'5' Calories Or 'Low' Calories? What Do We Know About Using Numbers Or Words to Describe Products and Where Do We Go From Here?

ABSTRACT - Marketing information about products is often conveyed by providing numerical or verbal information along specific attributes. Such information is the basic input to consumer decision making that is utilized to make higher-level decisions. This paper reviews empirical work on numerical and verbal information with the aim of synthesizing past research in terms of what we know and where we go from here. In keeping with this goal, the review of empirical research is organized in terms of different elements of decision making, specifically, information search, comparisons, memory, and evaluations. Details on the empirical design of each study reviewed here are provided to enable comparisons across studies. Insights drawn from each area are synthesized in a discussion of theoretical implications and future research directions in terms of dimensions along which numerical and verbal information differ and the impact of ability and motivation to process information.

Citation:

Madhubalan Viswanathan and Terry L. Childers (1997) ,"'5' Calories Or 'Low' Calories? What Do We Know About Using Numbers Or Words to Describe Products and Where Do We Go From Here?", in NA - Advances in Consumer Research Volume 24, eds. Merrie Brucks and Deborah J. MacInnis, Provo, UT : Association for Consumer Research, Pages: 412-418.

Advances in Consumer Research Volume 24, 1997      Pages 412-418

'5' CALORIES OR 'LOW' CALORIES? WHAT DO WE KNOW ABOUT USING NUMBERS OR WORDS TO DESCRIBE PRODUCTS AND WHERE DO WE GO FROM HERE?

Madhubalan Viswanathan, University of Illinois, Urbana-Chamapign

Terry L. Childers, University of Minnesota

ABSTRACT -

Marketing information about products is often conveyed by providing numerical or verbal information along specific attributes. Such information is the basic input to consumer decision making that is utilized to make higher-level decisions. This paper reviews empirical work on numerical and verbal information with the aim of synthesizing past research in terms of what we know and where we go from here. In keeping with this goal, the review of empirical research is organized in terms of different elements of decision making, specifically, information search, comparisons, memory, and evaluations. Details on the empirical design of each study reviewed here are provided to enable comparisons across studies. Insights drawn from each area are synthesized in a discussion of theoretical implications and future research directions in terms of dimensions along which numerical and verbal information differ and the impact of ability and motivation to process information.

This paper reviews empirical work on numerical and verbal information with the aim of synthesizing past research in terms of what we know and where we go from here. The review of empirical research is organized in terms of different elements of decision making, specifically, information search, comparisons, memory, and evaluations, in order to isolate factors that affect one or more elements of the decision making process and provide insight into different elements of consumer decision making. Insights drawn from each area are synthesized in a discussion of theoretical implications and future research directions. This review is selective in nature and does not include several studies in consumer behavior that have not directly compared numerical versus verbal information such as research on nutrition information (cf., Levy et al., 1985), pricing (Mazumdar and Monroe, 1990), and alpha-numeric brand names (Pavia and Costa, 1993).

INFORMATION SEARCH

Several studies have examined differences in the information search process for numerical versus verbal information. Stone and Schkade (1991) argued that it is easier to perform certain operations such as computing differences on numerical labels (i.e., numbers on a rating scale) when compared to verbal labels (e.g., the meaning of the subtraction between one verbal label and another such as 'good’ and 'very good’ is not clear). Therefore, a greater degree of attribute-based processing (i.e., search across brands within an attribute) as well as less processing time was predicted for numerical information. The authors used a task where subjects were required to choose the best information system from a set of alternatives described along four attributes (Table). Five levels of a rating scale either in numerical (i.e., 2, 4, 6, 8, and 10) or in verbal (i.e., very poor to excellent) modes were used to describe product attributes and subjects were instructed about the correspondence between numerical and verbal labels. Moreover, the study manipulated the mode of information between tasks in that the same subjects participated in two sessions scheduled six days apart, one involving numerical information and the other, verbal information. Task complexity (i.e., 2, 4, or 8 alternatives) as well as similarity of alternatives (high versus low levels of similarity) were manipulated within subjects. The study was administered using computers with subjects being able to access a piece of information from a matrix. Concurrent protocols were collected for several sessions. The authors found a greater degree of attribute-based search for numerical when compared to verbal information as well as directionally less processing time for numerical information and directionally fewer pieces of information examined for numerical information (Table). These results were consistent with the rationale that less effort is required to process numerical information.

