HCI Bibliography Home | HCI Conferences | BCSHCI Archive | Detailed Records | RefWorks | EndNote | Hide Abstracts
BCSHCI Tables of Contents: 858687888991929394959697

Proceedings of the HCI'86 Conference on People and Computers II

Fullname:Proceedings of the Second Conference of the British Computer Society Human Computer Interaction Specialist Group
Note:People and Computers: Designing for Usability
Editors:Michael D. Harrison; Andrew F. Monk
Location:University of York
Dates:1986-Sep-23 to 1986-Sep-26
Publisher:Cambridge University Press
Standard No:ISBN 0-521-33259-1 (out of print); hcibib: BCSHCI86
  1. Invited Papers
  2. Case Studies in Usable Design
  3. Iterative Design and Evaluation: Rapid Prototyping
  4. Office Automation
  5. Intelligent Interfaces
  6. Structuring Interaction
  7. Formal Design Methods
  8. Knowledge of the User as a Design Guide
  9. Case Studies: Evaluation
  10. Display Based Systems: Evaluation

Invited Papers

People and Computers: Designing for Usability: An Introduction to HCI-86 BIBA 3-23
  John Long
This paper provides a general introduction to HCI-86. First, the theme and aims of the conference are elaborated and the state-of-the-art of HCI assessed, as reflected in the presented papers. Then, areas of HCI poorly represented in the papers are identified to aid delegates make good the omissions by their active participation in the interactive sessions offered by the conference. The different groups making up the HCI community are identified and discussed in terms of their background disciplines. Ways of promoting understanding between the groups are proposed. The importance of the conference as a means for advancing understanding of HCI within the community is emphasised. The wider context within which HCI developments are likely to occur is briefly described.
Formal Methods and the Design of Effective User Interfaces BIBA 24-43
  Bernard Sufrin
Designing an effective user interface to a complex information system is difficult, since it cannot be done in isolation from the design of the information system itself. All too often the effectiveness of a system is compromised by the intrusion into the user interface of details of the system's implementation. We conjecture that it is only by understanding the essence of the purpose of an information system -- abstracting from the details of any proposed implementation -- that one can begin to judge the validity of design choices concerning the user-system interface. We also conjecture that the language of mathematics is a useful medium both for explaining our understanding and recording design choices. In this paper we offer support for our conjectures by discussing the design of part of an office system. We use the Z notation [Morgan & Sufrin 84, Morgan 85, Sufrin 85, Spivey 86] which is based on the language of mathematics -- extended slightly to facilitate the description of large systems.
Ergonomics in Design for Usability BIBA 44-64
  B. Shackel
In this paper some approaches to designing for usability are described which have been developed in Ergonomics (Human Factors) from experience and from empirical work reported in the literature. The basic approach of ergonomics to design problems is first summarised. Next some implications of the system design life-cycle are considered, and the multiplicity of interests and criteria during system design are noted. Then a definition of usability is proposed in operational terms and is illustrated. Finally, some precepts are offered to aid the process of design for usability.
Understanding the Nature of the Office for the Design of Third Wave Office Systems BIBA 65-77
  Niels Bjorn-Andersen
It is argued that technologists have a far too analytical, rationalistic and one-dimensional perception of what constitutes an office. This means that many office systems fail or that they only survive thanks to substantial contextualization processes on the part of the users. The desktop metaphor was a large step in the direction of improving the user interface. However, it is based exclusively on the level of the individual with his/her machine. It is suggested that we should develop a richer understanding of the office drawing much more on social science disciplines. Fragments in such a richer view of the office are suggested and some implications for office systems design are suggested.
Ease of Use -- The Ultimate Deception BIBA 78-94
  Harold Thimbleby
A correspondence is drawn between the historical development of mathematics and the development of users' conceptual models of interactive computer systems. Many mathematical concepts took centuries to resolve but computer users are often expected to handle comparable issues much more rapidly.
   Insights into user interface issues are drawn from non-standard analysis and non-Euclidean geometry. Mindful of Godel, I argue that if a system is sufficiently powerful to be 'easy to use' this implies it is sufficient to confuse.

