Information Systems Research Vol. 15, No. 1, March 2004, pp. 60���86 issn 1047-7047 eissn 1526-5536 04 1501 0060 informs �� doi 10.1287/isre.1040.0017 �� 2004 INFORMS Does Animation Attract Online Users��� Attention? The Effects of Flash on Information Search Performance and Perceptions Weiyin Hong Department of Management Information Systems, University of Nevada���Las Vegas, 4505 Maryland Parkway, Las Vegas, Nevada 89154, whong@unlv.nevada.edu James Y. L. Thong, Kar Yan Tam Department of Information and Systems Management, School of Business and Management, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong {jthong@ust.hk, kytam@ust.hk} Tresources. he proliferation of information on the Internet poses a significant challenge on humans��� limited attentional To attract online users��� attention, various kinds of animation are widely used on websites. Despite the ubiquitous use of animation, there is an inadequate understanding of its effect on attention. Focusing on flash animation, this study examines its effects on online users��� performance and perceptions in both task- relevant and task-irrelevant information search contexts by drawing on the visual search literature and two theories from cognitive psychology. In the task-relevant context, flash is applied on the search target while in the task-irrelevant context, flash is applied on a nontarget item. The results of this study confirm that flash does attract users��� attention and facilitates quicker location of the flashed target item in tightly packed screen displays. However, there is no evidence that attracting attention increases recall of the flashed item, as is generally presumed in practice, and may even decrease the overall recall. One explanation is that when users have to use their limited attentional resources on suppressing the distraction of flash, they will have less mental resources to process information. Moreover, the results suggest that processing information about an item depends not only on the attention it attracts per se, but also on the attention that other items on the same screen attract. While flashing an item may not increase the recall of that item, it can reduce the recall of other items (especially the nontarget items) on the screen. Finally, flash has negative effects on users��� focused attention and attitude towards using the website. These results have implications for website interface design, online product promotion, online advertising, and multimedia training systems, among others. Key words: flash animation attention online information search visual search central capacity theory associative network model laboratory experiment website interface design History: Jane Webster, Associate Editor. This paper was received on January 23, 2002, and was with the authors 12 months for 3 revisions. What information consumes is rather obvious: It con- sumes the attention of its recipients. Hence a wealth of information creates a poverty of attention. ���Herbert Simon 1. Introduction With the rapid development of multimedia and Internet technologies, information systems (IS) designers have a rich collection of tools available when designing both traditional and Web-based information systems. Among the various multimedia technologies (e.g., audio, video, and animation), animation has received much attention from both IS academics and practitioners. A review of the human-computer interaction (HCI) literature found that animation is often adopted in IS for three functions: (1) ���look and feel,��� e.g., novelty and enter- tainment (e.g., Dholakia and Rego 1998, Thomas and Calder 2001), (2) information visualization to increase comprehension (e.g., Baecker 1988, Mackinlay et al. 1994), and (3) attracting users��� attention to spe- cific information on the screen (e.g., Chimera and 60

Hong, Thong, and Tam: Does Animation Attract Online Users��� Attention? Information Systems Research 15(1), pp. 60���86, �� 2004 INFORMS 61 Shneiderman 1994, Nielsen 2000). While the first two functions of animation are generally supported by the literature, the third function is in need of a more thorough investigation as suggested by a number of researchers. In a study on user engagement in multimedia training systems, Chapman et al. (1999) concluded that it is unclear whether multimedia elements can help to maintain users��� attention on a learning task or distract them from the task, and the authors called for more research in this area. When applying animation techniques together with group support systems to support user involvement in organizational change processes, Vreede (1997���1998) acknowledged that if improperly used, animation may overwhelm the users and turn them into inter- ested viewers rather than critical thinkers. As Lim and Benbasat (2000, p. 464) pointed out, ��� the new challenge is to derive a set of guidelines on how and when to use this rich array of tools (text, graphics, audio, video, and animation) ��� Research on simple forms of animation (e.g., blink coding), as part of system interface design, first appeared as early as the 1960s. It was found that a moving or flashing element is useful for attracting attention to a certain part of the screen because of its visual distinctiveness (Cropper and Evans 1968, Smith and Goodwin 1971). However, it was also rec- ognized that such attention-attracting coding can be ���distracting,��� ���obtrusive,��� ���disruptive,��� and ���fatigu- ing��� (McCormick 1970, Stewart 1976). On one hand, these findings remain insightful because of the slow physiological evolution of humans. On the other hand, the computing environment has been undergoing dra- matic changes in the past couple of decades (e.g., the replacement of monocolor, small screen cathode ray tubes with multicolor liquid crystal displays). Hence, findings in these early studies may not ade- quately address the issues brought about by the new computing environment. With advances in graphical technologies in the 1980s, interface designers and IS researchers started to focus on the use of tables and graphics in supporting human decision making (see DeSanctis 1984, Jarvenpaa and Dickson 1988 for reviews). Other related interface variables, such