ARTICLE IN PRESS

Applied Ergonomics 37 (2006) 55–79

www.elsevier.com/locate/apergo

Hierarchical task analysis: Developments, applications,

and extensions
Neville A. Stanton
BITlab, Human Factors Integration Defence Technology Centre, School of Engineering and Design, Brunel University, Uxbridge,
Middlesex UB8 3PH, UK
Received 21 April 2004; accepted 6 June 2005

Abstract

Hierarchical task analysis (HTA) is a core ergonomics approach with a pedigree of over 30 years continuous use. At its heart,
HTA is based upon a theory of performance and has only three governing principles. Originally developed as a means of
determining training requirements, there was no way the initial pioneers of HTA could have foreseen the extent of its success. HTA
has endured as a way of representing a system sub-goal hierarchy for extended analysis. It has been used for a range of applications,
including interface design and evaluation, allocation of function, job aid design, error prediction, and workload assessment.
Ergonomists are still developing new ways of using HTA which has assured the continued use of the approach for the foreseeable
future.
r 2005 Elsevier Ltd. All rights reserved.

1. Origins of task analysis and Lillian Gilbreth observed that bricklayers tended to
use different methods of working. With the aim of
According to Kirwan and Ainsworth (1992), hier- seeking the best way to perform the task, they developed
archical task analysis (HTA) is the ‘‘best known task innovative tools, job aids, and work procedures. These
analysis technique’’ (p. 396). It is probably a special case innovations included: scaffolding that permitted quick
in the ergonomics repertoire of methods. Since the first adjustment, shelves for bricks and mortar, and methods
paper written on the specification for the method in 1967 for getting the bricks and mortar to the bricklayers by
by Annett and Duncan, the past 38 years have seen lower paid labourers. The net effect of these changes to
many developments in ergonomics research and meth- the work meant that the laying of a brick had been
ods but HTA has remained a central approach. It is reduced dramatically from approximately 18 move-
fitting to review the current state of the art to help take ments by the bricklayer down to some four movements.
stock of where HTA has come from, the contemporary The task was therefore performed much more efficiently.
issues, and the potential for the future. The principle underlying this work was to break down
The origins of all modern task analysis techniques can and study the individual elements of a task. The
be traced back to the scientific management movement individual elements (called Therbligs—a reversal of
in the early 1900s (Annett and Stanton, 1998, 2000). The Gilbreth) such as ‘grasp’ and ‘assemble’ were recorded
three figures that stand out from this time are Frank and against time, hence the phase ‘time-and-motion’ study
Lillian Gilbreth and Frederick Taylor. The Gilbreths (Gilbreth, 1911). Annett (2000) notes that whilst most of
sought to discover more efficient ways to perform tasks. the therbligs refer to physical movement, there were
By the way of a famous example of their work, Frank some ‘cognitive’ therbligs, such as ‘search’, ‘select’, and
‘find’. The scientific management community, with
Tel.: +44 1895 265543; fax: +44 1895 232806. which Frederick Taylor’s name is inextricably linked,
E-mail address: [email protected]. sought to apply the rigour of scientific method in the

0003-6870/$ - see front matter r 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apergo.2005.06.003
 ARTICLE IN PRESS
56 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

analysis of work. At the heart of this approach was understanding of cognitive tasks. With greater degrees
serious analytical critique of the details of methods for of automation in industrial work practices, the nature of
working: How was the work performed? What was worker tasks was changing in the 1960s. Annett argued
needed to perform the work? Why was the work that as these tasks involved significant cognitive
performed in this way? How could the working methods components (such as monitoring, anticipating, predict-
be improved? Modern task analysis methods have ing, and decision-making), a method of analysing and
retained this general approach to task critique. Annett representing this form of work was required. Existing
(1996) has certainly argued that HTA encourages the approaches tended to focus on observable aspects of
analyst to consider not only what should happen, but performance, whereas HTA sought to represent system
also what does actually happen and how this can go goals and plans. At the time of the late 1960s this was a
wrong. He suggests that these questions will arise radical departure from contemporary approaches. The
naturally as the analyst seeks to discover the indicators ‘cognitive’ revolution had yet to happen in mainstream
for success and failure of each of the sub-goals. psychology and the ‘behaviouristic’ paradigm was
The scientific management approach has been criti- dominant. At that time it was considered ‘unscientific’
cised for failing to consider the psychological make-up to infer cognitive processes, and academic psychology
of work (e.g., Hackman and Oldham, 1980). Accounts focused principally on observable behaviour. HTA,
of efficiency drives and job simplification may lead one however, offered a means of describing a system in
to suppose that it fails to take the effects on an terms of goals and sub-goals, with feedback loops in a
individual person into account. Certainly, Taylor’s nested hierarchy.
(1911) (in)famous book on ‘The Principles of Scientific The influence of the control theory of human
Management’ does little to dispel this idea, which behaviour as proposed by Miller et al. (1960) can clearly
contains capitalistic political overtones and provides be seen in HTA. Central to this theory are the twin ideas
accounts on the laziness of the working classes. The of a test–operate–test–exit (TOTE) unit and hierarchical
Gilbreth’s work, however, seemed to be focused on the levels of analysis. The classic example of a TOTE unit is
well-being of the person as well as the effectiveness of the explanation of hammering a nail flush with a piece of
the work. This may well have been influenced by Lillian wood. This is illustrated in Fig. 1.
Gilbreth’s profession as a psychologist. This latter The three units of analysis are a TEST (where the goal
approach is much closer to the heart of modern is to see if the nail is flush with the surface of the wood),
ergonomics. In the century that has passed since these if the nail is not flush then an OPERATION is
original pioneers of task analysis, several important performed (i.e., striking the nail with the hammer), then
changes have taken place. Annett (2000) cites several the TEST is performed again. If the nail is flush, then the
influences that have contributed to early thinking in operator can EXIT this activity. We can imagine a
HTA. In the 1950s, ergonomics was emerging as a situation where the nail is already flush, so the analysis
distinct discipline, but drawing on contemporary trends would comprise just the TEST and EXIT components,
in psychology and engineering. The 1950s gave rise to or other situations where multiple TESTS and OPERA-
new theories of human performance in systems and new TIONS are performed prior to the EXIT. In TOTE
ways of assessing human activities in system design. terms, these would be TE and TOTOTOTOTE,
Whilst it is difficult to pinpoint all of the possible factors respectively. The important aspects of the TOTE
that could have led to the development of HTA, some of analysis are that it implies some level of information
the main influences are likely to include: the break down feedback, a system of control, and it offers hierarchical
of tasks into their elements, the questioning of human analysis of systems. Miller et al. (1960) illustrated
performance in systems, a need to understand both the hierarchical analysis, by showing how the operation
physical and cognitive activity, a desire to represent the
analysis in a graphical manner, and a need for an
underpinning theory of human behaviour. One of the
most influential ideas for HTA was the identification of nail flush
Test nail Exit
error variance in system performance from systems
theory (Chapanis, 1951). Annett (2004) states that the
top-down systems approach taken by HTA enables the
analyst ‘‘to identify and deal first with factors generating nail sticks up
the largest error variance’’ (pp. 68–69). The error
variance could be generated by either humans or
machines, or an interaction between human and Hammer nail
machines. operate
Annett (2004) points out that the initial development
effort in HTA was in response to the need for greater Fig. 1. A TOTE unit for making a nail flush with the surface.
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 57

0.
nail sticks up Test nail nail flush
Exit Make nail flush Plan 0.
Is yes
nail 1 nail exit
sticks flush?
up Exit
no
2
HAMMER hammer
NAIL down

Exit
Test hammer Test hammer
hammer 1. 2.
up Test nail Hammer nail
hammer hammer
down Plan 2.
up
Is yes
2.1 hammer 2.3
up?
Operate Operate
Lift hammer Strike nail no exit
2.2

Fig. 2. The hierarchical plan with ‘hammer nail’ re-described.

2.1 2.2 2.3

Test hammer Lift hammer Strike nail
in Fig. 1 could be further investigated, as illustrated
in Fig. 2.
Miller et al. (1960) point out that the hammering of a Fig. 3. HTA for goal of ‘Make nail flush’.
nail only serves as an example and one might not
attempt to analyse all tasks down to this level of detail.
The analysis does show how it is possible to develop a structures that govern the sequence of the sub-goals.
more detailed system view of the control structures These are precisely the same control structures as those
within a hierarchical analysis. Any system could, in the TOTE analysis. The two forms of analysis are
potentially, comprise of hierarchically arranged TOTE highly compatible.
units. As Miller et al. put it: In one of the earliest papers leading up to the
specification for HTA, Annett and Duncan (1967) show
Thus the compound of TOTE units unravels itself
their concern with the adequacy of the description. The
simply enough into a co-ordinated sequence of tests
idea of a hierarchical description comprising subordi-
and actions, although the underlying structure that
nate operations is proposed with rules governing the
organises and co-ordinates the behaviour is itself
level that this is taken to. They argued that some aspects
hierarchical, not sequential. (Miller et al., 1960, p. 34)
of a task might be taken down several levels of re-
The example in Fig. 2 shows how the operation of a description, whereas others will not. The decision to re-
hammer comprises a test of its position and then the describe the task will depend upon estimates that the
operation of striking the nail. If the test of the hammer’s task can be performed adequately at that level of
position shows the hammer to be in the down position, description. The authors proposed that this estimate was
then the operation of lifting the hammer is triggered. If likely to include decisions about cost-critical aspects of
the hammer is already in the up position, then this the performance and difficulty of the task.
operation is omitted. There are many parallels with this
form of analysis of control structures and the triggering
of operations with the representations used in HTA. An 2. Development of HTA
illustration of the hammering task analysed by HTA is
presented in Fig. 3 for comparison with the TOTE In the original paper laying out the approach for
analysis in Fig. 2. conducting HTA, Annett et al. (1971) make it clear that
As shown by comparison between Figs. 2 and 3, the the methodology is based upon a theory of human
sub-goal and plans from HTA in Fig. 2 map onto the performance. They proposed three questions as a test
TOTE units in Fig. 1. The hierarchical systems analysis for any task analysis method, namely: does it lead to any
and control structures, whilst represented differently, positive recommendations, does it apply to more than a
the two systems of analysis are comparable. A super- limited range of tasks, and does it have any theoretical
ordinate goal of ‘‘Make nail flush’’ has been added in justifications? Perhaps part of the answer for the
the HTA, but this was implicit in the hierarchical plan longevity of HTA is that the answer to each of these
shown in Fig. 2. The plans in HTA act as the control questions is positive. More modern methods might well
 ARTICLE IN PRESS
58 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

