Measuring Communication-Evoked Imagery Processing

Pam Scholder Ellen, Georgia State University
Paula Fitzgerald Bone, West Virginia University
ABSTRACT - During the 1980s, interest among consumer researchers in imagery processing increased. In order for this line of research to continue to be productive, it is important that a measure of imagery processing which is evoked by a particular communication be made available. This paper details the initial stages of the development of a multi-item imagery scale which appears to be a reliable and valid scale of communication-evoked imagery.
[ to cite ]:
Pam Scholder Ellen and Paula Fitzgerald Bone (1991) ,"Measuring Communication-Evoked Imagery Processing", in NA - Advances in Consumer Research Volume 18, eds. Rebecca H. Holman and Michael R. Solomon, Provo, UT : Association for Consumer Research, Pages: 806-812.

Advances in Consumer Research Volume 18, 1991      Pages 806-812


Pam Scholder Ellen, Georgia State University

Paula Fitzgerald Bone, West Virginia University


During the 1980s, interest among consumer researchers in imagery processing increased. In order for this line of research to continue to be productive, it is important that a measure of imagery processing which is evoked by a particular communication be made available. This paper details the initial stages of the development of a multi-item imagery scale which appears to be a reliable and valid scale of communication-evoked imagery.


During the 1980's, consumer researchers became more and more interested in imagery processing and its relative merits over purely discursive processing. Imagery processing is defined as the representation of any sensory experience in working memory (MacInnis and Price 1987). While most empirical work in imagery processing has focused solely on visual imagery, imagery processing may be any multi-sensory experience (i.e., sight, taste, sound, smell, or touch) evoked by a message (MacInnis and Price 1987; Childers and Houston 1982). Of specific interest in this research is memory imagery, or imagery recalled or created in the absence of the actual sensory stimulus (Richardson 1969).

Consumer researchers have examined the effects of imagery-inducing strategies on various consumption-related outcomes such as brand name recall, information search, behavioral intentions, and to a limited extent, attitudes. (See MacInnis and Price 1987 for a review). In doing so, imagery processing is often treated as a distinct process from discursive processing although the researchers acknowledge that the two processes coexist. Additionally, few of these studies have actually assessed the form of processing. Instead most studies have inferred that imagery processing, as opposed to verbal processing, occurred by examining observed differences in these consumption-related variables.

For these reasons, it seems appropriate to develop indicators of the extent or degree of imagery processing evoked by a communication. The purpose of this research is to describe the initial development of such measures which may be used across studies. Reliable indicators with evidence of validity may be useful either as manipulation checks, criterion or predictor variables in subsequent imagery research. In the following section, we review existing imagery measures and describe the proposed measures.


A number of scales currently exist to measure individual differences in imagery processing abilities and preferences (see MacInnis 1986 for a review). Some of the measures assess a person's ability to engage in imagery (e.g. Betts Questionnaire Upon Mental Imagery (Betts 1909; Sheehan 1967)) while others assess individual differences in preferred processing style (e.g. Style of Processing Questionnaire (Childers, Houston and Heckler 1985)). These scales were designed to measure general traits rather than imagery states induced by a communication. "The use of these [trait] scales to predict specific consumer behaviors requires careful consideration" (MacInnis 1986, p.92). As MacInnis (1986) points out, one individual may have a greater score on an imagery ability or preference scale, yet another individual may experience more imagery because of greater knowledge or familiarity with a specific domain. Thus, the imagery evoked by a specific communication is a function of both one's ability and the stimulus itself and is not adequately captured by the trait scale.

