Accepted for publication in the Proceedings of Graphics Interface 2000 - PDF document

Accepted for publication in the Proceedings of Graphics Interface 2000, Montreal, May 2000.1Affordances: Clarifying and Evolving a ConceptJoanna McGrenere Department of Computer ScienceUniversity of TorontoToronto, OntarioCanada M5S 3G4joanna@dgp.toronto.eduWayne HoUser-Centered DesignIBM Software Solutions Toronto Laboratory1150 Eglinton Ave. East, Toronto, OntarioCanada M3C 1H7who@ca.ibm.comAbstractThe concept of affordance is popular in the HCIcommunity but not well understood. Donald Normanappropriated the concept of affordances from James J.Gibson for the design of common objects and bothimplicitly and explicitly adjusted the meaning given byGibson. There was, however, ambiguity in Norman’soriginal definition and use of affordances which he hassubsequently made efforts to clarify. His definitiongerminated quickly and through a review of the HCIliterature we show that this ambiguity has lead to widelyvarying uses of the concept. Norman has recentlyacknowledged the ambiguity, however, importantclarifications remain. Using affordances as a basis, weelucidate the role of the designer and the distinctionbetween usefulness and usability. We expand Gibson’sdefinition into a framework for design.Keywords: Affordance, usefulness, usability, design.1 IntroductionThe affordance concept was popularized in the HCIcommunity through Donald Norman’s book ThePsychology of Everyday Things (POET) [14]. The wordaffordance was new to the HCI vocabulary and theconcept seemed somewhat novel: an affordance is thedesign aspect of an object which suggests how theobject should be used [14]. It is not widely known thatthe word affordance was first coined by the perceptualpsychologist James J. Gibson in his seminal book TheEcological Approach to Visual Perception [5]. Gibsonand Norman appear at first glance to have similardefinitions of the concept. Gibson intended anaffordance to mean an action possibility available in theenvironment to an individual, independent of theindividual’s ability to perceive this possibility.Norman’s definition spread quickly and some inherentambiguities have lead to widely varying usage in theHCI literature. This inconsistent usage motivated a morethorough look at the similarities and importantdifferences between the two definitions.We first look at affordances as they were originallydefined by Gibson. We turn next to Norman’sintroduction of affordances into the HCI community andhis subsequent coverage of the concept. The differencesbetween the two uses are identified followed by a briefsurvey of the use of the concept in the HCI literature.We clarify a number of ambiguities that remain todayincluding the meaning of affordances in applicationsoftware. Lastly we provide a design framework thatextends Gibson’s definition of affordances.2 Gibson’s AffordancesGibson’s academic career centered on the field of visualperception [5]. He deviated from the classical theoriesof perception that were based on physics and physicaloptics because he felt that physics provided aninappropriate frame of reference for visual perception.Gibson made it his life’s work to describe anappropriate ecological frame of reference. He believedthat studying the animal’s visual perception in isolationfrom the environment that is perceived resulted in falseunderstandings. Gibson claimed that we perceive at thelevel of mediums, surfaces, and substances rather thanat the level of particles and atoms and, in particular, wetend to perceive what the combination of mediums,surfaces, and substances offer us. Thus “…theaffordances of the environment are what it offers theanimal, what it provides or furnishes, either for good orill. [5, p.127]”There are three fundamental properties of anaffordance:1. An affordance exists relative to the actioncapabilities of a particular actor.2. The existence of an affordance is independent ofthe actor’s ability to perceive it.3. An affordance does not change as the needs andgoals of the actor change.To elucidate the first property Gibson gives the exampleof a horizontal, flat, extended, and rigid surface thataffords support. A given surface that provides supportfor one actor, may not provide support for another actor(perhaps because of a differential in weight or size).There is only one surface in question here, yet theaffordance of support exists for one actor whereas itdoes not exist for another. Note that the affordance isnot a property of the experience of the actor but ratherof the action capabilities of the actor. Also note that Accepted for publication in the Proceedings of Graphics Interface 2000, Montreal, May 2000.2InformationIndividualOpticsEnvironment tobe PerceivedAffordanceActionDirectPerceptionFigure 1: Direct perception is the act of picking upinformation to guide action.even if the surface is not intended to provide support, ifit does in fact support a given actor, then the affordanceof support exists. The second and third properties pointto the