Content-Type: text/html

Running Head: Web Context Effects
Web Context Effects on Perceptions of Product Attributes














Jung-Gyo Lee

Doctoral Student
School of Journalism
University of Missouri-Columbia
Columbia, MO 65211












A manuscript submitted to the Advertising Division of AEJMC for possible presentation at the convention to be held in Miami, Florida, August 7-10, 2002.

All correspondence to:

Jung-Gyo Lee
3705 Forum Blvd. #513
Columbia, MO 65203
Email: <a href="/cgi-bin/wa?LOGON=A3%3Dind0209C%26L%3DAEJMC%26E%3D8bit%26P%3D2889653%26B%3D--%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D%253D_11591531%253D%253D_%26T%3Dtext%252Fplain%3B%2520charset%3Diso-8859-1" >[log in to unmask]</a>
Tel: 573-817-3234
Web Context Effects on Perceptions of Product Attributes

This research reports how product attributes primed by contextual Web pages affect consumers' judgments of products presented in subsequent Web pages in terms of product quality.  It also looks at the moderating effects of individuals' own levels of product involvement on assimilation and contrast processes.
As expected, extremity of the context and ambiguity of the target brand were found to exert significant influences on judgment of product information.  Product involvement moderated this effect.  The results are discussed in terms of assimilation and contrast theory that has been developed in applications to social judgment, but here is shown to be applicable to a consumer domain, Web context.
Running Head: Web Context Effects


Web Context Effects on Perceptions of Product Attributes

Introduction
The Internet experienced exponential growth in penetration rate and technological evolution, with 154 million users in the U.S. and 407 million users worldwide as of November 2000 (Nua Internet Surveys 2000). As the Internet grows, Net-based marketing activities, facilitated by commercial Web sites, are redefining the concept of consumer shopping and information search behaviors. Owing to its unique features such as convenience, interactivity, user-friendliness, rich information, and versatile multi- media capabilities, the Web is becoming a viable medium for purchasing products as well as gathering product information. According to the report released by ActiveMedia (2000), online sales revenues derived from the Web was estimated at $ 56 billion in 2000, which is growth of 103 percent from 1999. This number is projected to reach $ 1.1 trillion by 2010.
As more and more consumers go to the Web to look for product information and purchase online, increasing importance is placed on the issues regarding how consumers judge product information on the Web and what factors are involved in this process. Because technological advances in the Web make it possible for marketers to incorporate promotion and transaction capabilities, effective marketing communications can generate immediate sales online.
One possible influence on a consumer's judgment of product information on the Web is the context in which the judgment is made. Empirical findings in marketing and advertising domains suggest that the context in which an object is presented can exert a considerable impact on an individual's judgment of that object by priming certain category attributes (Herr1986; Martin et al. 1990; Meyers-Levy and Sternthal 1993; Yi 1993). Because our interpretations and judgments of objects (e.g., products) are experienced in a certain context, it seems that there would be very little judgments that occur independent of context. A product is "good", a price is "high", and a program is "interesting", but only in the context of other products, prices, and programs.
In a similar vein, advertisements or product information on the Web exist within a context. When consumers search product information on the Web, they are often exposed to a series of Web pages. Because the interactive and hyper-linked nature of the Web allows consumers to access a variety of product information at a relatively low cost and to compare different product offerings from different sources regardless of time and distance, information flow on the Web can be conceived of more context-driven than that of traditional media. This implies that more information on alternative products can be involved in consumers' product evaluation and decision-making processes. However, for marketers, this means more clutter of competing messages and greater price competition.
A key research issue of interest to this study revolves around the Web context in which consumers judge product information (see Figure 1). In particular, this study attempts to examine how product attributes primed by contextual Web pages (i.e., preceding Web pages) affect consumers' judgments of products presented in subsequent Web pages in terms of product quality. In this study, "context" refers to a Web page that consumers encounter initially, whereas "target" denotes the Web page to which they are exposed subsequently.
Prior studies in the priming paradigm suggest unobtrusive exposure to an object or stimulus in a certain category can increase the likelihood of retrieval of information relevant to that category from memory and subsequent use (Herr 1989; Srull and Wyer 1980). Empirical evidence lends credence to the view that two types of priming effects exist. First, if a target stimulus is judged in the same direction as the context, this phenomenon is referred to as assimilation (Herr et al. 1983). On the other hand, if the judgment of the target stimulus is made in the opposite direction from the context, this phenomenon is described as contrast (Herr et al. 1983).
This study attempts to illustrate the extent to which the assimilation and contrast process applies to a real world milieu, especially information search and evaluation context on the Web (see Figure 2). Because this study attempts to apply the priming paradigm to a more naturalistic setting (most priming studies have been conducted in a very unnatural context), the findings not only shed light on the generalizability of priming effects, but also enrich our understanding of the role of context in consumers' decision-making process.
This study also attempts to contribute to theory by examining the possibility that individual differences in involvement with a product class will play a role in moderating the assimilation and contrast processes. Product involvement has been found to affect the way that consumers process and judge product information (Celsi and Olson 1988; Laurent and Kapferer 1985; Park and Mittal 1985; Petty et al. 1983). Thus, another goal of the present study is to explicate the moderating impact of product involvement on assimilation and contrast processes.