A few researchers have attempted to assess the effects of imagery-inducing strategies on actual processing. Lutz and Lutz (1977) used a self-report to assess whether subjects used visual or verbal processing. While this type of measure provides an indication of the type of processing, there is no indication of the extent of imagery processing. In addition, this measure is limited to visual imagery. Smith, Houston and Childers (1983) and McGill and Anand (1989) attempted to measure the extent of imagery processing by asking subjects to provide written protocols as they considered a situation. Smith, Houston and Childers (1983) also asked subjects to indicate whether each activity described was visualized or verbalized and if visualized, how vivid or clear the images were. The primary disadvantage of such measures is that they require the subjects to translate sensory experiences into verbal protocols. Morris and Hampson (1983) argue that sensory experiences may be difficult to verbalize. Note that these measures are also limited to visual imagery and exclude the other senses. Thus, it appears that there is a need for a measure of the extent of imagery processing evoked by a communication which can be used across studies, can incorporate visual as well as other sensory experiences, and does not depend on the subject's ability to verbalize the imagery content.

The first step in developing such a measure was to examine the existing imagery literature to identify possible dimensions of communication-evoked imagery. The following section addresses that issue. Next, a multi-item scale developed to tap the various dimensions is introduced. Then, empirical work which investigates the dimensionality, internal consistency, sensitivity to processing differences, and discriminant validity is presented.


Imagery-evoked by a communication may be reflected in several different dimensions. The dimensions discussed here are the vividness and/or clarity, quantity, ease, and links experienced as a result of the message.

Morris and Hampson (1983) identify vividness as the major dimension of imagery. While vividness is by far the more prevalent characteristic ascribed to differences in imagery processing (Betts 1909; Cartwright et al. 1978; MacInnis and Price 1987, Marks-1973), assessing the vividness of evoked images assumes that the vividness of images varies and indicates the quality of the imagery (Morris and Hampson 1983). However, attempts to measure vividness are often confounded with measures of other dimensions, especially clarity. While they are likely related dimensions, clarity refers to the detail of the images while vividness is more closely associated the intensity of the images. Morris and Hampson (1983) argue that there is not sufficient empirical or theoretical support for the assumption that these dimensions are the same therefore it is important to consider these descriptors as separate initially.

While vividness and clarity tap qualitative aspects of evoked-imagery, Smith, Houston and Childers (1983) suggest that there may also be quantitative differences in imagery processing. McGill and Anand (1989) found that the use of imagery instructions resulted in a greater number of evoked scenes and a greater number of inferred attributes. The quantity or number of different images created may or may not be related to the vividness or clarity of the images. One person may evoke a single, very vivid image while another may experience numerous images which may be much less vivid.

Ease of imagining has been suggested by several researchers as an important aspect of imagery processing (Anderson 1983; Paivio 1968; Sherman et al. 1983). The more available that information is concerning the subject to be imagined, the easier it should be for the individual to engage in imagery processing. For instance, much of the discussion of effects of concrete words and sentences on recall are attributed to the ease with which associated images can be elicited.

The final dimension of imagery is imagery links (Kisielius and Sternthal 1986; Lord 1980; MacInnis and Price 1987). One of the touted advantages of imagery processing over verbal processing is its greater ability to link or activate other information in long term memory. This activation occurs because imagery processing allows activation of stored information through a number of means (i.e., sensory experiences) other than simply semantic links. Such links should ultimately result in greater elaboration and therefore greater availability of the information at judgment or decision-making time.

In sum, there have been at least five dimensions of imagery processing discussed in the literature. In the next section, we present a set of measures developed to measure these dimensions: vividness, clarity, quantity, ease and links.


To develop the items to measure each of the dimensions, we first examined the approaches used in previous research to describe aspects of imagery and measure imagery processing abilities. We developed a list of items or descriptors used in previous research (c.f., Paivio 1968) as well as synonyms. For vividness, consistent with Morris and Hampson (1983), we included descriptors assessing both intensity and clarity in these measures.

In sum, nineteen items were developed to assess communication-evoked imagery: eleven vividness/clarity items, three quantity items, three ease items and two link items. The items are presented in Table 1. Given differences in item scaling (5-point vs. 9-point), all items were standardized prior to any analysis. We new turn our attention to determining whether these are indeed separate dimensions.


