oturn home > Theory of presentation: overview > §6 Articulation

A theory of presentation and its implications for the design of online technical documentation
©1997 Detlev Fischer, Coventry University, VIDe (Visual and Information Design) centre

6    Navigation

Through navigation, users turn latent resources into presented resources. Navigation responds to situated presentation needs by displaying the anticipated resource as quickly and directly as possible. Navigation time and effort can be conceptualised as navigation distance. ‘Quick access’ topped the list of requirements established in a brainstorming with service engineers (cf. appendix IV-Brainstorming summary). Users' requirement for minimal navigation distance is a reaction to the weight of presentation in the face of breakdown causing delay and repair costs (cf. section 3.6 Presentation weight in chapter 3–Context).

Although the term ‘navigation’[1] has mainly been applied to describe the orienting interaction and accessing of nodes in on-line documents [2], the concept of navigation distance suggests a broader use [3]. Navigation can be described on three levels: delegating, switching, and noticing. Between the minimal distance of noticing and the maximal distance of delegating there is the moderate distance of switching between units of resource such as pages or documents. Navigation, then, includes scanning diagrams, browsing manuals, fetching folders, clicking buttons to access nodes in on-line documents, searching lists or indices, or contacting colleagues on the phone. It spans all levels of structure and resource: organisations, buildings, rooms, filing cabinets, files, documents, pages, paragraphs, words and icons.

The display consists of presented units of resource on different levels of abstraction. While being displayed is a precondition for noticing which turns units into tokens, noticing is a precondition for problem aggregation and understanding (cf. section 4.3 Tokens and granularity chapter 4–Problem). The focus of noticing is the moving threshold of awareness that constantly cedes its particulars to the process of aggregation in which tokens turn into dependent aspects of the problem. Aggregation and navigation mutually afford each other: both are inseparable moments of presentation.

Delegating describes navigation that crosses setting boundaries. Users delegate presentation to other users in other settings. Examples of delegating are requesting a report or asking a colleague to navigate to some remote resource. A service engineer, for example, may ask the laboratory for a spectral analysis of debris, or request air flow tests to be carried out at the engine test bed. As a result of delegating, new resources for navigation will arrive with some delay.

Switching causes objective changes to the available display. Page-turning and accessing nodes in on-line documents simultaneously reveals and hides views; unfolding a diagram hides documents underneath. Display is here taken to include all currently available resources, including those in other sensory spaces such as audible, olfactory, or tactile-kinetic spaces [4].  The extension or ‘real estate’ of the display, whether the size of a wall, desk, page or screen, constitutes the physical boundary of visibility. The audible display has less precise boundaries expressed as earshot.

Noticing describes the sequence of subjective changes of focus through resonance as users perceive and make sense of the current display (cf. section 4.2 Resonance in chapter 4–Problem). All levels of navigation are highly recursive, in that noticing affordances in the display flips over to switching which shows new affordances for noticing, or to delegating presentation to other settings.

[diagram showing navigation moves on the levels of noticing, switching and delegating]

Figure 6.1. Overview of navigation. Navigation causes moves to targets on different levels which recursively influence aggregation.

Acts of navigation produce discrete changes of focus according to operational protocols (cf. figure 6.1). A recurring figure of navigation includes the user who activates a pointer that invokes an operational protocol that leads to a target. The target of navigation becomes—or becomes part of—the new focus. The figure can be traced on the three different levels: An diagram arrow leads to noticing the token it points at; an entry in a list of contents or a button on a screen leads to switching to a previously hidden view; a telephone number or message leads to delegating presentation to another user in another setting.

There is a sense of discreteness, or flip-over, in all operational protocols: in walking from one room to another; choosing one document or another; switching between pages or system views (and thereby hiding other pages or views); noticing a new display token (and thereby rendering peripheral the previous token). The discontinuity of navigation flip-over is counterbalanced by the accumulating moment of aggregation. The problem pattern and the projection it affords contextualise and relate the source and the target of flip-over.