Huber (1980) used a task where subjects chose the best candidate for a post using a design where the mode of information (i.e., numbers on a rating scale from 1 to 9 or verbal labels such as bad and optimal), the number of alternatives, and the number of dimensions were manipulated. Subjects completed several tasks in random order with each of the factrs mentioned above being manipulated across these multiple tasks. Concurrent protocols were collected. Comparisons such as calculating differences or finding the maximum value within attributes were performed more frequently on numerical information while evaluations were made more frequently on verbal information.

Viswanathan and Narayanan (1992) argued that the conclusions drawn from the Stone and Schkade (1991) paper may not generalizable to the kind of numerical information often used in marketing, i.e., unit-specific numerical information (e.g., 200 calories). They argued that, unit-specific numerical information unlike numerical ratings on a scale, do not necessarily convey equal intervals at a psychological level. Therefore, the computation of differences may not be as meaningful an operation to perform on such information. Moreover, unlike ratings on a scale, unit-specific numerical information has meaningful reference points that consumers can relate to in order to interpret numerical values (e.g., interpreting "200 calories" by using a meaningful reference point rather than comparing across brands). The authors used a design where task (i.e., choice versus learning) was manipulated between subjects and information mode (numerical versus verbal; e.g., display width of '12 digits’ versus 'wide’) was manipulated within subjects. Information was presented on four fictitious brands of calculators along four attributes using a matrix display on a computer. The sample consisted of undergraduate students who were likely to be knowledgeable about the product category. In contrast to the Stone and Schkade study, neither a greater degree of attribute-based processing nor less processing time was found for unit-specific numerical information when compared to verbal information (Table).

Research on information search brings out the importance of the distinction between numerical information on specific units of measurement and artificial or preprocessed numerical information (Figure). This research also suggests the importance of consumers’ knowledge in assessing numerical labels (Figure). When consumers are knowledgeable about product attributes, they may be able to interpret numerical labels without engaging in attribute-based search (i.e, Viswanathan and Narayanan’s (1992) sample). However, lower levels of knowledge may necessitate attribute-based processing for unit-specific numerical information as well.

COMPARISONS

The comparative judgment task from research in psychology (Banks, 1977) requires subjects to compare stimuli on a dimension and make judgments based on the magnitudes of the stimuli along that dimension. Several robust findings from research on comparisons in cognitive psychology across a range of dimensions include the distance effect (i.e., faster and more accurate comparisons with increasing distance between pairs of stimuli being compared).

TABLE

SELECTIVE SUMMARY OF EMPIRICAL STUDIES

In a study which directly compared numerical versus verbal labels, Jaffe-Katz et al. (1989) examined comparisons of pairs of numerical labels, pairs of verbal labels, and pairs of numerical/verbal labels. Subjects were asked to choose the higher (or lower) of a pair of labels. Faster comparisons were observed for pairs of numerical when compared to verbal probability expressions (Table). Such a finding was argued to occur due to the the relatively precise nature of numerical expressions which leads to lesser overlap between a pair of numerical expressions. Several robust findings from research on comparisons in cognitive psychology were also found such as the occurrence of the distance effect. The authors explain their findings in terms of a modified version of the reference point model (Holyoak, 1978) which suggests that comparative judgments of a pair ofstimuli are based on the ratio of the distance of each stimulus from a reference point. The distance effect occurs due to multiple comparisons of distances of a pair of stimuli from a reference point being made until a decision is reached, with the instructions for choosing higher or lower stimuli leading to the high or low ends of the continuum serving as reference points, respectively. As the distance between a pair of stimuli increases, fewer comparisons will be required to reach a decision.

