Geert.de.Haan (AT) upcmail.nl -
Home Page -
Curriculum Vitae -
PhD Thesis -
The Green Estate
Everything you always wanted to know about Extended Task-Action Grammar but were
afraid to ask.
What is ETAG?
Extended Task-Action Grammar (ETAG) is a formalism proposed by
Michael Tauber (and developed by the uppersigned) to represent user-
interface designs. More specific, ETAG represent the knowledge which a
perfectly knowing user has about a particular user interface. It does not
specify how users represent knowledge; merely what they should know about
both the interaction language (the knowledge "how-to") and about the
interactive device (the knowledge "how it works").
ETAG is a formal model, and can thus be used for user interface analysis,
even before a single line of code has been written. As a formal
representation of a user interface; including the knowledge a user needs
about the application, it can be used as a source for automatic user
documentation and on-line help generation. In the same way, it has been used
as a (though rudimentary) system prototyping tool. Finally, since ETAG
describes user knowledge about user interfaces, it is an especially suitable
basis for a user interface design method: sufficiently formal to bridge the
'gap' with software engineering, while sufficiently exact to serve as a
communication means, and sufficiently user-oriented to ensure usable
computer systems. What is not sufficient -as yet- is the usability of the
ETAG notation for designers; something that will be addressed through the
use of design tools that hide the peculiarities of the notation.
What kind of Formal Model is ETAG?
For those familiar with formal modelling in HCI: ETAG is Task-Action
Grammar (TAG, Green and Payne) with a device model build on top of it. It
is like Command Language Grammar (CLG, Moran) but without the ad-hoccery
and its redundancy. In comparison with Cognitive Complexity
Theory (CCT, Kieras and Polson) and GOMS (Goals, etc.; Card, Moran and
Newell) it is a pure competence model instead of a performance model.
In comparison to the work of the York group, Petri-nets and the like, ETAG
is not concerned with software engineering characteristics of a user interface,
but with user-knowledge, building upon an ontology of the 'virtual
machinery' of the interface (Jackendoff and Sowa) and employing conceptual
In terms of design methods based on formal representations, ETAG-based
design comes closest to Muse*/JSD (Long and Lim) and Adept (Johnson and
Wilson). Through the use of JSD, Muse*/JSD may be easier to connect to
software Engineering design, and Adept seems more advanced in terms of
the tools it provides, but ETAG-based design is based on a superior
knowledge representation scheme.
What does an ETAG representation look like?
This is an extremely tiny example; only to get the 'taste' of ETAG's formalism.
It shows the type definition of a Clipboard and how it is used in a Cut_String
event as a temporal storage space. The Cut_String event is invoked by a user
task with (accidentally) the same name.
type [OBJECT = CLIPBOARD]
supertype: [SINGLE_OBJECT_BOARD] ;
themes: [STRING: *s] | [WORD: *w] | [PICTURE: *p] | ... ;
instances: [CLIPBOARD: #clipboard] ;
type [EVENT = CUT_STRING]
description: event.MOVE_TO [STRING: *s, OBJECT: #clipboard] ;
precondition: state.IS_AT [STRING: *s, OBJECT: #document] ;
state.HAS_VAL [STRING: *s, ATTRIBUTE: selected] ;
clears: state.IS_AT [STRING: *s, OBJECT: #document] ;
postcondition: state.IS_AT [STRING: *s, OBJECT: #clipboard] ;
state.HAS_VAL [STRING: *s, ATTRIBUTE: purgable] ;
comments: "move a string from a document on to the clipboard"
system: PLACE = clipboard, PLACE = document
T1 [EVENT = CUT_STRING, OBJECT = STRING: *s, ATTRIBUTE = selected]
comment: "move a selected string from a document to the clip"
What's written about ETAG?
The publications listed below are available as html files.
The papers are listed in more-or-less chronological order and each deals with a
particular aspect of ETAG. The last entry, PhD thesis, gives the best overview of
ETAG, in terms of background and motivation, content and application, the design
method and conclusions about ETAG.
This paper discusses the reasons to use formal models of user knowledge in
human-computer interaction. Four classes of formal models are identified:
models for task environment analysis, models to analyze user knowledge,
user performance prediction models, and representation models for design
The main problems mentioned in the literature as regards the use
of these models will be discussed. Finally, the contribution of our research
group to the NFI-project "Systematic Design of User Interfaces" is presented,
which is directed at the Extended Task-Action Grammar (ETAG, Tauber,
1988, 1990). This includes applying ETAG to the presentation component
and the semantics of the user interface, providing tools to increase ETAG's
usability for designers, and performing validation studies.
Appeared in: van der Veer, G.C., White, T.N. and Arnold, A.G. (1993,
eds.) Human-Computer Interaction: Proc. Preparing for the Nineties, 95-112.
Stichting Informatica Congressen, Amsterdam. ISBN 90-5005-039-5.