Case Studies in Usable Design

Human Factors in the Columbus Space Station BIBA 97-114
  Ian Alexander; Ged Morrisroe; Pat Norris; Andrew Tindell
The Columbus space station is intended to provide a working environment for scientists and engineers for up to a year each, carrying out experiments such as pharmaceutical processing, isotopic separation, and protein crystallisation. The majority of the crew will not be trained astronauts. Special attention has therefore been paid, throughout the design of Columbus, to human factors. The crew in the Pressurised Module are to have 'a comfortable shirt sleeve environment'. Given the complexity of Columbus, much of the crews' work will involve interacting with the computers that control the data processing, onboard environment, experiments, communications, and so on. Work is currently in progress on all aspects of human-computer interaction.
   Four examples of Logica's work are described here. The Anthrorack prototype modelled the MMI of the GRiD computer to be flown in Spacelab. The CISE experiment will assess the difficulty of using different forms of MMI on the Skynet flights. An Adaptive User Interface may provide Pressurised Module crew with "intelligent" access to the many services offered. The Data Management System is being designed with the user in mind.
Tools for Management and Support of Multiple Constraints in a Writer's Assistant BIBA 115-131
  Claire O'Malley; Mike Sharples
Conventional text editors have limitations that make them ill-suited to composing and manipulating large pieces of structural text. Recently, several researchers have begun to develop a new generation of text editors that act as "writer's assistants", allowing the user to view and alter the organisational structure of text, as well as providing tools to support related tasks, such as spelling correction and proofreading. These systems represent major progress in the development of writing, as opposed to editing, tools. However, they have two important limitations. Firstly, they do not appear to be derived from an explicit model of the writing process -- one that is based on observations of expert writers. Secondly, and as a result, they are limited in their ability to represent and satisfy the constraints involved in writing. This paper sets out a framework for the design of a Writer's Assistant, based on a model derived from current research on the writing process. The system would give support to the user engaged in a variety of writing tasks, by explicitly representing the structure of text at various levels, by providing tools for the manipulation and transformation of such structures, and by taking over some of the demands of planning and constraint satisfaction.
MacCadd, An Enabling Software Method Support Tool BIBA 132-154
  John Jones
Designers of computer hardware have had the benefit of computer aided design for many years. More recently the software industry is looking to computer applications to help in the design of software systems. This requires support for published methods for creating designs. The methods for "structured" software analysis and design often make use of diagrams as the notation for presenting, communicating and thinking about the design. There is therefore the need for design tools which support a graphical interface in an intelligent and enabling way. This paper described the goals and features of MacCadd, a tool with a graphical interface for support of specific methods using hierarchical network and tree diagrams. The paper presents the interface idioms which have been found to be desirable and the user reaction to MacCadd. In conclusion there are the open questions which suggest the future directions which could be taken, and the problems they raise in the interface.

Iterative Design and Evaluation: Rapid Prototyping

ECS -- A Technique for the Formal Specification and Rapid Prototyping of Human-Computer Interaction BIBA 157-179
  Heather Alexander
Increasingly, formal specification and rapid prototyping are recommended as techniques to be used in developing software. In particular they are appropriate when developing user interfaces, given the increased demand for sophisticated interactive software and the difficulty of producing it. Formally specifying the user interface allows the designer to reason about its properties in the light of the many guidelines on the subject. Early availability of prototypes of the user interface allows the designer to experiment with alternative options and to elicit feedback from potential users. This paper reports an extension to an existing formal specification and prototyping method, called me too, to handle the interaction required for an application.
Rapid Prototyping of Dialogue for Human Factors Research: The EASIE Approach BIBA 180-195
  Allan MacLean; Phil Barnard; Michael Wilson
Facilities for the rapid prototyping of dialogue are an extremely important component of a successful User Interface Management System (UIMS). Exactly how the UIMS should be optimised will depend on the type of application being developed and the environment in which it is being used. This paper focusses on the support for dialogue construction provided in EASIE (Experimental Applications System for Integrated Environments). EASIE is specifically designed to support human factors research into the human interface of so called 'integrated systems' by providing both flexibility and simplicity in the construction and modification of the dialogue. This is done by treating the dialogue specification at two distinct levels. A Dialogue Script (DS) text file contains the minimum amount of information necessary to define and modify the dialogue, and a separate Dialogue Interface (DI) maps the DS onto the basic functionality of the application.
The Role of Iterative Evaluation in Designing Systems for Usability BIBA 196-214
  Thomas T. Hewett
Increasingly it appears that design of interactive computing systems should be an iterative process of design and re-design. One factor which appears to be a driving force in successful iterative design is iterative evaluation -- evaluation as part of each design cycle. One type of evaluation -- formative -- involves monitoring the process and products of system development and gathering user feedback for use in refinement and further system development. A second type of evaluation -- summative -- involves assessing the impact, usability and effectiveness of the system. Different cycles in the evaluation-design process require different types of evaluation, or require different mixes of the two. An extensive example illustrates some of the ways in which the nature and purposes of evaluation may change during different cycles of evaluation.

