A Pilot Study to Examine the Mobility Problems of VisuallyImpaired Users Travelling the Web

Simon HARPER, Carole GOBLE, Robert STEVENS

Computer Science DepartmentUniversity of ManchesterManchester M13 9PL UK

simon.harper@mcc.ac.uk http://www.man.ac.uk/towel/

ABSTRACTThe ‘Towel’ project seeks to find solutions to problemsencountered by both visually impaired and sighted userswhen travelling in the World Wide Web by leveragingsolutions found in real-world mobility and applying them tothe virtual world. Visually impaired users find mobility onthe Web particularly difficult because of the reliance ofhypermedia on visual layout and large viewable areas thatfacilitate and enhance sighted mobility. Hypertext designand usability has traditionally concentrated upon navigationto facilitate this mobility; consequently other aspects oftravel are neglected and web mobility has suffered.Similarly, the Web Accessibility Initiative (WAI)Guidelines do not take a holistic view of travel andtherefore in both these cases a fully rounded view ofmobility cannot be formulated. This paper presents thebasis for these assertions by drawing analogies betweenreal-world and virtual-world mobility, and then attempts tosubstantiate these analogies by conducting a pilot study ofvirtual mobility on a focus group of both sighted andvisually impaired web users. Knowledge of the differencesin travel between visually impaired and sighted people willenable the design of better user agents and web content forvisually impaired users as well as the sighted community.

KeywordsTravel, mobility, navigation, orientation, complexinterfaces, hypertext, visually impaired, preview,knowledge, empirical study, focus group.

INTRODUCTIONMany modern user interfaces are difficult to interact withdue to the size, complexity of interface, and complexity ofthe presented information. These include non-standard userinterfaces, such as personal organisers, mobile telephones,and complex information systems combining a fixed userinterface and hypertext content, like the World Wide Web[12]. This potentially complex and difficult interaction for asighted user is further complicated if the user is visuallyimpaired, because the richness of visual cues presented to asighted user are not appropriate or accessible to a visuallyimpaired user. While movement can be thought of as just

the mechanics of journeying (i.e. the locomotion), mobilityis movement within the context of travel, and includes thewhole experience of successfully moving from one place toanother using context, actions and objects to facilitate theonward journey. Movement through and around complexhypermedia environments, of which the web is the mostobvious example, is a major issue in the hypermedia designand usability field [13, 14]. Although there are a number ofspecialist browsers that support the reading task for visuallyimpaired people, efforts have focused on supporting the‘sensory translation’ of visual textual content to eitheraudio or touch (through Braille) rather than enhancing webtravel [11]. Moreover in virtual environments where screensize, colour, resolution, contrast, brightness, and the visualor audio rendering may all vary, hypermedia designed for aspecific client side device may also be visuallyhandicapping for sighted users [12]. This means that virtualenvironments can 'level the field' in terms of travel andmobility support for both sighted and visually impairedusers, especially with the increase in disparate client sideaccess devices (e.g. phones, PDAs, and embeddedsystems). These factors suggest that explicit and fullmobility support is badly needed as the number of implicitvisual cues, and areas of information that can be seen at aglance are all decreasing, implying that reliance on theseimplicit cues must also be decreased or replaced. Manyvisualisation and modelling techniques [13, 14] alreadyexist that try to overcome these difficulties but thesesystems do not:

1. Look at travel from a holistic point of view andtherefore miss useful mobility information.

2. Address the issues of mobility support at the hypertextcomponent, page, and to a lesser extent the journeylevel.

3. Include the client side interface as part of the hypertextexperience and therefore do not pay attention to theproblems of mobility support within that interface.

4. Support the dynamic creation of multiple and differentuser specific views and memories such as combined'seed' trails, overviews, maps etc.

5. Address the needs of visually impaired users in respectof both the type and format of the informationreturned, in effect the appropriateness of the feedback.