Literature Review
Category Accessibility and Judgment
Many studies in consumer psychology and marketing have demonstrated that a consumer's judgment of a product or advertising can be affected by the context in which the judgment is made (Herr et al.1983; Herr 1989; Meyers-Levy and Sternthal 1993; Meyers-Levy and Tybout 1997; Yi 1990; 1993). In particular, prior research suggests that the influence of context is more pronounced when consumers evaluate ambiguous or unfamiliar objects (Herr 1986; Yi 1990). Some researchers attribute different perceptions or interpretations of ambiguous objects among individuals to a categorization process (Higgins and King 1981). According to this view, people attempt to categorize an object into an existing category in their memory. Categorization is made on the basis of associations with other relevant concepts representing the category in memory. How an object will be interpreted or judged is in part a function of the accessibility of any relevant categories or concepts stored in memory at the time
information is received (Srull and Wyer 1980). If a certain category in memory is more accessible at the time an object is interpreted, there is more likelihood that features or concepts representing that category will be used for the judgment of the object.
Priming theory suggests that unobtrusive exposure to information will enhance the accessibility of that category in memory and increase the likelihood of subsequent use in making a judgment of an object (Herr et al. 1983; Srull and Wyer 1980). For instance, Srull and Wyer (1980) found that participants exposed to exemplars of a particular trait category of people were more likely to use that trait category in judging subsequently presented stimuli (e.g., behavior of an ambiguous person). Higgins et al. (1977) demonstrate a similar process. Participants were unobtrusively exposed to one of four sets of trait terms describing people's personality. After the priming tasks, participants took part in reading tasks that involved judgments of four ambiguous descriptions of a person's behavior. Each description could be interpreted as either positive or negative (e. g., self confidence or conceit). The results indicated that people primed with positive trait terms (e.g., self-confidence) evaluated the person's behavior to be more positive than those primed with negative trait terms (e.g., conceit). These findings imply that a particular trait category activated by the priming task increases the accessibility of that category in memory and subsequent use. If a primed category is
perceived as relevant to subsequently presented stimuli, there is a great likelihood that information related to the primed category will be used for evaluating incoming information (Higgins et al. 1977; Srull and Wyer 1980).
Assimilation and Contrast Effects
Once a particular category or concept has been activated by exposure to a certain context, it may then affect the judgment of incoming information relevant to that category. Empirical findings to date have shown that priming can result in two opposite directions of judgments: assimilation and contrast.
Imagine a situation in which a college student is interested in buying a notebook computer. He or she may look for product information on the Web by browsing relevant Web sites such as manufacturers' sites, computer review sites, online PC magazines and news, and Web sites of computer interest groups, to find the best product. Suppose he initially visits a Web site that promotes very high quality products or brands targeted for professionals. He is not much interested in those computers and may look for other Web sites. Then, he finds another Web site that promotes moderately high quality brands targeted for general users, but the product qualities are relatively lower than those in the first site. He may be more interested in the products appeared in this site than those in the first visited site.
In this situation, his judgment of product information offered by the latter Web site (e. g., product quality) can be affected by the information offered on the former Web site. Put another way, product information offered in the first visited Web site may serve as a criterion for comparison for the judgment of the product category promoted in the subsequently visited Web site. As a result, our hypothetical individual may judge the qualities of computers offered by the latter Web site to be lower, although the actual product qualities are moderately high in a laptop category. On the contrary, if he initially visited a Web site promoting very low quality computer brands, he might rate the moderately high quality products appeared on the second Web site as higher quality products.
Phenomena like this are often referred to as "contrast effects". The intellectual strands of such phenomena are found in assimilation and contrast theory, originating from the classic social judgment theory (Sherif and Hovland 1961). According to this theory, there are two kinds of relationships between a context and a target. First, if there is a positive relationship between the value consumers place on a context and the value they place on a target object that follows the context, this phenomenon is described as assimilation (Martin et al. 1990; Meyers-Levy and Sternthal 1993). In this case, a target object is judged in the same direction as the context. On the other hand, if there is a negative relationship between the value consumers place on a target stimulus and the value they place on a context that precedes the target, this phenomenon is referred to as contrast (Martin et al. 1990; Meyers-Levy and Sternthal 1993). In this case, the target object is judged in the opposite direction from the context.
In some instances, consumers assimilate their judgments of target objects toward the characteristics associated with accessible contextual cues, whereas in other cases, they judge incoming information in the opposite direction from the preceding context. An important question, then, is what factors determine whether assimilation or contrast will occur?  Prior studies on priming effects offer several explanations for the conditions in which assimilation or contrast occurs.
One school of thought suggests that whether assimilation or contrast will occur is in part determined by the degree of feature overlap between a context and subsequent target object (Herr et al. 1983; Herr 1986; 1989). Assimilation effects have been found to be more pronounced when the target is ambiguous or unfamiliar. Contrasts effects have been reported when the target is unambiguous or familiar (Herr et al.1983). Researchers in this stream purport that people tend to judge an unfamiliar or ambiguous object by categorizing it conceptually on the basis of the most accessible information stored in memory at the time the information is received (Srull and Wyer 1980). In evaluating an ambiguous or unfamiliar object, people are likely to use information offered by the context because little information about the target is available in their memory. If there exists considerable feature overlap between contextual cues and target object, the target object will be interpreted as an inst
ance of the category activated by the contextual cues (Herr 1986). As a result, the judgments of the target object will be made toward the primed category, and assimilation will occur.
The assimilation effect has also been found to be more pronounced when an individual is primed with a moderate category rather than a extreme category due to more likelihood of feature matching between a category activated by contextual cues and ambiguous target stimuli (Herr 1986; Meyers-Levy and Sternthal 1993). For instance, Herr (1989) found that people who were primed with a category representing moderate features or values evaluated the attributes of subsequently presented ambiguous target objects (e.g., fictitious cars) in the same direction as the context. On the contrary, people exposed to a category exemplifying extreme attributes or values judged subsequently presented ambiguous targets in the opposite direction from the contextual stimuli. Because the overlap in features between the contextual cues and the target stimuli was small, contrast effects were found.
        The recent work of Martin and his co-workers (1990) represents another explanation of assimilation and contrast processes. Assimilation and contrast can occur without variations in the feature overlap between contextual cues and target objects. Whether assimilation or contrast will occur is in part a function of cognitive efforts made to a judgment task (Martin et al. 1990). In their seminal work, Martin and his co-workers operationalized cognitive resources devoted to the judgment task in terms of individual's need for cognition and tested how this factor played a role in determining the type of context effects (i.e., assimilation and contrast). The results indicated that participants with higher need for cognition tended to show contrast effects. Those with lower need for cognition exhibited assimilation effects (Martin et al. 1990). These results can be accounted for by noting that people with higher need for cognition made considerable cognitive efforts to hold back associati
ons prompted by contextual cues, and judged the target object more critically, whereas participants with lower need for cognition adopted the assimilation strategy requiring less cognitive resources by simply applying the category activated by contextual cues to the judgment of the target (Martin et al 1990).
The preceding discussion leads to a couple of expectations in the present study. First, with regard to the extremity of a category primed, it is expected that consumers who are exposed to context Web sites promoting extreme product categories  (e.g., very high quality or very low quality notebook computers) will be more likely to judge the attributes of products (e.g., quality of a notebook computer) presented in subsequent Web sites in the opposite direction from the context. Because of the extreme nature of brands advertised in the context sites, there is less likelihood of feature matching between product attributes activated by context Web sites and attributes of products advertised in the target sites. In this case, the brands advertised in the target sites are unlikely to be perceived as exemplars of brands advertised in the context sites. Consequently, consumers should use information activated by the context sites as a comparison standard rather than a judgment frame, and contrast effects will occur. On the contrary, consumers who are exposed to Web sites promoting moderate product categories in terms of quality
(i.e., moderately high or moderately low quality laptop brands) will be more likely to judge products presented in subsequent Web sites in the same direction as the context due to the more likelihood of feature matching between product attributes activated by context Web sites and those of products advertised in the target sites, and assimilation is expected.
 Second, with respect to the ambiguity of products advertised in target Web sites, it is expected that consumers' judgments will be more likely to assimilate toward product attributes presented in the context Web sites when they judge unfamiliar products advertised in the target sites. When consumers have little information about the products advertised in the target sites, the most accessible information in memory at the time of judgment (i.e., category attributes recently activated by the context sites) would guide the interpretation of products advertised in the target sites. In this case, consumers will categorize ambiguous products presented in the target sites as instances of products advertised in the context sites and use the product attributes prompted by the context sites as interpretation frames in evaluating the brands advertised in the target sites. Accordingly, following hypotheses are posited:
H1-a: Consumers who are exposed to Web sites promoting a very high-quality brand category will judge the quality of the brand presented in subsequent Web sites to be lower (contrast effects).
H1-b: Consumers who are exposed to Web sites promoting a very low-quality brand category will judge the quality of the brand presented in subsequent Web sites to be higher (contrast effects).

