The Role of Mouse Movements in Interactive Search

The Role of Mouse Movements in Interactive Search Anna L. Cox and Mariana M. Silva anna.cox@ucl.ac.uk, mariana.silva@ucl.ac.uk UCL Interaction Centre, Dept of Psychology, UCL 31/32 Alfred Place, London WC1E 7DP UK Abstract This paper describes two experiments that aimed to investigate the relationship between mouse movements and eye movements in an interactive task. In Exp 1, participants were free to move the mouse whilst performing a single-page web menu search task. The results show that eye movements depended on the quality of the distracter items present in the menu. Similarly, the frequency of mouse movement patterns changed according to the quality of the distracters that surrounded the target. When more distracters competed with the target for selection, participants’ used the mouse to “tag” potential targets whilst their eyes were free to scan the rest of the menu. In Exp 2, mouse movements were restrained. Results show that when people were not allowed to move the mouse, choice accuracy and search time decreased and eye- movement patterns became independent of the quality of the distracters present in the menus. The results are discussed within a rational analysis framework (Anderson, 1990) and compared to existing models of interactive search. Keywords: Human-Computer Interaction; Rational Analysis Introduction Research has shown that when searching for information in a web—based menu people do not always scan every item (e. g. Pierce, Parkinson & Sisson 1992; Brumby and Howes, 2003, 2004). Furthermore, this behaviour has been attributed to the pattern of relevance of the items in the menu with regards to a pre-specified task goal. Interactive search on the World Wide Web involves searching for information whilst interacting with an interface, usually in the form of the manual selection of links using the mouse. The mouse cursor is a visual object which has the potential to interfere or assist with visual search. Few investigations of interactive search have tried to understand whether people make use of the mouse to assist their task. Byrne et al (1999) investigated mouse movements in the use of click—down menus. They identified at least two strategies employed by participants: a single move once the target item had been located by visual search, and the many-move strategy that trails the eye-movements. Further evidence that the mouse is used to aid interactive search was found by Mueller and Lockerd (2001), in a study that identified correlations between eye and mouse movements. They report the same two strategies of mouse use as Byrne et al. Similarly, Chen et al (2001) have argued that gaze position and mouse cursor position are correlated. In summary, eye and mouse movements appear to be linked. While there appears to be converging evidence that people use the mouse in particular ways during interactive search tasks, its role is often overlooked. In this paper, we report the results of two experiments designed to test whether using the mouse is important to search efficiency. We hypothesise that mouse pointing is likely to aid interactive search by enabling the user to visually “tag” certain parts of the display while the eyes are free to move elsewhere and access the “knowledge in the world”. In order to develop a comprehensive understanding of interactive search, mouse behaviour patterns should not be overlooked. In Experiment 1, participants were able to make free use of the mouse whilst performing interactive search tasks on single level menus. Several eye—movement metrics were analysed in order to characterise eye—movement behaviour when the relevance of the items contained in the menus was manipulated. The number of items skipped (items that are not fixated) is likely to depend on the cost and the expected gain of scanning all items in the menu when compared with the cost and gain of selecting an item. According to Brumby and Howes (2003, 2004), searches can be self-terminating (when a target is selected as soon as it is viewed) or redundant (when more items are scanned after the target has been viewed). In light of this, the number of post-target visits was taken as a measure of how sure people are when they make a link selection. In addition, mouse—movement patterns were identified so they could be related to eye- movement behaviour. Accuracy and inspection time were also measured since they are likely to depend on the quality of the distracter items that surround the target. Experiment 2 replicated the first experiment with new participants who were restrained from using the mouse during search. By comparing their performance and eye—movement behaviour to those of the participants in Experiment 1, we hoped to identify the benefit that particular mouse—movement strategies have on task performance. If mouse pointing is used to assist search, one would expect that when participants cannot move the mouse, searches are likely to be longer and more redundant. This is even more likely in the conditions where more distracters are compete with the target, than on the “easier” conditions (where all distracters are irrelevant). Experiment 1 Method Participants. Ten adults, seven male and three female (mean age of 26.3 years) from the University College London Interaction Centre volunteered to participate in the experiment. All participants reported normal or corrected- to-normal vision and all had more than 5 years of computer experience, specifically on internet searching. 1156