Research already undertaken suggests that these issuesneed to be addressed with regard to all users and not justvisually impaired people [9]. Current research divides

LEAVE BLANK THE LAST 2.5 cm (1”) OF THE LEFTCOLUMN ON THE FIRST PAGE FOR THE

COPYRIGHT NOTICE.

movement through the user interface into landscapemetaphors [13] and path metaphors [14]. It is howeverasserted here that such a separation is unhelpful as realworld travel is accomplished with no such segregation [10]and so a holistic and inclusive approach to travel should beadopted. This is because a journey is made up of manytravel tasks with each tasks having a preferred solution [9,10].

MOBILITY PRIMERMany important lessons can be learnt from knowledgegained in assisting visually impaired travellers in real worldsituations. This knowledge can be applied to both visuallyimpaired and sighted web travellers. Travel can be thoughtof as the whole experience of moving from one place toanother, regardless of whether the destination is known atthe start of travel or if the journey is initially aimless.Travelling also involves orientation, environment, purposeand mobility, the latter defined as the ability to movefreely, easily and confidently when travelling. In thiscontext, a successful journey is one in which the desiredlocation is easily reached. Conventionally, travel ormobility can be separated into two aspects: Those ofNavigation and Orientation [7].

• Orientation - can be thought of as knowledge of thebasic spatial relationships between objects within theenvironment [6]. It is used as a term to suggest acomprehension of a travel environment or objects thatrelate to travel within the environment. How a personis oriented for travel is crucial to successful travelling.Information about position, direction, desired location,route, route planning etc. are all bound up with theconcept of orientation.

• Navigation – in contrast, suggests an ability ofmovement within the local environment [8]. Thisnavigation can be either by the use of pre-planningusing maps or fore-knowledge, or by navigating 'on-the-fly' and as such a knowledge of immediate objectsand obstacles, of the formation of the ground (holes,stairs, flooring etc.), and of dangers both moving andstationary are all required.

There are additional features of mobility by visuallydisabled people that are pertinent to web travel. Thesefeatures mainly relate to cognitive mapping and howvisually impaired people mentally represent the world. Theknowledge is useful because tailored feedback to enhancethese mental representation characteristics [16] can besupplied. Many visually impaired people have a tendencyto think of the real world in a ‘egocentric’ manner, suchthat descriptions of distance and journey become associatedwith the traveller and not the environment [16]. A sightedperson may say “walk to pedestrian crossing and thencontinue on to the bank” where as a visually impairedperson may say “walk 20 metres ahead, then from thetactile surface walk 10 metres to the North West of thatposition and you are at the Bank”. It can be seen that thespecification of distance and direction is far more exacting

and the traveller relies on a limited amount of externalinformation to reach the destination. Visually impairedtravellers also break their journey into shorter stages andorientate themselves within the journey a greater number oftimes. On average, a visually impaired traveller orientatesthemselves every 40 metres, compared to a sighted travellerwho does so every 100 metres. The mental maps created byvisually impaired travellers therefore have a tendency to beegocentric, exact, and divided into smaller moremanageable steps. Tailoring feedback to enhance thesetraits would therefore enhance the mapping process forvisually impaired (and sighted, in reduced visibilitysituations) travellers [16].

THE W3C WEB ACCESSABILITY INTIATIVE (WAI)The Web Content Accessibility Guidelines were formulatedby the World Wide Web Consortium (W3C) to make allweb content accessible to people with disabilities. Theseguidelines focus to a large extent on provisions for visuallyimpaired people, as it is this user group who currently havemost difficulty interacting with world wide web content.This is because of its predominant use of visual informationto relate content, context, structure, and layout.

