PSYCHOLOGICAL SCIENCEResearch ArticleREFLECTIONS OF THEENVIRONMENT IN MEMORYJohn R. Anderson and Lael J. SchoolerDepartment of Psychology, Carnegie Mellon UniversityAbstract—Avallahility of human memories for specific itemsshows reliable relationships to frequency, recency, and patternof prior exposures to the item. These relationships have defieda systematic theoretical treatment. A number of environmentalsources /New York Times, parental speech, eleetronic mail}are examined to show that the probability that a memory will beneeded also shows reliable relationships to frequency, recency,and pattern of prior exposures. Moreover, the environmentalrelationships are the same as the memory relationships, ft isargued that human memory has the form it does because it isadapted to these environmental relationships. Models for boththe environment and human memory are described. Among thememory- phenomena addressed are the practice function, theretention function, the effect of spacing of practice, and therelationship between degree of practice and retention.The title of our paper is inspired by the following remark inShepard (1990): "We may look into that window on the mind asthrough a glass darkly, but what we are beginning to discernthere looks very much like a reflection of the world" (p. 213).He was commenting on how the principles of perception areexquisitely tuned to the features of the environment in whichwe live. Basically, Shepard's thesis is that perception has beenoptimized through evolution to make the best possible infer-ences about the world given the perceptual input. Recently,Anderson (1989. 1990) has suggested that the same might betrue about human memory.Many people hold the bias that human memory is anythingbut optimal. They point to the many frustrating failures of mem-ory. However, these criticisms fail to appreciate the task beforehuman memory, which is to try to manage a huge stockpile ofmemories. In any system responsible for managing a vast database there must be failures of retrieval. It is just too expensiveto maintain access to an unbounded number of items.Given the initial bias against human memory, it would beparticularly compelling if we could show that human memorywere optimal. How does a system behave optimally when it isfaced with a huge data base of items and cannot make all ofthem instantaneously available? It would be behaving optimallyif it made most available those items that were most likely to beneeded.In this paper we explore the issue of whether human mem-ory is behaving optimally with respect to the pattern of pastinformation presentation. Each item in memory has had somehistory of past use. For instance, our memory for one person'sname may not have been used in the past month but might havebeen used five times in the month previous to that. What is theprobability that the memory will be needed (used) during thecurrent day? Memory would be behaving optimally if it madethis memory less available than memories that were more likelyto be used but made it more available than less likely memories.In this paper we examine a number of environmental sourcesto determine how probability of a memory being needed varieswith pattern of past use. However, we first review how avail-ability in human memory varies with pattern of past use. Someaspects of this problem have been extensively studied in em-pirical studies of human memory.FORM OF THE MEMORY FUNCTIONSTwo of the most basic statistics we might gather about pat-tern of past use are how often a memory has been practiced andhow long it has been since it was last practiced. Learning func-tions and retention functions to describe these two aspects ofhuman memory have been collected since the original experi-ments of Ebbinghaus (1885/1964). Figure 1 shows the retentionfunction and practice function obtained by Ebbinghaus.The Retention FunctionEbbinghaus measured retention in terms of the percent sav-ings in relearning a list of nonsense syllables. The functionshows the classic negative acceleration typical of such retentionfunctions. In order to be able to compare this memory functionto the environment, we need to decide how to characterize theforgetting function. Some (e.g., Loftus, 1985) have suggestedthat these functions satisfy an exponential formula:P = Ae-bT(1)where P is the performance measure, T is the delay time, and Aand b are parameters of the model. The intuitive appeal of anexponential function probably explains why it is so often sug-gested. It implies that during each unit of time, the memoryloses a constant fraction of what is left. This process evokesimages of radioactive decay, an analogy often used to describeforgetting. One can investigate whether this function holds byperforming a log transformation of the performance scale. If theunderlying relationship is exponential, a linear relationshipshould obtain between log performance and time:log P = Jog A - bT.(2)A precondition to performing an adequate test of such a func-tion is that we have a large manipulation of the time scale.396Copyright *& 1991 American Psychological SocietyVOL. 2, NO. 6, NOVEMBER 1991PSYCHOLOGICAL SCIENCEJohn R. Anderson and Lael J. Schooler60 r50(a) Ebblnghaus's Retention data605040S 3020-10200 400 600Hours of DelayBOO(b) Ebbinghaus's Practice Data2 3 4 5Days 01 PracticeFig. 1. (a) Ebbinghaus's (1885/1964) retention function showing percent savings as a function of delay.Ebbinghaus used delays from 20 minutes to 31 days, (b) Ebbinghaus's practice data showing total number oftrails to master a set of lists as a function of number of days of practice.Ebbinghaus's data certainly satisfy this precondition, as he var-ied retention intervals from 20 minutes to 31 days.Figure 2a illustrates the Ebbinghaus data with the perfor-mance scale transformed. As may be observed, the resultingfunction is anything but linear. Thus, despite its popularity, thehypothesis of an exponential forgetting function is not sup-ported. Wickelgren (1976), using a d' memory measure and de-lays from 2 minutes to 14 days, found evidence for a powerfunction relating delay to retention.' A power function has theform:P = AT (3)1. Actually, Wickelgren's theory also had an exponential componentthat would dominate the power component at very long delays.4.24.03.8 u(a) Ebbinghaus's Retention Datawith Log Transformation ofthe Performance Scalelog P = 3.B62- 0.126 log DR'V2 = 0.9764.24.03.6200 400 600Hours of Delay8003.43.23.0(b) Ebbinghaus's Retention
View Full Document
Unlocking...