fact that an affordance is invariant.Defined in this way, affordances cut across thesubjective/objective barrier. They are objective in thattheir existence does not depend on value, meaning, orinterpretation. Yet they are subjective in that an actor isneeded as a frame of reference. By cutting across thesubjective/objective barrier, Gibson’s affordancesintroduce the idea of the actor-environment mutuality;the actor and the environment make an inseparable pair.Gibson focussed his work on direct perception, aform of perception that does not require mediation orinternal processing by an actor. Direct perception ispossible when there is an affordance and there isinformation in the environment that uniquely specifiesthat affordance (see Figure 11). For example, one willperceive that one can walk forward when one sees asolid, opaque surface that extends under one’s feet. Theaffordance is walkability and the information thatspecifies walkability is a perceived invariantcombination of a solid, opaque surface of a certain sizerelative to oneself. Direct perception depends on theactor’s “picking up” the information that specifies theaffordance and may depend on the actor’s experiencesand culture. Let us be clear, the existence of theaffordance is independent of the actor’s experiences andculture, whereas the ability to perceive the affordancemay be dependent on these. Thus, an actor may need tolearn to discriminate the information in order toperceive directly. In this way learning can be seen as aprocess of discriminating patterns in the world, ratherthan one of supplementing sensory information withpast experience.Given that the existence of an affordance and theinformation that specifies the affordance areindependent, there are cases where an affordance existsbut there is no information to specify the affordance.Take, for example, a hidden door in a paneled room.The door affords passage to an appropriately sizedindividual even though there is no information tospecify that passage is in fact an action possibility. Heredirect visual perception is clearly not possible.There are two properties of affordances that Gibsonimplies but never directly states. The first is thataffordances are binary; they either exist or they do notexist. For example, a stair is climbable by a particularindividual or it isn’t. Gibson does not address the grayarea where an action possibility exists but it can only beundertaken with great difficulty: for example, a stair that 1 This diagram is a simplification of Gibson’s view of directperception. See Gibson, 1979 [5] for a more complete description.is climbable but only with great difficulty. Second,Gibson implies that affordances can be nested when anaction possibility is composed of one or more actionpossibilities. For instance, an apple affords eating, buteating is composed of biting, chewing, and swallowing,all of which are afforded by the apple. Gibson describesthe environment as being composed of nested objectsand he describes the nesting of information thatspecifies affordances but he never specifically uses theterm nested affordances.3 Norman’s AffordancesAffordances, as Gibson described them, can becontrasted with Norman’s affordances introduced inPOET. Norman described affordances as follows:…the term affordance refers to the perceived and actualproperties of the thing, primarily those fundamentalproperties that determine just how the thing couldpossibly be used. A chair affords (‘is for’) support and,therefore, affords sitting. A chair can also be carried. [14,p.9]This quotation points to some apparent differencesbetween Norman’s affordances and Gibson’saffordances. Norman talks of both perceived and actualproperties and implies that a perceived property may ormay not be an actual property, but regardless, it is anaffordance. Thus, he deviates from Gibson in thatperception by an individual may be involved incharacterizing the existence of the affordance. Further,Norman indicates that an affordance refers primarily tothe fundamental properties of an object. Gibson, on theother hand, does not make the distinction between thedifferent affordances of an object. Another importantdifference is that for Norman there is no actor as aframe of reference.Norman makes clear in an endnote in POET that heis deviating from the Gibsonian definition ofaffordances:The notion of affordance and the insights it providesoriginated with J.J. Gibson, a psychologist interested inhow people see the world. I believe that affordances resultfrom the mental interpretation of things, based on our pastknowledge and experience applied to our perception ofthe things about us. My view is somewhat in conflict with Accepted for publication in the Proceedings of Graphics Interface 2000, Montreal, May 2000.3the views of many Gibsonian psychologists, but thisinternal debate within modern psychology is of littlerelevance here. [14, p. 219]This quotation identifies another difference betweenGibson and Norman. Gibson claims that the existence ofaffordances is