Impact of Stimulus Variables on Exploratory Behavior

Stephen A. Goodwin, State University of New York at Buffalo
ABSTRACT - This paper examines the impact of three collative stimulus properties (blurredness, complexity, and incompleteness) on the attentional responses of consumers. The collative properties were each operationalized at three quantitative levels (low, intermediate and high) and embedded in a print ad for a new brand of a frequently consumed convenience good. The major proposition tested is derived from Berlyne's (1963) Theory of Specific Exploratory Behavior, and states that attention is an inverted U-shaped function of collative property intensity. The results of the 3 x 3x 3 factorial experiment conducted provide partial support for Berlyne's proposition. The study also provides a perspective which might be useful to the advertising practitioner interested in developing copy designed to capture and hold the audience's attention.
[ to cite ]:
Stephen A. Goodwin (1980) ,"Impact of Stimulus Variables on Exploratory Behavior", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 264-269.

Advances in Consumer Research Volume 7, 1980     Pages 264-269


Stephen A. Goodwin, State University of New York at Buffalo


This paper examines the impact of three collative stimulus properties (blurredness, complexity, and incompleteness) on the attentional responses of consumers. The collative properties were each operationalized at three quantitative levels (low, intermediate and high) and embedded in a print ad for a new brand of a frequently consumed convenience good. The major proposition tested is derived from Berlyne's (1963) Theory of Specific Exploratory Behavior, and states that attention is an inverted U-shaped function of collative property intensity. The results of the 3 x 3x 3 factorial experiment conducted provide partial support for Berlyne's proposition. The study also provides a perspective which might be useful to the advertising practitioner interested in developing copy designed to capture and hold the audience's attention.


The general area of novelty-seeking behavior has received little notice by consumer researchers. Recent articles in the JCR (Faison, 1977 and Rogers, 1979) provide brief updates on the nature and status of novelty-seeking, demonstrating the potential of this theoretical area in helping those interested in consumer research to probe and further their understanding of such topics as innovation proneness, brand switching, and receptivity to mass media advertising.

While novelty-seeking behavior has been of theoretical and empirical interest at least since 1954 (c.f., Berlyne, 1954, 1960, 1963, 1965, 1968; Goodwin, 1977; Maddi, 1968; Morrison and Dainoff, 1972), formal frameworks have only recently been developed in the area of Consumer Behavior which are capable of providing the necessary reference point for systematic empirical research (c.f., Hansen, 1972; Howard and Sheth, 1969; Venkatesan, 1974). These particular frameworks have all relied, to a great degree, upon the seminal theoretical position and empirical tradition of D. E. Berlyne.

Berlyne's work in this area is substantial, spanning the past 25 years; unfortunately, obvious space restrictions here preclude a thorough summary of his theory. (The interested reader may refer to Venkatesan (1974) for a terse review, or to Goodwin (1977) for a thorough summary). Nonetheless, since Berlyne's propositions form the theoretical backbone of the study to be described in this paper, a very brief overview will be attempted.


At the outset, it is very important to note that Berlyne's theory accommodates two major types of exploratory behaviors (also referred to as novelty-seeking behaviors): specific and diversive.

Specific exploration is behavior aimed at, and reinforced by, the prolongation or intensification of stimulation from particular sources. It is exemplified by the behavior of an animal that sees something novel in the distance and advances to give it a close examination, or by that of a man who turns his head and moves his eyes to fixate somebody who addresses him.

Diversive exploration, on the other hand, has the function of introducing stimulation from any source that is "interesting" or "entertaining." It is exemplified by the various activities through which human beings seek "amusement," "diversion," or "aesthetic experience," as well as by the behavior of a rat that presses a bar to produce a momentary change in illumination .... Whenever the external environment is inordinately dull or monotonous, diversive exploration is likely to occur. (Berlyne, 1965, p. 254)

Both types of behavior are triggered by a change in the arousal state of the individual (i.e., heightened physiological and mental reaction to some stimulus situation) which departs from the "optimal" ("non-zero") stimulation level each person is assumed to be comfortable with. Heightened arousal acts as a drive state which therefore implies that the person so aroused will be motivated so as to reduce the arousal-induced tension state and reattain the original "optimal" stimulation level.

