Effects of Perceived Interactivity on Web Site Preference and Memory: Role of Personal Motivation


  • Hwiman Chung,

    Corresponding author
    1. Ph. D. (University of North Carolina at Chapel Hill) is an assistant professor at New Mexico State University. He has worked in an advertising agency for six years and handled clients in industry ranging from aviation to computing as an account executive. His research interest is in e-commerce, with an emphasis on interactive advertising strategy. His projects have explored online consumer behavior and media usage, effectiveness of Web ads, perceived intrusiveness of rich media, measuring Internet access, and impact of sponsorship on perceived credibility of Web sites. His writings have appeared in such journals as International Journal of Advertising, Southwestern Mass Communication Journal, and Journal of Advertising (Korean)
    Search for more papers by this author
  • Xinshu Zhao

    Corresponding author
    1. Ph.D.(University of Wisconsin at Madison). Zhao's writings have appeared in such journals as Asian Journal of Communication Asian Survey, Communication Research, Comparative Education Review, Harvard International Journal of Press/Politics, International Journal of Advertising, International Journal of Public Opinion Research, Journal of Advertising Research, Journalism and Communication (Beijing, China), Journalism and Mass Communication Quarterly, Journalism Practice (Beijing, China), Journalism Research (Shanghai, China), and Public Opinion Quarterly, plus a number of book chapters and monographs. His research activities have been funded by a dozen organizations such as the Luce Foundation and the Rockefeller Foundation. Zhao's research interests include advertising, interactive media, and political communication in the United States and in China.
    Search for more papers by this author

Address: Department of Journalism and Mass Communications, MSC 3J, P.O. Box 30001, New Mexico State University, Las Cruces, NM 88003-8001. Tel: 505-646-1539.

Address: School of Journalism and Mass Communications, University of North Carolina at Chapel Hill. Tel: 919-962-1465.


The main purpose of this study is to explore theoretically and empirically the effects of consumers' different surfing behaviors in terms of advertising effectiveness in the new media context. This study attempts to answer two primary questions: (1) What effect does interactivity have on attitude and memory? (2) What is the role of individual motivation on clicking behavior on the Web site?

In this study, perceived interactivity was also found to influence consumers' attitudes toward the ad and memory for its contents. This finding is consistent with literature about the effects of interaction on attitude and memory. Results of this study showed a positive impact of perceived interactivity on both attitude and memory.


Over the past few years, the Internet, in particular the World Wide Web (hereafter Web), has been one of the hottest areas in advertising and communication research. During these years, advertising and communication scholars have studied every aspect of the Internet ranging from potentiality of the Internet (Dreze & Zufryden, 1998; Hoffman, Kalsbeek & Novak, 1995), and Web advertising measurement systems (Dreze & Zufryden, 1998; Leckenby & Hong, 1998; Wood, 1998), to the effects of new features of Web ads on memory and attitude (Bezjian-Avery, Calder, & Benzian-Avery, 1998; Coyle, 1997; Fortin, 1998; Geissler, 1998; Ghose & Dou, 1998).

Recently, scholars focused on the effects of interactive features of Web ads on subjects' attitudes and memory by manipulating different features of the Web ad. For instance, Coyle (1997) and Benzian-Avery et al. (1998) studied the differences between interactive Web ads and non-interactive Web ads in terms of consumers' Web site preferences and memory for Web site contents. By manipulating video and audio as important factors for interactivity, Coyle (1997) found that interactive Web ads are more effective in terms of consumers' attitudes and memory. However, Benzian-Avery et al. (1998) found different results. Manipulating interactivity by having an ad that gave consumers selections in choosing their own actions, Benzian-Avery et al. (1998) found that the traditional ad is better than the interactive ad because the interactive ad could inhibit consumers' cognitive processing. While these studies focused on the effects of interactive aspects of Web ads, advertising scholars have used different design tools for manipulating interactive aspects or features in Web ads. This difference is due mainly to a different conceptual definition of interactivity in a Web ad. For instance, Coyle considered ‘vividness’ as the most important dimension of interactivity; thus, he manipulated video and audio to make consumers interact with the Web ad. Benzian-Avery et al., on the other hand, defined interactive ads as those that give options for choosing and controlling information flow, and non-interactive ads as those that do not give options for consumers to control information flow.

Although those studies provide some meaningful insight regarding the effects of interactivity in the new media context, the existing literature is limited in two ways. First, previous literature did not explore consumers' surfing behavior on the Web. Since consumers' perceived interactivity with the Website will basically be based on consumers' clicking behaviors on the Web, no matter how we define interactivity on the Web, the basic concept of interactivity could be related to consumers' actual behaviors on the Website. In this sense, understanding consumers' surfing behaviors on the Website is important to further understand the concept of and effects of interactivity. Second, the studies have used different design tools to test the effects of interactivity based on their own definitions of interactivity. In some sense, those results were regarding the effects of different design tools on consumers' attitude or memory, not regarding the effects of perceived interactivity on attitude and memory. Consequently, in this study, we first try to explore the conceptual definitions and dimensions of interactivity (perceived interactivity), consumers' surfing behavior on the Website and effects of interactivity on consumers' Web site preferences and their memory of Web site contents. We also include in this studyinvolvement as a moderating variable of consumers' surfing behavior because, theoretically, involvement has been found to moderate the effects of various advertising executional cues (Petty, Cacioppo, & Schumann, 1983) and because involvement itself has been reported to explain up to 80% of the variance in ad effectiveness, e.g., attitude and memory (Ducoffe, 1996). Using a laboratory experiment, this study tries to answer the following research questions: 1) whether consumers have different surfing behavior (clicking behavior) as a function of their involvement; 2) whether perceived interactivity is related to consumers' surfing behavior (clicking behavior); and 3) whether perceived interactivity has a positive influence on consumers' attitudes and memory.