independent of an actor’s experience andculture. Norman, on the other hand, tightly couplesaffordances with past knowledge and experience. Theframe of reference for Gibson is the action capabilitiesof the actor, whereas for Norman it is the mental andperceptual capabilities of the actor.It is important to clarify Norman’s position thataffordances are perceived properties. He states thataffordances “provide strong clues to the operations ofthings” [14, p.9] and that they “suggest the range ofpossibilities” [14, p.82]. He argues that when designerstake advantage of affordances, the user knows what todo just by looking. Although complex things mayrequire supporting information, simple things shouldnot. If they do, then design has failed.In more recent books, Norman stresses theimportance of perceived affordances [15, 16, 17] anddifferentiates them from real affordances:It’s very important to distinguish real from perceivedaffordances. Design is about both, but the perceivedaffordances are what determine usability. I didn’t makethis point sufficiently clear in my book and I have spentmuch time trying to clarify the now widespread misuse ofthe term. [17, p. 123]This clarification will likely help to mitigate futuremisuse, but it still does not clearly separate theaffordance from the information specifying theaffordance.In a recent article on the topic of affordances [18],Norman begins to separate affordances from theirvisibility and thus deviates from his original usage.Unfortunately, some misconceptions about affordancesand the role of the designer remain in that article. Weaddress these in the discussion section.4 Highlighting and Interpreting the DifferencesWe will use what has become the canonical example ofaffordances in the HCI literature, namely the affordanceof a door, to elucidate the differences between Gibson’sand Norman’s original use of the concept. Consider adoor with no handle and no flat panel. Without priorknowledge of how the door operated, an actor wouldfind it difficult to know the direction of opening.Following Gibson’s definition, the fact that the door canbe opened by a given actor is sufficient to determinethat it has an affordance. (Perhaps the door can bepushed and it will swing away from the actor or theactor can grasp the door edges and pull.) There does notneed to be any visual information specifying the correctdirection to the actor for there to be an affordance.According to Norman’s use, on the other hand, theaffordance would only exist if there was information tospecify the possibility for action and the actor hadlearned how to interpret the information. In this case,there would need to be a door handle that signaled thedirection of opening to the actor. If we were to redrawFigure 1 using Norman’s definition, the two sections onthe right, Optics and the Environment to be Perceived,would be collapsed into a single section.Table 1 highlights the different meanings assigned toaffordances by Norman and Gibson.Gibson’s Affordances· Offerings or action possibilities in the environment inrelation to the action capabilities of an actor· Independent of the actor’s experience, knowledge,culture, or ability to perceive· Existence is binary – an affordance exists or it does notexistNorman’s Affordances· Perceived properties that may or may not actually exist· Suggestions or clues as to how to use the properties· Can be dependent on the experience, knowledge, orculture of the actor· Can make an action difficult or easyTable 1: Comparison of affordances as defined byGibson and Norman.The most fundamental difference between the twodefinitions is that for Gibson an affordance is the actionpossibility itself whereas according to Norman’s use ithas been both the action possibility and the way that thataction possibility is conveyed or made visible to theactor. Norman’s “make it visible” guideline actuallymaps quite nicely to Gibson’s statement that there mustbe perceptual information that specifies the affordancefor the affordance to be directly perceived. We believethat this difference has caused confusion in the HCIcommunity. In his original definition, Norman collapsedtwo very important but different, and perhaps evenindependent, aspects of design: designing the utility ofan object and designing the way in which that utility isconveyed to the user of the object. Because Norman hasstressed (but not entirely limited himself to) perceivedaffordances, he has actually favored the latter of thetwo. In Gibsonian terms, these two aspects are labeled:design of the affordances of an object and design of theperceptual information that specifies the affordances.It is important to note that Norman and Gibson hadtwo related yet different goals. Gibson was primarilyinterested in how we perceive the environment. Heacknowledged that both people and animals manipulate(that is, design) their environment to change what it Accepted for publication in the Proceedings of Graphics Interface 2000, Montreal, May 2000.4affords them, but the manner of manipulation