H2-a: Consumers who are exposed to Web sites promoting a moderately high- quality brand category will judge the quality of the brand presented in subsequent Web sites to be higher (assimilation effects).
H2-b: Consumers who are exposed to Web sites promoting a moderately low-quality brand category will judge the quality of the brand presented in subsequent Web sites to be lower (assimilation effects).

H3-a: Assimilation effects on quality perception will be more pronounced when consumers judge ambiguous or unfamiliar brands advertised in target sites.
H3-b: Contrast effects on quality perception will be more pronounced when consumers judge unambiguous or familiar brands advertised in target sites.

As we have seen, assimilation and contrast are influenced by the extremity of contextual cues and the ambiguity of the target. However, this may not invariably be the case. Instead, it seems that the magnitude of priming effects may vary among individual consumers. Indeed, a number of studies have suggested that an individual's involvement with a product class is likely to exert a considerable influence on consumers' processing and evaluation of advertising messages. Unfortunately, however, little empirical research has addressed the moderating role of product involvement in assimilation and contrast process. Thus, this study attempts to examine the possibility that the magnitude of assimilation and contrast effects will be in part contingent on individual differences in involvement with a product class.
Product Involvement and Information Processing

The concept of involvement has received considerable attention in the disciplines of psychology, marketing, and advertising because of its potential for influencing consumer's processing of advertisements, information search behavior, and the decision making process (Laurent and Kapferer 1985; Mittal and Lee 1989; Petty et al.1983). Involvement has been defined in terms of personal relevance or importance of an object (Zaichowsky 1985), a goal-directed arousal (Park and Mittal 1985), motivation to process information (Celsi and Olson 1988; Petty and Cacioppo 1986) and interest in an object or activity (Celsi and Olson 1988; Mitchell 1981).
Some researchers narrowed the concept of involvement to product involvement (Bloch 1981). Zaichowsky (1985) defined product involvement as an individual's "perceived relevance" with a product class, which is derived from intrinsic needs, values, and interests. Lastovicka and Gardener (1979) interpreted product involvement in terms of two underlying components: "normative importance" and "commitment". According to these researchers, normative importance refers to how relevant or important a product category is to an individual's values, whereas commitment denotes "the pledging of binding of an individual to his/her brand choice". Mittal and Lee (1989) conceptualized product involvement as "the interest a consumer find in a product class". They contend that such an interest is based on the extent to which the product class meets the consumer's goals and values. From this perspective, product involvement can be described as an internal state representing individual differences among
consumers (Mitchell 1981; Laurent and Kapferer 1985). According to Mitchell (1981), an involved state possesses two major properties: intensity and direction. Intensity designates the individual's level or degree of involvement, while direction denotes an object with which an individual is involved (Mitchell 1981).
In addition to intensity and direction of involvement, some researchers have focused on delving into a property of product involvement in terms of persistence (e.g., Houston and Rothschild 1977). Persistence can be described as the duration of involvement (Andrews et al. 1990). For instance, Houston and Rothschild (1977) classified product involvement into two types: enduring and situational involvement. The former reveals a general and ongoing concern with a product category, whereas the latter reflects a particular situation such as purchase occasion (Houston and Rothschild 1977).   By the same token, some researchers (e.g., Mittal and Lee 1989) have distinguished between product involvement and purchase involvement. Product involvement entails an interest in ownership or utilization of a product, while purchase involvement concerns an interest in the selection of a brand (Mittal and Lee 1989).
Although varied approaches to product involvement have been discussed by different researchers, there is consensus that involvement with a product class is an enduring trait of an individual, and consumers possess fairly stable levels of involvement with a certain product class (Celsi and Olson 1988). From this perspective, Bloch (1981) conceived product involvement as "a construct which affects consumer behavior on an ongoing basis". In this study, product involvement is operationalized as an ongoing concern with and interest in a particular product class based on an individual's inherent values, goals and needs across all purchase situations (Bloch 1981; Celsi and Olson 1988; Houston and Rothschild 1977).
        It should be noted that product involvement is a distinct construct from product knowledge.  Product knowledge can be defined as the amount of knowledge about a product class that has been stored in an individual consumer's long-term memory (Alba and Hutchinson 1987; Cameron 1990). This concept often has been operationalized in terms of familiarity, expertise, and experience (Alba and Hutchinson 1987). According to Cameron (1990), prior knowledge about a product yields extensive networks of concepts or impressions pertaining to that product in individual's long-term memory. If memory networks are well activated, this will result in a high level of involvement. This notion implies that prior knowledge about a particular product class may sometimes result in a higher level of involvement with that category. However, product knowledge is not a necessary but a sufficient condition for product involvement. Put another way, a high level of involvement can be derived without a great deal of product knowledge (Cameron 1990).
