This article was downloaded by: [University of Sussex Library]

On: 09 February 2015, At: 12:12

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Ergonomics
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/terg20

Relations between task and activity: elements for

elaborating a framework for error analysis
a
J. LEPLAT
a
Ecole Pratique des Hautes Etudes (Dept. No 3) , Laboratoire de Psychologic du Travail,
CNRS GDR 111, 41, rue Gay-Lussac, 75005, France
Published online: 20 Jan 2011.

To cite this article: J. LEPLAT (1990) Relations between task and activity: elements for elaborating a framework for error
analysis, Ergonomics, 33:10-11, 1389-1402, DOI: 10.1080/00140139008925340

To link to this article: http://dx.doi.org/10.1080/00140139008925340

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 ERGONOMICS, 1990, VOL. 33, NOS IOfll, 1389-1402

Relations between task and activity:

elements for elaborating a framework for error analysis

J. LEPLAT
Ecole Pratique des Hautes Etudes (Dept. No 3),
Laboratoire de Psychologie du Travail, CNRS GDR III,
41, rue Gay-Lussac, 75005 Paris, France

Keywords: Human error; Error modelling; Task analysis; Driver behaviour.

Downloaded by [University of Sussex Library] at 12:12 09 February 2015

The notion of error, when applied to an activity or the result of an activity, implies
the notion of task: it expressesthe deviation between the activity and the task being
consideredfrom an angle whichisjudged to be relevant.The task and the activityare
the object of representations for the analyst (or specialist)and for the driver. Four
representations are dealt with in this paper: the task and the activity for the
specialist and the task and the activity for the driver. An interpretation is proposed
for these tasks,and they are illustrated using some of the work already carried out in
this field. The signification of deviations between these representations is then
discussed, together with the advantage of studying these deviations in order to
clarify error-producing mechanisms. Analysis in terms of task and activity raises
methodological and practical problems whichare touched upon; it does not exclude
referring to psychological theoretical frameworks to which it is worthwhile linking
it. This perspective raises questions which make it possible to enhance the study of
errors: it could be completed at a later date by extending it to include other
representation categories.

1. Introduction
The notion of error when applied to an activity or its result, implies the notion of task. It
expresses a deviation between the task to be carried out and the activity really
implemented by the individual to accomplish this task. Error analysis therefore refers
back to the task analysis and activity analysis: this paper will be devoted to showing
that by acquiring a better knowledge of the relationship between these two analysis
categories it is possible to better understand the mechanisms which produce the error.
To do this, we shall first define the notion of task and activity; polysemantic words of
uncertain theoretical status. This first analysis leads us to differentiate between different
task categories, and the deviations between these tasks constitute a similar number of
guidelines for error analysis. The task and the activity involved in the task, can each be
represented from the viewpoint either of the analyst (or specialist) or from that of the
driver. This gives rise to four representations for a single event, the task seen from the
point of view of the analyst, the task seen from the point of view of the driver, the
activity seen from the point of view ofthe analyst, and the activity seen from the point of
view of the driver. An attempt will be made to show that the proposed analysis schema
can be implemented by using psychological studies of driving behaviour. Lastly, some
of the problems raised by this method of activity and error analysis will be examined,
and we shall indicate the correlations that exist with more traditional methods of
analysis. This paper will therefore be of interest from a methodological point of view. It
suggests perspectives which remain to be studied and which are outlined in certain
work analyses. The relevance of these in the study of driving errors is still to be tested
more systematically.
0014-0139/90 S3-00 © 1990 Taylor & Francis Ltd.
 1390 J. Leplat

2. Representations of the task as activity models

It is essential to differentiate between the notion of task and activity. A task can be
defined as a goal to be attained within set conditions. This definition was initially put
forward by Leontiev (1972), but later taken up with some variations by several other
authors who were apparently unaware of its original source. Activity is the response
that the individual gives to a task that has been imposed on him or that he wishes to
carry out. When applied to work situations, where these distinctions are often used, the
notion of a task, prescribed by the organization or by a hierarchical superior, is often
raised. It underlines the fact that the activity to be carried out in a work situation has
been finalized, or, as referred to be certain authors, 'goal-oriented'.
The task can be defined in several ways (Leplat and Hoc 1983). It can be done so in
terms of goals or sub-goals, and the idea of goal structure is often referred to. It can be
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

defined by the execution conditions to be taken into account: technical and

organizational rules. It can also be defined by the procedure to be implemented. All
these definitions are more or less complete, more or less in-depth, but finally part of
them is always implicit.
Several task representations will be differentiated (there are other possibilities), and
first of all it will be shown how they can be-or have been-studied in the field of
driving: these representations refer respectively to the task to be carried out, the task to
be redefined, the actual task, the actual task for the subject. In more banal terms, it can
be said that they correspond to the following questions. What must be done? What
does the driver want to do? What did the driver really do? What did the driver think he
or she did? These deviations beween the different representations categories will be
used to analyse the error and its origin.
The analysis framework that is proposed is of a methodological nature: it is not
connected to any particular theory, but does leave room for different types of studies,
which could be coordinated at a later date.