In the case of diversive exploratory behavior, the individual's arousal level is heightened due to a boredom state--and novelty seeking behavior is implemented in order to reduce the tension associated with being bored. When an individual is seeking the "novel," the "entertaining," any relief from boredom, etc., any suitable stimulus might be sought. Thus, diversive exploration, while occasioned by an aversive drive state (boredom), is instituted by the individual organism. In the current vernacular, this is what most writers refer to as novelty-seeking behavior.

However, in the case of specific exploratory behavior, we are dealing with behavior directed towards a specific stimulus array. Berlyne states that stimuli in the environment have associated with them various properties, some of which can be motivating. Examples of such properties have been identified as ambiguity, incongruity, incompleteness, blurredness, surprisingness, complexity, and incongruous juxtaposition. Collectively, Berlyne refers to these as collative stimulus properties, and uses this term to reflect the fact that it is not the stimulus or stimulus properties in isolation which elicits an organism's exploratory behavior. On the contrary, it is the collation or comparison of the stimulus with the environment, the context of the situation, the previous experience of the individual, etc.

Although for convenience we often speak of novelty, complexity, and the like as attributes of external stimulus patterns, they are actually relations between physiochemical and statistical attributes of external events and attributes of the subject. They depend On interactions between what is outside and what is inside, the most important internal factors in this regard being traces left by prior encounters with relevant events in the past. Something may be highly novel, surprising or complex for an individual on one occasion, but a physically identical pattern will be far from novel, surprising, or complex for the same individual on another occasion. (Berlyne, 1968, p. 258)

The various collative stimulus properties themselves are assumed to have the ability to raise arousal, and once heightened beyond the "optimal" level of the individual, acts as a drive state--which therefore implies that the person so aroused is motivated to behave (i.e., explore the stimulus array) so as to reduce the arousal-induced tension state. It is the kind of behavior that we perform when we are said to be "'looking for" or "taking a closer look" at "something in particular." (Berlyne, 1963, pp. 289-290).

The magnitude of exploratory behavior is influenced by the intensity of the collative stimulus property embedded in the stimulus array. Berlyne states that as the intensity of, for example, blurredness increases, exploratory behavior will increase then decrease. In other words, exploration will be "maximized" at intermediate levels of collative property intensity.

Hopefully, this very brief overview has given the reader a feel for Berlyne's propositions. Central concepts discussed are five-fold: arousal, collative stimulus properties, collative property intensity, diversive exploratory behavior and specific exploratory behavior. The research reported in this paper is predicated solely upon the propositions regarding specific exploratory behavior. The purpose of the study and further elaboration of the conceptual foundation will be discussed in the next section.


Once arousal is raised above the "optimal stimulation level" for an individual (due directly to the intensity of the collative stimulus property [Howard and Sheth (1969) use the term "stimulus ambiguity'' in their theory of search; it is synonymous with collative property intensity.]), arousal causes the sensory receptors to open and thus increase attention. (Howard and Sheth, 1969, p. 159). Berlyne proposes that arousal is the antecedent condition for attention (a particular manifestation of specific exploratory behavior), and the major antecedent condition of arousal is collative property-bearing stimuli. The outcome of these interconnections is the following: attention should conform to an inverted U-shaped function of collative property intensity; thus, attention should be maximized at intermediate levels of intensity (e.g., moderate levels of blurredness, incompleteness, etc.) This is the major proposition tested in the study to be described.

While attention is but one index of specific exploratory behavior, at the same time it may be one of the most important: particularly from an advertising practitioner's point of view. Advertising strategy often includes development of particular copy that is designed to capture and hold the audience's attention. This may be especially critical when introducing the consuming public to a new brand on the market. Obviously there are a multitude of other communication goals possible (e.g., attitude change, knowledge enhancement, reinforcement of attitude, predisposition-to-switch, etc.), but especially for a new brand on the market, design of copy that can capture and hold the audience's attention may be a major goal during the early introduction stage of the product life cycle. Two prominent authors put it this way:

An advertisement is intended to communicate--to transfer knowledge, to create or change an image, to arouse needs or desires, to generate interest in a product class attribute, or to precipitate action. However, before a message can communicate--can accomplish any of these things--it must enter the mind of the receiver. (Aaker and Myers, 1975, p. 270)

Berlyne's concept of specific exploratory behavior would appear to be eminently applicable to just such a symbolic stimulus domain: mass media advertising. His major proposition (viz., that attention will be maximized at moderate levels of collative stimulus property intensity) suggests that

(1) if collative properties could be meaningfully embedded in a mass media advertisement for a new brand, and

(2) manipulated at three intensity levels (low, moderate and high), and

(3) if the consumer's attention could be monitored, then

(4) we should find that attention is at its highest when collative stimulus properties are presented at intermediate levels.