Often, there is a strong interaction between navigated material resources and confluently articulated resources (cf. section 7.1 Display confluence in chapter 7–Confluence). As navigation triggers articulations which relate the display to the problem, these articulations in turn produce new navigational moves. Examples are verbal pointers [5] such as questions which cause delegation to another user and possibly occasion navigation of further resources. Such delegating may be intended or volunteered depending on the visibility of presentation for others. A user may intentionally select another user in earshot, or the latter may overhear articulation and volunteer to take over presentation.

The three levels of navigation can be compared in terms of availability, agency and scope of control. The agent of noticing exhausts given availability through eye movements across the visible display (or focusing attention within the audible space). Control is immediate (though largely unconscious), but limited: noticing does not materially affect or extend the display. The agent of switching mediates availability by changing the display through presentation of previously hidden resources. The scope of control is extended, but it is less immediate: the feedback loop must now accommodate discrete and possibly unexpected display changes. Finally, the agent of delegating passes on agency to a black box of another remote setting. Visibility altogether ceases; the scope of control is further extended, but latent in that it only intervenes when the dependent presentation fails to satisfy relevant protocols (e.g., if it fails to respond within temporal tolerances or in the predicted way). It is clear that all three levels of navigation are intimately transmitted and synthesised in actual presentation; the distinction however is useful for analytic purposes.

6.1    Delegating

In delegating, an agent in one setting triggers a dependent presentation in the black box of another setting. The setting may be another department or organisation or an algorithm encoded in software. Delegating indicates that presentation can have several agents on different levels. In delegating, agency is therefore mediated. An example is the mediated agency of user (A) who interrupts the prior activity of user (B) with a request that triggers a new presentation in (B)'s setting. This presentation closes out when its result is returned to the initial setting. On (A)'s level, presentation continues, (B)'s presentation being a single navigational step.

The structural and power differential  in and between organisations specifies delegating protocols and rationales. It defines presentation weight, scope for decisions, and scope of compliance. Through delegating users optimise costs for their setting. They need a protocol or rationale for delegating the trigger, task, or person (say, a visitor) to another level or type of setting. Certain trigger types invoke protocols that clearly map the trigger onto specific settings. Requests for issuing a repair scheme will be directed to repair engineering, job applications to the personnel department, etc.

Users may refuse to be the target of delegating if the protocol is invalid: Service engineers turn down requests for help by small operators who have bought second-hand aircraft for which no service contract exists. On the other hand, perceived problems of important customers may cause the delegation of presentation (e.g., explorations of engine modifications) which engineers consider unnecessary.

In the absence of unambiguous protocols, delegating will depend on some negotiated rationale which is likely to measure presentation cost (e.g., invested time) against presentation benefits. Benefits may be predictable (signing a contract), probable (talking to a potential customer), or unknown (meeting a research student). The case for delegating is strong if the target setting is less powerful and costs are lower or more difficult to measure: the researcher delegated by management to attend a course in the training school does not incur additional costs, since the course is run anyway. Social costs clearly exist, e.g., noticeable confusion and embarrassment caused by the injection of an irregular participant, but these costs are difficult to measure and therefore do not add up to a rationale for refusal. Difficult triggers may float in limbo, being passed up by settings claiming lack of authority or passed down by settings claiming lack of familiarity.

Delegating may be mediated through the design of documents which act as agents scheduling users' activity. In evaluations, the evaluation plan schedules users' presentation. CAL courseware code schedules learners' navigation and even articulation through the use of input templates. A work card schedules mechanics' maintenance actions.

6.2    Switching

In switching, users replace or transform the available view through actions such as clicking buttons, page-turning, or handling parts of equipment. Switching is needed when a resource cannot be displayed in one view. Such view may be a node in an on-line document, a diagram or page, or a certain aspect of equipment. In documents, this leads to resource segmentation into discrete views which are held together by the resource architecture: each document view shows only a part of the overall resource (cf. section 8.3 Document system design in chapter 8–Design). A borderline case between noticing and switching exists if the target of navigation is the equipment. Here, views may be discrete (e.g. when inspecting a range of parts, one after the other), or continuous and directly coupled to user actions (e.g., when inspecting one part from all sides through turning it in one's hands).