2.1. The task 10 be carried out, as a first reference in the definition of the error
The notion of prescribed task, familiar in work psychology, has no equivalent in
driving, except for professional drivers, for whom driving is an essential aspect of the
task ascribed to them. For other drivers, the prescriptions are derived from more
general tasks prescribed by external factors (having to be at work on time), or by those
that are self-imposed (going on holiday to a specified place). It was therefore decided to
replace 'prescribed task' by 'task to be carried out'. This can be interpreted as the model,
for the expert, of the task to be carried out. The driving task is first of all defined as a
combination of conditions which indeed have the status of prescriptions, as they are set
independently of the driver. It is possible to differentiate between:
• technical conditions: relating to the characteristics of the vehicle, the infra-
structure and their interactions (e.g., the effect of braking on a certain type of
road);
• general environmental conditions: driving at night, fog, rain, etc.;
• traffic conditions: the presence of other cars on the road, pedestrians, etc.;
• regulatory conditions: as defined in the Highway Code, in a more or less
operational manner.
These conditions are explained verbally to the driver (rules set out in the Highway
Code), imposed upon him or her (climatic conditions), or are left for him or her to
discover (the dynamic characteristics of the vehicle).
 Task, activity, and error analysis 139\

The driving task can also be defined by the procedures it requires the driver to carry
out. The procedure specifies the actions to be carried out, in the conditions likely to
occur. This is done more or less in depth. The more the procedure is sketchily set out,
the greater the skill required on the part of the driver to carry out the task. The
procedures are made explicit when learning to drive. They are also made explicit when
an accident is studied, or when an expert can retrospectively determine the conditions
of the task and the ways they can be processed.
Examples of this were proposed by Girard (1989) and Hale et al. (1988) for the task
of crossing controlled intersections. Here is an example, taken from Girard, of several
items contained in the procedure which was drawn up:
7a: Go to the cenral reservation area and stop the car; go on to 8a.
8a: Look to the right:
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

(a) if there is nothing coming, gap D is sufficient; go on to lOa;

(b) if a vehicle has been observed, go to 9a.
9a: Evaluate gap D:
(a) If gap D is too short, wait until the vehicle has crossed the intersection, and
go back to 8a;
(b) If gap D is sufficient, go on to lOa. [etc.]

There will be often be several possible procedures, especially when the task is complex.
It will always be of interest to define the criteria of choice, the relationships, the possible
known parts.
The definition of the task to be carried out, which is indispensable when defining
error, occurs in different situations: at the time of training, when accidents are being
reconstructed and, finally, when using accident data. In the last instance, it is obviously
important to be able to determine the task categories, in order to assign risk
coefficients, so as to be able to analyse the sources of this risk and envisage correct
measures to reduce it.
A certain number of task categorizations based on principles which are varied and
more or less clearly explained, can be found in the literature. Lourens (n.d.) has talked
about, and criticized some of these attempts, and has disputed the relevance of the level
of categorization. This led him to propose 14 categories which 'clearly indicate our
intended manoeuvre'. This categorization was intended for use during training.
Malaterre and Peytavin (1985), using accident analysis, proposed a categorization
based on multiple criteria, which takes into account the direction of the traffic, the type
of action and avoidance carried out, depending on whether they are of a differentiating
nature or not (cf', also Malaterre 1990).

2.2. The redefined task

This is the task the driver sets him or herself. It corresponds to the question 'What do
you want to do?' This question can be put in different ways: what do you want to do in a
specific situation, or in one particular type of situation? The driver is expected to reply
by stating the goals and sub-goals of his or her action, the conditions he or she proposes
to take into account, the procedure he or she is prepared to use: the questions could be
directed towards one or other ofthese aspects. This mental representation of the task to
be carried out is only relatively clear in the driver's mind. Thus, in the case of an
experienced driver, automatic reactions play an important part, and he or she would
find it hard to describe them. For the expert, the way he or she describes his or her task
 1392 J. Lep/at

would, in fact, be a way of evaluating his or her level of automatic reaction as shown by
Leontiev (1972, 1975).
As a constituent part of the redefined task, it is worthwhile making explicit the
classification of the driver's goals, as it determines a certain number of activity
limitations. In this way, the general goal 'getting there before such a time' implies time
limitations which have a direct influence on the activity, as is often shown in the study of
accidents. There can also be several goals which are more or less compatible and which
can be the cause of conflict. Thus, the need to arrive before a given time, reduce petrol
consumption, minimize work load, can be goals which generate conflict in the sense
that carrying them out cannot be jointly maximized and forces the driver to make
compromises which correspond to a differential weighting of these goals. It is useful to
know this weighting (e.g. what is the weight given to risk-taking in relation to arriving
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

late?) when analysing the activity.