as color (Benbasat and Dexter 1986) and spatial layout (Umanath et al. 1990) have also been examined. How- ever, given the difficulties in applying animation in the traditional IS (Thomas and Calder 2001), none of these studies have examined animation as an interface design variable. With the widespread adoption of Internet technolo- gies such as Java and Virtual Reality Modeling Lan- guage, animation has become much easier to create and increasingly popular on the Web (Spool et al. 1999, Zhang 2000). One often encounters flashing objects, popups, and moving text when surfing the Internet. A major reason for using animation is to attract users��� attention. Attention is a scarce resource on the Web (Davenport and Beck 2001, Glazer 1998), because of the vast amount of information available (Alba et al. 1997, Jarvenpaa and Todd 1996���1997) and the limited attention span of humans (Broadbent 1954, Lachman et al. 1979, Van der Heijden 1992). While there is enthusiasm for Web animation as being ���cool,��� ���engaging,��� and ���entertaining��� (Dolgenos 1996, McFarland 2000, Rewick 2001), it is balanced by the recognition that it can also be ���annoying,��� ���irritat- ing,��� and even ���evil��� (McGalliard 1998 Nielsen 1996, 1997 Spool et al. 1999). Sharples (1999) suggested that animation may help keep users��� attention and even boost sales at e-commerce websites when it is relevant to the users��� tasks. Given the promising Web anima- tion technology and the scarcity of human attention on the Internet, a more complete understanding of the effects of animation on attention will inform the design of websites. In the marketing literature, researchers are inter- ested in the use of animation for banner ads on websites to attract users��� attention and increase the ���click-through��� rate. Li and Bukovac (1999) found that animated banner ads can be more quickly iden- tified and better recalled as compared to still banner ads. Cho et al. (2001) concluded that a higher degree of forced exposure to animated banners ads will yield a higher click-through rate and more favorable atti- tudes among users. On the other hand, Tuten et al. (2000) noted that whether animation will provide an edge in attracting attention and generating click- through rates depends on the users��� tasks. Users who are surfing for fun or for relaxation are more likely to be attracted by animated banner ads than users who are searching the Internet for specific informa- tion. Bruner and Kumar (2000) found that experienced users are less likely to be distracted by competing

Hong, Thong, and Tam: Does Animation Attract Online Users��� Attention? 62 Information Systems Research 15(1), pp. 60���86, �� 2004 INFORMS stimuli on the Web. Dahlen���s (2001) study provides further evidence that the effectiveness of animated banner ads depends on both user experience and brand familiarity. In general, marketing researchers recognize that animation has the potential to attract users��� attention but its effects seem to vary under different conditions. The mixed results point to the hypothesis that the effect of animation on attention, recall, user attitude, display characteristics, and tasks are interrelated and warrant more thorough analyses. Also, these studies only focused on the banner ads and did not investigate them in a broader context (e.g., effects on other items on the screen or percep- tions of the websites using the animation), which is of interest to IS researchers. Surprisingly, empirical research on animation in the IS literature is scarce. A notable exception is Zhang���s (2000) study on the effects of animation on infor- mation search performance. The author found that irrelevant animation can reduce the performance of information seeking because it distracts users��� atten- tion from the core task. Zhang���s (2000) study can be extended in a number of ways. First, the information- seeking tasks used in Zhang���s (2000) study, i.e., searching for target strings of letters among an array of meaningless strings of letters, are significantly dif- ferent from the typical Web information search tasks (e.g., searching for a product). Second, the study used irrelevant animation as stimuli, while in the real Web environment, animation can be applied to promo- tional items or to highlight important information (e.g., British Broadcasting Corporation (BBC) weather center), which are potentially relevant to the view- ers. Finally, the relatively large number of tasks (20) assigned to each of the 24 subjects might have an effect on the subjects��� concentration and affected their performance. We believe an empirical study of anima- tion in a more realistic e-commerce environment, in a wider application context (i.e., both relevant and irrel- evant to users��� tasks), and with a more manageable number of tasks using a larger sample will contribute to a more comprehensive understanding of the effects of animation on online users. Among the different types of animation, flash1 is a simple but widely used form of animation on the Web 1 In this study, ���flash��� does not refer to the commercial software ���Macromedia Flash��� but to a blink coding. (Rewick 2001). In addition, as flash bears the basic characteristic of animation (i.e., a constant change in the visual field), findings derived from flash can be extended to other types of animation that share this basic characteristic.2 Hence, we will investigate flash animation in this study. The online information search environment is chosen for three reasons. First, infor- mation search is a major activity that online users per- form (Park and Kim 2000, Smith et al. 1997). Second, information search is the main context in which prior visual search research has been conducted. Third, the results involving information search will build on Zhang���s (2000) findings. Therefore, information search serves as a useful starting point for elucidating the effects of flash on websites. To gain a more complete understanding of the con- sequences of Web animation, this study will exam- ine the effects of flash in an online information search environment under both task-relevant and task-irrelevant conditions. When flash is applied on the