FIGURE

SUMMARY OF PROPOSED RATIONALE FOR FINDINGS

Viswanathan and Narayanan (1994) examined labels describing product attributes using product information about attributes of calculators (e.g., display width of '12 digits’ versus 'wide’). Subjects (undergraduate students) compared pairs of labels describing calculators along specific attributes, a product category that they were likely to be knowledgeable about. Similar to the Jaffe-Katz study, the authors found that comparisons of pairs of numerical labels take less time than comparisons of pairs of verbal labels or numerical/verbal label-pairs. However, unlike the Jaffe-Katz study as well as past research in cognitive psychology (cf., Banks, 1977), several robust effects from past research were not found for comparisons of numerical labels. For example, the distance effect occurred consistently only for verbal labels, and not for numerical labels (Table). On the basis of this difference, the authors argue that the results for comparisons of numerical labels cannot be explained in terms of the reference point model (Holyoak, 1978). Rather, they argue that numerical labels may have been compared based on the sizes of numbers involved without attention to the distances from the labels to the reference point implied in the instruction.

These findings suggest that even though subjects were knowledgeable about the product category and had the ability to interpret the numerical labels, they may have engaged in surface level processing (Figure) in contrast to studies on comparative judgments in psychology. Viswanathan and Narayanan (1994) suggest that product dimensions may be distinct from other dimensions used in psychology such as magnitudes of digits and probability expressions in that numerical labels describing product attributes are generally bounded by a narrow context (i.e., the product category in question) rather than being relevant across a broad range of situations as may be the case for the dimensions used in psychology. Consequently, consumers may have less experience with and less knowledge of labels describing product attributes than with dimensions such as digits and probability expressions though they may have high knowledge about a product category in a relative sense. Therefore, the effect of product knowledge may be moderated in consumer settings by motivation to expend effort required to interpret a label’s meaning.

MEMORY

Some research has examined memory differences between numerical and verbal information. Childers et al. (1992) used a design where subjects learned fictitious brand information about calculators with numerical versus verbal information being manipulated within subjects. Subjects then performed a speeded recognition task where the mode of information was again manipulated to be either numerical or verbal. Numerical information was recognized faster than verbal information when both types of information were presented in the same mode at recognition as well as in a different mode at recognition (Table).

Viswanathan and Childers (1995) argued that numerical information, such as 32 mpg., is a number in the context of a unit of measurement and consequently more specifically linked to an attribute through its unit of measurement (e.g., mpg. for mileage). Verbal information, such as 'high mileage’, is a generic descriptor that readily conveys meaning (i.e., degree of highness on mileage). Consequently, numerical information was argued to be easier to distinuish, i.e., less likely to be confused with similar information on another attribute, and, therefore, easier to encode and/or retrieve than verbal information. They used a design where task (i.e., learning versus choice or judgment) was manipulated between subjects and information mode (numerical versus verbal information) was manipulated within subjects. Four fictitious brands of calculators were described along four attributes, with equal proportions of numerical versus verbal information along each attribute. Information was presented to subjects sequentially in a brand-based format using computers. Numerical information was found to require less time to process than verbal information during a learning task, and was subsequently recognized faster and more accurately, and recalled more accurately than verbal information. Several of these differences persisted for a choice and a judgment task. Additional experiments showed that differences between numerical and verbal information during learning were decreased or eliminated if information along an attribute was presented either numerically or verbally for all brands (i.e., because such a presentation clearly linked a particular verbal anchor (e.g., narrow-wide) to a specific attribute (e.g., display width)) or by presenting numerical information in a more generic form that is similar to verbal information (i.e., numbers on a rating scale) (Table).

Viswanathan (1994) assessed the influence of summary information (e.g., the mean of all brands along an attribute) in facilitating the usage of numerical nutrition information by consumers. The design involved four groups of subjects assigned to the following conditions; numerical nutrition information, numerical nutrition information with the mean as summary information, numerical nutrition information with the range as summary information, and verbal nutrition information. Information on fictitious brands of cereal along three attributes was presented using computers. Following exposure to information on each brand in one experiment and following exposure to information on all brands in another experiment, subjects completed scales relating to evaluations of each brand. Subjects then completed a free recall task and a recognition task. The provision of summary information along with numerical nutrition information led to greater accuracy of recall, as well as greater recall of numerical information in a verbal form. Verbal presentation of information also led to greater recall accuracy than numerical presentation (Table). Verbal presentation of information also led to greater recognition accuracy than numerical presentation with or without summary information. These results represent a reversal of the advantages for numerical information reported by Viswanathan and Childers (1995).