was nothis focus. Norman, on the other hand, is specificallyinterested in manipulating or designing the environmentso that utility can be perceived easily. We speculatethat, given Gibson’s focus, he made the simplifyingassumption that affordances are binary. Recall theexample of a stair being climbable or non-climbable bya particular individual. Reality obviously isn’t this blackand white; a gray area exists that is meaningful to thestair climber. For a particular individual one stair maybe climbable with great difficulty whereas a differentstair may be climbable with ease. Gibson doesn’taddress this range; they are both climbable and thusthey both qualify as affordances. From a designperspective, an affordance that is extremely difficult toundertake versus one that is undertaken with ease canhardly be put in the same category. In the design ofeveryday things, the goal should be to designinformation that uniquely specifies an affordance andalso to design useful affordances that can be undertakenwith ease.Warren [26], an ecological psychologist, movesbeyond binary affordances. He defines p numbers to bedimensionless ratios that provide measurements of theactor in relation to the environment. He has donedetailed analysis of the affordance of stair climability,for which he uses p = R/L as the intrinsic measure,where R is the riser height of the stair and L is theclimber’s leg length. For climbers of different heights,Warren was able to determine a single optimal point(p0) at which the energy expenditure required to climbthrough a given vertical distance is at a minimum and asingle critical point (pmax) at which point a stairbecomes impossible to climb bipedally. Using Warren’sterms, the goal of design should be to achieve theoptimal point for the target user.5 Affordances as They Appear In the HCI CommunityIn order to understand how the affordance concept hasbeen adopted by the HCI community we conducted asurvey of the literature. We focussed mostly on theproceedings from the annual CHI conferences2 becausewe felt these proceedings to be generally representativeof the HCI literature. Nineteen papers were reviewed.The goal was to identify and loosely categorize how theterm affordance has been used. Three high-levelcategories emerged:· 8 papers adhering to Gibson’s definition – an actionpossibility or offering [1, 2, 4, 6, 20, 22, 23, 27] 2 Papers were selected using the ACM Digital Library and GaryPerlman’s HCI Bibliography with the search string “affordance.” Allthose papers that appear in the CHI proceedings have been reviewedand a few others were also selected based on availability.· 6 papers adhering to Norman’s original definition –a perceived suggestion [3, 7, 8, 10, 13, 19]· 5 papers deviating from both Gibson and Norman[11, 12, 21, 24, 25]For reasons of brevity we only highlight a couple ofpapers in each category.5.1 Gibson’s Affordances - An Action PossibilityPapers that used Gibson’s definition fall into twocategories: the affordances of software applications [1,2, 4, 23] and the affordances of physical objects [6, 20,22, 27].Action Possibility in Software ApplicationsGaver [4] published the first paper in the CHIProceedings that included the concept of affordances.This paper goes beyond the mention of affordances; it isspecifically about affordances. Because Gaver’scontribution is substantial, we discuss his work in depthat the end of this section. Another example of a paper inthis category is by Smets, Overbeeke, and Gaver [23].They show how the design of forms can conveycomplex non-visual information such as sound, taste,smell, and texture. They postulate that this researchcould be applied to the design of icons that representcomplex information and activities and thereby improvethe information that specifies the affordance.Action Possibility in a Physical ObjectZhai, Milgram, and Buxton [27] document a study thatstrongly suggests that high-degree-of-freedom inputdevices should be designed so that they can bemanipulated by the fingers because finger movementsoften provide more accurate control than do armmovements. Thus, these input devices should be shapedand sized so as to afford finger manipulation.5.2 Norman’s Affordance – A Perceived SuggestionMihnkern [10] describes affordances as the means ofcommunicating a design model to the user. He says thatwhen a metaphor is applied to a system, it gives thesystem a particular set of affordances and that themetaphor inevitably breaks down leaving some of thesystem’s features affordance-less or invisible. [InGibsonian terms, even if there is no information tospecify the affordance, it still exists.]Johnson [7] compares a number of techniques forpanning, in particular, moving the scene under thewindow or moving the window over the scene (GUIs dothe latter):… it is clear that the appearance of the touch-display caninfluence what people suggest [is the panning method].This is what Gibson and, later, Norman refer to as an‘affordance’: when an aspect of an artifact’s designsuggests how it is to