We present here a general theory of human reaction time (RT) that covers the three kinds of RT experiments originally described by Donders. The theory is not a stage model as originally envisioned by Donders and his successors. Rather, the strength of the excitatory tendency leading to response evocation grows continuously as a function of the time since stimulus onset. Response occurs when this strength reaches the value of a decision criterion that is normally distributed over trials. This leads to a Thurstonean measurement model that facilitates the analysis of data and has led to the development of the present theory.Excitatory strength may consist of either one or two processes: detection information, dependent on sensory strength, and associative strength. In simple RT only detection is involved, but in choice RT both are present. The analysis further reveals that in disjunctive RT, the Donders c reaction, there are two modes of performance, or "strategies," used by different subjects. Thus, correct performance may depend on the detection process alone or on the associative process as well. The theory also deals with the dynamics of making and avoiding errors and, consequently, with the speed-accuracy tradeoff. Associative inhibition attenuates incorrect tendencies resulting from detection, and associative discrimination attenuates generalized associative strength for the incorrect response. Both processes grow with time.In this article the theory is applied to a series of five experiments with auditory stimuli differing in intensity. Intensity was the basis of differential response in disjunctive and choice experiments. The theory successfully describes the data of the individual subjects in all experiments as well as the group data. Subjects apparently adjusted their criteria between experiments in disjunctive RT and between responses in choice RT as a means of reducing short latency errors to the more intense stimulus.The theory provides an analysis of the nature of individual differences, which also reveals further features of information-processing dynamics. An important one of these is the role of criterion variability in causing errors.

A theory of choice reaction time (CRT) is presented which is based on a variable criterion model assuming the decision criterion to be normally distributed. The theory provides functions of time describing the growth of component processes following stimulus onset. For correct responses, these processes are sensory and associative strength. For errors, the processes are sensory strength, generalized associative strength, and associative inhibition. A model for separating the effects of these processes from that of response competition is presented. The theory describes, with great accuracy, the CRT distributions of correct responses and errors, not only for the experiment for which it was derived, but also for a second experiment in which the criterion was experimentally manipulated. It accounts for the basic facts of the speed-accuracy tradeoff, not only with respect to the two experiments with different speed requirements, but also with respect to the variation of speed over trials within each experiment. While derived from group data, the theory also describes the performance of individual subjects. The mean and variability of the criterion distribution and the ability to utilize associative inhibition are identified as the major sources of individual differences. Methodological problems involved in the scaling analysis of group data are also discussed.The many years of continued interest in reaction time (RT) research lies largely in the fact that RT measures the time required for the processes involved in response evocation. Through experimental and theoretical analysis, it has been the goal to identify these processes and to specify their temporal and other properties. One approach, instituted by Donders and pursued more recently with increasing sophistication, is that RT is merely the sum of the times required for each of the necessary component -processes. An alternative is that each of the processes involved may develop as a separate function of time following stimulus onset, and that these functions may, at least in part, grow simultaneously. It is this type of interpretation which has resulted when variable criterion theory (Grice, 1968) has been applied to the analysis of latency phenomena. According to this view, response evocation will result when the combined strength of the processes satisfies a decision criterion.In analysis using the variable criterion principle, we assume the criterion to be a normally distributed random variable, under homogeneous conditions of performance, and treat it as the only source of random variability. This is in contrast to other models

The theory presented by Grice, Nullmeyer, and Spiker (1977) has been applied successfully to experiments with three levels of auditory similarity in choice reaction time. A defect in the original statement of the relation between theoretical probabilities and response probability is corrected. Similarity affects both positive associative strength and associative inhibition. The short-latency process of associative inhibition is especially sensitive to increasing levels of similarity. Descriptions of the speed-accuracy tradeoff as it depends upon stimulus similarity are presented. The forms of stimulus generalization gradients are shown to depend upon response latency. Common sets of theoretical functions described both group and individual subject latency distributions. As before, three individual difference variables are identified as mean criterion level, amount of criterion variability, and inhibitory ability. Multiple regression analysis of these three variables and error rate reveals interesting and interpretable features of information processing dynamics. One of these features is that the amount of criterion variability may be more important in determining error rate than the mean level of the criterion. Also included are comments on a critique by Link (1979) and a discussion of the relation between this approach and more formal deductive models.Early applications of variable criterion analysis led to the development of a theory of simple reaction time (SRT) (Grice, Hunt, Kushner, & Morrow, 1974) and disjunctive reaction time (DRT) of the c-reaction type (Grice, Hunt, Kushner, & Nullmeyer, 1976). More recently, we have presented an empirically derived theory of choice reaction time (CRT) (Grice, Nullmeyer, & Spiker, 1977). The result is a structure of some generality in that, when the same stimuli are used, functions obtained in SRT and DRT experiments become components of the CRT theory. According to this view, the excitatory strength (E) leading to response evocation increases as some deterministic function of time following stimulus onset. Response evocation occurs when the strength of E reaches a preset criterion value (C) which is specific to the trial. Over trials, the criterion is a normally distributed random variable with mean C and standard deviation a. The mean and variability of the criterion distribution are determined by task variables, instructions, set, attention, motivation, adaptation, and individual differences.As a scientific approach, our theoretical research is, at present, somewhat unconventional, and also, we fear, not fully understood. This is exemplified in a recent critique by Link (1979). While we believe his criticism to be correct in one important respect to be discussed later, it appears that his more general recommendations are based on a complete misunderstanding of our scientific goals and strategy. Our approach is not that of a simple, purely deductive model of the kind now most common in mathematical psychology. In addition to the usual deductive aspect of theory, we s...

