Goal-Orientation, Goal-Setting, and Goal-Driven Behavior in Minimalist User Instructions

IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 50, NO. 4, DECEMBER 2007 295

Goal-Orientation, Goal-Setting, and Goal-Driven Behavior inMinimalist User Instructions—HANS VAN DER MEIJ

Abstract—This paper opens with a summary of minimalist design strategies that aim to optimize user instructions.Next, it discusses three recent research efforts to further improve these strategies. The common focus in these effortsis the attention to people’s goal-related management and control of attention, time, and effort. First, a comprehensiveframework for designing procedures—the Four Components model—is described. The design principles for thegoal component focus on supporting the user’s goal orientation and goal setting. Second, two experiments arereported that studied the problem of when it is best to present conceptual information. When instructions employed alearning-by-doing approach, users clearly preferred a work-flow mode of presentation. This mode optimally exploitsthe user’s momentary interest in conceptual information during goal-driven task execution. The third research effortconcentrates on user affect in instructions. The main idea is that motivation and emotion play a key role in taskappraisals and corresponding actions. After discussing theories and design approaches, an experiment is discussedin which instructions were optimized for affect. Good results for perceived relevance and self-confidence were foundin all conditions. Presence of an affect-oriented co-user did not enhance these effects. The paper concludes that thecontribution of the efforts extends beyond the minimalist framework from which they originated.

Index Terms—Affect, design, just-in-time delivery, procedural instructions.

The minimalist design approach has played animportant role in optimization of user support inlearning-to-use software. This approach originatedwith John Carroll and his colleagues at IBM whowere looking for a way to assist people’s actionsand sense-making efforts [1]–[3]. The designstrategies employed in the approach have evolvedover the years into the four key principles and theirunderlying heuristics presented in Table I [4].

Various changes over time in these design strategiesdid not touch the core of this approach, but theydid help designers by giving more precise guidancefor understanding and applying minimalism, andthey offered opportunities for experimental researchon its distinct features. The design strategiesdescribed in Table I have functioned well for a longtime in research and practical efforts to implementminimalism. However, new developments and otherdemands have prompted further explorations. Inthis paper, I discuss three recent research projectsconducted in collaboration with colleagues andstudents. The theme that connects these studiesfocuses on goal-related resource managementstrategies. Each study examined what users do

Manuscript received July 3, 2006; revised January 30, 2007.The author is with the Faculty of Educational Scienceand Technology, Department of Instructional Technology,Twente University, 7500 AE Enschede, the Netherlands(e-mail: [email protected]).Color versions of one or more of the figures in this paper areavailable online at http://ieeexplore.ieee.org.

IEEE 10.1109/TPC.2007.908728

TABLE IMINIMALIST DESIGN PRINCIPLES AND HEURISTICS

to manage and control their attention, time, andeffort, and how design can support these processes.

I begin by introducing the Four Componentsmodel, a comprehensive and concise frameworkfor analyzing and designing procedures. Themodel was described in detail recently in IEEETRANSACTIONS ON PROFESSIONAL COMMUNICATION [5].

0361-1434/$25.00 © 2007 IEEE

296 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 50, NO. 4, DECEMBER 2007

Only the design strategies for the goal componentwill be discussed here.

Next, I discuss an empirical study on the designprinciple of just-in-time information. It revolvedaround the question of whether users benefitmore from having conceptually linked informationpresented together, or from having such informationpresented as needed during task execution. Thetested prediction was that people are goal-focusedand highly selective in the information they process.They are likely to adopt a parsimonious resourcemanagement strategy by allocating attention, time,and effort primarily to information that is pertinentfor immediate task execution.

The third research effort focused on useraffect. Theoretical models of affect indicate thatpeople’s goal orientation and goal setting are keydeterminants of their attitudes and actions inself-directed learning, and covary with resourcemanagement strategies. Design approaches try totune into this insight in several ways, by proposingstrategies that reduce the negative impact (e.g.,avoidance reaction) of unpleasant emotions. Afterdiscussing two such approaches, an experimentin which a special attempt was made to influenceuser affect in “minimal” manuals is summarized.The tested prediction was that designing for useraffect positively influences perceived relevanceand self-confidence. The study also exploredwhether there was a trade-off with training time,perceived task-difficulty, and learning, such thatresource management strategies for cognitiveinformation-processing are adversely affected.This paper ends with a short discussion of thecontribution of these efforts to instructional designapproaches for user documentation in general.