If this is found to be the case, advertising practitioners might have an additional tool available by which to increase the chances of consumers' paying attention to their ads--and if this occurs, increasing the chances of obtaining other communication goals (such as buying the new brand in order to try it).

Accordingly, the study reported in this paper is a first attempt to test Berlyne's "inverted U-shape" proposition by operationalizing three collative stimulus properties (at three levels each), embedding them in a print advertisement for a new brand (convenience item), and monitoring the attentional response of consumers to the print advertisement. Such an empirical test will not only provide insight into the above (as well as Berlyne's theory of specific exploratory behavior), but will also shed light on the rather neglected area of attention--and especially that facet of attention related to "enduring predispositions" which refer to relatively involuntary rules of attention (see Bettman, 1975 for more discussion of this point).


[The study described here is one part of the author's Ph.D. dissertation (Goodwin, 1977).]

The Stimulus Material

Three collative stimulus properties were selected for this study: blurredness, complexity, and incompleteness. Blurredness is a collative property which refers to the lack of clarity, or degree of "fuzziness" of the stimulus array. Complexity relates to the structure of the stimulus array and refers to the degree of intricacy, and/or the number of varied, interrelated parts, patterns, or elements. Incompleteness is a collative property describing the instance where facets of a stimulus array are unfinished, not fully developed, or not having all essential elements or parts.

Naturally, operational counterparts for these three concepts are necessary, as well as the selection of the basic stimulus object. Accordingly, a print advertisement for a new (at the time of this study) brand of instant coffee was selected as the basic stimulus. Selection of a new brand was deemed desirable in order to avoid certain biases due to predisposition toward current, on-the-market brands. Once the basic print ad was selected, there remained the task of selecting portions of it in order to implement the three collative properties. Toward this end, the print ad's headline was selected for manipulating degree of incompleteness. Blurredness was operationalized in the context of the brand name displayed on the container in the print ad. Complexity was operationalized in the context of a background characteristic (the number of clouds in the background) of the print ad.

These three collative properties were manipulated at three quantitative levels each: low, moderate and high. A professional artist was hired to build the twenty-seven (derived from the full 3 (headline incompleteness) x 3 (brand name blurredness) x 3 (background complexity) factorial design) mock print ads. At this stage, every attempt was made to represent distinctly different, and monotonically increasing, intensity levels for each of the three collative properties. Nonetheless, to insure isomorphism between the manipulanda and the underlying concepts, a thorough pretest was conducted. [Each of 54 students subjects received a single mock-up ad representing one of the 27 treatment combinations. All 27 mock-ups were used, and a full replication was implemented. The critical dependent measure gathered for this manipulation check was the specific (paper and pencil) measure of each subject's perception of the intensity of headline incompleteness, background complexity and brand name blurredness (11 point semantic differentials were used here). Three one-way analyses of variance were performed on the data, and the results indicated the desired isomorphism: subjects' perceptions were consistent with the "objective" collative property-level manipulations created by the researcher. A more detailed discussion of this pretest can be found in Goodwin, 1977.]

Once the manipulation checks were completed, the 27 mock-ups were photographed, and 27 black and white slides were prepared. These slides constitute the complete set of stimulus material used in the study described in this paper.


All subjects were female adults. A convenience sample (N=120), drawn from the female population of a large community church in a major Northeastern metropolitan area, participated in two rooms provided by the church for that purpose. The church was paid $3.00 per participant upon completion of the study. Due to the nature of the sample, wide generalization of results would clearly be speculative, at best. Consequently, the focus here is upon the internal validity of the study.

The procedure consisted of two separate stages, and each subject participated individually (one at a time) in each stage.

Stage I.  In the first stage, the subject was ushered into a room and was seated. Slightly behind the subject, on a table, a Kodak Carousel Slide Projector was positioned. The subject was told that she would be exposed to a number of slightly different pictures of a print advertisement for a new, and as yet unmarketed brand of instant coffee.

