fear. INCONGRUITY: THE EFFECT OF SYSTEMATICALLY ALTERED DRAWINGS ON ADULTS' RATINGS OF INCONGRUITY, UNFAMILIARITY, AND RELATED DIMENSIONS By Richard Robert Knight A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1970 447926; ACKNOWLEDGMENTS I am pleased to have the opportunity to eXpress my appreciation to peOple who have been important in my development, though what I write here cannot adequately express it. Lauren Harris, chairman of the guidance committee, has been an unfailing ideal which I can only hope to approximate. Stanley Ratner is the person who first excited me about psychology and its unanswered questions. Martin Balaban first introduced me to the developmental vieWpoint and the experi- mental precision that it requires. John McKinney has tempered all of this with a decidedly human, social point of view. All can see some part of themselves in this thesis. Finally, David Wright has been a patient, efficient liaison with computer facilities, and I am grateful for his help. ii TABLE OF CONTENTS LIST OF TABLES INTRODUCTION DRIVE REDUCTION AND 'LUDIC' BEHAVIORS 'COLLATIVE' VARIABLES: THE RESEARCH OF BERLYNE COMPLEXITY The effect of complexity upon visual eXploration College students: Children: The effect of complexity upon ratings and "preferences" NOVELTY The effect of novelty upon visual exploration College students: Children: The effect of novelty upon ratings and "preferences” INCONGRUITY Development of fear in chimpanzees and children The effect of incongruity upon visual exploration College students: Children: The effect of incongruity upon ratings and ”preferences" Definitions and theoretical treatments of incongruity SUMMARY OF RESEARCH ON VISUAL EXPLORATION AND "PREFERENCES" iii page vi 14 14 14 19 23 27 29 29 3O 32 33 33 37 37 40 43 47 53 I I III. III. I II | I ll ll [.1 All. i'. iv THE CONCEPT OF PREFERENCE THE PROBLEM METHOD Subjects Stimulus figures Rating scales Procedure RESULTS LEVEL OF SIGNIFICANCE CONFOUNDING VARIABLES l (a). Consistency of ratings within each scale 1 (b). Ratings compared across scales 2. Serial position effect 3. Sex effect RATINGS OF UNALTERED FIGURES EFFECT OF ORDER OF PRESENTATION, MANIPULATION, AND CLASS OF FIGURE 1. Summary: The effects over the six scales 2. Effects within each scale (a) Incongruity (b) Complexity ' (c) Interestingness (d) Pleasingness (e) Unfamiliarity (f) Conflictfulness CORRELATIONS AMONG THE RATINGS COMPARISON OF INCONGRUITY AND UNFAMILIARITY RATINGS 54 58 6O 60 6O 65 66 72 72 72 73 73 75 77 78 78 78 8O 8O 83 87 89 89 93 95 95 DISCUSSION Consistency of the ratings Serial position Sex The unaltered figures Order of presentation THE MEANING OF "INCONGRUITY" Variables which affect incongruity ratings Relations among the six rating scales Summary FURTHER RESEARCH REFERENCES APPENDICES APPENDIX A. "Incongruous" figures used by other investigators APPENDIX B. A conceptual scheme for incongruity APPENDIX C. Stimulus figures APPENDIX D. Ratings of the altered figures 102 102 102 104 104 104 105 106 113 118 119 122 133 136 137 151 Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table LIST OF TABLES Subjects Analysis of variance over all scales . Summary of serial position effect analyses of variance . Block means for serial position effect Mean ratings of unaltered figures Summary of six separate analyses of variance Analysis of variance of incongruity ratings Mean ratings for each manipulation . Mean ratings for each class of figure Mean ratings for twelve categories of figures Analysis of variance of complexity ratings Analysis of variance of interestingness ratings Analysis of variance of pleasingness ratings Analysis of variance of unfamiliarity ratings Analysis of variance of conflictfulness ratings Intercorrelations of the six scales Analysis of variance of incongruity-unfamiliarity Mean ratings for Group X Class interaction Mean ratings, Group X Manipulation X Class Incongruity ratings Complexity ratings . Interestingness ratings vi page 61 74 76 76 79 81 82 84 84 85 86 88 90 91 94 96 98 100 100 151 152 153 vii Table D4. Pleasingness ratings 154 Table D5. Unfamiliarity ratings 155 Table D6. Conflictfulness ratings 156 INTRODUCTION The period from the late 19403 to the present has seen great changes in psychological conceptions of motivation. The single most important factor driving these changes has been the recognition that certain behaviors resist explanation by traditional motivational theories. These behaviors, called "ludic" behaviors (from the Latin ludus, meaning play or game), generally are classifiable as explora- tion and play. Their most damaging feature, as far as the drive- reduction theories of the 19303 and 403 were concerned, was that they served no obvious biological purpose and reduced no biological or learned drive. Animals played and explored most insistently precisely when biological drives had been satisfied. The drive-reduction theories' failure to explain these ludic behaviors was generic; that is, the theories, given their fundamental premises about the bases of behavior, were incapable of explaining how painful stimulation, changes in homeostatic conditions, or previously neutral stimuli associated with either of these, could account for the intensity and direction of ludic behavior. The following review of literature examines a sample of research on exploratory behavior which figured in the abandonment of the tra- ditional drive-reduction model. Exploratory behavior in human subjects has been an important part of this research, and has been dominated by Daniel E. Berlyne. This work has been largely concerned with the conditions of the stimulus situation in which visual exploration occurs. Another approach has been to obtain subjects' verbal eval- uations of stimuli, relating those evaluations to duration of visual exploration. Many of the studies in which evaluations have been obtained have been introduced as studies of "preference" for various stimulus conditions. The concept of "preference," however, requires careful examination. Research on stimulus correlates with visual exploration and "preference," initiated by Berlyne's early work and using many of his stimulus materials, has later been extended to children, though with few implications for developmental theory. The intent of this review is to demonstrate that, across a variety of species, across age and experience, and in many different experimental tasks, certain features of the environment are likely to elicit exploration, manipulation, and perhaps play. Such gener- ality of findings suggests that these characteristics of the environ- ment, interacting with ongoing processes in the individual, may be potentially powerful in explaining other kinds of behavior. These qualities of the environment have been variously named, but can generally be referred to as complexity, novelty, and incon- gruity. Of the three (actually many; see discussion of Berlyne's "collative" variables, which follows), the least is known about incongruity, and the term has been carelessly used. An attempt will be made to come to a better understanding of what is meant by ”incon- gruity," and how it is related to other variables. The review of literature is limited in several ways. First, the dominant interest of the present research is in humans, and, although it is acknowledged that this research area is indebted to the study of nonhuman species, the more recent study of nonhuman exploratory, mani- pulatory, and play behaviors will not be reviewed. Second, the em- phasis in the present research is on characteristics of external stimulation which elicit exploration and verbal evaluation. Studies of physiological correlates of these conditions and the responses which they evoke, are clearly of interest, but not of primary interest here. The present research is eclectic rather than particular in theoretical orientation. Third, and this is dictated by the available research using human subjects, only studies of visual exploration will be of present concern. DRIVE REDUCTION AND 'LUDIC' BEHAVIORS Since a number of excellent surveys of this literature, largely based on nonhuman subjects, are available (Dember and Earl, 1957; Glanzer, 1958; Cofer, 1959; White, 1959; Berlyne, 1960, Chs. 5&6; Dember, 1961; Welker, 1961; Hunt, 1963; Butler, 1965, and Lester, 1967), still another comprehensive review will not be undertaken here. Rather, studies will be mentioned to illustrate the problems raised for drive- reduction theories, and to provide a context for discussion of sub- sequent studies of visual exploration in humans. These studies are presented chronologically, but they belong to two conceptual groups: the first has to do with learning under conditions in which none of the "normal" reinforcers seemed to be Operating. Learning seemed to take place when the reward was simply the opportunity to explore or to man- ipulate. Second is a group of studies based on a logical extension of drive-reduction theory, that organisms sated in every possible way will have no drives to reduce and therefore will be quiescent. As Berlyne (1966 , p. 26) has said, this reasoning represents "a view that anybody who has had to handle a child 'with nothing to do' must have been tempted to question." For many years, two investigations stood almost alone in the psychological literature. In 1925, Dashiell reported that sated rats would explore extensively in the maze that bears his name. In 1930, Nissen reported that rats would cross a shock grid to reach an area which contained novel objects. There being no other possible rein- forcer for this behavior, Nissen posited an exploratory drive. Pavlov (1927) observed the "orienting reflex" in his experimental dogs, and commented on the place of that reaction to sudden or new stimuli in a motivational scheme (2. also Razran, 1961). In addition, observation of "unlearned" fears were made during this early period. K6hler (1927) and Valentine (1930) observed a variety of situations which aroused fear in chimpanzees and young children, and argued that there had been no Opportunity for these fears to be learned. Reduction of some drive-state, the explanation for learning, did not help to explain the deve10pment of these fears. During the late 19403 and early 19503 the study of exploratory behaviors was intensified. Harlow (1950) and Harlow, Harlow, and Meyer (1950) found that food- and water—satiated rhesus monkeys would disassemble complicated puzzles for no other apparent reward than to take the puzzle apart. Berlyne (1950) reported that rats approached and sniffed at novel objects more than familiar ones, and that this exploration decreased ‘with successive presentations but was reinstated by introduction of another novel object. Later (1955) he found similar results for complex versus simple stimuli, and (Berlyne and Slater, 1957) showed that rats choose a more complex or more spacious Y-maze arm. Montgomery (1952, 1953) began a series of experiments showing that the alternation behavior of rats in a T-maze is not based on response inhibition, but rather on choice of the unfamiliar place. Glanzer (1953) independently performed a similar experiment with the same result. Butler (1953) found that monkeys would learn a discrimination problem when the reward was the opportunity to look through a window at the laboratory area, and (1954), that the rate of lever pressing depended on what was available to look at, a changing scene resulting in a greater rate of pressing than an unchanging scene. Another aspect of stimulation, change or variation, has a powerful influence on exploratory behavior. The now classic studies on stimulus deprivation (Bexton, Heron, and Scott, 1954; Heron, Doane, and Scott, 1956) have clearly shown that low levels of stimulation are not tol- erated for long by well-paid human subjects, and that they will do virtually anything to gain stimulation, from pressing a button for stockdmarket reports to hallucinating auditory and visual experiences. The implication is clear. Not only present stimulation, but stimula- tion experienced and perhaps adapted to at some time in the past, are powerful determiners of present exploratory behavior. Finally, the major part of this review, concerning visual ex- ploration and evaluation of complexity, novelty, and incongruity, con- stitutes clear evidence that some human behaviors are motivated by factors other than drive-reduction alone. The above research shows that the drive-reduction theories were incomplete at best. This body of research between 1925 and the mid 19503 occurred in a changing Zeitgeist, but in the years 1949-1951 the cli- mate changed in significant ways. A3 White (1959) has suggested, drive-reduction theory had been in trouble for some time, and its problems appear to have come to the fore during these years. Donald Hebb's Organization of Behavior appeared in 1949, with his well-grounded call for a serious consideration of the behavioral and physiological manifestations of attention and its influence on perception and learning. Moruzzi and Magoun's work on the brainstem reticular formation also appeared in 1949, and dealt the reflex arc, and by implication the drive-reduction model, its final blow. Finally, during this same period of time Lorenz (1950) and Tinbergen (1951) were arguing for a biolog- ical, evolutionary analysis of unlearned behaviors, thus diverting attention from and creating alternatives to the notion that motivation and resultant behavior could be explained by learning alone. In an address given in 1951, Harlow (1953) concluded that the drive-reduction concept of motivation as it had been applied to learn- ing theory was grossly inadequate in all but the very simplest of cases. We do not mean to imply that drive state and drive-state reduction are unrelated to learning; we wish merely to em- phasize that they are relatively unimportant variables. Our primary quarrel with drive-reduction theory is that it tends to focus more and more attention on problems of less and less importance" (p. 27). Harlow's call for the disavowal of drive-reduction, together with his plea for the study of stimulus prOperites influencing motivation undoubtedly had a seminal influence on conceptions of motivation, when taken in the context of Hebb, Moruzzi and Magoun, and the ethologists. The final recognition of the existence of these pervasive 'ludic' behaviors coupled with drive-reduction theory's inadequate explanation of them has led a number of investigators to seek other explanations. These have centered around the "intrinsic" motivating properties of the behaviors themselves, of internal states such as conflict, arousal, or adaptation level, or of certain properties of external stimulation. Leuba (1955) suggested that an optimal level of stimulation exists, which the organism strives to maintain. The question of what is optimal, however, is still open. Hebb (1955) suggested an optimum level of arousal, and Leuba's formulation can be interpreted in those terms. McClelland, Atkinson, Clark, and Lowell (1953) proposed an Optimal discrepancy between the stimulus input of the moment and an adaptation level. Hunt (1963, 1965) similarly has posited an Optimum of discrepancy between the stimulus input and some internal standard based on genetic factors and previous experience. Berlyne (1960, 1963b) has posited an optimal level of "arousal potential,” the interaction between certain characteristics of stimuli and the momentary organismic state. Maddi (1961) has suggested an Optimum of variation in stimulation, and Munsinger and Kessen (1964, 1965) have suggested an Optimum of uncertainty. All of these interpretations involve an op- timum (adaptation level, standard, "normal" level) of something in- volving change. The only major way in which they differ is the label by which change is specified. There are few other differences among them, and a choice of one rather than another would be largely idio- syncratic. The events considered to have the potential to motivate organisms have been significantly broadened. However, it is unparsimonious to have one strong theory explaining consummatory behaviors and another set of theories to explain other behaviors. Berlyne (1960, 1963b) has attempted to integrate the theories, but more work, especially of a physiolo- gical nature, remains to be done. 