Search

This article is concerned with four major propositions: 1) The next major thrust in long‐range planning will come in the public sector, and is a thrust long overdue. 2) Over the past two decades we have significantly improved corporate planning in the private sector. We still have much to learn, but compared with the early efforts we are now high on the learning curve. 3) The fundamental lessons we have learned in the private sector in doing comprehensive long‐range planning are applicable to the introduction, development, and success of aggregate integrated long‐range planning in the public sector. 4) It is imperative that a dialogue begin immediately between practitioners of long‐range planning in the private sector and those in the public sector who are directly involved in the development of this process. Both groups have much to learn from each other, but it is particularly important that those in the public sector be exposed to the first‐hand experience of those in the private sector.

SUMMARY A three parameter time constant learning curve model is applied to four case studies and the optimum parameters determined in the least sum of error-squared sense. The error residuals are then analysed using correlation, regression, and cusum techniques. A global range for learning curve scatter is derived from grouped data. One case study establishes a sinusoidal trend additional to wide-band noise. In all cases studied there is correlation between successive error residuals so that there are many less sign changes than predicted from an assumption of normally distributed, uncorrelated scatter. A simulation model is documented which generates artificial learning curves with the requisite degree of correlation as evident in terms of the number of sign changes. The results of the paper may be used to establish the a priori scatter estimates necessary in various parameter estimation techniques developed for on-line prediction of human operator future performance.

The primary objective of Experiment M151 was to study by means of time and motion analytic techniques the inflight adaptation of Skylab crewmen to a variety of task situations involving different types of activity. A parallel objective was to examine astronaut inflight performance for any behavioral stress effects associated with the working and living conditions of the Skylab environment. Training data provided the basis for comparison of preflight and inflight performance. Efficiency was evaluated through the adaptation function, namely, the relation of performance time over task trials. The results indicate that the initial changeover from preflight to inflight (or, from 1-G to zero-G) was accompanied by a substantial increase in performance time for most work and task activities. Equally important was the finding that crewmen adjusted rapidly to the weightless environment and became proficient in developing techniques with which to optimize task performance. By the end of the second inflight trial, most of the activities were performed almost as efficiently as on the last preflight trial. In addition, the analysis demonstrated the sensitivity of the adaptation function to differences in task and hardware configuration. The function was found to be more regular and less variable inflight than preflight. Translation and control of masses (large or small) were accomplished easily and efficiently through the rapid development of the arms and legs (and the entire body) as subtle guidance and restraint systems. Finally, the adaptation function provided no evidence of behavioral stress effects attributable to the Skylab environment.

SUMMARY Four subjects were submitted to an experiment to learn two-handed symmetrical simultaneous assembly work Even after several months, a final performance could not be determined. Particularly in the early stages of training, the learning curve showed peaks at the beginning of the week, after training-free days. Financial incentive resulted in the increase of the final performance; fatigue, however, caused decreasing performance. Training of the difficult motion elements such as “grasp” and “assemble”, the variability of which accounted for almost all individual differences of performance, did not lead to a lasting increase of performance. The use of the coefficient or variation as a measure for the objective degree of difficulty of a motion element is considered.

To determine scientifically how the lecture demonstration method compares with the individual-laboratory method of teaching, when geometric material is presented as a laboratory science, is the problem. The pro cedure consisted of forming two parallel groups of pupils which were nearly alike in mental equipment as measured by those factors which were considered likely deter minants for successful achievement in school subjects, and which were about at the same point on the learning curve in terms of achievement in geometry. The group which was taught by the lecture-demonstration method was designated the control group and the one taught by the individual-laboratory method was designated the experimental group.

This paper analyzes the educational process, using graphical methods wherever possible. In its simplest aspect education is represented by a growth curve. If restricted to a single subject the curve is a two-dimensional one and may include both learning and forgetting. The mechanism of learning appears to be quite different for accumulative subject matter such as mathematics or physics from that in nonaccumulative fields such as political theory or business management. Combining the learning curves for several subjects one obtains an educational pattern which may be directly related to particular job requirements or to college admission requirements.

1. When an octopus is faced repeatedly with the situation crabattached-to-a-bar there are progressive changes in the characteristics of the attacks. 2. The total time to attack decreases and also the variation between times for attacks. 3. The reduction in time is due partly to reduction in the time of construction of the command before the animal emerges from the home (“first time delay”). 4. There is also an elimination of the slow initial part of the attacks that is at first present (“second time delay”). 5. There is little change in the duration or speed of the final dash by which the prey is captured (“the final pattern of acceleration”) 6. The stage of learning reached is a function of the number of occasions of training and not of the number of days in the tank. 7. Days without food distort the learning curve but do not accelerate its progress. 8. The changes in line of attack are thus the expression of a positive learning process and not simply of “acclimatization” to the new environment.