This paper is a theoretical contribution, elaborating the concept of models as
used in Cognitive Ergonomics. A number of formal modelling techniques in
human-computer interaction will be reviewed and discussed. The analysis
focuses on different related concepts of formal modelling techniques in
human-computer interaction. The label "model" is used in various ways to
represent the knowledge users needs to operate interactive computer systems,
to represent user relevant aspects in the design of interactive systems, and to
refer to methods that generate evaluative and predictive statements about
usability aspects of such systems.
The reasons underlying the use of formal models is discussed and a review
is presented of the most important
modelling approaches, which include External- Internal Task Mapping
Analysis; Action Language; Task-Action Grammar; the Goals, Operators,
Methods and Selection model; Command Language Grammar and Extended
Task-Action Grammar. The problems associated with applying the present
formal modelling techniques are reviewed, and possibilities to solve these
problems are presented. Finally, we conclude with a discussion of the future
work that needs to be done, i.e., the development of a general design
approach for usable systems, and the need to focus attention on the practice
of applying formal modelling techniques in design.
Appeared in: Acta Psychologica, 78, nos. 1-3, 26-76. Also appeared in: Van
der Veer, G.C., Bagnara, S. and Kempen, G.A.M. (1992), (eds.), Cognitive
Ergonomics - contributions from experimental psychology. North-Holland,
Amsterdam, pp. 27-68.
In this paper we discuss the psychological basis of Extended Task-Action
Grammar (ETAG). ETAG is both a method to represent the knowledge a
competent user of a user interface has about the interface and a method to
represent the user interface for the purpose of design. The psychological
consideration underlying the format of the model are: the use of a feature
grammar to represent the interaction language, and the distinction of the
description into four levels of abstraction, similar to those in (natural)
The psychological choices underlying the content of the
model consider the choice of the canonical basis. This is the most abstract
level of representation in ETAG. The concepts defined at this level are most
general and necessary to acquire any knowledge of the world, and which
serve as the source from which all other concepts in ETAG are derived. The
concepts in the canonical basis are derived from psychological research into
semantic memory by Klix and colleagues, and from psycho-linguistic
analyses into the semantics which underlie language by Sowa and Jackendoff.
The paper concludes with an example of an ETAG representation and a brief
comparison between ETAG and other design models in Human-Computer
Unpublished manuascript (1995).
In this paper Extended Task Action Grammar (ETAG; Tauber, 1988, 1990)
is applied, as a formal method to represent the knowledge a competent user
has about the structure of a user interface. This paper consists of three parts.
First, ETAG itself will be discussed, including a brief treatment of the
considerations underlying its development and the purposes of ETAG, an
explanation of what an ETAG description consists of, and some remarks
about the similarities and differences between ETAG and other formal
Secondly, the models are represented which resulted from applying
ETAG to the design problems formulated in the Macinter modelling contest
(MacInter 1990), supplemented with remarks about specific problems and
choices taken. This part will be directed at both evaluating the design
alternatives of the contest and a preliminary evaluation of ETAG as a design
Finally, using the results of the previous section, the general
problems, possibilities and shortcomings that were encountered in applying
ETAG will be discussed in an attempt to formulate requirements for the
future development of the method.
Appeared in: Zeitschrift für Psychology, 200, 1992, 135-156.
In this paper we will discuss the use of Extended Task-Action Grammar
(ETAG) as a tool to analyze user interfaces. An ETAG representation is a
formal model of the knowledge of a competent user about how the computer
system works, and about how to perform tasks.
The structure of ETAG representations is briefly explained, and methods of
applying ETAG analyses are discussed. Several studies are reviewed in
which ETAG is applied to different types of computer systems, and to
variants of the same system, in order to reach conclusions about the validity
and the general applicability of ETAG, and to determine which aspects of the
method may be improved, and how this may be done.
Appeared in: Proc. of the 13th interdisciplinary workshop on Informatics and
Psychology - Task-Analysis in Human-Computer Interaction, Schärding,
Austria, 9-11 June, 1992.
ETAG is a method for representing interactive computer systems for design
purposes, on the basis of a representation of what a competent (perfectly
knowing) user knows about the structure and use of a system.
An ETAG representation is a conceptual model which contains all
information a user might want to have of a computer system. As such,
ETAG representations may serve as the basis for intelligent help facilities.
A number of studies will be reviewed in which ETAG was used to build non-
interactive and interactive facilities providing help information about
In order to provide static -non changing- information about the computer
system, the ETAG representations contain sufficient information.
To provide dynamic information about interacting with the system as well,
the information in ETAG representations needs to be supplemented with
In order to cope with dynamically changing information, it proved to be
necessary and more widely advantageous to use an intermediate (Prolog)
representation of ETAG and history information. In different respects,
following a modular system architecture proved to be advantageous.
Appeared in: van der Veer, G.C., Tauber, M.J. and Bagnara, S. (eds.).
Proc. ECCE-6. Human-Computer Interaction: Tasks and Organisation, 271-
283. CUD, Rome, Italy, 1992. ISBN 88-7721-232-2.
This paper describes ETAG-based design, as an approach to design user
interface on the basis of a formal model of the knowledge, a competent user
needs to use the interface to perform tasks with it. The approach is based on
the (obvious) idea that user interfaces are for people to use to perform tasks.
