Procedures of Mind

JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR 23, 425--449 (1984)

Procedures of Mind

PAUL A. KOLERS* AND HENRY L. ROEDIGER IIIt *University of Toronto and ~Purdue University

The prevailing metaphor for studies of learning and memory emphasizes the acquisition, storage, and retrieval of "information"; within this framework, mind is often treated as if it were a physical object and information similarly is assumed to have physicalistic prop- erties. Evidence that supports a more process-oriented view of information processing is offered. Mind is described in terms of skill in manipulating symbols and the notion of skills is shown to provide a useful framework for accounting for significant aspects of cognitive processes. Evidence supporting the procedural view includes studies that show that the means of acquisition of information form part of its representation in mind, that recognition varies with the similarity of procedures in acquisition and test, and that transfer between tasks varies with the degree of correspondence of underlying procedures.

Thirty years ago the central textbook of and its operations. In the present paper we human learning was McGeoch and Irion's first survey some of the assumptions that (1952) The Psychology of Human Learning. underlie modern work in learning and The primary concerns of the book were the memory and point out what we consider to acquisition, retention, and transfer of be some of their weaknesses--particularly verbal lists. There is almost nothing in the the assumption of static representations. book about mind, mentation, information We then propose an alternative view based processing, cognition, or even memory. on the functions or activities of mind as The contemporary scene is different. The forming the basis of knowledge. In the end intervening years have seen the develop- we apply this view to a number of memory ment of richly elaborated models of mind, phenomena of interest to contemporary in- information processing, language, and the vestigators. Many of the ideas expressed in like--a development that might be consid- these proposals, and the theory of symbols ered almost a reaction to the spare and cog- on which they are based, were developed nitively desiccated notions of McGeoch more formally in a previous paper (Kolers and Irion. & Smythe, 1984). Here we are more con- We believe that the reaction may have cerned with empirical justification of the gone too far and that current concerns with claims. cognition and information processing have developed too rich a description of mind CONVENTIONAL ASSUMPTIONS OF We thank the editor, E I. M. Craik, for the invita- MEMORY THEORIES tion to prepare this article. Its preparation was sup- If one asked an intelligent layperson what ported in part by Grant A7655 from National Science memory was, more than likely the answer and Engineering Research Council Canada to P.A.K. and in part by Grant R01 HD15054 from the U.S. Na- would be something like a place in the mind tional Institute of Child Health and Human Develop- where information is stored. This actually ment to H.L.R. Requests for reprints should be ad- is the implicit theory embodied in everyday dressed to Paul A. Kolers, Department of Psychology, language (Lakoff & Johnson, 1980; Roe- University of Toronto, Toronto, Ontario M5S 1A1, diger, 1979). Mental processes are often de- Canada, or Henry L. Roediger III, Department of Psy- chological Sciences, Purdue University, West Lafay- scribed as objects or events in an actual ette, IN 47907. physical space, as when we speak of

storing and organizing memories, of 2. Structure-Process Distinction searching through them, or of holding or Tulving and Bower (1974, p. 265) re- grasping ideas in our minds; like objects, marked that "A long tradition has held that memories may be lost or hard to find, and 'contents' of the mind in some sense exist so on (Roediger, 1980b). (Whorf, 1956, pro- independently of the 'processes' that create posed that such descriptions might be in- the contents, change them, and make use trinsic to European languages.) In this sec- of them. Thus, in studying memory we are tion of the paper we discuss these and some concerned with two broad questions: what other assumptions prevalent in many of the the structure of a memory is, and what pro- theories of memory developed since publi- cesses operate upon it." The structure- cation of McGeoch and Irion's (1952) book. process distinction follows from a notion of Some of these assumptions have a much a trace stored in a place, but it has not al- longer history than is represented by theo- ways been a productive metaphor and may ries of the modern era, but we will confine even be unnecessary, as we show later. our remarks largely to contemporary accounts. 3. Structure When hypothetical structures were first 1. Spatial Storage and Search reintroduced into psychology they were About 75% of the analogies used as used sparingly and only after several lines models of memory assumed spatial storage of behavioral evidence for them were de- and search (Roediger, 1980a), from Aristot- veloped (Garner, Hake, & Eriksen, 1956; le's notion of memory as a wax tablet on Miller, 1959). Such restrictions have often which experience writes, to James's (1890) been abandoned in recent work and mental / analogy between remembering something structures are introduced with little if any and searching a house for a lost object, and supporting evidence. The upshot is that Murdock's (1974) notion of memory as a cognitive psychologists stuff the human conveyor belt on which information is mind with metaphorical structures. A re- stored in packets that recede in time. The cent example, from a text by Dodd and computer analogy to mind is still another White (1982), is reproduced in Figure 1. Ac- example (Simon, 1979). The spatial array cording to the text caption it represents "a and various buffer storage systems of the working general model of the human infor- computer have been co-opted as models of mation processing system." It should be re- the human mind, where "information re- alized, however, that nothing "works" in trieval" is accomplished by search of the such a system; the boxes contain only memory stores (Anderson, 1983; Atkinson names, not descriptions of processes or re- & Shiffrin, 1968; among others). A distinc- lations among them. Kolers and Smythe tion between place and content, common (1984) called such instances nominal to computer structures, is also found in models, in contrast to relational and sys- these models. However, the fact that some tematic models, where more effort is di- structure is used as part of a computer's rected at operationalizing terms. Many operation is not a compelling reason for other instances exist in the current litera- supposing that it appropriately describes ture. For example, in Raaijmakers and Shif- functions in the human mind. Analogy is frin's (1980, 1981) search of associative not by itself an explanation. There is little memory (SAM) model the second of five reason to think that a computer's electronic "guiding principles" is the following state- circuitry provides a hardware model of the ment: "Long term memory is a richly in- brain, and there is no more reason to as- terconnected network, with numerous sume that its software models the mind. levels, stratifications, categories and trees MENTAL OPERATIONS 427

Envi...... i ~ r~*-~ r'~ RESPONSE SENSORy. ,I .... .---L---

_=:___ ...... A

PERMANENT MEMORY ] WORKING MEMORY (Long Term) ~. / I {ShortTerm) S..... it-Concepts, ~1 L-,I Attention Relationships ¢-~1 ] I --I v~ CurrentlyActivated Auditory-Dictionary, Elements .... L Music II ~ Visual Fatterns, Conce Current Subgoal Placein Plan Skills-Language,Motor ] I Processing Programs

Values

SSOR I~ Queueof Goals ~ P,~n, _

CONTROL F~G. 1. An information processing representation of the flow of information in mind. Repro- duced, with permission, from Dodd and White (1980). that contain varieties of relationships, jects or traces deployed in a space leads schemas, frames and associations. Roughly also to the assumption that processes act speaking, all elements of memory are con- upon these structures to accomplish the nected to all others, directly or indirectly work of remembering and thinking. The op- (though perhaps quite weakly)" (1981, p. eration and coordination of these processes 120). If everything is connected to every- is often charged to an "executive," but few thing else, "directly or indirectly," it is not if any constraints are imposed on the op- clear even what needs representing. All eration of the agents, nor does evidence such descriptions are little more than fan- typically require postulating them. ciful nominalizations, irrespective of the detail embodied in the accompanying prop- 5. Memory as Conscious Recollection ositions. Concerned that models are be- The methods used to study memory usu- coming top-heavy with mental structures ally test recall or recognition accompanied that often bear little relation to their data by the conscious experience of recollec- base (Underwood, 1972), some psycholo- tion-the rememberer knows that he or she gists have recently urged theorists to adopt is remembering the experiences in ques- a more functionalist approach (Jacoby, tion. Related investigations that study the Bartz, & Evans, 1978; Watkins, 1981; feeling of knowing experience (Hart, 1965) among others). or the tip-of-the-tongue phenomenon (Brown & McNeill, 1966) also presuppose 4. Process conscious monitoring of mental states. A belief that experience can be repre- These techniques reveal only limited as- sented in mind as a collection of static ob- pects of learning, however; more nearly 428 KOLERS AND ROEI)IGER complete understanding may require the this view a distinct contrast is proposed be- use of other methods. tween mental 'content and mental process; Ebbinghaus's (1885/1964) savings tech- between "information" (in picture, word, nique is one. Kolers (1976) found savings or other medium) and operations upon it. in speed of reading even when subjects Below we review evidence that challenges failed to recognize the reread pages amid these views. others from the same source; other inves- Actually, as Kolers and Smythe (1984) tigators have reported related phenomena bring out, certain forms of experience do (Cohen & Squire, 1980; Tulving, Schacter, not even lend themselves to linguistic de- & Stark, 1982). These examples reveal a scription; they are dense symbols. The per- dissociation between conscious remem- ceptual representation of a tree moving in bering of information and performance the wind might in some theoretical way be measured in other ways, and are discussed attained by appeal to propositional descrip- more fully below. Conscious recollection tion. Smells and tastes, however, which may be of only limited importance to many form so rich a part of personal experience, performances in our daily lives, however may fail even this loose criterion. Neither much it is emphasized in many current the- the mechanisms of smell nor an analytical ories. vocabulary are available for general use; most tastes and smells are referred to by 6. Semantic primacy appeal to other objects, such as, "This An assumption made in many memory smells (or tastes) like a . . . "; and when theories is that the representation of the lin- descriptive terms are used, they are vague guistic meaning of events is primary and and imprecise, such as acrid, sweet, foul, that other aspects of experience are not tangy, foxy, fruity, and the like (Dravnieks, coded with, or as durably as, meaning. 1982; Moskowitz, 1981). It is true that we Thus Craik and Lockhart (1972) proposed can often pick out by touch shapes that we that memory was a byproduct of stages of have seen or pick out by visual inspection analysis ranging from surface features that shapes that we have touched. But it is also were "shallowly encoded" to semantic true that we cannot tell how things will processing of language that was "deeply taste by looking at them, nor how heavy encoded." Similarly, theorists concerned they will be, nor what they will sound like with perception of or memory for con- or smell like--unless, that is, we have ex- nected discourse have proposed that what perienced those alternatives. The point is is remembered over longer time periods is that resort to propositional representation the meaning of the material read, divorced encounters difficulties in dealing with or- from the specific means by which it was dinary perceptual experience. acquired (Anderson, 1983; Just & Car- These six assumptions seem to be in- penter, 1980; Kintsch, 1974; Sachs, 1967, grained in conventional ways of thinking among many others). Others propose that about memory. The alternative ideas pro- faces (and, by extension, other pictures) posed below do not use any of these as- may be conceptualized as if they were sen- sumptions, and yet can account in a general tences or paragraphs, from which a "gist" way for many of the important memory or like precis can be extracted (Bower & phenomena currently being investigated. Karlin, 1974). An even stronger claim made by many A PROCEDURAL ACCOUNT contemporary authors is that all knowledge OF PERFORMANCE is amenable, if not to linguistic description, Research on learning and memory has to a language-like propositional represen- preserved a distinction between motor pro- tation or other such rule-based system. In cesses and processes based on language or MENTAL OPERATIONS 429

