by James Reason
Not only is the issue of volition fundamental to notions of criminal responsibility, it is also crucial to the psychological definition of human error. Thus the term error can only be applied to intentional actions. It has no meaning in relation to nonintentional behaviour because error types depend critically upon two kinds of failure: the failure of actions to go as intended (slips and lapses) and the failure of intended actions to achieve their desired consequences (mistakes). To clarify these basic error forms, we need to consider the distinction between intended and unintended actions.
4.2. Distinguishing between intended and unintended actions
Assuming a prior intention and/or an ‘intention in action’, it is possible to build a psychologically meaningful taxonomy of error and correct performance on the basis of the answers to the second and third questions listed above: Did the actions go as planned? And did the actions achieve their desired consequences? Since our primary concern is with error, we will look first at those cases where actions deviate from intention: unintended actions or actions-not-as-planned.
4.2.1. Unintended actions
Actions that deviate from intention fall into two classes: those that nevertheless achieve their intended goal and those that do not. While just conceivable, the former are highly unlikely. Searle (1980) provides an example of one such ‘successful yet unintended action’: A man intends to murder someone by shooting at him. He misses, but the shot stampedes a herd of wild pigs, which tramples the intended victim to death. I will not consider these curiosities further. Psychologists (as distinct from lawyers, philosophers or even theologians) are more interested in the ‘acts of man’ than in the so-called ‘acts of God’.
Common among the acts of humans are moments of absent-mindedness when we become aware that our actions have strayed from their intended path. Two conditions appear to be necessary for the occurrence of these slips of action: the performance of some largely automatic task in familiar surroundings and a marked degree of attentional ‘capture’ by something other than the job in hand (see Reason, 1979; Norman, 1981).
4.2.2. Intended actions and mistakes
Even when the intended actions proceed as planned, they can still be judged as erroneous if they fail to achieve their intended outcome. In this case, the problem resides in the adequacy of the plan rather than in the conformity of its constituent actions to some prior intention. Errors of this kind are termed mistakes (Norman, 1981,1983; Reason & Mycielska, 1982).
Norman (1983) summarised the distinction between mistakes and slips (or lapses) very succinctly: “If the intention is not appropriate, this is a mistake. If the action is not what was intended, this is a slip.” Mistakes involve a mismatch between the prior intention and the intended consequences. For slips and lapses, however, the discrepancy is between the intended actions and those that were actually executed.
Another way of distinguishing these two basic error forms is as planning failures (mistakes) and execution failures (slips and lapses). This categorization emerges not only from the logic of the three-question algorithm, but also from a consideration of cognitive levels. Planning failures are likely to arise from higher-level processes than either slips or lapses.
5. Some working definitions
We are now in a position to sketch out some working definitions of error and its principal types. A working definition is serviceable rather than ideal. It tries to embody the essential psychological characteristics of the phenomena without struggling too hard to mark out their exact semantic boundaries. The study of error, being largely an inductive mode of enquiry, does not demand precise axioms and definitions at the outset, as do the deductive sciences. The following working definitions have proved their utility in a psychological rather than a philosophical sense.
Error will be taken as a generic term to encompass all those occasions in which a planned sequence of mental or physical activities fails to achieve its intended outcome, and when these failures cannot be attributed to the intervention of some chance agency.
As we have seen, a series of planned actions may fail to achieve their desired outcome because the actions did not go as planned or because the plan itself was inadequate. It is also possible that both types of error could occur within the same sequence of planning, storage and execution. This distinction gives rise to two further working definitions.
Slips and lapses are errors which result from some failure in the execution and/or storage stage of an action sequence, regardless of whether or not the plan which guided them was adequate to achieve its objective.
Whereas slips are potentially observable as externalised actions-not-as-planned (slips of the tongue, slips of the pen, slips of action), the term lapse is generally reserved for more covert error forms, largely involving failures of memory, that do not necessarily manifest themselves in actual behaviour and may only be apparent to the person who experiences them.
Mistakes may be defined as deficiencies or failures in the judgemental and/or inferential processes involved in the selection of an objective or in the specification of the means to achieve it, irrespective of whether or not the actions directed by this decision-scheme run according to plan.
It is evident from this definition that mistakes are likely to be more subtle, more complex and less well understood than slips. As a result, they generally constitute a far greater danger. By their nature, mistakes are also far harder to detect. Consciousness is specifically tuned to picking up departures of action from intention (Mandler, 1975), but mistakes can pass unnoticed for lengthy periods (Woods, 1984), and even when detected they sometimes remain a matter of debate. Not only is the quality of the plan open to a diversity of opinion, it is also something that can be judged at two distinct stages: before and after it has been implemented. Prior to its execution, it can be assessed according to whether or not it has sufficient contingencies, displays soundness of judgement, imagination, flexibility, awareness of detail and the like. But once put into action, we judge it primarily according to how well it achieved its stated objectives.
