Processing Complexity in Magnitude Versus Category Scaling

ABSTRACT - This study supplements previous research on measurement qualities of magnitude versus rating scales. Its primary purpose is to determine the complexity in cognitive processing between these two scaling methods on responding to advertisements and erotic pictures. Secondly. it is to ascertain the conceptualization of collative properties of stimuli (e. g. the complexity and novelty of an object) as a more cognitive judgment compared to affective responses. The results suggest that magnitude scaling allows for fewer cognitively controlled responses and the conceptualization of collative responses as a cognitive judgment which could be empirically supported.

Citation:

Bruno Neibecker (1986) ,"Processing Complexity in Magnitude Versus Category Scaling", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 286-290.

Advances in Consumer Research Volume 13, 1986      Pages 286-290

PROCESSING COMPLEXITY IN MAGNITUDE VERSUS CATEGORY SCALING

Bruno Neibecker, University of the Saarland (Fed. Rep. of Germany)

[The author wishes to express his gratitude to Professor W. Kroeber-Riel and Professor Jerry Olson for ideas on the subject and helpful editorial comments. Author's address: Institut fur Konsum- und Verhaltensforschunx. 6600 Saarbrucken 11. FRG.]

ABSTRACT -

This study supplements previous research on measurement qualities of magnitude versus rating scales. Its primary purpose is to determine the complexity in cognitive processing between these two scaling methods on responding to advertisements and erotic pictures. Secondly. it is to ascertain the conceptualization of collative properties of stimuli (e. g. the complexity and novelty of an object) as a more cognitive judgment compared to affective responses. The results suggest that magnitude scaling allows for fewer cognitively controlled responses and the conceptualization of collative responses as a cognitive judgment which could be empirically supported.

BACKGROUND

The effects of affective factors in consumer behavior are now gaining closer attention. A splendid overview about the importance of consumers moods and emotions on their behavior is given in the 1984 ACR-Proceedings (Vol. XI. pp. 525-547). Let me summarize the main thoughts of the mood debate as far as the research conducted here is concerned. In emotion psychology at least two trends can be distinguished. On the one side. it is assumed that affect and cognition are under the control of separate and partially independent systems that may influence each other in a variety of ways (e.g. Zajonc 1980). This position is supported. at least partially. by research on hemispheric specialization of affective processes (Tucker 1981; Bryden and Ley 1983) and research on the biological determinants of human-behavior. called "human ethology" (Eibl-Eibesfeldt 1980).

On the other side. it is argued that thought is a necessary condition of emotion (e. g. Lazarus 1982). This position is based on the concept of cognitive appraisal which means that the way ore interprets one s own state at any given moment is a crucial determinant of the emotional response.

Since humans have the ability for autonomic discrimination without awareness. it is even reasonable to assume that perceptions can be global rather than built up sequentially of structural elements. Therefore, emotionally relevant meanings could be triggered by inputs. denotations of which had not been achieved completely.

However, the crucial question of this debate, whether the original affective reactions themselves - on which evaluative judgments are presumably based - must also contain cognitive components cannot be answered empirically given the current state of knowledge and research technology (Zajonc and Markus 1982. p. 124).

Of central interest for our further argumentation is the assumption. which may rely on both families of emotional theories. that emotional reactions, require a minimum of cognitive participation. at least on the self-report level (Lazarus, 1982). At the same time, there is support that affective reactions are difficult to verbalize and rely much more on nonverbal channels (Zajonc 1980: Ekman and Friesen 1969).

If we consider these two extremes - i.e.. affective and cognitive behavior - not as dichotomous types but as ends of a continuum. it is possible to distinguish between different self-report methods along this continuum

Recently, the validity of an alternative direct scaling procedure. called magnitude scaling or "new" psychophysics has been shown for emotional stimuli (Neibecker 1984; see also Stevens 1975; Marks 1974). In this paper it will be argued along with others (Kroeber-Riel 1984; 1985) that magnitude scaling allows for responses which are less cognitively controlled because subjects can express the magnitude of their sensory experiences on nonverbal continua. such as the length of a line, the duration of pressing a button, the loudness of a tone. etc. These analogue and nonverbal response modalities are quite different from the conventional rating-scale/category task. where subjects are asked to express their impressions of sensation intensities by choosing among a limited number of categories labelled with such cues as agree/disagree.