Agents of switching are user actions or encoded schedules, or a mix of both, as when schedules prompt or suggest user actions. This is common in CAL courseware. The encoding can be weak as in books which suggest but do not pace the scheduled order, or strong, as in slide shows, classroom-based training, and videos.

In the segmented resource of switching, target tokens are no longer embedded in one display as in noticing: they are spatially and temporally distinct. Turning a page hides the previous page; bringing up a new view replaces the previous view. This poses the problem of reversibility of focus, or more generally, that of token retrieval. Retrieval requires access structures [7] which show tokens in context and provide pointers to hidden tokens. Examples of access structures are cardboard boxes filled with equipment parts, lists of content, iconic overview maps, or on-line search engines.

In simple cases, content and access structure are just different aspects of the same units. Units are resolved as content tokens insofar as they act as immediate resource for presentation, and as pointer tokens insofar as they lead to further hidden content. The content realised through the inspection of a part, for example, may turn into a pointer to further resources, as when metal residues found on a master chip detector point to consequent laboratory tests and to additional checks of other detectors.

In the case of documents, access structures reach into the content display, which often leads to a clear separation of display units into content and pointer units. The illustrations in the Illustrated Parts Catalogue (IPC), for example, are content, while the marginal ATA references and figure numbers are pointers to other views in ATA manuals or in the IPC parts list. In on-line documents, content and pointer token merge in the concept of hot text (or hot spots in graphics).

The lack of visibility of the target views of switching means that for the user they exist in a twofold sense. Before switching the user may to some extent predict, i.e. project, remember or infer, the target view; after switching, the actual view is then perceived in relation to the predicted view. When the resource is very familiar, the focus extends temporally so that both predicted and actual view may become confluent. For example, in the case of frequently used stable views such as computer dialogue windows popping up in the same position and requesting confirmation of user actions, users may already position the cursor near the predicted position of a pointer before the dialogue window is displayed.

With unfamiliar resources, predicted and actual view may interfere and impede understanding. For example, when clicking a button labelled ‘Show buttons’, one cinegram user expected the appearance of an additional range of buttons along the bottom of the screen. Instead, the system simply outlines existing pointers within the current view. The user's prediction not only failed to materialise; the perturbation created by the gap between predicted and actual view also prevented an understanding of the system response.

Switching limits confluence between views within the same unit of resource (cf. section 7.3 Beyond the display in chapter 7–Confluence). The transient boundary of focus depends on the speed of fading of the replaced display, which in turn depends both on the strength of resonance, link type inferences, and the degree of resonance or interference between displays.

6.3    Noticing

Noticing is the basic operation of a user interacting with a displayed resource. Noticing ‘takes in’ the current display as it is drawn to aggregate those display tokens and relations that appear through resonance on display and patterns or are produced through articulation. Noticing navigates the visible display through discrete eye movements. The (largely transparent) operational protocol of noticing dynamically couples the gaze to tokens in the display. The process is highly recursive: tokens may be realised as pointers leading to targets which become new pointers to further targets.

The most restricting resource substrate for noticing is linear text where the top-left corner of a paragraph can be seen as the pointer invoking a continuous scheduled navigating activity. The operational protocol  is transparent: it organises the saccades that progressively move from current to future fixation, i.e., from pointer to target [8]. The relative stability and regularity of the protocol is owed to the homogenous, sequential, and unidirectional form of the substrate, text—and behind the text, speech.

