Novelty, complexity, and hedonic value

Attention, Perception, & Psychophysics, Sep 1970

Two experiments, in which Ss were exposed to sequences of colored shapes, investigated effects on ratings of “pleasingness” and “interestingness” of variables that had previously been shown to affect ratings of “novelty.” The results indicate, on the whole, that both pleasingness and interestingness increase with novelty. These findings run counter to those of experiments indicating an inverse relation between novelty and verbally expressed preference. Two further experiments examined effects of some variables that might account for this apparent discrepancy. Homogeneous sequences declined in judged “pleasantness” more than sequences in which several stimuli were interspersed, and simple stimuli became less pleasant as they became less novel, while complex stimuli declined less or became more pleasant. The findings are related to hypotheses regarding mechanisms of hedonic value. Two crucial predictions were confirmed in a fifth experiment.

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

http://link.springer.com/content/pdf/10.3758%2FBF03212593.pdf

Novelty, complexity, and hedonic value

I 0 I' 0 0 Perception & Psychophysics , 1970, Vol. 8 (SA) Two experiment.~. in which Ss were exposed to sequences of colored shapes. investigated effects on ratings of "pleasingness" and "interestingness" of variables that had previously been shown to affect ratings of "novelty. " The results indicate. on the whole. that both pleasingness and interestingness increase with novelty. These findings nm counter to those of experiments indicating an inverse relation between novelty and verbally expressed preference. Two further experiments examined effects of some variables that might account for this apparen t discrepancy. Homogeneous sequences declined in judged "pleasantness" more than sequences in which several stimuli were interspersed, and simple stimuli became less pleasant as they became less novel, while complex stimuli declined less or became more pleasant. The findings are related to hypotheses regarding mechanisms of hedonic value. Two crucial predictiollS were confirmed in a fifth experiment. - Two previous experiments (Berlyne & Parham, 1968) investigated determinants of subjective novelty and revealed several factors that can cause a simple. nonrepresentational, visual stimulus (a colored patch of irregular shape) to be rated more or less novel. The experiments to be reported in this paper studied effects of some of the same factors on rated "pleasingness" and "interestingness." Novelty is one of the "collative" stimulus properties whose importance for motivation theory is becoming more and more apparent, especially in such areas of research as exploratory behavior, experimental aesthetics, developmental psychology, and personality study (Berlyne, 1960, 1963a, 1966). There is now a sizable body of experimental literature (e.g., Berlyne, 1963b; Munsinger & Kessen, 1964; Day, 1965; Berlyne, OgilVie, & Parham, 1968) on verbal judgments applied to visual patterns of differing complexity, another collative property. Determinants of judged complexity have been identified, and the influence of complexity on judgments of pleasingness (liking, preference) and of interestingness have been studied. The extension of this kind of investigation to novelty seems warranted, and two of the experiments to be reported studied how novelty and complexity interact in determining how pleasing a visual pattern will be rated. The present state of knowledge with regard to effects of novelty on preference is rather puzzling. Zajonc (1968) reports some experiments of his own, and reviews a fair number of experiments by others, that indicate an inverse relation between novelty and verbally expressed preference. On the other hand, Cantor (1968) reports an experiment with children in which verbally expressed preference increased with novelty. Certainly, novelty tends in many conditions to increase the probability or duration of self-ex posure to a . stimulus object or manipulation of it, which many would regard as a manifestation of preference or liking. In everyday life, there are times when something is more attractive than something else because it is more familiar, just as there are times when something is found particularly appealing because it is novel. Some resolution of these apparently divergent findings is obviously called for and was one of the objectives of the present investigation. Stimulus Material In both experiments, colored patches of irregular shape on a white background were presented in six classes of sequences, and every S went through two sequences belonging to different classes in turn. The sequences were selected from those used in the two experiments reported in the earlier article (Berlyne & Parham, 1968), and incorporated variables that those experiments showed to have significant effects on judged novelty. The sequence classes can be represented as in Table I. In Sequence Classes I and 2, X differed from Y in color only, whereas X differed from Y in both color and shape in Sequence Classes l' and 2'. 