casson - schemata in cognitive anthropology

7/30/2019 Casson - Schemata in Cognitive Anthropology

1/36

Ann. Rev. Anthropol. 1983. 12:429-62Copyright 1983 by Annual Reviews Inc. All rights reserved

SCHEMATA N. COGNITIVEANTHROPOLOGYRonald W. CassonDepartment f Sociologyand Anthropology,Oberlin College, Oberlin, Ohio44074

The aim of this review is to examine he "schema" oncept being developed ncognitive science from the perspective of cognitive anthropology. Cognitivescience is the inter-discipli.nary field that wasoriginally formed round he jointinterests of cognitive psychology and computer science and now includescognitive anthropology nd cognitive linguistics as well. Schematathe plural),which are knowledge tructures that are "the building blocks of cognition,"pervade theorizing about cognitive organization and function in cognitivescience (173). "Schema" s the most widely used term for these knowledgestructures (7, 22, 39-41, 51, 80, 95, 96, 102, !20, 121, 142, 144-149,151-154, 170-175,205,209, 212), but they are also referred to as "frames"(138, 219), "scenes" (69), "scenarios" (69., 159), "scripts" (1, 2, 183--188),"gestalts" (117-119), "active structural networks" 174), and "memory rgani-zation packets" (185)."Schema" nd these other terms, although they are conceptualized somewhatdifferently by different writers, dependingon their particular aims and in-terests, bear a striking family resemblance o each other. The emphasisherewill not be on differentiating amonghese telms and concepts, but rather onassemblinga compositeconceptualization that incorporates the most importantaspects of all these variants. Because"schema" s the most commonlyccur-ring term, it will be used for this composite oncept, and the other terms will bereserved for distinguishing types of schemata and someof their interrela-tionships.The schema notion and its importance in cognitive research have beendescribed and examined in numerous previous reviews. These have beenreviews concernedeither with cognitive science as a whole 21, 94, i44)with work n cognitive psychology 100, 101,142), cognitive social psycholo-

4290066-4294/83/1015-0429502.00

www.annualreviews.org/aronlineAnnual Reviews

Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo

ntificiaUniversidadCatolicade

Chileon09/12/07.

Forpersonal

useonly.
http://www.annualreviews.org/aronlinehttp://www.annualreviews.org/aronlinehttp://www.annualreviews.org/aronline


2/36

430 CASSONgy (89, 92,203), or computerscience (83,140, 158). To date there has beenreview that examined he importanceof the schemanotion in cognitive anthro-pology. Following eneral discussion of the schema oncept n the first part ofthe review, attention will be focused in the second part on what DAndrade57)has termed "the cultural part of cognition."SCHEMATASchemata re conceptual abstractions that mediate betweenstimuli received bythe sense organs and behavioral responses (212). They are abstractions thatserve as the basis for all human nformation processing, e.g. perception andcomprehension, ategorization and planning, recognition and recall, and prob-lem-solving and decision-making (173). Schemata heory developed in reac-tion against associationist theories, whichposited mental representations thatdirectly reflected the external world(89). As Tyler, among thers, has pointedout, "the structure of knowledgeannot consist of a merepicture of the worldoreven of a set of conceptswhich efer to or stand in a one-to-onerelation withelementsof the external world" 209, p. 98). Iconic representations of this sortare simply not rich enough o account for the complexity of human ehavior.Much sychological research, however, continues to be conductedon the basisof associationism. See, for example, Wickelgrens recent review of learningand memoryesearch, whichadopts a strictly associationist perspective, main-taining that "there is no evidence supporting the hypothesis of unitary schemanodes, although they could exist" (216, p. 37).Bartlett, who s generally credited with being the first to use the term schemain its contemporary ense [although Kant used the term in much he samesensein his Critique of Pure Reason 175)], argued that "the past operates asorganized mass rather than as a group of elements each of which retains itsspecific character" (7, p. 197). Remembering,artlett maintained, is construc-tive. Not all stimuli are stored in memory; ather, schemataare employed oprovide "a general impression of the whole"and to construct (or reconstruct)"probabledetails" (7, p. 206). Much f the criticism directed at associationismhas been concernedwith demonstrating that there are significant differencesbetweenextemal stimuli and mental representations, with showing hat sche-mata mayomit much etail or include more nformation than is contained in thestimulus (89).Basic to these criticisms is the view hat schemata ccurat differing levels ofabstraction. At relatively low levels of abstraction there are schemataforperceivinggeometricalfigures, colors, faces, etc, while at higher levels thereare schemata for comprehending omplexactivities and events. There are noimportant differences in kind betweenschemata or perception and comprehen-sion: "perception is comprehension f sensory input" (175, p. 110). Schemata


Annu.

