Transfer of training

For more information on transfer of training, visit Britannica.com.

The effect of one form of training on another form of training. See also transfer of learning, transfer principle.

It is often evident in transfer situations that a very considerable amount of learning has been carried from one task to another. For both theoretical and practical reasons, it is desirable to be able to measure the amount of this transfer.

Let us consider a typical transfer experiment or situation. An experimental group of subjects learns some task or skill A. Initially, as we might expect, their errors are many, but after some time, or number of trials, they reach a satisfactory and stable level of performance characterized by few errors. They then attempt some new task B, and again proceed, more or less rapidly, from making many errors to making only a few. (The mean performance of this group is shown, smoothed for the sake of argument, as the clear line in Fig. 1, whose notation will henceforth be used.) A control group has learned task B ab initio (dark in figure), and it is generally the case that the final error levels of performance of the two groups (S and S¹) are virtually indistinguishable — and may, indeed, be zero.

All the quantities labelled in the figure can be measured. F, C, L, T, S, and S¹ are all measures of performance in terms of error, while m, n, and r are either numbers of trials or elapsed time. Since all these are available, it is natural to ask how they may best be used and combined to express the amount of learning transferred by the experimental group from task A to task B.

The number of expressions and formulae that have been proposed is very large (Gagne, Foster, and Crowley 1948, Murdock 1957), but they all fall into one of two categories: there are those that measure the saving of training time, or number of trials (i.e. which in some way compare r with n), and those that are concerned with the initial performance of subjects immediately after transfer (i.e. which evaluate T). The former class we may call savings measures, the latter we may call first-shot measures. Unfortunately, although the contrary was often tacitly assumed, these two classes of measures can give sharply contrasting impressions, for it is quite possible to save a great deal of training time on the second task (r ≤ n) while still finding a marked transient decremental effect of transfer (T ≥ C) (Hammerton 1967, Hammerton and Tickner 1967). The measure selected must answer the question the user wishes to put, and he must always be aware that it may not answer any other question.

The simplest question is: what proportion of training time in task B is saved by prior training on task A? This is clearly and straightforwardly measured by the quantity a, given by:

a = (n − r)/n

(1)

It is worth pausing to note how difficult it is to apply even the simplest psychological findings in a practical situation. Suppose someone was being trained to use some complex industrial equipment (task B) by practice on a simulator (task A). The proportional saving of time is given by (1), but in practice the following questions arise: what is the running cost per hour of the simulator as compared to that of the real equipment? Does even a poor operator of the real equipment produce some useful output which can be set against cost? Is he, on the other hand, likely merely to waste raw materials or damage the equipment itself? Evidently the practical utility of a training programme or device cannot be measured by any simple expression, but is a complex psychological and economic question.

If we recall that there is often a brief but marked impairment in performance immediately after transfer, it will be evident that there are many practical situations in which some measure of first-shot transfer is necessary. Again, more than one question may be asked about this. One is: how does the learning immediately after transfer compare with that required by the control group to reach a stable performance? This comparison may be given by:

b = (F − T)/(F − S)

(2)

But a user may raise a practical objection to this, for he may not be concerned with the comparative measurement of learning, however interesting that may be theoretically. He generally needs to know whether expensive equipment is liable to be misused when first handled: in other words, how does first-shot transfer performance (T) compare with the stable performance of the control group (S)? The information required may be given as a simple percentage:

c = 100 (1 − T/S)

(3)

Another question which can arise in practice is this. Suppose that you have no control group — i.e. suppose that all trainees start on some simulator device, and that no man is allowed to begin with task B. How much of the skill acquired with the simulator (A) will be retained upon first transfer? This is not the question answered either by (2) or by (3) above, and it is more precisely answered by:

d = (F − T)/F − L)

(4)

However, in pursuit of understanding, a quite different question may be put, namely: how does the learning shown by the first-shot performance of the experimental group compare with that acquired by the control group? A measure for this is:

e = (C − T)/(C − S)

(5)

It is by no means claimed that the expressions given here are the only, or necessarily the best, answers to the questions discussed, and certainly many other questions can be put, of both theoretical and economic significance. As we have shown, it is important to be clear about what question is being asked, and especially important to be clear about the extent and limitations of the information obtained from the answer. For example, an answer to a question on saving of training time — such as (1) — gives no information at all about first-shot performance: the one may be excellent and the other deplorable. In one particular case (Hammerton and Tickner 1967) the several expressions given here yielded values (rendered as percentages) from 92 to minus 54. If any one had been taken as 'the' measure of transfer, the oversimple question 'How much learning was transferred?' could have received any answer in the range from 'almost all' to 'less than none'.

Evidently there is no such thing as 'the' measure of transfer, but if a specific question concerning transfer is precisely formulated, expressions exist for answering it, and a useful and informative numerical measure can be obtained.


Fig. 1. Curves showing form of typical transfer experiment.


(Published 1987)

— M. Hammerton

Bibliography
• Gagne, R. M., Foster, H., and Crowley, M. E. (1948). 'The measurement of transfer of training'. Psychology Bulletin, 45.
• Hammerton, M. (1967). 'Measures for the efficiency of simulators as training devices'. Ergonomics, 10.
• — —  and Tickner, A. H. (1967). 'Visual factors affecting transfer of training from a simulated to a real situation'. Journal of Applied Psychology, 51.
• Murdock, B. B. (1957). 'Transfer designs and formulas'. Psychology Bulletin, 54.

---

Note: click on a word meaning below to see its connections and related words.

The noun has one meaning:

Meaning #1: application of a skill learned in one situation to a different but similar situation
  Synonyms: transfer, carry-over

---

This entry is from Wikipedia, the leading user-contributed encyclopedia. It may not have been reviewed by professional editors (see full disclaimer)