Literature Review

Types of Internet Advertising

There are two distinguishable types of Internet advertising according to the way they can be accessed. One is passive advertising, and another is intrusive advertising. The common form of Internet advertising is passive advertising in which the user voluntarily accedes to the advertising, for example a company's home page on the Web. Intrusive advertising is advertising via email or Usenet. Broadly, Internet or Web advertising can be viewed as any kind of selling messages on the World Wide Web. According to Ducoffe (1996), over 75% of respondents surveyed considered the following messages on the Web to be advertising: free sample or trial offers; branded banners; branded messages; on-line catalogs; billboard-type logos; graphical displays of products; shopper guides; and sponsor identifications for Web sites. However, most researchers and practitioners agree that the two current dominant forms of Web advertisements, classified by Hoffman and Novak (1995, 1996) are 1) banner ads, and 2) target ads or linked sites from banner ads. In this study, we consider the World Wide Web as a major source for advertisements, and use a company's Web site or company's specific product Web site as Web advertising in the World Wide Web.1

Concepts and Definitions of Interactivity in WWW

The term interactivity was widely used in various disciplines long before new media came into being. However, it is usually agreed that the major difference between new media and traditional media is interactivity (Morris & Ogan, 1996; Pavlik, 1996; Rafaeli & Sudweeks, 1997). It is also becoming an increasingly important concept or characteristic of marketing, as illustrated in the concept of one-to-one marketing (Peppers & Rogers, 1993). Previous studies have defined interactivity from different perspectives. Rafaeli (1988) defined it as “a variable quality of communication settings” (p. 111) based on the assumption that a reciprocal, two-way communication is a common desire of both the communicator and the audience. Formally stated, it is an expression of the extent that in a given series of communication exchanges, any later transmission is related to the degree to which previous exchanges referred to earlier transmissions. For full interactivity to occur, communication roles between sender and receiver need to be interchangeable. Therefore, to Rafaeli, bi-directionality, quick response, user control, and feedback are not interactivity, since these types of activity do not contain full responsiveness (Rafaeli 1988).

However, contrary to Rafaeli's definition of interactivity, other scholars defined interactivity based on the notion of control. Williams, Rice and Rogers (1988) suggest that interactivity can be defined as a three-dimensional construct. It includes control, exchange of roles and mutual discourse. Similarly, Neuman (1991) refers to interactivity as the “quality of electronically mediated communications characterized by increased control over the communications process by both sender and receiver.”

Steuer (1992), in his work on virtual reality, also addresses the notion of interactivity. In his proposed model, interactivity and vividness contribute to the so-called “telepresence” experience by which one feels present in the mediated environment. His definition identifies three factors that contribute to interactivity: speed, range, and mapping. Among the three factors, range and mapping are closely related to the concept of control. Range refers to the number of possible actions at a given time, so the greater the range is, the higher the interactivity a user can feel in communication. Mapping refers to the ability of a system to map its controls to changes in the computer-mediated environment in a natural and predictable manner. That is, if mapping is offered to a user, a user has control over his or her communication activities in the computer-mediated environment, and therefore a user can experience greater interactivity.

Heeter (1989) provides a comprehensive understanding of interactivity, defined as a six-dimensional concept. According to her, the first dimension of interactivity is complexity of choice, or “selectivity.” This dimension concerns the extent to which users are provided choices of available information. So, the more choice the user has or the more choice the medium provides, the higher the interactivity of the user or the medium. A second dimension of interactivity is related to the effort that users must exert to access information. A high-interactive medium allows users to access information more easily than a low-interactive medium. A third dimension of interactivity is “responsiveness to the user.” Responsiveness is defined as “the degree to which a communication exchange resembles human discourse” (p. 223). Therefore, humanlike responsiveness is the highest level of interactivity, and if media have higher interactivity, they react to a user like a human. The fourth dimension of interactivity is the “potential to monitor system use.” In a high-interactive medium, user selection of information can be monitored across the entire population of users. The fifth dimension is the degree of “ease of adding information.” In a high-interactive medium, a user can add information to the system that a mass, undifferentiated audience can access. The last dimension, according to Heeter, is the degree of interpersonal communication that a medium can offer. The high-interactive medium can facilitate interpersonal communication among users. Although Heeter's six dimensions of interactivity are not perfectly applied to current new media like the Web, they still offer a good overview. Heeter (1989) also points out that as technology is continuously developing, users have much more control over the information they wish to be exposed to, which is a form of selective exposure. So, among the six dimensions of interactivity, selective exposure is becoming a more important factor to give users a feel of interactivity in a new medium environment.

Ha and James (1998) also suggested five dimensions of interactivity. Using Rafaeli's and Steuer's approaches to the concept of interactivity, they define interactivity from two perspectives - interpersonal and mechanical. From an interpersonal perspective, they use Rafaeli's definition of interactivity, that is, “the extent to which messages in a sequence relate to each other, and especially the extent to which later messages recount the relatedness of earlier messages” (Rafaeli & Sudweeks, 1997 p. 3). From a mechanical perspective, they use Steuer's definition of interactivity, that is, “the extent to which users can participate in modifying the form and content of a mediated environment in real time” (Steuer, 1992 p. 84). The first dimension is “playfulness.” They suggested that the playfulness dimension of interactivity is within oneself rather than with another person, but because the communication need of an audience member on many occasions represents a desire to communicate with oneself rather than with others, they suggested that playfulness should be included as one dimension of interactivity. The second dimension of interactivity is “choice.” This dimension of interactivity has the same meaning as Heeter's “complexity of choice” dimension. Providing a choice of several options that users can choose can increase the perceived interactivity between users and the medium, in this case the Web. The third dimension of interactivity is “connectedness.” Unlike others' proposed dimensions of interactivity, Ha and James (1998) focused on the Web. They suggested that because the Web can offer diverse connections through hyperlinks, users can have more interaction with a Web site. The fourth dimension of interactivity is “information collection.” Unlike the previous three dimensions, the fourth is based on the perspective of Web site providers, not users. From the perspective of a provider, information about the users is the most important, especially when the Web sites are commercial in nature. In this case, willingness to provide the information depends on the users' free will. So, if the user provides the information, he willingly interacts with a Web site or a medium. The final dimension of interactivity is “reciprocal communication.” The more reciprocal the communication between the site visitor and the Web site provider or owner, the more the site can respond to the particular needs of visitors. So, the perceived interactivity can be increased.

Shih (1998) focused on the degree of control of the medium as the main dimension of interactivity. Control is defined in his study as “the ability to modify the causal relation between a person's intentions or perceptions and the corresponding events in the world” (p. 657). To Shih, the degree of interactivity depends on whether the user can control the flow of information. In this sense, the hyperlinks in a Web site could be an important tool for control, and therefore mean higher interactivity. If there are many hyperlinks in the Web site, users can control their own behavior through clicking or not clicking and can have greater interaction with the Web site.