Rev.

Anthropol.1983.1

2:429-462.Downloadedfroma

rjournals.annualreviews.org

byPontificiaUniversidadCatolicade

Chileon09/12/07.

Forpersonal

useonly.


3/36

SCHEMATAN ANTHROPOLOGY 431at particular levels of abstraction are not necessarily sensitive to schemata tother levels. Neisser(142, pp. 21-22)gives a particularly felicitous illustrationof this. Observingsomeone miling, he says, may nvolve only the perceptionof the shapes of teeth or changing positions of lips, or it may nvolve moreabstract comprehensionf a meangingful ultural act of "smiling," from whichinsights into the smilers mood e.g. happiness, cheerfulness, polite indiffer-ence) maybe gained.Schemata,unlike associations, are organic wholes comprisedof parts thatare oriented both to the wholeand to other parts (209, p. 109; 117, p. 246).Tyler states, they are indexical representations "foundedn holistic simultanei-ty," whereas ules are symbolic epresentations "founded n linear sequential-ity" (209, p. 100). Schemata are autonomousand automatic---once setmotion hey proceed to their conclusion--and hey are generally unconscious,nonpurposive, nd irreflexive; rules, in contrast, are conscious, purposive, andreflexive, i.e. they have feedback oops that enable self-modification (209, p.117). Tyler discusses the nature of rules at some ength, distinguishing "fivemajorbranchesof rule concepts": nstructions, precepts, regulations, uniformi-ties, and axioms (209, pp. 122-29).StructuresProbably he most influential discussion of schema heory is Minskys rticle,"A Framework or Representing Knowledge" (138). Minsky uses the term"frame" in discussing knowledgeepresentations, but he places frames in theschema radition stemming rom Bartlett. A frame, according to Minsky 138,p. 212), is "a data-structure for representinga stereotyped ituation, like beingin a certain kind of living room,or going to a childs birthday party" (138, p.212). This aspect of the schema oncept is elaborated by Rumelhart,who tatesthat a schemas "a data structure for representing he generic concepts tored inmemory," nd that "there are schemata representing our knowledge bout allconcepts: those underlying objects, situations, events, sequencesof events,actions and sequences of actions" (173, p. 34; see also 170, 171, 175). Thisrepresentation may e thought of as "a networkof nodes and relations" (138, p.212). It is the "network f interrelations that is believed o normally old amongthe constituents of the concept n question" (173, p. 34); moreover, t accountsfor "any situation that can be consideredan instance of the general concept trepresents" (171, p. 266).The highest levels of schemataare fixed and represent invariant aspects ofconcepts, whereas ower levels have erminals, or "slots," that mustbe filled byspecific instances of data, that is, they have variables that are associated with,or "boundby" elements in the environment n particular instantiations of theschema 138, p. 212; 173, p. 35). "Instantiation," a term used widely in theschema iterature, refers to the binding of particular elements to particular