Visually impaired users have a number of difficulties wheninteracting with this predominantly visual information. Forexample, a sighted user will be able to assimilate the pagestructure and visual cues on that page within a few seconds.This information is also continually present (on the page)for refreshing the memory of the user quickly whennecessary. Imagine now that all you hear in those first fewseconds are about 20 spoken words. This is what anexperienced visually impaired person hears using a browserand a screen reader. Pages are typically read from the topleft to the bottom right, one word at a time. Only a smallportion of the screen is viewable at any point. It is difficultto form or maintain an overall picture. All layout, style andfont information is lost, and contextual information is easyto lose. No familiarity with layout and style is applicable,so every page is as if it has never been seen before.

The W3C Web Accessibility Initiative recognises some ofthese difficulties, tackling the issue from two directions:improving the overall design of web pages and therecognition, appropriate interpretation and presentation ofsuitably marked-up pages by user agents [17]. User agentsare defined as browsers and assistive technologies such asscreen readers and Braille displays that together inter-operate to render the information to the user.

The web content accessibility guidelines [17] are presentedin two themes of (i) graceful transformation (of content,structure and presentation) and (ii) making contentunderstandable and navigable. The first is mainlyconcerned with the supply of text alternatives and thecorrect use of mark-up, and is really oriented to the bettersupport of 'sensory translation'. Of the 14 guidelines, 12 arein this theme. The remaining two make some reference tomore profound navigation and orientation issues, but anoverall systematic analysis of how the users travel around

pages is missing, and only two travel specific guidelinesattract a top priority.

Guideline 12 addresses the issue of orientation; however,this is mainly from a site and frame perspective, and doesnot address the more complex issues of orientation withinthe page. It suggests descriptions should be placed in either'longdesc' attributes or by using a 'd' description link. Theformer method is at present unsupported on many browsersand user agents and therefore requires a specialist webbrowser. The latter is unlikely to be used as inserting 'd'links means that designers have problems maintaining thevisually integrity of the site. An additional problem existsin that many specialist browsers separate all hyperlinksfrom the content making an un-descriptive 'd' link as un-navigable as a 'click here' link.

Guideline 13 addresses the issue of navigation.Unfortunately this guideline is primarily concerned withnavigation of the site, and does not look at this activity aspart of an integrated travel framework. The guidelinecheckpoints address issues such as: metadata addition, linkdescription, site maps, and navigation bars. While all thesemechanism's are useful site navigation tools, they do notaddress the specific mobility needs of many visuallyimpaired users strongly enough.

In summary, the guidelines for user agent accessibilityemphasise system interpretability [17], however, thenavigation and orientation mechanisms are limited,although the overall vision of these mechanisms is good.The guidelines and priority balance suggest that there ismore work to do on understanding how visually impairedusers really move around and between web pages. The webaccessibility initiative alludes to mobility but it is notsystematically addressed, burying it in a series of checklistsand design features.

TRAVELLING THROUGH HYPERMEDIAWe are concerned with the pragmatics of easily travellingaround the environment and to the information in order thatit can be assimilated. There has been extensive work on theusability of hypermedia (e.g. [3]), sometimes using thephysical world as a model [2, 1]. The focus is on sightedusers. Web usability studies have found that organisation ofcontent and navigation paths are the most important factors[4, 5], followed by link effectiveness, link differentiationand destination prediction. This usability research toucheson mobility, but conflates it with navigation, in the sameway that utility is conflated with usability [2]: just becauseit is possible to navigate does not mean that it is easy orobvious to do so. Well-organised content is only of benefitif the reader is able to move around it with accuracy andagility, and be able to quickly discover and absorb itsorganisation.

The systematic hypermedia evaluation methodology, SUE[3], includes accessibility (how easy it is for users to locateinformation) and orientation (a user’s understanding oftheir current location and their own movements, and auser’s grasp of their current navigation context) amongst its

efficiency measures. However, accessibility concentrateson navigational richness, and link completeness rather thanmobility, assuming that the user can easily travel within theweb site or page. SUE ignores presentation and orientationand presumes a speed of information assimilationunavailable to unsighted users. Although targeted at noviceusers, SUE’s learn-ability measures apply directly to non-novice visually impaired users. Criteria include consistency(both structural and dynamic) and predictability, wherepredictability is focused on the user being able to identifythe meaning of a structure or foresee the results of aninteraction. The visually impaired reader is required torelearn afresh the page as it is presented, as if they havenever seen it before. A visually impaired user has a set ofassumptions, information or preconceptions, part of ourwork is to encoded and encapsulated these into the “user’sknowledge conformance” [3].