To obtain evidence concerning the dependence of choice probability upon anticipated reward, as opposed to the cumulative effects of previous rewards, college student subjects were run on two successive sets of two-choice, differential reward problems under a noncorrection procedure. Relations between the two sets were such that transfer from first to second set learning could occur only via the learning of probabilistic contingencies between reward values. Significant transfer was observed, and error frequencies on the second problem set were monotonically related to differences in anticipated reward values rather than to differences in reward magnitudes actually received for particular responses. A four-state Markov model, embodying the assumption that reward values affect performance, especially in the intermediate states, but not the rate of associative learning, accounts reasonably well for the observed learning curves and several statistics of error data for the second problem set.

Several sets of learning data furnished by I. Krechevsky have been analyzed in terms of meaningful parameters of the learning curve, and the changes in the frequency distributions of these parameters with changes in the experimental conditions have been studied. One of the parameters represents the animal's initial preference for the light or dark, the other represents learning ability. The analysis shows that destruction of about ten or fifteen per cent. of the cortex, increases the animal's preference for the light and decreases the learning ability slightly. By ordinary methods of analysis, it is not possible to discover thatboth initial preference and learning ability have been changed by any given factor.

Learning curves of ten groups of male subjects and ten groups of females (16 cases of constant age within each group) were obtained on two large muscle motor coordination tasks. One of these required the maintenance of immobility of a stabilometer; the other involved climbing a free-standing ladder. The theoretical curves fitted to the data were of the exponential type, yielding mathematical coefficients for rate as well as amount of learning in addition to measures of initial skill and plateau level. Variance analyses showed that rate of learning was not influenced by age or sex over the range 6 to 26 years. The amount of learning was unrelated to sex; it was observed to increase during the adolescent period. Performance level (skill) varied considerably with age and was relatively poor in postadolescent females.

n/a

Robbi is an ER 56 computer programmed to play a simple nXn (2≦n≦9) board game against human opponents or against itself, that is, to play both sides of the game. Of special interest in the playing program is the learning routine, which improves the performance of the system as it gains experience with the actual game. The efficiency of the machine-learning procedure combined with various tactics adopted by the players has been studied in some detail. The results of each run are tabulated, i.e. the number of plays and the machine-playing time necessary to achieve the criteria of learning, the number of losses, the number of corrections, and the number of board positions stored in the memory of the computer during a run together with their associated lists of possible alternatives, the number of which is reduced by those which have been proved to be incorrect responses. Both the efficiency of the learning system and the characteristic form of the learning curves depends on the combination of tactics used by the players.

Ss were required to select previously exposed pictures of common objects from among series of alternative pictures graded in similarity to the prototypes. Response frequencies were plotted in the form of generalization gradients, and such gradients were obtained following 4 stages of training and 3 retention intervals. In Part II, Ss were trained by exposing the same prototype stimuli, but recognition tests consisted of alternatives at 1 of 3 homogeneous levels of similarity to the prototypes. Learning curves based upon the 3 types of tests differ markedly in slope, reflecting the differential sensitivity of various dichotomous tests to the changes in the discriminability function. It was shown that the slope of each curve could be predicted accurately from the gradients obtained in Part I. Thus, generalization gradients were shown to be sensitive, parsimonious representations of the recognition learning process. The purpose of this study was to obtain a series of generalizatio n gradients to reflect the acquisition and retention of a visual discrimination skill. A number of previous investigators (Hovland, 1937; Jensen & Cotton, 1961; Margolius, 1955; Razran, 1949) have obtained gradients of stimulus generalization at several stages of training. As Brown (1965) and Prokasy and Hall (1963) have pointed out, however, such gradients do not necessarily reflect discriminability functions.

This paper considers the transfer of tracking performance in manual control as a function of plant dynamics, where four different plants, i. e., 1/s, 1/s (s+1), 1/s (s+0.1), and 1/s2 are chosen. Using a one-dimensional compensatory tracking system with random function input, the transfer is investigated according to the response patterns, learning curves and indices of transfer.Going from an easy task to the most difficult task (plant 1/s2) yields positive transfer when the two tasks vary little in difficulty, but it does not yield remarkable transfer when the two tasks are widely different in difficulty. Going from a difficult task to the easiest task (plant 1/s) always yields negative transfer regardless of the difference of two tasks.