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


4/36

432 CASSONvariables on particular occasions (173, p. 36). The schemaunderlying commer-cial events in our culture may e taken as an example 66, 69, 173). The schemahas the variables BUYER, SELLER, MONEY,GOODS, and EXCHANGE(upper case is used to distinguish conceptual units from words). BUYERsperson who possesses MONEY,he medium of exchange, and SELLERs aperson who possesses GOODS,he merchandise for sale. EXCHANGEs aninteraction in which BUYERives MONEYnd gets GOODS, nd SELLERgives GOODSnd gets MONEY. n event is understood as a commercialtransaction whenpersons, objects, and subevents in the environmental situa-tion. are bound o appropriate schemavariables.Variables, or slots, have associated conditions that restrict the elements hatmaybe bound o them. These conditions, known s "variable constraints," areknowledge bout typical values of variables (simple conditions requiring thatthe boundelement be a person, an object, or a subevent), and knowledge boutinterrelationships among ariables (more complex onditions requiring certainrelations among lements bound o sets of variables). Variables in the commer-cial event schemaare constrained by knowledge hat BUYERnd SELLERrenormally persons, that MONEYs generally currency, that GOODSre usuallyinanimate objects, and that EXHANGESnvolve transfers between partici-pants of objects that they ownor possess, as well as by knowledge hat theMONEYnd GOODSariables are interrelated: the value of MONEYovarieswith the value of GOODS173, p. 35).Variableconstraints not only restrict variable binding but also assign "defaultvalues" to variables for which no matchingelements are found in the environ-mental situation. Default values are expectations or "best guesses" determinedby the typical or normalvalues associated with variables. Because hey are notspecifically justified by the situation, they are only weaklybound o variablesand are easily detached if additional information is revealed that makesmoreaccurate value assignment possible (138, pp. 212, 228). In a transactioninterpreted as a commercialevent, a transfer of moneymaybe inferred evenwhenno moneys seen to have changedhands on the basis of the default valuesassociated with the MONEYnd EXCHANGEariables in the schema. Notethat this illustrates the point madeearlier that a schemamaycontain moreinformation than the stimulus environmentwith which t is matched.The use ofschemataas a basis for making nferences about stimulus situations is a topicthat will be taken up below.FramingThe term "frame" has recently been used by linguists, principally Fillmore(66-70) and Chafe (39-41), in discussing the relationship between tructureslanguage and underlying schemata. Language tself is, of course, a knowledgesystem--a system of conceptual abstractions, both unconscious, holistic sche-


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


5/36

SCHEMATAN ANTHROPOLOGY 433mata and conscious, reflexive rules (209, pp. 119-21). Althoughmost contem-porary linguistic analysis treats language olely as a system of rules, omittingany consideration of the role of schemata n languagestructure and process, aparticularly lucid analysis of both the rules and schematanvolved n past tenseformation was recently published by Bybee& Slobin (33). The frame conceptof Fillmore and Chafe s not concernedwith the organizationof language tself,but rather with how exical and grammatical orms both structure and expressunderlying schematicrepresentations (69, p. 127; 39, p. 46; see also 50, 64,102, 120, 121).The basic notion in framing is that lexical items and grammatical ategoriesand rules are associated in memorywith schemata and parts of schemata.Framesand schemata activate" each other: linguistic forms bring schemata omindand schemataare expressed in "linguistic reflexes" (66, p. 124; 67, p.25). Lexical and grammaticalforms are the meansby which schemavariablesare labeled and verbalized. "Frame" s an appropriate term because what theselanguage structures do is organize schemata for verbalization by focusingattention on certain variables and not on others. For example, a numberoflexical items can activate the commercial vent schema,e.g. buy, sell, pay,cost, spend, charge. Each of these words selects particular aspects of theschema or highlighting or foregrounding, while leaving others in the back-ground unexpressed 66, p. 25; 69, p. 103). Buyfocuses on the exchange romthe buyers perspective, andsell from he sellers perspective; cost focuses onthe money-goodselationship, and so forth. Further examplesof framing havebeen discussed in tworecent articles: Langacker 120) describes how he singlewordorphan rames a "functional assembly" f kin relations and the life cycle,and Quinn(159; see also 160a) discusses how he key wordcommitmentramesthree polysemous meanings, PROMISE, DEDICATION,and ATTACH-MENT,n the cultural scenario (i.e. schema) for Americanmarriage.The frame concept is generally used in a wider sense, originally formulatedby Bateson (12) and later elaborated by Goffman 82). Frake, in his recentdiscussion of the dangers inherent in plying frames, assumes his Batesonianview in describing frames as the basic units of interpretative context, as themeansby which"people organize their conception of what is happening t thistime" (74, p. 4). This more general notion of framing maybe dubbed"contex-tual framing" to distinguish it from "conceptual framing," the latter being aspecial case of the former (67). Contextual framing, like conceptual framing,involves associations betweenunderlying schematic representations and themeansof their expression. In contextual framing, the same reflexive rela-tionship holds betweenschemata and frames, but the framing of contexts isaccomplished ot only on the basis of linguistic forms but also nonlinguisticcommunicativemeans, such as tone of voice, facial expression, appearance,gestures, and body postures and movements 74). Underlying conceptions


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


6/36

434 CASSONevents are associated with the verbal and nonverbaldevices by which hey areexpressed, and these devices shape and organize the conceptions they express.