EXAMINING MOBILITY SUPPORT ON THE WWW BYDIRECTED TASKS PERFORMED BY A FOCUS GROUPFormally examining mobility on the World Wide Web isby no means a trivial task. However, a first step in lookingat web mobility is the formulation of a pilot study on asmall focus group as a precursor to a larger study. In thisway the validity of the premise of the 'Towel' project [9,10] itself could be supported and the relevance of theanswers to the pilot study questions can be examined, sothat useful modifications to the larger-scale survey can bemade. These questions, which examine virtual travel from aholistic point of view, seek to address the issues of:

• mobility support at the hypertext component, page, andto a lesser extent the journey level.

• mobility support within that interface.• views and memories such as combined 'seed' trails,

overviews, maps etc.• appropriateness of the feedback

These are not already addressed by current visualisationand modelling techniques. These issues are addressed byexamining the directed user tasks of:

• initial orientation within the environment;• orientation on the page;• orientation to a specific and disjoint page component

(which maybe a cue);• reorientation to a familiar environment;• direct searching for both information and structural

page components;• navigation;• and obstacle detection and avoidance.

The pilot study was based on the Internet Movie Database(Figure 1) site because it is an example of a classical,popular and well-organised commercial web site. The coreof the site is the search capabilities over a variety ofinformation on over 200,000 movie & TV titles arecatalogued. The site is dynamically generated using Perl asthe Server Side Scripting language, linked to a relationaldatabase.

Table 1: Pilot Study Questions and Issues Addressed

The Question: Seeks to Address the Issue of(see Introduction):

By Examining:

1) Do you consider yourself to be visually impaired? User Type Vision

2) Go to 'www.imdb.com'

a) How do you know the page is loading? Client Interface / Feedback Initial Orientation

b) How do you know the page has loaded? Client Interface / Feedback Initial Orientation

3) Draw or describe the layout of the screen in rough areas?

a) How long did it take to find this out? Mobility Page Orientation

b) How would you label each area? Mobility Page Orientation

c) Is the layout familiar, and in what way? Views and Memory Re-Orientation

d) If each of the areas you've identified above have different sections what are they? Mobility Page Orientation

4) Can you search for a movie title?

a) If you can, how do you know? Mobility Component Orientation

5) Search for the Movie 'Heart Beat'.

a) How long did it take from starting the search to 'seeing' the cover graphic did it take? Feedback Directed Searching

b) Is the page layout (excluding sections) maintained? Views and Memory Re-Orientation

6) Where was Heart Beat Filmed?

a) How long did it take to find this out? Mobility / Feedback Obstacle Detection

7) What is the run-time?

a) How long did it take to find this out? Feedback Obstacle Detection

8) Are all the options on the left-hand side selectable?

a) How long did it take to find this out that the options are not selectable? Mobility Navigation

b) If not, which is the first option that is not selectable? Feedback Obstacle Detection

c) How long did it take to find this out? Mobility Navigation

The pilot study (as reproduced in above) was produced to learn how different people travel around the web, and to ascertain the differences and problems encounteredby sighted and visually impaired people. The pilot study attempts to quantify the difficulty of performing a directed task by relating it to both time (self-timed by thesubject) and solicited comments on the experience. While the questions above may not immediately seem concerned with mobility, they have all been formulatedbecause they directly address mobility within the real world (see the Introduction) and examine these real world issues by examining more specific areas. Thequestions are directed so that they can be split into groups addressing different issues within the mobility field, but are slightly modified to take into account thetransition from the real to virtual world.