Newhagen et al. (1995) offered a different approach to interactivity. They proposed the concept of perceived interactivity. They conceptualized perceived interactivity based on efficacy, which is “a two-dimensional construct: internally-based self-efficacy and externally-based system-efficacy” (p. 166). For a Web user, internally-based efficacy can be translated into his or her perceived control over where he or she is and where he or she is going, while externally-based efficacy can can correspond to his or her sense of how responsive the Web as a system is to his or her actions. Under Newhagen et al.'s two dimensions of interactivity, Web users can find their internally based efficacy in their navigation through cyberspace.

Cyberspace navigation includes “virtual movement through cognitive space made up of data and the knowledge emerging from those data” (Whitaker, 1998; p. 63). Web usability research reveals that easy navigation is critical to the success of a Web site (Kanerva et al., 1998). Web users' externally-based efficacy finds direct expression in how responsive a system external to themselves is toward their actions. The system may consist of a machine, messages, and an imagined receiver. Therefore, the perceived speed and amount of any change a Web user can produce will directly influence the level of responsiveness. In sum, the perceived interactivity can be defined as two dimensions, navigation and responsiveness.

Even though definitions and dimensions of interactivity differ across previous studies, perceived interactivity should be based on consumers' actual interactions with the stimulus. Interaction with the Website means that consumers have perceived control over information and communication flow. Therefore, a Website, which can allow consumers to seek and gain access to the information on demand where the content and sequence of consumers' surfing is under their own direct control, can be perceived to give greater interactivity to consumers while they are surfing. If a Website presentsconsumers with difficulty in gaining or accessing the information that they want, then consumers may have a lesser degree of perceived interactivity with the Website. In this study, we manipulated the Website in this way, so that consumers' have different degrees of gaining or accessing the information that they want.

Roles of Involvement in Advertising Context

One of the most pervasive intervening variables (both as a mediator and a moderator) in the communication, attitude, and persuasion literatures to date is involvement. Krugman (1965) first raised the importance of understanding varying levels of involvement as researchers struggled to describe the effects of advertisements in the relatively new medium of television. He introduced the concept of involvement into the marketing literature to help explain the different levels of processing varied advertisements appeared to receive. Since Krugman's seminal argument about television advertising, the construct of involvement has emerged as an important factor in studying advertising effectiveness (Greenwald & Leavitt 1984; Krugman 1967; Petty & Cacioppo, 1981; Petty et al. 1981; Petty et al. 1983; Rothschild & Ray 1974; Wright 1973). Involvement has been found to relate to advertising effectiveness (Greenwald & Leavitt, 1984), reaction to advertising stimuli (Laczniak, Muehling, & Grossbart, 1989), media characteristics (Krugman, 1965), and consumption (Hirschman, 1981; Holbrook & Hirschman, 1982).

From the perspective of information processing, involvement is related to elaborative processing and the amount of attention dedicated to advertising messages (Gardner, Mitchell, & Russo 1985). Involvement also affects the processing and storage of information for recall and retention (Salmon, 1986). Researchers have found that in a high-involvement situation, individuals are aroused to process stimuli (ad messages) more attentively and more systematically (Houston & Rothschild, 1978; Petty, Cacioppo, & Goldman, 1981). Celsi and Olson's study (1988) supports the notion that involvement affected the direction and focus of subjects' attention and comprehension processes.

Involvement has also been one of the most important moderating variables used by ELM researchers. In the ELM context, involvement refers to “the extent to which the attitudinal issue under consideration is of personal importance” (Petty & Cacioppo, 1979, p. 1915). So, when consumers have high MAO (motivation, ability, and opportunity) to process communication, consumers are willing or able to exert a lot of cognitive processing effort, which is called high-elaboration likelihood. In this process, consumers' attitudes are formed and changed through the central route, and central cues such as existing beliefs, argument quality and initial attitude influence attitude change and formation. On the contrary, when MAO is low, consumers are neither willing nor able to exert a lot of effort. In this low-elaboration process, peripheral persuasion cues such as attractive factors, music, humor, and visuals are determining factors of persuasion effects.

Theoretical Framework and Hypotheses

Role of Involvement in Consumers' Surfing Behavior

When consumers are highly involved with the stimulus (high personal relevance or high product involvement), they have strong motivation to process that stimulus. Therefore, their attention level will be higher than consumers with low involvement, and highly involved consumers will be more likely to voluntarily process the information given by the Web ad. That is, highly involved consumers are more likely to search the information in the Web ad than less involved consumers, since more involved consumers are willing or able to exert more cognitive processing effort. In this situation, consumers are more likely to demand greater information to satisfy their intrinsic need for information and cognition; that is, they are more likely to request and search for more information they are involved with by clicking related hyperlinks given in the Web ad in order to see detailed information or to search other related information.

In contrast to a high-involvement situation, consumers in low-involvement situations (low personal relevance or low product involvement) have lower motivation to process the advertising message. Therefore, they are less likely to request product-related information (i.e., less likely to click hyperlinks about product information in the Web ad than highly involved consumers). According to ELM (Petty & Cacioppo, 1986; Petty et al., 1983), low-involvement consumers are more likely to be influenced by the peripheral features in the Web site. Therefore, low-involvement consumers are more likely to click non-product related hyperlinks in the Web site since they are not interested in product information. Therefore, consumers' clicking behaviors will be different according to their different personal motivations. So, the following hypotheses are suggested:

H1: Consumers in a high-involvement condition will click product-related hyperlinks more than consumers in a low-involvement condition.

H1-a: Consumers in a low-involvement condition will click non-product-related hyperlinks more than consumers in a high-involvement condition.