ETAG-based design is a user-centered, task-centered and (conceptual)
knowledge-based method to design user interfaces. In design, these principles
are embodied within the standard model of design (the received view), by
using the ETAG formalism to specify a conceptual model of the user's task
world: the Virtual Machine.
First, the received view of user interface design is discussed, and it is
concluded that this view lacks guarantees that user interfaces are designed for
people to use.
Second, ETAG is briefly presented. ETAG is a formal model to specify the
user interface from the point of view of a user's competence knowledge.
Thirdly, we discuss ETAG-based design and show how it is used within the
received view of user interface design, to kindly persuade designers to take a
user (task) oriented stance.
Finally, we draw a few conclusions regarding the near- and far future of
ETAG-based design, or more in general, user-model-based design.
Appeared in: Proceedings of the 15th interdisciplinary workshop on
"Informatics and Psychology" - Interdisciplinary approaches to system
analysis and design, Schärding, Austria, 24-26 May, 1994.
Note: the material listed above was presented in a handout entitled: "All you
ever wanted to know about Extended Task-Action Grammar but were afraid to ask"
at the Seventh European Conference on Cognitive Ergonomics, ECCE-7, September 5-8,
1994, Bonn, Germany. The material listed below was added to this document later on.
Extended Task-Action Grammar (ETAG) is a modelling notation to formally
describe user interfaces in terms of the knowledge, a competent user would
have in order to use the interface for task performance. After introducing the
most important ideas underlying ETAG and, briefly, its place in the historical
context of formal modelling in HCI, the structure, contents and how to create
ETAG models are extensively discussed. Hereafter, we briefly mention some
of the uses of the notation, and the structure of ETAG-based user interface
design. After discussing the status of ETAG-based design in software
engineering, several conclusions are drawn.
Appeared in: Cognitive Systems 4, 3-4, 353-379, 1997.
Extended Task-Action Grammar (ETAG) is a formal language to represent
user interfaces in terms of the knowledge (the mental model) that a perfectly
knowing user would have about performing tasks. In ETAG-based design,
user interface design is regarded as the incremental specification of the
mental model of a perfectly knowing user.
The design process is structured
after the ETAG model into discrete steps, each covering a specific set of
design decisions. ETAG-based design is an example of model-based design,
and it shows what an HCI specific design, aiming at usable computer systems
should look like.
Appeared in: Palanque, P. and Benyon, D. (1996)(eds.) Critical Issues in
User-Interface System Engineering, Springer Verlag.
This paper discusses ETAG, a formal model for design representation, and
ETAG-based design, a method for user interface design, and the two principal
facilities they provide to accommodate different types of users in the design
of user interfaces. The paper starts with a description of the purpose and the
structure of ETAG. This is followed by a description of ETAG-based design and
using the notation to represent relevant aspects of the work context.
Thereafter, in relation to differences and similarities in required task
knowledge, the two main facilities in ETAG-based design for addressing different
types of users are discussed and exemplified. Finally, using evidence from
several design projects, conclusions are drawn about addressing differences
among users in ETAG and ETAG-based design.
Appeared in: Proceedings of HCI International 2001, 5-10 August 2001, New Orleans, USA.
Formal models for user-interface design are important because formality provides a number of advantages, such as
early use in the design process, precise prediction of usability aspects, precision for design and communication, and
enabling automatic generation of user interfaces.
Within the domain of formal models, the thesis attempts to answer two main questions:
(1) what is a good formal specification model for user-interface design?
(2) what is a good method for user-interface design, based on such a model?
To answer the first question, a number of criteria are developed for evaluating the available formal models such as
ETIT, TAG, CLG, GOMS, and ETAG. Mainly on the basis of the completeness of the conceptual specification and
the psychological validity, ETAG (Extended Task-Action Grammar) is selected for further development. Specifically
aimed at validating the notation, the thesis discusses the psychological basis and it provides a manual for creating the
The second question is answered from a radical user-centered perspective: the user interface is defined as everything a
user must know in order to perform tasks, software design is treated as subordinate to user interface design.
ETAG-based User Interface Design follows the common scheme for user interface design, except that ETAG is used
as the main representation method, and iteration between design stages is avoided as much as possible. Aimed at
validating both the notation and the design method, the thesis discusses task-analysis, user-interface analysis, and the
generation of help information and user interface prototypes.
The main conclusions from the study are as follows. ETAG and ETAG-based User Interface Design perform very well.
There are two main areas for further development. To address the relevance of perceptual information to task
performance, methods and tools are required to describe and analyse the perceptual interface. To address the
relevance of the dynamical aspects of the interaction to task performance, such as learning, it is necessary to bridge the
gap between static knowledge models, like ETAG, and dynamic knowledge models, like SOAR.
Appeared as PhD thesis, Department of Mathematics and Computer Science, Free University
Amsterdam, 10 October 2000.
Geert.de.Haan (AT) upcmail.nl -
Home Page -
Curriculum Vitae -
PhD Thesis -
The Green Estate