be used. We thought that adding abrightly colored border around the displayed image might Accepted for publication in the Proceedings of Graphics Interface 2000, Montreal, May 2000.5suggest ‘touch here’ to users, and might therefore suggestTouch Edge panning (camera or background). [7, p.219]5.3 Neither Gibson’s nor Norman’s AffordancesWith the exception of the first paper in this category[11], the use of affordance in the papers is unclear [12,21, 24, 25].An Interface ObjectMohageg et al. [11] equates an affordance with aninterface object: “all of this functionality is mapped ontoa single affordance on the dashboard.” [11, p.468] Here,they are referring to a virtual joystick.Unclear UsageVaughan [25] provides a confusing account ofaffordances. She seemingly identifies the affordance ofmovement. She talks about the movement of a butterflyaffording chasing and that when movement becomesmore prominent the affordance of emotion becomesmore evident. She cites Gibson, yet her use ofaffordances appears different from both Gibson’s andNorman’s.5.4 Acknowledging Gaver’s ContributionAs noted above, there are a number of authors who areaware that affordances originated with Gibson and haveread Gibson’s work. Yet most who cite Gibson andperhaps even quote him resort to using the meaninggiven by Norman. One author in particular, Gaver,makes a significant attempt to bring Gibson’s ideas intothe HCI community in his paper entitled “TechnologyAffordances”, [4] which is illuminating and thereforeneeds to be outlined in some detail. This was the firstpaper in the CHI Proceedings that discussesaffordances; but, it has gone largely unnoticed.Gaver’s discussion of the door example illustratesthat his understanding of affordances differs fromNorman’s. Where Norman and all who followed talkedabout the affordance suggesting the action, Gaver talksabout the design that suggests the affordance of thedoor. Here he uses the term design as the informationthat specifies the affordance. He uses the door exampleto demonstrate nested affordances, which he defines as“affordances that are grouped in space.” The affordanceof pulling a door handle is nested within the affordanceof opening the door. Gaver recognizes the importanceof distinguishing two aspects of design:Distinguishing affordances and the available informationabout them from their actual perception allows us toconsider affordances as properties that can be designedand analyzed in their own terms. [4, p. 81]Gaver identified apparent affordances:In general, when the apparent affordances of an artifactmatch its intended use, the artifact is easy to operate.When apparent affordances suggest different actions thanthose for which the object is designed, errors are commonand signs are necessary. [4, p.80]These match what Norman has termed perceivedaffordances. Gaver provides a framework for separatingaffordances from the perceptual information availableabout them (Figure 2). Note that Gaver’s perceptibleaffordance is not the same as his apparent affordance orNorman’s perceived affordance, as we have shown byoverlaying the latter two on Gaver’s framework.Perceptible Affordance and Hidden Affordancemake sense but False Affordance is problematic. It isnot the affordance that is false; rather, it is theinformation that is false. Gibson uses the termmisinformation to describe this phenomenon. Whenmisinformation is picked up by an actor, thenmisperception results. Gibson acknowledges that the“line between the pickup of misinformation and thefailure to pick up information is hard to draw.” [5,p.244]Interestingly, Gaver does seem to contradict himselfpart way through his paper when he finally gives aconcrete definition of affordances:The concept of affordances points to a rather specialconfiguration of properties. It implies that the physicalattributes of the thing to be acted upon are compatiblewith those of the actor, that information about thoseattributes is available in a form compatible with aperceptual system, and (implicitly) that these attributesand the action they make possible are relevant to a cultureand a perceiver. [4, p.81]Here he seems to be lumping in the information thatspecifies the affordance with the affordance itself.Gibson’s affordances only cover the first of these threepoints.Gaver then addresses the problem of complexaffordances. He extends the notion of affordances toexplicitly include exploration. He introduces theconcept of sequential affordances, which refers tosituations in which action on a perceptible affordanceleads to information indicating new affordances (e.g.,falseaffordanceperceptibleaffordancecorrectrejectionhiddenaffordancenoyesaffordancenoyesperceptualinformationperceived/apparentaffordancesFigure 2: Separating affordances from the perceptualinformation that specifies affordances (adapted from [4]). Accepted for publication in the Proceedings of Graphics Interface 2000, Montreal, May 2000.6after mousing-down on the scrollbar, it can then bedragged). Sequential affordances explain howaffordances