Even though there are many definitions of interactivity and dimensions of interactivity, communication scholars usually agree that companies' Websites can give some degree of perceived interactivity by offering users diverse hyperlinks to click. In this paper, we identified the control of information flow as the most important dimension of consumers' interactivity with the Website. This sense of interactivity is also closely related to Website visitors' controlling their movement or clicking behavior around Web ad pages. According to Ariely (1998), consumers' satisfaction with the site increases as the flexibility of the site increases. That is, if Websites allow consumers greater flexibility in their search, they are likely to be more satisfied with the Website. In the Web ad context, consumers can choose where they would go/visit next by clicking or not clicking the hyperlinked texts or pictures on the company's Web site. Therefore, clicking provided hyperlinks might increase consumers' perceived interactivity with the Web site. Clicking the hyperlinks provided means that the consumer already decided to be voluntarily exposed to other materials linked to that hyperlink. And this voluntary clicking behavior is more likely to yield active and intensive information processing. Therefore, more clicking may mean greater interaction with the Website. And in turn, these increased clicking behaviors may increase consumers' perceived interactivity with the Website. Therefore, a Website which can allow consumers to seek and gain access to the information on demand where the content and sequence of consumers' surfing is under the direct control of the consumers, can be perceived to provide greater interactivity to consumers while they are surfing. Thus, the following hypothesis can be offered:

H2: Consumers' perceived interactivity with the Web site will be positively associated with their clicking behavior.

Effects of Perceived Interactivity on Attitude and Memory

The general impact of interactive media was first addressed by communication scholars who investigated its influence on communication patterns (Ball-Rokeach & Reardon, 1988; McQuail, 1984), human-computer interaction (Penniman, 1975), and the social psychology of computer-mediated communication (Kiesler, Zubrow, Moses, & Geller, 1984). Most of the literature on interactivity focuses on its conceptual underpinnings but very few studies have actually tried to measure or use the construct in experimental settings. Bailey (1992) examined the effects of three levels of interactivity in an interactive video lesson on achievement of lesson concepts among physics students. Interactivity was defined as the level of opportunity for interaction between the user and the video program. He used three variations of the same software program based on: no interactivity, low interactivity (opportunities for interaction limited to control of pace, embedded questions, feedback and remediation of an incorrect response), and high interactivity (increased opportunities for interaction including control of sequence, pace, videodisc controls, embedded questions, feedback and choice for remediation). However, he could not find a significant effect of interactivity on the dependent variable, students' lesson achievements.

Using the definition of Williams, Rice and Rogers (1988), Shaw, Arnason and Belardo (1993) manipulated interactivity in three different levels (low, medium, high) to measure its effects on creative idea generation. The stimuli consisted of 18 pages of information displayed on a computer screen. In the low interactivity condition, subjects had no control over the information sequence and time to view the information. In the medium condition, subjects were allowed to navigate freely between the screens and in the high interactivitycondition, subjects were allowed to type their ideas at any time and return to browse. They also did not obtain significant results showing a link between interactivity and divergent or creative thinking. Similarly, Jaffe (1996) tested the impact of user-sequencing control on knowledge gain in a computer-mediated environment. In his study, interactivity was similarly defined as the control of the sequence of information. And he manipulated three different levels of interactivity (low, medium, high). However, he failed to find significant effects for interactivity on knowledge gain.

Recently, several researchers studied the effects of interactivity in the realm of computer communication, i.e., Internet advertising, World Wide Web advertising, virtual reality, etc. (e.g., Cho & Leckenby, 1998, 1999; Steuer, 1992; Wu, 1999). Cho and Leckenby (1999) found that the perceived level of interactivity has a positive effect on attitude toward the advertisement (target ads in their study), on attitude toward brand, and on purchase intention. Wu (1999) used perceived interactivity to see if there was a relationship between the users' attitude toward the Web sites and the level of perceived interactivity. Wu (1999) found that users' attitudes toward the Web sites are positively related to the perceived interactivity of the Web sites. Ghose and Dou (1998) found that the level of interactivity also has positive effects on the appeal of Internet Presence Sites (IPS). That is, the greater the degree of interactivity, the more likely that site is considered to be a top Web site. Also, the degree of interactivity has a significant effect on Web site attractiveness, and the increase in site interactivity can help elevate site recognition. Cho and Leckenby (1998) also tested the effects of clicking behavior on attitudes. They found that clicking banner ads on a Website gives a greater feeling of interaction with the site and this feeling of interaction in turn creates a more favorable attitude toward the linked ad. As we defined the perceived interactivity above, the consumer's perceived interactivity might be closely related to her/his clicking behavior on the Website. This can include voluntary activities on the Website. This voluntary activity on the Website is more likely to yield active and intensive information processing than are passive activities on the site. That is, more interaction with the site might yield more intensive and active Website processing, and this active processing with the Website might result in more favorable attitudes toward the site. If consumers' perceived interactivity is highly related to their clicking behavior on the Website, this clicking behavior is not forced by advertisers, but voluntary by consumers. Such voluntary clicking behavior on the Website could result in more favorable attitudes toward the Website. Therefore,

H3: The level of perceived interactivity with the Web site will be positively associated with consumers' attitudes toward the Web site.

Rafaeli (1988) argued that interactivity actually enhanced learning, mastery, thoughtfulness and care among users. He classified past research about the interactivity of new media into three categories: obvious, less obvious, and least obvious. As to the obvious category, he suggested that acceptance and satisfaction were associated with increased interactivity. In the less obvious category, he suggested that increased interactivity had a positive impact on motivation, performance, cognition, learning, sense of fun, and extremism. Lastly, in the least obvious category, Rafaeli (1988) suggested that increased interactivity could give communicants a sense of control, which, in turn, was likely to encourage cognitive processing. This increased cognitive processing, he said, would help users to learn more and master more content from the medium. Hoffman and Novak (1996) also argued that when consumers are highly interacting with the Web ad and enter the state of “flow”, they are more likely to learn more about the products and the company itself.

However, in an advertising context, the empirical results about the effects of interactive aspects of Web advertisements on memory do not support Rafaeli's and Hoffman and Novak's arguments regarding the effects of interaction with the medium. For example, according to Benzjian-Avery et al. (1998), interactive advertising in new media is more effective in general than traditional advertising in terms of attitude toward the ad. However, according to the results, under certain situations, interactive aspects of new media advertising inhibited consumers' information processing, so consumers remembered less about the advertising message. Similar results came from Mohageg's study in the hypertext context (1992). Therefore, Mohageg's results can possibly be applied to the Website because the Website uses a structure similar to that in hypertext. Moreover, in the hypertext area, many researchers have found that including hyperlinks in the hypertext could decrease users' ability to remember the document in hypertext. For example, Thüring et al. (1995) and Wright (1989) found that the presence of hyperlinks in hypertext decreased users' memory of the document because users exposed to hyperlinks clicked them and moved to another space. This movement causes users not to focus on one text and to face another information space. Therefore, users who are exposed to many hyperlinks will have lower recall about the initial document. In some sense, it will be obvious that if subjects have control over their information flow, their memory will increase. In another sense, however, as some scholars in hypertext communication argue, subjects who are exposed to a large number of hyperlinks would possibly have some feeling of being lost, which in turn would hinder their ability to remember the contents. Therefore, in this study, two research questions regarding memory were addressed to test the effects of perceived interactivity, based on the definition of consumers' clicking behavior, on consumers' memory of the Web site contents.