'COLLATIVE' VARIABLES: THE RESEARCH OF BERLYNE Discussion of exploratory and evaluative reactions of human beings to various qualities of their environment must start with the work of D. E. Berlyne. All other investigators cited in this present review refer to Berlyne, and many of the stimulus materials used by them are based on Berlyne's stimuli, first used in 1957. In the context of the study of exploratory behavior just reviewed, Berlyne (1960) called for a reconsideration of the drive-reduction model's utility. He reports, however, that by the time he finished Conflict, Arousal, and Curiosity (1960), he was convinced that a mod- ified form of the drive-reduction model was the best means to explain both consummatory behaviors and hitherto unexplained exploratory behaviors. In order to retain the drive-reduction theory, Berlyne found it necessary to posit two further drives: a curiosity drive explaining diversive exploratory behaviors, that is, behaviors aimed at no SPGCifiC stimulus, and a boredom drive, elicited under conditions of low or unvarying stimulation, and reduced by the arousal generated by exploration, involving kinesthetic feedback and a changing stimulus field. The curiosity drive is more complex to analyze. Berlyne suggests that objects with "collative" prOperties (novelty, complexity, uncertainty, incongruity, ambiguity, surprisingness) elicit conflict, arousal, and curiosity. Berlyne calls these relational aspects of the stimulus situation "collative variables" because, in order to evaluate them, it is necessary to examine the similarities and differences, compatibilities and incompatibilities between elements-~between a present stimulus and stimmli that have been experienced previously (novelty and change), between an element of a pattern and other elements that accompany it (complexity), between simultaneously aroused responses (conflict), between stimuli and expectations (surprisingness), or between simul- taneously aroused expectations (uncertainty)" (1960, p. 44). Curiosity, a drive state, is accompanied by a general arousal and elicits exploration. Exploration reduces the arousal by reducing the curiosity (the object is somehow assimilated or recognized). Berlyne calls the increase and subsequent decrease in arousal to the optimal level the "arousal jag," and maintains that the exploratory behavior is reinforced by the decrease in curiosity and arousal. It is in this sense that reduction of the curiosity drive is reinforcing. Only in his most recent major paper (1967) has Berlyne acknowledged that the increase in curiosity and arousal is a necessary concomitant of this arousal jag, and the effects of the decrease and increase in arousal have not, as yet, been separated. Perhaps the reinforcement of exploratory behavior is drive induction, or induction followed by reduction, or simply arousal change, and not drive reduction alone. After all, the organism behaves in such a way as to put itself in new situations as was shown earlier; it is more parsimonious to assume that a change in arousal is reinforcing. Berlyne's treatment of exploratory behavior can be criticized on a number of other grounds. First, he has not attempted to quan- tify the collative variables, despite the fact that his theory is specific in predicting greater "attractiveness” of stimuli having greater amounts (up to some point) of the collative variables. Second, Cantor (1963) has pointed out that Berlyne predicts that stimuli which produce moderate conflict will be approached and in- vestigated. Berlyne does not explain why these conflict-inducing 10 stimuli are not instead avoided. Third, the collative prOperties are circularly defined: their presence is indicated by responses (orientation, approach, investigation, manipulation) which are theorized to be the result of arousal changes generated by the collative prOperties themselves. Related to this is the lack of any definition of the responses independent of the stimulus situa- tions presumed to bring them about. Last is a problem by no means unique to Berlyne, brought about by the naming of specific drives. In Berlyne's case, this consists of postulating two drives (curiosity and boredom) to account for exploratory behavior. Brown (1961) has noted that "drive" must be defined independently of the situation which it is assumed to cause, and which it is used to explain. "Thus the presence of a drive to explore is sometimes inferred from, and at the same time used to explain, behavior of moving from one place to another, especially if there is no other apparent reason for the movement" (Brown, 1961, p. 334). Such a circular line of reasoning is of questionable value as an explanation of behavior, regardless of how difficult the be- havior is to explain. Hunt(1965) also comments on several unfortu- nate explanations for unexplained behavior: The first consists in the naming of drives to account for each of such varied activities as play, exploration, manipulation, and curiosity; in the naming of such needs as those for stim- ulation and variation; and in the naming of such urges as those for contact and locomotion. Such naming of drives, needs, and urges seems to revisit the instinct-naming of McDougall (1908). After our soul-searching excursion into theoretical methodology in the thirties and forties, we should know better. Insofar as the drives, needs, and urges named are accepted as explana- tions of the activities, they are theoretically unfortunate" (Hunt, 1965, p. 196). 11 Of more specific interest is Berlyne's (1957) first report of the use of a series of stimuli varying in complexity and incongruity, to study perceptual curiosity and visual exploration. In their final form (Berlyne, 1963a) they constitute nine different cate- gories, each having two or more pairs of patterns. Each pair is made up of a less irregular pattern (LI) and a more irregular pat- tern (MI). According to Berlyne, all of the L1 figures elicit little conflict and are highly redundant, whereas the MI figures elicit conflict (via competing responses), are ambiguous, and evoke un- certainty, the result being the arousal of a curiosity drive and subsequent exploration to reduce that drive. The nine categories are: A. Irregularity of arrangement; B. Amount of material; C. Heterogeneity of elements; D. Irregularity of material; E. In- congruity; F. Incongruous juxtaposition; XA. Number of independent units; XB. Asymmetry, and XC. Random redistribution. The collative variables, at least as represented in these stimuli, have been shown to influence: 1) duration of exploration (Berlyne, 1957, 1958a, b; Berlyne and Lewis, 1963; Berlyne and Lawrence, 1964; Clapp and Eichorn, 1965; Day, 1965; Minton, 1963; Smock and Holt, 1962) 2) choice of stimulus to be viewed again (Berlyne, 1963a; Berlyne and Lewis, 1963; Day, 1965; Boats, Miller, and Spitz, 1963) 3) verbal ratings (Berlyne, 1963a; Berlyne and Lawrence, 1964; Berlyne and Peckham, 1966; Berlyne, Ogilvie, and Parham, 1968; Day, 1965, 1966) 4) paired-associate learning (Berlyne, Borsa, Craw, Gelman, and Mandell, 1965) 12 5) the incidence and content of children's questions (Berlyne and Frommer, 1966) 6) incidence and amplitude of the galvanic skin response (Berlyne, Craw, Salapatek, and Lewis, 1963) 7) duration of EEG desynchronization (Berlyne and Borsa, 1968; Berlyne and McDonnell, 1965). Clearly, a great deal of attention by experimenter and subject alike has been paid to these stimuli. For the purposes of the present discussion it will be most meaningful to consider three dimensions Of stimulation as represented in these and similar stimuli, namely, complexity, novelty, and incongruity. These variables are not difficult to define if one can be satis- fied temporarily with limited definitions. Physical complexity has been defined by Attneave (1957) in terms of characteristics of the perimeter, angles, and number of elements. It is probably unwise, however, to assume that subjective complexity varies in a one-to-one manner with physical complexity. Few psychOphysical relations are so simple. Day (1967) reported a monotonically increasing, though not one-to-one relation between objective complexity (number of turns) and judged complexity. Also, Johnson (1968) reported that familiar- ization did not affect looking time equally to three levels of com- plexity: The objectively most complex polygons (31 turns) resulted in a greater decrease in looking time with familiarization than did the least complex polygons (8 turns), indicating that objective and sub- jective complexity differ. A functional, subjective definition of complexity depends not only on stimulus characteristics and method of measurement, but also on the subject's past experience, leading 13 us to expect both individual differences and strong deve10pmental differences in subjective complexity. Novelty is more difficult to appraise, because complete his- tories of human subjects are never available. However, if the eXper- imental treatment is preceded by familiarization with some stimuli, short-term, relative novelty is readily defined. Novelty will be defined in this way in the studies to be reviewed here. The "incongruity" variable is not so easily defined. The figures used by Berlyne, and modifications used by others, are clearly novel in a long-term sense. However, Berlyne and others Often imply that incongruity is something more than novelty pg; g9. Inspection of the figures (described in Appendix A) reveals that they are without exception familiar objects that have been altered in some way. One's expectation or prediction of the identity of the figure based on one part may be unfounded. A word of caution is in order. The term "incongruity" is a poor definition of stimulus characteristics, because it implies internal processes that are assumed to be aroused by the stimuli, and therefore is circular. Ideally, the definition of stimulus conditions should be in terms of the characteristics of the stimuli alone, especially when those stimuli are assumed to affect in some lawful way a series of organismic processes. A better term for these figures would be "altered" stimuli, and they will be described in that way here. We now turn to discussion of the various effects of complexity, novelty, and incongruity. These are discussed separately for clarity of exposition, but it will become clear that they are strongly related. l4 COMPLEXITY A number of experiments using nonhuman subjects provide evidence that a stimulus' complexity has motivating prOperties. Among the ex- periments already cited, Dashiell (1925), Nissen (1930), Butler (1954), and Berlyne (1955) can be interpreted to mean that exploration is elicited under conditions of greater stimulus complexity. It is im- portant to note, however, that complexity, novelty, and simply change are not independently defined and manipulated in these and other studies and the observed exploration could result from manipulation of any of these variables. Dember, Earl, and Paradise (1957) have shown that the rat's ex- ploration is influenced by the complexity of its surroundings. Rats which initially chose the less-complex leap of a figure eight maze (horizontal stripes) chose the more complex loop on subsequent days (vertical stripes). One might also suggest that the more complex pattern was chosen as it lost novelty. The most abundant evidence for the motivating effects of complexity comes from studies of visual exploration and verbal evaluations, to which we now turn. Because a number of potentially important age differences have been obtained in these studies, the results for college students and children will be discussed separately. The effect of complexity upon visual exploration College students: Berlyne (1957) devised a series of stimuli with which to determine the properties of visual stimulation which influence investigation. College students pressed a lever which exposed a figure in a 15 tachistosc0pe for 0.14 second. They were told that they could look at each figure until they had seen enough of it, and that they would be asked no questions about the figures. The subjects took part in four experiments: experiments 1 and 3 were concerned with incongruity and surprisingness, and will be discussed later; experi- ments 2 and 4 were manipulations of complexity. In all cases, figures 'with more complexity (absolute or relative uncertainty) elicited signif- icantly more lever presses than did less complex figures. Berlyne (1960, p. 162) notes that subjects reported that they pressed the lever until they had identified the figure, then stOpped. Apparently the identified figures aroused, in Berlyne's terms, little curiosity. A common problem in studies using tachistoscopic presentation is that it is not clear why the subject exposes the figure several times. This may be a subtle effect of the instructions to the subject and the fact that he is in an experiment. He may look at the figures more in order to identify them, as Berlyne implies. Lastly, and this would seem to be what Berlyne is interested in, the subject may look at the figure several times aftgr identifying it because of some quality which arouses him with the result that he explores it. Tachistosc0pic pres- entation does not allow separation of these potential factors. An experiment is needed to determine whether subjects continue to look at stimuli having collative properites 3:53; they have identified them. Berlyne's experiment was replicated in all essentials by Minton (1963). Berlyne may have recognized these methodological problems, because in a subsequent experiment (1958a) he examined the influence of the collative variables on orientation (selective attention), not inves- tigation, and enlarged his pool of stimulus figures. College students 16 were presented the pairs of figures for ten seconds per pair, and the duration of eye fixation to each member was recorded. Over all categories, the more complex figures were fixated for a significantly greater proportion of the ten-second period than were the less complex figures. In order to rule out the possibility that ten seconds was not enough time to identify the figures, Berlyne (1958b) replicated the eXperiment with a two-minute viewing time per pair. The same results were obtained. Thus, recognition or identification may not be a factor in the greater prOportion of time spent looking at the more complex figures. Unfortunately, Berlyne did not in- struct his subjects to report when they recognized the patterns. Another change in method took place in a 1963 study (Berlyne, 1963a). In this case, college students were shown the two members of the pair in succession, then allowed to choose one to look at again. When exposures were short (0.5 - 1 second), more "more irregular" 0M1) choices were made. However, at longer exposures (3 - 4 seconds), the subjects made more LI choices. In view of the ratings which also were obtained (the MI figures were more interesting but the L1 figures were more pleasing), it would appear that the interestingness of a figure is related to choice at short exposures. At long exposures, however, the choice of which to look at again is a function of how pleasing the figure is. Perhaps the motivating property of interestingness decays rather rapidly as time passes, while that Of pleasingness rises more slowly but is longer-lasting. Day (1967) has shown such a relation for four- to 160-sided polygons, but in terms of number of sides, not exposure time. Berlyne's results are potentially confounded however, because the ratings 17 were made only at a three-second exposure time. In order to com- pletely relate "choice to look at again" with interestingness and pleasingness, the ratings must also be made at the short exposure time. Berlyne and Lawrence (1964) allowed subjects to control their own exploration of the stimuli, and the length of time between the subjects' button-presses (which actuated the slide changer) was re- corded. The previous results were replicated with this changed pro- cedure. An attempt was also made to determine whether these stimuli changed the subjects' arousal state, as indicated by galvanic skin response, heart-rate change, or finger blood volume. There were no changes in any of these measures, either over three presentations of the same pattern or betweeen more- versus less-complex patterns. Berlyne, Craw, Salapatek, and Lewis (1963) repeated the GSR part of the Berlyne and Lawrence experiment, but with three-second ex- posures. There was a higher incidence of GSRs to the more complex or incongruous patterns than to the L1 members of the pairs, but only when the subjects were "extrinsically motivated." (Subjects were instructed, "You will have to recognize these later.") Berlyne and McDonnell (1965) used EEG desynchronization as the measure of arousal, and found longer desynchronization to MI patterns than to LI patterns. Berlyne and Lewis (1963) allowed subjects to control their dura- tion of exposure, and also used the choice task used by Berlyne (1963a). Three groups of subjects underwent conditions which were meant to induce higher arousal: shock expectation; memory test expectation, and white noise. A control group received the standard instructions. GSR changes indicated that the instructions did l8 affect the subjects' arousal state. The effect of heightened arousal was to increase the duration of exploring all patterns, and to increase the number of LI choices to be seen again. Over all groups, and over all categories but one, the duration of exposure was greater to MI patterns than to LI patterns. Thus, the heightened arousal tends to reduce the probability of choosing a more complex or incongruous pattern, but activates behavior such that any figures are looked at longer. Day (1966) suggests that there is an Optimum level of complexity as reflected in looking-time scores. While the X series patterns were explored longer than the Non-X patterns, the MI patterns in the X series were explored less than the LI patterns in that series. No attempt has been made to quantify the complexity of these stimuli, however, so Day's conclusion is based on his own judgment of the rela- tive complexity of these figures. Four studies have used stimuli different from Berlyne's (photo- graphs of landscapes and objects, and randomly-generated polygons). Leckart (1966), Faw and Nunnally (1967), Johnson (1968), and Willis and Dornbush (1968) have all found that college students look longer at (or choose more frequently) more complex patterns than less com- plex ones. In summary, the complexity of a visual array is clearly related to the time a person looks at it, and to the choice of a pattern to be seen again. In general, more complex patterns are looked at longer and chosen more frequently than are less complex patterns, although there is a hint that some optimum of complexity may exist. With regard to arousal mechanisms, Berlyne, et a1. (1963) and Hunt 19 (1965) have commented that complexity and other collative variables seem to have a greater effect on exploration than on physiological processes. Children: The research on infants' responses to complexity is in a state of flux. Until recently it was dominated by poor methods, and it is still related to studies concerned with responses to faces or face-like figures. It may be the influence of this latter that often results in an unsystematic selection of stimuli, some intui- tively more complex than others. Cantor (1963) presents an excellent critical review, now outdated however, of the responses of infants and children to complexity and novelty. Berlyne (1958c), Fantz (1958) and Spears (1962) all presented pairs of patterns to infants of ages up to nine months. They re- corded either the pattern which was fixated first (Berlyne) or the prOportion of time each was fixated (Fantz and Spears). It appears that patterns of greatest complexity were fixated first or longer. These were a checkerboard pattern, a computer-generated pattern of random dots, and a bull's-eye. These results were questioned by Hershenson (1964), who analyzed motion pictures of infants' eye fixations, a more precise technique than those used by Berlyne, Fantz, or Spears. He found that two- to four-day-old infants fixated most on a 2 X 2 checkerboard, versus 4 X 4 or 12 X 12 patterns. Contrariwise, an inverted-U relation be- tween looking time and complexity was obtained by Hershenson, Mun- singer, and Kessen (1965), who presented randomly-generated polygons of 5 - 20 turns to new-born infants. The results of these two 20 studies may not be comparable, because checkerboards may represent a different level of