In noticing pictures, the resource substrate cannot enforce such constraints. Nevertheless,   pictorial notation [9] provides directional and hierarchical cues for ‘reading’ by restricting the vocabulary to few clearly differentiated unit types which are arranged, linked and grouped by standardised means. However, the simultaneous availability of a multitude of units implies that users' response presentation is unpredictable, although certain commonalties of patterns of inspection can be established [10]. Given the same material display, different domains and problems can lead to very different noticings. Put positively, one image or text can have many relevant usages. Put negatively, noticing may lead to an incorrect understanding of the display, where correct and incorrect can only be decided on the basis of a validation context.

Noticing is always confluent and temporally extended, i.e., it includes more than one discrete token. The identification of noticing with the particular content token [11] on which a user currently focuses is a common abstraction convenient for the representation of problem solving as action sequence [12],  but it fails to capture the complexity of presentation. First of all, confluent articulation equally contributes to the focus. But even on the level of noticing, the token is never seen on its own; instead, resonance and projection renders the token as a dependent moment of some larger aggregate.

6.4    Search

The discussion of navigation would be incomplete without search. Search is projection onto the resource, i.e. the activation of a search pattern looking out for display resonance. The search pattern may be a clearly formulated or a more vague need that is not tied to a precise lexical form. For example, the need to get back to the start screen may activate an unspecified search pattern hovering over the display and ‘latching on’ to any semantically related token. Unsaturated, the need may articulate ad hoc descriptors of the navigation gap [13] such as ‘I cannot see how to get back to the main drawing’. Such articulation may activate a lexical cluster which types particular display tokens and hides others (cf. section 4.6 Projection in chapter 4–Problem).

Articulation qualifies the resource in concurrent talk and thereby updates the search pattern emphasis on the fly during navigation. The search descriptor is not always ‘search for X until  are met’, but is updated recursively, i.e. ‘search until a sub-search or parallel search takes precedence’, and without guarantee that the initial search is resumed [14].

6.5    Choice criteria

The investigation of actual navigation should afford an understanding of users' choice of particular resources over others. Choice reflects users' criteria [15] for relevance, which hinge on a host of situational factors.

The choice triggering navigation to the hidden resource may be grounded in emergent needs or in scheduled intention. Human agents have emergent needs, schedules execute the encoded intentions of their designers (e.g., the instructor's plan, the film maker's plot, or the evaluator's method). Emergent needs range from specific to vague; they may afford a specific choice which predicts the target, or undirected serendipitous activity.

Emergent choice according to situated relevance must be distinguished from constrained choice. While emergent presentation does not constrain choice (although users' actions often follow established patterns), schedules may constrain the choice space (as in most CAL software) or both choice space and choice time (as in video games). For example, the RB 211-535-E4 maintenance training courseware produced by Rolls Royce's Customer Training School prescribes the spatial choice space, e.g. by asking the user to click on a set of maintenance steps in the right order, while the user decides when to trigger the prescribed navigation. Of course, the choice space only exists on the respective level of abstraction. The user may choose to abandon the CAL session and have a coffee break, or change his or her job.

Users' choice criteria are often transparent in that users rarely make explicit choices between equally possible navigation options. Life is no controlled experiment; instead, domain patterns and users' history of experience (whether in the space of years of working life or in the space of minutes and hours of evaluation) suggest the most appropriate choice, even if careful analysis would prove advantages of another option [16]. Choice appears mainly when it is articulated, e.g. when options appear in menus, or when the inactive of two evaluation users proposes an alternative navigation option to the active user, or when the evaluator asks a user to explain a transparent choice. In the latter case, users' explanations regularly account for their actions without claiming exhaustive rationality. Instead, they openly admit the limitations of their judgement. Their choice patterns prefer available, i.e. familiar, resources, and shy from the effort of learning and adapting to new, still unpredictable, resources.

Why users choose particular resources over others depends on a host of qualifying criteria. On can distinguish three different types of choice criteria: there are (1) situational criteria which together define the weight of presentation in general and navigation in particular, (2) operational criteria which together define the navigation distance of the resource; (3) rendering criteria which together define the navigation pleasure. Presentation weight has already been covered in chapter 3–Context. Here, I will turn to navigation distance and navigation pleasure.