Viswanathan (1995) used a within subjects design where subjects were shown information on four fictitious brands of cereal along four attributes, two of which were presented numerically and two verbally, and asked to provide judgments. Therefore, numerical versus verbal information were in direct competition. Following exposure to information on each brand, subjects completed scales relating to evaluations of each brand. Subsequently, subjects completed an attribute rating task where they rated each fictitious brand along each attribute, and then a free recall task. Three groups of subjects were assigned to different conditions where the information on numerically presented attributes was presented without summary information, with the mean, or with the range of all brands as summary information. Accuracy of ratings of brands based on attribute information, and accuracy of recall of brand information were higher for the attributes presented verbally rather than numerically (Table). These advantages appeared to persist even when numerical information is presented with summary information to facilitate its interpretation.

Several differences and similarities in these studies on emory are noteworthy. Comparing the Viswanathan and Childers (1995) experiments where verbal information involved the use of unique anchors for each attribute (e.g., narrow-wide for display width, long-short for battery life, etc.) to the nutrition studies (Viswanathan 1994; 1995), it is possible that the description of all information on an attribute verbally in the nutrition studies may have contributed to the reversal of effects. However, the use of non-specific verbal labels across all attributes in the nutrition studies suggests that the effect of this factor may have been minimal.

The nutrition studies involved instructions to perform a judgment task similar to some experiments reported by Viswanathan and Childers (1995) where the advantages for numerical information persisted for a judgment task as well. Moreover, the reversal of effects in the nutrition studies were found when judgment ratings were collected after exposure to each brand, as well as after exposure to all brands as in the Viswanathan and Childers (1995) experiments. Therefore, task instructions and sequencing of evaluations may not have contributed to the reversal of effects.

A key difference between the studies that may explain the results is the subjects’ level of prior knowledge about cereals versus calculators and their consequent ability to interpret numerical information (Figure). Because subjects (undergraduate students) are likely to know about and own a calculator, they may have been able to interpret the numerical information provided to them on attributes when compared to subjects in the nutrition studies and their ability to interpret numerical labels describing, say, fiber content, of cereals. Consequently, they may have used numerical information to a lesser degree than when it was provided with summary information or when nutrition information was presented verbally leading to lesser subsequent recall and recognition for numerical information.

Motivation to process numerical information appears to be an important factor in understanding memory for numerical versus verbal information (Figure). As with the studies on comparisons, even with the availability of summary information to interpret numerical information, verbal presentation led to better memory suggesting that consumers were more likely to use verbal information rather than expend the effort to interpret numerical labels and extract meaning from it.

USAGE AND EVALUATION

Several researchers have examined the impact of numerical versus verbal information on product evaluations. Beltramini and Evans (1985) compared quantitative versus qualitative information on believability. A national sample of auto registrants completed a questionnaire containing print ads for a car with quantitative (e.g., "60 out of 100 consumers prefer ...") versus qualitative (e.g., "most consumers prefer ...") information. Unexpectedly, qualitative information was found to be significantly more believable than quantitative information. Yalch and Yalch (1984) argued that, because a quantitative message is more complex than a qualitative message, it may reduce consumers’ motivation to process the message and increase the likelihood of processing based on peripheral cues (see also Witt, 1976). They examined the effect of source expertise on attitudes for numerical versus verbal information by using a design that manipulated source expertise at two levels (expert versus non-expert) and the degree of message quantitativeness at two levels. Subjects viewed cmmercials inserted within a program which discussed the advantages of the automatic teller at a local bank with message quantitativeness being manipulated by the amount of numerical data (quantitative-"...many people do 95% of their banking..." versus qualitative-"...many persons do virtually all their banking...."). Subjects completed scales relating to their attitudes toward the message. An interaction between source expertise and quantitativeness was found, with no difference in attitudes for expert versus non-expert sources for qualitative messages and a significantly more favorable attitude toward the message for expert when compared to non-expert sources for quantitative messages (Table).