complexity than do randomly-generated polygons. In fact, the physical measures of complexity would seem to be dif- ferent for the two kinds of stimuli, the checkerboards involving internal change, and the polygons involving change at the border. The relation between age and visual responses to complexity was made clearer by Brennan, Ames, and Moore (1966). These investigators found, as did Hershenson, that S—week old infants fixated longest at a 2X2 checkerboard. Fourteen-week-old infants, however, looked longest at a 24x24 pattern, and 8-week-old infants fixated an 8X8 pattern the longest. Thus, there appears to be a direct relation between age and "preference" (indexed by visual fixation time) for complexity. This direct relation has been confirmed by Munsinger and Weir (1967) using children 9 to 41 months of age. In summary, it may be speculated that complexity and looking- time are related by a curvilinear function (inverted-U), and that this function varies monotonically with age, at least during infancy. Since the research on preschool and elementary school children appears to have been initiated by Berlyne's research with adults, his stimuli or variations on them are often used. Unfortunately, while there is some continuity across stimuli, there is little similarity in the methods used to determine how long children explore various visual arrays. Smock and Holt (1962) allowed first grade children to press a button, which projected a figure for 250 milliseconds. The number of presses to each figure was recorded. As in Berlyne's studies, the stimulus slides were designed in pairs, one member 'more complex' 21 (by unknown criteria) than the other. Despite a poor analysis of the data, it is clear that the children gave more responses to the 'more complex' figures. There was also a sex interaction: girls responded more to heterogeneity of dispersion, and boys responded more to greater amount of contour. Cantor, Cantor, and Ditrichs (1963) presented six triads of stimulus figures to preschool children, the members of each triad varying in complexity according to some un3pecified, intuitive defi- nition. The subjects sat before three boxes, each one containing one member of the triad. By pressing his forehead against a switch on the box, the subject turned on a clock timer and a light in the box which illuminated the figure. Each triad was presented for 60 seconds, during which time the child could view any of the three figures for as long as he desired. More time was spent viewing the high-complex figures than the medium- or low-complex figures. Thomas (1966a) introduced another technique. The child's head is positioned on a chin rest, and movement of the head to the left or right actuates switches which turn on projectors behind two viewing screens. Thus when the left projector is actuated, one can be fairly certain that the child is looking at the left screen and not somewhere else. In an extensive study (1966b), Thomas used this device to present all possible pairs of 3-, 6-, 10-, 20-, and 40-turn randomly-generated polygons to children of ages 6, 7, 8, 9, and 12 years. The duration that each figure of the pair was il- luminated was recorded, but the time data were not used, and "the stimulus displayed the longest was taken as the preferred stimulus." Thomas does not explain the basis for that decision, 22 whether on raw scores or on proportion of total time. Treating the data in this way, Thomas obtained the equivalent of paired- comparison judgments, which were scaled. Over all age groups, scaled scores increased with increasing complexity. Other studies (May, 1963; Pielstick and Woodruff, 1964; Faw and NunnalLy,1968a;JOhnson, 1968; Willis and Dornbush, 1968) illustrate the continuing variety of stimuli and responses. More- over, they all lead to the same conclusion: that children (pre- school through sixth-grade) choose to look at more complex patterns, and look at them longer than less complex patterns. The evidence is not so simple as the above studies would seem to indicate, however. Lore (1965) used five- and six-year-old children from middle- and lower-class homes, and allowed them to control the length of exposure of the slides by means of a lever. The 20 stimuli comprised four categories: incongruity, asymmetry, complexity, and number of turns. The complexity stimuli were modi- fications of those used by Cantor et a1. (1963). No subjects looked longer at complex than at simple figures, or at figures having a greater number of turns. Only the middle-class children showed longer viewing times to asymmetric versus symmetric figures. These results were replicated in all respects by a 'reduction of hand activity' measure in a second experiment. At least for these stimuli, complexity had little effect on children's viewing time. This is not the only such finding, however. Hoats, Miller, and Spitz (1963) have reported similar negative results, but their scanty analysis of the data and failure to apply any significance tests make the results very difficult to interpret. The instructions 23 may have contributed to this. "You can look at the picture you choose for as long as you like” might have been interpreted by the children in terms of pleasingness, and Berlyne (1963a) showed that the less irregular patterns were rated (by adults) as more pleasing. Clapp and Eichorn (1965) have also failed to obtain differential visual exploration of complex versus not-complex figures, with four- to five-year-old children. The stimulus patterns were taken from Berlyne (1957 ), along with another series of meaningful objects. Tachistosc0pic presentation was used, with the variation that the child did not press the button, but told the experimenter when to stOp pressing and change to another slide. With neither the geometric figures nor the meaningful figures did the children respond at other than a chance level to the more complex stimuli. These results are difficult to interpret, also. Having the child tell the experimenter when to stop may have introduced a social factor into the experi- ment. Themaare, after all, few times when a young child is asked to tell a strange adult when to stop doing something. In summary, greater complexity usually results in longer visual exploration, depending on the stimuli and methods used, and the age of the children. The results are not as clear as those obtained with adults. Clearly, a systematic developmental study, using well-defined stimuli and both choice and duration-of-exploration tasks remains to be done. The effect of complexity upon ratings and Hpreferences" The research cited in this section is concerned with verbal evaluative responses to stimulus complexity. Increased exploration of more complex situations may be explainable in terms of the 24 subject's evaluations of the stimuli, although it is just as likely that exploration and evaluation are merely two reflections of a common underlying process, changes in which bring about changes in both exploration and evaluation. That is, some figures may be looked at longer because (or while) the subject finds them "likable" or pleasant or interesting. Other figures may be looked at less because (or while) the person finds them uninteresting or unpleasant or "unlikable." In terms of arousal processes, these ratings might be considered to be the verbalizable emotional concomitants of arousal changes brought about by certain stimulus conditions. In a sense, they represent somewhat crude estimates of what is happening to a person who encounters complexity. Though certainly cruder than direct measurement of physiological processes, they are much easier to obtain. Berlyne's figures have been used in only a small number of the studies of preferences for varying degrees of complexity (Berlyne, 1963a; Berlyne and Lawrence, 1964; Berlyne and Peckham, 1966; Berlyne, Ogilvie, and Parham, 1968; Day, 1966). Berlyne and Lawrence found that college-age subjects expressed a preference for the less- irregular figures over all categories. Berlyne (1963a) made this preference clearer, showing that more-irregular patterns were rated more interesting, but less-irregular patterns were rated more pleasing. Berlyne and Peckham used all stimulus categories except "Incongruity" and "Incongruous juxtaposition” and related pleasingness ratings to the Semantic Differential. They found a bimodal distribution relating either 'evaluative' or 'potency' dimensions to complexity, and a curvilinear (inverted-U) relation between 'activity' and complexity. These functions depend, however, on the experimaters' subjective 25 ordering of the stimulus figures as increasing in complexity. There is no other information to support such an ordering. Day (1966) asked his adult subjects which of the X and Non-X series figures they liked. He found an increasing proportion of "like" ratings with increasing complexity (to the extent once again, that degrees of complexity can be intuitively determined from Berlyne's figures), though the rating scores seemed to reach an asymptote at the high level of complexity. Finally, Berlyne, Ogilivie, and Parham (1968) applied a multi- dimensional scaling technique to ratings of the complexity, interest- ingness, and pleasingness of 16 of the figures. They found significant correlations between ratings of complexity and interestingness (+.54), between complexity and pleasingness (+.4l), and between interestingness and pleasingness (+.23). This suggests that interestingness and pleasingness are related to (and may be part of what we mean by) complexity, though they are not strongly related to each other. The problem of obtaining some physical measurements of complexity is evident in Berlyne's figures, and was overcome by Munsinger and Kessen (1964) and Munsinger, Kessen and Kessen (1964), through the use of randomly shaped polygons of various number of sides (Attneave and Arnoult, 1956). In these stimuli, complexity is defined in terms of the number of independent turns or sides. Adults (Munsinger and Kessen, 1964) and children of ages 6 through 15 years (Munsinger, et a1., 1964) were shown all possible pairs of 12 stimuli, which varied in number of sides from 3 to 40. Up to age 9 or 10, the relation between complexity and stated preference was monotonic, the most complex stimuli were the most preferred. 26 This is the same relation as has been obtained using visual eXplor- ation as the response. After age 9 or 10, the relation becomes curvilinear, the highest preference occurring to stimuli of inter- mediate complexity, around 10 turns. Additionally, the lO-turn stimuli were highly preferred by all subjects. These studies prompted Thomas (1966b) to replicate, with some improvements. Specifically, Thomas suggested that the unique results Obtained with the lO-turn stimuli might be due to stimulus sampling error. To test this he constructed 4 stimuli at each of 5 complexity levels (3 through 40 turns). Subjects of ages 7 through 19 years were presented with all pairs of complexity levels and were asked to choose which figure of the pair they liked best. Thomas found a monotonic relation between preference and complexity until age 15 or 16, after which an intermediate level of complexity was preferred. The different ages obtained by Munsinger, et a1. and by Thomas, at which the shift in preference from most to intermediate complexity takes place cannot be reconciled at this time. This age difference requires further investigation. Finally, Thomas did not find any outstanding convergence of preferences across all ages for the lO—sided figures, as Munsinger and Kessen had found. Stenson (1966) correlated adults' judgments of the complexity of random forms with 24 different physical measurements of complexity, and factor analyzed the resulting matrix of intercorrelations. Four measures accounted for most of the variance in the ratings: number of turns, length of perimeter, ratio of perimeter squared to area, and variance of internal angles. Thus there is some correspondence between 27 (arbitrary and complicated) physical measures of complexity and sub- jective measures. Finally, Day (1967) presented all possible pairs of random polygons of 4 to 160 sides to adult subjects, and obtained judgments of complexity, pleasingness and interestingness. While subjective complexity increased monotonically as the number of sides increased, interestingness ratings slowly reached an asymptote. Pleasingness ratings rose more quickly, reaching a maximum at between 20 and 40 sides, and then fell steadily. Thus, at a high number of sides (100- 160) the patterns were rated interesting but not pleasing. At a low number of sides (4-20) the patterns were judged pleasing but only slightly interesting. This result is clearly related to Berlyne's (1963a) finding, that more-irregular figures were rated more interest- ing, but less-irregular and less complex patterns were rated more pleasing. Clearly, a variety of judgments of figures change as the physical complexity and to some extent, the subjective complexity changes. Just as clearly, complexity appears to have similar effects on both verbal evaluations and looking time. The work of Berlyne and of Thomas indicates that these ratings and preferences are related to the duration of time that adults and children explore those patterns. NOVELTY A wide variety of experiments with non-human subjects have implicated novelty as a motivating factor, so many that the relation between novelty and exploration hardly requires substantiation. Dashiell (1925), Nissen (1930), Berlyne (1950), Montgomery (1952, 1953), and Glanzer (1953b) all provide abundant evidence that novelty influences 'ludic' 28 behaviors. Denny and Leckart (1965) have shown that rats avoid the side of an E maze to which they had been forced and rewarded. Even when the free-choice trials were given one trial per day, avoidance of the previously forced and rewarded side still occurred after seven days. Furthermore, food was available at the previously forced and rewarded side during the free-choice trials. When the rat's behavior is viewed as approach to the unfamiliar side the effect seems more powerful, because the hungry rat is avoiding food. The authors conclude, "The traditional notions of reinforcement are challenged because learning occurred without the presence of an identifiable reinforcer, directly in the face of food reinforcement for the Opposite response." (p. 232) There have been observations of both withdrawal and approach in a novel situation (Berlyne, 1950; Melzack, 1954; Welker, 1956 a, b, 1961; Mason, 1965). Welker presented chimpanzees with novel moving toys. The first response was fear and withdrawal, later followed by alternating approach and withdrawal, and eventually leading to mani- pulation and subsequent habituation. Such observations are not uncommon in children. Mason (1965) has reviewed his own work with chimpanzees, which shows that play behavior is most likely to occur in conditons of moderate novelty. Unfortunately, Mason's judgment of the relative novelty of situations is largely subjective. Still, this research and Welker's is the only real response to Beach's (1945) plea for a concerted study of the nature and function of play. The out- standing conclusion to be drawn from Beach's review of attempts to explain play is that chaos abounds when a behavior is found for which 29 there is no manifest biological purpose. Beach pointed out that the existing "theories" of play were circular and non-explanatory. These theories resulted when a class of behavior was examined to which a need-drive formulation patently did not fit. But, and this is the most telling point in Beach's paper, there was little to fall back on when the drive-reduction model failed. The relation between novelty and fear is amply supported by Gordon Bronson (1968a, b) in his recent reviews, and also by Berlyne (1960, Ch. 5) and Hunt (1963). That novelty and fear are related is reflected in Bronson's (perhaps unfortunate) definition of fear as a response to visual novelty, and Mowrer's (1960) analysis of exploratory behavior as determined by fear. Although sporadic and Often crude observations of fear in human infants had been made in the 19303 (X. Bronson, 1968a), the work of Kahler (1927), Valentine (1930), Hebb and Riesen (1943) and Hebb (1946) were especially important. There is reason to suspect, however, that novelty was nOt the only cause of the observed fear, so this work will be discussed in the section on "incongruity." The effect of novelty upon visual exploration College students: Novelty, at least as defined in terms of familiarization trials, has not been greatly studied in adults. Many of the stimuli used by Berlyne and others to study other collative variables are novel in a long-term or absolute sense, but the term "novelty" could and should be used more presisely. It is of some interest that there are consid- erably more studies of the effect of novelty on children than on adults. Berlyne (1958a) presented pairs of stimuli side-by-side to 30 college students. Over 7 trials the pattern on one side did not change, while a different pattern on each trial was shown on the other side. Subjects looked significantly longer at the side where the novel or varying stimuli appeared than at the side where the same figure appeared on each trial. Leckart (1966), in the study referred to previously, manipulated familiarity, giving different groups of subjects 0, 10, or 20 seconds of familiarization to each pattern. After a delay of either 0 or 48 hours (which had no effect), the subject looked at the same patterns again for as long as he desired. There was a decrease in looking time with increasing familiarization, implying that greater novelty resulted in longer visual eXploration. Johnson (1968) found that both adults' and children's looking time decreased over 45 familiarization trials, but that the decrease was not the same for high-, medium-, or low-complexity stimuli. Children: A variety of children's responses to novelty have been studied, and the literature is both recent and extensive, enough so that the review by Cantor in 1963 is very much out of date. This literature will not be reviewed here in its entirety. However, the experimenter's notion of the ubiquity of novelty in the young child's world is worthy of some attention. The influence of novelty on children's choice of toys has been clearly shown by Harris (1965, 1967a, 1967b), Endsley (1967), and Mendel (1965). Green (1964) has shown the distracting effect of a novel stimulus in a discrimination learning task, and Ellis and Arnoult (1965) have shown that children tracing a T shape alternate less than chance when the situation changes from trial to trial. 