Philosophies that attributed human consciousness to incorporeal entities whose rates of apprehension were infinitely fast ('the speed of thought') delayed recognition of this simple fact until very late in human history. In 1799, Nevil Maskelyne, then British Astronomer Royal, sacked his assistant Kinnebrook because their timings of stellar transits always disagreed. Friedrich Bessel (1784–1846), a more considering scientist, noted that discrepancies between measurements made by individual astronomers were widespread and substantial. For the first time someone saw the necessity of calibrating human observers, as well as the equipment that they used. Each astronomer, thenceforth, determined his personal, characteristic lag and used it as a corrective constant in his work.
By 1858 Helmholtz had recognized a further implication of this lag, seeing that one factor limiting human reaction time must be the speed with which electrical impulses travel along nerve fibres. To measure this speed he asked a human observer to bite on a contact switch as soon as he felt an electric shock, which might be delivered either to his foot or to his face. Responses to foot shocks were slower. Assuming that this difference occurred because impulses from the foot have further to travel before they reach the brain, Helmholtz measured the difference to obtain the first estimate for the velocity of transmission of nerve impulses. The phrase 'swift as thought' took on a potentially exact, if rather humble, value.
Helmholtz's brilliant insight, that measurements of human reaction times may be used to make deductions about the nature of otherwise unobservable neural processes, provided impetus for a next step. In 1868, F. C. Donders, a Dutch ophthalmologist, reasoned that a human must recognize a new stimulus before he can begin to organize a response to it. As we shall see, this is not quite true. However, Donders argued that, since a stimulus must be recognized before a response to it can be chosen and executed, lags for these two processes must seem to give overall observed reaction times. By appropriate techniques reaction times may therefore be decomposed to give independent estimates for times taken to recognize signals and for times required to choose responses to them.
To obtain such estimates Donders required his observers to carry out each of three different tasks. In the a task (simple reaction time task) they always had to make the same response as fast as possible every time a single signal occurred. In the b task (choice reaction task) any one of five different signals might occur in unpredictable order and the observer made a different response to each. In the c task (Go/No Go task) any one of the same five stimuli might occur, unpredictably, but the observer responded only to occurrences of one of them and ignored the others. Donders argued that differences in observed reaction times c−a would provide estimates of how much longer a person needed in order to distinguish among five different signals than to make a response without having to discriminate which signal had occurred. Similarly, the difference b−c would give an estimate of how long an observer needed to select among five different responses, rather than merely always to choose whether or not to make a particular response, after making the same discrimination. Donders's experiment gave him estimates for perceptual discrimination times (less than 50 milliseconds) which were far shorter than his estimates for response selection times (over 150 milliseconds), and he began to hope that his technique would allow independent estimates for other, more complex 'unobservable internal mental processes'.
Donders's results were reasonable in principle, but the logic of his experiment was flawed in two ways: first, in his c task his observer never had to discriminate each of five signals from all of four others, as in the b task. In the c task the observer merely had to discriminate one 'Go' signal from all of another set of four 'No Go' signals. There was thus no requirement to discriminate any 'No Go' signal from any other 'No Go' signal. Thus the b task involved both a more complex perceptual discrimination and a more complex response choice than the c task, and the b−c estimate for discrimination time confounded these factors.
A more interesting weakness of Donders's experiment was that, like all other successful animals, human beings overcome their temporal lag behind events in the external world by learning to predict what will happen next. When they can do this they can have reaction times of zero milliseconds, responding to events just as soon as they occur, or they can even take leaps into the future, gaining an edge over a rapidly changing environment, or over rapidly moving adversaries, by anticipating events which have not yet occurred. In Donders's a task his observer could set himself, in advance, to identify the same, recurrent, signal on every trial, and on every trial he could also safely prepare to make the same response to it. Thus the neural processes timed in the a task have much more to do with anticipation and expectancy than with the processes of discrimination and choice that underlie the b and c tasks, in which observers must always wait until after something happens before their sense organs and brains can begin to work out what has occurred.
It is now widely recognized that reaction-time measurements make little sense unless we bear in mind that, although all living things can experience only the past, animals like ourselves have gained our success in the world by continuously, and accurately, predicting the very immediate future. A good cricketer has to predict, before the ball leaves the bowler's hand, the precise point in space at which it will be intercepted by his bat. A squash player who could not anticipate would founder hopelessly behind the game. Reaction times do not provide us with measurements of the time necessary for sets of nerve impulses generated in the sense organs to activate those parts of the brain that, in turn, activate our muscles. They rather measure the duration of operation of processes of active, predictive control, by means of which we organize responses that anticipate, and pre-empt, very fast changes in the world.
This point was only very slowly recognized. In Leipzig, in 1879, Wundt set up the first laboratory of experimental psychology. His research programme was largely concerned with criticisms and extensions of Donders's work. It is sad that his many years of heroically dull experimentation now merely provide a textbook moral that conscious introspection can tell us little about the nature of decisions that may take us less than a fifth of a second to make.
As is often the case in experimental psychology, unpretentious calibrational studies proved the most fruitful in the long run. In 1892, von Merkle showed that an observer's reaction times increase regularly with the number of signals among which he has to discriminate, and with the number of responses among which he must, consequently, choose. In 1931 Henmon showed that, as pairs of signals become more similar to each other, people take longer to discriminate between them. Observations such as these are very simple, and unsurprising, but their consequences are not at all trivial. From experiments based on von Merkle's results, Hick (1952) developed models for human reaction times that were the first, and the most influential, demonstrations that the new sciences of cybernetics and information theory, which had been developed during the 1940s, have very fruitful applications as abstract descriptions of the way in which the human brain arrives at decisions. From data similar to those gathered by Henmon, later workers such as Audley, Crossman, Falmagne, Laming, Luce, and Vickers have shown how reaction-time measurements may be incorporated into a new psychophysics of signal detection (see bit). Non-specialists may capture something of the importance of this endeavour if they reflect that human experimental psychologists can only measure two things about human performance: how accurately people do things and how long they take to do them. Metrics and models that allow us to discuss speed and errors as complementary indices are basic tools of research.
(Published 1987)
— P. A. M. Rabbitt
- Bibliography
- Hick, W. E. (1952). 'On the rate of gain of information'. Quart. J. Exp. Psychol., 4.