A second point of interest relevant for this research is derived from environmental and emotional psychology. The dimension theorists in emotional psychology have conceptualized a small set of independent dimensions that best describe the differences in emotional behavior (Mehrabian and Russel 1974; overview by Ekman et al. 1982). The most widely accepted dimensions are pleasantness-unpleasantness and activation or arousal. In addition, Ward and Russell (1981) discuss the collative properties of stimuli. This construct is based on research of Berlyne (1974) and Mehrabian and Russel (1974) and more generally known as the "information rate-arousal hypothesis". Collative properties of stimulus patterns are structural or formal characteristics measured with items like familiar-novel, simple-complex, expected-surprising etc. Ward and Russell (1981) have conceptualized this collative construct as a cognitive judgment about the complexity and novelty of an object.

The usefulness of the collative dimension is supported by research on TV-commercials (e. g. Aaker and Bruzzone 1981). But this prior research in consumer behavior lacks a clear theoretical foundation and uses the novelty component only as an empirically derived construct. The psychological background given here helps to interpret the different sets of adjectives developed to measure consumers reactions to advertising.

HYPOTHESES

Based on these theoretical considerations, two hypotheses were formulated. One concerned the measurement qualities of magnitude scaling as a more nonreflective and spontaneous measurement tool compared to rating scales. The second concerned the collative properties of stimuli which postulates that this construct is accompanied by more cognitive processing than the affective dimensions pleasantness and activation.

It was hypothesized that:

H1: Nonverbal responses (e.g. magnitude scaling) allow for less cognitively controlled responses than verbal responses (e.g. rating scales) when measuring emotions.

H2: Judgments about collative properties require more extensive cognitive processing than do affective responses.

METHOD

The experimental data reported herein were supplied by an extensive laboratory experiment conducted at the Institute for Consumer and Behavioral Research, University of the Saarland, FRG. The study was performed on a random sample of 136 persons with a quota sampling of 50 per cent students, the remainder being adult men and women aged 25 to 54 (non-students). Four stimuli were chosen, including advertisements and erotic pictures. The ads were selected from current image ad campaigns of two leading German chemical enterprises. Pictures 1 and 4 were erotic pictures showing a woman with little clothing; picture 2 was an ad with a woman s and a man s hand touching in an affectionate manner with a small amount of copy below; picture 3 was an ad showing three children sitting in a small pasteboard house with copy below. Because of time limitations, subjects were divided into two groups, one consisting of 69 subjects evaluating pictures 1 and 2 using category and magnitude scales and the other made up of 67 subjects evaluating pictures 3 and 4 using category and magnitude scales.

To reduce the time needed to collect and process data and reduce method-dependent artifacts, magnitude scaling was integrated into a computer-controlled questionnaire facility. The interview was conducted via a computer-controlled system. Respondents sat directly in front of a cathode ray tube (CRT) terminal and answered questions by pressing keys on the terminal. For our experiment, the following modalities were implemented: visual length (line length), time duration with simultaneous presentation of a white noise, and brightness by a regulated lamp.

First, the subjects received a description on the CRT screen of how to use the apparatus. These instructions included a learning session on how to use the equipment for the magnitude task (about four minutes). Subjects were ascertained initially to make approximate proportional judgments. The first group made their judgments with rating scales and the magnitude modalities line length and brightness for pictures 1 and 2, the second group with rating scales and magnitude modalities line length and duration for pictures 3 and 4.

The rating scales were of the semantic differential type. The typical scale had the following form:

Picture x is

slightly dynamic(1) (2) (3) (4) (5) (6) (7)very dynamic

For magnitude scales, only a mnemonic statement was given. The typical statement for the modality line-length had the following form.

Picture x is

slightly dynamic (short line)                        very dynamic (long line)

where the length of the line was under the control of the respondent.

The sequence of the pictures as well as of the items and methods were rotated randomly by the computer and only one item was shown at a time. To produce a line, the subjects pressed a button on the CRT and the computer started to draw a line on the screen. For the modality duration, subjects pressed a separate button. In addition a white noise was presented. The brightness could be regulated by turning a knob (potentiometer) until the adequate lamp brightness was reached. At the beginning of a response, reaction time was measured unobtrusively (non-reactive) by the computer. This was the time starting with the appearance of the item on the screen until a response was given.