THE FOUR COMPONENTS MODEL

In an insightful paper, Farkas suggests that thedesign of procedural instructions can be basedon combining the perspectives of systems theoryand rhetoric [6]. Goals and goal-directed actionsplay a vital role in systems theory, which arguesthat a system is a collection of highly integratedparts set out to realize an overall goal. The inputsof such a system go through certain processesto produce outputs that accomplish the overalldesired goal for the system. Farkas translates thesebasic notions from systems theory into an overviewof the main actions and states in human–computerinteractions. The descriptors for the various statesthat he distinguishes explicitly refer to goals. Forexample, the desired state is the goal presented to

the user and the interim state is the intermediatestate or subgoal. Actions by the user, the system,or outside sources can support or obstruct goalachievement.

The design of procedural instructions can only bepartly based on this logic, however, Farkas argues.The designer also needs to take the context andthe user characteristics into consideration. It isimportant to communicate in such a way thatthe audience both understands the message andis willing to act when instructed to do so. Forexample, a user may perceive and understandinformation about a desired state in the title ofa procedure, but may not be convinced that thisis a desirable state that he or she should try toaccomplish. To be convincing, the designer needsto communicate clearly and to establish sourcecredibility (i.e., “selling oneself”) and productcredibility (i.e., “selling the domain”). A rhetoricalapproach can support the designer in addressingissues of meaning and persuasion.

Building from Farkas’ analysis, we have proposed aFour Components model [5], [7] that describes themain building blocks or components in proceduresand their corresponding theory-based or empiricallytested design guidelines. According to the model,procedures always consist of one or more ofthe following components: goals, prerequisites,actions and reactions, and unwanted states. Theguidelines vary from abstract proposals to concretesuggestions, as illustrated in the ones for goals (seeTable II).

TABLE IIDESIGN GUIDELINES FOR GOALS

VAN DER MEIJ: GOAL-ORIENTATION, GOAL-SETTING, AND GOAL-DRIVEN BEHAVIOR 297

Some guidelines link factors that influence theuser’s goal achievement. For example, Guideline3 (to sell the goal in its description) aims tosupport goal orientation and refers to the needto communicate to the user the general reasonsfor engaging in a task. Likewise, the advice tobreak down certain goals into subgoals reflects thenotion that users may lose sight of their main goalwhen too many actions intervene. Titles should betask-oriented to achieve goal setting; task-orientedtitles establish an entry and objective for the user’sactions.

The Four Components model complements theminimalist approach. It offers the designer moreguidance than before in designing procedures inminimal manuals. In addition, the componentsand their corresponding guidelines have somethingto say about procedures in other manuals aswell. Although originally intended to improveprocedures in minimal manuals, this modelcan also be employed fruitfully to analyze and(re)design procedural information in other manuals.Systematic studies to validate its claims have yet tobe conducted, but we have already found the modeluseful in audits for various clients who asked us toevaluate their documentation.

DESIGNING IN A JUST-IN-TIME DELIVERY MODE

The minimalist Principle 1 (choosing anaction-oriented approach in Table I) directlyaddresses the user’s goal orientation. It is basedon the finding that people turn to documentationand support primarily for task completion. Theirgeneral purpose is to act, to do. The minimalistPrinciple 2 (anchoring the tool in the task domain inTable I) is linked with goal setting. It addresses thefundamental issue of finding ways to address theuser’s specific interest in the software as a meansto accomplish personal objectives. Realizing suchobjectives is much more difficult to accomplishthan action-orientation. This is nicely illustratedby the blooming market for dedicated (sections in)manuals for specific audiences. Together, the twoprinciples lead to documentation and support thatrevolve around authentic tasks.

In recent instructional design theories, one sees thesame focus on engaging people in authentic tasksin doing and learning [8], [9]. There are two riskswith these approaches to self-regulated learning.One is that they often do not make immediateactivity a high enough priority. For example, theinstructions may begin with an orientation to thesemantics of a domain. Such an explanation is

valuable to the user, of course, but it constitutes adistraction when positioned at the very beginning.The other risk is that these approaches do notgive the user’s motivation a high enough priority.Authentic tasks are employed primarily becausethey help the user integrate the knowledge,skills, and attitudes necessary for effective taskperformance processing. The motivational appealand benefits of such tasks are secondary, at best.

A general risk of presenting authentic or real-lifetasks is that they may overwhelm the user. Toreduce the possibility of cognitive overload fromtask complexity, designers can follow a number ofdesign strategies. For example, they can scaffoldwhole-task practice by presenting the learningtasks in a simple-to-complex order and they can usea fading approach in which explicit, action-orientedsupport is gradually reduced. Another popularsolution is the adoption of worked-out examplesthat take the user by the hand in completing a task.

Yet another design strategy is that of just-in-timedelivery of information. Minimalism has alwaysespecially advocated this strategy, often incombination with the strategy of giving “justenough information.” One of the most hotly debatedaspects of the implementation of these strategies iswhat Carroll and Rosson have dubbed the “paradoxof the active user” [10, p. 80]. In essence, thisparadox argues that people need to do in orderto learn, while they also need to learn in order todo. (The paradox is sometimes also interpretedas a reference to the persistent application ofinefficient procedures by users, even experiencedones; see [11].) In the practice of designing userdocumentation, the paradox often translates intothe question of when it is best to present conceptualinformation to the user.