Contextual framing by meansof linguistic devices has been most extensivelyexplored in "The Ethnography of Speaking," a tradition of sociolinguisticresearch originated by Hymes97), whoremains its leading spokesman. n thisarea of research, whichwas reviewed not long ago by Bauman& Sherzer (13),speechevents and the language sed in realizing themare described n their fullreflexivity, e.g. poetry in a particular society is found to be expressed inlanguage characterized by such formal devices as rhymeand meter, and theselinguistic properties are showno be whatprovides or its recognition as poetry.Tannen 201) adopts a morestrictly linguistic approach n attempting to answerthe question, "Whats n a frame?" In analyzing oral narratives told by Greekwomenn recounting the contents of a film, she discusses 16 types of surfacelinguistic features in relation to underlying structures of expectation (i.e.schemata).Fine-grained analysis of contextual framing by nonverbal meanshas beendone by McDermott nd his colleagues (135), who examined the body posi-tionings, or postural configurations, assumed y a group of school children andtheir teacher during a reading lesson. Their analysis shows hat the teacher andchildren, in taking up and leaving positionings, both express the current state ofthe reading context and provide each other with the means or organizing theirinteraction in the ongoing lesson. Bassos (10) description of WesternApacheportraits of "the Whiteman"epresents somethingof a culmination of this lineof research. It is a seriously amusing account of how nteractions that arepotentially insulting and therefore dangerousare framed hrough a combinationof both verbal and nonverbal devices (e.g. a joking variety of WesternApacheEnglish, discourteous repetitions of requests and demands, a loud tone ofvoice, and repeated handshaking and backslapping) as nonserious "whitemanjoking"--as play in which"the Whiteman"s brought to life in vivid, whollyunflattering portraits.PrototypesA schema is also a prototype (51, 66-70, 102, 117, 173, 175). It isstereotypic, or generic, representation of a concept hat serves as a standard forevaluating the goodness-of-fit betweenschemavariables and elements in theenvironment.Variable binding is constrained by typical or average values ofschemavariables, not by absolute values (175, p. 105). Conditionson variablesspecifying that a boundelementmustbe, for example,a person or an object of aparticular sort define prototypical instantiations, but less typical elementsmaybe bound f they bear a family resemblance o it, i.e. if they are sufficientlysimilar to it (163-165,167). Similarity is, of course, a matter of degree: someelements resemble prototypes more closely than others. Default values are