The pilot study questionnaire (see Table 1) was sent tovarious mailing lists on the Internet and required that anindividual complete a series of mobility tasks and answerquestions about each task completed. The questionnairewas completed by twelve individuals, six who werecompletely blind or visually impaired and six who weresighted. The equipment used varied from a standardcomputer set up for the sighted subjects to the use ofspeech and Braille translation devices for those who whereblind or visually impaired. The method of output is nottaken as a factor in the analysis of the figures as thesesystems all perform the task of information output and arenot concerned with structural and navigational information.

Although the IMDB is a dynamic site (i.e. one that isgenerated from a database, when the page is requested by auser) the questions where also formulated so as to be validfor all delivery mechanisms (be it dynamic or static). Thesequestions also cross-reference each other so that any onequestion is not entirely relied upon to give a single(possibly misleading) result. The questions are all based onreal world surveys mainly undertaken with visuallyimpaired subjects, and they all relate to ascertainingjourney information that can be divided into a number ofreal world inspired mobility groups

Pilot Study SummaryThe successful completion of these mobility questions

relies on both information present within the'environment' (i.e. web page or site) and informationin 'memory' (either user memory or memory on theclient side device) that can be referred to as requiredto aid mobility through the 'environment'. Theresults are grouped by the objective of the directedquestioning, are discussed in detail below and arealso summarised in Table 2.

Initial OrientationSighted users seem to be visually cued to theloading of the page by the progress bars and variousrevolving title-bar icons and graphics with only onerespondent stated the rate that the screen layout wasrendered provided a visual cue to the status of theloading page. Users are cued to the completion ofthe page loading by seeing that the title-bar iconshave stopped moving and that the status barnormally says 'Done'. They also use the full screenrendering (when the page is seemingly complete) todecide if the page has loaded. However the mainreason why the status bar was most used, is becausecontent may be still loading off the viewable area.Visually impaired users are not cued to the pageloading by the application they are using; rather theyuse implicit auditory cues, like the hard disk drivenoise, or the sound of the modem. Some individualsdo, however, seek explicit cues in the form ofcontinued probing of the application status to see ifanything has been loaded. These users notice thatthe hard disk drive or modem has stopped makingnoises, or that a manual probing of the applicationreturns a successful result. Some applications alsoprovide an auditory cue to signify that a page hasbeen loaded.

Page OrientationSighted respondents quickly assimilate the screenlayout. This group labelled each area consistently asbanner or advertising, menus, and content. Whilethis did not vary some respondents listed a numberof subsections within the previously defined areas.Visually impaired users however took noticeablylonger, in fact 33% of the respondents did notcomplete this task at all. However if these are given

Figure 1: The Internet Movie Database Site

the average score of the other 66% of respondents then theaverage time taken was 100 seconds with a median score of120 seconds. All respondents made an attempt to label theareas of the page, however 50% gave up as the layout wastoo complex and disorientating, and 50% labelled someareas correctly and some areas incorrectly.

Component OrientationSighted users all agreed that a user can search for a movietitle as well as a number of other options that are available.All stated that they could see the search box at the top leftof the screen, and so the spatial orientation of a user on thepage and their understanding of the page elements wereconsistent. However when the component was not withinthe viewable area some confusion ensued, and both theaverage and mean times increased substantially. Likewiseall visually impaired respondents agreed that they couldsearch for movie titles because the search box is one of thefirst items to be spoken in an auditory system due to thefact that it is near the top left of the page. Otherorientational tasks to specific components that occurredfurther along the HTML data stream where notaccomplished with as much success and in some cases notachieved at all.

Re-OrientationSighted users took a median time of 10-15 secondsexcluding download time to re-orient themselves to thepage layout on a different page. This re-orientation speedwas in part due to the familiarity of the page as all usersstated that the page layout was maintained correctly andtherefore the orientation process for each individual was

maintained also. Visually impaired users faired differently,they took an average time of 90 seconds and a median timeof 120 seconds (both excluding download time) to re-orientthemselves. However, only 33% of respondents agreed thatthe layout was maintained, where as the other 66% ofrespondents did not know either way.