We examined this issue in two studies, both inthe context of minimalist instructions that taughtusers how to program a machine with ComputerNumerically Controlled (CNC) language [12]. A CNCprogram can direct a milling machine in creatingobjects such as a chess piece by carving away slicesof wood or steel. CNC is difficult to learn because ithinges on getting to know a large set of seeminglyarbitrarily numbered CNC-codes (henceforth simplyreferred to as codes) that must be entered in a fixedsequence in a program. Besides knowledge of thecodes and language in general, learning to use CNCalso involves getting to know the two main factorsinvolved in working with the machine. One of theseis the dimensional aspect, which refers to the shapeof the work piece. The other is the technical aspect,


which refers to the process of machining. Thisaspect includes different types of cutting operations(e.g., leveling, rough-turn, and grooving), cuttingconditions, and tool types, among others.

The main research question was whether usersbenefitted more from having conceptually linkedinformation presented together, or from havingsuch information presented as needed during taskexecution. Just as in all minimalist designs, thebasic design of the instructions in both conditionsin the experiments took a learning-by-doingapproach. Variations were created by presentingconceptual information (i.e., codes) either inconceptual mode or in work-flow mode. Theprediction was that the work-flow mode would bebest attuned to the user’s goal achievement, leadingto proper attention for this information.

In conceptual mode, the presentation wasoptimized for reading for understanding. That is,explanatory sections included the relevant code(s).Such sections always appeared right before theaction steps. To draw the user’s attention to thecodes and to enhance their accessibility, codeswere highlighted with a colored contour (see Fig. 1).The presentation mode was expected to connectexplanations and codes better, and to improve theuser’s understanding of the links between relatedcodes.

In work-flow mode (i.e., just-in-time delivery), thepresentation of codes was optimized for processingthe instructions one at a time. The critical code(s)for a job appeared exactly when users needed themduring task execution. That is, codes appearedright before the action steps that included them (seeFig. 2). A superficial inspection of the instructionsconveyed an image of codes scattered randomlythroughout the manual.

Participants in both studies were students (ages16–24) from secondary vocational technicaltraining schools with little or no knowledge of CNCprogramming. The primary task of the participants,who were randomly assigned to conditions, was towrite a CNC program using the instructions forsupport. The training session took a maximumof 90 minutes. Participants were instructed tothink aloud during training and were observedby researchers to capture how they processedinstructions. For these observations we used theISTE (Information SubTypes Effect) approach inwhich the observer records whether informationtypes are: (a) attended to, or (b) processed asintended [13]. For example, we recorded whether

participants skipped, scanned, or read all units inthe manual that presented conceptual information.Learning was assessed two weeks after trainingwith a paper-and-pencil test that asked participantsto write parts of a CNC program without support.

Our prediction was confirmed in both studies. Userswith instructions in work-flow mode processedsignificantly more information from the manualthan users with instructions in conceptual-mode(81% versus 70%). They also processed instructionsmore accurately. For example, compared to theusers working with instructions in conceptualmode, work-flow mode users more often readthan scanned conceptual information, and theymore often performed rather than simply readaction steps. The outcomes for learning were alsosignificantly higher. Participants who had worked

Fig. 1. Instructions with conceptual information (i.e.,CNC-codes, circled for the reader) in conceptual mode.


with instructions in work-flow mode scored betteron the macro and meso-structure (i.e., lines ofcodes) of their CNC program.

This finding seems to conflict with the advice fromresearch taking a learning-by-reading approach,which suggests that linked concepts shouldbe presented before, rather than during, taskexecution. The argument is that a work-flowdelivery of linked concepts can overtax a userwho already has to cope with a complex task[14]. This seeming conflict can be resolved whenone considers the difference in overall designapproach. In learning-by-reading approaches, aconceptual mode of presentation makes sense.However, this is not the favored approach ofpeople who consult a tutorial to learn how touse an application. In such a situation, people

Fig. 2. Instructions with conceptual information (i.e.,CNC-codes, circled for the reader) in work-flow mode.

prefer and adopt a learning-by-doing approach.The two reported experiments indicate that inthis situation, conceptual information is bestgiven in a just-in-time fashion (i.e., in work-flowmode), because such a strategy optimally exploitsthe moment at which the user is motivated toprocess such information. Users are most likely toattend to and spend time and effort to understandconceptual information when it is critical for theimmediate goal they are pursuing.