target of search, it is considered as task rele- vant when flash is applied on a nontarget item, it is considered as task irrelevant. Whether flash attracts users��� attention is examined in terms of response time, which is defined as the time taken to locate the tar- get item. Response time is generally used as an indi- cator of information search efficiency in traditional visual search research. Presumably, if flash attracts users��� attention, it will shorten their response time when task relevant, and increase their response time when task irrelevant. If flash does attract attention, a subsequent question is whether attention leads to better recall. Recall is defined as a person���s ability to retrieve information from memory that has been earlier acquired and retained (Large et al. 1994). It is of particular interest to online retailers and advertis- ers, as they want their websites to effectively convey product information to users. Davenport and Beck (2001, p. 37) noted that recall can be a good measure of attention as it indicates the quality of attention. To answer these questions, we make reference to the visual search literature (e.g., Yantis and Egeth 1999), 2 Note that the opposite may not be true. For example, findings on a cute cartoon animation may not be applicable to a simple flash animation, as subjects may hold positive attitude toward the cartoon character.

Hong, Thong, and Tam: Does Animation Attract Online Users��� Attention? Information Systems Research 15(1), pp. 60���86, �� 2004 INFORMS 63 the central capacity theory (Kahneman 1973), and the associative network model (Collins and Loftus 1975) in understanding the effects of flash on response time and recall. Besides the performance measures, users��� perceptions of their Web experience are also central to the success of a website (Singh and Dalal 1999). Hence, this study also examines users��� focused atten- tion,3 which is the centering of attention on a lim- ited stimulus field such as on a computer screen (Csikszentmihalyi 1977), while completing the infor- mation search tasks and their attitudes toward using the website as measures of their Web experience. By including both objective and subjective performance measures, we aim to derive a more complete under- standing of the effects of flash on online users��� infor- mation search. The next section presents the theoretical back- ground of this research, including visual search research, central capacity theory, and the associative network model, and develops the hypotheses. The experimental design is then described in ��3. Section 4 presents the results of data analysis. Section 5 dis- cusses the findings, limitations, and directions for future research. Finally, ��6 summarizes this paper and provides implications. 2. Theoretical Background and Hypotheses HCI researchers have developed models such as GOMS (Card et al. 1983) and EPIC (Kieras et al. 1997) to simulate human information processing and predict human behavior when interacting with com- puters. However, these models do not account for the effects of special visual elements such as ani- mation. Based on an extensive review of the litera- ture on cognitive psychology, attention theories, and visual search studies, we decided to draw on the visual search research (e.g., Yantis and Egeth 1999) to examine how flash affects information search effi- ciency, the central capacity theory (Kahneman 1973) to understand allocation of attentional resources in the 3 Here, focused attention comes from the flow literature and is a self-reported measure of subjects��� degree of concentration when performing a given task. It should be differentiated from the ���focused attention��� variable in the cognitive psychology literature, where it is a synonym for ���selective attention��� and refers to the ability to pick out some information from a mass of data. presence of flash, and the associative network model (Collins and Loftus 1975, Nelson et al. 1993, Quillian 1969) to understand the interaction between target and nontarget items on the screen. 2.1. Visual Search Research���Salience and Task Relevance Human attention is considered to be limited, and hence is allocated selectively to objects in the visual field (Lachman et al. 1979, Van der Heijden 1992, Vecera and Farah 1994). Visual search research sug- gests that the ability to draw attention depends on both the salience of visual objects and whether they are relevant to the information search tasks. Salience refers to the phenomenon where one���s attention is differentially directed to portions of the environment (Taylor and Thompson 1982). Salience can be con- ferred by local contrast in any of the basic visual fea- tures such as color, size, or motion. Support for the ability of salient objects to attract human attention and shorten information search time can be found in various visual search theories, including similarity- based theory (Duncan and Humphreys 1989), local feature contrast theory (Nothdurft 1993), and the guided search model (Wolfe et al. 1989). Meanwhile, there is also experimental evidence indicating that salient objects may fail to attract attention when they are irrelevant to the search task (Hillstrom and Yantis 1994, Jonides and Yantis 1988, Lamy and Tsal 1999, Todd and Kramer 1994). Visual search researchers explain the above phe- nomenon by arguing that salient features only attract attention when they are relevant or perceived to be relevant to the search tasks (Bacon and Egeth 1994, Folk et al. 1992, Yantis and Egeth 1999), i.e., when applied to the target item, or perceived to be relevant to the target item. In other words, when subjects have reasons to believe that some salient features are totally irrelevant to their search tasks, they are capa- ble of ignoring them in their information search to a certain degree.4 Hence, the following discussion on the effects of flash on information search differenti- ates between task-relevant and task-irrelevant condi- tions. In the task-relevant condition, a salient feature (i.e., animation) is applied to the search target and in 4 For example, it may be more difficult to ignore a moving salient feature than a color salient feature.