"thought," usually with a pejorative impli- common form. We claim that knowledge of cation for the former. The procedural, often objects is specific to the means of experi- regarded as a motor skill, is thought of as encing them. more mechanical and uninteresting; the lin- In asserting our claims we do not deny a guistic or related aspects of performance contrast between procedural and declara- are thought of as more intellective and im- tive knowledge, but rather inquire into its portant. Procedural or skill learning was present relevance to psychology.a A partic- often tested by having people trace stars ular feature of the propositionalist view that or write while seeing their hands in a we dispute is the claim that all knowledge mirror, or the like (Hovland, 1951). Lan- can be represented in the language-like guage skills were tested by having people symbols appropriate to a computer pro- learn passages of text, lists of words, as- gram (as in Newell, 1980). Indeed, An- sociates, and the like. The two enterprises derson (1983) among others seeks to reduce were thought of as separable, although it all procedural knowledge to the proposi- was never made clear how motor processes tional. This approach misses the funda- were to run off without cognition, or intel- mental point, however, for there is no trick lective processes without their motor to propositionalizing. In principle, propo- expression. sitions can be written to any degree of re- Something of the same dichotomy is pre- finement, and without regard to number, to served in the current contrast between de- model any process; it is merely a matter of clarative and procedural knowledge, or de- adding statements, of saying something scription and action. A debate about consists of a and of b and of c and of d and whether they are fundamentally the same of e..., but without necessarily capturing or whether the two forms of expression its structure in an intellectually illuminating differ has exercised a number of investiga- or efficient way. Our proposal is to do the tors of artificial intelligence (Winston, opposite, that is, to accommodate declara- 1977). We note in a related vein that many tive knowledge in operationalizable terms cognitive scientists seek to represent be- of actions--the procedures that charac- havior in a symbol system like that of com- terize a person's acquisition and use of puter programs (Anderson, 1976, 1982; knowledge. Newell, 1980); indeed, the fundamental The contrast between declarative and idea is to reduce knowledge to a set of prop- procedural knowledge was developed by lo- ositions that can be entered into the appro- gicians to describe forms of statements. priate simulation or program. In that sort Some psychologists suppose that the terms of theorizing a sharp contrast is made be- define kinds of memory. Declarative knowl- tween knowledge and its processing, form, edge is identified with semantic memory or expression. (Collins & Loftus, 1975) or, in Tulving's It is our claim that such contrasts are ill- (1983) system, with semantic and episodic founded for psychological theory. We will memory, and procedural knowledge is iden- show in some detail that distinctions between mental representation and mental ~The contrast was introduced in modern times by Ryle (1949) and amplified by Scheffler (1965), among process, between '°symbol'' and "skill," others, as between declarative knowledge (or knowing are of questionable worth for psychology that) and operational knowledge (knowing how). and may indeed actually misrepresent psy- Scheffler's useful discussion is based on language chological processes. People can know forms; for example, upon the way "I know..." is something visually or tactually or motori- completed for statements regarding the value of the square root of 16; or regarding use of a typewriter: the cally, on our view, without the various first takes that and the second takes how to and the kinds of knowing being mappable into each terms are not interchangeable without altering other, let alone being reducible to a meaning. 430 KOLERS AND ROEDIGER

tiffed with "skills," which when they do should play a prominent role in forming the not require "effort" or "attention" are representation of the message in memory. usually thought to run off "automatically" However, during the 1960s and 1970s a (Hasher & Zacks, 1979; Posner & Snyder, number of experimenters claimed that fea- 1975; Wood, 1983). In contrast our aim is tures of language related to appearance, to carry forward the attempt to account for such as sound, typography, or the like, all of a person's capabilities within the were stored only briefly in short-term vi- framework of skills or procedures. We will sual or acoustic memory, whereas the se- argue that knowledge is a matter of skill in mantic aspects were stored in long-term operating on symbols, that the latter are of memory (Bransford & Franks, 1972; Sachs, many kinds, that the kinds are not perfectly 1967). Primacy of the semantic component correlated, and that knowledge is, as a con- was especially visible in the notions of sequence, means dependent. In the next Craik and Lockhart (1972), in the popular section we describe some mental events in concern with semantic relations, and in the those terms. elaborate logical constructions for summar- izing the claims due to Anderson (1976), EVIDENCE FOR A PROCEDURALIST VIEW Anderson and Bower (1973), Collins and Reading a text and learning something Quillian (1969), Kintsch (1974), Norman from it are surely dependent upon "very and Rumelhart (1975), and Smith, Shoben, many of the most intricate workings of the and Rips (1974), among others. 2 human mind" (Huey, 1908/1968, p. 6). This This concentration upon the semantic as- level of complexity of cognitive perfor- pect of language had a particular perversity mance seems to many investigators to re- to it, however. For one thing, its strong quire a fundamentally different account form stripped experience of its perceptual from that assigned to motor skills. attributes and left the person unable to re- On our account, however, cognitive pro- cord in memory the smell of a flower, the cesses may be well accommodated in pro- look of a face, the sound of music, or other cedural terms. We will support our claim such experiences that resist or often exceed by identifying procedural components in the requirements of a language-like descrip- behaviors that are of interest to contem- tion. For a second thing, this view assumed porary accounts of learning and memory. that its dictionary definition was the most The data are derived from experiments important aspect of a word, and alternate using many different dependent measures, ways of relating to words were usually ig- including speed of reading, accuracy of re- nored. In some cases, however, "superfi- port, recognition, recall, perceptual identi- cial" features of words are of the greatest fication, word completion, and lexical de- interest to understanding, and the words cision among them. By "semantic" we mean the relation between a symbol and its referent. This relation is always specific Refutation of Semantic Primacy to a system of symbols. For example, the semantic On a procedural account, knowledge ac- aspect of the word "chair" is not wholly the same as quired through reading or listening depends the semantic aspect of a picture of a chair or of some particular chair or collection of chairs. Our percep- on cognitive skills of the perceiver. Appli- tions in the different circumstances are symbolic, of cation of these skills is often directed at fea- course; the symbols do not overlap perfectly. All of tures of the message that most theorists the instances, moreover, have unique uses and capa- consider superficial--cadence or pitch of a bilities. Some psychologists use "semantic" to mean voice, typography, spacing, or orientation world knowledge or general interpretation, conceived broadly. Such a usage begs the question of the means of a written text, and the like. Thus, ac- of acquisition of particular knowledge and of the re- cording to the procedural view, and in con- lations that develop between systems of symbols, such trast with some other views, these features as between different perceptual modalities. MENTAL OPERATIONS 431 themselves only one of the means by which were only the carrier of information but such features are instantiated (Eisenstein, played no role in its longer term represen- 1979; Friedman & Friedman, 1957; Kahn, tation, and the associated notion that only 1967). meanings but not means were important-- Semantic primacy was refuted by certain were brought to test by Kolers and Ostry sorts of data, as well. An issue of some in- (1974). Building on an earlier study, they terest was the duration of the alleged required college students to read a large memory that recorded superficial features number of sentences singly and, at a later selectively, and students debated whether time, to read those same sentences mixed material in various short-term memory sys- in with an equal number not read before. tems endured for 0.5, for 3, for 10 seconds, On the second occasion the students clas- and so on (see Crowder, 1976). Several sified the sentences as to whether they had studies carried out at about the same time been read before and, if yes, whether their challenged these claims for a temporal appearance was the same as previously. bound on memory for superficial features. Appearance was established by orientation Craik and Kirsner (1974) found that people of the type. On the first reading half the were more likely to recognize words re- sentences were right side up, in normal ori- peated 2 minutes later if they recurred in entation, N; the other half were turned top the same voice than in another voice; and to bottom around a horizontal axis, I. On Kirsner (1973) showed that printed words the second reading, half of each kind of the were more likely to be recognized if they first set appeared in the other orientation, recurred in the same typeface than in alter- making NN, NI, IN, and II pairs, where nate typeface s. Rothkopf (1971) pointed out the letters indicate orientation on each of that people often remembered a number of the two occasions of reading. superficial or "irrelevant" features of On the theory that a superficial feature something read, such as the color of a like typography is stored briefly and only book's binding or the location on a page of meanings are stored in long-term memory, a particular item. Geiselman and Bellezza all of the pairings should have yielded (1976, 1977) demonstrated that people often equivalent results for memory; and on the encoded voice or location of a sound as part assumption that normal typography pro- of an auditory experience. Several other in- vided readier access to the semantics or gist vestigators demonstrated superior memor- of the sentences, those initially in normal ability of orthographic or phonemic fea- typography should have been recognized tures compared to semantic features, more readily on the second reading than Morris, Bransford, and Franks (1977), those that were initially inverted. The out- Fisher and Craik (1977), Stein (1978), and come of the tests yielded results contrary Zechmeister (1972) among them. Fisher to these expectations, however. The find- and Cuervo (1983), using bilingual mate- ings were that recognition of sentences rials, showed that both the language spoken varied greatly with their orientation, and and the gender of the speaker figured in the the sentences initially inverted (II, IN) memory of a passage. In all of these cases were recognized more frequently, and with location, voice, typography, color, and the greater accuracy, than the sentences ini- like aide instances of irrelevant or superficial tially in normal orientation (NN, NI). In ad- features, compared to the allegedly deeper dition, both inverted and normally oriented semantic features, and whose existence in sentences were recognized better when memory is somewhat incompatible with they reappeared in the same orientation theories of semantic primacy. than in the other one (II better than IN, NN These issues--the duration of short-term better than NI). The interval of time be- memory, the notion that superficial features tween first and second readings was not 3 432 KOLERS AND ROEDIGER or 10 or 30 seconds, moreover, as in the (Kolers, 1975a). But this issue too is com- studies of the duration of "short-term plex. memory," but ranged from about 1 minute The notion of disparity of the stimulus to 32 days. requires careful definition. Not all differ- The data were accounted for in terms of ences matter, and too much detail can in- analytical activities and procedures. The capacitate the person by overloading the proposal was that information is acquired system with masses of poorly referenced from an object by means of pattern ana- analysis. For example, if skilled readers of lyzing operations that are trained through English who are ignorant of a particular for- successive encounters with like objects, eign script are asked to scan or examine it, and that change with acquisition of skill in they find its disparities too many or too de- a particular analysis. The less skilled a tailed to encode usefully, and subsequent person is, the more detailed analytical work recognition suffers (Kolers, 1974; see Gold- he or she must do to acquire the stimulus; stein & Chance, 1971, for related observa- the more analytical work done, the more tions). The way that memorability changes extensive the person's representation of the as a function of a person's acquiring skill stimulus. Normally oriented text is almost with a symbol set remains to be worked transparent to the skilled reader in respect out. In light of experts' performance, it to its familiarity as a visible pattern; in- seems plausible that memorability is a U- verted text, by virtue of its lesser famil- shaped function of skill, increasing again iarity, requires substantially more analyt- as skill becomes expertise. Studying ical work from the reader than normal text transfer of training can play an important does. The assertion is that disparities in fa- part in determining these changes, as we miliar objects attract analytical encoding show later. operations. Disparities in normally oriented text are Means-Dependent Knowledge usually in its novel message or content, the The line of experimentation that empha- semantic domain, and so it is that its se- sized the encoding of semantic attributes in mantic aspects are usually what are re- preference to physical details of the stimuli called. By turning things around and made the claim especially strongly that making the graphemic features the more people learn from various media by ab- disparate, Kolers and Ostry also made stracting and storing their contents, mean- them memorable. Experiments that show a while discarding source information about superior memorability for semantic features the media except perhaps for "footnotes" may only be special cases of the more gen- (Bransford & Franks, 1972; Kintsch & eral principle that people particularly en- Monk, 1972; Kolers, 1966; Sachs, 1967). code disparities in familiar objects. Dis- This line of work characteristically ignored parity or difference of the stimulus from the the means by which information was ac- practiced skill is thus shown to be an effec- quired, but a number of studies since have tive variable controlling performance. shown that means and content cannot be On this account knowledge is expressed wholly distinguished (Kolers, 1978); rather, in activities, techniques, procedures-- the means of acquisition, even motoric skilled ways of relating to the stimulus-- means, often form a part of whatever a and these ways or activities change as a person knows. function of practice or exercise. A counter- We take a few examples from the litera- intuitive prediction from these claims is ture on perception and literacy. A demon- that as skill at encoding increases, memor- stration by Wertheimer (1912/1961) was ability of the encoded items should de- modified by Pomerantz (1970) to illustrate crease. Exactly this outcome was obtained that motor actions are integrated into per- MENTAL OPERATIONS 433 ceptual experiences. If the two diagonals of as in R of Figure 2, the rotated letters were an x are flashed alternately with a concen- named more rapidly leftward than right- tric plus sign (+) at the appropriate rate, ward. the entire shape seems to rotate in a com- Gonzalez and Kolers (1982) examined pelling motion. Pomerantz added circles at the issue further in a simple experiment on the end points of all the lines and instructed mental arithmetic. A propositionalist would the subjects to scan from circle to circle in affirm that numerical quantities are oper- a clockwise direction, in a counterclock- ated upon in mind in some abstracted rep- wise direction, or to fixate the center of the resentation; for example, the reader would display. Perceived direction of motion of evaluate 2 + 3 = 5 by abstracting twoness, the display tended to follow the direction threeness, and fiveness from the numerals of scanning eye movements. The display and testing the sum in some abstract form seemed to rotate clockwise when the of arithmetic. Gonzalez and Kolers sug- people scanned clockwise, to rotate coun- gested, to the contrary, that the operations terclockwise when they scanned oppo- that mind carries out depend in part on the sitely, and to "flap its arms" when people symbol system within which information is fixated the center. Here is evidence that the represented. On a propositionalist point of motor process of directed eye movements view, the equation above is identical to II entered into the perceptual experience. + III= V, and processing the two forms Kolers (1969) required people to name would be composed of separate stages of words of scrambled sentences leftward or acquisition of the symbols and operation on rightward. The scrambling distorted syn- their abstracted content. The latter opera- tactic sequences, and so people were tion would be identical for the two equa- obliged to look at the words one at a time. tions and differences in processing would Nevertheless, the words were named far be attributed only to differences in speed of more rapidly in the familiar rightward di- encoding the symbols. Gonzalez and Ko- rection than in the less familiar leftward di- lets showed that differences in processing rection, The influence of direction of scan- the equations were attributable not only to ning was so strong that even when the word acquisition but as well to differences in the order was nonsense rightward and sensible operations that were carried out on analog leftward people could still read more rap- roman symbols that tally quantities and on idly in the familiar rightward direction than digital arabic symbols. For example, in in the unfamiliar leftward direction (an ex- equations with mixed symbols (2 +III = ample is provided by rM in Figure 2). 5), the placement of the disparate symbol The perceptuomotor sequencing devel- affected performance in important ways. oped by people who read English is not a The argument was that the symbols refer in fixed or rigid pattern, independent of ori- different ways, that mind operates upon entation, however. Although rightward is them differently in correspondence, and the skilled direction for normally oriented that the operations are carried out upon the English, appropriate manipulation of the symbols rather than upon contents ab- stimulus induces an appropriate transfor- stracted from them. This interactionist mation of the analytical strategy. Kolers view stands in contrast to one that main- and Perkins (1969) required students to tains a structure-process dichotomy and name letters that had been rotated around that distinguishes means from knowledge. the principal axes of space. A line of nor- Data are not lacking within traditional mally oriented letters, spaced singly as r v paradigms to reveal a means dependency of t d g h, etc., was named reliably more rap- knowledge. Kirsner and Smith (1974) found idly rightward than leftward; but when the that two presentations of a word one of letters were rotated in the plane of the page, which was heard and one read were not as 434 KOLERS AND ROEDIGER