'Stimuli denoted by different letters differed in both color and shape in Sequence Classes 3 and 4. Experimental Design (see Table 2) The Ss of each experiment were divided into eight equal groups, containing as far as possible equal numbers of males and females. The key to the notation can be found in Tables I and 3 of the previous article (Berlyne & Parham, 1968). Groups 5 and 6 underwent the prefamiJiarization (F) treatment, whereas the remaining groups did not. Pre familiarization consisted of exposing Ss, at the outset of the experiment, to a slide showing all the stimuli to be used in the experiment and asking them to look at the slide until they felt "reasonably well acquainted with the range of pictures shown here" and able to recognize them if they saw them again. "Subscripts attached to the same letter represent successive appearances of the same stimulu~ 5 (F) and 7 (NF) 6 (F) and 8 (NF) .. For key, see Tables 1 and 3, Berlyne & Parham, .1968 F ; with prefamiliarization; NF = without prefamiliarizatlOn Discussion It will be seen that not all the variables that were found to influence judged novelty significantly in the previous investigation had significant effects on judged pleasingness or interestingness. There was even a hint that interestingness may reach a maximum with a certain degree of familiarity in the unexpected finding of a significantly higher mean rating (Experiment II) in Ss who had had the prefami1iarization treatment than in those who had not (XL, Sequences 3 and 4) had this treatment. Nevertheless, the results tend strongly to favor one of the two divergent views mentioned earlier over the other: Both pleasingness and interestingness appear to increase with novelty. There is a marked general resemblance between the graphs in Fig. I and those representing mean novelty ratings that were presented in the previous article (Berlyne & Parham, 1968). In particular, pleasingness and interestingness, like subjective novelty, decline with prolonged repetition of a particular stimulus, and they both rise temporarily when a monotonous sequence is interrupted by a single appearance of a con trasting stimulus. In experiments on complexity (Berlyne, 1963; Day, 1965, 1967, 1968), conditions have been found in which relatively high pleasingness coincides with relatively low interestingness, namely when complexity is particularly low. But as far as novelty is concerned, nothing of this sort seems to occur. Pleasingness and interestingness seem generally to go together, although the curves in Fig. I suggest that interestingness may decline faster. EXPERIMENT III Aims The results of Experiments I and II, like those of Cantor's (1968) experiment, indicate a direct relation between novelty and hedonic value, whereas the experiments reported and reviewed by Zajonc (1968) suggest an inverse relation. So we must look for differences in material or procedure that might account for these discrepant outcomes. Three possibilities suggest themselves, and some attempt was made in Experiment III to investigate all of them: 22222< 22222 2222!.'j f - - - - - - - - - f - . , ,~c~~1 ,~~6789~11~~~~~17~ 2'.~'rI9~1"21)~~~~ "'fQlI3lQll1i1Olts~lItJiIOlt.1i1O"gfORJ~ ~~~~~~~I~~~~~~~~~ I IDM'glOlII5R3IQ"'ltJlt!liIOfQlOl'I:JfQ!} 'U'"t:!.1'I1lll'N'lljltlIt!ItIt'lI"""tlll"'!! I I I I ," I " r' I I 1"3 2 22 2 2 Z 2: 2 2 :z 2 2 Z % ZI! , 4 ~ I'; I' I 9oQII12rJM~t'17YJ . ,. :5 6 r I ') 10 I1I2I3Me" I EXP II (INTERESTINGNESS) 3F'~~~: , , 3F , 2' ,4~>I- ~ - """"-'----- - - - I !G2~:?<:.1GO'~G'E~>G2::~'i.'C>.!~.<~;~ ?~!!GH02ijZlN!G2G2G2G2G2G2G2G2!! , .!5 6 , (I ') 10, f2 '3MI!II6'7!