A number of empirical studies have suggested that an individual's involvement with a product class can exert a significant influence on consumers' processing and evaluation of advertising messages. In particular, high involvement has been shown to yield more attention to and better memory of advertisements, better comprehension of messages, greater time devoted to information search, deeper processing of information and more message elaboration (see Celsi and Olson 1988; Laurent and Kapferer 1985; Park and Mittal 1985; Petty et al. 1983). These findings imply that a consumer with high involvement is more likely to process product information actively and deeply than a consumer with low involvement. From this perspective, Laurent and Kapferer (1985) contend that an individual's level of involvement in a product category determines the extent to which he/she will extend processing of advertising messages. In other words, if an individual's involvement with a product category is high
, the amount of cognitive effort spent on processing product information will be substantial. On the other hand, if an individual has a low level of involvement with a product class, he or she will not spend as much cognitive resources on processing advertisements of that product class.
The Moderating Role of Product Involvement
There is support for the idea that there are multiple mechanisms by which information processing ensues (Chaiken 1980; Petty and Cacioppo 1986). One of the most widely embraced theoretical frameworks in this stream is dual process model such as the elaboration likelihood model (Petty et al. 1983), the systematic/heuristic processing model (Chaiken 1980), and the automatic/strategic processing model (Grunert 1996). This school of thought suggests that there are two distinct strategies to process persuasive messages: systematic or central processing versus heuristic or peripheral processing. The use of a systematic strategy involves central and intensive processing of messages, based on critical thinking and extensive elaboration of message claims. The employment of a heuristic strategy leads to peripheral and superficial processing of advertising messages, which is often based on exposure to easily accessible contextual cues such as attractiveness of endorsers and background music
(Petty et al. 1983).
According to the elaboration likelihood model, whether systematic or heuristic processing strategy is adopted depends on two main factors: an individual's motivation and ability to process information (Petty and Cacioppo 1986). Many researchers conceptualize motivation to process information in terms of an individual's involvement (Celsi and Olson 1988; Houston and Rothschild 1977). Petty and Cacioppo (1986) suggest that characteristics of individuals (e.g., issue involvement) play an important role in determining the extent to which they are motivated to process information (Petty et al. 1983; Petty and Cacioppo 1986; Meyers-Levy and Prashant 1999). For instance, Petty and his colleagues (1983) found that the strength of advertising claims had a greater influence on persuasion under high involvement condition than low involvement condition, whereas contextual cues such as the attractiveness of endorsers exerted a greater impact on persuasion under low involvement condition than h
igh involvement condition. Chaiken also implicitly envisions the role of involvement in employing a particular processing strategy. According to her, whether the systematic or heuristic processing strategy is adopted counts on the extent to which recipients perceive messages as personally important (Chaiken 1980). She contends that the systematic strategy is more likely to be employed when recipients feel that messages are personally important to them.
From the above discussion, a couple of additional expectations can be articulated. It is posited that context effects will be more pronounced when consumers have low levels of product class involvement. When consumers hold high levels of product class involvement, priming effects would not be as substantial.
If consumers are highly involved in a product class, they would attempt to process product information presented in both context and target Web sites actively and deeply. Because products advertised in both Web sites are personally relevant and important to consumers' interests, needs and values, consumers will be more motivated to attend to and process product information appeared in both context and target Web sites. Such systematic processing involves much more thoughtful, elaborative and analytical processing of product information (e.g., product quality) than heuristic strategy, so cognitive resources required for this processing will be substantial. In addition, because priority is placed on the accuracy of the judgment when consumers are highly involved with a product class, they are more likely to make extensive and critical category-based comparisons, which are based on prior experiences as well as knowledge of the product category. In other words, consumers with high inv
olvement are more likely to suppress or ignore associations activated by contextual cues.
On the other hand, if consumers are low involved with a product class, they are likely to process product information presented in both context and target Web sites passively. Because products advertised are irrelevant and unimportant to consumers' interests, needs and values, they will not be as much motivated to attend to or process advertising messages actively and intensively as high involvement counterparts. Such heuristic strategy leads to simple, convenient, and intuitive processing of product information. In addition, relatively little cognitive resources would be devoted to processing and judging the product information presented in the target site because little priority is placed on the accuracy of the judgment. In this case, consumers are more likely to apply the associations prompted by the preceding Web site to the judgment of the target. As a result, judgments of product information will be more influenced by the primed context.
Accordingly, the following hypothesis is posited:
H4: Product involvement will moderate context effects in terms of assimilation and contrast. Context effects on perceptions of product qualities will be more pronounced among consumers with lower product class involvement than those with higher product class involvement.