DESIGNING FOR USER AFFECT

In reaction to the dominant research on “cold”cognition, an increasing number of educationalresearchers have begun to examine how affect(i.e., motivation and emotion) influences doingand learning, and how affect can be influenced byinstructions. I briefly discuss these two strands ofresearch on “warm” cognition below. For theories,I provide an overall framework in which two mainviews are distinguished. Each view subsumesa great number of varied and detailed theorieson affect and its impact on people’s actions. Fordesign, there are several approaches that offeradvice on how to influence motivation, but thereare just two design approaches that give guidelineson how to support or enhance both motivation andemotion. I discuss only the latter.

One theoretical view focuses on the joint effectof motivational processes and mental resourcesmanagement on people’s actions [15]. Accordingto this view, goals play a key role in motivation.People’s goal orientation and goal setting influencestheir intention to engage mental resources inhandling a task. Goals may also further influencethe direction of these efforts. For example, researchon goal orientation indicates that people who desireto learn to master a task respond to task challengeswith increased mental effort, whereas people witha performance orientation more often react bylowering their goals or by task avoidance. Similarly,research on goal setting shows that people regularlycheck whether their expectancies and valuescompare favorably with their initial intention towork on a task. If the outcome is favorable, peopleare likely to continue to engage mental resourcessuch as attention and monitoring to allow them tokeep working on the task until it is successfullycompleted. With an unfavorable outcome, suchactivities are either reduced or stopped.

Another theoretical view concentrates on thecomplex interactions between perceptual, cognitive,and emotional processes in doing and learning


[16]. The assumption is that task completionfollows a sequence of perceptual-cognitive,cognitive-emotional, and emotional processing.First, perceptual-cognitive processes yield anappraisal of a task. A key role is played in thisphase by factors such as relevance, valence, anddifficulty. RELEVANCE refers to the importanceof the task to the person’s goals or concerns.VALENCE is its pleasantness or unpleasantness.DIFFICULTY concerns the person’s assessment ofhis or her ability to deal with the task. Second,cognitive-emotional conditions of feelings arecontrolled and processed during task execution.For example, a person may engage in emotionaltracking to assess whether emotional goals such assatisfaction, relaxation, or stimulation are possibleor achieved. Third, emotional processes result ingeneral feelings towards the task. For example,aesthetic feelings may emerge from an evaluationof the task, along with self-related feelings ofmotivation and self-confidence.

In both views the main outcomes that areassessed before, during, and after task executionare (de)activation and (un)pleasantness. Theseoutcomes are predicted to have importantconsequences for what follows. For example, ifa person feels positively activated (e.g., by beingpleasantly surprised or finding something exciting),he or she probably continues task execution untilits completion and he or she may also feel inclinedto tackle a new task. Boredom, annoyance, andother negative emotions can reduce such actions,or even stop them altogether.

Two design approaches for affect have recentlybeen advanced [17], [18]. Both approaches takeoff from the idea that beneficial effects on doingand learning can be realized by strengtheningpositive affect and avoiding or reducing negativeaffect. These approaches therefore aim forinstructions that: (a) assist people in capitalizingon the beneficial influence of positive motivationand emotions, (b) minimize the risk that peopleexperience negative affect, and (c) moderate theimpact of unpleasant events that do occur. The twodesign approaches have little in common beyondshared general aims. That is, they differ in boththe design strategies that they advocate and themotivation and emotions that they address.

As presented by Astleitner, FEASP aims to reducethe negative consequences of Fear, Envy, and Angerand to enhance the positive feelings of Sympathyand Pleasure during doing and learning [17]. Itproposes a different set of design strategies for

each emotion. For example, to reduce fear, thefollowing strategies are advocated: ensure successin learning, accept mistakes as opportunities forlearning, induce relaxation, and be critical butsustain a positive perspective. Astleitner suggeststhe following strategies to increase pleasure:enhance well-being, establish open learningopportunities, use humor, and install play-likeactivities. These design strategies proposed inFEASP have a certain ad hoc character, because,as the author acknowledges, “it was not possible todeduce strategies from one consistent theoreticalframework, not even within one category ofemotion” [17, p. 192]. The FEASP-model has notbeen empirically tested to my knowledge.

The other design approach that explicitly seeks toinfluence affect is called ECOLE, an abbreviationfor Emotional and COgnitive aspects of LEarning[18]. This approach attends to students’ emotionsof interest, well-being, anxiety, boredom, andachievement. The following five general educationalguidelines are proposed to influence these:self-regulated learning activities, arrangementof competence experiences, affordances forsocial interaction, structured instructions andlearning materials, and authentic tasks or topics.Each guideline is further detailed in one or twospecific instructional strategies. For example, toprompt self-regulation the authors suggest usingstudent-centered activities such as projects orlaboratory-based work. ECOLE was tested in 37elementary school classrooms (8th and 9th grade)in 12 to 18 lessons in biology, German, and physics.In comparison to control classrooms in whichteaching was more traditional, ECOLE classroomsproduced significantly higher achievement in alldomains. Although this is an important finding,the outcome of the study was nevertheless adisappointment because effects on the students’emotions were low to nonexistent.