Literature on this subject contains a certain number of attempts which have been
made to deal with the redefined task, the model of the task the driver has decided to
carry out. One ofthe ways this can be analysed is by evaluating the driver's knowledge
of driving conditions. There are also studies on knowledge of signposting, knowledge of
the vehicle (the way it functions, its dynamic capacities). Research can be carried out on
the driver's knowledge of general cases, or of specfic tasks during which this knowledge
is implemented.
It is possible to obtain information on the principles that govern carrying out the
tasks and, here again, for specific tasks or for categories of tasks. An excellent example
of this was provided by Moget-Monseur and Biecheler (1984) who defined a certain
number of basic activities and asked the driver to note whether he or she carried out
each activity or not and, ifhe or she answered in the affirmative, how often, and in which
case. Here are several items (p. 12) which serve as an example:
• overtaking when it is thought to be 'a close thing';
• exceeding the speed limit in daylight;
• accelerating to get over an intersection when the traffic lights have just turned to
yellow;
• crossing a continuous white line.
The statistical treatment of the collected data makes it possible to identify the types of
patterns which can be considered as the overall principles characterising the task
redefined by the driver.
It can be noted that the notion of redefined task has recently been used by authors
who are interested in computerising certain tasks [cf.,e.g., Sebillotte 1988).The way in
which the operators expressed the tasks they had to carry out was analysed in terms of
abstraction level of representation, and the planned nature of this representation was
noted. It is indeed this redefined task that cognitive scientists interested in the design of
expert systems try to express.
If verbalization is a privileged way of gaining access to the redefined task, it is not
the only way, as is shown by psychological work analysis methods (Leplat and Cuny
1984).
The redefined task expresses the way in which the driver plans his or her activity
and the way he or she sets up its organization and its sequence. It could be characterized
by, in particular, its temporal span. This temporal span looks to the past, i.e., the
conditions governing past activity which are taken into consideration ('I have been
behind this lorry for a long time and I want to overtake it') and how far this situation
 Task, activity, and error analysis 1393

goes back into the past, but it also concerns the future, that is to say taking into account
the conditions which may arise in the near or not so near future ('After this bend the
road gets bad, I must slow down'). Here we come across the problem of anticipation,
which is one of the most important aspects of driving. The temporal span of task
redefinition raises the problem of actualizing the redefinition in relation to the
conditions actually encountered, whilst at the same time, dealing with the problem of
the more or less adaptable (vs rigid) nature of this redefinition.
Determination of the redefined task for accident analysis raises specific problems.
Tndeed, the driver's responses cannot easily be independent of the action he or she has
actually undertaken, and it is unlikely, in particular, that he or she will infringe the
regulations set out in the Highway Code. This limitation, even if it must remain ever-
present in the mind of the analyst, should not discourage the driver from searching for
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

the redefined task. The way he or she describes how the task should have been carried
out is always a source of extremely useful information on the mechanisms that govern
driving.

2.3. The actual task as a pivot in error analysis

The driver's activity represents the actual task; it does not necessarily correspond to the
task to be carried out. The expert or the analyst will set up a model which, for the sake of
con venience, will also be termed actual task.
The actual task is the task that should have been given to the driver to obtain the
activity, if the driver had carried out the task correctly. Although the task to be carried
out precedes and orientes the activity, the actual task is inferred when studying the
activity. This inference also makes use of the knowledge of the task to be carried out,
and the redefined task, together with psychological and technical knowledge acquired
from other sources. Whereas, in conventional laboratory studies, the drawing up of the
activity model is the basis for the design of the experimental situation where the activity
will be carried out, here the activity model is drawn up-inferred-using this activity
and this is what, for the analyst, will be termed 'actual task' or activity model. This is
one of the main characteristics of a field study compared with the laboratory study.
This model can be more or less precise or detailed, and it can only form an
approximation to the activity. A model such as this always supposes a certain skill on
the part of the driver.
There is no point in dwelling upon this drawing up of the actual task studied in
work psychology [cf., e.g., Leplat and Cuny 1984): this work is based on many models,
some of them extremely conventional. It should always be based on very careful
behaviour observation. Previous examples of driving situations (e.g., Quenault 1968),
show the importance of these observations. The identification of traffic conflicts
(Muhlrad 1988) is another example of systematic observation.
The reconstitution of the actual task, when it has not been possible to observe it
directly, raises certain specific problems, dealt with in clinical accident studies,
(Malaterre and Peytavin 1985) which emphasize the importance of using traces
collected after the accident, and which require considerable technical skill.