The data from two separate studies were used to assess the proposed measures. College students (n=179 for Study 1 and n=144 for Study 2) were processed individually in an audio-visual lab, and each subject was randomly assigned to one of the seven professionally-produced advertisements for a fictitious brand of popcorn, a product selected for its high sensory attributes. In each study, subjects listened to either one of several high-imagery radio advertisements or a low-imagery control ad. Participants listened to the ad twice to allow adequate time for creating images as well as comprehending the message.

All six high-imagery ads used concrete words, actionable sentences, present tense, and instructions to imagine in order to facilitate imagery processing in general (Sherman et al. 1983; Paivio 1976; Alesandrini and Sheikh 1983; Carroll 1978). The ads contained no music or sound effects, and all were similar in that different types of sensory experiences were described in each: sights, sounds, aromas, tastes, and body movement. The character and situation imagined differed in the ads; however, these differences are unimportant for this discussion and are not addressed here.

In addition to the high-imagery ads, one low-imagery control ad was used. The ad was for the same product and contained the same number of mentions of brand name as the high-imagery ads, however, the advertisement contained no instructions to imagine and no concrete words or sensory descriptions. After listening to the ad, subjects completed the measures of communication-evoked imagery and the short version of Betts' Questionnaire upon Mental Imagery (QMI) (Sheehan 1967; Betts 1909).




The nineteen items were subjected to confirmatory factor analysis using three, four and five factor solutions. It was expected that there would be one factor each for quantity, ease and links and either one vividness factor or separate vividness and clarity factors. Results for both studies are presented in Table 2.

The results from the two studies are remarkably similar. In both cases, the most interpretable solution was a four factor solution. The solution for both studies indicated a joint quantity and ease factor, a vividness factor, a paleness factor and a links factor. Although the x2 for all three models was significant, the four factor solution was significantly better than the three factor (one dimension for both vivid and pale) with a x2 difference of 124.46 (df=16, p<.01) and 255.09 (df=16, p<.01) for Study 1 and 2, respectively. The four factors accounted for 62.2% and 62.4% of the variance, respectively. In each case, there is a quantity/ease factor, explaining 29.9% and 18.9% of the variance, respectively, in the imagery measures. There are two separate factors accounting for the vividness/clarity items. The first factor, vividness, explains 19.9% (Study 1) and 30.9% (Study 2) of the variance. The second factor includes items indicating the paleness (or lack of clarity) in the imagery, explaining 7.6% and 7.8% of the variance in the two studies. The final factor, the links items, explain 5.7% and 4.8% of the variance in the two datasets.

The factor loadings in each case were quite clear and very similar. Specifically, factor loadings for the individual items exceeded 0.50 on the appropriate factor in all but one case (vivid on the vividness factor in Study 1). In every case, the loading was significantly larger on one factor than on the remaining three.

The results indicated that the number of images imagined and the ease with which they are imagined are highly correlated and in this case are captured by one factor. (The correlation between the separate summated indicators of the quantity and ease items is .71 and .74 in the two studies.) Vividness and paleness appear best handled as separate dimensions. The three factor solution results in a single dimension for all eleven items but the fit is substantially worse for that model. Finally, the links factor, while explaining the least amount of variance, still represents a significant and potentially important indicator of overall imagery processing. The factor analysis provided some evidence of the unidimensionality of the different measures.



The next step in the validation of these measure was to look at their reliability. Coefficient alpha was used for the quantity/ease measure and the vividness and paleness measures, while Pearson product moment correlation was used to assess the internal consistency of the link items. The reliabilities for each dimension are high--all exceed 0.83. For the quantity/ease dimension, coefficient alphas are 0.88 and 0.91 for Study 1 and Study 2, respectively. The reliabilities for vividness are 0.88 (Study 1) and 0.87 (Study 2) and for paleness are 0.89 and 0.84. The correlation between the two link measures is 0.91 for Study 1 and 0.93 for Study 2. One should note that the reliability indicators are quite similar between the two studies. Thus, the empirical evidence supports the claim that the four indicators of message-evoked imagery dimensions are internally consistent.