*Expectations can also mislead us; the unexpected is always hard to

perceive clearly. Sometimes we fail to recognize an object because we

~qW sea ~I •~q O~ sax~p aq se ~Z~ I s~ SX U~m AX~A~ ~q~ p~es O~UO UOSX~,

s~ ~I "~Mem O~ p~Dodx~ aq wqe~ um~znd ptreibux m~N ~ a~Ws~m Jo pu~M

~&ps~e g~s pn~ g te~ ot Cps ~egaoua ~o~ psyTeaTua ¢~ g l~zaou cgUuo¢

p6 CO~BG[O~a O~ ~[I ~[S ~e~gI b~o~es~es" ~s.ih o~psE Eeg~ou~ cg~ p6

spasL s as ypoLodoyeq Rsdoss~ odw ~o~sRoaq sops sTssy Ls~evs~* M 9onepil~s~,~ ~se~p ~o ~sm p~uoy s ~n~eia~e a~d las o~ b~d y~la~ev~m/

*On die ~l~aa bsy in ~oqey-Juavy i~b de ws~ ~doaoupdiy blao~lsnJsb. rN Hi~ ~99~ ws~s 5dove di~ dssbl de d~5 ~o ae~od ~o~ ~dsm wden ~9

rR g hlszem e~e~ ~cs]po ug IT ~s enspep o~ ~mse~ qI~pc dunoA hzeA V, ~IJ~no[cT~bgc MS[A ~0 qle[~ sp; ~GAgS I qug aLegue ;g~ saslg[ ign~[A

,dsA~qolob~ q~oe~ ~u axdeaxwau~I saTeuDe pnx~u6 ~q~ olosxu5 pa~epas oj rl ~q~ u~ua~a~uwq o~u~naA' e~ e ~T~a Mq~u ~nxoda~u ~qone~ M~s pomxu~ap qk

elap a no elcric der thgirb a swohs enO .serutcip tnereffid owt enigamI* rM .dlluorgkcab yarg a no elcric neerg thgirb a rehto eht ,dnuorgkcab wolley FIG. 2. Eight examples of geometrically rotated texts. The asterisk shows where to begin reading each pair. Reprinted, with permission, from Kolers and Perkins (1975). effective for memory as two presentations tations are in the same modality or same both read. Equivalent limitations on inter- language (Kolers & Paradis, 1980). Jacoby modal or interlingual stimulation were re- and Witherspoon (1982) helpfully survey ported by Kolers (1975b) and by Jacoby some more recent evidence to show depen- and Dallas (1981). In a lexical decision task dence of recognition upon means of acqui- Scarborough, Gerard, and Cortese (1979) sition (analysis) of the stimulus. It is not presented the printed name of objects or clear to us how such outcomes can be ac- line drawings of them and then tested for commodated within the theory of amodal speed in deciding whether a subsequent representation of information; but all of letter string was a word. They found con- these outcomes are compatible with the no- siderable facilitation from one presentation tion that the operations carried out in ac- of a word to another, but no positive influ- quiring a stimulus or the means exercised ence of the drawing upon the subsequent in doing so are part of what one knows response to the word. In a variety of cir- about it. cumstances, we may say, authors have re- Kolers and Smythe (1979), in an ex- ported lesser influence of one means of pre- tended review of the imagery debate, also sentation upon recognition in another, or proposed that the operations of mind varied lesser influence of information presented in with the contents operated upon and that it one language upon decisions made in an- was mistaken to try to distinguish a repre- other, than is obtained when both presen- sentation from a process operating upon it. MENTAL OPERATIONS 435