RQ1: What is the relationship between consumers' memory level of Web contents and how many times they click on hyperlinks?

RQ2: Are the effects of perceived interactivity on consumers' memory about the Web ad contents different across consumers' level of involvement?



This study used a 2x3 between-subjects design to answer research questions and test hypotheses. The first factor was personal motivation, which was manipulated into two levels (high vs. low). The other factor, different number of hyperlinks, was manipulated into three levels to represent different degrees of consumers' ability to gain or access the information that they wanted to see, to see whether consumers' clicking behavior and their perceived interactivity are different when they are exposed to different types and numbers of hyperlinks on the Website. The minimum number of hyperlinks was 6, which did not give consumers a full chance to gaining or access the information, and the maximum number of hyperlinks was 48, which gave consumers a chance to gain or access a high level of information on the Website.

Product Selection and Treatment Stimuli

Previous studies have shown that product involvement would influence consumer behavior (Block, 1981; Raman, 1996). Therefore, it was decided to use a product category that would have moderate involvement for most individuals in order to obtain sufficient variance across the product involvement construct. The Zaichowsky Scale II (1994) was administered to 28 undergraduate students from a media planning class on eight different product categories — calculator, answering machine, cordless phone, printer, 35 mm camera, cereal, computer and digital camera. Results show that the camera category had moderate levels of product involvement among the subjects. Therefore, the camera was chosen as the product category for this study, and 35mm point-and-shoot camera and digital camera were used as products for a Web site.

The stimuli for this study were created with two concerns: (1) creating a realistic Web site and (2) determining the amount of information. With respect to the first concern, creating a realistic Web site, many studies have explored consumer behavior in the Internet context by using laboratory experiments; however, those studies have used very limited mock-up Web sites. Since the Web is totally different from traditional media in that the Web can give users freedom to explore, using a very limited mock-up Web site does not look real to subjects, especially to those who have a lot of experience surfing the Internet. With respect to the second concern, the amount of information, Wright and Lynch write, “A fair test of direct experience versus advertising effects should attempt to equate the informational content of direct experience and advertising” (1995, p. 710). Thus, it is important to keep levels of information as equal as possible. To do this, the level of pages in the Web site was controlled. So, when a subject clicked a hyperlink on the page, she or he could move to the next page, and when she or he clicked a hyperlink in the second page, she or he could move to the third page. The main reason to keep the level of pages in the Web site constant was to keep constant the amount of information available. Other factors that could have increased the complexity of the Web ads, such as font, page length, and other visual components, were consistent across stimuli.

The stimulus material for this study was created using HTML with a fictitious product name and company name. The very first page of the Web site was a “log-in” page. Subjects had to type their log-in ID to continue surfing. The second page of the Website contained all the hyperlinks, including product-related and other hyperlinks such as information about the company, message from the CEO, other corporate companies (under the same group), and additional features (e.g., language options, information about the group). These hyperlinks were not directly related to the product (35 mm camera and digital camera), but they were included in this study to make the Web site look real and to give subjects more options to click, especially those in the low-involvement condition. The number of hyperlinks was manipulated into three different levels, 6, 24, and 48.2 If subjects wanted to get information which was not given by the hyperlinks, subjects could look for that specific information. So, in the six hyperlinks conditions, subjects would have more difficulty gaining or accessing the information that they wanted to explore than subjects in the other conditions.

Subjects and Experimental Procedure

A total of 180 (30 for each cell) undergraduate and graduate students from a large southern state university participated in the study for extra course credit. The final experiment with students was conducted in a computer lab at the same university. All participants viewed the stimulus materials on IBM computers with identical 17” monitors, using Explorer version 5 Web browser. Prior to the experiment, the resolution of monitors and the window size of Explorer 5 were set to the same level on all computers.

Upon arriving for the experiment lab, each subject was asked to randomly select one card from a pool of 180 cards. The selected card contained the subject's log-in ID for the study and this log-in ID could randomly assign each subject to one of six experiment conditions. For example, subjects with user IDs 1 - 30 were assigned into the high involvement and six hyperlinks condition; subjects with user IDs 31 - 60 were assigned into the high involvement and 24 hyperlinks condition, and so on. After selecting his or her ID card, subjects were asked to sign the consent form for the experiment. After signing the consent form, the experiment coordinator read participants the instructions for the study. After receiving brief instructions for the study, subjects were asked to access the study's home page and type their log-in ids and start surfing the Web site. By using a special software program designed specifically for this study, subjects' clicking behavior and movement during surfing, plus all dependent variables (e.g., number of clicks, type of clicks) were automatically recorded in log files attached to the Web site. After 10 minutes of surfing, subjects were asked to stop surfing and move to the questionnaire site to answer the study questions. When subjects finished, they needed only to click the submit button at the bottom of the questionnaire. When subjects submitted the questionnaire, answers were automatically stored in log files. After subjects submitted the questionnaire, they were debriefed and dismissed.

Involvement Manipulation

This study used the same involvement manipulation used in the previous study (Petty et al., 1981). Involvement manipulation was measured by a three-item involvement measurement: while going through the ad, I was (1) very involved/very uninvolved; (2) concentrating very hard/concentrating very little; (3) paying a lot of attention/paying little attention. These three indexes displayed excellent scale reliabilities (alpha = .87), but the involvement manipulation failed to produce a significant difference in the pre-test (t= .796, p> .05). Since the results of the pre-test showed weak manipulation between high and low involvement, a “Lottery Winner” manipulation was added for the final experiment. Therefore, higher involvement was stimulated in two ways: (1) subjects in the high-involvement condition were told that this company and product were real, that the company would soon introduce this product in the market, and that it would also launch its Web site on the Internet; (2) each subject was given two choices – entering the lottery or getting cash (see Appendix 1 for involvement manipulation).


Dependent and independent variables were measured witn an online questionnaire. Some covariates such as Web skill, computer skill, Web involvement, etc., were included in the questionnaire.