2.4. The driver's representation of the activity carried out

This is also what could be termed the actual task for the person who carried it out. It is
the model ofthe activity carried out, drawn up by the driver. The expression of this task
raises methodological problems which are discussed in studies on consecutive
verbalization (protocol analysis). The driver's representation of the task carried out will
 1394 J. Lep/at

be one of the factors used by the analyst to reconstitute the accident situation.
The study of the representation the driver makes of his or her own activity could be
clarified by certain studies carried out within the framework of attribution theories (cf.,
e.g., Jones et at. 1972), theories which attempt to determine the mechanisms which
control the attribution made by subject of the causes of his or her own actvity.

3. Representation deviations as factors for the analysis of driving errors

After defining the different task representation categories we shall examine how they
can be used to help in erroranalysis. To achieve this, we shall rely essentially on the
study of the deviations between the different representation categories. Figure I shows
the deviations which will be given particular consideration, with their general
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

significance. The 'deviation concept' (Kjellen 1984)and 'reasoning through deviations'

(Crozier and Friedberg 1977) are in fact particularly relevant in showing up critical
points in the carrying out of tasks and therefore in the origin of errors. Any deviation
between a pair of representations forms a question put to the analyst. This search for
the reply leads to the examination of the way in which representations are drawn up
and their relationship with the activity. Relationships between these representations
are not simple, as some of them are conditioned reciprocally. This is why the driver's
representation of the task to be carried out depends, when he or she has already carried
out a similar task, on the latter's representation. Inversely, the representation of the
carrying out of a task is not independent from the definition the subject has of his task.
The main object of the analysis comprises what we have termed the error for the
expert. The deviations between the different representations are to be compared to this
error, and t heir interpretation will lead to hypotheses on the sources of error. By going
back to the representations of the action, an attempt will be made to clarify the
mechanisms and understand the way in which the error was produced. The error forms
a deviation category which should be shown in relation to other deviation categories
which contribute to its production. A policy of prevention or reliability will be
instrumental in reducing these deviations.

3.1. The basic error: the error for the expert

The error for the expert is the deviation between the actual task corresponding to the
activity actually carried out ..and the task to be carried out. It has been said that the task
can be defined in different ways, more or less precisely and in-depth. The same applies
to the error or deviation between two tasks. The error can relate to a general goal, as
indicated in the Highway Code: e.g., 'Always remain in control of the speed of your
vehicle'. It may also relate to the conditions to be taken into consideration when
carrying out the task and which have not been considered: (e.g., not respecting
regulations with regard to a stop sign of a continuous white line) which corresponds to
the idea of a traffic offence (cf., Reason et at. 1990).
It may also relate to the procedure which is implemented. Once the specialist has
defined the procedure to follow when processing a given situation, he or she may
consider any deviation from this procedure to be an error. In this way there will be as
many error categories as ways of task characterization. The notion of driving error
must always be specified by what it refers to: error in relation to what? The error
therefore depends on both the way the task to be carried out is defined, and the way the
actual task, based on the activity, is characterized.
 Downloaded by [University of Sussex Library] at 12:12 09 February 2015

Sources of errors Deviation in the task: design Deviation in the execuaon Deviation in the evaluation

~
to
Types of tasks r:
'~"
§:
9
Error for the expert ...::s'"
Types of errors

Error for the driver ...~

::s
'"
Figure l. The fourtypes of representations '"
~
0:;'

~
CD symbol for the difference between a and b, c: result of the difference.

w
'C
V>
 1396 J. Leplat

An underlying problem for any study of the error, is how important should a
deviation be before it is considered to be an error. This is sometimes interpreted as an
error acceptability limit, when a deviation error is identified.
This problem is clearly raised and discussed in work dealing with traffic conflicts
where safety scales have been drawn up (Muhlrad 1988). It is possible to more or less
deviate from the task to be carried out, and it will often be of use to have a scale
denoting the seriousness of resulting errors.
When a procedure has been defined (Girard 1989), it is of particular interest to pick
out the stages in the procedure when activity deviates. For a given driver, these
deviations constitute factors for accident or near-accident diagnosis. Used
statistically, they make it possible to identify the characteristics of the situation which
constitute risk factors (e.g., because they make it difficult to carry out the procedure).
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

This method is only justified insofar as there is a certain compatibility between the
task to be carried out and the task which has been carried out.