Rather, as the experiments cited above sug- The findings may be summarized by gest, acquisition and processing of symbols saying that people read a passage faster are part of a single indivisible operation. A than they named its letters (not a surprising distinction between structure and process finding), that naming letters provided only has prevailed in psychological theorizing modest transfer to reading, and that reading for quite a long time, but evidence against provided even less transfer to naming in- its legitimacy seems very strong. dividual letters. These differences can be seen by comparing points in Figure 3. After Specificity of Transfer naming 15 passages of letters, the students Transfer of training is a venerable topic read their first page of inverted text at in psychology, but its course has taken two about the speed the text-trained group different directions. One set of students has achieved initially after reading only 4 studied transfer of motor skills from one pages; and after reading 25 pages of text the event to another (Adams, 1976; Magill, text-trained students named a passage of 1980). The other line of endeavor studied letters at about the speed the letter-trained transfer of training as a function of the group achieved after about 3 passages. The formal similarity of items, usually two lists. results (like many others) decisively chal- Investigators counted discernible or de- lenge the theory of reading based on the nominated features in lists of nonsense syl- alphabetic principle, that reading is largely lables, and measured transfer effects from a matter of concatenating letters; more to training on one list to learning another the present point, the results show that (Martin, 1965; Osgood, 1949). letter identification and word identification We use the notion of transfer of training are based on particular and specific aspects in still a third sense, as in the following de- of pattern recognition, and with only lim- scription. One group of students read 15 ited transfer between them. passages of inverted text (I in Figure 2) and Neither exercise of muscle groups nor a second group named its letters, which had formal similarity of items constitutes the been scrambled haphazardly (Kolers & basis of transfer in the present case for, Magee, 1978). The two groups were thus surely, reading an unfamiliar typography is exposed to exactly the same set of mis- not principally a matter of muscle control oriented characters during training with ex- or of learning lists of nonsense syllables. actly the same frequency. Following the We propose that positive transfer occurs in training, both groups read 10 passages of relation to the similarity of procedures that text as a test and then named one passage two tasks exercise, and we explore aspects of scrambled letters. The curves in Figure of the specificity in the next section. 3 bring out several relations. One concerns speed of performance during training and Transfer versus Dissociation of Skills test: The group trained on text improved in On the classical theory of memory a speed of performance throughout and read single trace was laid down that represented the 10 test pages, passages 15 to 25 on the an experience wholly (Gomulicki, 1953. abscissa, faster than the group trained on The notion that experiences can be repre- letters. A second relation of interest is that sented in mind on a unitary dimension is the group trained to name letters named a still preserved in some theorizing (Wickel- final passage of letters at their former gren, 1973). This approach seems unable to speed, whereas the group that had just read deal with asymmetrical transfer, the un- 25 pages of inverted text named individual equal influence of training tasks on each letters only a little faster than the letter- other (tbr example, Kolers & Perkins, trained group had done initially--26 on the 1975). An alternative contemporary view abscissa. proposes that events are decomposed by 436 KOLERS AND ROEDIGER

22 TRAINING GROUP TASK • READING TEXT READINGTEXT 20 & " " NAMING LETTERS o NAMING LETTERS NAMINGLETTERS 18 • ,, ,, READING TEXT 09 LLI 16 I-- ,2D 14 Z~

-- IO LLI 8 I"- 6

I I I Illllltllllllllllllllll I 2 2, 4 5 6 7' 8 9 I0 II 12 12, 14 15 16 17 18 19 20 21 22 22, 24 25 26 SUCCESSIVE PASSAGES FIG. 3. Differential training and differential transfer after exposure to "identical" stimuli. Copy- right (1978) Canadian Psychological Association. Reproduced, with permission, from Kolers and Magee (1978). sensory processes into distinguishing fea- periences train skills selectively. We dis- tures, and the features are stored as aspects cuss this broad proposition by reference to of the memory trace of the event, usually some special studies. The issue of interest in a language-like form (Bower, 1967; Un- concerns dissociability of components of derwood, 1969). Other approaches have performance. If mind contained a single studied false recognitions (Underwood, trace of an experience that varied only in 1965), release from proactive interference strength, then various measures of memory (Wickens, 1972), and effectiveness of re- or knowledge would be affected in the same trieval cues (Tulving & Watkins, 1975), way. The finding has been, rather, that dif- among other tasks that are thought to referent test performances reveal differential veal the selective influence of past experi- memorability of an experience. For ex- ence on perception and memory (Tulving & ample, Kolers (1975a, 1976) studied peo- Bower, 1974). Within this history of en- ple's acquisition of skill at reading geomet- deavor, no agreed upon technique has yet rically inverted text and its savings and been established for specifying the means transfer after a year. Some of the passages by which features are selected for memory, read on the second occasion had been read nor has a technique been devised for con- also on the first occasion, whereas others, clusively determining the attributes of taken from the same sources, had not been memory traces. The reason for these fail- read before. Analysis revealed that the re- ures may be that memory has been studied read pages were read reliably faster than mainly in terms of descriptions of knOwl- the companion pages read for the first time. edge regarded as the contents of mind The students subsequently classified a large rather than in terms of the procedures used number of these and other pages as new or to acquire or express knowledge. as having been read before; and if read be- Our view is that the effects of experi- fore, when and how often. Thus three in- ences depend upon the procedures used to dependent measures of performance and an realize them rather than upon some de- associated one were obtained: speed of scription of them, and that particular ex- reading, and judged familiarity, frequency, MENTAL OPERATIONS 437 and recency of encounter. People were ical level. For example, Tulving et al. (1982) often able to reread a page faster than they found that an episodic presentation facili- read a companion page but not recognize it tated performance on a semantic word as having been read before, or people knew completion test, but as the advantage did when or how often a page had been read not decay with time in a manner common but without any corresponding change in to other episodic tasks, the behavior Was their speed of reading. The various perfor- attributed to a third, "procedural," mances were thus said to be dissociated, in memory system. Tulving (1983) invokes a analogy to the distinction drawn in the clin- distinction between procedural and declar- ical literature (Rapaport, 1951). The disso- ative systems, nests episodic and semantic ciations in the present case were between memories within the declarative system, descriptions of a text or of encounters with and assigns "skilled" aspects of perfor- it, and skill in reading it. mance to the procedural system; but even Jacoby (1983) provides another instance. this complicated arrangement seems poorly He found that antonyms generated by a justified. Unless the boundaries of tasks person were more likely to be detected in said to be episodic, semantic, or other can a recognition test than were the same words be established, little conceptual gain ac- read from a screen (as did Slamecka & crues from Tulving's analysis, and we may Graf, 1978); but words that had been read anticipate the invention of still more were more likely to be identified later from memory systems to explain still other dis- a brief visual presentation than were the sociations encountered experimentally. In same words that had been generated and illustration, Moscovitch (1981) argues for a spoken originally but not read. multiplicity of dissociations among various The notion of dissociation of task perfor- capabilities, and Cohen (in press) argues for mance has attracted considerable attention a multiplicity of memory systems. It is not in the modern study of memory, both certain that much is gained by postulating among clinical investigators (Squire, 1982; independent "systems". responsible for Wanington & Weiskrantz, 1970, 1974) and each sort of memory or dissociation. others (Jacoby & Dallas, 1981; Tulving et al. Zola-Morgan, Squire, and Mishkin (1982) 1982). How best to understand such disso- assume that the procedural and the declar- ciations is currently under discussion. One ative are two fundamentally different ways way of interpreting them is by attributing of representing experience, each with its different aspects of performance to dif- own type of dissociation. A related pro- ferent memory "systems." Tulving (1983) posal has been urged by Cohen (in press). recommends this approach and attributes These authors ignore in this approach the some aspects of performance to an "epi- fact that creating statements or descriptions sodic" system and some to a "semantic" for oneself--the declarations of a declara- system. If the occasion of learning forms tive system--is as much a matter of exe- part of the knowledge, the knowledge is cuting a procedure as is walking across a said to be in the episodic system, whereas room or throwing a ball or reading a text if the knowledge is independent of the oc- aloud. Statements or declarations, there- casion of its acquisition, the knowledge is fore, do not fail of procedural representa- said to be in the semantic system. tion; indeed, it would be exceedingly diffi- Unfortunately for the argument, this ap- cult for anyone accepting this dichotomy to proach lacks a clearly derived means of dis- specify what aspects of reading aloud are tinguishing episodic from semantic memory only procedural and what aspects of tasks; moreover the contrast of occasion- making statements are not procedural. dependent and occasion-free knowledge Below we will argue that dissociation does lends itself to confusion even at the empir- no t require a special accounting. 438 KOLERS AND ROEDIGER

The perceptuomotor skill of reading were, however, that hearing the words or words is not perfectly correlated with their reading them in translation transferred only recognition as having recently been read; weakly to the later reading; reading the independence of the two skills is quite very words but in a different typography marked (Cohen & Squire, 1980; Jacoby & transferred better; reading the very words Dallas, 1981; Kolers, 1976). The claim for in the very typography transferred best. dissociation supposes that different effects The experiment thus created a ranking of of performance identify different memory influences: a cross-modal presentation systems (Moscovitch, 1981; Tulving et al., (hearing to reading) or a cross-language 1982; Zola-Morgan et al., 1982). It could presentation (French to English)--the se- therefore be argued that the perceptuo- mantic component--yielded least transfer; motor aspects of reading are in a different a greater influence was due to the lexical "system" from the linguistic. Tulving component--the same words read but in a (1979, 1983), actually has defeated such an different typography; and the greatest argument; he reviewed many instances of transfer occurred when the same words "dissociation," all within an episodic were read in the same typography. Reading memory system, ff dissociations are found the very same typography activated nearly among tests allegedly tapping the same the very same procedures; reading the memory system, then the discovery of dis- same words in a different typography acti- sociation between tasks cannot be taken as vated fewer procedures in common; and evidence for different memory systems. reading translations of words or acquiring In contrast to a view that emphasizes a the words through a different medium ac- proliferation of memory systems, we rec- tivated still fewer similarities of procedure ommend that dissociation phenomena be across the two readings. Thus skill was viewed as still another instance of the spe- manifested across occasions in correspon- cifity of learning and transfer. On our view dence as procedures were similar across "dissociation" is the natural state of af- tasks. fairs, not what needs explaining. This view Because procedures are task specific, is illustrated by the following study (Ko- they can readily appear to be "disso- lers, 1975b). ciated," as the following shows. Seven The experiment required students to r~ad groups of students each read 24 pages in a long list of sentences. In the critical task, one of the transformations illustrated in the sentences were inverted typographi- Figure 2 (Kolers & Perkins, 1975). After cally and reading speed was measured. As training on the 24 pages of a single trans- "practice" for reading each test sentence formation, all the students read 2 pages in each student read the very same sentence each of the eight orientations illustrated, in in the same typographic orientation, or the appropriately counterbalanced order. The same sentence in normal typography, or a question of interest was the extent of close translation of the sentence in French transfer from training on one transforma- (for bilingual subjects), or the words of the tion to test on the others. Those percent- sentence were played through earphones. ages are shown in Table 1. Transfer from A propositionalist account would suppose training to test on the same transformation that reading a sentence or hearing its words was always 100%; if performance on the would transfer equivalently to aid a later test transformation was less than perfor- reading, as would reading a sentence in mance on the training transformation, French for a skilled bilingual. After all, the transfer was less than 100%, otherwise it variants preserve the same semantic con- was greater. tent and, except for the French translation, The rows of the table show the training in exactly the same words.The findings transformation, and the columns of the MENTAL OPERATIONS 439

