Learning 10000 pictures (1973) [pdf]

Quarterly Journal of Experimental Psychology (I 973) 25, 207-222 LEARNING 10,000 PICTURES LIONEL STANDINGt Department of Psychology, Bishop’s University, Lennoxville, P.Q., Canada Four experiments are reported which examined memory capacity and retrieval speed for pictures and for words. Single-trial learning tasks were employed throughout, with memory performance assessed by forced-choice recognition, recall measures or choice reaction-time tasks. The main experimental findings were: (I) memory capacity, as a function of the amount of material presented, follows a general power law with a characteristic exponent for each task; (2)pictorial material obeys this power law and shows an overall superiority to verbal material. The capacity of recognition memory for pictures is almost limitless, when mea- sured under appropriate conditions; (3) when the recognition task is made harder by using more alternatives, memory capacity stays constant and the superiority of pictures is maintained; (4)picture memory also exceeds verbal memory in terms of verbal recall; comparable recognition/recall ratios are obtained for pictures, words and nonsense syllables; ( 5 ) verbal memory shows a higher retrieval speed than picture memory, as inferred from reaction-time measures. Both types of material obey a power law, when reaction-time is measured for various sizes of learning set, and both show very rapid rates of memory search. From a consideration of the experimental results and other data it is concluded that the superiority of the pictorial mode in recognition and free recall learning tasks is well established and cannot be attributed to methodological artifact. Introduction Human memory can store both abstract information (letters, words, numbers) and concrete stimuli (objects, scenes, sounds). Abstract memory can involve only that limited number of stimuli which through the subject’s prior experience have already acquired a high degree of symbolic meaning. However, although the possible range of concrete stimuli appears to be much wider, nearly all studies of memory are traditionally confined to abstract material drawn from this restricted set. The neglect of concrete learning tasks is unfortunate because experimental evidence suggests that picture memory, which represents one form of concrete learning, is a strikingly efficient process. Shepard (1967), extending a similar study by Nickerson (1965), has found that immediately following a single exposure of 612 picture stimuli, for about 6 s each, subjects could select the correct picture in two-alternative recognition tests with 98% success. (Similar tests using single words and short sentences as stimuli, produced 90% and 88% success, respectively.) Pictures also show excellent retention over time in memory, as Nickerson (1968) has demonstrated. Seeking the limits of picture memory, -f Now at Psychology Department, University of St Andrews, Fife. 207Downloaded by [University of Washington Libraries] at 16:46 15 March 2013 208 LIONEL STANDING Standing, Conezio and Haber (1970) gave subjects a single presentation of a sequence of 2560 photographs, for 5 or 10 s per picture. Their subjects then scored approximately 90% correct with pairs of photographs (one previously seen, one new), even when the mean retention interval was 1.5 days. T h e first three experiments reported here examine systematically the storage capacity of picture memory, while the fourth examines the rate at which material may be retreived from it. A major consideration in each case is to evaluate picture memory against the standard of verbal memory. I n each case the subjects were experimentally naive students (non-psychologists), aged 18-25. T h e subjects were paid on an hourly basis. Experiment I I t is desirable first to examine picture memory capacity in more detail than has been achieved by Nickerson, Shepard or Standing et al. in order to find the general relationship between the number of stimuli presented and the number retained in memory, which cannot be inferred from these studies. This experi- ment therefore follows the basic pattern of the above studies but systematically varies the number of stimuli shown to the subject over a wide range, the dependent variable in each case being the number of items that are retained in his memory. T h e experiment compares recognition memory for pictures (both vivid and normal) with that for words. Method Stimuli These were 35 mm transparencies; 92% were coloured. The method of assembling this population from various sources is outlined by Standing (1971~2). No simple metric exists for specifying picture stimuli; however, these photographs may be characterized as resembling a highly variegated collection of competent snapshots, and will be referred to subsequently as Normal pictures. Another population, of 1200 striking pictures, was selected from an original pool of about fifteen thousand photographs by three judges. This process of