3.2. Deviation between the task to be carried out and the redefined task
This deviation can be defined in as many ways as a task may be defined. Three large
interpretation categories are possible:
(I) M isappreciation or incomprehension ofthe task to be carried out. This suggests
that the task has been badly explained or the subject is not sufficiently skilled to
understand the requirements, as they have been presented. For example, the
driver thinks, wrongly, that he or she has right of way at an intersection.
(2) The level of explication. Deviations noted between the two tasks sometimes
result from a level different from the level at which they were explained. There is
an implied part in any task description, and this may be different in the two
cases. It must then be determined whether the two descriptions are compatible
(3) A deliberate refusal to accept certain rules which govern the task to be carried
out, deliberately breaking the law when this concerns legal regulations. This
would be exemplified by a driver who knows the speed limit is 90 kmh, but who
travels nonetheless at 110 kmh (Moget-Monseur and Biecheler 1984). This can
be recognised as a typical case of 'violation' in Reason's terminology (1988).
These sources of deviation are very different, the first two are cognitive the third
concerns attitude and motivation, and the remedial measures for these two error
sources will not be the same. This is a problem whieh is often discussed in relation to
risk: does the driver taking the risk ignore the consequences or is he or she fully aware of
them? (Saad 1988). The deviation between the task to be carried out and the redefined
task, if it is considered to be an error, should be categorized as a 'mistake' (Reason
1987): the driver does not set him or herself the correct task.
When the definition of the redefined task is based on a collection of closed
questions, as in the survey carried out by Moget-Monseur and Biecheler (1984) referred
to above, it is easy to evaluate the deviation between the task to be carried out and the
redefined task, on a very general level and not in a specfic situation. Moget-Monseur
(1984)was therefore able to define a 'deviation scale', and put forward the idea that this
deviation could also be evaluated in relation to what is termed the 'mean of declared
behaviours', i.e., a sort of mean redefined task. These deviations could later be
compared to actual behaviour and accidents.
 Task, activity, and errpr analysis 1397

3.3. Deviation between the redefined task and the actual task
This deviation shows that the driver has not done what he set out to do. It may have
several sources.
(I) Lack of skill. This deficiency may be found at a sensorimotor level (lack of
control when handling the vehicle), or at a more specifically cognitive level
(error when representing the situation, resulting in unsuitable driving). It can
also stem from an unsuitable instruction or, more radically, from human
operational limitations (the best visual or sensorimotor capacities do not
exceed a certain level).
(2) Unforeseen events. The activity is not carried out as planned because a non-
anticipated event has modified the procedure initially chosen. The driver does
not carry out the redefined task because it is not a good model for the situation
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

in hand.
(3) Incorrect evaluation of his or her own task. The driver may consider, wrongly,
that he or she has carried out the task he defined, because he or she evaluated
the task incorrectly.
If this deviation is considered to be an error, it will be categorised as a slip: the driver has
not carried out what he wished to do (cf., Reason 1988).

3.4. Deviation between the actual task and the driver's representation ofhis or her activity
This deviation is that of two representations or activity models, the representation
made by the analyst- the actual task-and the representation the driver makes him or
herself. Here we have all the problems facing the subject with regard to the anaylsis he
or she is likely to make of his or her own activity. Some features relate more to the
driver's activity. This deviation shows that the driver evaluates his or her own action
badly, that the feedback he or she uses is biased. The sources of this inadequacy, which
are also often those of the driving error, are multiple. They result essentially from the
nature of driving activity, a considerable part of which is carried out automatically. It is
therefore difficult for the driver to account for this part of the activity. Thus the driver
may see a signal, and reply to it correctly without remembering he or she has seen it.
One source of deviation is also linked to the complexity of the task, and to a possibly
severe time restrictions in which to perform the task, limitations which act as
identification obstacles such as the memorizing and exact restitution of activity
charactersitics.

3.5. Deviation between the redefined task and the way in which the driver represents his or
her own activity
This deviation forms the error for the driver. Figure I clearly shows how this error
differs from the error for the expert: it is differentiated by the reference task and the
activity representation. Should these two errors not coincide e.g., the driver does not
acknowledge he or she has committed the error the specialist claims he or she has, this
may be because the task to which he or she is referring is not the same (for example, the
driver fails to recognize, or interprets badly, a provision of the Highway Code), or
because the activity as it is modelled by the specialist (actual task), does not correspond
to the model the driver has made (task to be carried out). The error diagnostic for the
expert should result in the determining of the part played by these two sources of
deviation.
 1398 J. Leplat

The error for the driver can itself be attributed to the difficulty he or she has in
carrying out the task he or she has set him or herself.This difficulty may be linked to a
deviation between knowledge and know-how: the driver knows he or she must stop at a
stop sign, but he or she brakes too late or badly, and his or her car runs over onto the
main road. This difficulty may also be because the task defined initially or at some point
during task performance, proved to be inadequate in relation to the actual conditions:
the driver did not do what he or she wanted to do because of conditions he or she did
not take into consideration, and could not therefore control.
The deviation between the redefined task and the driver's representation of his or
her own task can be categorized as a slip: not carrying out what he or she intended to
do. It is therefore necessary to differentiate between two categories of slips: those which
result from a deviation between the intention and the actual task, and those which
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

result from the deviation between the intention and the task which the subject thinks he
or she has carried out. This last category would seem best suited to the standard
definition of slip.