} 2 1" !.I 16 r e 9 I() I 12 '3 t4 ~ Ifi!! 2 , Fig. l. Experiments I and II: Mean ratings. (1) The stimuli used in Experiments I and II, like the black-and-white patterns from the Welsh Figure Preference Test used by Cantor, were rather simple, whereas those used in the experiments performed and cited by Zajonc (word!t, paralogs, Chinese ideograms, photographs of faces) were rather complex. If stimuli sometimes become less appealing and sometimes become more appealing as they become familiar, this could be explained by the interaction of two antagonistic processes, of which sometimes one prevails and sometimes the other. As a stimulus is repeated or prolonged, its appeal may gradually succumb to a "tedium" factor. Preference may, on the other hand, rise with repetition or prolongation through the action of a "positive-habituation" factor. It may reflect increased liking for a pattern as infonnation is absorbed from it and it becomes assimilated or organized. Subjects Forty-eight students, for the most part not students of psychology or of fine arts, were recruited in the corridors of the University of Paris-Nanterre. They participated in the experiment in parties of two to six. Stimulus Material Four nonrepresentational (N) patterns were taken from a set (Berlyne, 1963b) that has been used over the years in a number of experiments studying the effects of complexity variables on First or Second Appearance of X NumbcrofYs Before first X Reccncyof Prcvious X Degree of Change from Y to X Homogeneous or Heterogeneous Sequence Before X Prefamiliarization 7.02 1,36 < .025 17.76 1,24 <.001 4.97 1,36 <.05 4.48 1,24 < .05 A. Orthogonal Dyadic Comparisons Pleasingness (Experiment I) Interestingness (Experiment II) B. Linear Components of Trends Pleasingness (Experiment I) Overall Decline Overall Decline 4.23 1,24 -0.5 Interestingness (Experiment 11) Interestingness (Experiment 11) Distribu tion (Withou t Ties) Rubens's La Kermesse. The remammg items were designated by the number 2 and used for the heterogeneous sequence. They consisted of the less complex member of the fourth pair in Category A, the more complex member of the second pair in Category XA, Raeburn's Portrait of a Man and Rubens's Massacre of the Innocents. Experimental Design and Procedure Ss were divided into eight groups of six. Groups 1-4 went through a homogeneous sequence followed by the heterogeneous sequence, while Groups 58 had the heterogeneous sequence before the homogeneous sequence. The homogeneous sequence comprised 10 successive presentations of Item NS t for Groups 1 and 5, of NC 1 for Groups 2 and 6, of RS 1 for Groups 3 and 7, and of RC 1 for Groups 4 and 8. The heterogeneous sequence was the same for all groups. It consisted of 10 presentations each of Items NS2, NC2, RS2, and RC2, randomly intermingled with the restriction that every item appear once in each consecutive set of four. Each presentation consisted of a 4-sec projection on a screen, and there was a 4sec interval between consecutive presentations. An interval of 1 min 30 sec intervened between the two phases, and there was a break of 1 min between the first and the second halves of the heterogeneous sequence to allow the slide magazine to be changed. S had in front of him a booklet, each page of which bore a horizontal line divided into seven compartments, with the word "Desagreable" (unpleasant) to its left and the word "Agreable" (pleasant) on the right. The scale thus resembled those used for the semantic differential (Osgood, Suci, & Tannenbaum, 1957). After each presentation, S had to mark one of the seven compartments of the appropriate scale. He had been told that the scale went from "extremely unpleasant" to "extremely pleasant" and that the center represented the point of indifference "that is to say, neither pleasant nor unpleasant." Results Separate analyses of variance were carried out on the ratings for the representational-complex (RC) items were homogeneous sequence and heterogeneous crowded canvases replete with a multitude sequence. The results are shown in Fig. 2. of human figures and other details. Homogeneous sequence. The N items One NS, one NC, one RS, and one RC were rated higher than the R items-means item were selected by tosses of a coin to be = 3.7, 2.8, F(I ,40) =6.11, P < .025. This designated by the number 1 and to be unexpected finding is peripheral to the elements of the homogeneous sequences. main objectives of the experiment. It is They consisted ofless complex members of presumably connected with the fact that the third pair in Category C and the more contemporary taste in painting, complex member of the first pair in particularly in intellectual circles, favors Category XC (Figs. 1 and 2, Berlyne, the nonrepresentational. And