  • •Number and Types of Clicks

The total number and types of hyperlinks clicked by subjects was automatically calculated by the software in the Website.

  • •Perceived Interactivity

A 5-item scale containing items which were used in Fortin's study (1998) and Cho and Leckenby's study (1999) was used to measure perceived interactivity.

  • •Attitude toward the Website (Awad)

A multi-item, 7-point semantic differential scale which was used to measure attitude toward the Website in several other studies and has been proven generally reliable (e.g., MacKenzie & Lutz 1989; MacKenzie, Lutz, & Belch 1986; Mitchell & Olson 1981) were used to measure attitudes toward the Website. In this study, 4 items which were most often used in the previous studies (see for example, Geissler, 1998), were selected from MacKenzie and Lutz's (1989) study. Those include the following anchors: (un)/favorable, like/dislike, (un)/interesting, and (un)/appealing.

  • •Memory

Memory was measured in two different open-ended questions. First, subjects were asked to write down everything they could remember about the Web site during their surfing. Second, subjects were also asked to write down all the information about products (product name, features, etc.).


Sample Size and Data Screening

As reported above, subjects were 180 undergraduate and graduate students. However, two subjects did not submit their questionnaires, which was administered on-line, making the study have slightly unequal cell sizes. Average age of subjects was 20.32, and the median age was 20. Subjects were mostly female (120, 67%) and white (154, 85.6%). Most were undergraduates in mass communications.

The data collected were examined for violations of normality and outlier contamination so that, if necessary, appropriate data transformations could be executed to correct for abnormal skewness and kurtosis levels. First, univariate normality was checked by examining univariate skewness, kurtosis, and outlying cases. Further, bivariate scatterplots were used to check for outliers and relationships among variables. Some variables seemed to be slightly skewed (e.g., memory and the number of clicks by subjects), but those skewnesses were not outside the +3/-3 range of ratio of statistics to standard error within each involvement and hyperlinks. Therefore, it seems to be normal. And finally, multivariate outliers were also checked by using Mahalanobis' Distance (critical value for Mahalanobis' distance χ2= 54.23, DF = 24). Cases 137, 87, 161, and 21 were identified as outlying cases in terms of Mahalanobis' Distance; however the range seemed to be reasonable. Therefore, for the final analysis, all cases were included.

Scale Reliability

All scales used in this study either as dependent variables or as covariates were tested and optimized for internal consistency using the Cronbach's alpha reliability procedure (Cronbach, 1951). Table 1 displays descriptive statistics and the results of the Cronbach's alpha tests. All variables, dependent or covariates, have Cronbach's alphas ranging from .85 to .97 except for the measure of camera familiarity(Table 1). That measure had a Cronbach's alpha below .3, so it was not used as a covariate in final analysis.

Table 1.  Descriptive statistics for scales used in the experiment
ScaleMeanSt. D.SkewnessKurtosisCronbach's Alpha
Perceived Interactivity 2.89 0.98 0.52−0.35 0.94
Web Ad Involvement 3.55 1.17 0.14−0.46 0.94
Involvement with 35 mm Camera 4.59 1.30−0.50 0.00 0.96
Involvement with Digital Camera 5.02 1.22−0.81 0.59 0.95
Attitude Toward the Web Site 3.51 1.43 0.01−0.85 0.96
35 mm Camera Familiarity 4.59 1.60−0.41−0.72 0.87
Digital Camera Familiarity 2.51 1.50 0.74−0.40 0.94
Perceived Web Skill 5.83 0.97−1.50 3.23 0.94
Perceived Computer Skill 2.01 0.81 0.91 0.95 0.82

Manipulation Checks

The involvement manipulation was measured by three items (while going through the ad, I was: very involved/very uninvolved; concentrating very hard/concentrating very little; paying a lot of attention/paying little attention). These indexes displayed excellent scale reliabilities (α= .87), but the involvement manipulation failed to produce a significant difference in involvement levels (t= .796, p > .05) in the pre-test. Since the t-test shows that involvement manipulation could not create a significant difference between two groups, in the final experiment, the lottery manipulation was added to create a significant difference. As Table 2 shows, the lottery device successfully created a larger difference in involvement levels between two groups. Average mean score was 4.45 for high involvement and 2.85 for low involvement, which was also statistically significant (t[176] = 9.735, p <.01).

Table 2.  Involvement scale: Pre-test and experiment
ItemInvolvement Scale ItemPre-TestExperiment
 While going through this Web site, I was:MeanSDMeanSD
1very involved / very uninvolved4.06.793.741.47
2paying a lot of attention / paying little attention4.11.853.821.43
3concentrating very hard / concentrating very little3.83.873.701.49
 Cronbach alpha.8667.9358
  High     LowHigh     Low
 Mean score4.07     3.994.45     2.85
 Significance test (t-test)t= .796, DF = 34, p >.05t= 9.735, DF = 176, p <.01

Hypotheses Tests

Hypotheses 1 and 1-a.

Table 3 shows descriptive statistics by levels of involvement. The total number of clicks is not different by involvement; however, the types of clicks are different by level of involvement. Subjects in the high-involvement condition clicked an average of 7 hyperlinks to request product information and 2 hyperlinks to see other information. Subjects in the low-involvement condition clicked an average of 4 hyperlinks for product information and 5 hyperlinks for other information.

Whether consumers' clicking behavior was different across each group, multiple regressions on the number of product-related clicks and the number of non-product-related clicks were run again. Regression results clearly show that subjects in the high-involvement condition would click product-related hyperlinks to request more information about the product. A significant coefficient for involvement means that the coefficient is significantly different from that of the comparison group, in this case the low-involvement group (p< .001). Involvement itself explains almost 16% of individuals' product-related clicking behavior (R2= .158, p < .001). Hypothesis 1-a was also tested through multiple regression. Individuals' levels of involvement had a significant effect on non product-related clicking behavior. As expected, the coefficient for the high-involvement group was negative and significant (p < .001), which means that the non-product-related hyperlinks were more likely clicked by subjects in the low-involvement condition. Individuals' involvement with the Web site had a significant coefficient (p < .01), but variation explained by Web site involvement is not statistically significant (R2= .021, p > .05). A possible explanation is that Web site involvement affects individuals' product-related information surfing behavior, not other clicking behavior (see Table 4).