And what about minimalism? How does minimalismfit into this picture? Minimalism has always fused“cold” and “warm” cognition in its design. As ause- and user-centered approach, minimalism hasconsidered both how to create instructions thatassist people in learning to use new software aswell as how to accommodate their propensitiesand needs. Numerous illustrations can be givento exemplify this view, but one of the mostinnovative ideas advanced by minimalism is itsdesign principle of supporting error recognition andrecovery.


This principle not only acknowledges that peoplefrequently face failures, breakdowns, or goalobstructions in learning to use software, but it alsoproposes that these realities need to be addressed ininstruction. Accordingly, minimalism offers variousdesign strategies for mitigating such moments. Itaddresses potentially frustrating experiences bystressing that mishaps are to be expected, explainsthat they may arise from different causes, andsupports error recovery. Because people are ofteneager to learn more about the software duringfailures or breakdowns, minimalism also exploitsthese moments to present background informationto which users would otherwise not attend. Thereis a striking but unnoticed convergence betweenminimalism and FEASP on this desire to mitigateand leverage user mistakes.

A review of the empirical research on minimalistinstructions indicates that they have yieldedmixed results for user affect. Ramsay and Oatleyreport greater satisfaction overall, and a higherpreference for minimalist instructions in re-useconditions compared to other instructions [19],but neither Lazonder and Van der Meij [20] norVan der Meij [21] found any difference betweenminimalist and conventional instructions on userattention, relevance, confidence, and satisfaction.The only study reporting absolute figures forminimalist instructions on perceived relevanceand self-confidence mentions moderate scores ofaround 6.6 on a 10-point scale [22].

These findings leave room for improvement.Therefore, in a recent experiment, we set out toexamine whether stronger effects on user affectcould be obtained if we took into account theaforementioned insights on (designing for) useraffect [23]. Participants in the control condition inthis study worked with a regular minimal manualto which we added a touch of “warm” cognitionthrough relevance organizers and modifiederror-information. This basic manual condition wascontrasted with an experimental condition in whicha co-user was added. The co-user manual featureda person who sided with the user, serving as anemotional buddy.

The research questions of the study were thefollowing: (1) Does the basic manual yield high(absolute) scores on user affect? (2) Does theco-user manual yield higher affect scores? and (3)Is there a trade-off between the two manuals onfactors of training time, perceived task-difficulty,and learning?

The two special measures to improve basicconditions for user affect in a minimal manualwere relevance organizers and modifiederror-information. Relevance organizers presentan annotated before–after display to increasetask-relevance of the training goal (see Fig. 3). Intheatrical plays, this presentation technique isknown as a “late point of attack,” and is used forengaging and persuading the audience [24]. Thevisual images of the organizers were expected todraw the users’ attention and positively influencetheir goal orientation as well as their goal setting.

Minimalist designs have generally focused onoptimizing the user’s cognitive processing ofmistakes; all four heuristics for error-handling areprimarily oriented towards user cognition. Thefollowing is a typical expression: If your screenremains empty, you may have made a typing error.Type the name of the file again and press the Enterkey. Users may perceive this as useful help, but thefocus is a cognitive one and the influence on useraffect is indirect at best.

The new error-information design explicitly soughtto address user affect in dealing with mistakes. Thatis, we adapted the vocabulary and also insertedsentences that directly addressed user affect.Vocabulary was chosen in accordance with Ortony,Clore, and Foss, who suggest that people often useadverbs and adjectives to express emotions [25].Thus, we occasionally inserted affective words toadd a soothing tone to the cognitive information(e.g., if you accidentally clicked, and not serious).In addition, we sometimes added a sentence topositively influence the user’s motivation andfeelings (e.g., This may be odd, but it’s just howthings are, and Don’t worry).

Participants in the experimental condition workedwith the basic manual to which we added a co-userwho served as a buddy. That is, the co-user wasportrayed as another student (female) whoseexpressions had been recorded while processingthe instructions. The co-user mainly vents herchanging moods during and after task execution.Her primary role is that of enhancing users’ positivefeelings and moderating frustrating experiences.She consistently tries to interact with the real userregarding motivational and emotional issues thatarise during task completion. At the start of a task,she expresses her appraisal. During task execution,she models emotional tracking by venting herfeelings. After task completion, she expresses hergeneral feelings towards the task and mentionsincreased self-confidence. Co-user presence links


Fig. 3. Relevance organizer with a before–after display.

with the FEASP principle to “establish cooperativelearning structures” with peer helpers serving asbuddies [17, p. 187].