such forms 1963b), Ramsay's Norman McLeod, and of representational painting as are now in First 12 of (1) and 0') First 15 of (3 F) and (3NF) First 15 of (4F) and (4NF) exploratory behavior and other processes. Two simple (NS) items and two complex (NC) items represented, respectively, the lower and upper reaches of the subjectivecomplexity continuum, as scaling experiments (Day, 1965; Berlyne, Ogilvie, & Parham, 1968) have confirmed. The four representational (R) patterns were blackandwhite reproductions of paintings. The representationalsimple (RS) items were portraits of single figures on a plain background, and the Overall Decline Steeper Decline in (3F, 3NF) than in (4F, 4NF) Xl > Preceding Y Xl> Succeeding Y XI, > Preceding Y Xl > Preceding Y Xl> Succeeding Y XL> Preceding Y XL> Preceding Y C. Sign Tests Pleasingness (Experiment I) Distribution (Without Ties) HOMOGENEOUS SEQUENCE HETEIlOGENOUS SEQUENCE A-_~~=~-~--o-...o NC~ PRESENTATIONS Fig. 2. Experiment III: Mean ratinp. vogue (e.g., Pop Art, Nouvelle Figuration) are certainly very different from our Baroque and Neoclassical R items. A trend analysis showed the linear component of the overall declining trend to be significant: F(I,432)= 52.51, P < .001. The linear component was steeper for N items than f'" R items-F( 1,432) = 4.49, P < .0:' ~nd steeper for S items than f~,r C items-F( 1,432) = 11.61, p < ,005. The declines were, however, significant for the N items alone (F = 42.94, P < .001), the R items alone (F = 13.03, P < .ool), the S items alone (F = 55.81, P < .001), and the C items alone (F = 7.25, P < .01). None of the comparisons or interactions involving the phase difference (whether the homogeneous series occurred first or second) reached significance. Heterogeneous sequence. In the heterogeneous sequence, the overall mean rating was once again higher for N items than for R items-means = 4.23, 3.24; F( 1,1728) =481.44, P < .00 I-and higher for C items than for S items-means = 4.03, 3.44; F = 173.82, P < .001. There was, however, a significant interaction between these two dichotomies: F = 97.32, P < .001. It turned out that the NC item was rated significantly higher than the NS item-means = 4.7, 3.7; F = 26.56, P < .00l-but the means for the RC and RS items (3.3, 3.2) were close together. Turning to the trend analysis, we find both the overall linear component-F(I,1794) = 31.21, p < .00 I-and the quadratic component-F = 9.29, P < .005-to be significant. The linear component was significantly steeper for the S items than for the C items-F = 10.60, P < .005. It was also steeper in Phase II (i.e., in groups having the heterogeneous sequence second)-F = 15.35, P < .001. There was, however, a significant triple interaction involving the R/N, SIC, and phase variables-F = 12.91, P < .001. For this reason, the results for the two phases are plotted separately in Fig. 2. Inspection of the curves shows that the triple interaction resulted from the greater dissimilarity of trends in Phase I than in Phase II. So the data for this phase were subjected to a separate trend analysis. In Groups 58, which had the heterogeneous sequence first, the overall linear component was not significant. The interaction between the linear component and the R/N and SIC variables was significant-F(I,864) = 9.65, P < .005. Data for the Rand N items were therefore examined separately. The linear components for the RS and RC items did not differ significantly, and there was a significant decline over both R items taken together-F = 7.70, P < .01. Linear components differed significantly for the NS and NC items-F = 29.22, p < .00 I. The NS item produced a significant decline -F = 10.01, P < .005, while the NC items produced a rise with a significant linear component-F = 19.20, P < .00 I. Discussion Our hypothesis predicting a decrease in the hedonic value of simpler patterns and an increase in that of more complex patterns as repetition reduces novelty receives partial confirmation. This is precisely what happened with the N patterns of the heterogeneous sequence in Phase I. And in both sequences, there was a significantly steeper decline with the S stimuli than with the C stimuli. Nevertheless, ratings of both Sand C stimuli declined in the heterogeneous sequence when it came second and in the homogeneou s sequence. We must, therefore, suppose that the homogeneity-heterogeneity variable played some part and that monotony favors the tedium factor. Furthermore, Ss were apparently more susceptible to the tedium