Artz and Tybout (1991) argued that numerical versus verbal information may require different source characteristics in order to be believable. Because numerical information is precise in nature, source expertise is likely to enhance believability. Because verbal information is evaluative, trustworthiness of source is likely to enhance believability. The authors used a between subjects design where information mode (numerical-"...60% reduction in the computer delays..." versus verbal-"...significant reduction in the computing delays..."), source expertise (expert/non-expert), and source trustworthiness (trustworthy/untrustworthy) were manipulated. Subjects read a brief introduction to a micro-computer utility and then indicated their attitude toward the product. Consistent with the predictions, expertise had a significant effect on attitudes only for numerical information as in the Yalch and Yalch (1984) study and trustworthiness had a significant effect on attitudes only for verbal information (Table).

Scammon (1977) compared adjectival versus percentage of Recommended Daily Allowance presentations of nutrition information using a between subjects design. Subjects viewed 30 second commercials for two brands of peanut butter with nutritional labels from packages being superimposed on the last 6 seconds of each commercial, and then completed a questionnaire about the information they saw. Scammon (1977) found that adjectival rather than percentage descriptors of nutritional information led to more accurate identification of most nutritious brands (Table). Greater decision satisfaction and less need for information were found with the use of percentage information. The author points out that percentage versus adjective information differ in the extent to which percentage information is unprocessed and adjective information is preprocessed.

Viswanathan (1994; 1995) compared verbal versus numerical presentations in facilitating the usage of nutrition information. The valence of numerical versus verbal information was manipulated for each brand such that it was positive for attributes presented verbally and negative for attributes presented numerically or vice versa. Differences between brands on overall ratings were used to assess the weight given to numerical versus verbal information. Results of several experiments suggested that verbal information may be given greater weight in judgments of healthiness. These advantages persisted even when numerical information was presented with summary information to facilitate its interpretation (Table).

Svenson and Karlsson (1986) examined attractiveness of student apartments using a design where subjects were shown information along three attributes with the assignment of numerical versus verbal information to attributes manipulated between subjects. They found weak support for more weight being given to numerical information in overall judgments, with this effect being obtained for poor alternatives.

In summary, past research suggests that the believability of numerical versus verbal information may be moderated by source characteristics (Figure). Moreover, verbal information which directly conveys meaning may enhance the accuracy of overall judgents. More weight may also be given to verbal information in overall judgments. These results appear to hold when consumers’ knowledge about a product category is not high. Consumers who lack the ability to interpret numerical information in terms of its meaning may be motivated to process verbal information which is preprocessed (Scammon 1977), when compared to numerical information.

THEORETICAL IMPLICATIONS AND FUTURE RESEARCH

Two categories of factors are noteworthy in examining the research across different elements of decision making; differences in numerical and verbal information along certain dimensions, and ability and motivation to process numerical information in terms of its meaning. Research on information search brings out the distinction between unit-specific numerical information and numerical ratings along the dimension of ease of computations (Figure) as well as the importance of knowledge about a product category and the consequent ability in assessing numerical magnitudes along attributes. Research on comparisons of product labels suggested that the effect of product knowledge may be moderated by motivation to expend effort required to interpret a label’s meaning.

Research on memory suggests additional dimensions that distinguish between numerical and verbal information; the degree to which a magnitude is linked to a specific attribute and the degree to which a magnitude readily conveys meaning (Figure). Consumers’ ability to interpret numerical labels appears to be central to the different results found across the studies. Advantages for numerical information may occur when consumers’ have the ability to interpret numerical labels and may be reversed when consumers do not have such ability. Even with ability, motivation to process numerical information in terms of its meaning is an important factor. Hence, advantages for verbal information were found even when numerical information was presented with summary information to facilitate its interpretation. Research on evaluations suggests that more weight may also be given to verbal information in overall judgments when consumers’ ability to interpret numerical labels is low (Figure).