Since the purpose of this stage of the study was to examine free-response looking-time (a useful surrogate measure for attention to a stimulus object--see Leckart, 1969 or Morrison and Dainoff, 1972 for a discussion of the looking-time measure as an appropriate index of onset attention), it is important that each subject is unaware of this objective. Accordingly, a ruse was initiated: each subject was advised that the purpose of this stage of the study was to monitor her psychophysiological response to the stimulus material by use of a galvanic skin response (GSR) device. The experimenter verbally summarized how the GSR device worked, the subject was shown the device, the diode was then attached to her left palm, and she was reminded that the GSR responses were registered automatically on the machine in the back of the room [Pretests on this procedure indicated that subjects were fairly apprehensive; to help alleviate this, the GSR diode was attached by using a small dab of standard toothpaste (drawn directly from the tube in front of the subject); each subject was told that use of a paste-like substance increased the sensitivity of the diode. This ruse appreciably reduced subject apprehension.] (of course, in actuality the GSR device was not hooked up at all), and that these responses were not controllable at the conscious level. Thus, the subject was simply asked to sit back, relax, examine each slide for as long as she cared to, and to push the remote control button (held in her right hand) [The button was connected to a "looking-time device" which was the portable Oscillograph manufactured by Sperry Rand. This is an analog device which operates by running a continuous stream of paper upon which is recorded the point when a subject pushes the remote control button to initiate the next slide in the Kodak carousel. The Oscillograph was thoroughly cleaned and checked by a licensed electrician, pretested, and was found to be extremely reliable. Subjects' looking-times were recorded in equivalent distance units (where 250 millimeters of distance equaled ten seconds of looking-time).] when finished with a slide so as to go on to the next one.

Each subject looked at all 27 slides. The initial order selected for presenting these slides was determined by a random number table. Also, 27 different orders were developed so as to counterbalance the likely learning effects due to repeated exposure to "similar" stimuli. The upshot of this procedure is that 27 distinct sets of subjects (four to five per set) initially viewed a different treatment combination such that each one of the 27 treatment combinations was viewed first by separate small sets of subjects. This procedure permits a direct analysis of variance on the looking-time data. This particular analysis is the sole concern in this paper.

Stage II.  Upon completion of the looking-time task, each subject was ushered into a different room where she filled out a questionnaire designed to measure standard socioeconomic factors and a variety of pre-dispositional and personality factors. The latter set of factors were measured in order to control for "extraneous variation" in the looking-time data and included: instant coffee consumption (in cups) during a typical seven-day week; education level attained (in years); tolerance for uncertainty (Cox, 1967); cognitive clarity (Kelman and Cohler, 1957) and Arousal-Seeking Tendency (Mehrabian and Russell, 1974). The latter personality measure was used as an index of each subject's "optimal stimulation level"; as such, it is a surrogate measure for the arousal level that each respondent normally feels comfortable with. Clearly, it is important to control for this individual difference information in light of the theoretical arguments reviewed in the second section of this paper (i.e., "Berlyne's Complexity Theory").

Upon completion of this task, each subject was thanked for her participation. The church which recruited the subjects was sent a three page summary highlighting the nature of the study. Thus, debriefing was a collective one. Data collection took almost three full weeks to complete.


Effect of Collative Properties on Attention

The experimental procedure permits a "direct" analysis of variance of the looking-times at the first slide presented. Since each of the 27 slides (ads) was examined first by 27 different sets of four to five subjects, a between-subjects factorial analysis of variance was performed. [Since it is desirable to have equal cell sizes for the ANOVA (nj=5), this would have required 135 subjects. However, budget restrictions resulted in the participation of only 120 subjects. Consequently, in order to ensure equal cell sizes, for those 15 treatment conditions where only four subjects actually examined a given ad, an additional average subject was created. Thus, the analysis reported here is based upon a full 3 (Incompleteness) x 3 (Blurredness) x 3 (Complexity) factorial design, with cell sizes equal to 5, and total sample size equal to 135. This "adjustment" is but one of many available (Myers, 1972), and in retrospect may not have been the best one to use.]