31 In a similar study, Harris and Granskog (1968) found just the opposite result, the more the situation varied from trial to trial the more likely were the children's maze-tracing responses to be more variable. The results of May's (1963) study, already discussed in the section on complexity, could also be interpreted to mean that the children chose the more novel and complex 8- and 12-rectangle cards, after familiarization with the cards bearing five rectangles. In two studies, Cantor and Cantor (1964a, b) have shown that kindergarten children look longer at novel than at familiar figures, regardless of the delay between familiarization and self-controlled visual exploration. This result holds whether the figures are from the Welsh Figure Preference Test, or are cartoon drawings from a children's toy (Kenner Give-a-Show). Leckart, Briggs, and Kirk (1968) showed four- and five-year-old children pairs of pictures taken from children's books. For 20 trials the picture on one side did not change but the picture on the other side changed on each trial, similar to Berlyne (1958a). When asked which side "they would enjoy looking at most," the novel side was chosen most frequently, and this novel choice increased over trials, and increased more for girls than for boys. Similar results have been reported by Lewis, Goldberg, and Rausch (1967), though a sex difference was not found. In summary, novelty even in this restricted Operational sense, exerts a strong influence on the exploratory behavior of children and adults alike. The strongest support for this conclusion comes from Harris (1965). Children not only chose the novel toy overwhelmingly 32 in excess of chance expectation, but chose the novel toy when it meant giving up two familiar toys, even when the novel toy was broken. The effect of novelty upon ratings and "preferences" Subjective novelty has been little studied. The only investi- gation is an extensive one by Berlyne and Parham (1968), in which adult subjects rated the novelty of colored random shapes under eight conditions of successive presentation. Berlyne and Parham found that a stimulus X is rated less novel: 1) when X was experienced recently 2) on the second presentation of X 3) when the subject was familiar with all the stimuli 4) with repeated presentations. The stimulus was rated more novel: 1) with the number of repetitions of another stimulus Y before X 2) with the number of ways X differed from Y. They found no evidence that novelty varies with the time since the last presentation of X, with the number of prior presentations of X, or with the number of kinds of stimuli preceding the first presenta- tion of X. These last are surprising findings, and, as the authors note, should be investigated further. The only study in which novelty was manipulated and ratings on some dimension obtained is that of Cantor (1968). Figures from the Welsh Figure Preference Test were shown to fifth- and sixth-grade children. Ten figures were presented in familiarization trials, then the children rated those ten and ten others which they had not seen 33 on a lO-centimeter scale, from "strongly dislike" to "strongly like." The unfamiliarized stimuli received higher ratings. Cantor interprets this to mean that children look at novel stimuli longer than familiar stimuli because they like them more, not because they contain more information, as Berlyne argues. Neither Cantor nor Berlyne suggest that both "liking" and information content may determine eXploration, and Cantor fails to ask what is the basis for the preference for the novel figures. It is possible, of course, that children like these figures because they contain more information. INCONGRUITY In contrast to complexity and novelty, relatively little can be learned about incongruity from previous studies of nonhuman animals. The only study using rats as subjects is that of Haywood and Wachs (1967), and the results of this study are not clear in many ways. Their definition of incongruity as 'not experiencing shock where shock was experienced before' is likely a better definition of sur- prise. They did find that rats who received shock or "shock incon- gruity" chose a novel versus familiar endbox at a chance level, while control rats chose the novel endbox on 25 of 32 trials. Haywood and Wachs interpret these data in terms of Hebb's theory, that there is some Optimal level of arousal, and that situations that are highly arousing (beyond the optimum in this case), such as shock or white noise, will result in fewer choices of novel stimuli. Development of fear in chimpanzees and children The research discussed in this section is important because it represents the earliest discussion of incongruity as a motivational 34 factor, and because it provides descriptions of situations that might be called "incongruous,” and represents a crude operational definition of the term. The findings of Hebb and Riesen (1943) and Hebb (1946) are now classic, largely because they showed that fear in young chimpanzees could not be explained by conditioning, which Watson (1920, 1924) maintained was the means by which all but the most rudimentary, un- learned fears (to loud noises and loss of support) were develOped. In the 1943 study, Hebb and Riesen observed that chimpanzees do not show fear of strangers until about four months of age, under care- fully controlled laboratory conditions in which the animals' histories were known and any association of a stranger with a noxious event could be ruled out. Hebb (1946) later showed that laboratory-reared chimpanzees showed withdrawal and fear responses to such stimuli as the sculpted head of a chimpanzee with the body missing, an anesthe- tized infant chimpanzee, a painted human eye and eyebrow from a mannequin, or the experimenter in the familiar animal caretaker's coat. Because Hebb knew the histories of his animals he could rule out learning as an explanation for these fear responses. He hypo— thesized, that fear occurs when an object is seen which is like familiar objects in enough respects to arouse habitual processes of perception, but in other respects arouses incompatible processes" (1946, p. 268). Observations similar to Hebb's had been made a number of years earlier by KOhler (1927) and Valentine (1930). K6hler observed that his chimpanzees showed extreme fear to primitive stuffed toys, "...caricatures of oxen and asses [with which the chimpanzees were familiar], though most drolly unnatural" (1927, p. 320). 35 One day I entered their room with one of these toys under my arm. Their reaction times can be very short; in a moment a black cluster, consisting of the whole group of chimpanzees, hung suspended from the farthest corner of the wire-roofing, each individual trying to thrust the others aside and bury his deep head [sic] in among them" (p. 321). The chimpanzees also showed fear when KOhler approached them wearing a demon mask. K5hler uses these observations to argue against the proposition that experience (presumably learning) is necessary for the deve10pment of fear, because the chimpanzees had never encountered any of these things before, and yet he notes: It is too facile an explanation of these reactions to assume that everything new and unknown appears terrible to these creatures. Any geometrical figure of wood found standing or lying about, though it represents something quite new to them, rouses no such convulsions of terror, even though in the first moment it is rather cautiously examined. New things are not necessarily frightful to a chimpanzee, any more than to a human child; certain impressive qualities are requisite to produce this special effect" (pp. 321-322). Kahler suggests that these impressive qualities are surprise and the unknown. Valentine (1930) took Watson to task for oversimplifying the deve10pment of fear, and suggested that general fears or predispo- sitions to emotional arousal were based on maturation. He observed that his daughter (age 12 months) showed fear of her brother in a cat costume, and (age 14 months) when her brother put a paper bag over his head. Further observations, between 17 and 29 months of age, revealed fear to: her father crawling toward her on hands and feet, "monkey-wise", with head down; a doll whose head Opened back- wards; a nearly decapitated doll, and the doll's eyes joined by a wire. Valentine refers to these events as "uncanny", in the sense of 36 strangeness associated with the very familiar. He concludes: Even assuming that the fear of animals is always due to some chance association with loud noises or withdrawal of support or to suggestion, we can scarcely imagine that such weird combinations -- that of a boy with a paper bag on his head or a boy dressed as a black cat -- can become fearsome by association. These and other examples in the records of careful observers, of fear of uncanny strange things, make it certain, as it seems to me, that we must assume an innate tendency to fear under certain conditions. Not that there is a specific definite fixed innate fear of, say, all furry animals; what the facts suggest to my mind is that there is, first, a general tendency to fear the very strange, especially when closely associated with the familiar" (p. 409). Although KOhler and Valentine did not have control over the histories of their subjects as Hebb did, there is striking similarity in the events that aroused fear both in chimpanzees and in young children. In each case the Object or situation is familiar, but has been changed in some way. It is not clear whether the change is from familiar to unfamiliar, or familiar to 'familiar-but-changed', or perhaps 'familiar-but-unfamiliar'. Hebb interpreted his findings not in terms of novelty alone, but rather in terms of "the familiar in an unfamiliar guise." Hebb's theory, later elaborated (1949) held that while the fear of novelty may be unlearned, a certain amount of experience is necessary to encode the familiar before an organism can identify a pattern as novel. "The familiar in an unfamiliar guise" is perhaps as good a defi- nition as any for the variable which Berlyne and Others have referred to as "incongruity." As will be shown in the sections to follow, incongruity results in visual exploration in humans, both children and adults, and there are only valentine's suggestive observations of his own child as evidence of fear or withdrawal. As has been shown, in- congruity presented to nonhuman animals at a variety of ages results 37 in withdrawal and fear, and approach or exploratory responses have not been observed. Such a clear dichotomy is not common, and re- presents an interesting problem in itself. The effect of incongruity upon visual exploration College students: At this juncture, it is well to call attention again to the point made earlier, that ”incongruity" or "incongruous" are poor terms to use to describe stimulus characteristics. ”Incongruity” seems to refer to the presumed competing responses aroused by these figures, rather than to the characteristics of the figures themselves. Henceforth in this discussion, stimulus figures which have been called "incongruous" or "congruous" will be labeled "altered" or "unaltered." "Incongruity" may well be a useful term, but not when applied to stimulus characteristics. With only three exceptions (Bruner and Postman, 1949; Connolly, 1969; Nunnally, Faw, and Bashford, 1969), the stimuli and methods used to investigate incongruity as a motivating variable have been those used by Berlyne. The stimulus figures that have been used and referred to as "incongruous” are described in Appendix A. Bruner and Postman's investigation was prompted not by questions about exploratory behavior, but rather by questions about the nature of perceptual organization when expectations were not confirmed. They altered six playing cards such that the color was reversed, e.g., a black four of hearts. College students were presented 1, 2, 3, or 4 of these cards along with normal cards in a tachistoscope, and recog- nition thresholds were determined. The thresholds for altered cards were significantly higher than for normal cards, and decreased de- pending on how many altered cards had been seen previously. Also, 38 the number of prerecognition responses that were compromises ("brown spades," "purple," "rusty black") was greater to the altered cards. Bruner and Postman interpret these compromises as resis- tance to stimulation that fails to confirm the observer's expecta- tions, that is, is unexpected or incongruous. The increased recog- nition thresholds, they maintain, are the result of this resistance. The existence of any compromise reactions has been questioned by Kempen, Hermans, Klinkum, Brand, and VerHaaren (1969). They suggest that the results obtained by Bruner and Postman may have been due to the red paint used to alter the playing cards, a darker, less yellow shade of red than on normal cards. Kempen et al. suggest that the 'incongruity' is not only the red-black inversion, but also the unusual (for playing cards) red color. They had playing cards made with the red-black inversion, but in the usual colors, and replicated the Bruner-Postman experiment exactly. Their results agreed con- siderably with Bruner and Postman's, with one exception. They found no compromise reactions, although they do not report how their sub- jects were encouraged to verbalize what they were seeing in the tach- istoscope. It is not clear why the recognition thresholds are higher. The recognition threshold method has been amply and justly criticized (y. Dember, 1960; Eriksen, 1960), because it is not clear that the subject reports at the same moment that he recognizes. He may recog- nize the figure, but not report it, perhaps thinking "I couldn't have seen that," or "I can't say that to him (or her)" (the experi- menter). Thus the longer durations required for correct recognition of the altered figures may be the result of a conceptual conflict 39 rather than of perceptual processes alone. There is clearly a possibility that higher order, more complex processes than perception alone may influence the results of such experiments. In addition, the results are potentially confounded with the social relation that exists between subject and experimenter. The studies by Berlyne which are germane to this discussion also have involved novelty or complexity, so they have been discussed in previous sections. When responses to altered and unaltered figures are compared, altered figures 1) are viewed more frequently in a tachistosc0pe (Berlyne, 1957 ; Minton, 1963) 2) are chosen more frequently to be seen again (Berlyne, 1963a) 3) are fixated for a greater proportion of a lO-second or 2-minute period (Berlyne, 1958a, b) 4) are explored fin: a longer period of time (Berlyne and Lewis, 1963; Berlyne and Lawrence, 1964) 5) result in greater incidence and amplitude of GSR (Berlyne, et al., 1963) and longer EEG desynchronization (Berlyne and McDonnell, 1965). Adult males fixate longer on some of Berlyne's altered figures than at the unaltered versions (Faw and Nunnally, 1967), and adult females look longer at 'more incongruous' figures than at 'less incon- gruous' ones (Nunnally, Faw, and Bashford, 1969). However, the latter finding is not without question and will be discussed more fully in the section on "'preference' for incongruity." Clearly, adults respond differently to familiar figures that have been altered than to their unaltered counterparts. No study has 40 failed to obtain these results. Children: Much of the research to be reviewed here has examined novelty, complexity, and incongruity in the same experiment, and so has been discussed in previous sections. The studies which show that altered figures are explored longer are those of Smock and Holt (1962), Lore (1965), Clapp and Eichorn (1965), Faw and Nunnally (1968a), Connolly (1969), Dodd and Lewis (1969), and Nunnally, Faw, and Bashford (1969). In the case of the Smock-Holt and Clapp-Eichorn studies, the altered figures were part of Berlyne's "animals" and "birds" series. Thus these positive results are based on only four figures: a bird with three heads; a bird with the lower body of an elephant; a dog with an elephant's head, and a camel with a lion's head. Smock and Holt found a strong sex interaction, the boys showing a larger difference score than the girls between the altered and the unaltered figures. Clapp and Eichorn found no such sex difference. Lore (1965) presents results which largely support those above. Similar to the results for complexity, only the middle-class children reSponded significantly more to the altered than to the unaltered figures. It is possible that the effect was masked in the lower- class children by the elaborate apparatus, being "finger printed” for the GSR measure, and by the attention of the experimenter. Attempts to determine social-class differences in reactions to figures such as these will have to be more carefully controlled, to the ex- tent that machinery, explicit instructions, and an adult's avid atten- tion can be expected to have differential effects on children from 41 different social classes. Faw and Nunnally (1968a) presented pairs of altered figures to seven- to 13-year-old boys, and measured the prOportion of time that each figure was fixated. Two pairs of stimuli were taken from Berlyne's "Incongruous juxtaposition" category, and two other pairs were made up by the authors. The altered figures were fixated sig- nificantly more (58.7%) of the total time that either figure was fixated. In experiment II, an attempt was made to scale "novelty" (the authors' description of their stimuli) by designing other altered figures. The stimulus pairs were presented as before to 33 nine- and ten-year-Old girls, and fixations were recorded. After the fixation trials the subjects were asked to rank the stimuli on the basis of their familiarity. The two sets of figures are shown in Appendix A. A monotonic increasing relation obtained between the scaled fixation score and the mean rank assigned to the figure. That is, altered figures rated as less familiar were fixated a greater percentage of the time. This result has been replicated twice (Faw and Nunnally, 1968b; Nunnally, Faw, and Bashford, 1969). Dodd and Lewis (1969) assessed the reactions of 60 two-year-Old children to altered and unaltered figures. The figures were photo- graphs of: a man; the same man standing on both sides of himself, resulting in three identical men; a man with three heads, and a man with an upside-down head. A variety of measures of attention were made: fixation time; direction of first fixation; number of vocali- zations; arm movements; pointing; smiling; changes in heart rate. In general, the children pointed more, moved their arms more, and looked first and longer at altered figures than at the unaltered figure. 