fail some of these criteria. To paraphrase Annett et al.’s described in the worked examples shows how the
words, the theory is based on goal-directed behaviour analyst works in a process of continual reiteration and
comprising a sub-goal hierarchy linked by plans. Thus, refinement. To start with the goals are described in
performance towards a goal can be described at multiple rough terms to produce an outline of the hierarchy. This
levels of analysis. The plans determine the conditions allows further clarification and analysis. Progressive re-
under which any sub-goals are triggered. The three main description of the sub-goal hierarchy could go on
principles governing the analysis were stated as follows: indefinitely, and Annett et al. (1971) caution that
knowing when to stop the analysis is ‘‘one of the most
1. At the highest level we choose to consider a task as
difficult features of task analysis’’ (Annett et al., 1971, p.
consisting of an operation and the operation is
6). The criterion for stopping the analysis was deter-
defined in terms of its goal. The goal implies the
mined through the probability of failure (P) multiplied
objective of the system in some real terms of
(
) by the cost of failure (C) to an acceptable level,
production units, quality or other criteria.
known as the P
C rule. Annett et al. (1971) admit that
2. The operation can be broken down into sub-
it is not always easy to estimate these values and urge
operations each defined by a sub-goal again measured
task analysts not to pursue re-description unless it is
in real terms by its contribution to overall system
absolutely necessary.
output or goal, and therefore measurable in terms of
The stopping rule is simple enough in its conception:
performance standards and criteria.
P
C is acceptable then stop the task analysis; if P
C
3. The important relationship between operations
is unacceptable, then the analysis should continue. Under
and sub-operations is really one of inclusion; it is a
most situations, the probabilities and costs are not known
hierarchical relationship. Although tasks are often
and the analyst has to apply an approximation, although
proceduralised, that is the sub-goals have to be
it may not be clear what they are basing this judgement
attained in a sequence, this is by no means always the
on. Stammers and Astley (1987) point out that the
case. (Annett et al., 1971, p. 4)
stopping rule has remained a problem area for HTA. The
It is important to fully digest these three principles, P
C rule attempts provides an economy of description.
which have remained unwavering throughout the past There is no need to re-describe every sub-goal down to the
34 years of HTA. In the first principle, HTA is proposed most basic, elemental, level if failure to perform that sub-
as a means of describing a system in terms of its goals. goal is inconsequential. Piso (1981) notes that the P
C
Goals are expressed in terms of some objective criteria. criterion is complicated and time consuming. His
The two important points here are that HTA is a goal- proposed solution to this problem is to continue the
based analysis of a system and that a system analysis is analysis until the sub-goal is clear to both the analyst and
presented in HTA. These points can escape analysts who subject-matter expert(s). Annett (2004, pers. commun.)
think that they are only describing tasks carried out by has pointed out that ‘‘it is important to think of the P
C
people, whereas HTA is quite capable of producing a criterion as a statement of principle rather than an exact
systems analysis. Therefore, HTA can be used to calculation’’. The role of the P
C rule seems to be to
describe both team work and non-human tasks save the analyst time in analysing tasks where the error
performed by the system. HTA describes goals for variance would be inconsequential, and to guide more
tasks, such that each task is described in terms of its exploration where the error variance would be intolerable.
goals. ‘Hierarchical Sub-Goal Analysis of Tasks’ might The original hierarchical number scheme for HTA
be a better description of what HTA actually does. required that every sub-goal was uniquely numbered
In the second principle, HTA is proposed as a means with an integer in sequence. Each sub-goal was further
of breaking down sub-operations in a hierarchy. The identified by stating its super-ordinate goal and its
sub-operations are described in terms of sub-goals. This position under that sub-goal. This arrangement is
reiterates the point above, that HTA is a description of a illustrated in Fig. 4. The overall goal of ‘Operate
sub-goal hierarchy. Again the sub-goals are described in radiator line’ is numbered ‘1’ as the super-ordinate
terms of measurable performance criteria. The final goal. The immediate subordinate goals are numbered
principle states that there is a hierarchical relationship 2–7 (only 2–4 are shown in Fig. 4). The sub-goal
between the goals and sub-goals and there are rules to ‘Operate control panel’ has additional numbering of
guide the sequence that the sub-goals are attained. This ‘1,1’ to denote that it is the first sub-goal of super-
means that in order to satisfy the goal in the hierarchy ordinate goal 1. ‘Control cross welder’ is denoted 3/1, 2
its immediate sub-goals have to be satisfied, and so on. to show that its unique identifier is sub-goal 3, and that
The sequence with which each sub-goal is attained is it is the second sub-goal of super-ordinate goal 1.
guided by the rules that govern the relationship between Likewise, sub-goals 8, 9, and 10 show their relationship
the immediate super-ordinate goal and its sub-ordinates. to their super-ordinate goal 3.
In their original paper, Annett et al. (1971) present As well as the hierarchical diagram, Annett et al.
some industrial examples of HTA. The procedure (1971) specified the production of a table for representing
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 59

task relevant information as illustrated in Table 1. The sharing rule required some sub-goals to be performed
numbers in the left-hand column identify the goals in the in tandem. Annett et al. (1971) argued that if the HTA is
hierarchical diagram (although this is a different task conducted properly it could be applied immediately to
being analysed to that in Fig. 4). The next column training design.
(Description of Operation and Training Notes) contains Some criticisms of the original specification of HTA
the goal name, an ‘R’ if it is to be re-described elsewhere were made by Shepherd (1976), who proposed enhance-
in the table, and notes relevant to training performance, ments to the tabular format as used to supplement the
methods, and constraints (as HTA was original devised hierarchical diagram, but he identified some potential
to address training specification). The column titled ‘I or weaknesses with the original table layout. His objections
F’ would contain an ‘X’ if there were any Input or were: the remoteness of plans; combining information
Feedback difficulties found in performance of the task. on operations, plans, and training notes into one
Similarly, the column titled ‘A’ would contain an ‘X’ if column; and the usefulness of the two columns for
there were any Action difficulties found in performance sensory information/perceptual feedback (I/F) and
of the task. motor action (A). Shepherd argued that close proximity
Annett et al. (1971) intended the ‘I or F’ and ‘A’ of plans to the operations to which they refer would
columns as memory aids for the analyst. They suggest reduce confusion for the analyst and anyone else who
that the analyst should ask of every task if there were has to interpret the analysis. He proposed that the plans
any difficulties with the input–action–feedback cycle of and operations should be grouped together so that the
behaviour. The order of the sub-goals was governed by a control structure governing the sequence of operation is
rule determining their exact sequence. In the original easy to refer to. Shepherd also argued that the plans,
specification of HTA, three types of rule were identified: operations, and training notes should not appear within
procedure or chain, selection, and time sharing. The the same column. To deal with these criticisms,
procedure or chain rule required that the sub-goals were Shepherd proposed an improved tabular format to
performed in a fixed sequence. The selection rule overcome the problems as he saw them. An example of
indicated that the sub-goals were selected depending the revised tabular format is illustrated in Table 2.
upon the outcome of another sub-goal. The time- Further changes and simplifications have been pro-
posed over the years. For example, an HTA training
manual by Patrick et al. (1986) has proposed only three
columns, with the notes column used for suggestions on
1. how the analyst can improve the sub-goals. A variety of
Operate other formats for doing this have emerged over the
radiator
line years, such as separate columns for job design, job aids,
training, and procedures (Kirwan and Ainsworth, 1992).
The hierarchical numbering system proposed in the
original format was more complex than it needed to be
2/1,1 3/1,2 4/1,3
Operate Control Control and has subsequently been replaced with a decimal
control cross seam format. In the original proposal, all sub-goals were
panel welder welder numbered by integers from left to right, from 1 onwards,
with their relationship with their immediate super-
ordinate goal expressed underneath. Thus, sub-goal 2
8/3,1 9/3,2 10/3,3 in Table 2 was the first sub-division of super-ordinate
Carry Clear Recify goal 1, so was denoted 2/1,1., whereas sub-goal 3 was
out pile-ups faults
checks the second sub-goal of super-ordinate goal 1, so was
denoted 3/1, 2. The first number refers to the unique
Fig. 4. Numerical hierarchy system specified for HTA. number of the sub-goal, the next number refers to the

Table 1
Part of the original tabular format

No. Description of operation and training notes (R ¼ re-description) I or F A Re-described

1. Operate acid purification plant. R Instructions when to start-up or — X 2–4

shut-down the whole process given by supervisor.
2: Start-up plant. R Must memorise order of units, i.e., C10, R2, C12. — X 5–7
1;2
3: Run plant. R Log keeping and sampling tests for contamination at — X 8–10
1;2
intervals fixed by supervisor. Alarm signal dynamic failure.
 ARTICLE IN PRESS
60 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

Table 2
Part of the improved tabular format proposed by Shepherd

Super-ordinate Task component—operation or plan Reason for stopping the Notes on performance,
analysis training, and further
analysis

1. Operate acid purification plant

Plan 1: Instructions to start-up or shut-down are given by the plant
supervisor.
—————————————————————————————
2. Start-up plant
3. Run plant
4. Shut-down plant

2. Start-up plant
Plan 2: Units must be started up in the following order: first column Plan 2. The sequence must
10, second reactor 2, third column 12. be memorised
—————————————————————————————
5. Start-up column 10
6. Start-up reactor 2
7. Start-up column 10

Table 3
Part of the tabular format with hierarchical numbering

Super- Task component—operation or plan Notes

ordinate

1. Operate acid purification plant

Plan 1: Instructions to start-up or shut-down are given by the plant supervisor.
——————————————————————————————————
1.1. Start-up plant
1.2. Run plant
1.3. Shut-down plant
1.1. Start-up plant Plan 1.1. Provide a memory prompt for
the sequence
P1.1: 1-2-3-EXIT
——————————————————————————————————
1.1.1. Start-up column 10 //
1.1.2. Start-up reactor 2 //
1.1.3. Start-up column 10 //

super-ordinate goal, and the last number refers to the stage in the sub-goal re-description, all of these
position under the super-ordinate goal. Under the questions could have been asked—depending upon the
decimal system these sub-goals would be referred to as problem domain.
1.1 and 1.2, respectively, to show that they were the first The questions in the training and job design studies
and second sub-goals of super-ordinate goal 1. The were devised from the four-stage control loop model of
advantages of this newer system is that it makes it far performance (Piso, 1981): perception-decision-ac-
easier to trace the family tree of any sub-goal. Imagine tion-evaluation. This general model can be used to
trying to find the genealogy of sub-goal 1.3.2.4.6 under describe all tasks and is probably implicit in all HTA, as
the original system in the tabular format. An illustration it would be rather cumbersome to ask each of the
of how the newer system of hierarchical decimal questions explicitly at every single sub-goal.
numbering represents the sub-goal is illustrated in Table The enduring popularity of HTA can be put down to
3. two key points. First, it is inherently flexible: the
Some researchers have produced a semi-structured approach can be used to describe any system. Astley
approach to question the problem under analysis (Piso, and Stammers (1987) point out that over the decades
1981; Hodgkinson and Crawshaw, 1985; Bruseberg and since its inception, HTA has been used to describe each
Shepherd, 1997). Three examples of the problem new generation of technological system. Second, it can
domains are presented in Table 4, training design, be used for many ends: from personnel specification, to
interface design, and job design. Potentially, at each training requirements, to error prediction, to team
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 61