Numerical and verbal information appear to differ on some important dimensions that influence their processing including the degree to which a magnitude is linked to a specific attribute, and the degree to which a magnitude readily conveys meaning. Numerical information along a unit of measurement versus generic verbal descriptors represent extremes along these dimensions. Numerical ratings are like generic verbal descriptors, whereas percentage information, such as % of USDA, are similar to numerical information along a unit of measurement in not conveying meaning directly, and similar to generic verbal descriptors in not being specific to an attribute. Certain verbal magnitudes may be relatively more specifically linked to an attribute to the extent that they are exclusively associated with it (e.g., "roomy" interior for an automobile, on the attribute, interior space). Certain numerical magnitudes along a unit of measurement may convey meaning more directly (e.g., the "1 calorie" cola which directly conveys very low calorie content). Ease of computations is another factor on which verbal information may represent one extreme and numerical ratings the othe with unit-specific numerical information and % of USDA being in between. Believability is another dimension on which numerical and verbal information may differ that is contingent on factors such as source characteristics.

Ability to process numerical information as well as motivation appear to be central to understanding the studies reviewed here. Central to the processing of numerical information is the ability to derive meaning from it using prior knowledge or external reference information. Lacking this ability as well as the motivation to acquire and/or use reference information, consumers may depend more on verbal information. Consumers may also process numerical information at a surface level in terms of the sizes of numbers involved. On the other hand, with a high level of ability to interpret numerical labels almost effortlessly using prior knowledge, consumers may depend more on numerical information. With intermediate levels of prior knowledge where some effort is required in comparing numerical information to reference information available in memory, motivation may play a role in determining whether numerical information is processed in terms of its meaning. A similar scenario may occur when reference information is provided to consumers with low levels of ability to enable them to interpret numerical information. Such interpretation may require effort and consequently lead to greater dependence on verbal information if motivation is low.

A parallel can be drawn to the Elaboration-Likelihood Model (Petty and Cacioppo, 1981) wherein processing numerical information in terms of its meaning roughly overlaps with central processing and surface level processing or lack of processing of numerical information roughly overlaps with peripheral processing. The role of ability can be distinguished in terms of consumers’ prior knowledge versus the provision of external information that enables central (i.e., meaning level) processing, with the latter perhaps requiring greater effort. The role of motivation to engage in central (i.e., meaning level) processing in light of task characteristics are noteworthy. When the task can be accomplished without meaning level processing, consumers may not be motivated to do so even when they have the ability. For example, consumers may compare two brands of cereal with fiber content of 1 and 2 grams per serving and choose the one with 2 grams in the belief that it has "high" fiber content when it is below average on fiber content for cereals. Such "peripheral" processing may be particularly likely for consumers who lack ability and may persist even when external information is available to interpret numerical information. It may also occur when consumers have ability based on prior knowledge but need to expend some effort to interpret numerical information, i.e., engage in "central" processing.

Key avenues of future research include an examination of different types of numerical and verbal product information. Key characteristics of magnitude information suggested here should be considered in examining the processing of numerical versus verbal information. Numerical information ranging from ratings on a scale to unit-specific numerical information and percentage information should be examined as should descriptive verbal information (such as "high" calories), evaluative verbal labels (such as "good" on calories), and the use of unique versus generic verbal anchors to describe attributes. Whereas past research has focused on only one or a few elements of consumer decision making, all elements should be studied in future research. Research should also examine the impact of information load (i.e., the number of attributes and brands).

Future research should also examine two key aspects of processing; ability and motivation to process numerical information. A host of issues relating to different elements of consumer decision making need to be examined while mnipulating the level of product knowledge and the consequent ability to interpret numerical information along an attribute (i.e., the availability of a reference point). The effects of providing reference information to consumers with low knowledge need to be examined in terms of consumers’ motivation to interpret numerical information versus using simpler verbal information. Motivation also appears to be an issue with knowledgeable consumers who may still engage in more superficial processing rather than expend effort depending on the nature of the task.