Table 4.  Effects of involvement on the number and types of clicks *p<.05, **p<.01, ***p<.001 * Group 3 (42 Hyperlinks) and Low-Involvement Group are used as comparison groups * Number in parenthesis is standard error.
 Number of Product-Related ClicksNumber of Non-Product Related Clicks
Constant4.041  4.759  
I. Hyperlink      
Group 1−.374 (.514)−.080−.728−.092 (.474)−.022−.195
Group 2−.385 (.518)−.082−.744.551 (.478).1301.154
II. High Involvement3.056 (.515).3985.929***−2.885 (.475)−.419***−6.075***
III. Interaction      
Group1 × High Involvement−1.256 (.727)−.188−1.728.219 (.670).037.327
Group2 × High Involvement.422 (.730).063.578−.425 (.673)−.071−.632
R2 by Hyperlink.053*  .006  
R2 by Involvement.158***  .175***  
R2 by Interaction.014  .002  
Total R2.225  .183  
Adjusted R2.202  .159  

Hypothesis 2.

We have argued that consumers' interactivity would be based on their actual surfing behavior (clicking behavior in this case). Multiple regression with consumers' clicking behavior and other covariates were run to test hypothesis 2. Table 5 shows the results of multiple regression with subjects' clicking behavior and other control variables. The results show that consumers' perceived interactivity is closely related to their actual clicking behavior, not to their computer skill or Web familiarity, and that mere exposure to many different choices in the Web ad is not related to consumers' perceived interactivity with the Web site. Consumers' clicking behavior explains about 50.3% of variance of perception of interactivity with the Web site. Other control variables, such as computer skills or Web familiarity, did not have any significant impact on consumers' perceived interactivity. As this result implies, consumers' perception of interactivity with the Web ad is proven to be a sort of behavior-related dimension.

Table 5.  Relationship between clicking behavior and perceived interactivity: Multiple regression *p<.05: **p<.01; ***p<.001 * Group 3 was used as a comparison group * Number in parenthesis is standard error.
 Perceived Interactivity
I. Control Block   
Control 1 — Computer Skill.031 (.075).026.413
Control 2 — Web Familiarity.086 (.061).0861.405
II. Number of Hyperlinks   
Group1 (6 Hyperlinks)−.254 (.128)−.123−1.780
Group2 (24 Hyperlinks)−.194 (.129)−.094−1.508
Number of Clicks.140 (.011).70012.986***
R2 by Computer Skill.000  
R2 by Web Familiarity.006  
R2 by Number of Hyperlinks.012  
Total R2 by Control Block.018  
R2 by Number of Clicks.503***  
Total R2.521  
Adjusted R2.508  

Hypothesis 3.

To test hypothesis 3, multiple attitude scores and perceived interactivity scores were averaged to create single attitude and perceived interactivity scores for the analysis. A simple regression using perceived interactivity as an independent variable and attitude score as a dependent variable was run to test the relationship between perceived interactivity and the dependent variable. Table 6 shows the effects of perceived interactivity on attitude. As seen in the table, perceived interactivity significantly influenced subjects' attitudes toward the Web site. Perceived interactivity with the Web site explains 15% of subjects' attitude. Coefficients of perceived interactivity were significant (p < .01) and the regression model was also significant. The results from simple regression demonstrate that Hypothesis 3, suggesting a positive relationship between perceived interactivity and attitude, was supported.

Table 6.  Effects of perceived interactivity on attitude *p<.05, **p<.01, ***p<.001 * Number in parenthesis is standard error.
 Attitude Toward the Web Ad
Perceived Interactivity.575 (.102).3935.666***
Total R2.154  
Adjusted R2.149  

Research Questions 1 and 2. In this study, memory was measured by open-ended questions asking subjects to write down everything they could remember about site features and product information after surfing the Web site. Two graduate students coded those memories into two different measures: (1) a memory score about product information and (2) a memory score about the Web site. If subjects wrote the product name or any information related to the products, that information was coded as product-related memory. Other information about the Web site was coded as Website memory score. Then, one ‘point’ was assigned to each bit of information. If subjects wrote two things regarding the product, the product-related memory was two, and so on. Therefore, there were three different memory scores — memory scores for product-related information, memory scores for the Website and total memory scores (total of the product and Website memory scores). Table 7 presents the descriptive information about memory scores. The average memory score for product-related information was 1.46 (SD = 1.73) and for Web site features was .90 (SD = 1.47). These memory scores are highly positively skewed (the ratio of the statistic to the standard error was more than 8). In the high-involvement condition, subjects' average memory score for product information was 2.06 (SD = 2.02), and the average Web-site features score was .91 (SD = 1.50). In the low-involvement condition, those scores were .87 (SD = 1.11) and .90 (SD = 1.45), respectively.

Table 7.  Descriptive statistics of memory N=178
  MeanStd. DSkewnessKurtosis
 Product Memory1.461.731.283.978
 Web Features.901.471.9523.631
High InvolvementProduct Memory2.062.02.796−.456
 Web Features.911.502.064.267
Low InvolvementProduct Memory.871.111.2931.420
 Web Features.901.451.8633.161

To get answers to these research questions, simple regression using number of subjects' total clicks, product-related clicks and other clicks as the independent variables was run on subjects' memory scores. Table 8 shows the effects of total clicks on subjects' memory about product-related information and other information.

Table 8.  Effects of Total Clicks, Product Clicks and Other Clicks on Memory: Simple Regression *p<.05, **p<.01, ***p<.001 * Number in parenthesis is standard error.
 Product MemoryWebsite Memory
Constant−.0643  −.214  
Number of Total Clicks.172 (.023).4847.341***.126 (.021).4186.112***
Constant.0629  .602  
Number of Product Clicks.251 (.028).5588.916***.054 (.029).1421.906
Constant1.354  .290*  
Number of Other Clicks.032 (.038).064.855.185 (.029).4356.407***
R2 for Total Clicks.234***  .175***  
R2 for Product Clicks.311***  .020  
R2 for Other Clicks.004  .189***  

Table 8 shows that memory was positively associated with how many and what links subjects clicked. As seen in the table, what subjects clicked influenced what they remembered during their surfing. For example, the number of clicks regarding product-related information did not have a significant effect on subjects' memory about Web-site features (regression coefficient was .0323, and p > .05). The opposite was true regarding memory for product-related information. Thus, subjects' clicking behaviors were in some way related to their memory for content. To understand how subjects' clicking behavior affected their memory for content, multiple regression on total memory, controlling for subjects' perceived interactivity and levels of involvement, was run again. This time, total memory was used as a dependent variable because there was no point in categorizing memory into two different categories to test mere effects of clicks on memory, controlling involvement and perceived interactivity. Table 9 shows the results of multiple regression on memory, controlling subjects' involvement and perceived interactivity.