Office Automation

Towards the Successful Design and Implementation of Computer Based Management Information Systems in Small Companies BIBA 217-234
  Brenda Wroe
The coincidence of general economic depression, keen competition and the microelectronics revolution has presented small companies with the need and the opportunity to implement relatively sophisticated computer-based management information systems (CBMIS). However the implementation is rarely troublefree and the effectiveness of the resultant live CBMIS often fails to meet the expectations of the management and users. Research into the success of CBMIS development in the small business environment was recently undertaken in a longitudinal study of ten small U.K. construction companies. The primary aim was to identify the nature of problems experienced in the CBMIS development process, and suggest guidelines for achieving greater success in the small business sector. A model was derived to explain the interaction or organisational, application and system development process variables in determining the success of CBMIS.
A Study of Group Interaction over a Computer-Based Message System BIBA 235-248
  Sylvia Wilbur; Tony Rubin; Stephen Lee
This paper is based on a study of the use of a computer-based message system for collaboration among a team of researchers based at five different sites. A review is presented of similar work on messaging for group communication, and attention is drawn to the comparative paucity of empirical analysis of the strengths and weaknesses of computer-based systems in this respect.
   An analysis is made of messages sent over the period of a year, during which the team developed a joint proposal for a programme of research and successfully obtained approval for funding. Two methods have been used to carry out the analysis. First, a statistical analysis using the SPSS package was conducted, and these results are shown in figures within the paper. This was followed by an analysis of the contents of messages, and this was carried out manually, in the absence of a suitable application package. Since there were over 1400 messages to be examined, contents were categorised on the basis of a restricted set of criteria, chosen to reflect the team's objectives and with reference to other work in the area.
   The analysis allows us to make observations about the kinds of interaction that took place, and the extent to which the system was able adequately to support the communication requirements of the group. We believe that further research of this kind will result in the introduction of new structures and design features in advanced message systems of the future.
Usability Engineering in Office Product Development BIBA 249-259
  J. B. Brooke
Applying human factors to the development of computer systems is often left to be something of an afterthought. The application of usability engineering techniques can provide a number of benefits in ensuring the improvement of the human factors of a software product. Firstly, usability is modified from a vague wish to a clear statement of measurable goals in product requirements. This helps both the human factors engineer and the product development team. Secondly, the human factors engineer and the product developers must work closely together to develop the goals, which ensures that as many constraining factors on usability as possible are taken into account -- for example, the feasibility of building certain sort of interfaces, the target user population, requirements for backwards compatibility with other products are all things which might affect what is judged to be "good" human factors in absolute terms.
   The statement of clear measurable usability goals also implies that the product must itself be subjected to usability testing. This procedure allows the human factors engineer to identify those flaws in the software system which cause the greatest usability problems. Experience in testing a variety of office systems has shown that the problems are usually "trivial" flaws in design or implementation. Nonetheless, these tend to compound with each other and interact to cause major usability problems. Basic conceptual problems seem to be rare.

Intelligent Interfaces

Identifying the Knowledge Requirements of an Expert System's Natural Language Processing Interface BIBA 263-280
  Dan Diaper
It is now a common belief that expert systems will require an intelligent interface to facilitate the dialogue between such systems and their users. The desirable properties of intelligent interfaces are currently unknown. A methodology has been developed that simulates a future expert system that possesses a powerful intelligent interface. Dialogues have been collected between this simulated system and users and are analysed in terms of the knowledge that the interface requires to support the dialogues. A procedural specification of these knowledge requirements is offered.
Design and Evaluation of the AID Adaptive Front-End to Telecom Gold BIBA 281-295
  Peter Totterdell; Paul Cooper
This paper describes the design and initial evaluation of an adaptive front-end to the Telecom Gold electronic mail system. The system, developed as part of the Adaptive Intelligent Dialogues (AID) project, adapts along a number of select adaption dimensions: level of guidance, context switching, recognition of analogous mail system and user tailoring. A generalised architecture based on a dialogue controller, user model and application expert to support these dimensions is described. In the evaluation, members of the public used the system while playing the role of an office manager. This required them to perform a variety of tasks representative of a typical electronic mail system. Measures were taken of user satisfaction, efficiency, and effectiveness. In addition, an independent expert assessed the value and consequences of the system's strategy for adaption from observation of the system, and study of both concurrent and retrospective user protocols. The paper draws conclusions on the strengths and weaknesses of the exemplar, and the implications for the subsequent design of adaptive systems.
Plan Recognition for Intelligent Advice and Monitoring BIBA 296-315
  Colin Davenport; George Weir
In contrast to the conventional interface approach to command languages, we offer a plan-based system (PRIAM) which acts on a structured representation of the language, to provide intelligent monitoring and advice to users. We detail an experiment to assess the relative performance of a PRIAM-based interface and a more conventional interface design. Our results suggest that the PRIAM alternative offers considerable learning advantages for new users. The operations of the full PRIAM system are described and its potential benefits noted.