TABLE 1 dence for dissociation in some of the re- PERCENTAGE TRANSFER FROM TRAINING TO TEST ports mentioned above, asymmetry and

Test specificity of transfer, and interaction of Training training and test; but here the results are Training M R rR rN rI I rM average obtained within the same sensory modality I 95 118 97 80 84 100 52 89 and on the "same" task of reading misori- rI 92 112 90 82 100 78 72 89 ented text. From our point of view, it is rR 81 84 100 68 56 79 56 74 more appropriate to examine these results rM 88 67 70 57 37 44 100 66 in terms of the operations the reader per- R 87 100 70 37 58 44 41 62 M 100 69 71 72 38 28 52 61 forms to carry out the tasks than to invent rN 82 3 57 100 36 34 20 47 different perceptual states or memory systems to accommodate the "dissociations" Test average 89 79 79 71 58 58 56 in the data. On our view it is not dissociation that needs to be explained, for that is the natural state of affairs; it is the char- table indicate performance on the test acteristics of tasks--and relations among transformations. Scanning across columns their underlying procedures-- that needs within any row shows that the percentage explaining. Responses are specific to skills transfer to any test from any training varied in manipulating systems of symbols and with the test. Scanning down any column there is no necessary connection among re- shows that the different training transfor- sponses. mations affected test performance differentially. In column R, for example, transfer Consciousness ranged from 118% for training on I (that As we noted earlier, the greater part of is, I trained R better than R trained itself) contemporary research into learning and to near zero, 3%, for training on rN to memory seems to study conscious recollec- test on R. Readers thus do not merely learn tion or conscious experience. By this we some general strategy of coping with trans- mean that the primary measures have been formations or acquire some general skill of recognition or recall of materials con- reading them; rather, they acquire partic- sciously perceived or conscious decisions ular skills due to the procedures a transfor- about such materials, and mechanisms pro- mation activates, and transfer those skills posed have emphasized accumulation and to other transformations as particulars. The organization of information about stimuli "anomalous" transfer in which I and rI (Mandler, 1980; Tulving, 1983). trained R better than R trained itself is con- We find the emphasis upon description as sistent with the theory of component skills the basis of knowledge faulty on two that rationalized the data (Kolers & Per- counts. One is that descriptions of events kins, 1975, pp. 250f). rarely if ever tell a person what to do about Another fact of the table is the notable the events described. The second is that de- asymmetry of transfer; as already re- scriptions implicate both a describer and a marked, training on I transferred to test on reader or interpreter of the description-- R 118%, but training on R transferred to some ego or like processor. (An appeal to test on I only 44%; training on rM trans- an executive ego is also made by Baddeley, ferred 88% to test on M, but training on M 1982, among many others.) It is unclear to transferred only 52% to test on rM. Many us how such an executive improves at a other examples of asymmetry are present task such as reading, say, except in the also. doing, or why telling oneself to do better In sum, this body of data bears many of ,rarely by itself improves performance. In- the characteristics that are taken as evi- deed, for reading geometrically misoriented 440 KOLERS AND ROEDIGER texts, description of the transformations to describe features of a stimulus as incidental the reader or by the reader to himself or to the original task. Moreover, how a herself does remarkably little to aid perfor- person describes an event depends upon mance, compared to practice at the task it- the question the person is asked, the vo- serf. Moreover, people can often carry out cabulary the person has available for de- a task expeditiously while being unable to scribing the experience, the person's skill report the principle on which it was based in description, and the like. (Brooks, 1978; Hull, 1920), and, as we People do not encode all identifiable fea- noted in connection with Figure 3, the features of an event equally, but encode what tures of practice that are useful in one con- they have learned how to encode, and their text may transfer little if at all to perfor- performance changes as a function of en- mance in an apparently related task. Of counters with stimuli. (Training can feed course consciousness can be important to back into encoding, of course; for example, behavior and we do not gainsay its study, a linguist by virtue of training becomes sen- but we can emphasize that people carry out sitive to many formal "linguistic" features many tasks in which neither the perfor- of words.) mance itself nor the means of its improve- The nature of consciousness is not even ment is open to conscious description. certain; it has been debated for rather a Hence, on our view mind is better con- long time and been given many treatments strued in terms of what it can do than in (Humphrey, 1951). We believe that con- terms of descriptions of what it "knows." sciousness can be usefully conceived of in An instance of the deleterious effects of analogy to a sense organ that is responsive a concern with conscious content stems to changes in some internal states of the from the work on incidental and intentional individual. Consciousness, on this account, learning. It was assumed by many experi- does not report on what is in the world, but menters that an instruction to a subject to on how the person's sensing organs re- attend to a particular feature of a stimulus spond to events; and like a sense organ it meant that the subject attended only to that monitors selectively. We believe that an ap- feature; other aspects of the stimulus were proach to behavior put in terms of a per- thought to be peripheral or outside aware- son's skills would prove to be more useful ness (Craik & Lockhart, 1972). There is no inductively than an appeal to "machinery" certainty that a subject encoding stimuli is for manipulating the contents of conscious- sensitive to the same features or categories ness, or lists of descriptions of its contents. that interest the experimenter, and no reason to believe that only the feature to APPLICATIONS AND EXTENSIONS which attention is directed is encoded We have reviewed a number of condi- (Nelson, 1979). The very term "incidental tions that encourage a view of learning and learning" endows the psychologist with the memory in terms of the operation of ana- illusory power to decide what is central and lytical procedures directed at the stimulus. what is incidental to processing. Using the We turn now to other phenomena that in- term confuses the psychologist's goals in terest students of learning and memory and the experiment with a description of the show how some of the same notions can be nervous system's operations. What may be applied there. In doing this we do not affirm "incidental" to the psychologist may be a that a proceduralist account necessarily ac- fundamental constituent of encoding oper- commodates all phenomena of learning and ations, as size, orientation, or typography memory; rather, we recommend this gen- of a text have been shown to be. Only after eral account for its emphasis on measurable appropriate testing has revealed no resi- activities. We recommend it also on Oc- dues of their influence can one realistically cam's principle that simpler accounts MENTAL OPERATIONS 441 should be refuted before more complex by Kolers and Ostry (1974). Actually, the ones are entertained. very notion of level is suspect. Craik and Cognitive psychologists often factor their Lockhart (1972) implied the existence of a field of inquiry into encoding processes, re- fixed scale on which one moved from less trieval processes, and their interaction. It to more as one went from graphemic should be clear from what has gone before through syntactic to semantic. Of course, that such a division may not be warranted; no such dimension or scale exists indepen- for convenience, however, we discuss phe2 dently of circumstances; what is superficial nomena sometimes attributed wl~olly to en- and what is deep depends not on the stim- coding processes alone, then the conjec- ulus but on skill, purpose, and the way that tured interactions of encoding and re- the stimulus is "taken" (Kolers, 1975a). 3 trieval. The generation effect (Slamecka & Graf, 1978) is obtained when under instruction a Encoding Phenomena person carries out some productive opera- Under this rubric we discuss phenomena tion on a linguistic stimulus, in contrast to associated with the levels of processing only reading it, with a consequent advan- framework, the generation effect, and the tage to memory. Words that had been gen- effect of imagery instructions on memory. erated in response to a riddle, for example, The levels of processing notion has been were recalled better than words that were criticized on a number of grounds that do only flashed on a screen (Erdelyi, Buschke, not need review here (Cermak & Craik, & Finkelstein, 1977). As we mentioned ear- 1979). We mention the issue in the context lier, Jacoby (1983) has pressed the issue fur- of asserting that procedures undertaken to ther and shown that the actual advantage acquire a stimulus may fully account for the to reading or generating depends on the de- performances sometimes attributed to gree to which operations required initially "level" of processing. Actually, a number match those required at test. The advantage of psychologists have exploited in a number seems to come from repetition of mental of ways the fact that instructions or inten- actions, not from generating or reading per tions can "set~'~ people or direct their en- se. We believe the generation effect can be coding or analytical procedures to different accommodated by a proceduralist account aspects of the stimulus, as have Postman of learning and memory. and Adams (1956), Posner (1969), and Hyde A similar appeal to analytical operations and Jenkins (1969). seems to lend itself to many other circum- Craik and Tulving (1975) produced a par- stances. The more analytical work or the ticularly striking example of the effect of more nonstandard the analytical work a instructional set in showing that the fre- person is required to perform in acquiring quency with which words were recognized a stimulus (Bower, 1972), the greater the as having been presented earlier varied ac- likelihood of learning. A spectacular use of cording to the directions governing their en- procedures is found in many of the tech- coding. The argument of Craik and Tulving niques that are intended as mnemonic aids. was that, when judging words for their se- In the method of loci, one "situates" ob- mantic category, people encode them more jects as one walks along a real or imagined deeply and richly than when judging their path; the object is:recovered cognitively appearance as typographic objects. In contrast, and as We will discuss again below, 3 It is not even clear how to compare graphemic and Stein (1978) among others has shown that semantic patterns as burdens or costs to encoding. Lacking a scale that provides a measurement common by suitable definition of the task, the gra- to them both, we cannot directly compare the effi- phemic can be made the more important ciency of memory for such features, despite some ef- and the semantic less, a point made earlier forts to do so, as by Bransford et al. (1979). 442 KOLERS AND ROEDIGER when one recreates in mind the path taken longer function productively in recovering (Crovitz, 1970). Variations of the technique that knowledge. The difference is not in the are also useful to memory (Roediger, presence or absence of a source, but in the 1980b; Smith, 1982). Other remarkable multiplicity of procedural routes the person cases of memory also seem to yield to an can use to express the knowledge. analysis in terms of the operations or pro- To the degree that situational variables cedures that the rememberer engages in function as an intrinsic aspect of the acqui- (Hunter, 1957; Luria, 1968). It is not a sition of a stimulus, a number of context- matter merely of quantity of analysis or dependent phenomena also seem amenable procedures, as we mentioned earlier--re- to a proceduralist analysis. The more sim- quiring people to analyze stimuli at too de- ilar the procedures used during two occa- tailed a level does not yield superior per- sions, such as learning and test, the more formance (Kolers, 1974); rather, it is anal- successful performance will be at the later ysis of the stimulus in a manner appropriate occasion (Smith, 1982). This contingency to the question asked, that is significant. between learning and test has often been demonstrated in the verbal learning litera- Interaction of Acquisition and Test ture. Tasks such as paired associate On our view all performances express the learning and the learning of serial lists have means by which knowledge was acquired; been used as training and the skills acquired memory is never a matter of descriptions have been measured by transfer to other recovered from a static store. Tulving tasks. Hall (1971) surveys a body of litera- (1979, 1983) has discussed one aspect of the ture that shows first that people improve in interactive nature of encoding and re- task-specific ways across a number of trieval, and Bransford, Franks, Morris, and trials; that is, people take fewer trials to Stein (1979) have contributed the notion of reach a criterion on successive lists of serial transfer-appropriate processing to empha- or paired associate learning (learning to size their belief that relevant similarity of learn). The second point is that learning to conditions of learning and test is the im- learn is not some general process, but is portant criterion. Both of these accounts specific to the exercise of particular skills. are accommodated by the