4. Discussion and conclusion

The analysis model which has just been presented can be used to analyse human actions
which have led to an error or an accident. This use raises, however, problems relating to
the nature of the model and the situation in which it is implemented. Some of these
problems will now be examined.

4.1. The notion of task

One advantage of the model is that it proposes a common language to analyse the
different stages during which the action is performed. However, in every analysis of a
practical task, the difficulty is to define the analysis units, in this case to define the
elementary tasks to be differentiated in the overall task. The task can be divided up
without considering the activity related to it, by only taking into account logic
requirements or ease of division: divisions proposed in driving studies often fall into this
category. It is also possible to divide up the task in relation to the actions which perform
the activity; an action being defined by a conscious goal. Leontiev (1972, 1975)initially
proposed an approach of this type, which was taken up by research specialists working
on the goal direction action concept (von Cranach, 1982). The driver's explicitation of
the redefined task, would lead back to this type of approach. The division of the task is
then in relation to driver characeristics, in particular his or her level of skill and his or
her familiarity with the situation. Here, the operations which perform the action
represent the automated part of the activity.
The dividing up of the global task may be seen according to two dimensions, as
discussed by Rasmussen (1986), an aggregate dimension, and an abstraction
dimension. The first characterises the size, which can vary, of the units to be
differentiated (going to work ... , crossing an intersection ... , braking ...). Here, the
abstraction dimension would concern the scope of the category of situations the task is
referred to, from a single task (a certain driver crossing an intersection at a certain
moment) to a category of tasks (crossing an intersection). Defining the task category
under consideration will obviously depend on the aim of the study, but is will always be
important to state the levels.
One difficulty in structuring the task also stems from the continuity of the action
sequence when driving a car, and the transformations of the task resulting from
unexpected events (cf., notion of discontinuity, Malaterre 1990).
 Task, activity, and error analysis 1399

4.2. Hypothetical nature of the notion of task: methodological problems

The four tasks which have been differentiated here have different status. The task to be
carried out has an ambiguous status: it can be perceived as being constructed by the
expert using an a priori goal set by the driver. For example, in an in-depth examination
of an accident, the specialist may define a task which corresponds to the conditions
facing the driver. This will be what is sometimes termed the task to be carried out so
that there will not be an accident. The description of this task is based on hypotheses
pertaining to the driver's skill and the characteristics of the system of which he is a part.
The redefined task is the driver's mental representation, which is not directly
accessible and which the analyst will infer using observable data. This redefined task
inferred by the analyst will only be a model of the driver's redefined task.
The actual task, or the task which it can be said the activity of the subject performs,
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

is inferred by the analyst based on the driver's activity. This activity has an overt and a
covert facet. What can be seen of the activity is only an element used as a basis to
attempt to determine the cognitive functioning. The actual task is therefore constructed
by the analyst, his model of the subject's activity. This model retains from the activity,
characteristics which are more or less important and profound.
The driver's representation of his or her own activity is, like the previous tasks, and
with similar methods, inferred by analysis which defines a model-a representation of
the driver's representation of his or her activity.
This succesion of model constructions shows the difficulty of identifying activity
mechanisms in the driving situation and, what is even more difficult, identifying error
mechanisms. It must also be remembered that the analyst him or herself may make
mistakes when constructing these models.
It will not be forgotten that tasks can be characterized in several ways, more or less
complete: using their conditions, the organization of these conditions or the processing
procedures. The models may also be more or less in-depth approximations of the task
being considered.
The emphasis placed on representations in the conception presented in this paper,
shows that it is little adapted to the study of automatisms: it can identify them, but does
not provide a way of analysing them.

4.3. Comparison with other analysis categories

The analysis category proposed in this paper does not exclude other, more
conventional analysis categories to which it would be worthwhile linking it. The four
task categories mentioned here could be compared to the activity constituents: the task
to be carried out corresponds to the goal and task setting; the redefined task
corresponds to the conception or planning phase; the actual task to the task carried
out, the representation of the activity carried out to the evaluation or control task. It is
simple to recognize these phases in standard models, particularly in the layered
models described by Norman (1986) and Rasmussen (1986), especially in his latest
version of this (1988). This text contains the following processing activities, 'Definition
of task, planning or procedure, execution of acts, monitoring' which would seem to
correspond to the four tasks defined above.
The model presented in this paper shows two distinct features which are shown in
figure 2. It emphasizes both the differences between and also the relations between task
and activity. It must therefore emphasize the activity units which are meaningful for the
task to which they correspond. These activity units integrate the elementary actions
which are most often dealt with by standard models.
 1400 J. Leplat
.--Representetlon by
_------ the expert or the enellYSl
Tesk to ree]lze
-"
---~'-- ..---- Representotion oy
lhednver --

__ Representotion by
._--.----- the expert or the analYSt
1-'
Acl\vjl~
'--
..... Representatlon by
the driver

Figure 2. Task, activity, and their representations.