In designing the co-user, we first created a personalprofile, and then attended to the “when,” “what,”and “how” for presenting her voice [26], [27]. (Toachieve a uniform tone of voice in the co-usermanual, we also made a minor stylistic change inthe presentation of the action steps.) The co-user,Lineke, first presents herself in the introductionof the manual. She says she is a 15-year-old girlwho, like the target audience, must frequentlyhand in well-formatted reports, at which shefails. She further expresses her interest in theinstructions and says she hopes that the effortshe spends in processing them will improve herformatting skills—the topic of the instructions. Inthe remainder of the manual, the co-user mainlycomments on successes or failures (e.g., formattingproblems, new system states, (sub)task completion,and system feedback). The content of the co-usercomments varies systematically between goals andprocesses. That is, she addresses task-relevance(e.g., Oh yes. How funny. That’s handy.), or userconfidence (e.g., I can remember this.). In addition,she expresses important thought processes (e.g.,Just one word? All one by one?), feelings (e.g.,Cool, Careful, Awful), or their combination. Eachcomment is accompanied with a matching facialpicture [28]. These pictures show affect states

and frames of minds such as happiness, gloom,appreciation, confidence, and daring (see Fig. 4).

Participants (aged 15 years and 6 months) in thestudy were from third year vocational educationaltraining classes. They could all perform basic tasksin Word, but were unable to work with Word’sformatting options for systematically creatingmargins, indents, and columns that were taught inthe manual. For dependent variables, we lookedat training time and perceived difficulty duringtask execution, immediate and delayed learningoutcomes, facility in re-use conditions (i.e., solvingtasks with access to the manual), and outcomemeasures for affect (i.e., relevance and confidence).

For the first research question, the answerwas confirmatory. The manual with the basicinstructions yielded excellent results for perceivedtask-relevance and self-confidence. Mean scores forthese measures were well above 8 on a 10-pointscale, indicating very favorable affect scores.In other words, the basic measures combiningdesigning for “warm” cognition with minimalistinstructions had been effective.

The second research question was not confirmed.The outcomes for user affect were only slightlyhigher in the co-user manual condition.

The third research question yielded severalsignificant findings, all in favor of the basic manual.For example, although we found a very favorable


Fig. 4. Section from the co-user manual.

low overall rating for task-difficulty experienced(2.2 on a 10-point scale), and users of the basicmanual gave their manual a better rating (2.1versus 2.4). These users also needed less timeto complete training (60 versus 67 minutes). Inaddition, the basic manual better facilitated taskcompletion in re-use conditions. There was no effectof condition on learning outcomes. Users of bothmanuals performed equally well on the learning testimmediately after training, and on the delayed test.

All in all, the findings in the study make a strongcase for the basic manual. The design measuresof relevance organizers and adapted vocabularyin error-information sufficed to create a goodfusion of “cold” and “warm” cognition support forthe user. However, it is too early to conclude thatthe co-user did not have any effect and needsnot be studied further. One possibility is thatthere may be a ceiling effect for affect scores.Another possibility is that the timing of the co-userinteractions could be improved. In line with designguidelines from the design literature on pedagogicalagents, the co-user was a commentator, givingonly after-the-fact feedback on successes orfailures in goal achievement. The co-user did not

address the user’s affect states during the initialgoal orientation and goal setting that theoreticalviews advance as critical conditions for motivationand emotion—these were effectively covered bythe relevance organizers. In addition, we alsohave yet to examine the impact of the co-user onparticipants with low input values for motivation. Itis conceivable that the co-user is more importantand influential for them than for users who alreadyhave high motivation at the start of the training.

CONCLUSION

The three research efforts reported in this paperwere all conducted within the framework ofminimalism; but, as the integrative theme of thestudies indicates, the insights offered are moregeneral.

Neither the construction nor the outcome ofthe Four Components model was restricted tominimalism. We started this research endeavorbecause we felt that minimalism did not offersufficiently detailed and tested ideas for designingprocedural information. Accordingly, we examineda wide spectrum of relevant (minimalist and


non-minimalist) theories and practices. For theory,we studied the broad instructional design researchliterature. To find out about existing practices,we conducted an inventory study in which weperformed in-depth analyses of procedures sampledfrom 104 commercially produced manuals. Thecombination of these perspectives led to theFour Components model that now representsour best effort to introduce systematicity andtested principles into the design of proceduralinstructions. Goals form an essential component inthe model. As codified in the title, they establish avital entry and objective for the user’s actions.