can be revealed over time. As previouslymentioned, Gibson implies the existence of nestedaffordances but never actually identifies them. Gaver,on the other hand, specifically defines nestedaffordances to be affordances that are grouped in space.He provides the example of manipulating the scrollbarwidget as being nested within the affordance ofscrolling within a window. Here Gaver is exploringaffordances for low-level interaction in GUIs, which wedeal with in greater detail in the next section.6 Discussion6.1 Does it matter?In the end, does establishing a clear meaning ofaffordances really matter? We argue that it does matter.At the most basic level, establishing a concrete meaningwill prevent widely varying uses of the term. Norman, inhis latest article, also sees the need for clarification:“Sloppy thinking about the concepts and tactics oftenleads to sloppiness in design. And sloppiness in designtranslates into confusion for users. [18, p. 41]”In the same way, we are motivated to further clarifyaffordances in terms of design, and specifically in thearea of software design. To this end, we return toGibson’s definition of affordances and discuss itsimpact on design.6.2 Usefulness and UsabilityClearly differentiating the two aspects of design iscritical: designing affordances and designing theinformation that specifies the affordance should not beconfounded. Said in another way, designing the utilityof an object is related to but separate from designing theusability of an object. This is a distinction of usefulnessversus usability [9].The HCI community has largely focussed onusability at the expense of usefulness. Norman alsoemphasizes usability: “The designer cares more aboutwhat actions the user perceives to be possible than whatis true” [18, p. 39]. A designer must also be concernedwith creating the useful actions of the design, creatingwhat is truly possible in the design. A useful designcontains the right functions required for users toperform their jobs efficiently and to accomplish theirgoals. The usefulness of a design is determined by whatthe design affords (that is, the possibilities for action inthe design) and whether these affordances match thegoals of the user and allow the necessary work to beaccomplished. The usability of a design can beenhanced by clearly designing the perceptualinformation that specifies these affordances. Usabledesigns have information specifying affordances thataccounts for various attributes of the end-users,including their cultural conventions and level ofexpertise. Of course, usability is also enhanced byfollowing principles such as providing appropriatefeedback, being consistent, and providing errorrecovery. Figure 1 can be redrawn to show therelationship between usefulness and affordances and therelationship between usability and the informationspecifying an affordance (see Figure 3).6.3 Clarifying Affordances in Software DesignIt is necessary to clarify the meaning of an affordance inthe context of application software. There wasconsiderable ambiguity on this in the reviewed HCIliterature and there is additional confusion in Norman’slatest article [18]. An affordance is an action possibilityor an offering. Possible actions on a computer systeminclude physical interaction with devices such as thescreen, keyboard, and mouse. But the role ofaffordances does not end with the physical aspect of thesystem, as Norman implies [18]. The applicationsoftware also provides possible actions. A wordprocessor affords writing and editing at a high level, butit also affords clicking, scrolling, dragging anddropping. The functions that are invokable by the userare the affordances in software. Functions may includetext-editing, searching, or drawing. The information thatspecifies these functions may be graphical (buttons,menus) or it may not exist at all.Norman claims that a scrollbar is a learnedconvention and implies that it is not an affordance [18].We disagree. The fact that the object affords scrolling isan affordance that is built into the software. Theinformation that specifies this affordance is in fact alearned convention – we have all come to recognize ascrollbar.In general, an underlying affordance or function canstill exist regardless of correct interpretation or evenperception by the user. A low-level user action triggersthe execution of the function. The action could be theinput of some obscure command (e.g., “ls -la”) at aprompt or it could be clicking on a button in a GUI. Inthe first case, there is little or no information to specifythe affordance. In the second case, there is someUserInterfaceUsabilityFunctionalityUsefulnessAffordanceActionDirectPerceptionFigure 3: Usefulness and Usability.Information Accepted for publication in the Proceedings of Graphics Interface 2000, Montreal, May 2000.7information. This case relies on the notion of nestedaffordances. The button has a clickability affordance,which is specified by a raised-looking push button. Butusers are not interested in clicking on a button for