notion of a pro- Postman and Schwartz (1964), for example, cedural correspondence between learning showed that training on one sort of list and test circumstances, albeit from dif- learning task transferred better to the same ferent perspectives. sort of task than to others. Thus, specific One way of establishing the contrast be- skills must be invoked to account for this tween episodic and semantic memory in type of improvement, not just nonspecific Tulving's (1983) account has to do with the factors ("warm up") (Postman, 1982). A re- source of knowledge. Information in mind lated claim seems applicable to many other whose access requires recovery of time and studies of encoding and remembering place of the experience is regarded as epi- (Hung & Tzeng, 1981). sodic information, and is thereby assigned A final example of an apposite explana- to a specific memory system. Information tion of a phenomenon in terms of proce- in mind that is independent of the occasion dures is Slamecka's (1977) resolution of the of experience is attributed to another, se- problem of seemingly minimal transfer from mantic, memory system. There is nothing a serial to a paired associate list. Slamecka special about time and place as sources, (1977) argued that, in addition to simple however. On our view all knowledge is presentation of the prior item by the exper- source dependent. If some knowledge were imenter, the part of the stimulus complex acquired on a number of different occa- that guides responding in serial learning is sions, however, time and place might no the subject's act of recalling the item. The MENTAL OPERATIONS 443 cue to produce C in the serial associative fundament upon which to construct an em- chain is not simply the presentation of B, pirically based cognitive psychology. The but also is the recall of B by the subject. techniques would be applied as measures Slamecka proposed that the procedures in of skills acquired in one cognitive task and recall are a significant part of the associa- expressed in performance on another. De- tive complex that leads to recovery of the gree of transfer from one task to the second next item in a serial list. When Slamecka or, as in Nelson's (1978) work, the more (1977) arranged conditions to make com- subtle measurement of savings, can aid in patible responses include both the overt diagnosing the underlying cognitive opera- stimulus cues and the covert response pro- tions. The idea is that any complex event duced cues, great positive transfer was ob- is composed of a number of component ac- tained. Thus a classic problem in the verbal tivities, and the more alike they are, the learning tradition was solved by consider- more alike the behavior will be (Kolers & ation of the procedural cues in the situa- Perkins, 1975). Judicious experimentation tion; increasing the similarity of procedures may allow one to infer their identity. in the two tasks increased transfer, as in other examples we have discussed. REPRISE We have argued that an appropriate ac- Recovery count of mind should be based on the pro- We began by remarking on the changes cedures and skills a person brings to bear that have occurred in the study of learning on cognitive tasks, rather than be ex- and memory in the 30 years since publica- pressed in terms of hypothetical contents tion of the book by McGeoch and Irion or structures of mind. Broadbent (1958) (1952); changes that have seen a return of proposed that an appropriate analysis of interest in cognitive phenomena. We end mind would chart the flow of information now by pointing out that although much has through various "systems." In the 25 years been gained in the interval in respect to in- since, psychologists have produced interest in mental life as a topic for study, creasingly elaborate descriptions of such much, too, has been neglected. For ex- conjectured flows, as in Figure 1. Unfor- ample, in his review "Human Learning and tunately, there is little agreement on which Retention" Hovland (1951) included sec- proposed stores or boxes are necessary for tions on motor skills, transfer of training, the proper description, much less how they and savings, among other topics; relatively should be arranged. Even seemingly simple few contemporary works consider them. questions such as whether "mental search" Study of motor skills has become self-con- occurs in a serial or parallel fashion has tained and somewhat isolated from the turned out to be remarkably difficult to de- broader study of learning and memory cide (Townsend, 1974). Lack of cumulative (Kelso, 1982; Marteniuk, 1976; Schmidt, progress in the information processing tra- 1975; Stelmach, 1978). Savings and transfer dition may be due to an inappropriate of training seem to have faded from view, framing of the questions at the outset. perhaps unfortunately. (It is true that the Asking how information flows through var- greater number of older studies in which ious cognitive systems presupposes that these notions figured prominently consid- such systems exist and can be validated by ered them principally as methods for mea- behavioral evidence, yet the research re- suring the interaction in mind of list con- cord lacks convincing support for these tents, as in the studies manipulating formal conceptions. It may be more appropriate at similarity (Hall, 1971; Osgood, 1949).) On this point to admit that many of the as- our view, transfer of training and savings sumptions of the information processing methods, properly applied, constitute a approach are either wrong or are not test- 444 KOLERS AND ROEDIGER able, and to seek other ways of studying ficity limits the possibility of creating en- mind. The procedural approach, even in its compassing descriptive statements. current state, provides a useful alternative. "Adaptation," "evolution," and "pattern Some comment on the strengths and recognition" are similarly synoptic terms, weaknesses of this approach seems in suggestive of processes but not explicit order. It is common to lament the com- across instances. plexity of many phenomena in our field, a In regard to theoretical precision, we point that cannot be gainsaid (Jenkins, cannot justify being more specific than cur- 1979). However, treating cognitive pro- rent knowledge permits, and current cesses as skills that transfer differentially to knowledge in this area is limited to the new tasks can serve as a unifying principle work of a few investigators. Many contem- to organize patterns of behavior. Second, porary models are remarkably sophisti- complex skills can be conceived as com- cated with respect to precision of definition posed of simpler components. These com- of terms and the mathematization of their ponents may be trained differentially and relations. However, this precision is often organized in various ways as a person bought at a heavy price, for the models or copes with particular tasks. An important theories deal with such a circumscribed set area of study is the organization and allo- of conditions that general statements are cation of component skills as a function of precluded. Often the merest broadening of practice at a task. Third, the logic of conditions renders the model inoperative. transfer of training can be used to identify Estes (1982) devised a mathematical rep- and study the component skills, in ways al- resentation of similarity effects in letter rec- ready mentioned. Differential transfer after ognition, but without concern for size, face, mastery of a task can be used to identify orientation, or other such characteristics of the components (Kolers & Perkins, 1975), letters, yet these are known to affect rec- leading to a taxonomy of trainable capabil- ognition in important ways (Huey, 1908/ ities. This emphasis on transfer leads to a 1968; Tinker, 1963), as well as affecting fourth advantage: a procedural description judgments of similarity. Murdock (1982) de- of mental processes encourages focusing on veloped a model of recognition complex, observables, the actual behavior of a its author remarks, almost to the point of person performing a task. Such an ac- inutility but able to deal with only the very counting is preferable to description of be- simplest memorial circumstances. Mathe- havior in terms of metaphysical entities matical representation does not by itself ex- populating the mind. plain behavior; like any other system of The procedural approach we have rec- symbols, mathematics only represents re- ommended does have its weaknesses, as lations. It is not obvious at this stage of we are aware. Chief among them is the im- investigation that such mathematical for- precision of many key statements and the malisms possess any conceptual advantage rather loose definition of terms, including over a straightforward idea expressed in even the concept of skill and its compo- simple prose. However, there is an advan- nents. These may be defined well in a par- tage to working with a metaphor that cap- ticular situation but general definitions are tures fundamental relations in behavior difficult to achieve. Some colleagues have over a wide range of situations, albeit im- questioned whether the "theory" is in any precisely, for that metaphor can be used as real sense testable. a target for formal development and refine- In answer to such queries we make sev- ment. eral points. Regarding the definition of Regarding testability, we have reviewed skills, we have emphasized throughout that many sorts of empirical evidence that bear skills are situation specific. We may raise on and support the notion of mentation as as a conjecture the notion that this speci- an activity and remembering as a skill. This MENTAL OPERATIONS 445 evidence shows that the procedures of ac- BOWER, G. H. (1967). A multicomponent theory of the memory trace. In K. W. Spence & J.T. Spence quisition inextricably form a part of mental (Eds.), The psychology of learning and motivation representation, that procedures are remark- (Vol. I). New York: Academic Press. ably specific, that they transfer differen- BOWER, G. H. (1972). Mental imagery and associative tially to new tasks, and the like; such dem- learning. In L. W. Gregg (Ed.), Cognition in onstrations are the foundation of our claims learning and memory. New York: Wiley. BOWER, G. H., & KARLIN, M. B. (1974). Depth of pro- regarding the propriety of a procedural cessing pictures of faces and recognition memory. view. Journal of Experimental Psychology, 103, 751- The study of cognitive processes has be- 757. come one of the most popular psycholog- BRANSFORD, J. D., & FRANKS, J. J. (1972). The ab- ical subjects, especially for English-lan- straction of linguistic ideas: A review. Cognition, 1, 211-249. guage psychologists, and yet there is little BRANSFORD, J. D., FRANKS, J. J., MORRIS, C. D., & cumulative development in the field. It is STEIN, B. S. (1979). Some general constraints on very difficult, for example, to establish psy- learning and memory research. In L. S. Cermak chological principles that have generality and E I. M. Craik (Eds.), Levels of processing in across many circumstances. Behavior is human memory. HiUsdale, NJ: Erlbaum. BROADBENT, D. E. (1958). Perception and communi- situation specific, but its description clearly cation. London: Pergamon. should have some generality. Our proposal BROOKS, L. (1978). Nonanalytic concept formation is that a proper course for cognitive psy- and memory for instances. In E. Rosch & B. B chology is the study of mind construed as Lloyd (Eds.), Cognition and categorization. Hills- skill in manipulating symbols. Acquisition dale, NJ: Erlbaum. BROWN, R., & MCNEILL, D. (1966). The "tip of the of reference (symbol creation), skill in tongue" phenomenon. Journal of Verbal transforming and manipulating symbols Learning and Verbal Behavior, 5,325-337. (recognition and problem solving), and CERMAK, L. S., ~; CRAIK, F. I. M. (Eds.) (1979). their application to new occasions Levels of processing in human memory. Hillsdalc, (transfer) can provide a basis (Kolers & NJ: Erlbaum. COHEN, N. J. (in press). Preserved learning capacity Smythe, 1984). Mental life is intrinsically in amnesia: Evidence for multiple memory sys- symbolic. Study of the procedures realizing tems. In N. Butters & L. R. Squires (Eds.), The the manipulations must necessarily reveal neuropsychology of memory. New York: Guil- in a cumulative way the significant char- ford. acteristics of mind. COHEN, N. J., & SQUIRE, L. R. (1980). Preserved learning and retention of pattern-analyzing skill in amnesia: Dissociation of knowing how and knowing that. Science, 210, 207-210. COLLINS, A. M., & LOFTUS, E. E. (1975). A spreading- REFERENCES activation theory of semantic processing. Psycho- ADAMS, J. A. (1976). Learning and memory: An intro- logical Review, 82, 407-428. duction. Homewood, 1L: Dorsey. COLLINS, A. M., & QUILLIAN, M. R. (1969). Retrieval ANDERSON, J. R. (1976). Language, memory, and time from semantic memory. Journal of Verbal thought. Hillsdale, NJ: Erlbaum. Learning and Verbal Behavior, 8, 240-247. ANDERSON, J. R. (1982). Acquisition of cognitive skill. CRAIK, E I. M., & KmSNER, K. (1974). The effect of Psychological Review, 89, 369-406. speaker's voice on word recognition. Quarterly ANDERSON, J. R. (1983). The architecture of cognition. Journal of Experimental Psychology, 26, 274- Cambridge, MA: Harvard Univ. Press. 284. ANDERSON, J. R., & BOWER, G. H. (1973). Human as- CRAIK, E I. M., & LOCKHART,R. S. (1972). Levels of sociative memory. Hillsdale, NJ: Erlbaum. processing: A framework for memory research. ATKINSON, R. C., & SHIFERIN, R. M. (1968). Human Journal of Verbal Learning and Verbal Behavior, memory: A proposed system and its control pro- 11,671-684. cesses. In K. W. Spence & J. T. Spence (Eds.), CRAIK, E I. M., & TULVING, E. (1975). Deptl- of pro- The psychology of learning and motivation: Ad- cessing and the retention of words in episodic vances in research and theory (Vol. 2). New York: memory. Journal of Experimental Psychology: Academic Press. General, 104, 268-294. BADDELEY, A. D. (1982). Domains of recollection. CROVITZ, H. F. (1970). Galton's walk. New York: Psychological Review, 89, 708-729. Harper & Row. 446 KOLERS AND ROEDIGER