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


7/36

SCHEMATA N ANTHROPOLOGY 435prototypical or near prototypical values. In the schema or commercial vents,BUYERs prototypically a person, although it maysometimes e bound by acorporation or institution; SELLERs prototypically a person who ownsGOODS,lthough less typically it may be a salesman or middleman; andMONEYs prototypically cash, but it mayalso be a check, a charge card, or anI.O.U.This notion of prototype schematahas lent itself to the development f atheory of meaning, recently christened "prototype semantics" by Coleman&Kay 51). A basic tenet of this theory is that the prototypeschema or schemata)underlying a concept corresponds to the meaningof the concept. Lexicalmeaning consists in a cognitive prototype to whichvarious real and imaginedevents maycorrespond in varying degrees" (51, p. 26). Linguistic forms,stated in discussing framing, organize and express schematicrepresentations;the point to be added now s that, because the schemataunderlying conceptualcategories are taken to be prototypes, these categories have analog representa-tions (164). Meanings re determined on the basis of overall resemblanceprototype schemata.Elements hat possess some but not necessarily all) of theproperties defining the prototype are instances of the concept. Thus, mem-bership in semantic categories is a "more"or "less" matter and the boundariesof categories are frequently blurred or fuzzy; elements re instances of categor-ies to a degree and instances differ in the degree o which hey are members5 l,102, 104, 105, 109, 111, 112, 116, 164, 165). This approach to meaning soften referred to in anthropologyas "extensionist semantics."The prototype theory of meaningcontrasts with what Fillmore (66) labels"checklist" theories of meaning.These re theories that posit digital representa-tions of concepts (164), and have in commonhe view that meaningscanspecified in terms of conjunctions of discrete properties, called variously"distinctive features," "semanticcomponents," nd "criterial attributes." Ele-ments hat display the list of properties defining a conceptare instances of theconcept. Thus, membershipn semantic categories is a "yes" or "no" matterand the boundariesof categories are clear-cut and definite; elementsare eitherinstances of categories or they are not and all instances are equivalent asmembers. Componential nalysis," as practiced in anthropologyand linguis-tics, is an exampleof a checklist theory (cf 11, 84, 85, 122).Prototype theory is not new n anthropology, although it recently has beentaken up with considerable enthusiasm in other disciplines, most notablypsychology, whereworkby Rosch 163) has triggered a burst of research. Thispsychological work has been reviewed by Mervis & Rosch(136). Although notalways acknowledged, he development of the prototype concept began withLounsburys work on kinship semantics (125-127,190). Lounsburys seminalarticles demonstrated hat the primary meaningof kinship terms derives fromthe genealogically closest of the types of kin (kintypes) that are included


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


8/36

436 CASSONinstances of kin categories, and that this focal or prototypical sense is extendedto the more distant kintypes also included in kin categories. Shortly afterLounsburys work, the extensionist approach was employedby Berlin & Kay(20) in their pioneering study of basic color categories and by Berlin and hisassociates (15-18) in their work on ethnobotanical taxonomies.Berlin &Kays book 20) set off an explosionof research on color classifica-tion systems, whichhas shown ather convincingly that color categories areorganized around prototypes (9 l, 164), and that they are nondiscrete or fuzzycategories in which membership s a matter of degree of approximation toprototypes (103). A further particularly important finding of this research,established most thoroughly by Kayand McDaniel, s that color prototypes are"based on panhuman europhysiological processes in the perception of color"(103, p. 644; see also 220). Rosch(164) argues that color categories arespecial type of category. Theyare prime examplesof what she calls "attributecategories," biologically based categories of perceptual sensations that are"general attributes of concrete things" (164, p. 23). Other examples f attributecategories are geometrical shape categories, which have "wired-in" prototypesfor circle, square, and equilateral triangle (163), and categories of facialexpressions, whichhave physiologically determinedprototypes for happiness,sadness, anger, fear, surprise, and disgust (61).Rosch(164) contrasts attribute categories with "object categories." Theseare categories of concrete entities, such as plants, animals, vehicles, furniture,and so forth, that are not assumed to be biologically based but that arenonetheless organized in terms of prototypes and approximations o prototypes.A particularly popular and productive domain or investigating the internalstructure of object categories has been he domainof "containers for liquids."Kempton 104, 105), for example, has shown that categories of drinkingvessels (cups, mugs, coffee cups) are fuzzy categories organized in termsfocal membersand grades of membership see also I09, 11 l, 112). Theprototype notion has also been applied in studies of social categories. The lineof research on kin categories initiated by Lounsbury as been continued in anumberof studies (14, 36, 189). Kay (102) has analyzed the prototype seman-tics of Tahitian race and class categories. Cantor& Mischel(34) have reviewedresearch on prototype categorization in abstract person categories, such asextrovert, madman,and activist. And Coleman & Kay (51) have studiedprototypical lies, speechacts that are whollyabstract entities.Embedding and LinkingA schema generally includes a number of embedded ubschemata as consti-tuent parts, each of which interacts in its ownright with elements in theenvironment 142, p. 23). A schema, in other words, is most often a complexstructure in whichvariables are boundby subschemata 138). Overall, then,