Table 4
Questions for sub-goals

Training design Interface design Job design

Piso (1981) Hodgkinson and Crawshaw (1985) Bruseberg and Shepherd (1997)

What is the goal of the task? What are the sensory inputs? How does information flow in the task?
What information is used for the decision to How can the display of information be When must tasks be done?
act? improved?
When and under what conditions does the What are the information processing What is the temporal relation of tasks?
person (system) decide to take action? demands?
What are the sequence of operations that are What kind of responses are required? What are the physical constraints on tasks?
carried out?
What are the consequences of action and How can the control inputs be improved? Where can and cannot error and delay be
what feedback is provided? tolerated?
How often are tasks carried out? What kind of feedback is given? Where is workload unacceptable?
Who carries the tasks out? How can the feedback be improved? Where is working knowledge common to
more than one task element?
What kinds of problems can occur? How can the environmental characteristics Where do different tasks share the same or
be improved? similar skills?

performance assessment, and to system design. How- errors and quicker solutions, over the paper-based
ever, despite the enduring use of HTA, and the fact that counterparts.
the analysis is governed by only a few rules, it is In one of the few comparative studies, Miller and
something of a craft-skill to apply effectively. Whilst the Vicente (2001) compare HTA with the Abstraction
basic approach can be trained in a few hours, it is Hierarchy in the analysis of the DURESS II (DUal
generally acknowledged that sensitive use of the method REservoir System Simulation developed at the Uni-
will take some months of practice under expert guidance versity of Toronto) with the purpose of producing
(Stanton and Young, 1999). display requirements. Although they do not present the
A more recent innovation in HTA has been sub-goal output of the analysis in their paper, Miller and Vicente
templates (SGTs) to help formalise the process and help compare the types of information produced. They report
guide the novice analyst. Ormerod and Shepherd (2004) that the two methods produce different, but comple-
propose the adoption of sub-goal and plan templates to mentary, sets of display requirements. Their research
assist in the process of re-description in HTA. They points to some shortcomings of HTA, such as the lack
argue that these two tools could help make the process of representation of physical objects, propagation
of HTA less daunting and reduce the inevitable learning effects, causal understanding, and social-organisational
curve associated with acquiring a new analytical knowledge. These criticisms might have been withdrawn
technique. The SGTs comprise action templates (e.g., if they had used some of the extensions of HTA (as
activation, adjustment, and deactivation), exchange described in Section 5). Miller and Vicente argued that
templates (e.g., entering and extracting data), navigation HTA is a useful addition to the Abstraction Hierarchy.
templates (e.g., locating, moving, and exploring), and Some of their comments on the level and type of the
monitoring templates (e.g., detecting, anticipating, and analysis show that they are using HTA in a very
observing). A fuller description of the SGTs is provided constrained way. For example, they note that HTA
in Table 5. focuses on human action whereas the abstraction
Although the SGTs were developed with process hierarchy focuses on the whole system. Annett and
control operations in mind, they can be applied more others have argued that HTA can provide sub-goal
widely. As people start to use the SGTs in other hierarchies at many levels within a system. The analyst
domains, new SGT might become necessary. Within can choose to focus on the human agents, machine
each, the analyst may choose a plan template to help agents or the entire system. Thus, one is drawn to the
determine the sequence of the sub-goals. Ormerod and conclusion that some of the critique could be due to an
Shepherd (2004) proposed four plan templates, as incomplete understanding of HTA or the way it has
illustrated in Table 6. been portrayed in some of the materials they have cited.
Ormerod et al. (1998) report studies evaluating the In a comparison of different task analysis representa-
effectiveness of novice analysts performing HTA with tions, Stanton (2004) identified five forms that encom-
the SGT tools. They show that the SGT tools can help passed most methods: list, narratives, flow diagrams,
novice analysts, particularly with mastery of more hierarchical diagrams, and tables. In a comparison of 22
difficult analyses. Computerisation of the SGT tools methods, only three had three different forms of
led to even better performance, as measured by fewer representation. Most methods relied upon only one
 ARTICLE IN PRESS
62 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

Table 5
Sub-goal templates

SGTs Task element Context for assigning SGT and task element

Act: To operate as part of a procedure A1: Activate To make a subunit operational, e.g., to switch from an ‘off’ state to an ‘on’ state
A2: Adjust To regulate the rate of operation of a unit maintaining an ‘on’ state
A3: Deactivate To make a subunit non-operations, e.g., to switch from an ‘on’ state to an
‘off’ state
Exchange: To exchange information E1: Enter To record a value in a specified location
E2: Extract To obtain a value of a specified parameter
Navigate: To search for information N1: Locate To find the location of a target value or control
N2: Move To go to a given location and search it
N3: Explore To browse through a set of locations and values

Monitor: To monitor system state and look M1: Detect To routinely compare the system state against the target state in order to
for change determine the need for action
M2: Anticipate To compare the system state against the target state in order to determine
readiness for a known action
M£: Transition To routinely compare the rate of change during a system state transition

Table 6 principles mainly guide the progressive sub-goal re-

Plan templates description and nomenclature, although there is an
Code Plan type Syntax underlying psychological model of feedback and control
loops in the analysis. That said, the basic heuristics for
S1 Fixed sequence Do X, Y, Z conducting an HTA are as follows:
S2 Contingent sequence If (c) then do X
(i) Define the purpose of the analysis: Although the
If not (c) then do Y
S3 Parallel sequence Do together X, Y, Z case has been made that HTA can be all things to all
S4 Free sequence In any order do X, Y, Z people, the level or re-description and the associated
information collected might vary depending upon the
purpose. Purposes for HTA include system design,
interface design, operating procedures design, develop-
form of representation. It seems fair to suggest that ing personnel specifications, analysis of workload and
HTA benefits from multiple forms of representation, manning levels, and training design. The name(s),
and this is indicative of the flexibility of the approach. contact details, and brief biography of the analyst(s)
should also be recorded. This will enable future analysts
to check with the HTA originator(s) if they plan to re-
3. A framework for conducting HTA use or adapt the HTA.
(ii) Define the boundaries of the system description:
There do seem to be many different conventions for Depending upon the purpose, the system boundaries
expressing HTA that have developed from peoples’ own may vary. If the purpose was to develop a personnel
adaptation and mutations. It is difficult, therefore, to specification then the system boundary might be drawn
propose that there is one right way of doing this, around the tasks performed by that individual. If the
although some have tried (Shepherd, 1989, 2001; purpose of the analysis is to analyse co-ordination and
Annett, 2004). This section will follow the examples of communication in team work, then the entire set of
Stammers (1996) and Shepherd (1998, 2001) to propose tasks of a team of people would be analysed. If the
a framework within which HTA can be conducted, purpose of the analysis is to determine allocation of
allowing for personal adaptation for the purpose at system function to human and computers, then the
hand. whole system will need to be analysed. Both Shepherd
The number of guidelines for conducting HTA are (2001) and Annett (2004) emphasise the need to perform
surprisingly few. Annett (1996) has pointed out that the the analysis appropriate to the intended purpose.
methodology is based on some broad principles (as (iii) Try to access a variety of sources of information
detailed earlier), rather than a rigidly prescribed about the system to be analysed: All task analysis guides
technique. This fits well with Shepherd’s (2001) view stress the importance of multiple sources of information
that HTA is a framework for task analysis. The broad to guide, check, and validate the accuracy of the HTA
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 63

(Patrick et al., 1986; Kirwan and Ainsworth, 1992; the sub-goals between 4 and 8, but if there are more than
Shepherd, 2001; Annett, 2004). Sources such as ob- 10 sub-ordinates the analyst should check to see if any of
servation, subject-matter experts, interviews, operating the sub-goals can be grouped together under another
manuals, walkthoughs, and simulations can all be used super-ordinate. It is generally good practice to con-
as a means of checking the reliability and validity of the tinually review the sub-goal groupings, to check if they
analysis. Careful documentation and recording of the are logical. HTA does not permit single subordinate
sources of data needs to be archived, so that the analyst goals.
or others may refer back and check if they need to. (vi) Link goals to sub-goals, and describe the condi-
Annett (2004) points out that cross-checking the data tions under which sub-goals are triggered: Plans are the
between sources is the best guarantee that the informa- control structures that enable the analyst to capture the
tion is accurate. conditions which trigger the sub-goals under any super-
(iv) Describe the system goals and sub-goals: As ordinate goal. Plans are read from the top of the
proposed in the original principles for HTA, the overall hierarchy down to the sub-goals that are triggered and
aim of the analysis is to derive a sub-goal hierarchy for back up the hierarchy again as the exit conditions are
the tasks under scrutiny. As goals are broken down and met. Shepherd (2001) identified six basic types of plan:
new operations emerge, sub-goals for each of the fixed sequences, contingent sequences, choices, optional
operations need to be identified. As originally specified, completion, concurrent operations, and cycles. These
it is not the operations that are being described, but their different types of plans take the variety of different sub-
sub-goals (Annett et al., 1971). All of the lower level goal triggers into account. He states that complex tasks
sub-goals are a logical expansion of the higher ones will require combinations of these different sorts of
(Patrick et al., 1986). A formal specification for the plans. As each of the sub-goals, and the plans that
statement of each of the sub-goals can be derived, trigger them, are contained within higher goals (and
although most analyses do not go such lengths. Patrick higher plans) considerable complexity of tasks within
et al. (1986) describe the three components of these systems can be analysed and described. The plans
statements, as indicated in Table 7. Obviously, this is a contain the context under which particular sub-goals
trivial task, but it does show how the task statement can are triggered. This context might include time, environ-
be composed and describes its relationship with the goal mental conditions, completion of other sub-goals,
(Fig. 3). system state, receipt of information, and so on. For
As Table 7 shows, the goal is presented in the activity each goal, the analyst has to question how each of its
verb. The performance standards and the conditions immediate subordinates is triggered. Ormerod and
could be expressed in the notes section of the tabular Shepherd (2004) have proposed some basic plan
format. templates to guide this process (see Table 7). As well
(v) Try to keep the number of immediate sub-goals as identifying the sub-goal trigger conditions, it is also
under any super-ordinate goal to a small number (i.e., important to identify the exit condition for the plan that
between 3 and 10): There is an art to HTA, which will enable the analyst to trace their way back up the
requires that the analysis does not turn into a procedural sub-goal hierarchy. Otherwise, the analysis could be
list of operations. The goal hierarchy is determined by stuck in a control loop with no obvious means of
looking for clusters of operations that belong together exiting.
under the same goal. This normally involves several (vii) Stop re-describing the sub-goals when you judge
iterations of the analysis. Whilst it is accepted that there the analysis is fit-for-purpose: When to stop the analysis
are bound to be exceptions, for most HTAs any super- has been identified as one of the more conceptually
ordinate goal will have between 3 and 10 immediate troublesome aspects of HTA. The proposed P
C
subordinates. Patrick et al. (1986) recommend keeping stopping rule is a rough heuristic, but analysts may have

Table 7
The elements of task statements

Task statement element Questions Example

Activity verb Is it clearly defined? To make the nail flush.