Structuring Interaction

Application Modelling in a User Interface Management System BIBA 319-335
  J. L. Alty; P. McKell
The role of the Application Model (or Application Expert) in a User Interface Management System is discussed. It is postulated that it should represent an expert's view of the interface and that all other user related information should be housed in a separate User Model. By this approach the Application Model acts as an anchor point in the design of an Interface System. The Model is considered to have two main functions, as a custodian to the application and as a source of guidance and advice to the User Model. An outline architecture for an Application Model for command-driven systems is developed using UNIX as an example application. The output of user error log files is used to identify appropriate modules in an Application Model which, because of the general nature of UNIX ought to be applicable to other command-driven situations. A control strategy is developed for a blackboard-like approach and some examples or error detection by the Application Model are given. The work will now be applied to a more specific application used by Petroleum Engineers.
The Design of Two Innovative User Interfaces BIBA 336-351
  Harold Thimbleby
Two innovative user interfaces are described: one for an arithmetic calculator and one for a spreadsheet. The emphasis of the paper is on the designs themselves and on the underlying rationale. The interfaces were developed methodically, using a heuristic of property closure. User interface issues which arise are discussed and include: equal opportunity, declarative conceptual models, non-determinism, and implied task domain.
Principles and Interaction Models for Window Managers BIBA 352-366
  A. J. Dix; M. D. Harrison
Formal methods have been used to develop a prototype interactive editing system, in which different edits are viewed through separate windows. Designing the prototype has involved the development of a simple window management system. The design of the window manager was achieved with the assistance of an initial description using an abstract model of interaction. We argue that abstract interaction models clarify certain design issues. We discuss more complex properties of windowing systems including separability, sharing and interference. We formulate some simple generative user-engineering principles to support these properties.

Formal Design Methods

Modelling Generic User-Interfaces with Functional Programs BIBA 369-385
  Steve Cook
This paper outlines the development of a model of a simple interactive mouse-driven user-interface in a functional programming style. The intention is to demonstrate the possibility of using such a style of programming to describe formally components of such user-interfaces. It is also shown how generic components may be used to develop families of interactive applications with common user-interface characteristics. To describe such generic components, the type structure of the language used must admit polymorphism and in particular the notion of subtyping.
Text Representation and Manipulation in a Mouse-Driven Interface BIBA 386-401
  Roger Took
In many mouse-driven interface packages, the direct manipulation mode does not extend to fine-grained objects such as individual characters in text. This paper examines the benefits and problems of doing this, in the light of experiences with the Presenter, the interface manager to the Alvey ASPECT IPSE. The text-handling capabilities of the Presenter are described in the formal language Z.
Proving Properties of Interactive Systems BIBA 402-416
  Stuart Anderson
The problem of stating and proving properties of interactive systems is dealt with within a fairly concrete notion of interactive system specification. A small specification is presented and outline proofs of some simple properties are given. Though the proofs are simple they do illustrate some of the proof techniques available within the framework. We argue that within the framework presented here properties can be succinctly stated and easily proved and that it is possible to modularise proofs in such a way that the same techniques can be used on larger problems. In addition the possibility of synthesising systems from such properties is considered.
Where Do We Draw the Line? -- Derivation and Evaluation of User Interface Software Separation Rules BIBA 417-431
  Gilbert Cockton
The potential benefits of separating the user interface from the rest of the application are well known. Dialogue design tools, especially omnicompetent User Interface Management Systems (UIMS), are not viable if separation is impossible. Current characterisations of the separated agents are vague veneers. Some UIMS designers offer no definitions. Sound criticisms of the nature, practicality and possibility of user interface separation are commonplace. A Patterned Transition System (PTS) formalism is presented. It allows rigorous definition of separation as isolation. A decision procedure exists which states whether two subsystems are isolated. This proves separation is possible. However isolated subsystems cannot communicate. Communication must be restored by a further separate unisolated subsystem called a linkage. Properties required for a subsystem to be a linkage for two isolated subsystems are presented. The costs of linking isolated systems is measurable. A linkage also embodies attributes of the relationship between two linked subsystems, ending dilemmas on application dependent dialogue functions.