42 In a well designed and executed experiment, Connolly (1969) has shown that kindergarten, second-grade, and fourth-grade children look longer at, and show longer and more intense changes in facial expression to altered drawings of animals than to unaltered drawings. These results were independent of age. All of the figures were presented as a picture book, and E recorded the interval between page-turns. This is a particularly powerful technique to use with children because it rules out the possibility that the children are not responding to the stimulus figures, but to the operation of the tachistosc0pe (Clapp and Eichorn) or slide projector Smock and Holt). It should be pointed out, however, that Murray and Brown (1967), using adults, and Faw and Nunnally (1968b), using children, have found no difference in looking time at altered figures between the typical laboratory setting and a less formal "waiting room" setting, in which the subjects leafed through a booklet containing the figures. Only one study shows no difference in exploration of altered versus unaltered figures (Hoats, Miller, and Spitz, 1963). Certain data analysis problems of this study have already been discussed. In view of these problems, the results are difficult to interpret. In summary, the results would seem to indicate that infants, preschool, and elementary school children, like adults, look at al- tered figures more than at their unaltered counterparts. The results do not show when in the child's growth and experience such differences in responding begin to take place, though in Connolly's study there were no age differences. This is a matter of some interest from a deve10pmenta1 standpoint, especially for the development of cognitive processes, as Charlesworth (1969) has pointed out. 43 The effect of incongruity upon ratings and :preferences" Few attempts have been made to obtain subjects' evaluations of altered familiar figures. As noted previously, Berlyne and Lawrence (1964) found that adults preferred the unaltered forms to their altered versions, and Berlyne (1963a) found that the altered figures were rated more interesting but the unaltered figures were rated more pleasing. When considered in light of Day's (1967) results, this finding would suggest that looking time is more a function of inter- estingness than pleasingness, whereas the Opposite may be the case for verbal evaluations. It also might mean that altered figures are more complex than unaltered figures. The work of Nunnally and Faw at Vanderbilt University (Faw and Nunnally, 1968a, b; Nunnally, Faw, and Bashford, 1969)is highly germane to this discussion. This work attempts to relate ratings of altered figures to looking time as measured by photography of eye fixations. The altered figures are described in Appendix A. The three studies by these investigators are similar in many respects, and only the latter will be examined in detail. The hypothesis under investigation was concerned with the shape of the function relating level of incongruity to looking time. Following the reasoning of Berlyne and others, one would predict a curvilinear relation, on the argument that incongruity elicits conflict and arousal for which individuals seek some optimum level. Nunnally, Faw, and Bashford suggest an alternative explanation, that longer looking time at altered figures is the result of a greater number of competing interpretations of those figures, and that a person looks until the interpretations are resolved. This suggests a monotonic 44 (increasing) relation between level of incongruity and looking time. The authors designed four sets of one unaltered and three altered figures each, so that each set would represent four distinct levels of incongruity. Twenty-seven female college students and 54 third- and fourth-grade pgys were shown all possible pairs of figures within each set, a total of 24 pairs. Each pair was shown for 30 seconds and S3' eye fixations were photographed. After all pairs were pre- sented, the subjects ranked the four figures in each set from most familiar (l) to most unusual (4). The movie frames were analyzed, and eye fixations scored as Left, Right, ar Center. The percentage of time for Left and Right fixations was calculated, with the "Center" category disregarded. A "mean % viewing time" score was thus computed for each figure, for each of the three 10-second periods of the 30-second exposure. For both adults and children, significant F-ratios were obtained for "levels of incongruity" and the "levels of incongruity X time intervals" interaction. A monotonic relation between viewing time and level of incongruity occurred during the first and second 10- second periods for the adults, and only during the first lO-second period for the children. The results seem to show fairly clearly that the more "unusual" an altered figure is ranked the greater time a person will look at that figure. There are several reasons, however, to consider the above inter- pretation tentatively. The first concerns the "incongruous" figures and their evaluation by the subjects. Examination of the description of the figures (Appendix A) reveals a considerable haphazardness in the design of these figures. It would be helpful if the authors 45 would spell out the intuitive basis on which the figures were drawn. Why were these figures chosen to be altered in these ways? More importantly, the subjects were asked to rank the figures, not on a dimension of "incongruity," but in terms of their unfamiliarity. The two terms may not be synonymous, that is, incongruity may in fact be something other than or in addition to novelty, as much of the literature seems to suggest. But this is a moot point. The figures were Egg, in fact, ranked on the basis of their familiarity. They were ranked from.most familiar to most unusual, pg£_most un- familiar. Do "unusual" and "unfamiliar" have the same meaning for both female college students and third- and fourth-grade boys? If they do not mean the same, then no clear conclusions can be drawn from the rankings, and about their relation to viewing time. It is likely, therefore, that the research is mistitled. It is concerned with "Effects of degrees of incongruity on visual fixations in children and adults" only according to the authors' unstated criteria for incongruity. The subjects were not asked about incongruity, however, so it is more likely concerned with the effects of familiar- ity-unusualness on fixations in children and adults. The second reason why Nunnally et al.'s interpretation must re- main tentative is the manner in which the viewing time was scored. As was mentioned previously, Center fixations were discarded. To use the authors' example, "if of the total 30-second viewing period, the more incongruous stimulus was looked at for 15 sec., while the less incongruous stimulus was looked at for 10 sec., then the percentage scores for these pictures would be 60% and 40% respectively." Note that they are not 50% and 33.3%. This perhaps mistakenly assumes 46 that "center" looks are meaningless. It would be worthwhile to see the results analyzed with the "center" fixations included, be- cause the percentage of "center" looks may vary considerably from pair to pair. The third source of concern is based on the statistical ana- lysis. The research was based on an hypothesis about the nature of the relation between level of incongruity and looking time, mono- tonic versus curvilinear. Such an hypothesis would seem to require a trend analysis, at least some consideration of nonlinear components of variance. Yet the reported analyses of variance involve only linear components. No F-Ratios are reported for quadratic or cubic components. Finally, the subjects differed in both age and sex. While the literature on visual exploration would lead us to expect few sex differences, the developmental literature is replete with sex X age interactions, for example, deve10pmental changes in "field-dependence" (Witkin, 1960). Thus, the results of this research are highly suggestive, but inconclusive. In sum, however, verbal evaluations of the complexity and famil- iarity of stimulus figures made to vary in their complexity, novelty and "incongruity" would seem to be related to visual exploration of these figures in an exploration or choice task. As noted previously, the meaning of those ratings and preferences in terms of the physio- logical effects of various parameters of stimulation is a matter for further investigation. 47 Definitions and theoretical treatments of incongruity The term "incongruity" has not been of major importance in any theory of motivation, or attention, or deve10pment. The term has most frequently been used in conjunction with the terms "novelty" and "complexity", and has represented, in effect, an acknowledgment that novelty and complexity are not the only non-homeostatic, non-drive- reducing determiners of motivation and attention. When an attempt is made to explain exploratory, investigatory, and play behaviors, the first stimulus determiners invoked are novelty and complexity. Given such a situation, it is not surprising that incongruity has remained poorly defined and imperfectly integrated into any theoretical for- mulation. The three terms, novelty, complexity, and incongruity have a common base, however, in the role of experience in establishing the levels of these stimulus dimensions which result in different moti- vational and attentional states. Novelty and incongruity are explicitly based on experience; complexity is less 30. Furthermore, the three have in common the concept of anticipation or expectation. Expectation is used in two senses: either as a cognitive event based on past ex- perience with concepts and classes of objects, or as a learned (habitual) response to some specific situation. In either case, some disruption of the usual response to that situation is experienced or evidenced. While Berlyne (1960) has been prominent in the use of the term incongruity, its definition and place in theory have been discussed also by Hebb (1949), Piaget (1952), Bruner and Postman (1949), Maddi (1961), and Nunnally, Faw, and Bashford (1969). Each of these will be 48 taken up briefly, but it must be emphasized that treatments of this stimulus dimension have been uniformly speculative, rational in approach. There has been no empirical treatment of ”incongruity." Berlyne (1960) regarded incongruity as one of a number of stim- ulus variables ("collative" variables) which generate arousal and curiosity through conflict. The drive state curiosity is reduced through exploration. Berlyne defined incongruity as existing when a stimulus induces an expectation which turns out to be disappointed by the accompanying stimuli. Incongruity requires not only a combination of stimuli that is novel but a combination differing from, yet having components in common with, one that the organism has learned to treat as more likely. We can think of incongruity if we insist, as a special case of surprise, since the incongruous pattern contradicts expectations aroused by the whole mass of past experience” (pp. 24-25). Thus, in Berlyne's terms, incongruous figures are of necessity novel, but something has been added such that the result is incongruity. This is a temporal distinction: novelty is the result of ”successive" expectations and incongruity is aroused by'simultaneous" expectations. This "simultaneous-successive” distinction is more explicit in his discussion of incongruity and surprise. In both cases the distinction is notably weak. No two events in the same sense modality are ever simultaneous in terms of the operation of the nervous system. The eyes, in addition, are in constant motion changing the point of fixation. All events are functionally successive. It is very dif- ficult and arbitrary to decide when two events occur in such close succession that they can be considered simultaneous. Hebb (1946, 1949) treated incongruity on two levels, 49 neurological and observational. The neurological explanation is in terms of disruption of established 'associations' of neurons, with resultant emotional behavior, arousal, and exploration. Specifically, through perceptual experience mediated by eye movements and selective attention, neurons become functionally associated with each other (their thresholds for firing are reduced). These functional 'struc- tures', which Hebb referred to as cell assemblies, can be combined with each other and with motor processes to result in still more complex functional 'structures' called phase sequences. A stimulus situation which results in partial firing of a cell assembly, but which contains an element or elements not normally there (not 'eXpected' by the succeeding neurons in the chain) disrupts the organized act- ivity of the cell assembly. This disruption then leads to exploratory activity. A stimulus situation giving rise to such disruption would be called incongruous. Thus, incongruity is here defined in terms of disorganized activity of hypothetical neural structures. This is a difficult preposition to test. The cell assemblies and phase sequences remain hypothetical, and, as Hebb points out, are extremely difficult to identify, due to their experientially-based complexity. Hebb's (1946) observations, along with those of KOhler (1927) and Valentine (1930), provide a variety of descriptions of familiar sit- uations or stimuli that had been changed, and provoked fear responses. Unfortunately, there is little commonality among the situations that Hebb, KOhler, and valentine describe. The head (as distinct from the rest of the body) seems to be involved in many of them, but by no means all. In general the situations were familiar, but with some elements changed. In terms of frequency, many of the elements of the situation 50 were highly frequent. Some, however, were infrequent, either in an absolute sense, or in the context of the frequent elements. This may be an important distinction. If the changed elements are infre- quent in an absolute sense, the situation may, in fact, be novel. Incongruity may be the case when the changed elements are infrequent 13 the context of the high frequency elements. This may be what Hebb all- uded to with "the familiar in an unfamiliar guise." There is reason for caution, however. Fear was the reaction ob- served by Hebb, KBhler, and Valentine. Approach and exploration are the responses of children and adults to incongruity, as shown by the literature reviewed in the preceding sections. The generality of Hebb's, KOhler's, and Valentine's observations therefore may be ques- tionable. Piaget (1952, pp. 68, 276) has noted that the most effective elicitor of attention and exploration in infants is a moderate alter- ation of some familiar stimulus. This could be taken to mean in- congruity, though Piaget does not use the term. His treatment of stimulus characteristics that influence exploration and investigation is in terms of novelty. It should be noted that Hebb's and Piaget's discussions of novelty-incongruity are similar: both point to the im- portance of prior experience to establish standards by which to deter- mine whether a situation is familiar or unfamiliar. Novelty (and perhaps incongruity) are important in Piaget's theory of cognitive development. The child's tendency to orient toward, approach, and eXplore novel objects is a major source of cognitive deve10pment (y. Flavell, 1963), because it leads to a constantly en- larging and reorganizing store of representations of stimulus 51 characteristics and the child's responses to them. As a result of this exploration, the child's cognitive structure is expanded and reor- ganized, thus directing his behavior differently. He therefore en- counters other new objects which result in further reorganization, thus making different behaviors possible, and so on. This source of cognitive deve10pment is seen most dramatically in play, which Piaget (1962) regards as the means by which the child exposes himself to, and thus gains more cognitive representations of, a wide range of objects and their characteristics. This treatment of cognitive change as generated by novelty has been enlarged by Charlesworth (1969) to include surprise. Charles- worth distinguishes between novelty and surprise, however, on a number of grounds, and points out that surprise, as a potent deter- miner of attention, may be an important source of cognitive development. Bruner and Postman (1949) and Bruner (1951) have explicitly defined incongruity in terms of expectancy. Perceivers, they main- tain, bring with them to any situation a general readiness to respond selectively, an expectancy or hypothesis. Stimulus information is compared with the existing hypothesis, and the hypothesis is confirmed or not. In the latter case the hypothesis is changed according to personal, idiosyncratic factors, and according to experience. New information is sought out by which to test the modified hypothesis. Incongruity exists "when perceptual expectancies fail of confirmation" (Bruner and Postman, 1949, p. 208). In this sense, incongruity is the outcome of a comparison, and the outcome could be influenced by any stimulus conditions failing to confirm the hypotheses. Thus, incongruity is not a dimension of stimulation but a state of the perceiver that 52 has motivating prOperties, such that: 1) the hypothesis is changed; 2) contradictory information is not attended to (”defended against"), or 3) further information is sought to reduce the mismatch between hypothesis and information. According to this formulation of incon- gruity, the terms 'novelty', 'change', hurprise', and 'incongruity' (as a dimension of stimulation) would all be synonymous, because they all could result in defeated expectations. Maddi (1961) deals more directly with incongruity as a stimulus variable. ”An incongruous stimulus