Research should continue to uncover empirical effects that provide a basis for theorizing about numerical and verbal information. Boundary conditions as well as reversals of effects need to be identified for all elements of consumer decision making. If the research reviewed here is an indication, this area of research offers several interesting empirical findings that may be moderated by various factors. In conclusion, research on numerical and verbal information offers important insight into consumer decision making and interesting avenues for future consumer research.

REFERENCES

Artz, Nancy, and Alice Tybout (1991), "Numeric and Verbal Information and Source Credibility," Working Paper.

Banks, W.P. (1977), "Encoding and Processing of Symbolic Information in Comparative Judgments," In G.H. Bower (ed.), The Psychology of Learning and Motivation, 11, 101-159, New York: Academic Press.

Beltramini, Richard F., and Kenneth R. Evans (1985), "Perceived Believability of Research Results Information in Advertising," Journal of Advertising, 14 (3), 18-31.

Childers, Terry, Esra Gencturk, and Janette Shimanski (1992), Issues in the Organization and Representation of Consumer Information in Memory, Unpublished Manuscript.

Holyoak, K. J. (1978), "Comparative judgments with numerical reference points," Cognitive Psychology, 10, 203-243.

Huber, Oswald (1980), "The Influence of some Task Variables on Cognitive Operations in an Information Processing Decision Model," Acta Psychologica, 45 , 187-96.

Jaffe-Katz, A., D. Budescu, and T. S. Wallsten (1989), "Timed magnitude comparisons of numerical and nonnumerical expressions of uncertainty," Memory and Cognition, 17 (3), 249-264.

Levy, Alan S., Odonna Mathews, Marilyn Stephenson, Janet E. Tenney, and Raymond E. Schucker (1985), "The Impact of a Nutrition Information Program on Food Purchases," Journal of Public Policy and Marketing, 4, 1-13.

Mazumdar, Tridib, and Kent B. Monroe (1990), "The Effects of Buyers’ Intentions to Learn Price Information on Price Encoding," Journal of Retailing, 66 (1), 15-32.

Pavio, Teresa M., and Janeen Arnold Costa (1993), "The Winning Number: Consumer Perceptions of Alpha-Numeric Brand Names," Journal of Marketing, 57, 85-98.

Petty, R. E., and T. J. Cacioppo (1981), Attitudes and persuasion: Classic and contemporary approaches, Dubuque, IA: Brown.

Scammon, Debra L. (1977), "'Information Load’ and Consumers," Journal of Consumer Research, 4 (December), 148-155.

Stone, Dan, and David Schkade (1991), "Numeric and Linguistic Information Representation in Multiattribute Choice," Organizational Behavior and Human DecisionProcesses, 49, 42-59.

Svenson, Ola, and Gunnar Karlsson (1986), "Attractiveness of decision alternatives characterized by numerical and non-numerical information," Scandinavian Journal of Psychology, 27, 74-84.

Viswanathan, Madhubalan (1994), "The Influence of Summary Information on the Usage of Nutrition Information," Journal of Public Policy and Marketing, 13 (1), 48-60.

Viswanathan, Madhubalan (1995), "A Comparison of the Usage of Numerical Versus Verbal Nutrition Information by Consumers," Advances in Consumer Research, Vol. 23, Kim Corfman and John Lynch eds., Provo, UT: Association for Consumer Research, p. 277-81.

Viswanathan, Madhubalan, and Terry Childers (1995), "Processing of Numerical and Verbal Product Information," Working Paper, University of Illinois.

Viswanathan, Madhubalan, and Sunder Narayanan (1992), "Processing Numerical Versus Verbal Attribute Information: A Study Using Information Acquisition Patterns," Marketing Letters, 3 (2), 201-208.

Viswanathan, Madhubalan, and Sunder Narayanan (1994), "Comparative Judgments of Numerical and Verbal Attribute Labels," Journal of Consumer Psychology, 3 (1), 79-101.

Witt, William (1976), "Effects of Quantification in Scientific Writing," Journal of Communication, 26, 67-69.

Yalch, Richard F., and Rebecca Elmore Yalch (1984), "The Effect of Numbers on the Route to Persuasion," Journal of Consumer Research, 11, 522-527.

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