Table 1 presents the summary for the analysis of variance. Five covariates were employed, and while none of the five were statistically significant overall, inclusion did not change the pattern of results and at the same time lowered the error variation (as compared to the ANOVA without covariates) thus permitting increased significance of the results. [It is important to point out that the error degrees of freedom (88) and the degrees of freedom (Total) (119) presented in Table 2 are smaller than one would expect given that 135 subjects' looking-time data is subjected to analysis. Since five covariates have been introduced, error degrees of freedom are reduced by five; that leaves 15 degrees of freedom to be accounted for. The explanation is as follows: the fifteen "average" subjects which were created in order to achieve design proportionality were used only to achieve proportionality in the looking-time data. "Average" data on the covariates was not calculated. Since educational level and personality characteristics are especially peculiar to each of the 120 subjects in the study, creating "average" covariate information would not make much sense. Thus, the missing values option of the program was utilized to handle the missing covariate information. This then explains why error degrees of freedom are 88 and total degrees of freedom are 119.]

The summary table reflects analysis after transforming the looking-time data via a log10 transformation (which was necessary in order to achieve normality of the error terms).



The three-way interaction is not strong enough to achieve statistical significance. However, two of the two-way interactions (Headline Incompleteness X Background Complexity, and Brand Name Blurredness X Background Complexity) are strong enough to achieve statistical significance at or below normally accepted levels of Type I error. (Although some would argue that the p<.065 obtained for the latter interaction is unacceptable, this author believes it is sufficiently close to the "standard" .05 level to merit additional scrutiny).

When interactions are significant, main effects are typically of little interest as far as further analysis is concerned. [Thus, even though the Blurredness main effect is marginally significant (p < .081), since blurredness interacts with complexity (see Table 1), the main effect itself is of little interest. This is unfortunate, in a way, since interaction hypotheses were not entertained a priori. At the same time, it is revealing to note that collative properties, when jointly embedded in a stimulus array, may have a different impact on exploratory behavior from that which might occur were only one single collative property to be embedded in the same stimulus array. Also, by uncovering the source of the interactions, it will still be possible to ascertain whether or not the inverted U-shaped relation obtains.] The significant two-way interactions uncovered indicate that (1) the effects of headline incompleteness on looking-time are not constant over levels of background complexity; and (2) the effects of brand name blurredness on looking-time are also not constant over levels of background complexity. Figure I portrays these interactions graphically.



In order to fully interpret these significant interactions, simple-effects tests were performed. For the incompleteness x complexity interaction, one of the three simple-effects was highly significant (looking-times at the low complexity level varied dramatically over the three levels of headline incompleteness) at p < .001. A second simple-effect (looking-times at the moderate complexity level across all three levels of incompleteness) approached significance at p < .09.

For the blurredness x complexity interaction, only one of the three simple-effects was of sufficient strength to indicate statistical significance. This occurred at the low background complexity level across all brand name blurredness levels (p < .02). A second simple-effect (looking-times at the high complexity level across all three levels of blurredness) approached significance at p < .10. Table 2 presents the (transformed) looking-time cell means for both significant two-way interactions, and Table 3 presents the results of the appropriate post hoc multiple comparison tests (Scheffe, 1959).



Table 2 reveals that the mean looking-time values for low illustration complexity over the three levels of headline incompleteness are monotonically increasing. Attentional responses at low complexity level increases linearly from low headline incompleteness to high incompleteness. However, the difference in magnitude between the medium and high level is much smaller than the difference between the low and medium level of incompleteness. This is mildly suggestive of a response function which is tapering off; had a higher level of headline incompleteness been implemented, it is plausible to expect the downward sloping response surface. The mean differences were not all statistically significant. Only the low-medium and low-high differences were of sufficient magnitude to warrant such a conclusion (Table 3).



When background complexity is at the medium level, the results are quite different. The mean looking-times reflect a cup shaped parabolic function over the three levels of headline incompleteness. Further, the mean value for incompleteness (at its lowest level) is the biggest of the three. These results imply (since the low and high levels are not statistically different) that attention is maximized--when considering a medium background complexity level--at either the lowest or highest level of incompleteness.

Turning to the examination of mean looking-time responses for low and high background complexity over the three levels of brand name blurredness, it is evident that here the results are reasonably consistent with Berlyne's major proposition. In both cases (where simple-effects were found to approach statistical significance), an inverted U-shaped function between attention and brand name blurredness is found.

The magnitude of the mean looking-times at the low complexity level indicates that attention is lowest for the low level of brand name blurredness while attention is highest for the medium blurredness level. However, the mean difference between looking-times at the medium and high levels of brand name blurredness does not approach statistical significance (at low complexity level). (Table 3) Thus, while the array of means does approximate the hypothesized inverted U-shaped function, differences between all pairs of means are not significant.