 Method
Design
The study employed a 2 ( 2 ( 2 ( 2 between-subjects design with following factors: (1) extremity of a category: extreme and moderate, (2) direction of product quality: low and high (the first and second factors constituted four quality categories ranging from extremely low quality to extremely high quality), (3) ambiguity of target brand: fictitious and real notebook brands, and (4) involvement with a product class: high and low enduring involvement. Extremity of a category, direction of quality and ambiguity of target brand were manipulated variables. Product class involvement was a measured variable. The dependent variable of this study was quality perceptions of products advertised in target Web sites.
Pretest
The extremity of a category and direction of quality were manipulated by varying brands of notebook computers advertised in context Web sites so that they would prompt different impressions in terms of perceived product qualities. To establish four levels of contextual priming conditions in terms of quality ranges, a pretest was conducted one month before the main experiment. Fourteen notebook computer brands were pre-selected by three judges and tested with the intent of finding four groups of notebook brands that might differ significantly as to perceived product qualities.
137 student participants from the same population that was sampled for the main study were employed to select four categories of notebook brands in terms of product quality. The students rated the qualities of fourteen laptop brands on an 11-point scale of (1) very low quality to (11) very high quality (see Table 1). The students were instructed not to judge notebook brands with which they were not familiar. From this test, four sets of notebook brands were selected as exemplars of four different quality categories for context sites (see Table 2). To enhance the external validity of stimulus materials, two laptop brands were used for each quality category. The notebook brand that participants rated to be moderate in terms of product quality was selected as the target brand (i.e., Hewlett Packard) for the main experiment.
Stimulus Materials
Four online shopping sites specializing in computer-related products were created by a Web publishing specialist as context Web sites. To reduce the possibility that the contents and structures of these four Web sites might influence responses examined in the study, all four context Web sites were created identically in terms of organization and content, except for the brands advertised. Each of four brand categories representing distinct quality ranges was randomly embedded in each of the four test Web sites.
To manipulate the ambiguity of the target brand, two different versions of target Web sites were created using the same content. The only difference between these two sites was the use of either fictitious (Optima) or real brand (Hewlett Packard). These two Web sites were identical except the names of brands advertised. Table 3 summarizes the experimental stimuli.
Measure
Because product involvement operationalized in this study is enduring involvement with a product class, the revised version of Zaichkowsky's Personal Involvement Inventory was adopted for assessing participants' involvement with laptop computers (Zaichkowsky 1985; 1994). This scale was designed primarily to measure consumers' ongoing concern with and interest in a particular product class. The revised version of PII scale consisted of ten 7-point semantic differential items including: important/unimportant, boring/interesting, relevant/irrelevant, exciting/unexciting, means nothing/means a lot to me, appealing/unappealing, fascinating/mundane, worthless/valuable, involving/uninvolving, not needed/needed. Cronbach's ( coefficient for product involvement was .90.
Participants responded to product involvement items on seven point Likert scales.  Then, all the scores from product involvement scale items were summed and divided by the number of items and became the product involvement index scores. On the basis of the average index of the ten involvement items, participants were divided into two groups: low and high product involvement groups. Because literature suggests excluding the middle range index scores would give advantages such as a distinction between the two groups and not standing in the light of criticism of continuity (e.g., Haugtvedt, Petty and Cacioppo 1992), participants whose product involvement scores were placed in the bottom 30 percentiles of the distribution were chosen as a low involvement group, while those whose product involvement scores were placed in the top 30 percentiles of distribution were selected as a high involvement group.
Sample
160 participants were recruited from an introductory advertising class at a large Midwestern university.  Students went to an assigned URL at their convenience and participated in an experiment controlled completely at that URL.  Data were automatically entered into a database. The students were randomly assigned to the eight experimental conditions. The students were told that the purpose of the study was to evaluate a Web site. Immediately after being exposed to each treatment condition, a post-exposure questionnaire was filled in by participants assessing perceptions of product qualities advertised in the target Web sites.
Manipulation Checks
As manipulation checks, one-way ANOVAs and a post-hoc test were conducted for the eight test Web sites to confirm whether the extremity of a category, direction of quality, ambiguity of target object were successfully manipulated in the expected direction. As shown in Table 4, the post-hoc test revealed that mean differences in quality perceptions among the five-laptop categories were all statistically significant. Further, the familiarity score for the real brand  (M = 9.30) was significantly higher than that for the fictitious brand (M = 1.21, T = 13.35, p < .01, see Table 5).
Because the participants' level of prior knowledge about a product class could serve as alternative explanations for the effects being examined in the current study, this variable was measured before the actual experiment and treated as covariate in the model examined.