Directed SearchingAll sighted users responded that they could findinformation resulting from a directed search easily, with theaverage response time being 20 seconds. The medianresponse time was 10 seconds. However, after the initialsearch users who were directed to find differentinformation using the same search method reported an evenlower average time of 17 seconds. Some visually impairedusers reported a very long time to complete the directedsearch. However the average time was 200 seconds, with amedian time of 100 seconds. Interestingly, no visuallyimpaired user reported a decrease in time based on themore searches performed which suggests (as with the re-orientation results) that visually impaired users could noteasily recognise locations they have been to a number oftimes.

NavigationSighted users experienced no problems in finding outnavigational links based mainly on implied menus downthe left-hand side of the screen. 'Feints' and actual linkswhere easily distinguished, the average time being 17seconds with a median time of 10 seconds. This was alsothe case when navigational components were not displayedon the viewable area, but their existence was implied.

Sighted Visually ImpairedCategoryMeanTime(sec.)

User Comments MeanTime(sec.)

User Comments

Initial Orientation N/A Cued to icons / status bars. N/A Cued to random hardware noise.

Page Orientation 10 All agreed on layout. Userscorrectly labelled each area.

120 Agreement on layout was notreached. Users did not label eacharea correctly.

ComponentOrientation

N/A All users easily identified thesearch box component.

N/A All users easily identified thesearch box component.

Re-Orientation 15 The page layout is familiar. Allusers responded that the pagelayout was the same.

120 The page layout was not familiar.Only 33% of users recognisedthat the page layout was the same.

Directed Search 10 The information was easilyfound.

100 It was difficult to decide if theinformation was present.

Navigation 10 All users identified which menuitems could be traversed andwhich could not.

95 All users identified that somemenu items could be traversedand that some could not.

Obstacle Detection 5 All respondents correctlyidentified the obstacles.

95 No respondents correctlyidentified the obstacles.

Table 2: Mobility Questionnaire Summary

However the time was increased when navigationalcomponents where not on the viewable area and theirexistence was not implied. 66% of visually impairedrespondents thought that all the options where selectable,while 33% did not. It took on average 100 seconds to findthis out, with a median of 95 seconds.

Obstacle DetectionSighted users all stated certain options on a standard menuwhere not all selectable and that the first one that was notselectable was 'User Comments'. It only took an average of6 seconds to find out this information, with a slightlydifferent median time of 5 seconds. This suggests thatsighted users can easily detect visual obstacles, in this case'non-selectable' menu items. Of the visually impaired users,no respondents identified ‘User Comments', as the firstnon-selectable option. All stated, that some items where notselectable, the average time respondents took was 80seconds with a median time of 95 seconds.

AnalysisOur preliminary study suggests that visually impaired userstake far longer to orient themselves within, and navigatearound, a web environment. While sighted users couldorient themselves to the page quickly, it often took visuallyimpaired users roughly ten times as long to perform thesame task. Visually impaired users also had problemsdescribing the layout and structure of the pages accurately,and this included correctly distinguishing selectable andnon-selectable menu items. The main problems seemed tooccur due to the complexity of the screen format and theunavailability of a structural overview of both the page andthe site. There was, however, a correlation between theproximity of an element to the top left corner of the screenand the speed of a visually impaired users access of thatelement, which was on average faster than elements to thebottom right corner.

DISCUSSIONThe pilot study suggests 3 main aspects that visuallyimpaired travellers find more difficult to deal with thansighted users:

1. Navigation

2. Orientation

3. Layout and Structure

These 3 problem categories can be directly associated withthose encountered in the real world and are concerned withseeing into the distance and the appropriateness of theinformation either stored (as maps, descriptions, etc) orreturned from the environment as it is traversed.