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


9/36

SCHEMATA N ANTHROPOLOGY 437schemataare organized as hierarchical structures in which schemata at thehigher levels represent the most general concepts, and schemataat successivelylower levels represent moreand morespecific concepts. Schemata t the lowestlevel are atomic, i.e. primitive concepts hat are componentsf knowledgehatdo not break down nto constituent parts or further subschemata 174, p. 106).Taxonomiesre the type of hierarchical organization formedon the basis ofrelations of class inclusion. A commercial vent can be understoodas "buyingdinner in a restaurant" or as "buying a Whoppern a fast food restaurant," forexample. Other types of hierarchical organization of schemataare formed onthe basis of other types of conceptual relations, e.g. part-whole relations,causal relations, Fillmores case relations, among thers (65; see also 37,209).Understood generally, a commercial event framed as buying involves aBUYER who EXCHANGESMONEYwith a SELLER for GOODS. EX-CHANGE,owever, is not a primitive concept; its schematic representationcontains subschemata for DO, CAUSE, nd TRANSFER.t represents anevent in which a BUYERDOES omething that CAUSESwo TRANSFERSof possession, one of MONEYrom BUYERo SELLER nd another ofGOODSrom SELLERo BUYERthis example is simplified from (78), pp.212-25]. A commercialevent, then, maybe comprehendedither in general interms of a major schemaalone, omitting any consideration of its internalstructure, or more deeply in terms of embedded ubschemata 175, p. 106).

Schemata are also organized sequentially. Subschemata embedded n aschemamaybe ordered to represent changes over time or in location, cause-effect relationships, and sequencing f stages or actions in events (138, p. 234).Continuing the commercialevent example, the exchange of goods and moneyinvolves wostates of affairs, the state prior to the exchangend the state afterthe exchange.The transfer of possession that characterizes exchanges s repre-sented by two temporally ordered subschemata that share the same set ofvariables but in different arrangements: ne represents the earlier state in whichthe BUYER POSSESSES the MONEY nd the SELLER POSSESSESthe GOODS,nd the other represents the later state in which the BUYERPOSSESSES the GOODS and the SELLER POSSESSES the MONEY(138, p. 240).Schemata re not only organized into complex ierarchical structures, theyare also interlinked with other schemata o form still larger structures. In acommercial vent, the money ne participant gives another maybe specified asa tip, bribe, ransom, uition, retainer, change, rebate, etc (69, p. 114). Theframingof this money s a tip, for instance, serves to characterize it as moneygiven in exchangeor services (rather than goods)and, at the same ime, to linkthe commercial event schema with a wider schema, about which numerousinferences can be made,e.g. that the services involvedare those of a waiter orwaitress in a restaurant (66, p. 28).


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


10/36

438 CASSONLinkage such as this to wider schematadiffers from linkage that organizesschemata nto ordered sequencesor chains. "Eating in a restaurant," to take a

widely discussed example, is an event whose epresentation is comprisedof asequence of linked schemata--ENTERING, ORDERING, ATING,and EX-ITING 183,188). These schemata, termed "scenes," are in turn comprisedsequences of actions; TIPPINGhe waitress and PAYINGhe check are two ofthe constituent actions in the EXITINGcene, for example. The sequenceof actions defining scenes and larger-scale events is an elaborate causal chain:each action in the sequence esults in conditions that enable the next action andmust be completedbefore the next action can be started (188, p. 45). The term"script," fnst adopted by Schankand Abelson 1, 2, 183-188), has come othe standard label for these sequences of schemata. Scripts are considered atgreater length in the second part of this review.ProcessesSchemata re not only data structures, they are also data processors. As Neisserputs it, a "schemas not only the plan but also the executorof the plan. It is apattern of action as well as a pattern for action (142, p. 56). Schematare activeprocesses whoseprimary activity is the construction of interpretations ofexperience; they are procedurescapable of evaluating their owngoodness-of-fitto elements in the environmentand of thereby accounting for them(173, pp.37, 39).As stated, the top levels of schemataare fixed and invariant, whereas hebottom evels contain variables that are boundby elements in the environment.All the schemata interrelated in a complexschemaoperate individually inprocessing information at their respective levels, and supply each other withinformation for processing. Moregeneral higher level schemata direct theoverall processing of information, motivatingand coordinating the activities oflower level embedded ubschemata; more specific lower level schemata passinformation along to other low level schemata hat follow them in sequentialorder and feed information up to the wider schemata in which they areembedded 142, pp. 56, 124).There are two basic modes of processing. "Bottom-up processing," alsotermed"data-driven processing," is processing initiated whendata are bound ovariables in bottom evel subschemata hat moveupward o activate the higherlevel schemata in which the subschemata are embedded. "Top-downproces-sing," also called "conceptually driven processing," is processing initiatedwhen top level schemata activate embedded ubschemata in the expectationthat these subschematawill fit the data (22, p. 140; see also 21, 45,146-148).Data-driven processing movesfrom part to whole, and conceptually drivenprocessing moves rom whole to part (173). Data-driven processing is subcon-scious, automatic, and guided by the principle that "all the data must be