Results
Data Screening and Statistical Assumptions
Prior to the main analysis, the variables examined in the study were tested for accuracy of data entry, missing values, outliers, and the assumptions of the statistical analysis. Three randomly missing values on quality perception were replaced using Expectation Maximization methods (Tabachnick and Fidell 2001), while other variables, with missing values on more than 5 % of cases were deleted. Two cases with extremely high Z scores on items concerning product involvement were found to be univariate outliers, while no case was identified through Mahalanobis distance with p < .001 as a multivariate outlier (Osterlind and Tabachnick 2001). The two outliers were deleted, leaving 139 cases for analysis. Checks for assumptions of normality, linearity, multicolinearity, homoscedasticity, and homogeneity of error variance were satisfactory.
Extremity of Context and Judgment of Product Quality
The first and second hypotheses involved the influence of context representing four different quality ranges (i.e., extremity of a category ( direction of quality) on judgment of target brand in terms of product quality. Evidence for a contrast effect would be obtained either if participants rated the quality of the target brand to be lower upon exposure to very high quality brand category, or if participants judged the quality of the target brand to be higher upon exposure to very low quality brand category. To examine the effects of four quality categories on judgment of product quality, a 2 (extreme and moderate) ( 2 (high and low) analysis of variance was conducted.
Table 6 demonstrates significant interaction effects between extremity of quality and direction of quality on perceptions of product qualities (F = 4.863, p < .01). The result of an ANCOVA showed that no significant change in error was detected after prior knowledge was added to the model as a covariate (F = 2.93, p > .05).
The statistically significant interaction between extremity of a category and direction of quality corroborated the hypothesized patterns of assimilation and contrast in quality perceptions for all four brand categories. As shown in Table 7, participants exposed to very high quality brands rated the quality of the target brand much lower (M = 6.91) than those exposed to moderately high quality brands (M = 7.87). Likewise, participants exposed to very low quality brands rated the quality of the target brand higher (M = 8.39) than those exposed to moderately low quality brands (M = 8.25). The results also suggest assimilation and contrast effects were more pronounced among participants primed with high quality categories than those primed with low quality categories.
Ambiguity of Target Brand and Judgment of Product Quality
Hypothesis 3 suggested that contrast effects would be more pronounced when participants judged a familiar brand, whereas assimilation would be more expected when participants rated an unfamiliar brand. As shown in Table 8, there was a significant interaction between four quality categories and ambiguity of target brands (F = 2.106, p < .05). The means for this interaction are presented in Table 9.
Figure 4 demonstrates differences in priming effects between unfamiliar and familiar brand conditions. Participants' judgments of the familiar brand (i.e., Hewlett Packard) were more contrasted from the primed contexts than their judgments of the unfamiliar brand (i.e., Optima). In particular, differences in contrast effects were more pronounced when participants were primed with extreme brand categories (i.e., very high quality and very low quality brands). As shown in Figure 4 and Table 9, participants exposed to extremely high quality brand rated the quality of the familiar brand lower (M = 6.80) than that of the unfamiliar brand (M = 7.00). Likewise, participants exposed to extremely low quality brand rated the quality of the familiar brand higher (M = 8.47) than that of the unfamiliar brand (M = 8.31). Therefore, hypothesis 3 was supported.
Product Involvement and Judgment of Product Quality
  Hypothesis 4 posited that product involvement would moderate the priming effects. In particular, it was expected that priming effects would be more pronounced among participants with lower product involvement than those with higher product involvement. The moderating role of product involvement was assessed by following Baron and Kenny's recommendations (1986). According to them, moderator variables specify "when certain effects will hold", whereas mediator variables tell us "how or why such effects occur". A variable can be regarded as a moderator if there is a statistically significant interaction (see path C in Figure 3) between the independent variable and the moderator variable (Baron and Kenney 1986). In addition, it has been recognized that the correlation between the moderator and the independent variable as well as dependent variable should be weak to render a significant interactive effect (Baron and Kenney 1986).
As shown in Table 10, there were significant interaction effects between four quality categories and product involvement on judgment of product quality (F = 2.996, p < .01). The means for this interaction are presented in Table 11.
Figure 5 demonstrates differences in priming effects between low and high product involvement groups. Although assimilation and contrast effects did occur among both low and high product involvement folks, these effects appeared to be more pronounced among participants with lower product involvement than those with higher product involvement.
For instance, in low involvement group, participants exposed to very high quality brands rated the quality of the target brand much lower (M = 5.72) than those exposed to moderately high quality brands (M = 8.00). Likewise, participants exposed to very low quality brands rated the quality of the target brand much higher (M = 8.48) than those exposed to moderately high quality brands (M = 8.00). However, in high involvement group, mean differences between very high (M = 7.69) and moderately high (M = 8.00) quality conditions or very low (M = 8.75) and moderately low (M = 8.45) quality conditions were not as substantial as those of low involvement group (see Table 11 and Figure 5). Therefore, hypothesis 4 was also supported.
Discussion
The present study explores the role of context in judgment of product information on the Web. In particular, this study examines how product attributes primed by contextual Web pages affect consumers' judgments of products presented in subsequent Web pages in terms of product quality. The results suggest that context can exert significant influences on both categorization process and judgment by priming certain category attributes in a real world milieu.
The most important finding is that this study corroborated the relationship between category accessibility in human memory and judgment process noted by Srull and Wyer (1980) in a naturalistic setting (i.e., Web browsing). Priming theory suggests that unobtrusive exposure to information will enhance the accessibility of that category in memory and increase the likelihood of subsequent use in making a judgment of an object (Herr et al. 1983; Srull and Wyer 1980).
Consistent with previous findings in the priming domain (e.g., Herr et al. 1983; Herr 1986), the extremity of the contextual brands and the ambiguity of the target brand had consistently large influences on the assimilation and contrast processes in Web context. In particular, assimilation effects were more pronounced when participants were exposed to the moderate brand categories or the target brand was ambiguous, whereas contrasts effects were consistently reported when participants were primed with the extreme brand categories or the target brand was unambiguous. These results provide support for the Herr (1986; 1989)'s notion that the degree of feature overlap between a context and target object can play a key role in determining whether assimilation or contrast will occur.
The findings clearly indicate that if you are looking for an environment to advertise your product online, consider the context in which your product is advertised in terms of other products and the hyper-linked sequence of presentation so that products advertised are not suffered but benefited from the context. In particular, from an advertiser's point of view, the context that could render contrast effects turned out to be managerially important in the advertising media planning process on the Web.
The findings suggest that exposure to a very high quality brand can have a detrimental effect on the subsequent evaluation of a moderate quality brand, whereas exposure to a very low quality brand can have a beneficial effect on the subsequent evaluation of a moderate quality brand. This is critically important information for the thousands of companies that promote their products on the Internet because marketers always want to ensure that their products or brands are perceived favorably.
 The study also examined the moderating impact of consumer characteristics, namely product class involvement. In concert with findings from the information processing literature, product involvement significantly moderated the contextual priming effects on perception of product quality. As expected, the context effects were found to be more pronounced among participants with lower product involvement than those with higher product involvement.
This pattern of results makes sense from the perspective of the previous literature (e.g., Martin et al. 1990; Meyers-Levy and Tybout 1997) noting whether assimilation or contrast will occur is a function of cognitive efforts made to a judgment task. In this study, it stands to reason that participants with higher product involvement made considerable cognitive efforts to hold back associations prompted by contextual cues, and judged the target object more critically on the basis of prior experiences as well as knowledge of the product category because priority was placed on the accuracy of the judgment. On the other hand, participants with lower product involvement did not spend substantial cognitive resources and simply applied the category activated by contextual cues to the judgment of the target because products advertised were irrelevant and unimportant to them.
The problem for advertisers would be to determine how to identify consumers with different levels of product involvement in that they clearly seem to be more vulnerable to context effects, and then integrate this information with media planning process to generate more favorable evaluations of their products appearing on the Web.
Of course, the exact mechanism by which this occurs remains for future research.  In subsequent studies it would be valuable to test the context effect beyond the judgment of product quality (e.g., judgment of price and attitude toward the brand) and using more than one product category so we can be more assured of the generalizability of the results. And always, it would be important to test these results with an adult sample.