Summary The statistical relation between labour-productivity and production. Rising labour productivity can be induced by increases in output, or result from “autonomous” factors. Many induced increases will be irreversible, in particular those associated with “learning” processes. Under conditions of continuous economic growth the learning curve hypothesis leads to a simple exponential relation between labour demand and output in the long run. In the short run, current and past profits and the profitability of import substitution join output as explanatory factors of labour demand. In the long run, the effect of growth of capital must be taken into account. Autonomous factors can be represented by an exponential trend. With United States data, increasing returns to scale are still found. Under conditions of a constant savings ratio and continuous population growth a production function, which takes account of all these factors, can be shown to be compatible with a constant labour-output elasticity, however.

The startup or learning curve literature has in the past concentrated mainly on the algebraic power function or on versions based on this function. This concentration is not unusual because the power function has proven, in numerous studies, to fit empirical data quite well. However, other easy-to-apply algebraic functions should also be analyzed and considered. One such function, an exponential function, is discussed and is shown to provide a better fit to several sets of empirical data than the traditional power function. In addition to providing a better fit, this exponential function is easy to apply in practice.

A look at how behavioral objectives are related to learning curves. Definitions and major uses are noted. Havighurst's development tasks are presented as a model for providing better retention of learning. Also includes a brief look at the “infant science” of Psychometrics.

Summary. 100 boys and girls of 13+ years of age were given repeated trials on a simple musical theme. The performance on eight consecutive trials was related to performance on a battery of musical and other tests and the factors underlying performance were elicited. It was found that the task was complex and changed in complexity during the eight trials. The learning curve for a hypothetical performer was broken down into a series of other curves corresponding to the factors revealed in the earlier part of the study. It was found that these curves did not follow the overall learning curve although they behaved in a generally regular manner.

T HIS paper presents evidence derived from engineering data on economies of scale in manufacturing and processing plants. The evidence presented here relates solely to production cost. Costs associated with factors such as marketing, general overhead, transportation, dispersal of market, and raw materials are excluded from this study. The chief purpose underlying this research was to provide some empirical justification for investigations into the nature and extent of investment barriers created by economies of scale in less developed countries.' In light of this purpose, certain dynamic phenomena (such as learning curves) that have a scale effect will also be discussed briefly. Our main conclusion is that in many basic industries, such as petroleum refining, primary metals, and electric power, economies of scale are found up to very large plant sizes (often the largest built or contemplated). These economies occur mostly in the initial investment cost and in operating labor cost, with no significant economies observed in raw material cost. Scale economies can also result from learning curve effects, spreading of set-up costs, and certain stochastic processes associated with inventories. With some reservations, we feel that

G. A. Miller and W. J. McGill showed that free recall of verbal material could be analyzed in terms of a probability model. The model assumes that the probability of recalling a word on a given trial is completely determined by the number of its recalls on the preceding trials.Although the model holds only in a limited case of learning, it seems worthy of special mention, because the model describes the data from every aspect we can think of and in almost all aspects the curves derived from the model fit the data very well.As they themselves admitted, however, the two authors' method of estimating parameters, p0 and a, was not satisfactory. Hence, the present author used the maximum likelihood estimates for p0 and a which were obtained from the frequency distribution of the number of non-recalls for the word after it has been recalled k-times on the preceding trials (Equations 7 and 8). The table to be used in the estimation was given in an appendix. That is a supplement to the table already available (2) in the case where both p0 and a are larger than 0.5.Materials in Miller and McGill's experiments were simple English words read aloud to the subject. The order of the words in a list was randomized from trial to trial. In the first experiment of the present study, Japanese nonsense syllables were visually presented to the subject one at a time and results similar to Miller and McGill's were obtained. In other words, our data were well fitted by their model.The maximum likelihood method of estimation of the parameters can be also applied to each word by substituting m=1. In the second experiment, serial learning was done for five lists, each of which consisted of 20 nonsense syllables of 2 Japanese letters. The procedures were almost the same as those in the first experiment except that the word order in each list was fixed throughout the trials. A list was presented repeatedly until it was reproduced without errors, and a and p0 were estimated for each word. Then averages of a's and p0's for the five lists were computed and the a and p0 were expressed as functions of the position in the list. By applying the method of moving averages (span of 3) it became clear that both the initial and final effects appeared primarily in the p0-curve and slightly, if any, in the a-curve (Fig. 4).As trials were repeated, a kind of learning occurred in the skill of remembering and this was apparent in a but not in p0 (Fig. 6). In some cases, learning curves were S-shaped and in others, they had no inflection point. The latter shows only negative acceleration. It is to be noted that from the model under discussion the curve with no inflection point can be derived as well as the S-shaped curve by putting an appropriate set of values into a and p0 (Fig. 7).As to the data which were fitted by Miller and McGill by highly complex equations involving three parameters, it was shown that they could also be successfully explained by simpler equations with two parameters.