The questionnaire consisted of 10 emotion descriptive adjectives. Convergent and discriminant validity of the arousal measuring items were shown already elsewhere (Neibecker 1984). For this research, items were separated according to these results into three treatment factors: arousal, collative judgments and evaluation. The respective items were: dynamic, lively, tiring, lifeless (for arousal) / common, average (for collative properties) / beautiful, disgusting (for evaluation).

To operationalize the cognitive processes involved in responses the reaction time was measured. The faster responses are given, the fewer cognitive processes are assumed. This is a usual operationalization to determine the relative complexity in cognitive processing (e.g., Mandler and Shebo 1983; Paivio 1978).

RESULTS

For a descriptive overview, the mean reaction times (or response latencies) are given in seconds (see Table 1). Furthermore, they are divided into three different judgment tasks (rating, line length, duration / brightness), three response types (evaluation, collative judgments, activation), and two picture Categories (erotic pictures and advertisements).

TABLE 1

RESPONSE TIME IN SECONDS

As can be seen, reaction times for rating scales are longer than those for magnitude measures (in columns). Likewise, response times for collative judgments are longer than those for affective responses (evaluation and activation).

In a further step, two 2 x 3 x 3 analyses of variances were performed. The effects were stimulus x method x judgment task (or dimension). Stimulus was treated as between subjects factor and the others as within subject factors. The raw scores were rescaled according to MacLachlan (1977). By this rescaling procedure, typical skewness is diminished from 1.30 to about 0.40 (highest skewness 1.17) and typical kurtosis is diminished from 1.70 to about -0.70 to 0.70.

                    RT_i

ART_ik  =

                   RT_ik

(i = subject; k = item; RT_i = mean reaction time of subject i; ART = adjusted reaction time; RT = reachtion time)

Results are given in Table 2.

As can be seen, none of the stimulus factors (erotic pictures and advertisements) reached significance showing that no differences exist between the over-all reaction times of the two groups of subjects. In fact, the evaluated pictures in each group resembled each other strongly. The division into these groups was only necessary because of time limitations. Of central interest for our hypotheses are the factors method and judgment task (dimension). Each one of these factors is highly significant reflecting that affective responses are given faster than collative judgments, and nonverbal responses via magnitude scaling are given faster than those with rating scales. Up to now the three-way interactions are difficult to interpret. Figure 1 shows that this is due to the unsystematic deviation of one method in pictures 1, 3 and 4 from the over-all main effects. It reads as follows:

FIGURE 1

INTERACTIONS

TABLE 2

ANOVA FOR EROTIC PICTURES AND ADVERTISEMENTS

The upper left-hand side of Figure 1 represents the plot of the profiles for the Method x Dimension interactions for the two levels of the erotic pictures. On the upper right-hand side, the profiles of the overall Method x Dimension means are given. These three profiles are parallel, indicating that the Method x Dimension interaction is not significant. Since some profiles for each level of the picture factor are not parallel to the respective profile of the combined picture data the three-way interaction is significant.

For the advertisements, profiles are given at the bottom of Figure 1. Here, the profiles on the right side of the over-all Method x Dimension means are not parallel indicating a significant Method x Dimension interaction. Again, some profiles for each level of the advertisement factor are not parallel to the respective profile of the combined advertisement data. This finding implies a significant three-way interaction.

CONCLUSIONS

This research does not claim to answer the opposing question, whether or not there exist two distinct, qualitatively different systems for affect and cognition. Since this question cannot be answered empirically given the current state of knowledge and research technology, we proposed a more pragmatic position. Considering the requirements of the respondents by answering questions, we expected a more or less complex processing in making evaluative judgements. It can be assumed that the usual response situation, in which a questionnaire has to be completed, tends to be as least as complex as simple lexical decision judgements, for example.

On the other side, emotional reactions, even if cognitive mediation should be necessary, could be triggered by inputs whose full-fledged denotations had not yet been achieved. In addition, considering that emotional reactions rely more on non-verbal channels and are difficult to verbalize, we suggest that affective responses be measured more adequately and spontaneously by non-verbal response modalities.