CROWDER, R. G. (1976). Principles of learning and HASHER, L., & ZACKS, R. T. (1979). Automatic and memory. Hillsdale, NJ: Erlbaum. effortful processes in memory. Journal of Exper- DODD, D. H., & WHITE, R. M. (1980). Cognition: imental Psychology: General, 108, 356-388. Mental structures and processes. Boston: Allyn HOVLAND, C. I. (1951). Human learning and retention. & Bacon. In S. S. Stevens (Ed.), Handbook of experimental DRAVNIEKS, A. (1982). Odor quality: Semantically psychology. New York: Wiley. generated multidimensional profiles are stable. HUEY, E. B. (1968). The psychology and pedagogy of Science, 218,799-801. reading. Cambridge, HA: HIT Press (Original EBBINGHAUS, H. (1964). Memory (H. A. Ruger & work published 1908). C. E. Bussenius, Trans.). New York: Dover. HULL, C. L. (1920). Quantitative aspects of the evo- (Original work published, 1885). lution of concepts. Psychological Monographs, EISENSTEIN, E. L. (1979). The printing press as an 28. agent of change. London: Cambridge Univ. HUMPHREY, G. (1951). Thinking: An introduction to its Press. experimental psychology. London: Methuen. ERDELYI, M. H., BUSCHKE, H., 8¢ FINKELSTEIN, S. HUNG, D. L., & TZENG, O. J. (1981). Orthographic (1977). Hypermnesia for Socratic stimuli: The variations and visual information processing. Psy- growth of recall for an internally generated chological Bulletin, 90, 377-414. memory list abstracted from a series of riddles. HUNTER, I. M. L. (1957). Memory: Facts and falla- Memory & Cognition, 5,283-286. cies. Harmondsworth, Middlesex: Penguin. ESTES, W. K. (1982). Similarity-related channel inter- HYDE, T. S., • JENKINS, J. J. (1969). The differential actions in visual processing. Journal of Experi- effects of incidental tasks on the organization of mental Psychology: Human Perception and Per- recall of a list of highly associated words. Journal formance, 8, 353-382. of Experimental Psychology, 82, 472-481. FISHER, R. P., & CRAIK, E I. M. (1977). Interaction JACOBY, L. L. (1983). Perceptual enchancement: Per- between encoding and retrieval operations in cued sistent effects of an experience. Journal of Exper- recall. Journal of Experimental Psychology: imental Psychology: Learning, Memory, and Human Learning and Memory, 3,701-711. Cognition, 9, 21-38. FISHER, R. P., & CUERVO, A. (1983). Memory for JACOBY, L. L., BARTZ, W. H., & EVANS, J. D. (1978). physical features of discourse as a function of A functional approach to levels of processing. their relevance. Journal of Experimental Psy- Journal of Experimental Psychology: Human chology: Learning, Memory, and Cognition, 9, Learning and Memory, 4, 346-351. 130-138. JACOBY, L. L., ~ DALLAS, M. (1981). On the relation- FRIEDMAN, W. F., & FRIEDMAN, E. S. (1957). The ship between autobiographical memory and per- Shakespearean ciphers examined. Cambridge: ceptual learning. Journal of Experimental Psy- Cambridge Univ. Press. chology: General, 3, 306-340. GARNER, W. R., HAKE, H. W., & ERIKSEN, C. W. JACOBY, L. L., & WITHERSPOON, D. (1982). Remem- (1956). Operationism and the concept of percep- bering without awareness. Canadian Journal of tion. Psychological Review, 63, 149-159. Psychology, 36, 300-324. GEISELMAN, R. E., & BELLEZZA, E S. (1976). Long- JAMES, W. (1890). The principles of psychology. New term memory for speaker's voice and source lo- York: Holt. cation. Memory & Cognition, 4, 483-489. JENKINS, J. J.'(1979). Four points to remember: A tet- GEISELMAN, R. E., & BELLEZZA, E S. (1977). Inci- rahedral model of memory experiments. In L. S. dental retention of speaker's voice. Memory & Cermak & E I. M. Craik (Eds.), Levels of pro- Cognition, 5,658-666. cessing in human memory. Hillsdale, NJ: Erl- GOLDSTEIN, A. G., & CHANCE, J. E. (1971). Visual baum. recognition memory for complex configurations. JusT, M. A., t~¢ CARPENTER, P. A. (1980). A theory of Perception & Psychophysics, 9, 237-241. reading: From eye fixations to comprehension. GOMULICKI, B. R. (1953). The development and Psychological Review, 87, 329-354. present status of the trace theory of memory. KAHN, D. (1967). The codebreakers. New York: Mac- British Journal of Psychology Monograph Supple- millan Co. ments 29. Cambridge: Cambridge Univ. Press. KELSO, J. A. S. (Ed.) (1982). Human motor behavior: GONZALEZ, E. G., & KOLERS, P. A. (1982). Mental An introduction. Hillsdale, NJ: Erlbaum. manipulation of arithmetic symbols. Journal of Experimental Psychology: Learning, Memory, KINTSCH, W. (1974). The representation of meaning in and Cognition, 8, 308-319. memory. Hillsdale, NJ: Erlbaum. HALL, J. F. (1971). Verbal learning and retention. KINTSCH, W., & MONK, D. (1972). Storage of complex Philadelphia: Lippincott. information in memory: Some implications of the HART, J. T. (1965). Memory and the feeling-of-knowing speed with which inferences can be made. Journal experience. Journal of Educational Psychology, of Experimental Psychology, 94, 25-32. 56, 208-216. KIRSNER, K. (1973). An analysis of the visual com- MENTAL OPERATIONS 447