The studies on the just-in-time principle (i.e.,using a work-flow mode) were conducted withinthe overall framework of the learning-by-doingapproach advocated by minimalism. From itsconception, the minimalist design approach“emphasized encouraging and supporting workon realistic tasks from the start and throughouttraining: learning by doing rather than learningby reading” [29, p. 3]. Minimalism is not the onlyapproach advocating learning-by-doing. Practicalapproaches such as project-based education; thecase method; problem-based learning methods;and Schank, Berman, and MacPerson’s theoryof goal-based scenario [30] are all variants onthe learning-by-doing approach. The experimentssignal that these approaches should also considerpresenting conceptual information primarily inwork-flow mode. When conceptual information iswell-aligned with task execution, people are morelikely to perceive its relevance for goal achievementand give the information its proper attention.

The discussion of user affect began with twobrief theoretical accounts of how motivation andemotion may influence doing and learning. Indiscussing two design approaches on user affect,it was noted that designs based on strategies tomanage affect have yet to prove that they actuallyinfluence user affect; their link with theory alsoneeds further elaboration. We explored both issuesin our empirical study on minimalist instructions.The findings suggest that relevance organizersare an effective rhetorical tool for “selling thegoal” to users. These organizers, which can bepresented in any type of manual, directly addressthe processes of goal orientation and goal settingseen as key factors by motivation theory. Thefindings for co-user presence indicated that therewas no noticeable positive effect on user affect.Co-user presence even had a negative impact ontraining efficiency. Several reasons have already

been advanced why this does not mean that furtherresearch on co-user presence is without meaning.An additional argument is that our study did notdifferentiate between special design features (i.e.,co-user presence) and general design features(e.g., action-orientation and task-orientation) thatalso target user affect. In other words, our studyleft open the possibility that co-user presence isimportant for user affect in situations in whichgeneral design features that contribute to useraffect are missing. This is generally the case inhuman–computer interactions, which could explainthe recent surge in designs of affect-orientedpedagogical agents in such interactions [26], [28],[31].

To recapitulate, even though all of the reportedstudies were inspired by and situated withinminimalism, they address a common theme in alldesigning of documentation and support—namely,the optimization of people’s goal-relatedmanagement and control of attention, time, andeffort.

ACKNOWLEDGMENT

I very gratefully acknowledge the contribution ofthe following colleagues and students for theirhelp in conducting the various studies reported inthis paper: Hilde Agterbos, Peter Blijleven, MarkGellevij, Leonie Jansen, Ard Lazonder, Jules Pieters,Igo Weber, and Marja op de Weegh. I also thankthe anonymous reviewers of this paper for theirsupport and helpful suggestions on articulating theintegrative theme.

REFERENCES

[1] J. M. Carroll, The Nurnberg Funnel: DesigningMinimalist Instruction for Practical ComputerSkill. Cambridge, MA: MIT Press, 1990.

[2] J. M. Carroll, Minimalism Beyond the NurnbergFunnel. Cambridge, MA: MIT Press, 1998.

[3] J. M. Carroll, P. L. Smith-Kerker, J. R. Ford,and S. A. Mazur-Rimetz, “The minimal manual,”Human-Comput. Interaction, vol. 3, no. 2, pp.123–153, 1987.

[4] H. van der Meij and J. M. Carroll, “Principles andheuristics for designing minimalist instruction,”in Minimalism Beyond the Nurnberg Funnel, J. M.Carroll, Ed. Cambridge, MA: MIT Press, 1998,pp. 19–53.

[5] H. van der Meij and M. R. M. Gellevij, “The fourcomponents of procedures,” IEEE Trans. Prof.Commun., vol. 47, no. 1, pp. 5–14, Mar., 2004.

[6] D. K. Farkas, “The logical and rhetoricalconstruction of procedural discourse,” Tech.Commun., vol. 46, no. 1, pp. 42–54, 1999.


[7] H. van der Meij, P. Blijleven, and L. Jansen, “Whatmakes up a procedure?,” in Content & Complexity.Information Design in Technical Communication,M. J. Albers and B. Mazur, Eds. Mahwah, NJ:Erlbaum, 2003, pp. 129–186.

[8] M. D. Merrill, “First principles of instruction,”Educat. Technol. Res. Develop., vol. 50, no. 3, pp.43–59, 2002.

[9] J. J. G. van Merriënboer and P. A. Kirschner,“Three worlds of instructional design: State of theart and future directions,” Instructional Sci., vol.29, no. 4–5, pp. 429–441, 2001.

[10] J. M. Carroll and M. B. Rosson, “The paradox ofthe active user,” in Interfacing Thought: CognitiveAspects of Human-Computer Interaction, J. M.Carroll, Ed. Cambridge, MA: MIT Press, 1987, pp.80–111.

[11] F. Wai-Tat and W. D. Gray, “Resolving the paradoxof the active user: Stable suboptimal performancein interactive tasks,” Cognitive Sci., vol. 28, no. 6,pp. 901–935, 2004.

[12] H. M. Agterbos, H. van der Meij, and J. M. Pieters,“Mentact,” Twente University, Final Report forSVO-project 92504, 2003.