The theory of presentation suggests the hypothesis that provided that resources satisfy situated validation protocols, users prefer navigations with as little weight as possible; which require the shortest navigation distance; and which afford most navigation pleasure.

6.6    Navigation distance

The availability of navigable resources is inversely related to estimated navigation distance [17]. Conditions of availability are that pointer and target are provided (e.g. not misplaced or tied up by other users), pointers are resonant (familiar or inferable), and the target navigable at an effort which remains within situated temporal tolerances (cf. section 7.2 Temporal tolerances in chapter 7–Confluence).

Availability increases with reduced navigation distance. Distance is lowest in noticing units within the display. It increases when switching to adjacent units within the same resource that are possibly removed by several steps, and increases further when switching to other resources which may involve walking to shelves, other departments or other buildings.

Navigation distance accounts for the fact that slow systems such as Rolls Royce's electronic Engine Manual DRUID [18] are rarely used except for tasks where they offer definite advantages over the paper equivalent (e.g., for word or part number search). Delays are however accepted if the target is crucial for presentation—for example, requesting black box data after aircraft accidents.

Resource transformation is often motivated by users' wish to save navigation distance. Engineers may photocopy sub-tasks and attach them to problem-specific messages to customers in order to cut navigation distance in the target setting. Or they photocopy relevant pages from the technical manuals in order to minimise their own navigation distance (and to avoid tying up the whole manual while waiting for a customer phone call).

[diagram of dimensions affecting navigation distance]

Figure 6.2. An overview of the dimensions affecting navigation distance.

Navigation distance can be divided into spatial distance [19] to resource or resource interface; access distance from the point of entry to the target view; target predictability; and navigation effort (cf. figure 6.2).

Spatial distance interacts with resource-specific access protocols and the emergent complexity of use. For example, engineers often shout across several desks when they need just one detail from a colleague so that they anticipate resource use to be very brief. If the answer turns out to be more complex than expected or occasions further questions, the enquiring engineer will get up and walk over to the colleague's desk to adapt the protocol to the emergent requirements for a prolonged presentation.

Access distance may be measured in access complexity, step distance or delay. Access complexity is given in poorly documented login protocols. Visual access complexity exists in noisy views. An example is the ‘forest of pipes’ in the IPC. Step distance describes the number of individual actions for accessing the target view. On its own it is of little relevance since perceived step distance depends on the familiarity of the access protocol which may semantically merge separate operational steps (cf. section 3.2 Protocols in chapter 3–Context). Delay depends on resource type and volume. Accessing colleagues—tapping someone's brain—is usually quickest. Access to material resources shows exponentially growing search times the larger the set of units. The electronic equivalent are the impact of file or database size on processing speed, e.g., during searches. Other factors are bottlenecks such as data transfer rate.

Predictability and effort are important qualifiers of navigation distance. Unpredictable spatial or temporal distance, for example, through delegating presentation to a black box setting, negatively affects all advance co-ordination needed for resource confluence.

The predictability of the navigation target reduces distance so that an awkward but familiar navigation involving several steps is often preferred to one which is short and elegant but less familiar. For example, users in cinegram evaluations generally preferred midpoint navigation via the overview diagram to lateral navigation to adjacent component views via the buttons placed at the end of incoming and outgoing oil pipes.

Predictability increases with increased feedback and control granularity. Feedback through rapid system response means that navigation alternatives can be tried without loosing too much time. Feedback and fine control granularity are provided in direct-manipulation interfaces [20] and features such as incremental loading of on-line pages which allows interruption of navigation as soon as the display reveals the target as unwanted.