Hong, Thong, and Tam: Does Animation Attract Online Users��� Attention? 64 Information Systems Research 15(1), pp. 60���86, �� 2004 INFORMS the task-irrelevant condition, it is applied to a nontar- get item. We are interested in studying whether the use of animation on websites always attracts attention as generally believed, especially when it is unrelated to the search task. And if it does, how well can the animated item be recalled and how are unanimated items on the screen affected? 2.2. Does Flash Attract Attention? Response Time If flash attracts attention, it should have an impact on users��� information search efficiency. Response time, which is defined as the time taken to locate the tar- get item, is often adopted in traditional visual search research as a reflection of information search effi- ciency. It is also a common measure for IS perfor- mance (e.g., Benbasat and Dexter 1985, Jarvenpaa 1989). According to visual search research, a salient fea- ture that is applied on the target of search can be detected efficiently in visual search (Nothdurft 1993, Neisser 1967). Flash is a highly salient feature as it continuously changes the appearance of the visual field by shifting between ���on��� and ���off.��� When flash is applied on the target item that users are searching for on a website, it can direct their attention to the target item immediately without the need to scan the items on the screen. Therefore, compared to a static website, response time is likely to be shorter when the target item is flashed on a website. Hypothesis 1a. Response time will be shorter when the target item is flashed compared to when it is not flashed. When a nontarget item is flashed, the situation is more complicated. Visual search research suggests that whether a salient feature affects information search depends on both its task relevance and salience (Theeuwes 1990, Yantis and Egeth 1999). There is evi- dence that people are able to ignore some irrelevant salient features, such as unique color or brightness, but not others, such as onset (i.e., one element appear- ing abruptly in a previously blank location, which has some similarity to flash) when they are searching for a target item (Jonides and Yantis 1988, Theeuwes 1990). Girelli and Luck���s (1997) study provides a plausi- ble explanation for the phenomenon by showing that the attentional mechanisms used in color or orienta- tion are entirely different from those used in motion detection and that human attention is much more sensitive to motion than color or orientation. In the Web environment, when users visit a website that they are not familiar with, they will have difficulty judging in advance whether a flashed item is relevant or not to their search task. And even if they have rea- sons to believe that all animation is irrelevant (e.g., banner ads), to totally ignore it can still be difficult. In fact, Spool et al. (1999) noted that users often scroll the animation off the screen or cover it with their hands to focus their reading on the rest of the information on the screen. If users cannot ignore the irrelevant flash- ing item, visual search research suggests that their information search efficiency is likely to be compro- mised, resulting in longer response time in searching for the target item (Jonides 1981, Jonides and Yantis 1988). Hypothesis 1b. Response time will be longer when a nontarget item is flashed compared to when it is not flashed. Visual search research has also found that the effec- tiveness of salient features in facilitating target detec- tion increases with local density (Bacon and Egeth 1991, Bravo and Nakayama 1992). Local density mea- sures how ���tightly packed��� the items are over the information space (Tullis 1983).5 The closer the items are on the display, the higher the local density. Note that given the same amount of information displayed over an area, local density could vary (e.g., one with all items packed into one corner of the display, while the other with items evenly dispersed across the entire display). There are several explanations for why local density could affect the effectiveness of salient features in expediting information search. First, the deployment of focused attention to a salient feature is more efficient when local density of the feature con- trast is steeper (Bravo and Nakayama 1992). Second, it could be that as the average distance between non- target items decreases in high local density displays, increased grouping among the nontarget items will result, which, in turn, makes them easier to reject as 5 ���Local density��� needs to be differentiated from ���overall density,��� which is defined as the total amount of information displayed on a single frame (Tullis 1983), or ���the percentage of active screen area��� (Danchak 1976). As noted by Tullis (1983), it is possible to have two displays with the same overall density, but different local density. It should also be distinguished from ���grouping,��� which is often related to the relevance of items on a display in the HCI literature (e.g., similar items should be grouped together) (Stewart 1976).