factor during the second phase of the experiment. This is shown by the fact that ratings declined for all stimuli of the heterogeneous sequence in Phase H, including the NC stimulus that produced a rise in Phase I. The N/R variable also proved to hilVe some influence, but it took the opposite form to that suggested by the observation that experiments indicating an inverse relation between novelty and hedonic value have used meaningful, symbolic stimuli, while those indicating a direct relation have used meaningless stimuli. In both the homogeneous and the heterogeneous sequences, the representational stimuli made for a steeper decline than the nonrepresentational stimuli. EXPERIMENT IV Aims The results of the three experiments reported so far certainly cast doubt on the validity for all kinds of stimulus material of Zajonc's conclusion that hedonic value increases with familiarity. There is, however, one more factor that must be examined. In Zajonc's experiments, Shad simply to look at the stimulus patterns during the familiarization phase and did not have to record judgments until a later phase. In our experiments, a rating had to be made after every presentation. This might very well have made the procedure more laborious and therefore more tedious. If no ratings had been required during familiarization, it is possible that positive habituation would have come to the fore, producing a rise in preference even for the simpler stimuli as they became more familiar. Experiment IV was carried out to check this possibility. The necessity for a judgment after every representation might have affected the ratings by making Ss pay more attention to the stimuli. To see if this was an operative factor, half of the Ss in Experiment IV were given instructions designed to induce close attention, while the other half were not, but repeated judgments were required of neither group during familiarization. Subjects Forty-eight students were recruited as in Experiment HI. They all went through the heterogeneous sequence alone and had to rate stimuli only during the first four presentations of the sequence (dUring which each of the four stimuli appeared once) and during the last four presentations (during which again each stimulus appeared once). After the first four presentations had been concluded, 24 Ss, comprising the attention-instruction (AI) group, were told that they would see the pictures again a number of times and that they should look at them closely because they would later be questioned about them. The next 32 presentations (8 of each stimulus) then followed, and Ss were then instructed about the second set of ratings just before the last four presentations were given. The 24 Ss of the no-attention instruction (NAI) group were simply told after the first four presentations that they would see the stimuli again a number of times. Results Neither the main effect nor any of the interactions involving the presence or interaction involving these variables and the experiments variable are of in terest in view of the different results that were obtained with Nand R patterns when the heterogeneous sequence came first in Experiment III. However, neither of these interactions reached significance. GENERAL DISCUSSION Fig. 5. Experiment V: Mean ratings. further acquaintance with the pattern will provide scope for absorption of additional information and for the perceptual and ideational processing that enables uncertainty and conflict to be resolved as elements are discriminated, classified, recognized, and grouped together as subwholes. This can evidently be a source of pleasure, presumably dependent on the arousal-reduction mechanism. However, the arousal-increase mechanism to which the Wundt curve is relevant is also likely to come into play. A stimulus that is high in both novelty and complexity will be high in arousal value and thus correspond to a point on the horizontal axis in Region C. As it lost its novelty, the point representing its arousal potential would shift towards Region B accompanied by an increase in hedonic value. Skaife's (1967) findings with auditory stimulus material are consonant with ours and open to a similar interpretation. She used musical sequences varying in what she called "complexity," but, since it involved varying degrees of deviation from normal melodies, it might better be designated "surprisingness" or "uncertainty." Whatever name is applied to it, it can be recognized as a constituent of arousal potential. When repeated over 