ponent in recognition memory for verbal stimuli. MARTENIUK, R. G. (1976). Information processing in Memory & Cognition, 1,449-453. motor skills. New York: Holt, Rinehart, & Win- KIRSNER, K., & SMITH, M. C. (1974). Modality effects ston. in word identification. Memory & Cognition, 2, MARTIN, E. (1965). Transfer of verbal paired-associ- 637-640. ates. Psychological Review, 72, 327-343. KOLERS, R A. (1966). Interlingual facilitation of short- McGEOCH, J. A., & IRION, A. L. (1952). The psy- term memory. Journal of Verbal Learning and chology of human learning. New York: Long- Verbal Behavior, 5, 314-319. roans. KOLERS, E A. (1969). Clues to a letter's recognition: MILLER, N. E. (1959). Liberalization of basic S-R Implications for the design of characters. Journal concepts: Extensions to conflict behavior, moti- of Typographic Research, 3, 145-168. vation, and social learning. In S. Koch (Ed.), Psy- KOI.ERS, R A. (1974). Remembering trivia. Language chology: A study of a science (Vol. 2). New York: and Speech, 17, 324-336. McGraw-Hill. KOLERS, R A. (1975a). Memorial consequences of au- MORRIS, C. D., BRANSFORD, J. D., & FRANKS, J. J. tomatized encoding. Journal of Experimental Psy- (1977). Levels of processing versus transfer ap- chology: Human Learning and Memory, 1, 680- propriate processing. Journal of Verbal Learning 701. and Verbal Behavior, 16, 519-533. KOLERS, R (1975b). Specificity of operations in sen- MOSCOVITCH, M. (1981). Multiple dissociations of tence recognition. Cognitive Psychology, 7, 289- function in amnesia. In L. S. Cermak (Ed.), 306. Memory and amnesia. Hillsdale, NJ: Erlbaum. KOLERS, R A. (1976). Reading a year later. Journal of MOSKOWITZ, H. R. (1981). Psychophysical scaling and Experimental Psychology: Human Learning and optimization of odor mixtures. In H. R. Mos- Memory, 2, 554-565. kowitz & C. B. Warren (Eds.), Odor quality and KOLERS, E A. (1978). On the representations of ex- chemical structure. Washington, D.C.: Amer. perience. In D. Gerver & W. Sinaiko (Eds.), Lan- Chem. Soc. guage interpretation and communication. New MUROOCK, B. B., JR. (1974). Human memory: Theory York: Plenum. and data. Hillsdale, NJ: Erlbaum. KOLERS, R A., & MAGEE, L. E. (1978). Specificity of MURDOCK, B. B., JR. (1982). A theory for the storage ? pattern-analyzing skills in reading. Canadian and retrieval of item and associative information. Journal of Psychology, 32, 43-51. Psychological Review, 89, 609-626. KOLERS, R A., & OSTRY, D. J. (1974). Time course of NELSON, T. O. (1978). Detecting small amounts of in- loss of information regarding pattern analyzing op- formation in memory: Savings for nonrecognized erations. Journal of Verbal Learning and Verbal items. Journal of Experimental Psychology: Behavior, 13, 599-612. Human Learning and Memory, 4, 453-468. KOLERS, E A., & PARADIS, M. (1980). Introduction: NELSON, D. L. (1979). Remembering pictures and Psychological and linguistic studies of bilin- words: Appearance, significance, and name. In gualism. Canadian Journal of Psychology, 34, L. S. Cermak & E I. M. Craik (Eds.), Levels of 287-303. processing in human memory. Hillsdale, NJ: Erl- KOLERS, P. A., & PERKINS, D. N. (1969). Orientation baum. of letters and their speed of recognition. Percep- NEWELL, A. (1980). Physical symbol systems. Cog- tion & Psychophysics, 5,275-280. nitive Science, 4, 135-183. KOLERS, P. A., 8,: PERKINS, D. N. (1975). Spatial and NORMAN, D. A., & RUMELHART, D. E. (Eds.) (1975). ordinal components of form perception and lit- Explorations in cognition. San Francisco: eracy. Cognitive Psychology, 7, 228-267. Freeman. KOLERS, P. A. • SMYTHE, W. E. (1979). Images, sym- OSGOOD, C. E. (1949). The similarity paradox in bols, and skills. Canadian Journal of Psychology, human learning: A resolution. Psychological Re- 33, 158-184. view, 56, 132-143. KOLERS, P. A. 8,: SMYTHE, W. E. (1984). Symbol ma- POMERANTZ, J. R. (1970). Eye movements affect the nipulation: Alternatives to the computational view perception of apparent (beta) movement, Psy- of mind. Journal of Verbal Learning and Verbal chonomic Science, 19, 193-194. Behavior, 23,289-314. POSNER, M. I. (1969). Abstraction and the process of LAKOFF, G., & JOHNSON, M. G. (1980). Metaphors we recognition. In G. H. Bower & J. T. Spence live by. Chicago: Univ. of Chicago Press. (Eds.), The psychology of learning and motivation LURIA, A. R. (1968). The mind ofa mnemonist. New (Vol. 3). New York: Academic Press. York: Basic Books. MAGILL, R. A. (1980), Motor learning: Concepts and POSNER, M. I., ~f¢ SNYDER, C. R. R. (1975). Attention applications. Dubuque, IA: Brown. and cognitive control. In R. L. Solso (Ed.), Infor- HANDLER, G. (1980). Recognizing: The judgment of mation processing and cognition: The Loyola previous occurrence. Psychological Review, 87, Symposium. Hillsdale, NJ: Erlbaum. 252-271. POSTMAN, L. (1982). An examination of practice ef- 448 KOLERS AND ROEDIGER

fects in recognition. Memory & Cognition, 10, SMITH, E. E., SHOBEN, E. J., & RIPS, L. J. (1974). 333-340. Structure and process in semantic memory: A fea- POSTMAN, L., & ADAMS, P. A. (1956). Studies in inci- tural model for semantic decisions. Psychological dental learning: IV. The interaction of orienting Review, 81, 214-241. tasks and stimulus materials. Journal of Experi- SQUIRE, L. R. (1982). The neuropsychology of human mental Psychology, 51, 329-333. memory. Annual Review of Neuroscience, 5,241- POSTMAN, L., & SCHWARTZ, M. (1964). Studies of 273. learning to learn. I. Transfer as a function of STEIN, B. S. (1978). Depth of processing reexamined: method of practice and class of verbal materials. The effects of precision of encoding and test ap- Journal of Verbal Learning and Verbal Behavior, propriateness. Journal of Verbal Learning and 3, 37-49. Verbal Behavior, 17, 165-174. RAPAPORT, D. (1951). Organization and pathology of STELMACH, G. E. (1978). Information processing in thought. New York: Columbia Univ. Press. motor control and learning. New York: Academic RAAIJMAKERS, J. G. W., & SHIFFRIN, R. M. (1980). Press. SAM: A theory of probabilistic search of associ- TINKER, M. A. (1963). Legibility of print. Ames: Univ. ative memory. In G. H. Bower (Ed.), The psy- of Iowa Press. chology of learning and motivation (Vol. 14). New TOWNSEND, J. T. (1974). Issues and models concerning York: Academic Press. the processing of a finite number of inputs. In RAAIJMAKERS, J. G. W., & SHIFERIN, R. M. (1981). B. H. Kantowitz (Ed.), Human information pro- Search of associative memory. Psychological Re- cessing: Tutorials in performance and cognition. view, 88, 93-134. Hillsdale, NJ: Erlbaum. ROED1GER, H. L., III (1979). Implicit and explicit TULVING, E. (1979). Relation between encoding spec- memory models. Bulletin of the Psychonomic So- ificity and levels of processing. In L. S. Cermak ciety, 13,339-342. & E I. M. Craik (Eds.), Levels of processing in ROEDIGER, H. L., III (1980a). The effectiveness of four human memory. Hillsdale, NJ: Erlbaum. mnemonics in ordering recall. Journal of Experi- TULVING, E. (1983). Elements of episodic memory. mental Psychology: Human Learning and New York: Oxford Univ. Press. Memory, 6, 558-567. TULVING, E., & BOWER, G. H. (1974). The logic of ROEDIGER, H. L., Ill (1980b). Memory metaphors in memory representations. In G. H. Bower (Ed.), cognitive psychology. Memory & Cognition, 8, Psychology of learning and motivation (Vol. 8). 231-246. New York: Academic Press. ROTHKOPF, E. Z. (1971). Incidental memory for loca- TULVING, E., SCHACTER, D. L., & STARK, H. A. tion of information in text. Journal of Verbal (1982). Priming effects in word-fragment comple- Learning and Verbal Behavior, 10, 608-613. tion are independent of recognition memory. RYLE, G. (1949). The concept of mind. London: Journal of Experimental Psychology: Learning, Hutchinson. Memory, and Cognition, 8, 336-342. SACHS, J. D. S. (1967). Recognition memory for syn- TULVING, E., & WATKINS, M. J. (1975). Structure of tactic and semantic aspects of connected dis- memory traces. Psychological Review, 82, 261- course. Perception & Psychophysics, 2, 437-442. 275. SCARBOROUGH, D. L., GERARD, L., & CORTESE, C. UNDERWOOD, B. J. (1965). False recognition produced (1979). Accessing lexical memory: The transfer of by implicit verbal responses. Journal of Experi- word repetition effects across tasks and modality. mental Psychology, 70, 122-129. Memory & Cognition, 7, 3-12. UNDERWOOD, B. J. (1969). Attributes of memory. Psy- SCHEFFLER, I. (1965). Conditions of knowledge. Glen- chological Review, 76, 559-573. view, IL: Scott, Foresman. UNDERWOOD, B. J. (1972). Are we overloading SCHMIDT, R. A. (1975). Motor skills. New York: memory? In A. W. Melton & E. Martin (Eds.), Harper & Row. Coding processes in human memory. Washington, SIMON, H. A. (1979). Information processing models D.C.: Winston. of cognition. Annual Review of Psychology, 30, WARRINGTON, E. K., & WEISKRANTZ, L. (1970). Am- 363-396. nesic syndrome: Consolidation or retrieval? Na- SLAMECKA,N. J. (1977). A case for response produced ture (London) 228, 628-630. cues in serial learning. Journal of Experimental WARRINGTON,E. K., & WEISKRANTZ,L. (1974). The Psychology, 3,222-232. effect of prior learning on subsequent retention in SLAMECKA, N. J., & GRAF, P. (1978). The generation amnesic patients. Neuropsychologia, 12,419- effect: Delineation of a phenomenon. Journal of 428. Experimental Psychology: Human Learning and WATKINS, M. J. (1981). Human memory and the in- Memory, 4, 592-604. formation-processing metaphor. Cognition, 10, SMITH, S. M. (1982). Enhancement of recall using mul- 331-336. tiple environmental contexts during learning. WERTHEIMER, M. (1961). Experimentelle studien Memory & Cognition, 10, 405-412. ueber das Sehen yon Bewegung. In T. Shipley MENTAL OPERATIONS 449

(Ed. and Trans.) Classics in psychology. New WOOD, G, (1983). Cognitive psychology: A skills ap- York: Philosophical Library. (Reprinted from Zeit- proach. Monterey, CA: Brooks/Cole. schriftfuer Psychologie, 1912, 61, 161-265.) ZECHMEISTER, E. B. (1972). Orthographic distinctive- WHORF, B. L. (1956). Language, thought, and reality. ness as a variable in word recognition. American Cambridge, MA: MIT Press. Journal of Psychology, 85,425-430. WIEKELGREN, W. (1973). The long and the short of ZOLA-MORGAN, S., SQUIRE, L. R., & MISHKIN, M. memory. Psychological Bulletin, 80, 425-438. (1982). The neuroanatomy of amnesia: Amyg- WICKENS, D. D. (1972). Characteristics of word en- dala-hippocampus versus temporal stem. Sci- coding. In A. W. Melton & E. Martin (Eds.), ence, 218, 1337-1339. Coding processes in human memory. New York: Wiley. WINSTON, P. H. (1977). Artificial intelligence. Reading, (Received July 18, 1983) MA: Addison-Wesley. (Revision received January 17, 1984)