Is it differentiated?
Does it state the objective of the behaviour?
Performance standards Is the quantity or quality of the performance specified (e.g., speed, ywithout damaging the surface of a
accuracy, errors, etc.)? piece of woody
Conditions Are the conditions under which the task is to be performed described (e.g., yusing a hammer.
environment, tools, materials, etc.)?
 ARTICLE IN PRESS
64 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

trouble quantifying the estimates of P and C. Annett et START

al. (1971) proposed that it is likely to be preferable to
stop the analysis early than to continue it beyond the State overall goal
point at which it will be useful. The level of description
is likely to be highly dependent upon the purpose of the
State subordinate goals Select next goal
analysis, so it is conceivable that a stopping rule could
be generated at that point in the analysis. For example,
in analysing team work the analysis could stop at the State plan
point where sub-goals dealt with the exchange of
information (e.g., receiving, analysing and sending Check adequacy of
redescription
information from one agent to another). For practical
purposes, the stopping point of the analysis is indicated
by underlining the lowest level sub-goal in the hierarch- is N
redescription Revise
ical diagram, or ending the sub-goal description with a ok? redescription
double forward slash (i.e., ‘‘//’’) in the hierarchical list
and tabular format. This communicates to the reader Y
that the sub-goal is not re-described further elsewhere in Consider first/next
sub-goal
the document.
(viii) Try to verify the analysis with subject-matter
experts: Annett (2004) makes the point that it is is further
important to check the HTA with subject-matter redescription Y
warranted?
experts. This can help both with verification of the
Y
completeness of the analysis and help the experts N
develop a sense of ownership of the analysis. Terminate redescription any more
(ix) Be prepared to revise the analysis: HTA requires a of this goal goals?
flexible approach to achieve the final sub-goal hierarchy
with plans and notes. The first pass analysis is never N
going to be sufficiently well developed to be acceptable, STOP
no matter what the purpose. The number of revisions
Fig. 5. Procedure for breaking down the sub-goal hierarchy.
will depend on the time available and the extent of the
analysis, but simple analyses (such as the analysis of the
goals of extracting cash from an automatic teller
machine) may require at least three iterations, where
more complex analyses (such as the analysis of the found in Shepherd (2001), who devotes an entire chapter
emergency services responding to a hazardous chemical to plans in his book.
incident) might require at least 10 iterations. It is useful Some plans may use one of these basic types whereas
to think of the analysis as a working document that only others may be a hybrid combining two or more. The
exists in the latest state of revision. Careful documenta- three different representations of HTA are hierarchical
tion of the analysis will mean that it can be modified and diagrams, hierarchical lists, and the tabular format.
re-used by other analysts as required. Each of these is illustrated through a team work task
A procedure for development of the sub-goal from Baber et al. (2004). These examples show the
hierarchy with the plans is presented in Fig. 5. This compatibility of the three different representations. The
procedure only describes the steps (iv)–(vii) in the HTA was based upon the analysis of the emergency
guidance, but offers a useful heuristic for breaking the services responses to a hazardous chemical incident. In
tasks down into a sub-goal hierarchy. the scenario analysed, some youths had broken into a
The notation used by HTA analysts can be standar- farm and disturbed some chemicals in sacking. One of
dised to help to ensure the analysis can be interpreted by the youths had been taken to the hospital with
others (Patrick et al., 1986; Shepherd, 2001; Annett, respiratory problems, whilst the others were still at the
2004). Standard conventions tend to use either text or scene. The police were sent to investigate the break-in at
symbols (Shepherd, 2001). Examples of the text and the farm. They called in the fire service to identify the
symbols that have been used are indicated in Table 8. chemical and clean up the spillage. The overall analysis
The notation in Table 8 is used in the plans to indicate shows four main sub-goals: receive notification of an
the sequence, and trigger condition, for the sub-goals. incident, gather information about the incident, deal
Six different forms of plans with three different notation with the chemical incident, and resolve incident. Only
conventions are shown in Table 9. A more detailed part of the analysis is presented in Figs. 6 and 7, to
description of the forms that plans can take may be illustrate HTA.
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 65

Table 8 0.
Plan 0.
Notation used in HTA Wait until 1 then 2 then 3 Deal with
If [hazard] then 4 then chemical
incident
Text Symbols 5 then exit
Else exit
then 4-
and +& 1. 4.
2.
or / [Police Control] [Police Control] [Fire Control]
any of : receive notice gather deal with
from public about information chemical
decide ? incident about incident incident
if condition X then X?4
3.
Plan 2.
[Police Control]
Do 2.1 at any time
make decison
Do 2.2 then 2.3
about nature of
Then exit
incident

Table 9
Different plan types with three notation conventions 2.1. 2.2. 2.3. Plan 2.3
[Hospital] inform [Police Control] [Police Control] If hazards] then 2.3.1
police control of get a Police get a Police If [suspects] then 2.3.2
Type of plan Types of notation then 2.3.3
casualty with Officer to search Officer to report
respiratory scene of incident nature of incident Then 2.3.4 then exit
Linear 1424344 problems Else exit
Sequential plan 1 then 2 then 3 then 4
Plan 2.2.
Do in order Do 2.2.1
2.2.2. 2.3.1. 2.3.3.
Then 2.2.2.
[Police Officer] [Police Officer] [Police Officer]
Non-linear 1/2/3/4 Then 2.2.3
identify possible gather
Until arrive at scene
Non-sequential plan N/A [suspects] of incident hazard information from
Do in any order suspects
or
[hazards]
Simultaneous 1+2+3+4 Then exit
Concurrent plan 1 and 2 and 3 and 4 2.2.1. 2.2.3. 2.3.2. 2.3.4.
[Police Control] [Police Officer] [Police Officer] [Police Officer]
Do at the same time send Police search scene of capture inform police
Office to scene of incident suspects control of nature
Branching X? Y42 N43 incident of incident
Choice plan If X present then 2 else 3
Do when required
Fig. 6. Part of the hierarchical diagram for the goal of ‘‘Deal with
Cyclical 142434441y chemical incident’’.
Repetitious plan 1 then 2 then 3 then 4 then repeat from 1 until
Repeat the following until
Selection 1:2:3:4
Exclusive plan 1 or 2 or 3 or 4
become cumbersome and unwieldy. For these analyses
Choose one of the following
a hierarchical list approach might be more useful. The
same analysis in Fig. 6 is presented as a hierarchical list
in Fig. 7 for comparison.
The hierarchical diagram and hierarchical list present
In Fig. 7, the overall goal is shown at the top of the exactly the same information on the sub-goal hierarchy
hierarchy with the main sub-goals underneath. Plan 0 in two different forms. The advantage of the diagram is
shows the conditions under which each of the sub-goals that it represents the groups of sub-goals in a spatial
is triggered. As sub-goal 1 is not re-described, it has been manner which is useful for gaining a quick overview of
underlined. Sub-goal 2 is re-described, and has eight the HTA. The hierarchical lists show the same informa-
sub-goals of its own. Plan 2 refers to the conditions tion in a more condensed format, which is useful for
under which the sub-goals of super-ordinate goal 2 will very large analyses. It is possible to annotate the sub-
be triggered. As none of the sub-goals under super- goal hierarchy with the tabular format, as illustrated in
ordinate goal 2 are re-described further they have been Table 10.
underlined. The tabular format permits more detail of how the
As multiple agencies and people are involved in the emergency services deal with the incident. The analysis is
team task, they have been identified under each of the not exhaustive, nor is it complete. Rather it is presented
sub-goals. In Fig. 6, police control, fire control, the to serve as an illustration of how the three different
hospital and the police officer have all been assigned to representations of HTA present information on the
different sub-goals. same sub-goal hierarchy. HTA serves as a springboard
As Fig. 6 shows, the tree-like structure of the for a variety of other techniques. Once the sub-goal
hierarchical diagram makes it reasonably easy to trace hierarchy has been broken down, many other forms of
the genealogy of sub-goals for small scale analyses. For analysis may be carried out on it. This is the subject of
larger scale analyses, the hierarchical diagram can Section 4.
 ARTICLE IN PRESS
66 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

0. Deal with chemical incident In the large-scale design and development of a new
Plan 0: Wait until 1 then do 2 then 3-If [hazard] then 4 then 5 then exit - nuclear reactor, Staples (1993) describes how HTA was
Else exit
1. [Police control] receive notice from public about incident //
used as the basis for virtually all of the ergonomics
2. [Police Control] gather information about incident studies. The sub-goal hierarchy was produced through
Plan 2: Do 2.1 at any time if appropriate reviews of contemporary operating procedures, discus-
Do 2.2 then 2.3 sions with subject-matter experts, and interviews with
Then exit
2.1. [Hospital] inform police control of casualty
operating personnel from another reactor. Both the
with respiratory problems// hierarchical diagram and the tabular format versions of
2.2. [Police Control] get a Police Officer to search HTA were produced. The resultant HTA was used to
scene of incident examine potential errors and their consequences, the
Plan 2.2: D o 2.1.1 then 2.2.2 then 2.2.3
Until [suspects] or [hazards] then exit
interface design verification, identification of training
2.2.1. [Police Control] procedures, development and verification of operating
send Police Officer to procedures, workload assessment, and communication
scene of incident//
analysis. Staples argued that HTA is of major benefit in
2.2.2. [Police Officer]
arrive at scene of
system design as it makes possible a detailed and
incident// systematic assessment of the interactions between hu-
2.2.3. [Police Officer] man operators and their technical systems. As Annett
search scene of incident//
and colleagues have pointed out on many occasions,
2.3. [Police Control] get Police Officer to report conducting the HTA helps the analyst become familiar
nature of incident with the processes and procedures so that they can
Plan 2.3: If [suspects] then 2.3.1 critically assess the crucial aspects of the work. Staples
If[suspects] then 2.3.2. then 2.3.3
also notes that reference to the HTA for the analysis of
Then 2.3.4. then exit
Else exit all aspects of the system can highlight inconsistencies
2.3.1. [Police Officer] between training, procedures and system design. Staples
identify possible draw the general conclusion that the broad application
hazard//
2.3.2. [Police Officer]
of HTA can make it a very cost-effective approach to
capture suspects// system design.
2.3.3. [Police Officer] Most books containing descriptions of HTA also
gathe r information from contain examples of potential application areas, far
suspec ts//
2.3.4. [Police Officer] inform police
wider than the training applications for which it was
control of nature of incident// originally devised. Annett (2000) has pointed out the
HTA is a general problem solving approach, and
3. [Police Control] make decision about nature of incident// performing the analysis helps the analyst understand
4. [Fire Control] clean up chemical spillage
etc...
the nature of both the problem and the domain. An
5. etc... indication of some of the application areas is illustrated
in Table 11.
Fig. 7. Part of the hierarchical list for the goal of ‘‘Deal with chemical
There are at least 12 additional applications to which
incident’’.
HTA has been put. This list is not intended to be
exhaustive, rather it illustrates that HTA is a means-to-
an-end, rather than an end in itself (Stanton, 2004). The
4. Some applications of HTA reader is referred to the appropriate texts for examples
of the applications. Duncan (1972) has argued that a
Most, if not all, application areas in ergonomics task description should not be biased in terms of any
require some form of task representation. Kirwan and particular solution. An example of HTA applied to the
Ainsworth (1992) claim that HTA may ‘‘be used in evaluation of radio-cassette machines demonstrates this
almost every circumstance’’ (p. 29), offering a major cost point. In a study comparing a Ford and a Sharp in-car
saving in a system design program, rather than radio-cassette, Stanton and Young (1999) showed that
continually re-analysing the task for every different HTA of the drivers’ sub-goal task structure did not
type of application. Annett (2004, pers. commun.) has indicate which was a better interface. Rather the
made the point that the form of the HTA could vary analysis just showed that the sub-goal structures were
depending upon the application, so that the first or different for the two machines. To determine which was
subsequent drafts of HTA might not serve all purposes, a better interface required an extension of the analysis,
and some modifications might have to be made. This such as an examination of the error potential or task
view sits comfortably with Shepherd’s proposal of HTA time when interacting with the device. With the HTA
as a framework, a living documentation of the sub-goal completed first, the subsequent analyses are possible.
hierarchy that only exists in the latest state of revision. Many methods and techniques either depend upon
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 67