 Table 1
Quality Perceptions of Notebook Brands

IBM
Sony
Apple
Dell
Compaq
Gateway
Toshiba
Samsung
Hitachi
HP
NEC
Fujitsu
Acer
Emachines
Mean
8.08
7.62
6.98
6.96
6.75
6.73
6.53
6.48
6.45
6.35
5.87
5.53
5.01
4.80
SD
1.90
1.77
2.14
2.11
2.12
2.00
2.12
2.21
2.04
1.99
2.20
1.89
2.17
2.30

Table 2
Categories of Notebook Brands

Brand Category
Exemplar Brands
Mean
SD
Very High Quality Brands
IBM and Sony
7.85
1.46
Moderately High Quality Brands
Apple and Dell
6.97
1.69
Moderate Quality
(Target brand)
Hewlett Packard
6.35
1.99
Moderately Low Quality Brands
NEC and Fujitsu
5.70
1.75
Very Low Quality Brands
Acer and Emachines
4.90
1.81

Table 3
Experimental Stimuli

Participant Group
Context Site
Target Site
Group 1
Very High Quality Brands
Real Brand
Group 2
Moderately High Quality Brands
Fictitious Brand
Group 3
Moderately Low Quality Brands
Real Brand
Group 4
Very Low Quality Brands
Fictitious Brand
Group 5
Very High Quality Brands
Fictitious Brand
Group 6
Moderately High Quality Brands
Real Brand
Group 7
Moderately Low Quality Brands
Fictitious Brand
Group 8
Very Low Quality Brands
Real Brand

Table 4
Mean Differences in Quality Perception among Five Notebook Categories

          ***Pair
Mean Difference
df
T value
Sig.
A - B
.88
34
5.933
.000**
B - C
.62
34
2.401
.017*
C - D
.65
34
2.413
.018*
D - E
.80
34
4.730
.000**
                (* p < .05, ** p < .01)
        ***A = Very high quality category
B = Moderately high quality category
C = Moderate category (Target brand)
D = Moderately low quality category
E = Very low quality category
Table 5
Mean Difference in Perceived Ambiguity between Fictitious and Real Brand

     Ambiguity
Mean Difference
df
T value
Sig.
Real brand
M = 9.30
Fictitious brand
M = 1.21

8.09

68

13.355

.000**
(** p < .01)
Table 6
Effects of Four Quality Conditions on Judgment of Target Brand

Dependent
Source
Sum of Squares
df
Mean Square
F
Sig.
Quality Perception
Extremity ( Direction
47.264
3
15.755
4.863
.003**
Prior Knowledge (CV)
9.513
1
9.513
2.936
.089
(**p <. 01)

Table 7
Mean Scores for Quality Perceptions among Four Notebook Categories

Direction
Extremity
Mean
N
SD
Low quality
Moderate
8.39
36
.302
Extreme
8.25
35
.306
High Quality
Moderate
7.89
33
.316
Extreme
6.91
35
.306

Table 8
Interactive Effects of Four Quality Conditions and Ambiguity of Target Brand

Dependent
Source
Sum of Squares
df
Mean Square
F
Sig.
Quality Perception
Extremity ( Direction x Ambiguity
49.027
7
7.004
2.106
.047*
Prior Knowledge (CV)
10.586
1
10.586
3.183
.077
(*p < .05)













Table 9
Mean Differences in Quality Perception between Familiar and Unfamiliar Brands

Ambiguity of Target
Four Quality Conditions
Mean
SD

Very High Quality
6.80
.47
Moderately High Quality
7.81
.46
Familiar Brand
Moderately Low Quality
8.24
.40
Very Low Quality
8.47
.45
Very High Quality
7.00
.41
Unfamiliar Brand
Moderately High Quality
7.94
.45
Moderately Low Quality
8.27
.47

Very Low Quality
8.32
.43

 Table 10
Three-way Interaction between Extremity of Category, Direction of Quality and Product Involvement on Judgment of Product Quality

Dependent
Source
Sum of Squares
Df
Mean Square
F
Sig.
Quality Perception
Extremity ( Direction x Involvement
68.869
7
9.838
2.996
.008**
(**p < .01)
Table 11
Mean Differences in Quality Perception between Low and High Product
Involvement Groups

Ambiguity of Target
Four Quality Conditions
Mean
SD

Very High Quality
5.72
.54
Low Involvement
Moderately High Quality
8.00
.57
Group
Moderately Low Quality
8.00
.60
Very Low Quality
8.48
.45
High Involvement
Very High Quality
7.69
.52
Group
Moderately High Quality
7.90
.54
Moderately Low Quality
8.45
.54

Very Low Quality
8.75
.52
  Figure 1
 Consumer Buying Process on the Web
Problem Recognition






Information Search


Contextual
Priming




Information Evaluation





Decision Making




Post-Purchase Evaluation





Figure 2
Priming Effects in Information Evaluation Stage

Context sites

Target sites

Judgment



Extreme brands
(Very high quality)





Unfamiliar
brand


Familiar
 brand

Assimilation

Contrast


Extreme brands
(Very low quality)





Moderate brands
(Moderately high quality)





Moderate brands
(Moderately low quality)





Product involvement







Figure 3
 The Role of the Moderator Variable

Independent variables:
(Extremity, direction and ambiguity)

Independent variables
(
Moderator

C

B

A

Dependent variable
(Quality perception)

Moderator
(Product involvement)

















Figure 4
Differences in Priming Effects between Unfamiliar and Familiar Brand Conditions


  [--- WMF  Graphic Goes Here  ---]



 Figure 5
Differences in Priming Effects between Low and High Product
Involvement Groups


  [--- WMF  Graphic Goes Here  ---]




 References

ActiveMedia (2000), "B2C Worth USD 56 Billion in 2000," [URL: http://www.nua.ie/surveys/].