contains elements which are per- ceived as incompatible with each other on the basis of prior experience with these elements" (p. 256). The definition resembles Berlyne's (1960). In both, expectations are thought to be aroused, but are disconfirmed. The result is a change in arousal or activation (though Berlyne prefers arousal to be mediated through conflict), which leads to exploration. Berlyne seems to refer to the stimulus as a whole arousing an expectation which is disconfirmed by some part or parts, whereas Maddi refers not to the whole stimulus but to in- compatibilities among its parts. Maddi continues: Novelty, complexity, incongruity, and surprisingness can be fairly well defined in the abstract, but when an attempt is made to vary one of them while holding the others constant, the distinctions become blurred. The more complex, incongruous, and surprising a stimulus is, the more likely it is to be novel. The more incongruous a stimulus is, the more likely it is to be complex" (1961, p. 267). The final treatment of incongruity to be considered is by Nunnally, Faw, and Bashford (1969). In this investigation, incon- gruity was not explicitly defined. An empirical definition might possibly be obtained by examination of the kinds of alterations made on the stimulus figures, but cannot, because (as previously noted) 53 the design of the figures was haphazard. The authors Speculate, however, on the mechanism by which incongruity affects visual ex- ploration. They hypothesize that, effects on looking behavior are because of the competing possible interpretations of the incongruous figure rather than because Of conflict-induced states of arousal. The incongruous figure will tend to dominate visual orienting behavior until the conflicting interpretations are re- solved" (p. 361). If 'interpretation' can be taken to mean 'expectation', then this explanation is similar to those preceding. It is not clear that this is the case, however. This seems to refer to a multiplicity of meanings, in a sense conceptual complexity, with the implication that a great number of 'part-meanings' makes more difficult (and there- fore more time consuming) an association of meaning to the figure as a whole. It is clear that incongruity has received little more than spec- ulative definition. There has been no empirical treatment of the term other than the observations of KOhler, Valentine, and Hebb. While there is considerable variability in the specifics of defini- tion, all reflect the importance of prior experience, and most refer to alteration of arousal or activation state by this and other related variables. SUMMARY OF RESEARCH ON VISUAL EXPLORATION AND "PREFERENCES” It seems clear that novelty, complexity, and incongruity, all variously defined, are aspects of the subject-environment relation— ship which have the power to effect shifts in attention and prohnnged visual exploration, which are paralleled by changes in various evaluations. However, if one considers the age of the 54 subject the result is less clear, especially with respect to the influence of complexity. One of the strengths of this research is that, despite different definitions of the variables, despite different stimuli meant to depict such variables, and despite the variety of measures of ex- ploration that have been used, the major results still obtain: that situations that are novel, complex, or incongruous are explored longer than are familiar, less complex, or "congruous" situations. This research has a major weakness also. These results may be peculiar to the stimuli that have been used, and not generalizable to other stimulus situations. This is a serious criticism and is especially applicable to the incongruity dimension. Berlyne and others following him have taken a simple intuitive approach to incongruity. There is nothing wrong with an intuitively-based Operational definition, at least in the early stages of scientific work, but to regard incongruity as a phenomenon as simple as its crude definition may be a serious mistake. This problem of defining stimulus characteristics of "incongruity" is paralleled by the difficulty in defining the response, that is, "preference" in many of the studies already reviewed. THE CONCEPT OF PREFERENCE A word is in order concerning the use of the term "preference.” A number of investigators have asked subjects whether they preferred or "liked best" a stimulus figure. Others have measured the duration that a figure was explored and inferred that it was preferred if it was explored longer than other figures. Either procedure alone is 55 of questionable utility because different "preferences" may be obtained with the two methods. Kaess and Weir (1968) present evidence that the method of measurement influenced the degree of preference for figures dif- fering in complexity. They presented pairs of randomly-generated polygons to 54 29- to 66-month-old children and obtained, on suc- cessive days and in counterbalanced order, 1) stated preference, i.e., which member of the pair they "liked best," and 2) looking time, the figure looked at longer being scored as preferred. The relations between the complexity of the figures and each of the two measures differed considerably. Stated preference showed little change as number of turns increased from 5 to 40, but looking time increased monotonically. At the complexity level of 5 turns, the scaled score was higher for stated preference than for looking time, but at 40 turns the reverse was true. Age was not a contributing factor. The authors are apprOpriately tentative about stated preferences obtained from three- to five-year-olds, but present some evidence (split-half reliabilities from .45 to .65) that the children were not responding randomly. Comparison of the evaluation and looking time methods is not quite so clear as Kaess and Weir's study would indicate, however, because age is also a factor. Thomas (1966b) found no difference in results between the two measures, using six- to 12-year-old children. Berlyne and Lawrence (1964) used adult subjects and found a negative correlation between evaluation and looking time. It is difficult to say which of the stimuli in Kaess and Weir's study were "preferred," because the "preference" depended on the 56 method of measuring it. The confusing results already summarized in the section on "preference" for complexity may be explainable in terms of these different response measures. Experimenters have been understandably reluctant to ask young children which figures they liked best, and so "preference" has been based on looking time. Contrariwise, older children have been counted on to understand instructions and give trustworthy verbal responses, and their judg- ments of "preference" have tended to be based on such verbal eval- uations. Thus, age differences in "preference" have clearly been confounded with mode of response. There are further objections to the indiscriminate use of the term ”preference." Knowing that a figure is preferred over other figures (however measured) tells little about the subject's response to that figure. A more meaningful question is, What is the under- lying basis for the longer looking time and verbal evaluation? or, What are more specific responses that the subject makes to the stimuli? There seems to be an encouraging trend in this literature away from global "preference" or "liking" ratings toward more specific questions, for example, how interesting or pleasing the stimuli are, although these questions too are difficult to inter- pret. " have Furthermore, investigators interested in "preference virtually ignored organismic variables as a source of explanation of choice behavior, in favor of explanation in terms of stimulus characteristics. The only such organismic variable to receive any consideration has been age, and as was remarked above, age has been confounded with method of response. The possibility for systematic 57 individual variations in responsiveness and quality of response exists, whether or not we choose to use the term 'personality'. Individual differences in style of perception and cognition have been explored to some advantage by Witkin (1960; Witkin, et al., 1954) and by Kagan, Moss, and Sigel (1963). There is no reason why such a research approach could not help explicate the concept of preference. There is one further danger. That is the error of reification, that the name arbitrarily assigned to a class of behavior comes to be regarded as a £3123, a process or structure within the organism. To say that a subject prefers one thing to another is to describe his behavior. To say that he (therefore) has a_preference is to impute to him some unknown, perhaps unnecessary, and perhaps overly simplistic process or entity. Such an inference is unparsimonious and easily leads to circular reasoning. An analogy with the concept of instinct is compelling. Schnierla (1966) maintained that "'innateness' is nothing more than a poor hypothesis as to alleged types of ontogenetic origins" (p. 287). To say that a behavior is instinctive is to describe it according to some criteria, primarily, species-specificity and independence of learning. "Instinctive" is a description in the same sense that "preferred" describes choice behavior, either verbal or visual. However, Schnierla points out that it has been relatively easy to slip from "instinctive” to "instinct," implying that there exists some entity intervening between genetic mechanisms and behavior. In like fachion, a slip from "preferred" to "preference" may be just as easy and just as potentially misleading. Unexplained behaviors 58 must be described nevertheless, and "instinctive" and "preferred" or "preference behavior" are perfectly adequate labels, but they are only descriptive, not explanatory. THE PROBLEM A number of problems have been raised in this review. The current research focuses on a basic question: What does "incongruity" mean? That is, what kinds of figures, altered in what kinds of ways, are called ”incongruous?", and how do "incongruity" ratings relate to other possible ways of evaluating these same figures? As Appendix A shows, no systematic operations have been per- formed on figures that are called "incongruous,” or by others (Smock and Holt, 1962; Faw and Nunnally, 1967, 1968a, b; Nunnally, Faw, and Bashford, 1969), simply "novel." In fact, the only common- ality among the figures is that they are all altered forms of meaning- ful figures, the alteration often being one of substitution of various parts. In research of this type, in which subjects are exposed to various qualities of stimulation and in which the experimenter is called upon to devise stimuli which 13 £335 depict those variables, interpretation of the results depends intimately on the stimulus figures and the response measure used. It behooves the investigator to show that the effects of the manipulated stimulus variables are not specific to the stimuli used, and that the stimuli are a sample of all stimuli that would depict those variables. The aim of this research is essentially a clarification of the meaning of "incongruity" in terms of: the manipulations or oper- ations performed on stimuli that yield "incongruity," and verbal 59 evaluations on a number of other dimensions which can be expected to be related to "incongruity." Only after such an initial step to understand "incongruity" can this complex aSpect of stimulation be used in a meaningful way to investigate other complex classes of behavior. METHOD Subjects The subjects were 240 college students (147 female, 93 male) enrolled in the introductory and child psychology courses at Michigan State University during the summer term, 1969. The mean age of the group was 20.7 years. Table 1 lists the number of subjects, male and female, in each group and in each order of presentation. Stimulus figures The stimulus figures chosen in earlier studies to represent "incongruity" are diverse and unsystematically sampled, and they may not be representative of a general factor "incongruity," assuming for the moment that such a factor exists. The stimuli described in Appendix A are based on each previous experimenter's intuitive ideas about incongruity. There is nothing wrong with an intuitive approach, but evidence suggests that it is time to become more specific about what is meant by "incongruity." To this end the author has listed in Appendix B a variety of factors which may be thought to make an object or a two-object situation incongruous. For simplicity's sake, the factors are listed singly rather than in combinations or multiple interactions. This conceptualization, like previous unstated ones, is rational and in- tuitive, but also more systematic and comprehensive. It is by no means exhaustive, yet it serves to illustrate that the concept of incongruity is potentially far more complex than the figures used to date would indicate. We note too in comparing Appendices A and B 60 61 OON NOH NO ONH OO ON ONH NO NO Hmuoa OO ON OH ON OH O ON O NH mmmcHOHuuHHHcoo OO ON NH ON NH N ON OH OH NOHHOHHHEOHOO OO ON OH ON OH O ON O HH mmmawaHmmmHN OO ON OH ON OH O ON OH OH mwmamcHummumuaH OO ON OH ON OH O ON OH OH NOHNOHOEOO Os ON OH ON OH O ON OH OH NOHONmaooaH HeuoH OHNEOM mam: HmuoH OHoaom was: HmuoH oHoaom OHmz N Hoouo H Hoouo macaw Hmuoa Nouuo muomnnsm .H OHOOH 62 that the majority of incongruous figures used in past studies have involved only one operation, "wrong part.” This outline, besides exposing the conceptual complexity of incongruity, allows the construction of a series of stimulus figures which vary in specified ways. The figures used in this research are described in Appendix C. Two major factors were involved in the construction of these figures: Class of object (Living versus Non- living) and Manipulations performed on body part. The latter is comprised of: 1) Missing part; 2) Misplaced part; 3) Wrong part; 4) MiSplaced and Wrong part. Another potentially important variable is the part manipulated (head or leg). In figures used in previous studies, either has been manipulated. Therefore, in the current study all manipulations were made on both head and leg of the living figures. Such was not possible for the nonliving figures, since these figures, while having parts similar to legs, have nothing like heads. Thus, the Class factor has three values: 1) Living figures with Head manipulated; 2) Living figures with Leg manipulated, and 3) Nonliving figures with part manipulated. With considerable effort these diverse figures in Appendix C were made as much alike as their differences would allow. For in- stance, with the exception of the man, ape, and umbrella, the figures are oriented toward the right, either full-right (car, bicycle, and elephant) or approximately half-right, to allow all legs to be seen (dog, eagle, table, and chair). In all cases, the leg (or wheel) missing is in the right half of the drawing. When misplaced, the leg (or wheel) is moved left, toward the back of the figure. When a part is wrong, in both the living and nonliving figures, it is made the 63 size of the correct part (the ape's head on the elephant, the eagle's leg on the man, the chair seat on the car). When the head of the living figure is misplaced, it is moved down the body, whether that be gravitationally down (man, ape), or left (elephant, dog, eagle). Finally, the limb manipulated always implied dis- equilibrium, a loss or change of support. A number of potentially important factors were varied randomly or left uncontrolled in these figures. One is the angle of view of the figure; face front; profile, or half front. To make clear that a part was missing or misplaced and not merely obscured or a function of the perspective of the figure, it was necessary to vary the perspective. Eye contact is related to perspective and is es- pecially important for the human figure, but is also potentially effective in the other living figures. Another factor is whether the entire part is missing, misplaced, or wrong, or whether only part of the part is manipulated. Another is the symmetry of the figure. Yet another is whether any action is portrayed. Finally, the prOportional size of the "wrong" parts could contribute to a figure's judged incongruity. For instance, when an ape's head is put on an elephant's body, should the head be the size of an ape's head or an elephant's? These factors could be controlled, of course. Eye contact could be controlled by constructing four basic drawings: 1) man standing face front, facing front; 2) man standing face front, but looking to the side (head in profile); 3) man standing in profile, but looking face front; 4) man standing in profile and looking straight ahead (head in profile). Each of the four manipulations then could be 64 performed on the heads of each of the above four figures. Each manipulation (with the exception of "missing”) would be performed in two ways: with the manipulated head in profile, that is, looking away from the observer; and, with the manipulated head face front, looking at the observer. Similar controls could be exerted on other variables, though to allow generalization over figures and to keep the task reasonably short, such control was not exerted in the present study. The 70 figures listed in Appendix C were drawn in india ink on 9 X 12 inch layout paper. The unaltered versions of the figures were drawn first, and as many of the altered versions as possible were then drawn by tracing the unaltered versions. The 70 figures were drawn approximately the same size and have some inner detail, though they are basically outline drawings. The pen-and-ink drawings were photographed on Kodachrome X film and processed as slides. Because the lighting conditions were not Optimal, the projected slides had a definite yellow-brown tint. The slides were not true black on white, though they were uniform in this fault. The 60 slides of altered figures plus the 10 slides of unaltered figures were randomly ordered and placed in Kodak Carousel slide trays, alternating with black cardboard to darken the screen between ex- posures. The 10 unaltered figures were included in order to get estimates of their own incongruity, unfamiliarity, or conflictfulness as a base against which to interpret the ratings of the altered figures. It is possible that including the unaltered figures with the altered figures influenced the ratings of the altered figures, but this was 65 not considered to be a serious problem, since