A particular aspect of predictability is reversibility. Controls allowing absolute or relative reset-options in case of breakdown (‘Where is start?’, ‘How do I get back?’) decrease delay and repair costs. Reversibility is problematic in asymmetrical navigation such as zoom-in. For example, most cinegram users intuitively understood that clicking on map icons triggered zooming-in to the selected component. But there is no equivalent intuitive pointer to reverse zoom-in navigation, simply because the narrowing of the view cuts off the periphery. Reversal requires knowledge of operational protocols or formulation of a descriptor for searching for a fitting pointer.

Navigation effort increases with the complexity of operational protocols which may warrant auxiliary navigation such as accessing help with the risk of errors and mistakes. It decreases with increased resource visibility that affords recognition instead of necessitating recall [21].  Physical effort, such as having to walk down to the technical library or to the Production and Test facilities, is often offset by navigation pleasure, such a legitimised pause and serendipitous social contact, chats, etc. during navigation.

Navigation will proceed with the least possible effort, but only provided that the respective validation context is saturated. If the validation context produces a sharply defined need, the effort may verge towards the infinite. When travelling abroad, I must find my passport for the validation context of border control, and nothing else will do [22]. With more relaxed validation contexts, effort will settle for less. The validation context of novice users without engineering experience is lax since it emerges through the very presentation it tentatively validates. Questions saturate prematurely as soon as its descriptor can be matched to a display token. For example, a question may ask for the location of component (Y). Once inspection discovers (Y) on the currently open overview diagram, the answer is likely to be ‘After component (X) and before component (Z)’ with no further navigation effort, although an investigation of the component level would afford a more precise answer such as ‘Y is mounted at the inlet of Z’.

6.7    Navigation pleasure

Navigation pleasure describes users' sensory and social gratification through their navigating activity. Provided that navigation can be expected to saturate the validation protocol, for example, be precise enough and timely enough, users prefer more pleasurable navigations. Dimensions of navigation pleasure are diversion, serendipity, and concretion (cf. figure 6.3).

[diagram of dimensions affecting navigation pleasure]

Figure 6.3. An overview of the dimensions affecting navigation pleasure.

Diversion of navigation to less direct and more pleasurable tracks is justified by resources lacking quality. Diversion may involve a flip-over from navigation to articulation and user design. In the service engineering domain, the lack of a fitting removal procedure justifies the diversion of walking down to the Production and Test Facilities in order to talk informally to people with hands-on experience. Difficulties in navigating a report justify calling or meeting with someone in the setting where it originated to get qualification. An unstructured performance monitoring printout that is difficult to navigate justifies its transformation into a graph. The lack of a resource for converting times across the globe justifies the design of a software to this end by one service engineer. In evaluations, users lack of familiarity with the cinegram justified diversion from navigation to articulation, i.e., to extended talking and hypothesising, often with considerable problem drift.

However, it is clear that there is no simple inverse correlation between lack of resource quality and pleasure of diversion since often enough diversion may be burdensome or impossible in the face of response pressure. But equally, there is no unambiguous link between resource quality and navigation pleasure. As pointed out in chapter 3–Context, users' presentation interest may favour diversions which run counter to the business interest of maximising efficiency—an interest which motivates optimisation of resource quality.

Serendipity describes the peripheral surplus afforded by navigation. It is lowest in noticing, higher in switching, particularly when moving through space (for example, to fetch some document) since here, navigation serendipitously crosses trajectories of other users. Conversation is also rich in serendipity, i.e., unexpected news or unforeseen qualifications. Serendipity increases both navigation pleasure and resource quality. It creates confluent resources which resonate and remind users of aspects which one specific resource alone would not reveal.

Concretion expresses the preference for the tangible object over resources with greater reference distance. The equipment (e.g., a failed ball bearing) is the most concrete resource; it has closely coupled visual, tactile and kinesthetic [23] dimensions. Importantly, handling equipment allows pleasurable immediate and fine-grained control and feedback: it affords an infinite range of views and levels of magnification, and direct testing through manipulation such as scratching, shaking, knocking, bending, or moving to feel friction. For example, in identifying a bearing case fragment, service engineers prefer sifting through a box of spares in search of a possible match to delegating presentation to the lab which would identify the case through spectral analysis techniques.