Downloaded by [University of Sussex Library] at 12:12 09 February 2015

A second feature of this model is that it distinguishes the representations of the task
and driver's activity on the one hand, and the expert or analyst on the other. It has been
seen that the deviations between all these representation categories form a particularly
important source of information for the sudy ofthe activity and the origin of the error.
Describing these representations in terms of task, facilitates the expression of the
deviations and their interpretation in relation to the activity.
It would be interesting to discuss the now classical categories of errors defined by
Reason (1987, 1988) at the light of the distinction proposed here. For example: the
notion of mistake as planning failure raises the question 'failure in relation with what
type oftaskT In complex driving situations, the notion of rationality evoked by Reason
approximates the notion of task to realize. But this task, as with rationality, is alo
related, in some measure, to the driver's competence. There is, here a field for future
research: Reason has provided some clues on how to approach it.

4.4. Use of the model for error analysis

There are several ways of using a model such as this for error analysis. Using the four
task categories, it is possible to draw up a list of questions to act as a guide when
working out the representation of these tasks, and using their deviations. With regard
to task goal(s), for example, the subject will be asked to explain the goal to be reached,
the goal he or she sets him or herself, and the result obtained, in his or her opinion and
in the expert's opinion. The same will be done for all the factors used to characterize the
task. For conditions judged to be relevant, the state of a vehicle's tyres, for example, an
examination will be made of: the technical characteristics of the tyres, the driver's
representation of these tyres, the part played by the tyres according to the expert and
according to the driver. Questions will then be directed towards the comparison of
deviations (see figure 1).
It is also possible to use the proposed model for statistical purposes. The principle of
this model is to compare sources of error with the errors themselves. An example of
this, already reported, was provided by Moget-Monseur and Biecheler (1984).
In accident analysis, the difficulty of using such a model is that of collecting reliable
information with which to draw up the four representation categories. A posteriori,
these representations are distorted in a way which makes it difficult to characterise and
interpret the deviations.
Analysis of the deviations between task representations highlights problems which
cannot be solved by the model itself. This solution will have to be found in existing
 Task, activity, and error analysis 1401

knowledge and theories. The model cannot be substituted for these, but it can help co-
ordinate and integrate the contributions they make.

References
CROZIER, M. and FRIEDBERG, E. 1977, Uacteur et Ie systeme (Editions du Seuil, Paris).
GIRARD, Y. 1989, Analyse des dysfonctionnements observes dans une tache de traversee
d'intersections amenagees, INRETS, Arcueil, communication aune journee specialisee du
17 janvier.
HALE, A. R., QUIST, B. W. and STOOP, S. 1988, Errors in routine driving tasks: a model and
proposed analysis technique, Ergonomics, 31, 631--{j42.
JONES, E. E., KANOUSE, D. E., KELLY, H. H., NISBETT, R. E., VALINS, S. and WEINER, B. (eds), 1972,
Attribution: Perceiving the Cause of Behavior (General Learning Press, Morristown, NJ).
KJELLEN, U. 1984,The deviation concept in occupational accident control, Accident Analysis and
Prevention. 16, 289-323.
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

LEONTIEV, A. 1972, Le deoeloppment du psychisme (Editions sociales, Paris).