Knowledge of the User as a Design Guide

A Viewdata-Structure Editor Designed Around a Task/Action Mapping BIBA 435-446
  Richard M. Young; John E. Harris
This paper reports an exploratory attempt to design an editor for viewdata networks based explicitly around a task/action mapping. Task analysis reveals that one editing task can imply the need for another, leading to a view of the overall task as a set of frame-edits joined by a web of implications. A viewdata editor to capitalise on this task structure supplements the frame editor itself with a helpful "assistant" which keeps track of the implications, and at appropriate times reminds the user of the tasks remaining to be done and suggests one to work on next. The order in which tasks are offered to the user is crucial to the success of the system, and is determined by the interaction between four heuristic ordering principles.
The Use of Complexity Theory in Evaluating Interfaces BIBA 447-463
  George Kiss; Roy Pinder
The paper argues the case for using the techniques of computational complexity theory in the evaluation of user interfaces. Taking a task-oriented attitude, we interpret user effort as the computational work done by the user in terms of interface operations in carrying out tasks. Computational work is in turn identified with the size complexity of the algorithm through which the task is accomplished at the interface. We show how design decisions about primitives and tools determine the complexity of algorithms available to the user to carry out tasks. An illustration of the approach is given for direct manipulation graphics interfaces. The discussion indicates how the approach gives a reasonable interpretation to concepts like ease of use and tradeoff decisions.
User Programs: A Way to Match Computer Systems and Human Cognition BIBA 464-481
  Colin Runciman; Nick Hammond
To avoid bias towards machine issues in interactive system design it is proposed that processing to be performed by the human user should be expressed in user programs. The intention is to bring design closer to a human-centred approach without all the costs and uncertainties often associated with experimental prototyping. Topics discussed include the kind of cognitive architecture on which such programs might run, the kind of programming language in which they might be expressed, and the ways in which they might be developed and interpreted. Some specific proposals and examples are given, but these are intended as illustrative rather than definitive.
Using an Expert System to Convey HCI Information BIBA 482-497
  Michael Wilson; Philip Barnard; Allan MacLean
Where the focus is upon human cognition, guidelines and technical reports are an inadequate means of conveying information from the research to the design communities concerned with HCI. Automated databases or simple expert systems assist in accessing relevant information. They do not, however, readily predict behaviour in novel settings. This possibility is offered by expert systems that incorporate a cognitive analysis of user knowledge and human information processing activity. The present paper outlines an approach to Cognitive Task Analysis (CTA) in which a theoretical framework (Barnard, 1985; in press) is used to derive an explicit representation of cognitive activity associated with dialogue tasks. The representation constructed (or Task Model) includes a specification of mental processes; procedural knowledge; the contents of episodic memory; and a characterisation of the way in which the cognitive mechanism is controlled during task execution. Prespecified mappings from the contents of Task Models then predict aspects of user behaviour. Components of an example analysis, implemented in a working expert system, are used to illustrate the approach.