the differences between various classes of figures are unlikely to have been affected, and it is these differences in ratings, not their absolute values, that are important. Two orders of presentation were used. Order 1 was generated according to the following: The series of 70 eXposures was par- titioned into ten 7-digit sequences. From each of these sequences a number 1 through 7 was randomly chosen, and one of the ten ran- domly ordered unaltered figures was assigned to that position in the series. Thus, an unaltered figure appeared in every seven exposures, although any two were not allowed to occur in succession. The remaining 60 positions were filled by the altered figures, which were randomly assigned with the following restrictions: 1) no two alterations of the same basic figure (e.g., man) were presented in succession; 2) no manipulation was presented twice in succession, and 3) no more than three Living or three Nonliving figures were pre- sented in succession. Order 2 was the reverse of order 1. Ratingiscales The dimensions used in the rating scales were 1) incongruity 2) complexity 3) interestingness 4) pleasingness 5) unfamiliarity 6) conflictfulness. Incongruity is the variable of major interest, especially as respects its relations with the other five variables. As was suggested 66 earlier, "incongruous" figures are also unfamiliar, and differential reactions to incongruity (longer looking times, "preference") in fact may be reactions to novelty (V. Smock and Holt, 1962; Faw and Nunnally, 1967, 1968a; Nunnally, Faw, and Bashford, 1969). A similar case has been made for complexity (Maddi, 1961; Cantor, 1963), to the extent that "incongruous" figures are also more complex than their ”congruous" counterparts, and in view of the similarities of response to higher versus lower complexity and altered versus unaltered figures. The research of Berlyne (1963a; Berlyne, Ogilvie, and Parham, 1968), Cantor (1968), and, in particular, Day (1967) indicate that interestingness and pleasingness are related mono- tonically and curvilinearly, respectively, to complexity. Thus a consideration of complexity should include those variables. Finally, Berlyne (1957, 1960, 1963b) has named conflict as the major factor in any explanation of differential responding to var- ious forms of stimulation. The notion of conflict, whether arising through competing responses or physiological processes, seems im- plicit in the theorizing of Festinger (1957), Hebb (1949, 1955), and Hunt (1965). It is reasonable to expect that these six concepts, represented by the six rating scales, will be interrelated. Procedure The subjects were given one of the six instruction-answer booklets in the classroom. The instructions, read aloud by the experimenter, were as follows: Please state your age and sex in the spaces provided on the answer sheet. This is an experiment to discover how you react to various aspects of pictures. A number of figures will be projected on the screen at the front of the room. After you have looked 67 at each figure, please rate that figure by marking an X in one of the 7 spaces provided on the answer sheet. Just before each new figure appears, I'll say "Next picture" so be sure you are looking at the screen. You'll have 6 seconds to look. Then I'll turn the picture off, and you'll have 7 seconds to rate it on your answer sheet. There are 70 figures in all, so this will take approximately 17 minutes. There are a number of ways that you might react to each figure,... After reading this far, E asked the subjects to continue reading on their own, "because there are several sets of instructions." The subjects were also advised that the person sitting next to them was unlikely to have the same instructions. The instructions for each of the six groups were: 1) Incongruity: There are a number of ways that you might react to each figure, but your particular task is to rate the figure on how incopgruous it appears to you. Think of 'incongruous' as meaning that the parts don't seem to fit, or the figure doesn't seem to go along with your previous experience. If the figure appears to you to be very incongruous, you should give it a rating of "6" or "7". If it seems to you to be not at all, or just barely incongruous, you should give it a rating of "l" or "2". The 7-point scale is: not at all or barely moderately very incongruous incongruous incongruous 1 2 3 4 5 6 7 2) Complexity: There are a number of ways that you might react to each figure, but your particular task is to rate the figure on how complex it appears to you. Think of 'complex' as meaning that the figure is complicated, made up of many parts, perhaps difficult to understand. If the figure appears to you to be very complex, you should give it a rating of "6" or "7". If it seems to you to be not at all, or just barely complex, you should give it a rating of "1” or "2”. The 7-point scale is: 68 ‘not at all or barely moderately very complex complex complex 1 2 3 4 5 6 7 3) Interestingness: There are a number of ways that you might react to each figure, but your particular task is to rate the figure on how interesting it appears to you. Think of 'interesting' as meaning that the figure arouses your curiosity, attracts and holds your attention. If it appears to you to be very interesting, you should give it a rating of "6" or "7". If it seems to you to be not at all, or just barely inter- esting, you should give it a rating of "1” or "2". The 7-point scale is: not at all or barely moderately very interesting interesting interesting 1 2 3 4 5 6 7 4) Pleasingness: There are a number of ways that you might react to each figure, but your particular task is to rate the figure on how pleasing it appears to you. Think of 'pleasing' as meaning that the figure is satisfying, enjoyable, gives you a good feeling. If the figure appears to you to be very pleasing, you should give it a rating of "6" or "7". If it seems to you to be not at all, or just barely pleasing, you should give it a rating of "1" or "2". The 7-point scale is: not at all or barely moderately very pleasing pleasing pleasing l 2 3 4 5 6 7 5) Unfamiliarity: There are a number of ways that you might react to each figure, but your particular task is to rate the figure on how unfamiliar it appears to you. Think of 'unfamiliar' as meaning that the figure is out of your experience, unknown, or strange. If the figure appears to you to be very unfamiliar, you should give it a rating of "6" or "7". If it seems to you to be not at all, or just barely unfamiliar, you should give it a rating of "1" or "2". The 7-point scale is: 69 not at all or barely moderately very unfamiliar unfamiliar unfamiliar l 2 3 4 5 6 7 6) Conflictfulness: There are a number of ways that you might react to each figure, but you particular task is to rate how much you feel some conflict about that figure. That is, you may feel that you want to get closer to the figure or look at it, but at the same time you want to move farther away from it or look away. Such a situation can be thought of as conflictful. If the figure appears to you to be very conflictful, you should give it a rating of "6" or "7". If it seems to you to be not at all, or just barely conflictful, you should give it a rating of "1” or "2”. The 7-point scale is: not at all or barely moderately very conflictful conflictful conflictful l 2 3 4 5 6 7 All sets of instructions concluded with: Try to be thinking about how you would rate the figure while you're looking at it. Then, when the figure goes off, mark an X in the space on the answer sheet that best describes your reaction to that figure, in terms of how (incongruous, complex, interesting, pleasing, unfamiliar, conflictful) it is. There are no right or wrong answers; this research is concerned with your reaction to the figures. Remember, mark an X in one of the spaces "1" through "7", whatever you think applies best. When all of the figures have been shown, please make sure that you have marked your age and sex on the answer sheet. The instructions "Think of 'incongruous', 'complex', interesting', " were derived from con- 'pleasing', and 'unfamiliar' as meaning ... tent analysis in a previous pilot study, of 29 college students' most frequent responses to the instruction, ”Write in one or two short sentences what this word means to you." The present instruc- tions were elaborated, that is the terms were "defined" by several synonyms, for two reasons. First, the range of responses from the 29 students was large, suggesting that any particular person's 70 interpretation of these terms was likely to be general and some- what vague. The scales were expected to overlap, but the overlap should be attributable to common characteristics among the terms, not to misinterpretations of the terms. Essentially, then, the purpose was to reduce the variance within each of the six groups of subjects. Second, these 'definitions' were intended to allow the subjects to attend more to the figures as they were presented instead of thinking about the various meanings of the term by which they were to evaluate the figures. The slides were exposed by a Kodak Carousel 750 projector located 22 feet from a projection screen. The projected image was 26 X 34 inches in dimension. EXposure time was measured by a hand-held stOpwatch. The 120 subjects given order 1 made their ratings during a single lecture class. Ratings obtained under order 2 were obtained in one large class and four smaller classes. In summary, six groups of subjects (N = 40 per group; Order 1, n = 20; Order 2, n = 20) rated each of the 70 figures. Each subject rated each figure once on only one dimension. The resulting factorial design for each scale is 2 X 3 X 4, with factors: Order (2); Class of figure (3: Nonliving (NL); Living- Head manipulated (L-H); Living-Leg manipulated (L-L)), and Manipulation performed on part (4: Missing (M); Misplaced (MP); Wrong (W); Misplaced and Wrong (MP&W)). Subjects were nested within order, with repeated measures within subjects. The basic datum for the analyses of variance was the mean score per subject for the five figures which made up each of the 12 C X M combinations. The ratings 71 of the unaltered figures were not included in the analyses of variance. RESULTS The answer sheets were checked for mistakes, and the ratings punched on IBM cards for computer analysis. A total of 290 sets of ratings were collected. Of these, 19 were unscorable, usually because of two ratings for one figure and resulting confusion later in the list. The number of unscorable answer sheets in each group was: Incongruity -- 3; Complexity -- 3; Interestingness -- 5; Pleasingness -- 0; Unfamiliarity -- 2; Conflictfulness -- 6. 0f the remaining 271 scorable answer sheets, 31 were randomly discarded, so that 40 subjects would remain per group with roughly equal propor- tions of males and females in each group. LEVEL OF SIGNIFICANCE In the analyses of variance and related analyses reported below, a large number of significance tests have been performed on the same data. Since it is difficult to specify the true alpha level under these conditions (y. Hays, 1963), type I error was minimized by letting alpha = .01. Differences significant at alpha levels greater than .01 should be interpreted cautiously. QNFOUNDING VARIABLES The first step in the analysis was to determine whether the ratings were influenced by any of several potentially confounding variables. Those examined were: 1) inconsistency of the ratings, that is, whether the ratings were randomly distributed, either with- in or between scales; 2) a serial position effect across the 70 72 73 presentations, and 3) sex of the subject. 1 (a). Consistency of ratings within each scale. The aim of this analysis was to determine whether the subjects' ratings were sufficiently consistent (nonrandom) on each scale to make further analysis meaningful. Two-way analyses of variance were performed on the data of each of the six groups, and coefficients of reliability were calculated (Winer, 1962, pp. 126-128). They were: Incongruity, 0.94; Complexity, 0.91; Interestingness, 0.69; Pleasingness, 0.91; Unfamiliarity, 0.91; Conflictfulness, 0.86. According to this method the reliability of the ratings on each scale increases as the variance of the ratings between figures grows increasingly larger than the variance of the ratings within figures. 1 (b). Ratings compared across scales. The ratings were com- pared across all six scales. Nonrandomly disrtibuted ratings would indicate that the scales measured different things. At the least it would show that subjects were not all responding on the same (un- known) basis, that is, that the instructions had some effect. Accordingly, a four-way analysis of variance (summarized in Table 2) was performed, with factors: Groups (6); Order (2); Manipulations (4), and Class of figure (3). The "Groups" factor represents the six rating scales. In this analysis of variance, as in all subsequent ones, appropriate degrees of freedom, expected values of mean squares, and error terms for F ratios were determined according to the rules outlined by Millman and Glass (1967). The effect of Order, Manipulation, and Class will be examined in greater detail when each of the six scales is discussed separately. For the present, it is sufficient to note that the Group effect was 74 Table 2. Analysis of variance over all scales Source df MS F p Group (G) 5 754.887 58.12 (.001 Order (0) 1 228.488 17.59 (.001 G X 0 5 55.766 4.29 (.001 Subjects within GXO (S:GXO) 228 12.988 Manipulation (M) 3 101.775 231.05 <.OOl G X M 15 9.480 21.52 (.001 O X M 3 3.572 8.11 (.001 G X 0 X M 15 0.561 1.27 ’.10 (NS) S X M: G X 0 684 0.440 Class (C) 2 135.190 141.45 <.001 G X C 10 20.307 21.25 <.001 0 X C 2 1.004 1.05 >.10 (NS) G X 0 X C 10 0.810 4 1.00 NS S X C: G X 0 456 0.956 M X C 6 2.401 10.07 (.001 G X M X C 30 1.566 6.57 <.OOl O X M X C 6 0.483 2.02 >.05 (NS) G X 0 X M X C 30 0.856 3.59 (.001 S X M X C: G X 0 1368 0.238 Total 2879 3.246 75 significant (F=58.12, df=5, 228, p<3001), but interacted with Order, Manipulation, and Class. On the basis of the significant Group effect, it can be concluded that the different instructions had significantly different effects on the ratings. 2. Serial position effect. Before major analyses proceed, it must also be determined whether there was a serial position I ’- ___1-/ effect, that is, whether there were any systematic changes in the ratings as a function of figure position in the series of 70 pre- sentations. For example, habituation, boredom, or systematic changes .— -o-r—Vz'. in context could have influenced the ratings. [v_ To test for such an effect, an Order of presentation X Block analysis of variance was performed on the ratings on each scale in- dependently. The blocked factor consisted of 5 levels: the mean rating given to the first group of 14 figures presented, through the mean for the fifth group of 14 figures. For each subject a mean was calculated of the ratings of the first 14 figures (whatever figures they were), for the second 14, and so on. These five means for each subject were the basic data in the analysis of variance. The five levels of the blocked factor were crossed with the two orders of presentation. Six such analyses were performed, one for each scale, and are summarized in Table 3. Three scales showed significant (p<.01) differences among the five means: Incongruity (F=6.28, df=4, 152, p(.001); Interestingness (F=7.85, df=4, 152, p mo mommeam oumumoom New mo humaESm .o OHQMH Table 7. Analysis of variance of incongruity ratings Source df MS F p Order (0) 1 39.675 4.42 <.05 Subjects within Order (3:0) 38 8.982 Manipulation (M) 3 41.521 102.94 (.001 O X M 3 1.190 2.95 (.05 S X M:0 114 0.403 Class (C) 2 68.405 86.68 “.001 O X C 2 0.042 ‘1.00 NS 8 X C:O 76 0.789 M X C 6 2.486 10.18 <.001 O X M X C 6 0.662 2.71 «(.025 S X M X C:O 228 0.244 Total 479 1.726 83 for those figures (M) was significantly lower than the means for each of the other three manipulations. The Misplaced figures (MP) and Wrong figures (W) did not differ significantly. The Class effect can be better understood if we consider the mean rating for each class of figure (Table 9). The living figures with the head manipulated (L-H) were rated most incongruous. That mean (L-H) differed significantly from the mean for the nonliving figures (NL), and from the mean of nonliving figures (NL) and living figures with leg manipulated (L-L). The nonliving figures were rated significantly lower than either of the other two classes. The two classes of living figures (L-H vs. L-L) did not differ significantly. The Manipulation X Class interaction can be better understood by examining the mean rating for each category of figure on each scale (Table 10). It is clear that within each manipulation rated incon- gruity varied with the class of figure, and within each class the rating varied with the manipulation. The highest incongruity ratings were made to the living figures with the head misplaced and wrong (L-H, MP&W), the lowest to the nonliving figures with a part missing (NL, M). (b) Complexity. The analysis of variance of the complexity ratings (Table 11) disclosed significant effects only of Manipulation (F=67.55, df=3, 114, p(.001), Class (F=55.09, df=2, 76, p<.001), and Manipulation X Class (F=7.25, df=6, 228, p<.001). Tests of contrasts revealed that the significant Manipulation effect stemmed largely from the significantly lower rated complexity of the Missing figures (M) (Table 8), which differed from each of the 84 Table 8. Mean ratings for each manipulation Scale Manipulation Incon. Compl. Int. Pleas. Unfam. Confl. Missing (M) 4.56 2.92 2.55 2.02 4.05 3.25 MiSplaced (MP) 5.53 3.87 3.18 1.94 5.13 3.83 Wrong (W) 5.42 3.76 3.24 2.07 5.16 3.71 Misplaced and Wrong (MP&W) 5.96 4.17 3.14 1.72 5.46 4.00 Table 9. Mean ratings for each class of figure Scale Class Incon. Compl. Int. Pleas. Unfam. Confl. Living-Head (L-H) 5.99 4.07 3.18 1.80 5.44 4.12 Living-Leg (L-L) 5.43 3.89 3.16 1.80 4.90 4.04 Nonliving (NL) 4.69 3.08 2.74 2.22 4.52 2.94 85 Table 10. Mean ratings for twelve categories of figures Scale Category Incon. Compl. Int. Pleas. Unfam. Confl. L-H, M 5.43 3.12 2.46 1.69 4.96 3.58 L-L, M 4.33 3.15 2.76 1.97 3.86 3.58 NL, M 3.91 2.50 2.42 2.42 3.32 2.60 L-H, MP 6.12 4.26 3.35 1.74 5.32 4.21 L-L, MP 5.78 4.27 3.24 1.82 5.29 4.26 NL, MP 4.70 3.10 2.95 2.26 4.79 3.01 L-H, W 5.94 4.14 3.52 2.20 5.56 4.15 L-L, W 5.35 3.82 3.30 1.74 4.96 4.01 NL, W 4.98 3.31 2.90 2.28 4.97 2.97 L-H, MP&W 6.48 4.76 3.39 1.55 5.92 4.52 L-L, MP&W 6.24 4.33 3.33 1.66 5.48 4.31 NL, MP&W 5.16 3.42 2.70 1.94 4.98 3.16 86 Table 11. Analysis of variance of complexity ratings Source df MS F p Order (0) 1 0.290 < 1.00 NS Subjects within Order (8:0) 38 12.088 Manipulation (M) 3 34.026 67.55 ‘<.001 O X M 3 0.619 1.23 >410 (NS) 3 x M:O ' 114 0.504 Class (C) 2 44.623 55.09 <2001 O X C 2 1.867 2.30 >.10 (NS) S X C:O 76 0.810 M X C 6 1.611 7.25 (.001 O X M X C 6 0.408 1.84, >.10 (NS) S X M X C:O 228 0.222 Total 479 1.750 87 other three manipulations. The Misplaced and Wrong figures (MP&W) were rated most complex, but the mean rating of these figures dif- fered only from the Missing (M) figures, and from the mean of Missing, Misplaced, and Wrong figures. The Misplaced (MP) figures did not differ from the Wrong (W) figures. The Class effect (Table 9) is attributable largely to the sig- nificantly lower rated complexity for Nonliving (NL) figures than for either Living figures with head manipulated (L-H) or Living figures with leg manipulated (L-L). The latter two were not significantly different from each other. Comparison of the incongruity and complexity ratings shows that Manipulation and Class had virtually the same effect on each scale. Table 10 shows that the effect of either Manipulation or Class depended on the level of the other. While Nonliving (NL) figures as a group were rated low in complexity, the rating varied depending on the manipulation performed. Thus, the Nonliving figures with part Misplaced and Wrong (NL, MP&W) were rated more complex than Living figures with the head Missing (L-H, M), while the L-H figures in general received the highest complexity ratings of the three classes. (c) Interestingness. The factors Manipulation (F=31.l9, df=3, 114, p(.001), Class (F=8.06, df=2, 76, p<.001), and Manipulation X Class (F=3.84, df=6, 228, p<.005) were significant contributors to differences in interestingness ratings (Table 12). Order and all interactions but 0 X M X C (p<.05) were not significant. The effect of the different manipulations on rated interesting- ness is shown in Table 8. The Missing (M) figures were rated least ‘; ’1‘! ar~—_—_-sfl'vnl.