Table 9.  Effects of total clicks on memory controlling involvement and perceived interactivity *p<.05, **p<.01, ***p<.001 * Number in parenthesis is standard error.
 Total Memory
Control Block 1 - Involvement1.118 (.279).2314.003***
Control Block 2 - Perceived Interactivity.343 (.203).1381.687
Number of Total Clicks.248 (.041).4996.110***
R2 for Involvement.062**  
R2 for Perceived Interactivity.242***  
R2 for Total Clicks.123***  
Total R2.427  
Adjusted R2.417  

As seen in the table, the number of clicks had a positive impact on subjects' memory above/beyond their level of involvement and perceived interactivity. However, there was high collinearity between perceived interactivity and number of clicks, which means that the perceived interactivity and clicking behavior are actually representing the same dimension. Additionally, the positive impact of subjects' clicking behavior on memory was mediated by their perceived interactivity, not merely by subjects' clicking behavior. Therefore, results suggest that consumers' memory about the Website content (product and Website features) is positively associated with their level of perceived interactivity with the Website.

Conclusion and Discussion

This study was an attempt to understand the concept of interactivity in a computer-mediated communication context and to understand the effects of interactivity on advertising effectiveness, in particular memory and attitude toward the Web site. It was expected that the main dimension of perceived interactivity in Internet advertising would be consumers' actual clicking behavior on the Website, and this dimension of perceived interactivity was supported in this study.

Clicking Behavior and Role of Involvement

Regarding consumers' clicking behaviors during their surfing, this study found that those clicking behaviors are different according to consumers' different levels of involvement. These findings provide some support for the adaptability of the Elaboration Likelihood Model, which suggests consumers' behavior is different according to their level of motivation, to computer-mediated communication. As demonstrated in this study, consumers' clicking behaviors and surfing behaviors are different when their levels of involvement are different. Highly involved consumers showed that their clicking behaviors were focused on product-related information. On the contrary, consumers in the low-involvement condition exhibited different clicking behaviors.

However, these different clicking behaviors did not cause different levels of perceived interactivity with the Web site. Consumers have similar perceived interactivity when they interact with the Website by clicking given hyperlinks. That is, even though the types of hyperlinks clicked are different according to their levels of involvement, consumers' clicking behavior was interaction with the Website and this interaction with the Website actually increased consumers' perceived interactivity with the Website.3 This finding further confirms the “more-is-better” philosophy in Internet advertising. Since consumers' clicking behavior is different according to their personal motivations and different personal motivations did not impede consumers' surfing behavior in this study, providing more features to interact with increases consumers' overall clicking behavior during their surfing on the Web, which increases their overall interaction with the ad and, in turn, has a positive impact on their attitudes and memory. Therefore, in the Internet advertising context, many options and features will get consumers to become more involved with surfing Internet advertising.

Effects of Perceived Interactivity

In this study, perceived interactivity was also found to influence consumers' attitudes toward the Web ad and their memory of its contents. This finding is consistent with literature about the effects of interaction on attitude and memory (Hoffman & Novak, 1996; Rafaeli, 1988). Results of this study showed a positive impact of perceived interactivity on both attitude and memory. In particular, the positive impact of interactivity on consumers' memory contradicts the results of previous studies in hypertext communication (Mohagag, 1992), which argued that many hyperlinks inhibit consumer's memory because of information overload. However, this study found that even though total clicks can increase as the number of hyperlinks in the Web ad increases, consumers usually control their information flow by selectively clicking the links. This control facilitates increased memory, no matter how involved consumers are with the Web ad. Therefore, in some sense, more information on a Web ad provides more surfing options, which works best for some consumers.

Implications for Advertising Practitioners

This line of research has important implications for advertising practitioners. Advertisers need to become more knowledgeable about consumers' surfing behavior in the new media. From the results of this study, it appears that new media such as the Web have the capability of affecting attitude formation and change and, therefore, can be interesting and potentially powerful outlets for consumer communication. Even though previous studies argued that interactive features and design elements should be balanced and that a moderate level of features (or an optimal level) would work best in the new media context, this study found that the current “more is better” approach works best because consumers might have a greater interaction when they are exposed to more features in the Website regardless of their motivation. Therefore, advertisers should include diverse uncomplicated features in their Web ads in order to make consumers spend more time on their Web site.

Suggestions for Future Study

Several issues should be addressed by future research. First, this study did not consider the relevance of information to surfers. It is reasonable to think that when Web ads contain more relevant information, consumers will more easily process the information, stay longer, and click more hyperlinks. Therefore, the effects of relevance of information in Web ads should be explored. Second, this study used the involvement condition to manipulate personal motivation for surfing behavior. However, actual behavior may be different. For example, as Park and Mittal (1985) and Park and Young (1986) pointed out, a person can be highly involved with the advertising stimuli with different underlying motives. In their studies, they divided the high-involvement condition into two different types based on underlying motives, utilitarian and value-expressive involvement. In the Web ad context, consumers usually have two different motives for their surfing activity. One motive is to look for specific information, which is called “goal-directed” surfing behavior. These consumers are simply called “seekers.” The other motive is to seek for entertainment through the Internet, which is called, “wandering-through” surfing behavior. These consumers are simply called “surfers.” Even though their underlying motives are different, it is possible that they are both highly involved with the Web ad. Therefore, research should be conducted into whether the effects of hyperlinks are different between the different types of user, seeker and surfer. Finally, the effects of Web ad structure on consumers' motivation should be explored. In this study, the depth of Web ad was manipulated constantly across all materials; however, previous studies have found that the depth could cause different effects on consumers' surfing behaviors, and, in turn, on consumers' overall memory about the stimulus. So, the effects of structure on consumers' surfing behavior also should be explored.


  • 1

    In this sense, we use the terms Web site, Web ad, and Web advertising interchangeably.

  • 2

    These findings should be confirmed in future studies by using many other features, such as pictures, video, audio, etc. Since this study used pictures for product information, it is possible to extrapolate the findings.