Once dimensionality and reliability were assessed, we turned our attention to validity issues. We examined validity in two ways. The first was to assess whether the proposed measure is sensitive to processing differences and the second was to determine whether the measure differs from trait measures of imagery, specifically one's intrinsic ability to engage in imagery processing.

When assessing whether this measure is sensitive to differences in processing style, we relied heavily on the imagery literature in developing an advertisement which should evoke greater imagery and one which should evoke less imagery processing. Thus, the high-imagery ad used concrete words (Sherman et al. 1983; Paivio 1976), active voice (Alesandrini and Sheikh 1983), instructions to imagine (Carroll 1978; Paivio 1976) and interaction with the product (Alesandrini and Sheikh 1983), while the low-imagery ad did not contain these elements.

To assess whether the measures reflect differences in imagery-evoking strategies, we compared responses on each dimension for a high-imagery ad group and a low imagery ad group from each study. A MANOVA model was used to address the question at hand. The high vs. low imagery advertisement served as the independent variable and the four summated indicators of imagery processing (quantity/ease, vividness, paleness and links) served as the-dependent variables. The Wilks' lambda for Study 1 was significant (lambda= 0.72, F=4.44, p<0.01). Individual ANOVAs indicated that statistically significant differences existed between the two advertisements for the quantity/ease factor (F=17.06, p<0.01) and the vividness factor (F=9.57, p<0.01). The paleness factor was not significant (F=2.68, p=11) while the links factor was only marginally significant (F 3.11; p=0.08). Examination of the means indicated that the high-imagery ad resulted in greater quantity, easier imaging and more vivid images.



Similar results are found for the Study 2. The overall Wilks' lambda was marginally significant (lambda= 0.78, F=2.45, p .06). Differences were found between the two groups for the quantity/ease factor (F=7.23, p<0.05) and the vividness factor (F=5.42, p<0.05) but not for the paleness factor (F=1.5, p=.23) or the links factor (F=0.12, p=0.74).

This analysis suggests that the quantity/ease factors and the vividness factors are working as expected, but that the paleness and links factors are not. It appears that more research of these factors are needed. It may be that these inconclusive results are an artifact of the specific message content.


A critical issue in developing this measure is to deter nine whether the measure is sensitive to a particular communication or is simply measuring a person's innate ability to imagine. It must be shown that the measure of communication-evoked imagery does not duplicate measures of imaging ability, otherwise it provides no additional value beyond what is currently available.

Two methods of addressing the discriminant [ validity of the proposed message-evoked r communication measure were used. First, we examined the correlations between the shortened version of Betts' QMI (Sheehan 1967) and the communication-evoked imagery measures. Second, we examined whether the proposed measure of communication-evoked imagery accounted for differences between high and low imagery ad groups once the effects of an individual's ability to imagine were accounted for.

Correlational Results. One would expect that the ability to imagine and the amount of imagery evoked by a particular communication to be positively correlated, but only slightly so since the degree of imagery evoked by a message is also a function of the individual's familiarity and knowledge with the message content as well as imagining ability. On the other hand, one would expect stronger correlations among the different indicators of communication-evoked imagery processing. Table 3 shows the correlational results for the full datasets of both studies. As can be seen in Table 3, the correlations among the imagery measures were stronger than the measures' correlation with the QMI scale.

With respect to Study 1, we found that the correlations between the Betts' QMI (the measure of innate ability to imagine) ranged from -0.11 to 0.22 with the different indicators of communication-evoked imagery, while the correlations between the four measures were stronger, ranging in absolute terms from 0.34 to 0.61. The same pattern appeared when Study 2 is analyzed. Specifically, correlations between Betts' QMI and the indicators of communication-evoked imagery ranged from -0.18 to 0.23 and the correlations among the communication-evoked imagery measures ranged in absolute terms from 0.30 to 0.63. This analysis supports the contention that the communication-evoked imagery measure is tapping something different than simple ability to imagine.