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


11/36

SCHEMATAN ANTHROPOLOGY 439accounted or," while conceptually driven processing is conscious, purposive,and guidedby high level plans and goals (22, p. 148; see also 209, pp. 98-112).Top-down nd bottom-up processing occur simultaneously, and each re-quires the other. Expectations are combinedwith data in constructing inter-pretations of experience, and schemataare judged to account for elements inthe environmentwhen here is goodness-of-fit between expectation and data(22). An example liberally amendedfrom Rumelhart (173, pp. 42-43)make his somewhat bstract discussion of "mixed nitiative processing" moreconcrete. The event in question occurs in a large enclosure in which auto-mobiles are on display. A number f people are milling around, chatting, andexamininghe automobiles.Thesetting, people, objects, and activities suggestthat the automobiledealership schemamaybe relevant to achieving an under-standing of the event. Thusactivated by bottom-upprocessing, the high-levelauto dealership scheman turn sets off top-down rocessing of its subschemata.Becausean expectation associated with this high-level schema s that auto-mobiles will be bought and sold, one of its subschemata s the nowfamiliarcommercialevent schema. Top-down rocessing activates this subschemata,initiating searches for data that maybe bound o appropriate schemavariables.The GOODSariable may potentially be bound by the automobiles, and awell-dressed manengaged n kicking tires and slamming oors is a reasonablecandidate for the BUYERariable. The fact that the man is well-dressedsuggests wealth, a possible binding for the MONEYariable. And he presenceof a broadly smiling man in a checkered sportscoat who approaches thewell-dressed manand engages him in jovial conversation suggests a potentialbinding for the SELLERariable.

The binding of these variables activates new bottom-upprocessing, whichconfirms the expectations of the auto dealership schemaand eliminates fromconsideration alternative general schemata hat might have been nvokedat theoutset of processing, e.g. the auto show cheman which he expectation is notthat automobileswill be boughtor sold but that they simplywill .be exhibited.This examples a heuristic fiction, oversimplified o the point of artificiality,but it illustrates how op-downand bottom-up processing work in concert toevaluate the success of schemata n providing interpretations of experience.Morecomplexand realistic examples of schematic processing can be found inarticles by Rumelhart (170-175) and Bobrow& Norman 22, 144-146).MemoryThere is general agreement hat schemataare mental representations located inmemory--that they are frameworks selected from memorywhen new situa-tions are encountered (138). The structure of memorytself, however, hasreceived relatively little attention (151, 185). Discussionof memoryas for themost part been conducted in terms of Tulvings (207) distinction between