Case Studies: Evaluation

New Technology Work Aids for the Physically Disabled BIBA 501-526
  Kate Howey
A research study into the effects of new technology on job opportunities for the physically disabled was carried out on behalf of the European Commission during 1981-1982 at the Department of Human Sciences, Loughborough University of Technology.
   The aims of the research were, firstly, to investigate the effects of the increasing use of technology in many industries on changes in the nature of work and on work methods, especially the implications of these changes for the job opportunities of physically disabled people. Secondly, it was intended to report on the development of computer-based aids for the disabled workforce within the EEC.
   This paper discusses both issues, but because of the wide scope of the original study, it reports primarily on the range of available new technology devices designed to help the differing communication problems of physically disabled people at work. Within this context, requirements for the design of future systems for the disabled are highlighted.
Structural Visibility and Program Comprehension BIBA 527-545
  David J. Gilmore
This paper draws on work by Anderson and Jeffries (1985) which examined the cause of novice LISP errors, interpreting them in terms of a processing overload model. Two experiments are reported which ask about factors which influence processing demand. Factors examined include the visibility of program structure, the length of program transactions and the programming task.
   The results of these experiments suggests that effects due to structural visibility may be explicable in relation to processing overload, but the effect of transaction length on performance cannot be so explained. To explain the results it is necessary to realise that processing load is not the only thing affected by language features. For example, some language features (e.g. transaction length) change strategy not processing demand.
   This paper closes with a consideration of utility of processing demand models, and concludes that it is more profitable to consider the relationship between specific aspects of notations and particular psychological processes.
Voice versus Keyboard: Use of a Comparative Analysis of Learning to Identify Skill Requirements of Input Devices BIBA 546-562
  Peter Johnson; John Long; David Visick
This paper is concerned with the evaluation of alternative forms of input device, specifically voice recognition and keyboards. Four devices were tested experimentally in a data entry task. The aim of the assessment was to provide, on the basis of a comparative analysis of learning, information concerning the skill requirements of operators, which would allow the selection of a suitable device for parcel sorting. Learning was analysed in terms of two performance indices: percentage errors and time to complete the task or task component. An analysis of learning on each device was carried out. The analysis was used to identify the skill requirements of operation, and to make recommendations concerning the application of the devices.

Display Based Systems: Evaluation

Empirical Evaluation of Map Interfaces: A Preliminary Study BIBA 565-585
  Graham J. Hitch; Alistair G. Sutcliffe; John M. Bowers; Lucy M. Eccles
The use of spatial maps as human computer interfaces has been described by Sutcliffe (1985) who demonstrated, using the keystroke model of human performance (Card, Moran and Newell, 1980), that maps should have an operational advantage over menus. This paper reports an empirical evaluation of the use of menu and map interfaces to retrieve information from a hierarchically organized geographical database. Two task variables were manipulated; amount of practice and the provision of retrieval cues in the search query. The importance of the compatibility between the physical layout of the map and the organization of the user's knowledge was investigated by comparing performance with a "geographical" and a "random" map Search was slower and more error prone with a menu interface at all stages of practice, the advantage to map interfaces being greatest when there were no retrieval cues to assist search. The geographical map was no more efficient than the random one when users were totally naive; however, it became the more efficient of the two as a result of practice. These findings are discussed in relation to limitations on the practical utility of the keystroke model in predictive evaluation, and in terms of the ease and flexibility of user search processes guided by map interfaces.
Evaluating the Meaningfulness of Icon Sets to Represent Command Operations BIBA 586-603
  Yvonne Rogers
Iconic interfacing is becoming increasingly popular as a medium to present information about computer systems and their command operations. This paper considers the extent to which various icons that differ in the form of correspondence between referent and icon symbol can effectively represent a large number of abstract command operations typically used in a word processing environment.
   Six icon sets, depicting either abstract symbols, concrete objects operated on, concrete analogies associated with the action or combinations of these were constructed to represent 20 commands covering a range of word processing operation areas. Using a questionnaire, 60 subjects (10 for each set) were required to match the icons to the commands they thought they referred to. Significant differences were found between the icon sets. Specifically, the icon sets with the most direct mapping (i.e. those depicting concrete objects operated on) were found to have the highest number of correct matches, with over 85% of the icons being correctly identified. An interaction between icon set and type of command was also found indicating that some commands can be represented in a range of pictorial forms while for other commands the type of pictorial form is critical. The results from this experiment are discussed in relation to the demands made on the cognitive resources for the comprehension of visual symbols.
Optimizing the Usability of Computer-Generated Displays BIBA 604-613
  Thomas S. Tullis
Previous research indicated that the two best predictors of the time that it takes users to extract information from an alphanumeric display are the number of visual groups of characters on the display and the average visual angle subtended by those groups. As either of these values increases, search time increases. However, the number of groups of characters and their average visual angle are not independent of each other. Using mathematical modelling techniques, an exponential function was derived to describe the relationship between these two display measures. Combining that equation with a regression equation fitting user search time with the two display measures resulted in a U-shaped function relating search time to number of visual groups. The shortest search times were associated with a range of about 19 to 40 groups, which corresponds to an average visual angle of about 4.9 to 2.4 degrees. The results are interpreted as indicating that groups smaller than about 5 degrees allow for a more efficient pattern of visual search, in which the necessary information can be extracted from each group with only one fixation.