Experimental Results. The MANOVA model used in the previous section was re-run with the addition of the Betts' QMI as a covariate. For Study 1 (n=51), the Betts' covariate was not significant (F=.63, p=.65). After the covariate was factored out, the Wilks' lambda remained significant (lambda=0.71, F=4.66, p<0.01). Significant differences were found between the high and low-imagery groups on the quantity/ease factor (F=18.07, p<0.01) and the vividness factor (F=10.00, p<0.01). The paleness factor again was not significant (F=2.86, p=.10). The links factor, as before, was marginally significant (F=3.34, p=0.07).

For Study 2 (n=41), the Betts' covariate was significant (F=2.90, p<0.05). Once this influence was taken into account, the overall MANOVA model was still significant (lamda=0.76, F=2.75, p<0.05). In addition, the quantity/ease factor still showed differences between the two groups (F=8.90, p<0.01) as did the vividness factor (F=6.69; p<0.01). Neither the paleness nor links factor were significant (F=2.17, p=.15; F=0.22, p=0.64, respectively). Again, these results are similar to those found when the Betts' covariate was not included.

In sum, it appears that these indicators, especially the quantity/ease and vividness dimensions, are more than a measure of one's ability to engage in imagery processing. While they are positively related to the ability to imagine, they are also sensitive to the content or message elements.


This article describes the development of a set of reliable indicators of the imagery-evoked by a message such as an advertisement. These measures would provide researchers with items which can be used as indicators of the effects of different strategies designed to induce imagery processing. In addition, they may serve as manipulation checks for experimental research investigating imagery effects on consumption-related outcomes.

The results suggest that the self-report measures are reliable indicators of several different aspects of imagery processing. Factor analysis results indicated stable loadings of the items on four dimensions across the two studies. In addition, there is evidence that these measures have some validity in terms of their sensitivity to different imagery-inducing strategies. Finally, these indicators seem to be related to, but different from, indicators of imagery processing abilities.

Additional work still needs to be done to validate these items. Obviously many of the criticisms of self-report measures are applicable here. For instance, Morris and Hampson (1983) suggest that such measures may suffer from social desirability bias. There are the associated difficulties of tapping unobservable activities through any type of obtrusive verbal measure. These measures rely on the subject's awareness of his/her internal processes to reply. Certain physiological approaches may hold promise as alternative indicators of these internal processes. For example, Cacioppo, Petty and Tassinary (1989) have used surface electromyographic response to examine differences in subjects performing an activity and those simply imagining it for evidence of differences in cognitive and affective activity. While physiological measures may be preferable in some cases and may be useful for validating the indicators proposed here, in general, physiological measurement is often not practical as a regular means of imagery processing assessment. Thus, the self-report measures proposed have definite advantages over physiological measures in imagery research.

Tests of convergent and discriminant validity also should include other measures of imaging ability as well as preferences. Specifically, such tests should include assessing differences in processing preferences using the revised Style of Processing Scale (Childers, Houston and Heckler 1985).

Finally, differences in knowledge and familiarity with the products, scenes, schemas described in a message should be examined to determine their relative effects on the imagery-induced as measured by the indicators described here. In particular, these variables may-significantly affect the paleness and links dimensions. For instance, one would expect greater knowledge of a subject to evoke more links with existing information in long term memory. Since the procedures we used did not manipulate or control for knowledge, this could explain why we found no differences on the paleness and links factors.

As with all initial studies, the generalizability of this investigation is unknown. Future research should include different imagery-induction techniques, different subjects and different media. However, the indicators described here can provide researchers with a valuable tool for such assessments.