Unfortunately, the results of these two sets of judgements do not necessarily correspond. Plans rated as good prior to their execution can fail to attain their objectives; whereas plans judged as inadequate by any reasonable criteria can turn out successfully, though not necessarily to the credit of their originators. Consider, for example, the relative qualities and eventual outcomes of the German and French plans for waging the First World War. By most a priori standards, the Schlieffen Plan was a masterpiece of military thinking. It was bold, imaginative and worked out to the last logistic detail; but it ultimately failed. The corresponding Plan XVII was, by comparison, crude and ill-considered, but it eventually succeeded. Janis (1972, p. 11) put the matter succinctly in a discussion of high-level policy making: “Defective decisions based on misinformation and poor judgement sometimes lead to successful outcomes... we must acknowledge that chance and the stupidity of the enemy can sometimes give a silk-purse ending to a command decision worth less than a sow’s ear.”
6. The classification of errors
A central problem in error classification is the difficulty of reconciling the often highly specific contextual triggers of a particular error form with the fact that it may also manifest the influence of some very general adaptive process or basic error tendency. A classification that emphasises the former at the expense of the latter is likely to overlook the broad regularities in the more predictable forms of error. Yet one that neglects the local contextual factors (e.g., task and situational considerations) will not only be of limited use to the practitioner, it will also fail to accommodate the theoretically important causal interactions between basic error tendencies and immediate task features.
There is no universally agreed classification of human error, nor is there one in prospect. A taxonomy is usually made for a specific purpose, and no single scheme is likely to satisfy all needs. Nearly everyone who has published in this field has devised some form of error classi
fication. Consequently, the literature abounds with such taxonomies, reflecting a variety of practical concerns and theoretical orientations and ranging from the highly task specific to broad statements of underlying error tendencies.
Notwithstanding the number and diversity of existing error taxonomies, it is possible to penetrate beyond their surface idiosyncracies to distinguish three levels at which classifications are attempted: the behavioural, contextual and conceptual levels. These correspond approximately to the “What?”, “Where?” and “How?” questions about human errors.
6.1. The behavioural level of classification
At the most superficial level, errors may be classified according to some easily observable feature of the erroneous behaviour. These can include either the formal characteristics of the error (omission-commission, repetition, misordering), or its more immediate consequences (nature and extent of damage, injury). Classifications dealing with applied data at this behavioural level may also be concerned with such things as recoverability, human versus machine attribution, and operator versus design responsibility.
Most human behaviour is essentially serial in character, and there are only a limited number of purely formal ways in which a sequence of words or actions can deviate from intention. Therefore, it is not surprising that classifications of verbal and action slips framed at this level show a high degree of uniformity. There is also a large measure of agreement between judges in their allocation of slips to these limited behavioural categories (Reason, 1984a).
Despite the parsimony of such schemes and the appealing conformity of error-to-category assignment, there are strong grounds for believing that there is no simple and direct mapping of these behavioural error types onto more theoretical categories of cognitive failure. Rather, the evidence indicates that members of the same behavioural error class can arise from quite different causal mechanisms and that members of different behavioural categories can share common aetiologies (see Norman, 1981; Reason & Mycielska, 1982; Reason, 1984a).
6.2. The contextual level of classification
This level goes beyond the formal error characteristics and includes limited assumptions about causality. In most cases, these assumptions do not stray far from the ‘surface’ data. Many of the slips of the tongue and pen taxonomies are constructed at this level and include reference to such contextual triggering features as anticipations and perseverations.
Such categorizations are valuable because they draw attention to the complex interaction between ‘local’ triggering factors and the underlying error tendencies. They address themselves to the issue of what prompts an error to appear at a particular point in the behavioural sequence and so stress the importance of recording as much information as possible regarding the surrounding circumstances, both internal and external to the perpetrator of the slip. In short, they acknowledge the critical relationship between error type and the character of the situation or task in which it appears.
Yet even this useful level of classification has serious limitations. By themselves, contextual factors cannot explain why the same or very similar circumstances do not always trigger the same error forms. Freud (1922, p. 36) was well aware of this problem, as the following passage reveals:
The influence of sound-values, resemblances between words, and common associations connecting words, must also be recognized as important. They facilitate the slip by pointing out a path for it to take. But if there is a path before me, does it necessarily follow that I must go along it? I also require a motive determining my choice and, further, some force to propel me forward. These sound-values and word associations are, therefore, just... the facilitating causes of slips of the tongue, and cannot provide the real explanation for them.