[13] H. van der Meij, “The ISTE-approach to usabilitytesting,” IEEE Trans. Prof. Commun., vol. 40, no. 3,pp. 209–223, Sep., 1997.

[14] L. Kester, C. Lehnen, P. W. M. van Gerven, and P.A. Kirschner, “Just-in-time, schematic supportiveinformation presentation during cognitive skillacquisition,” Comput. Human Behavior, vol. 22, no.1, pp. 93–112, 2006.

[15] H. Astleitner and C. Wiesner, “An integrated modelof multimedia learning and motivation,” J. Educat.Multimedia Hypermedia, vol. 13, pp. 3–21, 2004.

[16] H. Astleitner, “Multimedia elements andemotional processes,” E-Journal InstructionalSci. Technol., vol. 7, no. 2 [Online]. Available:http://www.usq.edu.au/electpub/e-jist/docs/Vol7_no2/FullPapers/MMelements.htm

[17] H. Astleitner, “Designing emotionally soundinstruction: The FEASP-approach,” InstructionalSci., vol. 28, no. 3, pp. 169–198, 2000.

[18] M. Gläser-Zikuda, S. Fuss, M. Laukenmann,K. Metz, and C. Randler, “Promoting students’emotions and achievement—Instructional designand evaluation of the ECOLE-approach,” Learningand Instruction, vol. 15, no. 5, pp. 481–495, 2005.

[19] J. E. Ramsay and K. Oatley, “Designing minimalcomputer manuals from scratch,” InstructionalSci., vol. 21, no. 1–3, pp. 85–98, 1992.

[20] A. W. Lazonder and H. van der Meij, “The minimalmanual: Is less really more?,” Int. J. Man MachineStudies, vol. 39, no. 5, pp. 729–752, 1993.

[21] H. van der Meij, “A critical assessment of theminimalist approach to documentation,” in Proc.10th Annu. Int. Conf. Systems Documentation(SIGDOC92), 1992, pp. 7–17.

[22] M. R. M. Gellevij, “Visuals in Instructions:Functions of Screen Captures in SoftwareManuals”, Ph.D. Thesis Twente University,Enschede, The Netherlands, 2002.

[23] H. van der Meij, “Designing for user cognition andaffect in software instructions,” in Learning andInstruction. The Netherlands: Elsevier. [Online].Available: http://www.sciencedirect.com/science/journal/09594752

[24] D. Goodwin, “Emplotting the reader: Motivationand technical documentation,” J. Tech. WritingCommun., vol. 21, no. 2, pp. 99–115, 1991.

[25] A. Ortony, G. L. Clore, and M. A. Foss, “Thereferential structure of the affective lexicon,”Cognitive Sci., vol. 11, no. 3, pp. 341–364, 1987.

[26] A. Paiva, I. Machado, and C. Martinho. (1999).Enriching Pedagogical Agents With EmotionalBehavior: The Case of Vincent. Tech. Rep. ,INESC. Univ. Lisboa, Portugal. [Online]. Available:http:www//gaips.inesc.pt/gaips.inesc-id.pt/gaips/shared/docs/Machado99Enriching.pdf,Technical Report

[27] M. B. Coney and M. F. Steehouder, “Roleplaying on the web: Guidelines for designing andevaluating personas online,” Tech. Commun., vol.47, no. 3, pp. 327–340, 2000.

[28] S. Brave, C. Nass, and K. Hutchinson,“Computers that care: Investigating the effects oforientation of emotion exhibited by an embodiedcomputer agent,” Int. J. Human-Comput. Studies,vol. 62, no. 2, pp. 161–178, 2005.

[29] J. M. Carroll, “Reconstructing minimalism,” inMinimalism Beyond the Nurnberg Funnel, J. M.Carroll, Ed. Cambridge, MA: MIT Press, 1998, pp.1–17.

[30] R. C. Schank, T. R. Berman, and K. A.MacPerson, “Learning by doing,” in InstructionalDesign Theories and Models: A New Paradigmof Instructional Design Theory, C. M. Reigeluth,Ed. Mahwah, NJ: Lawrence Erlbaum, 1999, vol.2, pp. 161–181.

[31] R. W. Picard and J. Klein, “Computers thatrecognize and respond to user emotion: Theoreticaland practical implications,” Interacting WithComputers, vol. 14, no. 2, pp. 141–169, 2002.

Hans van der Meij studies information-seeking, instructionaldesign, minimalism, and usability testing. He is (co-)author ofmore than 100 publications in scientific books and articles. Hisawards include STC’s Frank R. Smith Outstanding Article Awardin 1996 and 1998, and IEEE’s Professional CommunicationSociety’s Best Transactions Article Award in 1997.

Documents

Goal-Orientation, Goal-Setting, and Goal-Driven Behavior in Minimalist User Instructions