Scaling analysis based on variable criterion theory has been applied to the c-reaction form of disjunctive RT. In addition to previously identified sensory growth functions, two associative processes have been identified and functions of time describing their growth have been obtained. Associative strength to the positive stimulus begins at about 200 msec, or after, and grows with initial positive and later negative acceleration. Associative inhibition to the negative stimulus begins earlier, shortly after the sensory detection functions, and grows rapidly with negative acceleration. Subjects may adopt strategies which emphasize the use of either of these associative processes. With the pure inhibitory strategy, they respond to the positive stimulus entirely on the basis of sensory detection, but associatively inhibit response to the negative stimulus. With either strategy, the speed-accuracy tradeoff was determined by the level of criterion adjustment.To describe the temporal growth in the strength of the sensory and associative information leading to response evocation is one of the goals of variable criterion theory. This type of analytic approach is made possible by the scaling properties of the model based upon the normality of criterion distributions. The logic involved and the methods by which such functions may be derived from cumulative RT distributions have been described by Grice (1972b) and Grice, Hunt, Kushner, and Morrow (1974). Considerable progress has been made in the description of the growth of the sensory component. These functions are obtained from simple RT experiments in which the associative factor may be assumed to be absent. The basic model here is E = V -T, where V is sensory strength, T is the mean of the criterion distribution, and E is excitatory strength measured from T. V grows in strength following stimulus onset according to a negatively accelerated function. The rate of growth is a function of stimulus intensity. The probability of response at any time is given by (E,a), where is the normal function and a is the standard deviation of the criterion distribution. The study by Grice et a1. (1974) included three simple RT conditions with differing catch trial conditions designed to affect the criterion level. The auditory signal was also presented at three intensities in an irregular order. For each subject a family of three negatively accelerated functions was derived indicating the growth of V with respect to time for each intensity. Estimates of T and a were also obtained which described the relations between conditions and between experimental sessions. This family of three functions, together with the cri· terion parameters, provided quite a satisfactory description of the data.The situation with respect to discovery of the growth functions for associative information in RT research is less advanced. The model which has been successfully applied in eyelid conditioning (Grice, 1972a) is E = H + V -T, where H is the associative factor; and it is the intent to apply th...

The 2012 Pilot Source Study (Phase III) was a continuation of the 2010 Pilot Source Study (Smith, Bjerke, NewMyer, Niemczyk & Hamilton, 2010), using the same research design with a new data set containing no duplicate records. University faculty and students assisted seven regional airlines to enter data on 4,024 pilots hired between 2005 and 2011. New-hire pilots' college and piloting backgrounds defined their input (Source) variables. Training and first year operations data defined the output (Success) variables. Identifying information for pilots and participating airlines was removed from the data sets, and records were combined into a single data set for independent analysis by five researchers. Results were verified by two independent researchers from the University of Central Florida (a non-affiliated university). Results showed considerable consistency between the 2010 Pilot Source Study and the 2012 Pilot Source Study regarding initial pilot training at a regional air carrier. The study found that pilots entering the industry with an aviation-specific college degree, particularly a degree from an AABI-accredited flight program, performed better in initial training than those with no degree or a non-aviation degree. The results also indicated that a pilot's background, such as having a CFI certificate and obtaining advanced training from a collegiate aviation program, is an indicator of success in training. One important result was that commercial pilots had more completions than pilots with an ATP certificate. On the other hand, total flight hours produced inconclusive results.