selection is des- scribed by Standing (1971b). I n brief, slides with definitely interesting subject matter (with or without technical excellence) were placed in this population, which will be re- ferred to as Vivid pictures. Ordinary photographs of dogs appeared in the Normal pop- ulation, but a picture of a dog holding a pipe in its mouth was assigned to the Vivid category; an aeroplane was generally Normal, but a crashcd place Vivid. A population of Word stimuli was produced by selecting English words randomly from a Merriam-Webster dictionary and printing them on 35 mm slides. This dictionary gives the 25,000most common English words (e.g., salad, ton, station, landholder, cotton, zoology, camouflage, reduce, well-worn, somehow). A population of 11,000photographic slides was first assembled. (The term is used simply as a convenient label.) Procedure test. photos comprised the stimuli; the group size was always 10in the latter case. was performed under group conditions. and to try to learn them in preparation for a memory test. The procedure involved a single-trial learning task, followed by a delayed recognition A different group of 5 subjects was used for each learning task, except when Vivid All testing Before the learning task, the subjects were instructed to attend closely to the stimuli, The importance of maintainingDownloaded by [University of Washington Libraries] at 16:46 15 March 2013 LEARNING 10,000PICTURES 209 strict concentration even during long sequences of stimuli was strongly emphasised by the experimenter. The subjects were then shown a set of stimuli randomly selected from one of the three populations (Vivid pictures, Normal pictures, or Words). This learning set was presented once only, at 5 s per item and with an interstimulus interval of 600 ms. A commercial Sawyer projector controlled by Hunter timers was used to present the slides under dark- room conditions ; they subtended a maximum visual angle of approximately 14’. On each trial two stimuli were presented side by side; one had been randomly selected from the learning set and then randomly allocated to the left-hand or right-hand position, while the other was new to the subject and had been selected randomly from the appropriate one of the three stimu- lus populations. Each subject wrote down an “L” or an “R” on each trial to indicate whether the left- or the right-hand stimulus looked most familiar to him. This task was forced-choice; unlimited time was allowed, but each trial in practice usually required only a second or two. The size of the learning set for each type of material was 20, 40, 100, zoo, 400,or 1000 stimuli ; however when Normal pictures were used, additional groups were tested with 4000 and 10,000items. The recognition test consisted of 80 trials, except when 4000 or 10,000 items were learned (160 trials), or with learning sets of twenty or forty (when 20 or 40 recognition trials were given). Rest pauses of 4 min were given during the learning task after every 200 items, and a I h break after 1000items, where applicable. While the recognition test was generally given exactly two days after the learning task, when 4000 or 10,000stimuli were used only the average retention interval could be set at z days, since only 2000 slides per day were shown in these cases. This was achieved in the former case by testing recognition 1.5 days after the second learning session, and in the latter by giving the recognition test immediately after the fifth daily learning session. Two days later the subjects performed a recognition test. Results and discassion The mean number of errors occurring in the recognition task under each con- dition is given in Table I. Also given is an estimate of the number of items ( M ) that have been retained in memory in each case; making the usual guessing correction, this is calculated as S(T-zE)/Twhere S is the size of learning set, E is the mean number of recognition errors, and T is the number of recognition test trials. TABLEI Mean errors in recognition test, Experiment I (standard deviations in parentheses). Material S T Vivid pictures Normal pictures Words 20 40 I 0 0 200 400 4000 I 0 0 0 10000 20 40 80 80 80 80 160 I60 20 0‘2 (0.45) 40 1.8 (1.1) 96 4.0 (2.2) 190 6.8 (1.3) 381 11.4 (5.8) 880 9.2 (3.0) 30.2 (16.4) 27.2 (6.1) 19.6 2.0 (1.2) 16 36.4 5.6 (1.3) 28.8 90 1 2 (1.9) 70 166 16.8 (4.5) 116 286 16.6 (7.1) 234 770 15’4 ( 5 . 5 ) 615 -2490 6600 - S is the number of stimuli presented in the learning set; T is the number of recognition test The third value within each cell is the estimated number of items retained in memorytrials. ( M ) . Each cell is based on 5 subjects (10 for Vivid pictures)Downloaded by [University of Washington Libraries] at 16:46 15 March 2013 210 LIONEL STANDING The mean number of items in memory (M> was then plotted against the number of stimuli presented (8)for each type of material, using log-log co-ordinates as shown in Figure I . 10,000 1000 2 IOC IC / 1 40 I00 400 1000 4000 10,000