- . ‘n . 88 Table 12. Analysis of variance of interestingness ratings Source df MS F p Order (0) 1 1.240 "1.00 NS Subjects within Order (3:0) 38 14.942 Manipulation OH) 3 12.393 31.19 ‘<.001 O X M 3 0.236 ( 1.00 NS S X M:O 114 0.397 Class (C) 2 9.545 8.06 (.001 O X C 2 0.340 .05 (NS) S X M:O 114 0.274 Class (C) 2 9.833 14.15 4 .001 O X C 2 1.629 2.34- >.10 (NS) S X C:O 76 0.695 M X C 6 1.218 7.44 (.001 O X M_X C 6 1.007 6.15 (.001 S X M X C:O 228 0.164 Total 479 0.812 .r- Li‘s-“CV .10 'r v I . 91 Table 14. Analysis of variance of unfamiliarity ratings Source df MS F p Order (0) 1 385.208 23.93 ‘(.001 Subjects within Order (8:0) 38 16.099 Manipulation (M) 3 45.965 72.30 < .001 O X M 3 1.845 2.90 (.05 S X M20 114 0.636 Class (C) 2 34.479 59.45 ‘<.001 O X C 2 0.961 1.66 >.10 (NS) 8 X C:O 76 0.580 M X C 6 3.547 12.44 (.001 O X M X C 6 1.527 5.36 (.001 S X M X C10 228 0.285 Total 479 2.971 _. _.a._.._..._..l ._._.1. ‘-.—V n- Aun~v‘. I 92 (F=72.30, df=3, 114, p(.001), and Class (F=59.45, df=2, 76, p(.001) were significant main effects. The interactions of Manipulation X Class (F=12.44, df=6, 228, p(.001) and Order X Manipulation X Class (F=5.36, df=6, 228, p(.001) were also significant. The Order X Manipulation interaction was significant with p(.05. The Order X Class interaction was not significant. The Order effect arose because the figures when presented in order 2 were rated less unfamiliar (Xe4.04) than when presented in order 1 (X=5.85) . As Table 8 shows, figures whose parts were MiSplaced and Wrong (MP&W) were rated most unfamiliar, though this mean differed signifi- cantly only from the Missing (M) figures, and from the mean of Missing (M), Misplaced (MP), and Wrong (W) figures. The Missing 0%) figures were rated least unfamiliar, and that mean differed signifi- cantly from each of the other three. NO other pairwise comparisons were significant. This pattern of Manipulation means is the same as occurred for the incongruity and complexity ratings. The significant Class effect can be attributed to the Living figures with head manipulated (L-H) (Table 9). They were rated significantly more unfamiliar than either the Living figures with leg, manipulated (L-L) or Nonliving (NL) figures. The L-L versus NL comparison was not significant. The effects of Class and Manipulation were dependent on each other, as Table 10 shows. While Nonliving figures (NL) were generally low in unfamiliarity, Nonliving figures with a part Misplaced and Wrong (NL, MP&W) were rated about as unfamiliar as Living figures with the head Missing (L-H, M). The Order X Manipulation X Class interaction 93 indicates that the above effect is also influenced by the order of presentation, that is, the ratings for any M X C combination were lower in order 2 than in order 1. (f) Conflictfulness. Analysis of the conflictfulness ratings (Table 15) showed that significant effects were Manipulation (F=28.54, df=3, 114, p(.001), Class (F=41.67, df=2, 76, p<.001), and Order X Manipulation (F=4.22, df=3, 114, p(.Ol). Order and all other interactions were not significant. The means for each manipulation are shown in Table 8. The Missing figures (M) were rated significantly less conflictful than any of the other three manipulations. While the Misplaced and Wrong figures (MP&W) were rated most conflictful, that mean differed significantly only from the Missing (M) figures. The MP&W figures, furthermore, were not significantly more conflictful than the mean for the other three manipulations. The Class effect is indicated in Table 9. The Nonliving (NL) figures were rated significantly less conflictful than either Living figures with head manipulated (L-H) or Living figures with leg mani- pulated (L-L). The L-H versus L-L comparison was not significant. It should be noted that the pattern of Manipulation and Class means for the conflictfulness ratings is similar to the pattern for incongruity, complexity, and unfamiliarity ratings. The Order X Manipulation interaction reflects higher conflictful- ness ratings to the four manipulations when presented in order 1 ($4.11) than in order 2 (£3.29). 94 Table 15. Analysis of variance of conflictfulness ratings Source df MS F p Order (0) 1 80.688 4.03 >.05 (NS) Subjects within Order (8:0) 38 20.024 Manipulation (M) 3 12.244 28.54 ‘.001 O X M 3 1.811 4.22 (.01 S X M:O 114 0.429 Class (C) 2 69.841 41.67 ".001 O X C 2 0.217 ‘<1400 NS 8 X C:O 76 1.676 M X C 6 0.342 1.38 >.10 (NS) 0 X M X C 6 0.494 2.00 >.05 (NS) 8 X M X 0:0 228 0.247 Total 479 2.634 95 CORRELATIONS AMONG THE RATINGS What was the relation between each figure's rated incongruity and its ratings on the other five scales? The ideal way to answer this question would be to have each subject rate each figure six times, once on each scale. Given the number of figures used, how- ever, such procedure would be likely to result in highly contaminated ratings. The data were analyzed as follows. Each of the 60 altered figures had six ratings associated with it, one for each scale, each being the mean of 40 83' ratings. Thus six measures were associated with each of 60 figures. Pearson product-moment coefficients were computed for each pairwise comparison of the rating scales. The correlation of each scale with the others is shown in Table 16. Incongruity ratings correlated positively with rated complexity, interestingness, conflictfulness, and highest with unfamiliarity (r=0.866). Pleasingness ratings were negatively correlated with in- congruity ratings (r=-0.645), and with all other ratings. With only one exception (the correlation of pleasingness with interestingness ratings), the obtained correlation coefficients are significant (by t-test) beyond the .002 level. However, the true alpha level for these 15 t-tests is not known, because only 3 of the correlation coef- ficients can be based on completely independent samples. COMPARISON OF INCONGRUITY AND UNFAMILIARITY RATINGS It was stated in the introduction that one way to explain the effects of incongruity is to assume that incongruous figures are really novel figures, and that it is novelty that attracts and holds 96 Table 16. Intercorrelations of the six scales Scale Scale Incon. Compl. Int. Pleas. Unfam. Confl. Incongruity 1.000 Complexity 0.817 1.000 Interestingness 0.660 0.780 1.000 Pleasingness -0.645 -0.546 -0.237 1.000 Unfamiliarity 0.866 0.718 0.601 -0.512 1.000 Conflictfulness 0.802 0.829 0.667 -0.727 0.674 1.000 97 .attention. Contrarily, incongruity may incorporate novelty but add sonma<3ften important element to it. Ideally, this could be determined through analysis of covariance, subtracting the effect of one factor (novelty) from the other (incongruity), and observing whether any differences remain. Unfortunately, such an analysis was not possible in the current study, because no subject rated any figure more than once. That is, there are not two or more scores associated with the same subject. An analysis using the 60 stimulus figures as 'subjects' is not possible, either, because while each figure does have six scores associated with it, the figures were systematically assigned, not randomly, to the Manipulation and Class conditions. Instead, a Groups (Incongruity, Unfamiliarity) X Order X Class X Manipulation analysis of variance was performed. This analysis is similar to that presented in Table 2, but it compares two groups rather than all six. The results are summarized in Table 17. The Groups factor is relatively meaningless, for reasons previously discussed. It was argued, however, that if the psychological var- iables underlying the incongruity and unfamiliarity ratings are differ- ent, then the same factor would influence them differently. That is, if "incongruity" and "unfamiliarity" represent different processes, those processes may be differentially influenced by the Class and Man- ipulation variables. Therefore, "Groups" would interact with those variables, and the Group X Class, Group X Manipulation, and Group X Class X Manipulation interactions would significantly affect the ratings. As Table 17 shows, the G X‘M interaction was not signif- icant (F=1.52, df=3,228, p>.10), but the G X C (F=5.32, df=2,152, p<.Ol) and G X M X C (F=4.20, df=6,456, p(.001) interactions were 98 Table 17. Analysis of variance of incongruity-unfamiliarity Source df MS F p Group (G) 1 41.834 3.34 >.05 (NS) Order (0) 1 336.067 26.80 (.001 G X 0 1 88.817 7.08 (.025 Subjects within GXO (S:GXO) 76 12.540 Manipulation (M) 3 86.695 166.72 (.001 G X M 3 0.791 1.52 >.lO (NS) 0 X M 3 2.985 5.74 ‘(.001 G X 0 X M 3 0.050 < 1.00 NS 8 X M: G X 0 228 0.520 Class (C) 2 99.237 144.87 ‘<.OOl G X C 2 3.647 5.32 (.01 O X C 2 0.575 < 1.00 NS G X 0 X C 2 0.428 (1.00 NS 8 X C: G X 0 152 0.685 M X C 6 4.919 18.56 (.001 G X M X C 6 1.114 4.20 <.OOl O X M X C 6 1.794 6.77 4.001 G X 0 X M X C 6 0.394 1.49 >.10 (NS) 8 X M X C: G X 0 456 0.265 Total 959 2.390 99 significant. The prediction is partly supported. The ratings of incongruity versus unfamiliarity depended upon the class of figure, and also upon the Manipulation X Class category. The mean ratings for the Group X Class interaction are shown in Table 18. Ppgg‘ppg comparisons (alpha=.01) were performed on these means according to the methods suggested by Marascuilo and Levin (1970). These authors point out that the usual pppp p93 methods applied to interaction means lead to unclear interpretations. Each of the means in Table 18 can be partitioned into 1) the grand mean; 2) the effect of Group; 3) the effect of Class; 4) the joint effect of Group and Class, and 5) error. If the usual pppp p92 methods were performed on any two of the means and the difference was shown to be significant, it could not be determined whether the difference was due to the in- teraction, or to either of the main effects. Marascuilo and Levin propose, therefore, that each mean be trans- formed into an interaction term by subtracting the two main effects and the grand mean from each cell mean. These interaction terms can then be compared according to Scheffé's method. Such an analysis was performed on the means in Table 18. It was found that no single interaction term contributed significantly to the over-all F ratio. However, comparison of pairs of interaction terms revealed that the only significant difference was between the interaction term in the Incongruity-Nonliving mean and the interaction term in the Unfamiliarity-Nonliving mean. Thus the Group X Class F ratio is attributable to the different effect of Nonliving figures upon incongruity and unfamiliarity ratings. The Nonliving figures were Table 18. Mean ratings for Group X 100 Class interaction Class Group L-H L-L NL Incongruity 5.99 5.43 4.69 Unfamiliarity 5.44 4.90 4.52 Table 19. Mean ratings, Group X Manipulation X Class Class Manipulation Group L-H L-L NL Missing Incon. 5.43 4.33 3.91 (M) Unfam. 4.96 3 86 3 32 Misplaced Incon. 6.12 5 78 4.70 (MP) Unfam. 5.32 5 29 4.79 Wrong Incon. 5.94 5.35 4.98 (W) Unfam. 5.56 4.96 4 97 Misplaced and Wrong Incon. 6.48 6.24 5.16 (MP&W) Unfam. 5.92 5.48 4 98 101 rated only slightly less unfamiliar than the Living figures, but the Nonliving figures were rated considerably less incongruous than the Living figures. The Group X Manipulation X Class means are shown in Table 19. Inspection suggests that, again, the interaction can be attributed to the different effects of the Nonliving figures upon the manipula- tions across the two sets of ratings. DISCUSSION Before turning to the major question of how the results help to explicate the concept of incongruity, we must consider certain methodological questions bearing on the interpretation of the rating data. Were the ratings consistent? Were there any effects of serial position, sex of subject, or presentation order? How were the un- altered figures rated? Consistency of the ratings The ratings proved to be fairly consistent on five of the six scales, with coefficients ranging from 0.86 (conflictfulness) to 0.94 (incongruity). The interestingness ratings were the least consistent (0.69). It is not clear whether this lower coefficient is an effect of sampling error, or an indication of greater indivi- dual differences among college students in what they consider to be interesting rather than incongruous, complex, pleasing, unfamiliar, or conflictful. Perhaps "interesting" is the most variably-defined, that is, the most idiosnycratic of these six terms. With the possible exception of the interestingness ratings, therefore, the ratings were sufficiently consistent as to be trustworthy. The significant F ratio for Groups (Table 2) indicates that the basis for the consistency was the instructions given to the six groups of subjects. Serial position The position effect in the series of 70 presentations was re- flected in higher ratings of incongruity and interestingness for the first 14 figures. After that (arbitrarily determined) point in the 102 103 series there was little systematic change in those ratings. Serial position influenced pleasingness ratings, but less systematically, and did not influence rated complexity, unfamiliarity, or conflict- fulness. This suggests that boredom, habituation, or progressive changes in judgmental context did not affect the ratings beyond the fourteenth figure. Such changes would be expected to occur early in the series as the subjects became familiar with the stimulus figures and with the task they were to carry out, and as they began to anticipate subsequent figures. It was informally observed, how— ever, that during the latter half of the series the subjects rated the figures more quickly and became more active in general, perhaps indicating that they were restless and bored. If this was the case, the ratings were not markedly sensitive to the effect. That there was little systematic change in the ratings contra- dicts the finding of nearly all previous studies of looking time, that looking time to all figures decreases as the stimulus series progresses. Perhaps the discrepancy is due to the large number of presentations (70) used in the current study compared with the much smaller number (10-30) usually used in studies of looking time. Over the first 10 to 30 figures in the present research, there was an effect of position in the series, at least on two of the six scales. In any event, the tasks in previous studies and in the current study (to look versus to rate) were different in a number of ways, which may account for the different results. It is noteworthy, however, that the unfamiliarity ratings were not affected by serial position. Ordinarily, the effect of unfamiliarity diminishes rather quickly as the number of presentations increases 104 (Berlyne and Parham, 1968). This did not happen with the present ratings, suggesting that the subjects were judging unfamiliarity on a long-term, personal life history basis, rather than on the short- term basis defined strictly by the presentations in the experiment. §e__x The only relation between the sex of the subject and ratings was on the complexity scale, with p<105. Previous studies of re- actions to complexity, novelty, and incongruity have rarely included sex as a factor, and rarely found it to be important when included. Only Smock and Holt (1962) and Leckart, Briggs, and Kirk (1968) have reported sex differences in looking time. The unaltered figures The unaltered figures (Table 5) were rated much less incongruous, complex, interesting, unfamiliar, and conflictful than were the altered figures (Tables Dl-D6). The unaltered figures were rated more pleasing, however. These results hold for all of the unaltered figures except the ape. The ape was rated one to two units higher than the other unaltered figures, i.e., more like the altered figures, on five of the six scales, and lower on the sixth (pleasingness). As will be shown shortly, manipulations performed on the ape resulted in some of the highest incongruity ratings. These ratings are difficult to interpret, however, because of the moderate incongruity of the unaltered ape figure. Order of presentation Order of presentation of the 60 altered figures influenced only the unfamiliarity ratings (Table 6), and so need not be considered a major factor in the interpretation of the effects of Manipulation and Class of figure. Unfamiliarity ratings were lower (p