Alba, J. W. and J. W. Hutchinson (1987), "Dimensions of Consumer Expertise," Journal of Consumer Research, 13 (4), 411-454.

Andrew, J. C., S. Durvasula and S. H. Akhter (1990), "A Framework for Conceptualizing and Measuring the Involvement Construct in Advertising Research," Journal of Advertising, 19(4), 27-40.

Baron, R. M. and D. A. Kenny (1986), "The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations," Journal of Personality and Social Psychology, 51 (6), 1173-1182.

Bloch, Peter H. (1981), "An Exploration into the Scaling of Consumers' Involvement in A Product Class," Advances in Consumer Research, 8, 61-65.

Cameron, G. T. (1990), "Involvement, Arousal, Attention, Salience, and So Forth: The Location of Involvement and Related Variables in the Spreading Activation Model of Memory," in Proceedings of the 1990 Conference of American Academy of Advertising, 117-123.

Celsi, Richard L. and Jerry C. Olson (1988), "The Role of Involvement in Attention and Comprehension Process," Journal of Marketing Research, 15 (2), 210-224.

Chaiken, Shelly (1980), "Heuristic Versus Systematic Information Processing and the Use of Source Versus Message Cues in Persuasion," Journal of Personality and Social Psychology, 39 (5), 752-766.

Grunert, K. G. (1996), "Automatic and Strategic Processes in Advertising Effects," Journal of Marketing, 60(4), 88-101.

Haugtvedt, Curtis P., Richard E. Petty and John T. Cacioppo (1992), "Need for Cognition and Advertising; Understanding the Role of Personality Variables in Consumer Behavior," Journal of Consumer Psychology, 1 (3), 239-260.

Herr, P. M. (1986), "Consequences of Priming: Judgment and Behavior," Journal of Personality and Social Psychology, 51 (December), 1106-1115.

Herr, P. M., S. J. Sherman and R. H. Fazio (1982), "On the Consequences of Priming: Assimilation and Contrast Effects," Journal of Experimental Social Psychology, 19, 323-340.

Herr, P. M. (1989), "Priming Price: Prior Knowledge and Context Effects," Journal of Consumer Research, 16 (June), 67-75.

Higgins, E. T., W. S. Rholes and C. R. Jones (1977), "Category Accessibility and Impression Formation," Journal of Experimental Social Psychology, 13, 141-154.

Higgins, E. T. and G. A. King (1981), "Accessibility of Social Construct: Information-Processing Consequences of Individual and Contextual Variability, in Personality," in Cognition and Social Interaction, N. Cantor and J. F. Kihlstrom eds., Hillsdale, NJ: Erlbaum, 69-121.

Houston, M.J. and M. L. Rothschild (1977), "A Paradigm for Research on Consumer Involvement," Working Paper 11-77-46, Madison, WI: University of Wisconsin-Madison.

Kim, Chung K. (1991), "Testing the Independence of Cognitive and Affective Involvement," in Developments in Marketing Science, R. King, ed., Fort Lauderdale, FL: Academy of Marketing Science, 71-75.

Krugman, Herbert E. (1965), "The Impact of Television Advertising: Learning without Involvement," Public Opinion Quarterly, 29 (Fall), 349-356.

Laurent, Gilles and Jean-Noel Kapferer (1985), "Measuring Consumer Involvement Profiles," Journal of Marketing Research, 22 (February), 41-53.

Martin, L., J. Seta and R. Crelia (1990), "Assimilation and Contrast as a Function of People's Willingness and Ability to Expend Effort in Forming an Impression," Journal of Personality and Social Psychology, 59 (1), 27-37.

Meyers-Levy, J. and B. Sternthal (1993), "A Two Factor Explanation of Assimilation and Contrast Effects," Journal of Marketing Research, 30 (August), 359-368.

Meyers-Levy, J. and A. Tybout (1997), "Context Effects at Encoding and Judgment in Consumption Settings: The Role of Cognitive Resources," Journal of Consumer Research, 24 (June), 1-14.

Meyers-Levy, J. and P. Prashant (1999), "Consumers' Processing of Persuasive Advertisements: An Integrative Framework of Persuasion Theories," Journal of Marketing research, 63 (5), 45-60.

Mitchell, Andrew A. (1981), "Dimensions of Advertising Involvement," Advances in Consumer Research, 8, 25-30.

Mittal, Banwari and Myung-Soo Lee (1989), "A Causal Model of Consumer Involvement," Journal of Economic Psychology, 10 (November), 363-389.

NUA Internet Survey (2001), "How Many Online?"
[URL: http://www.nua.ie/surveys/]

Park, C. W. and Banwari Mittal (1985), "A Theory of Involvement in Consumer Behavior: Problems and Issues," Research in Consumer Behavior, 1, J. N.  Sheth, ed., Greenwich, CT: JAI Press.

Petty, Richard E. and John T. Cacioppo. (1986), Communication and Persuasion: Central and Peripheral Routes to Attitude Change.  New York: Springer-Verlag.

Petty, R. E., J. T. Cacioppo and D. Schumann (1983), "Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement," Journal of Consumer Research, 10 (September), 135-146.

Sherif, Muzifer and Carl I. Hovland (1961), Social Judgment: Assimilation and Contrast Effects in Communication and Attitude Change, New Haven, CT: Yale University Press.

Srull, T. K. and R. S. Wyer (1980), "Category Accessibility and Social Perception: Some Implications for the Study of Person Memory and Interpersonal Judgments," Journal of Personality and Social Psychology, 38 (6), 841-856.

Tabachnick, B. G. and L. S. Fidell (2001), Using Multivariate Statistics, Needham Heights, MA: Allyn and Bacon.

Zaichkowsky, Judith L. (1985), "Measuring the Involvement Construct," Journal of Consumer Research, 12 (December), 341-352.

Zaichkowsky, Judith L. (1994), "Research Notes: The Personal Involvement Inventory: Reduction, Revision, and Application to Advertising," Journal of Advertising, 23 (4), 59-70.