itsown sake; they are interested in invoking some function.It is generally the icon or the label on the button thatspecifies the function to be invoked. Therefore, buttonclickability is nested within the affordance of functioninvokability. This is much the same as we woulddescribe a piano as having an affordance of musicplayability. Nested within this affordance, the pianokeys have the affordance of depressability.It is important to note that affordances exist (or arenested) in a hierarchy and that the levels of thehierarchy may or may not map to system functions. Inother words, affordances do not necessarily map one-to-one onto system functions. Taking a standard GUI-based word processor as an example, we can say that itaffords document editing. Editing includes affordancesfor text addition and deletion, margin adjustment, fontselection, and many others.As Gaver identified, there are also sequentialaffordances, that is, affordances that are only availableat certain points in time. Although such affordances alsoexist outside GUIs and applications, they are perhapsmore obvious here given the dynamic nature of softwareand the ability to update the display quickly. Theinformation that specifies an affordance can be updatedas new affordances become available. Once a userclicks a visible button, a drop down menu may appear,from which the user can then make a selection. This isnot to say that all applications update the visualinformation to specify the available affordances. TheUNIX text editor vi, for example, gives the user novisual information about whether text entry is possible.In command mode, a user must first switch to inputmode before entering text. It is impossible to discernfrom simply looking at the screen whether the system isin command mode or input mode.6.4 Affordances as a Framework for DesignTo use affordances to evaluate and improve design, it isuseful to think of the degree of an affordance. To regardaffordances as binary is to oversimplify them. Warren’s[26] work on p numbers, and specifically the optimaland the critical points, began to address what we call thedegree of an affordance. However, we still requirelanguage to describe affordances that exist betweenthese two points and we need to incorporate theinformation that specifies the affordance. We can thinkof a two-dimensional space where one dimensiondescribes the ease with which an affordance can beundertaken and the second dimension describes theclarity of the information that describes the existingaffordance. Each of these dimensions is a continuum.The goal of design is to first determine the necessaryaffordances and then to maximize each of thesedimensions. If both dimensions are of equal importancefor a given affordance, improvements in design shouldbe seen to move along the diagonal given in Figure 4.Note that while determining the necessary affordances isrelated to usefulness, making an improvement in eitherof these dimensions is related to usability.Personal customization of an interface provides agood example of how a user can improve the design of asystem to make the affordance easier to undertake. Forinstance, a user may make an alias for a long commandstring (for example, turning “lpr –Pmyprinter” into“lpm”) or may add a button to a toolbar for a frequentlyused command. Thus, an affordance is easier toundertake when the time to perform the action isreduced. It can also be made easier by increasing thephysical comfort or reducing the exertion required. Acommand that requires a single key to invoke isphysically easier than one that requires the simultaneouspressing of multiple keys.By comparing a GUI to a command-line interfacewe can understand how the degree of informationspecifying the affordance can be varied. Command-lineinterfaces often provide little or no information aboutthe options that are available to the user. GUIs, on theother hand, provide significant information. Despite theavailable information in a GUI, expert users tend toprefer command-line interfaces. Their preference can beunderstood in the context of this two-dimensionalframework; it is faster to enter a short command via thekeyboard than to move the hand to the mouse, positionthe pointer, and click. Expert users have committedthese commands to memory and so the visualinformation is clutter and the mouse access is a slow-down. For novice users, having visual information andmouse access is easier than committing a series ofcommand strings to memory. This same informationincreasingly easy to undertakeaffordanceDegree of Affordanceincreasingly clear informationDegree ofPerceptualInformationimprovementsin design thatmaximize bothdimensionsFigure 4: Representing the affordance and theinformation that specifies the affordance on a continuum. Accepted for publication in the Proceedings of Graphics Interface 2000, Montreal, May 2000.8comes at the cost of making the affordance moredifficult to undertake for expert users. Thus, the degreeof an affordance exists relative to a particular user.7 ConclusionWithout