Table 10
Part of the tabular format with hierarchical numbering for the goal of ‘‘Deal with chemical incident’’

Super- Task component—operation or plan Notes

ordinate

0. Deal with chemical incident This is a multi-agency task involving the police and fire
service as well as the hospital with a possible casualty
Plan: Wait until 1 then do 2—If [hazard] then 3 then 4 then exit—else
exit
—————————————————————————————
1. [Police control] receive notice from public about incident // The response to the incident is initiated by a phone call
2. [Police Control] gather information about incident
3. [Fire Control] clean up chemical spillage The re-description is missing, to shorten the example

2. [Police Control] gather information about incident

Plan 2: Do 2.1 at any time if appropriate
Do 2.2 then 2.3
Then exit
——————————————————————————————
2.1. [Hospital] inform police control of casualty with respiratory The hospital may call in about a casualty at any time, but it
problems // has to be linked with this incident
2.2. [Police Control] get a Police Officer to search scene of incident The police officer has to find his/her way to the scene of the
incident
2.3. [Police Control] get Police Officer to report nature of incident
2.2. [Police Control] get a Police Officer to search scene of incident
Plan 2.2: Do 2.1.1 then 2.2.2 then 2.2.3
Until [suspects] or [hazards] then exit
—————————————————————————————
2.2.1. [Police Control] send Police Officer to scene of incident // The police officer may have to find a remote location based
on sketchy information
2.2.2. [Police Officer] arrive at scene of incident //
2.2.3. [Police Officer] search scene of incident // The police officer has to search for signs of a break-in and
hazards
2.3. [Police Control] get Police Officer to report nature of incident
Plan 2.3: If [suspects] then 2.3.1
If [suspects] then 2.3.2 then 2.3.3
Then 2.3.4 then exit
Else exit
—————————————————————————————
2.3.1. [Police Officer] identify possible hazard // Any potential hazard needs to be identified, including the
chemical ID number
2.3.2. [Police Officer] capture suspects // Any suspects on the scene need to be identified
2.3.3. [Police Officer] gather information from suspects // Suspects need to be questioned about the incident
2.3.4. [Police Officer] inform police control of nature of incident // Incident details need to be passed on so that the clean-up
operation can begin

output from HTA or are made easier when HTA is (16 reports), human error identification (29 reports), and
performed first. systems assessment (55 reports).
Ainsworth and Marshall (1998) describe a survey of Table 12 shows that discussions and interviews with
reports on task analysis methods, including HTA, experts were a core source of data, supplemented by
conducted in the armed services and nuclear industries. walkthoughs, direct observation, questionnaires, and
The results of their survey showed that ‘‘HTA is perhaps scenario modelling, depending upon what was possible
the nearest thing to a universal task analysis technique’’ in the area of application. Ainsworth and Marshall
(Ainsworth and Marshall, 1998, p. 83). The areas that (1998) were critical about the quality of reporting in the
HTA was used for in the armed services and nuclear task analysis accounts. They state that the purpose of
power are presented in Table 13, together with the the analysis was not always clear and the sources of the
methods of data collection that were used and the data were poorly documented. They also note that some
numbers of reports analysed. In nuclear power, there of the analyses were very superficial and showed poor
were three other areas covered: allocation of function insight; many problems may have been overcome if the
 ARTICLE IN PRESS
68 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

Table 11
Application of HTA from ergonomics texts

Application Kirwan and Wilson and Stanton (1996) Annett and Shepherd (2001)
Ainsworth Corlett (1995) Stanton (2000)
(1992)

Table 12
Areas of application of HTA in the armed services and nuclear power

Area Data collection methods Armed services Nuclear power

reports reports

Systems Technical expert interviews, informal discussions, and scenario 2 (8%) 0 (0%)
procurement modelling
Manpower analysis Walkthoughs, interviews, and discussions with experts 7 (28%) 31 (15%)
and personnel
requirements
Operability Walkthoughs and discussions with experts 5 (20%) 53 (26%)
Interface design Walkthoughs, interviews, and discussions with experts 9 (36%) 95 (46%)
Training Direct observation, discussions with experts, and questionnaires 2 (8%) 27 (13%)
Total 25 206

analyst had been properly trained in HTA and had indicated in Table 11), rather the aim of this section
followed the guiding principles. is to indicate some of the variety of the extensions
to HTA.
Shepherd (2001) has numerous examples of the
5. Some extensions of HTA application of HTA, including one of the investigating
redesign opportunities in a batch control process. The
The whole point of conducting HTA to analyse tasks, tabular format he devised for investigating this problem
means that the HTA representation is the starting contains a task taxonomy that analyses the context and
point for the analysis rather than the end point. The constraints of the tasks and their associated sub-goals.
tabular format has enabled a mechanism for extending The taxonomy comprises 12 factors that need to be
the analysis beyond the system description provided in considered when investigating the adequacy of design
the sub-goal hierarchy and plans. It is perhaps decisions in support of task performance. These factors
ironic that, whilst initial developments in HTA sought are: the difficulty of the task; the predictability,
to simplify the tabular format, latter developments controllability, and frequency of events; the severity of
have sought to extend it. These extensions in HTA consequences of error and possibility for recovery;
have enabled the analyst to investigate design decisions, information representation and feedback; the presence
analyse human–machine interaction, predict error, of environmental, situational, and task stresses; access
allocate function, design jobs, analyse team work, to help and advice; physical and environmental con-
and assess interface design. It is impossible to cover straints; and legal, industrial, and cultural compliance.
all of the extensions to HTA (for that, the reader is Table 13 shows part of the analysis presented by
referred back to source books—some of which are Shepherd, to illustrate the task taxonomy; he argues
Table 13
An analysis of the contextual constraints and conditions for a safety critical task

Context and constraints

Task analysis Stop Task Predictability Control Frequency Severity of Information Task Recover- Stressors Access to Environ- Cost of Legal Comments

N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

analysis? difficulty of events of risks of events error representation feedback ability help mental training compli-
constraint support ance

1. Deal with

ARTICLE IN PRESS
emergencies
Plan 1: Do 1, then 2 or
3 as appropriate
1.1. Assess situation N hi lo lo lo hi lo hi hi hi
to establish the extent
of the emergency This entails
1.2. Deal with local Y having a
isolation good
1.3. Deal with Y understanding of
emergency systems to enable
evaluations flexible
operations.
1.2. Deal with local Analytical skills
isolation needed and
Plan 1.2: Do 1, then 2 intelligent
1.2.1. Assess extent of N hi lo lo lo hi lo hi hi hi planning aid may
problem help
1.2.2. Isolate affected Y
area

1.3. Deal with

emergency
evaluations, etc.

69
 ARTICLE IN PRESS
70 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

that the contextual constraints and conditions interact  fault diagnosis (determining the cause of a fault or
with the design decisions in a safety critical task. alarm);
As shown in Table 13, the context and constraints  fault detection (detecting that a fault or alarm has
have been estimated at the lowest level where the sub- occurred);
goal analysis has been stopped (indicated by ‘N’ in the  decision-making (choosing between alternate courses
‘Stop analysis?’ column). Shepherd notes that these of action);
estimates may be based on data or informed com-  problem solving (finding a solution to a problem);
ment from subject-matter experts. These contextual  operation (conducting manual control).
analyses can help guide the analyst to consider what
aspects of the task need to be improved and the form An example of the output of analysis for the coal
that those improvements could take. The design preparation plant operator’s task (Astley and Stammers,
hypotheses are presented in the ‘Comments’ section of 1987) is shown in Table 14, where the information flow
the table. In the first pass analysis, all relevant design to the human operator(s) from the technical system is
hypotheses should be included, for screening at a later shown by a right pointing arrow (-) and information
point in time. flow to the technical system from the human operator(s)
Stammers and Astley (1987) have shown how HTA is shown by a left pointing arrow (’). The plans are a
can be extended to help determine the information hybrid of symbols and text.
requirements for human–computer interface design. Astley and Stammers propose that an extended tabular
Their method extends the tabular format to include format can be used for underlying assumptions about the
three additional sections on information flow (i.e., the operators’ knowledge and skills, allocation of function,
information flow to and from the interface), information operator characteristics and training. The tabular format
assumed (i.e., information that is a prerequisite for the allows for scrutiny of the sub-goals and the format is easily
performance of the task), and task classification (i.e., a adaptable to many different types of analysis.
taxonomy of operations that will be performed). The HTA has also been used to assess the error potential
analysis of information flow can detail the information in tasks. Hellier et al. (2001) describe how they
necessary to perform each part of the task. The performed HTA of a chemical sample analysis proce-
taxonomy developed by Stammers and Astley was based dure by conducting observations and interviews with
on process control operations and comprised distinct chemists. They argued that HTA helped uncover the
types of task, namely: complexities of the task. Then they identified error
potential for each sub-goal, also using observation and
 monitoring (watching developments in the process); interviews. This time the HTA served as a frame for the
 procedure (following a set sequence of tasks); observational studies and interviews, through which