Based on these arguments, it has been shown that magnitude scaling with non-verbal modalities like "length of a line" or "duration of pressing a bottom" allows for shorter reaction times than category scaling with ratings. Since it is of central interest to measure the first affective reaction, which is not cognitively masked and less influenced by social desirability, quick responses are preferred. Thus, magnitude scaling seems to have a more direct (nonverbal) path to emotional processes which are probably more concentrated in the right hemisphere of the brain.

At the same time, if reaction time is an effective operationalization for the amount of cognitive processing, then our distinction between affective items and information rate items, conceptualized as a cognitive judgement concerning the complexity and novelty of a stimulus, should hold. This distinction could be empirically supported. Thus, with some modification of the prior theorizing of Berlyne (1960), who predicted a direct relationship between collative properties of a stimulus and activation (arousal), the conceptualization of Ward and Russell (1981) was supported.

Unfortunately, the main effects are overlapped by three-way interactions, which are due to a rest of unsystematic variations between methods and dimensions.

REFERENCES

Aaker, D. A. and D. E. Bruzzone (1981), "Viewer Perceptions of Prime-Time Television Advertising," in Journal of Advertising Research, 21, 15-23.

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

Berlyne, D. E. (1974), " The New Experimental Aesthetics," in Studies in the New Experimental Aesthetics, D. E. Berlyne, Et., Washington D. C.. Hemisphere Publishing, 1-25.

Bryden, M. P. and R. G. Ley (1983), "Right Hemispheric Involvement in Imagery and Affect," in Cognitive Processing in the Right Hemisphere, E. Perecman, Ed., New York, et al.: Academic Press, 111-123.

Eibl-Eibesfeldt, I. (1980), " Strategies of Social Interaction," in Emotion. Volume 1: Theories of Emotion, R. Plutchik and H. Kellerman, Eds., New York Ft al: Academic Press. 57-80.

Ekman, P. and W. V. Friesen (1969), "The Repertoire of Nonverbal Behavior: Categories, Origins, Usage, and Coding," Semiotica 1, 49-98.

Ekman, P., W. V. Friesen, and P. Ellsworth (1982), "What Emotion Categories or Dimensions can Observers Judge from Facial Behavior?," in Emotion in the Human Face, P. Ekman, Ed., Cambridge et al.:Cambridge University Press, 39-55.

Kroeber-Riel, W. (1984), " Konsumentenverhalten," 3rd ed., Munchen: Vahlen. (Consumer Behavior)

Kroeber-Riel, W. (1985), " Analysis of Non-cognitive Behavior, Especially by Non-verbal Measurement," Advances in Marketing Communication Research, R. P. Bagozzi, Ed., Greenwich, CT: JAI Press.

Lazarus, R. S. (1982), " Thoughts on the Relations Between Emotion and Cognition," American Psychologist 37, 1019-024.

MacLachlan, J. (1977), "Response Latency: New Measure of Advertising," New York: Advertising Research Foundation.

Mandler, G. and B. J. Shebo (1983), "Knowing and Liking," Motivation and Emotion 7, 125-144.

Marks, L. E. (1974), " Sensory Processes. The New Psychophysics," New York-London: Academic Press.

Mehrabian, A. and J. A. Russel (1974), "An Approach to Environmental Psychology," Cambridge Massachusetts: MIT Press.

Neibecker, B. (1984), " The Validity of Computer-Controlled Magnitude Scaling to Measure Emotional Impact of Stimuli," Journal of Marketing Research 21, 325-331.

Paivio, A. (1978), " Mental Comparisons Involving Abstract Attributes," Memory & Cognition 6, 199-208.

Stevens, S. S. (1975), " Psychophysics. Introduction to Its Perceptual, Neural and Social Prospects," New York, et al.: Wiley.

Tucker, D. M. (1981), " Lateral Brain Function, Emotion, and Conceptualization," Psychological Bulletin 89, 19-46.

Ward, L. M., and J. A. Russel (1981), "The Psychological Representation of Molar Physical Environments," Journal of Experimental Psychology: General 110, 121-152.

Zajonc, R. B. (1980), " Feeling and Thinking. Preferences Need no Inferences," American Psychologist 35, 151-175.

Zajonc, R. B., and H. Markus (1982), "Affective and Cognitive Factors in Preferences," Journal of Consumer Research 9, 123-131.

----------------------------------------