To accept Freud’s point does not necessarily require agreement with his theory of unconscious determination, but it does indicate the necessity of plumbing comparable ‘depths’ within the cognitive system in order to find a more solid bedrock upon which to construct a classificatory framework.
6.3. The conceptual level of classification
This third level is predicated on assumptions about the cognitive mechanisms involved in error production. In contrast to the other two, these classifications are based more upon theoretical inferences than on the observable characteristics of the error or its context. With each successive level of classification, we move further from the immediate ‘surface’ data and deeper into the realm of assumption and conjecture. Despite these problems, classifications based upon conceptual considerations are potentially the most fruitful because they seek to identify underlying causal mechanisms. At this point, it would be helpful to introduce the distinction between error types and error forms.
7. A distinction between error types and error forms
7.1. Error types
The term error type relates to the presumed origin of an error within the stages involved in conceiving and then carrying out an action sequence. These stages can be described under three broad headings: planning, storage and execution. Planning refers to the processes concerned with identifying a goal and deciding upon the means to achieve it. Since plans are not usually acted upon immediately, it is likely that a storage phase of some variable duration will intervene between formulating the intended actions and running them off. The execution stage covers the processes involved in actually implementing the stored plan. The relationship between these three stages and the primary error types is shown in Table 1.1.
For reasons that will be discussed later, mistakes can be further subdivided into (a) failures of expertise, where some preestablished plan or problem solution is applied inappropriately and (b) a lack of expertise, where the individual, not having an appropriate ‘off-the-shelf routine, is forced to work out a plan of action from first principles, relying upon whatever relevant knowledge he or she currently possesses. These two types of mistakes correspond closely to the rule-based and knowledge-based levels of performance as described by Rasmussen (1983).
Table l.l. Classifying the primary error types according to the cognitive stages at which they occur.
Cognitive stage
Primary error type
Planning
Mistakes
Storage
Lapses
Execution
Slips
7.2. Error forms
Whereas error types are conceptually tied to underlying cognitive stages or mechanisms, error forms are recurrent varieties of fallibility that appear in all kinds of cognitive activity, irrespective of error type. Thus, they are evident in mistakes, lapses and slips. Error forms are so widespread that it is extremely unlikely that their occurrence is linked to the failure of any single cognitive entity. Rather, this omnipresence suggests that they are rooted in universal cognitive processes, particularly the mechanisms involved in knowledge retrieval. Two such error forms will be considered at some length in this book: similarity and frequency biases (see Chapters 4 and 5).
8. Methods of investigating human error
8.1. Naturalistic methods or ‘corpus gathering’
For well over a 100 years, psycholinguists in particular and, more recently, cognitive psychologists in general have been collecting, analysing and classifying naturally occurring slips and lapses. The rationale for this type of investigation was eloquently expressed by Bawden (1900, p. 5) as follows.
Just as little irregularities in the road enable one accustomed to it to make his way in the dark, so to the student of human nature little inadvertencies of expression, aberrations in speech, lapses of thought, confusion of ideas, hitches or slips in speaking or writing are sometimes most useful and unerring guides in the understanding of mental process. Neglected trifles are sometimes suggestive of most fruitful problems for research.
Corpus gathering is concerned with the identification and description of naturally occurring phenomena, and is the first step in the process of classification. It has the merit of portraying the richness and variety of real-world phenomena an
d, given a large enough corpus, it provides a reasonably comprehensive qualitative account of the available species of error. From this analysis, it is possible to identify recurrent pattern of error across aspects of mental life that are often treated in relative isolation. Not only does corpus gathering satisfy the criterion of ecological validity, it also offers a much broader perspective on the mental landscape than can be obtained from necessarily focused laboratory studies.
But it is also clear that corpus gathering is not enough. The errors so noted may be natural and spontaneous, but these very qualities also mean that the investigator has little or no control over the circumstances of their occurrence. Without the possibility of systematically manipulating the various predisposing factors, it is difficult to achieve scientifically satisfactory causal explanations. In short, natural history techniques are excellent for providing a wide-angle view of the phenomena, but they often raise more questions than they can answer.
Methods used in the collection of naturalistic error data range from the kind of inspired self-observation employed by Freud (1914), to the more quantitative extended diaries of recent times (Reason, 1979, 1984a; Reason & Lucas, 1984b), as well as ingenious attempts to bring the rigour of laboratory measurement into the realm of everyday life (Wilkins & Baddeley, 1978).