Table 14
Analysis of human–computer interaction

Super-ordinate Plan Operations Information flow Information Task Notes

across interface assumed classification

0. 1-2 and ntil 4. 1. Start-up plant -Initiate start Start-up Procedure

procedure
Operate coal Do 5 as ’Plant items
preparation plant appropriate and 6 selected
at end of shift.
2. Run plant normally ’Plant Knowledge of Operation
operation and plant flows and
monitoring operational
procedures
-Control
information
3. Carry out fault ’Fault data Some Fault detection
detection and fault understanding of
diagnosis faults
4. Shut-down plant ’Initiate shut- Shut-down Fault detection
down procedures
5. Operate telephone ——————— Operational Procedure
and tannoy knowledge
Operation
6. Make daily reports ’Plant data for Reporting Procedure
log procedure
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 71

potential errors could be assessed. As well as observing Table 15

Error modes and their description
errors, HTA can be used as a basis for predicting errors.
Systematic human error reduction and prediction Error mode Error description
approach (SHERPA) for example, uses an error
taxonomy to predict potential errors from the HTA Action
A1 Operation too long/short
sub-goal hierarchy (Stanton and Young, 1999). The idea
A2 Operation mistimed
is that each task can be classified into one of the five A3 Operation in wrong direction
basic types. Each of these task types links with an error A4 Operation too much/little
taxonomy to identify credible errors associated with a A5 Misalign
sequence of human activity. In essence, the SHERPA A6 Right operation on wrong object
A7 Wrong operation on right object
technique works by indicating which error modes are A8 Operation omitted
credible for each task step in turn. This indication is A9 Operation incomplete
based upon the judgement of the analyst, and requires A10 Wrong operation on wrong object
subject-matter experts. The process begins with HTA.
Information retrieval
For the application of SHERPA, each task step from R1 Information not obtained
the bottom level of the sub-goal hierarchy is taken in R2 Wrong information obtained
turn. First each task step is classified into one of the R3 Information retrieval incomplete
following types from the taxonomy: Checking
C1 Check omitted
 action (e.g., pressing a button, pulling a switch, C2 Check incomplete
opening a door), C3 Right check on wrong object
C4 Wrong check on right object
 retrieval (e.g., getting information from a screen or
C5 Check mistimed
manual), C6 Wrong check on wrong object
 checking (e.g., conducting a procedural check),
Information communication
 information communication (e.g., talking to another
I1 Information not communicated
party), I2 Wrong information communicated
 selection (e.g., choosing one alternative over an- I3 Information communication incomplete
other). Selection
S1 Selection omitted
This classification of the task step then leads the S2 Wrong selection made
analyst to consider credible error modes associated with
that activity, as shown in Table 15.
The sub-goal hierarchy (without the plans) for the
task of programming a video cassette recorder is (C) Failing to select the program number.
presented in the left-hand column of Table 16. This (D) Failing to wait.
example is taken from Stanton (2003). Where the sub- (E) Failing to enter programming information cor-
goals are broken down further, the SHERPA analysis rectly.
has not been undertaken. This is in keeping with the (F) Failing to press the confirmation buttons.
general SHERPA approach. For each sub-goal that is
analysed, credible error modes (i.e., those judged by a The purpose of SHERPA is not only to identify
subject-matter expert to be possible) are identified and potential errors with the current design, but also to
labelled using the codes from Table 15. A description of guide future design considerations. The structured
the form that the error would take is also given. The nature of the analysis can help to focus the design
consequence of the error on the system is determined in remedies on solving problems, as shown in the remedial
the next column, as this has implications for the strategies column. As this analysis shows, quite a lot of
criticality of the error. The last four steps consider the improvements could be made. It is important to note,
possibility for error recovery, the ordinal probability of however, that the improvements are constrained by the
the error (high, medium, or low), its criticality (high, analysis. This does not address radically different design
medium, or low) and potential remedies. Again, all of solutions that may remove the need to program at all.
these analyses are shown in Table 16. Marsden and Kirby (2005) describe the application of
As Table 17 shows there are six basic error types HTA to function allocation. Allocation of system
associated with the activities of programming a VCR. function has been a problem that has challenged
These are: ergonomics researchers and practitioners for the past
50 years. Numerous methods have arisen, but all depend
(A) Failing to check that the VCR clock is correct. upon an adequate description of the system within
(B) Failing to insert a cassette. which functions are to be allocated to humans or
 72
Table 16
The SHERPA table

Sub-goal Error Error description Consequence Recovery P C Remedial strategy

mode

1. Prepare VCR N/A N/A N/A N/A N/A

1.1. Switch VCR on A8 Fail to switch VCR on Cannot proceed Immediate L L Press of any button to switch VCR on
1.2. Check clock time C1 Omit to check clock VCR clock time may be None L H Automatic clock setting and adjust via radio transmitter

N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

incorrect
C2 Incomplete check
1.3. Insert cassette A3 Insert cassette wrong way around Damage to VCR Immediate L H Strengthen mechanism
A8 Fail to insert cassette Cannot record Task 3 L H On-screen prompt

ARTICLE IN PRESS
2. Pull down front cover A8 Fail to pull down front cover Cannot proceed Immediate L L Remove cover to programming
3. Prepare to program N/A N/A N/A N/A N/A
3.1. Set timer selector to program S1 Fail move timer selector Cannot proceed Immediate L L Separate timer selector from programming function
3.2. Press ‘Program’ A8 Fail to press PROGRAM Cannot proceed Immediate L L Remove this task step from sequence
3.3. Press ‘On’ button A8 Fail to press ON button Cannot proceed Immediate L L Label button START TIME
4. Enter Program details N/A N/A N/A N/A N/A
4.1. Select channel N/A N/A N/A N/A N/A
4.1.1. Press ‘Channel up’ button A8 Fail to press UP button Wrong channel selected None M H Enter channel number directly from keypad
4.1.2. Press ‘Channel down’ button A8 Fail to press DOWN button Wrong channel selected None M H Enter channel number directly from keypad
4.2. Press ‘Day’ button A8 Fail to press DAY button Wrong day selected None M H Present day via a calendar
4.3. Set start time I1 No time entered No program recorded None L H Dial time in via analogue clock
I2 Wrong time entered Wrong program recorded None L H Dial time in via analogue clock
4.4. Wait for 5 s A1 Fail to wait Start time not set Task 4.5 L L Remove need to wait
4.5. Press ‘Off’ button A8 Fail to press OFF button Cannot set finish time L L Label button FINISH TIME
4.6. Set finish time I1 No time entered No program recorded None L H Dial time in via analogue clock
I2 Wrong time entered Wrong program recorded None L H Dial time in via analogue clock
4.7. Set timer A8 Fail to set timer No program recorded None L H Separate timer selector from programming function
4.8. Press ‘Timer record’ button A8 Fail to press TIME RECORD button No program recorded None L H Remove this task step from sequence
5. Lift up front cover A8 Fail to lift up front cover Cover left down Immediate L L Remove cover to programming
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 73

Table 17
Using HTA in function allocation

Super-ordinate goal Subordinate goal Human or computer?

1.1. Forecast demand H

1.1.1. Review regular sales H
1.1.2. Review demand from pub chains H
1.1.3. Review potential demand from one-off events H
1.2. Produce provisional resource plan H–C
1.2.1. Calculate expected demand for each type of beer H–C
1.2.2. Make adjustment for production minima and maxima C

1.3. Check feasibility of plan H–C

1.3.1. Do materials explosion of ingredients H–C
1.3.2. Do materials explosion of casks and other packaging C
1.3.3. Check material stocks H–C
1.3.4. Calculate materials required C
1.3.5. Negotiate with suppliers H
1.3.6. Check staff availability H
1.3.7. Check ability to deliver beer to customers H
1.4. Review potential impact H
1.4.1. Review impact of plan on cash flow H
1.4.2. Review impact of plan on staff H
1.4.3. Review impact on customer relations H
1.4.4. Review impact on supplier relations H

machines. Marsden and Kirby argue that the most After the initial functional allocation, Marsden and
suitable system description is that provided by the sub- Kirby recommend a review of the potential impact of
goal hierarchy in HTA, because it focuses attention on the allocations to consider the likely impact on job
the purposes (i.e., goals and sub-goals) of the system in satisfaction, human error, attention, workload, produc-
question. They suggest that many of the function tivity, and cost effectiveness. An overall analysis of the
allocation problems can be circumvented by getting proposed allocation may also reveal any potential
the ‘stakeholders’ to agree upon the description of the conflicts or incompatibilities in the allocations. When
purpose of the system. As with Duncan’s (1972) the allocations have been confirmed, more detailed
comments about the neutrality of HTA, Marsden and analyses may be undertaken to propose how the sub-
Kirby propose that the analysis should present the sub- goals may be achieved with the proposed resources.
goals of what should be done by the system, rather than Kirwan and Ainsworth (1992) report on how HTA
how it should be done. The former will enable impartial may be used to assess the adequacy of interface design.
function allocation whereas the latter may bias function The example is based on a study of tasks in emergency
allocation. They also suggest that the stopping rule shut-down procedures on an off-shore oil and gas rig in
should be replaced with a no-solution heuristic, i.e., the the North Sea (Pennington, 1992). The analyses suggest
sub-goal decomposition stops at a point just before an that good communications between the production
allocation of function solution would become apparent. and the drilling teams play an important part in
This is to prevent premature function allocation. The maintaining safety. An extract of this analysis is
sub-goal hierarchy for the goal of ‘‘Checking the presented in Table 18. The analysis sought to investigate
desirability of meeting a potential increase in demand’’ the adequacy of the input, action, and feedback cycle in
is presented in the first two columns of Table 17. the tasks. This analysis harks back to the original
Marsden and Kirby (2005) outline a number of formulation of HTA proposed by Annett et al. (1971)
criteria to be considered when allocating system func- some 20 years earlier.
tions to humans or computers, some of which are As Table 18 shows, the sub-goal of ‘‘Formulate action
contradictory. This means that there is considerable to control the incident [y]’’ is criticised for lack of
discretion on the part of the analyst to resolve ‘‘keeping feedback, whereas the sub-goal of ‘‘Stop the drill’’ is
the job as simple as possible’’ and ‘‘having a challenging shown to have adequate feedback. In the original
job to do’’. The function allocation in Table 17 has been example, Pennington (1992) presented a more detailed
coded for human only (H), human and computer with analysis of communications in a separate table. The
the human in control (H–C), computer and human with advantage of the tabular format is that it permits new
the computer in control (C–H), and computer only (C). columns to be added to the analysis and it focuses the
 ARTICLE IN PRESS
74 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

analysis on each sub-goal in turn. This leads to an audit

should be provided to
Some direct feedback
trail of tables that can be checked, verified, acted upon,
and archived.
Annett et al. (2000) have shown how HTA can be

the drill team

used to analyse team tasks, arguing that team goals must
Comment

share common performance criteria (i.e., the team

product) which the success or failure of the team can
be judged against. Team processes normally stress the
importance of communication and co-ordination activ-
Adequacy of Feedback

ities. In their model of team processes, Annett et al.

proposed that the communication activities are likely to
include information sent and received and discussions
Unacceptable

Acceptable

Acceptable

Acceptable

Acceptable
between team members. The same model shows co-
ordination activities as the collaboration and synchro-
nisation of team members towards the team goals. The
example is based upon the analysis of the tasks of an
anti-submarine warfare team and an extract of the
available to drill team of
location of the incident

hierarchical sub-goal analysis is presented in Fig. 8.