Alesandrini, Kathy Lutz and Anees A. Sheikh (1983), "Research on Imagery: Implications for Advertising," in A.A. Sheikh (Ed.) Imagery: Current Theory, Research and Application, New York: John Wiley & Sons, 535-56.

Anderson, Craig A. (1983), "Imagination and Expectation: The Effect of Imagining Behavioral Scripts on Personal Intentions," Journal of Personality and Social Psychology, 45, 2, 293305.

Betts, G. H. (1909), The Distribution and Functions of Mental Imagery, Teachers College, Columbia University, New York, New York.

Cacioppo, John T., Richard E. Petty and Louis H. Tassinary (1989), "Social Psychophysiology: A New Look," Advances in Experimental Social Psychology, 22, 39-91.

Carroll, John (1978), 'The Effect of Imagining an Event on Expectations for the Event: An Interpretation in Terms of the Availability Heuristic," Journal of Experimental Social Psychology, 14, 88-96.

Cartwright, Desmond S., Mary E.- Marks, and John H. Durrett, Jr. (1978), ''Definition and Measurement of Three Processes of Imagery Representation: Exploratory Studies of Verbally Stimulated Imagery," Multivariate Behavioral Research, 13, (October) 449473.

Childers, Terry L. and Michael Houston (1982), "Imagery Paradigms for Consumer Research: Alternative Perspectives from Cognitive Psychology," Advances in Consumer Research, 10, 59-64.

Childers, Terry L., Michael Houston and Susan Heckler (1985), "Measurement of Individual Differences in Visual vs. Verbal Information Processing," Journal of Consumer Research, 12 (September), 125-134.

Kisielius, Jolita and Brian Sternthal (1986), "Examining the Vividness Controversy: An Availability-Valence Interpretation," Journal of Consumer Research, 12 (March), 418-31.

Lord, Charles G. (1980), "Schemas and Images as Memory Aids: Two Modes of Processing Social Information," Journal of Personality and Social Psychology, 38 (2), 357-69.

Lutz, Kathy and Richard J. Lutz (1977), "Effects of Interactive Imagery on Learning: Application to Advertising," Journal of Applied Psychology, 62 (4). 493-98.

MacInnis, Deborah J. (1986), "Constructs and Measures of Individual Differences in Imagery Processing: A Review," Advances in Consumer Research, 14, 88-92.

MacInnis, Deborah J. and Linda L. Price (1987), "The Role of Imagery in Information Processing: Review and Extensions," Journal of Consumer Research, 13 (March), 473-91.

McGill, Ann L. and Punam Anand (1989), "The Effect of Imagery on Information Processing Strategy in a Multiattribute Choice Task," Marketing Letters, 1 (1), 7-16.

Marks, David F. (1973), "Visual Imagery Differences in the Recall of Pictures," British Journal of Psychology, 64, 17-24.

Morris, Peter E. and Peter J. Hampson (1983), Imagery and Consciousness, London: Academic Press.

Paivio, Allan (1968), "A Factor-Analytic Study of Word Attributes and Verbal Learning," Journal of Verbal Learning and Verbal Behavior,7, 41-49.

Paivio, Allan (1976), "Imagery in Recall and Recognition," in John Brown (Ed.) Recall and Recognition, London: John Wiley & Sons, 103-129.

Richardson, A. (1969), Mental Imagery, New York: Springer 1969.

Sheehan, Peter (1967), "A Shortened Form of Betts' Questionnaire Upon Mental Imagery," Journal of Clinical Psychology, 23, 386-389.

Sherman, Steven, Kim S. Zehmer, James Johnson, and Edward R. Hirt (1983), "Social Explanation: The Role of Timing, Set and Recall on Subjective Likelihood Estimates," Journal of Personality and Social Psychology, 44, 1127-43.

Smith, Ruth Ann, Michael J. Houston and Terry L. Childers (1983), "Verbal Versus Visual Processing Modes: An Empirical Test of the Cyclical Processing Hypothesis," Advances in Consumer Research, 11, 75-80.