The object is more concrete than its illustration, the illustration more concrete than a description. The pleasure of concretion applies notwithstanding the fact that there are many dimensions of objects which cannot be revealed by concretion, such as load or tolerances, composition of an alloy, or price.

In documents, concretion is afforded through the added realism of dimensions showing depth, process, degree of freedom or causal interdependencies of parts, whether in state views or animation. Concretion is related to serendipity in that it offers a surplus of dimensions in resources which may not be essential, but may at times provide important cues for differentiation that transparently help avoid errors. For example, the three-dimensional rendering of a component improves parts differentiation. Animated views afford picking up flow direction immediately and thereby reduce noticing complexity. Navigating rich views provides pleasure as a side effect, independent of navigation efficiency—an aspect repeatedly mentioned by cinegram users. An economic argument for rich resources affording pleasurable navigation is that it may ease users' acceptance of new document systems.

Articulation and navigation are intimately tied together in confluent presentation, which is the subject of the next chapter.


Footnotes to chapter 6-Navigation

[1] 'Cf. the classification in Pettersson & Kindborg (1991). The BS 7830 draft (1995 p71) treats navigation merely as implemented navigation feature; similarly, MIL-M-87268 (1992) treats it as implemented navigation function.

[2] The term 'online' is here used for all computer-based hypermedia documents regardless of whether the document content is stored locally or remotely.

[3] The methodological regime of most experimental investigations into aspects of what is here called navigation confines the experience of subjects to particular levels of navigation. Investigations of noticing limit research to a prescribed finite display (Hegarty 1988; Rogers 1995; Narayanan et al 1995 p510). Other approaches investigate switching and navigational distance, but only within one resource type (e.g., a computer system—Wright & Lickorish 1990) and without taking into account the semantics of display content (Seidler & Wickens 1992). This study proposes to treat navigation as one integrated activity which combines delegating, switching and noticing.

[4] Cf. Gaver's (1991) concept of affordances in media spaces, which is developed from Gibson's (1979) concept of affordance.

[5] Conversation analysis has described navigational moves within articulations under the heading of turn-taking (cf. Sacks & Schegloff 1974 or Stubbs 1983)

[6] The situation of service engineering is peculiar in that it is exempted from the internal market which normally tacks a price to any delegation.

[7] Cf. Waller's (1987) discussion of access structures encoded through typographic means.

[8] Cf. Smyth et al (1994 p33), or Landsdale & Ormerod (1994).

[9] Some authors have defined diagrams in terms of a notational system, or visual language (e.g., Marcus 1980; Waller & Anderson 1981; Guri 1984; Richards 1984). ‘To be notational a system must consist of separate and differentiated elements that correlate with equally separate and differentiated referents in the symbol system's field of reference’. (Guri 1984 p39) At the extreme end of such notationality, very restricted types, for example, derivatives of Venn diagrams in blackboard systems, can be treated as isomor-phic to propositional logic so that they are executable by a declarative pro-gramming language using graphics as native data type (cf. Lakin 1986).

[10] Narayanan et al (1995 p518), for example, measured that 39% of focus shifts in the inspection of schematic diagrams of simple mechanical devices went to connected components. The authors were however unable to account for 2/5 of all focus shifts, after attributing shifts to connectivity/contacts, visualisation, or search for information respectively (depending on verbal protocols and video data). Hegarty (1988) used eye-tracking experiments to establish in which way users understood diagrams of pulley systems accompanied by text. Her results show that users take their cues for the deciphering of diagrams from the text. She indicates however that ‘the locus of the difficulty in processing text and diagrams is in its reliance on familiarity with the domain or its spatial nature. In contrast, linguistic processing does not seem to be a source of difficulty’ (op cit p27).