Norman’s adoption of affordances in POETand his ongoing writing, affordances would likely beunfamiliar to many of us. It has been necessary for us tobe detailed with respect to Norman’s use of affordancesbecause otherwise it would not be possible to sort outthe misuse and the current confusions that remain. Weapplaud Norman’s efforts in bringing this importantconcept to our community and continuing to clarify it.As the concept of affordances is used currently, ithas marginal value because it lacks specific meaning.Returning to a definition close to that of Gibson’s wouldsolidify the concept and would also recognize thatdesigning the utility or functional purpose is aworthwhile endeavor in its own right. In order for theaffordance concept to be used fully in the design world,however, Gibson’s definition needs to incorporate thenotion of varying degrees of an affordance. We haveprovided a framework for design that is based on thisexpanded notion of an affordance.8 AcknowledgmentsWe are grateful to Kim Vicente for many engagingdiscussions on affordances and his thoughtful commentson earlier drafts of this paper. This research has beensupported by the Centre for Advanced Studies at theIBM Toronto Laboratory and by NSERC.References[1] Ackerman, M.S., and Palen, L. (1996). The zephyr helpinstance: Promoting ongoing activity in a cscw system.CHI 96 Conference Proceedings, 268-275.[2] Bers, M.U., Ackermann, E., Cassell, J., Donegan, B.,Gonzalez-Heydrich, J., DeMaso, D.R., Strohecker, C.,Lualdi, S., Bromley, D., and Karlin, J. (1998). Interactivestorytelling environments: Coping with cardiac illness atBoston’s Children’s Hospital. CHI 98 ConferenceProceedings, 603-610.[3] Conn, A.P. (1995). Time affordances: The time factor indiagnostic usability heuristics. CHI ’95 ConferenceProceedings. 186-193.[4] Gaver, W.W. (1991). Technology affordances. CHI’91Conference Proceedings. 79-84.[5] Gibson, J.J. (1979). The Ecological Approach to VisualPerception. Boston: Houghton Mifflin.[6] Harrison, B.L, Fishkin, K.P., Gujar, A., Mochon, C., andWant, R. (1998). Squeeze me, hold me, tilt me! Anexploration of manipulative user interfaces. CHI 98Conference Proceedings, 17-24.[7] Johnson, J.A. (1995). A comparison of user interfaces forpanning on a touch-controlled display. CHI’95Conference Proceedings, 218-225.[8] Kohlert, D.C., and Olsen, D.R. (1995). Pictures andinput data. CHI’95 Conference Proceedings, 464-471.[9] Landauer, T.K. (1995). The Trouble with Computers:Usefulness, Usability, and Productivity, Cambridge,MA: The MIT Press.[10] Mihnkern, K. (1997). Visual interaction design: Beyondthe interface metaphor. SIGCHI Bulletin, 29(2), 11-15.[11] Mohageg, M., Myers, R., Marrin, C., Kent, J., Mott, D.,and Isaacs, P. (1996). A user interface for accessing 3Dcontent on the world wide web. CHI 96 ConferenceProceedings, 466-472.[12] Moran, T.P., Palen, L., Harrison, S., Chiu, P., Kimber,D., Minneman, S., van Melle, W., and Zellweger, P.(1997). “I’ll Get That Off the Audio”: A case study ofsalvaging multimedia meeting records. CHI 97Conference Proceedings, 202-209.[13] Nielsen, J., and Wagner, A. (1996). User interface designfor the WWW. CHI 96 Conference Companion, 330-331.[14] Norman, D.A. (1988). The Psychology of EverydayThings. New York: Basic Books.[15] Norman, D.A. (1992). Turn Signals are the FacialExpression of Automobiles. Reading, MA: Addison-Wesley.[16] Norman, D.A. (1993). Things That Make Us Smart.Reading, MA: Addison-Wesley.[17] Norman, D.A. (1998). The Invisible Computer.Cambridge: MA, MITPress.[18] Norman, D.A. (1999). Affordance, conventions, anddesign. Interactions, 6(3), 38-42.[19] Perkins, R. (1995). The interchange online network:Simplifying information access. CHI’95 ConferenceProceedings, 558-565.[20] Schilit, B.N., Golovchinsky, G., and Price, M. (1998).Beyond paper: Supporting active reading with free formdigital ink annotations. CHI 98 Conference Proceedings,249-256.[21] Shafrir, E. and Nabkel, J. (1994). Visual access to hyper-information: Using multiple metaphors with graphicaffordances. CHI’94 Conference Proceedings. 142.[22] Sellen, A., and Harper, R. (1997). Paper as an analyticresource for the design of new technologies. CHI 97Conference Proceedings, 319-326.[23] Smets, G., Overbeeke, K., and Gaver, W. (1994). Form-giving: Expressing the nonobvious. CHI’94 ConferenceProceedings, 79-84.[24] Tamura, H., and Bannai, Y. (1996). Real3communication and aromatic group computing: HCI andCSCW research at Canon Media Technology Laboratory.CHI 96 Conference Companion, 131-132.[25] Vaughan, L.C. (1997). Understanding movement. CHI97 Conference Proceedings, 548-549.[26] Warren (1995). Constructing an econiche. In J. Flach, P.Hancock, J. Caird, and K. J. Vicente (Eds.), GlobalPerspectives on the Ecology of Human-MachineSystems, (pp. 210-237). Hillsdale, NJ: LawrenceErlbaum Associates.[27] Zhai, S., Milgram, P., and Buxton, W. (1996). Theinfluence of muscle groups on performance of multipledegree of-freedom input. CHI 96 ConferenceProceedings, 308-315.