RPM on digital and

Position indicators

Fig. 8 shows that the Anti-Submarine Warfare team

analogue displays

down ¼ drill up)

No information

have to simultaneously scan for threats and classify

Tactile (pedal

threats. They can either immediately classify the contact

Feedback

as a threat or investigate for local features (such as

Gauge

rocks, wreck, and pipelines) that could show as a

contact. If a threat is classified as high priority then the
team investigate the possibility that it is a submarine.
Production team assess
importance themselves

Annett et al. (2000) argued that it is important that

and if in drilling area

the task analysis captures the three principle compo-

contact drill team

nents of team work, namely the communication, co-

Turn handles
Turn handle

ordination activities as well as team goals. These

Method

Depress

Depress

components, and the corresponding activities, are

indicated in Table 19. They also analysed the compo-
nents of team work in a tabular form, as shown in
Tables 20 and 21.
Halt the progress of the
effective action as soon

Pick up drill string off

Table 20 shows that various members of the team are

Stop the rotary table
Inform appropriate

seeking data on contacts, activities being overseen by the

personnel to take

Shut off pumps

PWO. This activity is going on constantly. By contrast,

bottom of hole

Table 21 shows an activity that is triggered by discovery

as possible

of a new contact.
Purpose
Analysis of two sub-goals in emergency shut-down procedures

The team work described in Table 21 is rather more

drill

complicated than that in Table 20. This information could

have been expressed at deeper levels in the sub-goal
hierarchy, but the tabular format allows for the complexity
Brake on foot control

of the activity to be captured in a narrative form.

All of the analyses show specialised enhancements of
Clutch pedal

the basic HTA tabular format to perform specific

analyses, such as of human computer interaction, error
Handles
Control

Handle

analysis, allocations of function, and team tasks. It is

N/A

possible to combine these analyses into one large table,

as shown in Table 22. Gramopadhye and Thaker (1998)
the case of an emergency

illustrate the multiple column format that can be used to

control the incident in
Formulate action to

record several forms of analysis within the same

document. The table in this example of PC operation
Stop the drill

has been split into two parts because there were 16

shut-down
Sub-goals

analysis columns to support.

Table 18

As Table 22 shows, the analysis can be very

comprehensive, but the format may be unwieldy. It
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 75

might be difficult to perform all of the analyses as part commun.) has considered developing HTA into dy-
of a single study. Rather there are likely to be separate namic programmable models that may be used to
studies for allocation of function, interaction, error evaluate the performance of a system. He proposed
analysis, and knowledge requirements. Then each of that the methodology could be used for system design,
these separate analyses could be compiled into a single performance estimation, and allocation of system
table. What Table 22 does show is an overview of the function. Both of these approaches represent a depar-
task demands and constraints in one single source. ture from the table and taxonomy approaches that have
There may be occasions when all of this information is been used in contemporary extensions of HTA.
needed together.
Other extensions of HTA have considered the
possibility of modelling task scenarios. Baber and
6. Future requirements for software support of HTA
Stanton (1999) have proposed a method that combined
HTA with state-space diagrams and transition matrices
Attempts have been made to develop software
to model human–computer interaction as part of an
support for HTA, with varying degrees of success but
analytical prototyping process. The TAFEI (Task
none supports the full range of applications to which
Analysis For Error Identification) methodology com-
HTA may be put. The software tool of Bass et al. (1995)
bines scenario analysis, structural analysis and func-
was only developed to prototype form, to simplify the
tional analysis to test device design prior to development
production of the hierarchical diagrams and the tabular
of an operational prototype. Annett (2004, pers.
format by allowing direct manipulation of the data
objects and easy editing of the analysis. Other examples
have been developed for specific applications, such as
error prediction or workload analysis. Some analysts
have used outline processors, organisational charting,
1.1 and planning tools such as More, OrgPLUS, and
Identify Inspiration. Their tools tend to support some, but not
threats all, aspects of HTA. It is, therefore, proposed that any
[1 + 2] future software support for HTA should support the
wide range of applications, and needs to combine four
principal facets of HTA use:

1.1.1 1.1.2 (i) Support the development of the sub-goal hierarchy

Scan for Classify and plans in the three different formats of HTA
threats threats representation.
[1/2>3] (ii) Enable editing and verification of the analysis to
percolate through each of the representations.
(iii) Support extended analysis of the sub-goal hierar-
chy.
(iv) Enable further extensions of the analysis to be
1.1.2.1 1.1.2.2 1.1.2.3 added.
Classify Check chart Investigate
immediate for known possible
threat feature submarine
Each of these requirements will be dealt with in turn.
First, the software should support development of the
sub-goal hierarchy and plans. Examples of templates for
Fig. 8. Hierarchical diagram for the goal of ‘‘Identify threats’’. development of the hierarchy and plans have been

Table 19
Components of team processes

Team process Category Observable activities

Communication Send information Transmit data or comment to another party

Receive information Receive data or comment to another party
Discussion Discuss situation and/or options with other team members
Co-ordination Collaboration Share or rearrange work according a plan
Synchronisation Keep to planned time or event schedule
 ARTICLE IN PRESS
76 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

Table 20
Team work to identify and classify anti-submarine contacts

Criteria Description of criteria

1.1. Identify and classify all anti-submarine warfare contacts

Measure Contacts are identified and classified as quickly and accurately as possible
Teamwork Team compiles information from various sources. Principal Warfare Officer (PWO) monitors and directs the team
Plan [1.1.1.+1.1.2] Scanning all sources for information on potential threats is continuous [1.1.1.] classification procedures
[1.1.2.] follows identification as soon as possible

Table 21
Team work to check the chart

Criteria Description of criteria

1.1. 2.2. Check chart for known feature, such as rock, wreck, pipeline, etc.
Measure Chart checking procedures should be executed correctly and the conflicts resolved. All units should be informed of outcome.
Teamwork The sonar operator passes the information onto the Active Sonar Director who confers with the PWO. The PWO calls for a
‘‘Chart check, poss. sub BRG/RG’’ The Officer of the Watch plots the position to agreed margin of error. The Active Sonar
Director directs the Sonar Controller to investigate the location. The Action Picture supervisor inputs the data into the system.
The Electronic Warfare and Radar teams check returns on that bearing. The Officer of the Watch and Missile Director check
the bearing visually. All report the results of their checks.
Plan If the chart check is negative then go to Respond to Threats [1.2]. If the information is inconsistent the go to Investigate
Possible Submarine [1.1.2.3].

proposed by Ormerod and colleagues (see Tables 6 add symbols and taxonomies or elements also needs to
and 7) and also in Table 10. The questions to be be provided for. This is to allow maximum flexibility for
addressed in development of sub-goals are presented in the analyst. This means that after the sub-goal hierarchy
Table 5. Development of task statements was presented and plans have been constructed, the HTA can be
in Table 8. The software could support each aspect of subjected to further analysis, such as error potential,
the sub-goal and plan development following through allocation of function, and team work.
the stages outlined in Fig. 5. Each form of representa- Finally, the software should, as far as is reasonably
tion should interact with the others, so that as the practicable, enable further extensions to be added. A
hierarchical diagram (see Fig. 6) is developed, so is the simple means of adding additional functionality would
hierarchical list (see Fig. 7) and the tabular format (see be to allow the analyst to create their own templates for
Table 10). The development of the plans should be plans, tabular forms, taxonomies, and symbols. A more
automated as far as possible, using the templates as complex means of allowing additional functionality
suggested previously. would be to leave the software architecture open to
Second, as HTA involves reiteration and revision, the additional development so that different approaches,
software tool should enable the sub-goal hierarchy and such as TAFEI (Baber and Stanton, 1994) and Annett’s
plans to be edited with ease. Any change made in one (2004, pers. commun.) dynamic programmable task
form of the representation should propagate through models can be added at a future date.
the other forms of the representation. For example, a
change in the sub-goal hierarchy or in the plans in the
hierarchical diagram should work through to the 7. Some general conclusions
hierarchical list and tabular format automatically, and
vice versa. As far as possible, the editing of the sub-goal The future for HTA seems assured, at least in the
hierarchy and plans should be possible though direct short to medium term. The variety of domains and
interaction with the objects and text on the screen. applications that it has been used for is a testament to its
Third, the software should support extended analysis usefulness. The developments and extensions of the
of the sub-goal hierarchy and plans, as shown in (but approach suggest that it is likely to remain in the core
not limited to) the examples earlier in this paper. repertoire for ergonomists. HTA should also serve as a
Templates of the tabular formats, with their associated benchmark for all other ergonomics methods and
symbology and taxonomies, would need to be provided. approaches. The key features of the approach are that
The facility to edit the templates or remove columns and it was not only developed on strong theoretical
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 77

Failure to locate
foundations but also focused on solving real-world

Task criticality
Possible errors

Specify wrong

issuing wrong
problems. The approach was flexible enough to enable it

directory or

command
to be applied to a wide variety of domains and

switch
applications. It has continued to be developed, ex-

High

High
tended, and improved throughout the past 34 years, but
the original three guiding principles remain as true today
as when they were first put forward. One reason for the
endurance of HTA is that it can provide a comprehen-

Complexity
sive model of a sub-goal hierarchy in a system. This
Cognitive
demand

Simple

Simple
model could be of an existing system or one that is
Low

Low

anticipated. The sub-goal hierarchy lends itself to all

manner of analyses, which is the real point of HTA.
HTA was never meant to be the end point in the
analyses, just the start. The original aims of HTA were
quite modest. The authors of the first report hoped that
Coordination

it would spread the ideas of ‘‘a new approach to tasks

Skill level

analysis even more widely in British Industry’’ (Annett

Medium
None

None

et al., 1971, p. iii). It would be fair comment to say that

Low

they have exceeded these aims many-fold, to provide

ergonomists in industry and academia throughout the
world with a core approach to systems analysis of sub-
Computer input

Basic operation

Basic operation

goals.
Knowledge

of a PC

of DOS
N/A

N/A

Acknowledgements
Update switch

The author of this paper gratefully acknowledges the

Human input

helpful comments from Professor John Annett, Dr. Don

command
directory
position

Change

Harris and Dr. Karen Lane on an earlier version. This

Clerk

Clerk
Who

work from the Human Factors Integration Defence

Technology Centre was part-funded by the Human
Sciences Domain of the UK Ministry of Defence
Info. presented

Scientific Research Programme. The report on which

Switch label

this paper is based was used to develop a software tool

Once a day

Once a day
Frequency
directory

for HTA which is available free-of-charge from

Current

www.hfidtc.com.
Position of ON/

References
Task duration
Info. required

OFF switch

Ainsworth, L.K., Marshall, E., 1998. Issues of quality and practic-

o1 min

o1 min
Human

ability in task analysis: preliminary results from two surveys.