[11] Narayanan et al (1995 p505n) ground the focus of attention in ‘an element in the diagram…or a conceptual aspect of an element…that is being heeded at any given moment during problem-solving’. A unit-based concept of focus is also implicit in the method of eye-tracking (cf. Hegarty 1988, 1995). The identified unit however cannot reveal the degree of users' resonance and the crucial contribution of remembered and expected peripheral tokens. The realised focus always transcends the singular token, if only for the fact that its realisation is contingent on the problem pattern aggregated so far.

[12] The concept of action sequence simply reflects the temporality of descriptive language: it alphabetises perception. Another reason for rendering perception in terms of a sequential action sequence is the commitment to a modular view of cognition where percepts are seen as the building blocks of higher processes. Since the problem appears as a causal-incremental action sequence, the supposedly underlying perceptual building blocks are forced into the same sequential order. The action sequence may then be constructed through a mapping of observable units of speech (generated through a think-aloud protocol) onto discrete interactions between hypothesised cognitive modular processes, such as information flows between perceptual process, visual interaction process, and problem-solving process (cf. Rogers 1995 p496). The particular assignments between units of speech and cognitive processes, however, seem rather arbitrary; they can only rest on surface features of an utterance which, for reasons of the sequentialisation enforced by the medium of speech, may lag behind or precede—and thereby afford—the aggregated meaning.

[13] Cf. Dervin's (1983) concept of cognitive gap bridging between information needs and answers (quoted after Schamber 1991 p33).

[14] Cf. Hert (no date) who investigated OPAC search behaviour using a grounded theory approach.

[15] Schamber (1991) defines criteria as situated and subjective, as ‘part of users' perceptions of their overall information seeking and use situations’ (op cit p5). Criteria, in this view, arise from ‘a value judgement made on the quality of a relationship between information and information need…at a certain time in an information seeking and use situation’ (op cit p3). Schamber does not deal with the problem of translating subjective criteria into objective dimensions of designed systems. She expects criteria derived from users' ‘own words with regard to their own situations’ (op cit p7) to reveal ‘dimensions of relevance’ (op cit p11) which help ‘design…systems… to effectively deliver relevant information to users’ (op cit p10) [my italics].

[16] This may be explained by the pace of external processes forcing decisions which in turn limit the time spent on using resources and favours incomplete but strategically relevant information. Cf. Linder (1970), quoted after Lamb (1996).

[17] While Seidler & Wickens (1992) define navigation distance narrowly as the number of choice points lying between screens, de Young (1990 p241) takes it to be ‘the time it takes to go from source to destination’. Hutchins 1986 p93) qualifies that ‘…directness is never a property of the interface alone, but involves a relationship between the task the user has in mind and the way that task can be accomplished via the interface’.

[18] There are many substantially similar CD-ROM based electronic manual systems used by manufacturers and operators. At British Airways I saw the DISC system and the Pinpoint system. The latter affords updating with temporary revisions from a floppy disk. At Lufthansa I saw the BISAM system which integrates the manual with fault reporting and service history data. All mentioned systems mimic the page layout of the printed manual, with references turned into hot-text hyperlinks. There is a paucity of literature on these systems. The pilot of the DISC system is described in Jones (1991 pp101).

[19] Schamber (1991) claims that the extant literature on information relevance judgements has missed out the criterion of geographic proximity.

[20] The HCI literature has discussed display feedback under the heading of ‘direct manipulation’ (Shneiderman 1987) or ‘direct engagement’ in interacting with model world metaphor (Hutchins et al 1986 p94).

[21] Cf. the discussion of affordances for availability in display design in Lansdale & Ormerod (1994). Menus, for example, don't have to be learned because ‘they are always available for inspection. They constitute an external memory’ (op cit p90).

[22] I owe this example to John Lindsay (personal communication).

[23] ‘Kinesthesia feeds into motor memory…I strongly suspect that … the stability of the kinesthetic model is more important than that of the visual model’ (Tognazzini 1992 p137/138).

Last update: 08 November 2007 | Impressum—Imprint