The magnitudes of the looking-times at the high background complexity level suggest that attention is smallest for the high level of brand name blurredness, and attention is again at its maximum for the medium level of blurredness. However, the mean difference between looking-times at the low and medium levels of blurredness do not approach statistical significance at this highest level of background complexity (see Table 3). The only strong difference occurs between the medium and high levels of blurredness, where the intermediate level of blurredness dominates the high level. Again, the array of means approximates the hypothesized inverted U-shaped function, but differences between all pairs of means are not significant.


Unfortunately, presence of interaction in the looking-time data "clouds" the interpretation. Nonetheless, close examination of the significant 2-way interactions led to the following conclusion: attention to the collative property-bearing print ad for a new brand of instant coffee depended upon the joint functions of headline incompleteness and background complexity as well as brand name blurredness and background complexity; and, attention was maximized for these collative property implementations above the zero-order level. While Berlyne's major proposition was not supported in the case of complexity across incompleteness levels, it was (largely) in the case of complexity across blurredness levels. It was particularly interesting to note the confirmation of the inverted U-shape function for the significant simple-effects of the blurredness x complexity interaction (as determined by two trend analyses).

While generalization of these results would clearly be unwarranted, they are suggestive. For cases where preliminary marketing objectives are set in terms of maximizing attention or awareness to a new brand on the market, creative embedding of collative properties (at other than zero-order levels) in a print ad may help to attain such a goal. Also, in pretesting such ad(s), a free response looking-time paradigm might prove useful when attention maximization is a central concern. Still, the ubiquitous caveat applies here as everywhere: more research is in order.


Aaker, D. and Myers, J. (1975), Advertising Management, Englewood-Cliffs, New Jersey: Prentice-Hall.

Berlyne, D. E. (1954), "A Theory of Human Curiosity," British Journal of Psychology, 45, 180-191.

Berlyne, D. E. (1960), Conflict, Arousal and Curiosity, New York: McGraw-Hill.

Berlyne, D. E. (1963), "Motivational Problems Raised by Exploratory Behavior," in S. Koch, ed., Psychology: A Study of Science, New York: McGraw-Hill.

Berlyne, D. E. (1965), Structure and Direction in Thinking, New York: John Wiley and Sons.

Berlyne, D. E. (1968), "The Motivational Significance of Collative Variables," in R. Abelson, et. al., eds. Theories of Cognitive Consistency, Chicago, Rand McNally, 275-88.

Bettman, J. (1975), "Issues in Designing Consumer Information Environments," J. of Consumer Research, 2, 169-177.

Cox, D. F. (1967), Risk Taking and Information Handling in Consumer Behavior, Boston: Harvard University.

Faison, E. (1977), "The Neglected Variety Drive: A Useful Concept for Consumer Behavior," J. of Consumer Research, 4, 172-175.

Goodwin, S. (1977), "An Empirical Investigation On The Attentional and Perceptual Effects of Collative Print Advertising Properties," unpublished Ph.D. dissertation, The University of Iowa.

Hansen, F. (1972), A Cognitive Theory of Consumer Choice Behavior, New York: Free Press.

Howard, J. and Sheth, J. (1969), The Theory of Buyer Behavior, New York: John Wiley and Sons.

Leckart, B. (1969), "Reliability of Looking Time," Perception and Psychophysics, 5, 313-314.

Maddi, S. (1968), "The Pursuit of Consistency and Variety," in R. Abelson, et. al., eds., Theories of Cognitive Consistency, Chicago: Rand McNally, 267-74.

Mehrabian, A. and Russell, J. (1974), An Approach to Environmental Psychology, Boston: MIT Press.

Morrison, B. J. and Dainoff, M. J. (1972), "Advertisement Complexity and Looking Time," J. of Marketing Research, 9, 396-400.

Myers, J. L. (1972), Fundamentals of Experimental Design (2nd edition), Boston: Allyn and Bacon.

Rogers, R. D. (1979), "A Commentary on The Neglected Variety Drive," J. of Consumer Research, 6, 88-91.

ScheffT, H. (1959), The Analysis of Variance, New York: John Wiley and Sons.

Venkatesan, M. (1974), "Cognitive Consistency and Novelty-Seeking," in S. Ward and T. Robertson, eds., Consumer Behavior: Theoretical Perspectives, Englewood Cliffs, N. J.: Prentice-Hall.