LEONTIEV, A. 1975, Activite, conscience, personnalite (Editions du Progres, Moscow).
LEPLAT, J. 1985, Erreur humaine.fiabilite humaine dans Ie travail (A. Colin, Paris).
LEPLAT, J. and CUNY, X. 1984, Introduction a la psychologie du travail (Presse Universitaire
de France, Paris).
LEPLAT,1. and Hoc, 1. M. 1983, Tache et activite dans l'analyse psychologique des situations,
Cahiers de psycholoqie coqnit ive, 3, 49--{j4.
LOURENS, P. F. n.d., Pschogenesis of incorrect driving behaviour, unpublished report.
MALATERRE, G. 1990, Error analysis and in-depth accident studies, Ergonomics, 33,1403-1421.
MALATERRE, G. and PEYTAVIN, J. F. 1985, Exploitation de la banque de donnees detaillees
d'accidents, rapport ONSER, Arcueil.
MOGET-MONSEUR, M. 1984, Regles legales et normes sociales en matiere de circulation, rapport
interne ONSER, Monthlery,
MOGET-MoNSEUR, M. and BIECHELER, M. B. 1984, Le comportement de base du contucteur
automobile, rapport ONSER, Arcueil.
MUHLRAD, N. 1988, Techniques des conflits de trafic, synthese INRETS, II, Arcueil.
NORMAN, D. A. 1986, Cognitive engineering, in D. A. Norman and S. W. Draper (eds), User-
Centered System Design (Lawrence Erlbaum, Hillsdale, NJ), 31-62.
QUENAULT, S. W. 1968, Development of the method of systematic observation of driver
behaviour. Report LR213 Road Research Laboratory, Crowthorne, Berkshire.
RASMUSSEN, J. 1986, Information Processing and Human-Machine Interaction (North Holland,
New York), 215.
RASMUSSEN, J., 1988, Concepts for analysis of cooperative works, working paper Riso Project
5.17.5.6.07.
REASON, 1. 1987, A preliminary classification of mistakes, in J. Rasmussen, K. Duncan and J.
Leplat, New Technology and Human Error (Wiley, Chichester), 15-22.
REASON, 1. 1988, Human Error (Cambridge University Press, Cambridge).
REASON, J., MANSTEAD, A., STRADLING, S., BAXTER, J. and CAMPBELL, K. 1990, Errors and
violation on the roads: a real distinction? Ergonomics, 33, 1315-1332.
SAAD, F. 1988, Prise de risque ou non perception du danger, RT S Recherche Transports,
Securite, 18-·19, 55-62.
SEBILLOTTE, S. 1988,Hierarchical planning as method for task analysis: the example of office task
analysis, Behaviour and Information Technology, 7, 275-294.
VON CRANACH, M. 1982,The psychological study of goal-directed action: basic issues, in R. Harre
and M. von Cranach (eds), The Analysis of Action (Cambridge University Press,
Cambridge), 35-73.

La notion d'erreur, lorsqu'elle est appliquee a une activite ou au resultat d'une activite implique
la notion de tache: elle exprime la discordance entre l'activite etla tache consideree, so us un angle
juge approprie. La tache et l'activite sont l'objet de representations pour celui qui analyse (Ie
specialiste) et pour Ie conducteur. Dans cet article, on etudie quatre types de representations: la
tache et l'activite pour Ie specialiste ainsi que la tache et l'activite pour Ieconducteur. On illustre
et on interprete ces taches ala lurniere des travaux actuels dans ce domaine. La signification des
 1402 Task, activity, and error analysis

differences entre ces representations est ensuite argumentee et on met I'accent sur I'importance de
ces differences qui permettent d'elucider les rnecanismes generateurs d'erreurs. On aborde les
problernes methodologiques et pratiques de I'analyse en termes de tache et d'activite. Pour cela, il
peut etre utile de se referer a d'autre travaux theoriques existant dans ce domaine. De cette
maniere, il devient possible de promouvoir d'autres etudes sur les erreurs humaines. En
particulier des etudes ou d'autres categories de representations seraient prises en compte.

Der Begriff des Fehlers, wenn er auf eine Handlung oder auf das Ergebnis einer Handlung
bezogen wird, impliziert den Begriff der Aufgabe. Dabei wird die Abweichung des Winkels
zwischen der Handlung und der in Betracht gezogenen Aufgabe von einem als relevant
beurteilten Winkel ausgedriickt. Die Aufgabe und die Handlung sind Objekt der Repriisentat-
ionen fiir den Analytiker (oder Spezialisten) und fiir den Fahrer. Vier Repriisentationen werden
in diesem Artikel behandelt: Die Aufgabe und die Handlung fiir den Spezialisten und die Aufgabe
Downloaded by [University of Sussex Library] at 12:12 09 February 2015

und die Handlung fiir den Fahrer. Fiir diese Aufgaben wird eine Interpretation vorgestellt.
lIlustriert werden die Aufgaben anhand von Arbeiten, die auf diesem Gebiet bereits ausgefiihrt
wurden. Die Signifikanz der Abweichungen zwischen den Repriisentationen wird anschliel3end
zusammen mit den Vorteilen des Studiums der Abweichungen in Zusammenhang mit der
Kliirung von fehlergenerierenden Mechanismen diskutiert. Die Analyse an hand von Aufgabe
und Handlung bringt methodische und praktische Probleme mit sich, die angesprochen werden.
Das schlieBt nicht den Bezug zu theoretischen Rahmenmodellen aus dem psychologischen
Bereich aus, mit denen eine Verbindung herzustellen lohnend ist. Diese Betrachtungsweise wirft
Fragen auf, die es ermoglichen, das Studium von Fehlern zu erweitern: Zu einem spiiteren
Zeitpunkt konnte es mit der Ausdehnung auf andere Kategorien von Repriisentationen
vervollstiindigt werden.