Ergonomics 41 (11), 1607–1617 (Reprinted in: Annett, J., Stanton,
N.A. (Eds.), 2000. Task Analysis. Taylor & Francis, London, pp.
79–89).
Annett, J., 1996. Recent developments in hierarchical task analysis. In:
Unable to run

Unable to run
Consequences

Robertson, S.A. (Ed.), Contemporary Ergonomics 1996. Taylor &

the system

the system
Allocation

Francis, London, pp. 263–268.

Human

Human

Annett, J., 2000. Theoretical and pragmatic influences on task analysis

methods. In: Schraagen, J.M., Chipman, S.F., Shalin, V.L. (Eds.),
Cognitive Task Analysis. Lawrence Erlbaum Associates, Mahwah,
HTA for work design

NJ, pp. 25–37.

Annett, J., 2004. Hierarchical task analysis. In: Diaper, D., Stanton,
1.1.1. Place ON/

1.1.1. Place ON/

OFF switch in

OFF switch in
RTS directors

RTS directors

N.A. (Eds.), The Handbook of Task Analysis for Human–Com-

1.1.2. Access

1.1.2. Access
ON position

ON position

puter Interaction. Lawrence Erlbaum Associates, Mahwah, NJ, pp.

Table 22

67–82.
Task

Annett, J., Duncan, K.D., 1967. Task analysis and training design.
Occup. Psychol. 41, 211–221.
 ARTICLE IN PRESS
78 N.A. Stanton / Applied Ergonomics 37 (2006) 55–79

Annett, J., Stanton, N.A., 1998. Research and developments in task action. Lawrence Erlbaum Associates, Mahwah, NJ,
analysis. Ergonomics 41 (11), 1529–1536 (Reprinted in: Annett, J., pp. 347–365.
Stanton, N.A. (Eds.), 2000. Task Analysis. Taylor & Francis, Ormerod, T.C., Richardson, J., Shepherd, A., 1998. Enhancing the
London, pp. 1–8). usability of a task analysis method: a notation and environment for
Annett, J., Stanton, N.A. (Eds.), 2000. Task Analysis. Taylor & requirements specification. Ergonomics 41 (11), 1642–1663 (Rep-
Francis, London. rinted in: Annett, J., Stanton, N.A. (Eds.), (2000). Task Analysis.
Annett, J., Duncan, K.D., Stammers, R.B., Gray, M.J., 1971. Task Taylor & Francis, London, pp. 114–135).
Analysis. Department of Employment Training Information Paper Patrick, J., Spurgeon, P., Shepherd, A., 1986. A Guide to Task
6. HMSO, London. Analysis: Applications of Hierarchical Methods. An Occupational
Annett, J., Cunningham, D., Mathias-Jones, P., 2000. A method for Services Publication, Birmingham.
measuring team skills. Ergonomics 43 (8), 1076–1094. Pennington, J., 1992. A preliminary communications systems assess-
Astley, J.A., Stammers, R.B., 1987. Adapting hierarchical task analysis ment. In: Kirwan, B., Ainsworth, L.K. (Eds.), A Guide to Task
for user–system interface design. In: Wilson, J.R., Corlett, E.N., Analysis. Taylor & Francis, London, pp. 252–265.
Manenica, I. (Eds.), New Methods in Applied Ergonomics. Taylor Piso, E., 1981. Task analysis for process-control tasks: the method of
& Francis, London, pp. 175–184. Annett et al. applied. Occup. Psychol. 54, 347–352.
Baber, C., Stanton, N.A., 1994. Task analysis for error identification: a Shepherd, A., 1976. An improved tabular format for task analysis.
methodology for designing error-tolerant consumer products. Occup. Psychol. 49, 93–104.
Ergonomics 37 (11), 1923–1941. Shepherd, A., 1989. Analysis and training in information technology
Baber, C., Stanton, N.A., 1999. Analytical prototyping. In: Noyes, tasks. In: Diaper, D. (Ed.), Task Analysis for Human–Computer
J.M., Cook, M. (Eds.), Interface Technology: The Leading Edge. Interaction. Ellis Horwood, Chichester, UK, pp. 15–55.
Research Studies Press, Baldock. Shepherd, A., 1998. HTA as a framework for task analysis.
Baber, C., Walker, G., Salmon, P., Stanton, N.A., 2004. Obser- Ergonomics 41 (11), 1537–1552.
vation study conducted at the fire service training college. Shepherd, A., 2001. Hierarchical Task Analysis. Taylor & Francis,
Human Factors Integration Defence Technology Report (unpub- London.
lished). Stammers, R.B., 1996. Hierarchical task analysis: an overview. In:
Bass, A., Aspinall, J., Walters, G., Stanton, N.A., 1995. A Software Jordan, P.W., Thomas, B., Weerdmeester, B.A., McClelland, I.L.
Toolkit for Hierarchical Task Analysis. Appl. Ergon. 26 (2), (Eds.), Usability Evaluation in Industry. Taylor & Francis,
147–151. London, pp. 207–213.
Bruseberg, A., Shepherd, A., 1997. Job design in integrated mail Stammers, R.B., Astley, J.A., 1987. Hierarchical task analysis: twenty
processing. In: Harris, D. (Ed.), Engineering Psychology and years on. In: Megaw, E.D. (Ed.), Contemporary Ergonomics 1987.
Cognitive Ergonomics. Job Design and Product Design, vol. 2. Taylor & Francis, London, pp. 135–139.
Ashgate Publishing, Aldershot, Hampshire, pp. 25–32. Stanton, N.A. (Ed.), 1996. Human Factors in Nuclear Safety. Taylor
Chapanis, A., 1951. Theory and methods for analysing errors in & Francis, London.
man–machine systems. Ann. NY Acad. Sci. 51 (6), 1179–1203. Stanton, N.A., 2003. Human error identification in human computer
Duncan, K.D., 1972. Strategies for analysis of the task. In: Hartley, J. interaction. In: Jacko, J., Sears, A. (Eds.), The Handbook of
(Ed.), Strategies for Programmed Instruction: An Educational Human–Computer Interaction. Lawrence Erlbaum Associates,
Technology. Butterworths, London. Mahwah, NJ.
Gilbreth, F.B., 1911. Motion Study. Van Nostrand, Princeton, NJ. Stanton, N.A., 2004. The psychology of task analysis today. In:
Gramopadhye, A., Thaker, J., 1998. Task analysis. In: Karwowski, Diaper, D., Stanton, N.A. (Eds.), The Handbook of Task Analysis
W., Marras, W.S. (Eds.), The Occupational Ergonomics Hand- for Human–Computer Interaction. Lawrence Erlbaum Associates,
book. CRC Press, Boca Raton, FL, pp. 297–329. Mahwah, NJ, pp. 569–584.
Hackman, J.R., Oldham, G.R., 1980. Work Redesign. Addison- Stanton, N.A., Young, M.S., 1999. A Guide to Methodology in
Wesley, Mass. Ergonomics. Taylor & Francis, London.
Hellier, E., Edworthy, J., LEE, A., 2001. An analysis of human error in Staples, L.J., 1993. The task analysis process for a new reactor. In:
the analytical measurement task in chemistry. Int. J. Cogn. Ergon. Proceedings of the Human Factors and Ergonomics Society 37th
5 (4), 445–458. Annual Meeting—Designing for Diversity, Seattle, WA, October
Hodgkinson, G.P., Crawshaw, C.M., 1985. Hierarchical task analysis 11–15, 1993. The Human Factors and Ergonomics Society, Santa
Monica, CA, pp. 1024–1028.
for ergonomics research. An application of the method to the
Taylor, F.W., 1911. Principles of Scientific Management. Harper &
design and evaluation of sound mixing consoles. Appl. Ergon. 16
Row, New York.
(4), 289–299.
Wilson, J.R., Corlett, E.N. (Eds.), 1995. Evaluation of Human Work.
Kirwan, B., Ainsworth, L.K. (Eds.), 1992. A Guide to Task Analysis.
Taylor & Francis, London.
Taylor & Francis, London.
Marsden, P., Kirby, M., 2005. Allocation of functions. In: Stanton,
N.A., Hedge, A., Brookhuis, K., Salas, E., Hendrick, H. (Eds.),
Handbook of Human Factors and Ergonomics Methods. Taylor &
Francis, London, in press.
Miller, C.A., Vicente, K.J., 2001. Comparison of display require- Further reading
ments generated via hierarchical task and abstraction–decomposi-
tion space analysis techniques. Int. J. Cogn. Ergon. 5 (3), Annett, J., 2003. Hierarchical task analysis. In: Hollnagel, E. (Ed.),
335–355. Handbook of Cognitive Task Design. Lawrence Erlbaum Associ-
Miller, G.A., Galanter, E., Pribram, K.H., 1960. Plans and the ates, Mahwah, NJ, pp. 17–35.
Structure of Behaviour. Holt, New York. Carey, M.S., Stammers, R.B., Astley, J.A., 1989. Human–
Ormerod, T.C., Shepherd, A., 2004. Using task analysis for computer interaction design: the potential and pitfalls of hierarch-
information requirements specification: the sub-goal template ical task analysis. In: Diaper, D. (Ed.), Task Analysis for
(SGT) method. In: Diaper, D., Stanton, N.A. (Eds.), The Human–Computer Interaction. Ellis Horwood, Chichester, UK,
Handbook of Task Analysis for Human–Computer Inter- pp. 56–74.
 ARTICLE IN PRESS
N.A. Stanton / Applied Ergonomics 37 (2006) 55–79 79

Diaper, D., Stanton, N.A. (Eds.), 2004. The Handbook of Task temporary Ergonomics 1989. Taylor & Francis, London, pp.
Analysis for Human–Computer Interaction. Lawrence Erlbaum 36–43.
Associates, Mahwah, NJ. Richardson, J., Ormerod, T.C., Shepherd, A., 1998. The role of task
Kirwan, B., 1994. A Guide to Practical Human Reliability Assessment. analysis in capturing requirements for interface design. Interact.
Taylor & Francis, London. Comput. 9 (4), 367–384.
Kirwan, B., Reed, J., 1989. A task analytical approach for the Stammers, R.B., Shepherd, A., 1990. Task analysis. In: Wilson, J.R.,
derivation and justification of ergonomics improvements in Corlett, E.N. (Eds.), Evaluation of Human Work: a Practical Ergono-
the detailed design phase. In: Megaw, E.D. (Ed.), Con- mics Methodology, second ed. Taylor & Francis, London, pp. 144–168.