Seeing into the DistanceInformation with regard to upcoming and related eventsand components of the information at both the local level(on pages and frames) or at an overview level (informationresource wide), as well as on the client side component ofthe information system, should be provided. The user willthen know where they are and how to get to where they

want to be. This means that a preview should be given ofwhat will happen next and where the user will be taken to(within the information and interface), when the next userevent occurs. This increases both a user’s mobility andsense of security (that nothing unforeseen will occur).Information that is found further down the page (and istherefore out-of-view) should be linked to a preview nearthe start of the information, to increase a users structuralorientation to the page and the information containedwithin it. Previews should also occur on items that may baror hinder a users progress (called obstacles in mobilityparlance) and items that may actively facilitate orientationand navigation within an area (called cues).

Explicit Universal FeedbackThe issue of appropriate feedback of mobility and travelinformation should be addressed. This means that a userhas full access to the information the system designerwishes to present (feints, menus, different fonts etc). Thisfeedback should be universally available (using bothauditory and visual cues) and should also be explicit. Forexample, a colour change to red text may mean somethingvisually but means nothing in the auditory domain, and itsimplied meaning is therefore lost to a visually impaireduser [6]. Therefore, including the hidden components, likethe implied semantics and actions of the interface in anexplicit feedback mechanism, means that both the designersand users perception of the meaning of the information andhow the mobility components work, will match.

Mobility ObjectsTo recap, preview is the ability to look ahead so that thephysical or virtual environment

can be detected and

interacted with more easily. External memory is the abilityto recognise environmental and spatial information fromsources of learnt knowledge. Both preview and learntknowledge exist in the environment as cues, and external tothe environment as part of either internal or externalmemory such as maps and descriptions. Using thisinformation it is therefore possible to group mobilityobjects within the virtual environment into a number ofroles and sub-roles:1. Cues - are objects or combinations of objects that a

traveller actively uses to facilitate their onwardjourney.1.1. Navigational Cues - answer a traveller’s question

'Where can I?' and could be for example asignpost.

1.2. Orientational Cues - answer a travellers question'Where am I?' for example a unique combinationof objects.

2. Obstacles - are objects that inhibit a users onwardjourney, however under certain conditions (such asfamiliarity with the object) an obstacle can change to acue.

3. Memories - are either internal or external, and containthe knowledge that enables a traveller to decide if theobject is a cue or an obstacle.

3.1. Navigational Memory - answers a travellersquestion 'Where can I?', for example the results ofa search on a help system.

3.2. Orientational Memory - answers a travellersquestion 'Where am I?', for example a site map onthe world wide web.

4. Out-of-view - addresses the concept of preview. Validtravel objects can be present but out-of-view. They aretherefore not obstacles because they do not inhibittravel but they are not cues either because they do notfacilitate travel until they come into view. Out-of-viewobjects are those that take over approx. 20 seconds tobe spoken (remember in the pilot study, a sighted usercan get an overview of the a standard page within 20seconds) or are below the current viewable area, andobjects in the viewable area do not lead onto or suggestthe presence of out of view objects.

SUMMARYFrom the pilot study it can be seen that both the memory(or knowledge) and the out-of-view (or preview) rolesinfluence the decision as to whether an object encounteredduring travel is an obstacle or a cue. Hence the ease ofmobility can be predicted from the number of cues andobstacles. The amount of preview of information and theappropriateness of the knowledge feedback of thatinformation, are key aspects of a good user interface [6].Therefore, systems that enhance the informationpresentation, by directly addressing these issues, provide amuch better experience for the user than those that do not.This may seem like an orthodox viewpoint and in somerespects it is, however these problems have now beenmethodically identified in the context of visual impairmentand real world travel. Moreover, it is the assertion of a linkbetween real and virtual world mobility that makes thesesolutions both important and novel.