20 days, low"complexity" sequences that were initially rated pleasant gradually lost their pleasantness, whereas high"complexity" sequences that were initially given low ratings underwent a significan t rise in judged pleasantness. Our interpetation in terms of the Wundt curve implies that, if familiarization had proceeded further, ratings of the simple patterns would have continued to decline, whereas those of the complex patterns would have climbed to a peak and then dropped. Skaife (1967) recorded such a rise followed by a fall with some of her musical sequences and some of her Ss. So did Alpert (I953) with repetitions of an unfamiliar rhythmic sound pattern. It seemed desirable, however, to verify these predictions with our visual material. Experiment V was undertaken for this purpose. EXPERIMENT V Subjects Forty undergraduates from elementary psychology classes at the University of Toronto took part, two at a time. Procedure The procedure was the same as for Experiment IV, except that it was prolonged. S went through six tests, during each of which the four patterns were presented in a random order and had to be rated on a 7-point Osgood scale from "displeasing" to "pleasing." Between each test and the next, the four patterns were each presented eight times in the random order that had been used between the initial and final presentations in Experiment IV, but no judgments were recorded. Results The overall mean was higher for the R patterns (4.4) than for the N patterns (3.8): F( I,897) = 30.12, p < .00 I. This was opposite to the preference recorded in Experiment m. This difference might have reflected differences in taste between French and Canadian students. It might, on the other hand, have been due to the fact that the projector used in Toronto (a Kodak Carousel) produced brighter images than the projector used in Paris (a Liesegang nonautomatic), since the N slides consisted mostly of white background. There was also a significant interaction between the N/R and CIS dichotomies-F(I,897) = 15.26, P < .001. The means for the SN, SR, CN, and CR patterns were, respectively, 3.5, 4.6, 4.0, and 4.2. To turn to the comparisons that are germane to the chief purpose of the experiment, the Tests by CiS interaction was significant-F(5,897) =2.65, p < .025. But neither the Tests by N/R nor the Tests by CiS by N/R interaction approached significance. Consequently, curves for the two S patterns and the two C patterns are presented in Fig. 5. Ferguson's (1965) nonparametric trend analysis was used. Since the hypotheses under investigation implied a decline for S patterns and a rise followed by a decline for C patterns, one-tailed p values were relevant. With regard to C patterns, there was significant bitonicity-z = 2.19, p < .015. The monotonic and tritonic components were not significant. As for the S patterns, there was both a significant monotonic component-z = 4.30, Discussio11 The wo predictions from the interpret:- cion in terms of the Wundt curve thus received confirmation, and this explanation of the positive-habituation and tedium factors is therefore tenable. The C curve reached a maximum at the third test. After that, it dropped slowly and erratically. This may have been because rewarding effects of perceptual processing carne into play. The bitonic component in the predominantly downward trend of the S curve seems to represent a flattening out that is compatible with the shape of the Wundt curve (see Fig.4 and Berlyne, 1967), if we make the assumption that arousal potential drops linearly as number of presentations increases. The rise in the S curve over the last three tests was far from significant-z =0.52. Since the S curve sank only slightly below the indifference point (4.0), boredom seems to have played no more than a minor role. and JJ were collected by Jerry Colglazier, those of Experiments III and IV by Lydie Boudet, and those of Experiment V by Mary Louise King and Kathryn Davies. The author-is indebted to them for their abTe collaboration. Experiments Ifl and IV were carried out in the Laboratory of Experimental Psychology of the University of ParisN3Ilterre. thanks to the cooperation of Professor R-. "'ranees. 2. Address: Department of Psychology, University of Toronto, Toronto. Ontario. Canada.


This is a preview of a remote PDF: http://link.springer.com/content/pdf/10.3758%2FBF03212593.pdf

D. E. Berlyne. Novelty, complexity, and hedonic value, Attention, Perception, & Psychophysics, 1970, 279-286, DOI: 10.3758/BF03212593