542
"IHEMA 14
zusammen, dann lassen sie sich auf ein einziges Prinzip zur~ickffihren, n~indich das der Erreichung eines optimalen lnformationsflusses zwischen Umwelt und Organismus (unter jeweiliger Beriicksichtigung der nicht ver~inderlichen Or6ssen). Die Konstz~aktion solcher ,,Modelle" war Saehe der lnformationstheorie, die konkrcte Anwendung Sache des psyehologischen Experiments, die Erklilrung wartet auf den Physiologen, der dic (unangcnchme) Aufgabe hat, die gefundenen Resultatc als zwangsl'aufige Folgen ciner im Organismns befindlichen ,,Kolloid-P.cchenmaschinc" abzuleitcn.
I N F O R M A T I O N THEORY, R E C O G N I T I O N AND R E C A L L N. S. SUTHERLAND Oxford (Englarld) Two aspects of some work in progress on memory in rote learning tasks will he reported. The first is the development of some mathematical technlqucs whictl enable us to measure information Iransmitted in performance of such tasks. The second is the result of an experiment comparing the efficiency of recognition and recall in terms of information transmitted. (1) In the past it has been difficult to compare performance under different conditions in this type of task, because no rational measure of performance was available. For example it was impossible to say whether, if a subject is presented with a list of 15 items and recalls 10 correctly and 3 incorrectly, he performs better or worse than a subject who recalls 9 correctly and none incorrectly. It is clear that if we could measure the information transmitted under these conditions we would be able to compare such results meaninbffnlly. A second type of prohlem which has been raised, but not solved because of inadequate measurement techniques, is whether there is a difference between the system responsible for storing items in memory and the system responsible for storing the order of items: thus we cannot measure whether recall of order decays faster than recall of items irrespective of order. It will be shown that it is possible to develop information measures of: (a) The amount of infonnation transmitted when a subject recalls items independently of their order; and (b) Tl~e amount of information transmitted by the order in which a subject arranges a series of items inde-
WAHRNEHMUNG UND INFORMATIONSTI'IEORIE
543
pendently of information transmitted by the selection of the items themselves. These measures make possible direct comparison of performance under widely different conditior~s of recognition and recall, and should enable us to investigate many problems which were previously impossible to solve because of the lack of adequate techniques of measurement. (2) Measure (a) above has been applied to the results of an experiment on recognition and recall. It was hypothesised that the reason why subjects normally recognise items more readily than they recall them, is that in recognition they are selecting from a smaller ensemble. An experiment was performed on recognition and recall in which the number of alternatives from which subjects could select in both conditions was corttrolled. The information transmitted under different conditions was theu computed, and it was found that subjects transmit approximately the same amount of information in recall as in recognition.
HOW LONG IS A SENTENCE? GEORGE A. MILLER Cambridge, Mass. (USA)
The deveIopmcnt by Shannon in 1948 of a new unit to measure the amount of information in verbal messages led psychologists to ask whether the "bit" is also an appropriate unit for measuring psycholinguistie phenomena. The measure seemed to have interesting properties for describing perceptual and perceptual-motor phenomena, but its application to data for information storage seems less direct. In 1950 Miller and Selfridge reported that Ss can repeat more words from a passage that approximates customary usage than they can when successive words are chosen at random. Since random sequences of words are ]es,; redundant, the results r~li~t be interpreted to mean that the amount of information stored ~ p r o x i m a t e l y invariant, independent of coding. However, such an intd~,~tation conflicts with other evidence; e.g., Pollack's 1952 report that t1"ie span of immediate memory is not a constant number of bird. Thus perception seems to depend upon the amount of information in the stimulus materials, whereas memory, as measured by recall, does not; in 1958 Teghtsoonian demonstrated that recognition memory fell somewhere in between, depending upon the cognitive strategy that S adopted.