Defining a series of mobility roles is useful because itenables mobility objects to be characterised and negativemobility objects (such as obstacles) to be highlighted andchanged. However, the scale of this 'usefulness' could onlybe fully understood after an initial study had beencompleted to ascertain the level of the mobility problemsencountered by both visually impaired and sighted userswhile travelling the World Wide Web. Further work cannow be carried out to attempt to correct mobility problemsand another survey undertaken to check that these problemshave been addressed. Although providing appropriatefeedback and better previewing techniques should enhancethe manageability of an interface, further work is needed inidentifying the amount and type of preview users finduseful. These issues should be addressed by in depth usertesting, using well-documented techniques, as differentusers have different requirements. While further work isstill required, this pilot study has lead to the developmentof a framework for analysing hypermedia with regard tomobility. It has also lead to the creation of a number ofsmall plug-in tools, analysing XHTML, to assist in

rectifying deficiencies in mobility support for visuallyimpaired users travelling through hypermedia.

REFERENCES1. Darken R and Sibert J, Wayfinding strategies and

behaviours in large virtual worlds in Proc ACM CHI'96,pp:132-149, 1996.

2. Dillon A and Vaughan M ‘It’s the journey and thedestination’: shape and the emergent property of genrein evaluating digital documents in The New Review ofHypermedia and Multimedia Vol 3, pp: 91-106, 1997

3. Garzotto F and Matera M A systematic method forhypermedia usability inspection in The New Review ofHypermedia and Multimedia Vol 3, pp: 39-65, 1997

4. Nielson J How to Write for the Webhttp://www.useit.com/papers/webwriting/writing.html1999

5. Spool, JM. et al Web Site Usability: A Designer's Guide,User Interface Engineeringhttp://world.std.com/~uieweb/index.html 1997

6. Bentzen B. L., 1979, (Pg. 291-355), in Foundations ofOrientation and Mobility, Ed Welsh R. L. et al, 1979,AFB.

7. Farmer L. W., 1979, (Pg. 357-402), in Foundations ofOrientation and Mobility, Ed Welsh R. L. et al, 1979,AFB.

8. Brambring M., 1984, Mobility and Orientation Processesof the Blind, in Electronic Spatial Sensing for the Blind(Pg. 493-508), Ed Warren D. H. and Strelow E. R.,1984, Dordrecht; Lancaster: Nijhoff, 1985, USA

9. Harper S., Stevens R., Goble C., 1999, Towel: RealWorld Mobility on the Web (Pg. 305-314), InComputer-Aided Design of User Interfaces II, eds.Vanderdonckt J. and Puerta A., Kluwer Academic 1999.

10. Goble C., Harper S., Stevens R., 2000, Towel: TheTravails of Visually Impaired Web Travellers, In:Proceedings of the ACM Conference on Hypertext andHypermedia 2000, ACM Press New York 2000.

11. Chieko A. and Lewis C., 1998, Home Page Reader:IBM’s Talking Web Browser, in Closing the GapConference Proceedings 1998.

12. Dix A., et al 1993 (1st Edition), Human ComputerInteraction, Prentice Hall, London, 1993.

13. Chen C., 1997, Structuring and Visualising the WWWby Generalised Similarity Analysis, In: Proceedings ofthe Eighth International ACM Conference on Hypertextand Hypermedia, ACM Press New York 1997

14. Furuta R., 1997, Hypertext Paths and the WWW:Experiences with Walden’s Paths, In: Proceedings ofthe Eighth International ACM Conference on Hypertextand Hypermedia., ACM Press New York 1997.

15. Edwards, Alistair D. N., 1995, Extra-Ordinary Human-Computer Interaction, Cambridge University Press,Cambridge, UK.

16. Dodds A. G. et al, 1982, The Mental Maps of the Blind,in Journal of Visual Impairment and Blindness January1982 Pt 76 (Pg. 5-12), Ed Shively S., AFB Press, USA.

17. WWW Consortium - http://www.w3c.org