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


12/36

440 CASSONsemantic and episodic memory.The basic difference generally recognizedbetween these two types of memory s that schemata in episodic memoryrepresent specific knowledge, .e. knowledgehat is idiosyncratic, particular,and directly reflective of subjectively experienced timuli, whereas chemata nsemantic memoryepresent general knowledge, i.e. knowledge hat is encyc-lopedic, relatively permanent, and known ndependently of experience (151,175).Researchers have tended to emphasize one or the other of these types ofmeaning. Some 52, 53) have concentrated on abstract semantic categorieslinked in hierarchical networksby class inclusion relationships, e.g. animal,bird, robin. Theories of semantic memory ave recently been reviewed bySmith (195). Others (183,184, 188) have focused on experiential groupingsconcepts linked together by cooccurrence n events and episodes, e.g. hammer,wood,nail. Schank, he leading advocate of episodic memory, oes so far as toargue that memorys basically episodic in nature, that very few--perhaps nomore than ten--hierarchical supersets actually occur in memory183, p. 255).In recent work, however, Schankproposes different levels of memory truc-tures for representing different sorts of knowledge; hese levels--"event mem-ory," "general event memory," "situational memory" nd "intensional mem-ory" vary in the degree to which hey reflect particularexperiences (185, pp.259-61). The semantic/episodic memoryssue has not yet been resolved, but itis reasonable to conclude, even if only provisionally, that memory ustcontainschemata hat represent concepts abstracted from particular events and episodesas well as schemata hat represent concepts formed on the basis of particularexperiences.CULTURAL SCHEMATASchemata iffer in their distribution in populations: someare universal, someidiosyncratic, and somecultural. Universal schemataare uniform n the humanspecies because of innate faculties of the mindand/or inherent divisions in thenatural word; idiosyncratic schemata re unique to particular individuals as theresult of their personal histories and life experiences; cultural schemataareneither unique to individuals nor shared by all humans,but rather shared bymembers f particular societies (37, pp. 20-21). Cognitive anthropologists areconcerned with cultural schemata, with "the cultural part of cognition," asDAndrade57) has recently phrased it. Unlike cognitive psychologists, whowant to learn how he brain deals with all sorts of information, includingcultural information, cognitive anthropologists focus on how ystems of cultu-ral knowledge re constrained and shaped by the machineryof the brain. Theassumption ere is that, as the result of its regular transmission romgenerationto generation, cultural knowledge omes o be organized in such a way hat it


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


13/36

SCHEMATA N ANTHROPOLOGY 441"fits" the capacities and constraints of the humanmind (57, p. 182).succinctly stated by DAndrade,he program or studying cultural cognition isto "search for commonalitiesn knowledgeystems built into language n orderto find out basic characteristics of humanhought" (56, p. 3).The anthropological concern with the interface betweensystems of culturalknowledgend basic psychological actors is clearly illustrated by Lounsburysworkon kinship semantics. Lounsbury, t will be recalled from the "prototype"discussion, demonstrated hat categories in systems of kin classification areregularly organized in terms of prototypes (focal kintypes) and extensions(nonfocal kintypes that are equated with focal kintypes). This general findingabout kin class~ification systemspoints to the existence of underlyingpsycholo-gical mechanismshat interface with the prototypes-with-extensions organiza-tion of kin categories (57, p. 182). Lounsburys athbreakingwork,as alreadymentioned, has been followed up by a great deal of research in manynon-kinship domains. In psychology, Roschs work (163-168) has been particular-ly significant in spearheading efforts to elucidate the psychological factorsinvolved in prototype categorization. Lakoffs (117a) most recent work hasalso contributed importantly to this line of research.In the following sections, research exploring the organization of complexcultural schemata is reviewed. Attention is first focused on three types ofcomplex chemata ound in the organization of cultural content: object schema-ta, orientation schemata,and event schemata. Thenbrief surveys are presentedof research concernedwith the role of these complex chemata n two currentlyvery active areas of cognitive research, metaphorand narrative.Object SchemataRecent research has advanced understanding of the complexschemata under-lying a wide range of object classification systems. Object categories, asmentioned n discussing prototypes, are categories of concrete entities. Someof the object classification systems hat have been studied are classifications ofplants (15-20, 26, 28, 90), animals(27, 29, 93), manufactured bjects includ-ing vehicles, tools, clothing, and furniture (168), containers (104, 105,109),persons (34), kinsmen 189, 190), occupations (32), ethnic identitiespersonality descriptors (108, 215), illnesses (55), and emotions (128).Taxonomies,which occur widely in cultures and have been studied extensively by anthropologists, are a familiar type of classification system. Intaxonomicclassifications, wider, more general and narrower, more specificcategories stand to each other in class inclusion, or "kind of," relations (see 37,pp. 75-77). The English ethnobotanical category tree, for example, includesthe oakcategory, andoakis a kindof tree. Other categories included n the treecategory are maple, pine, elm, spruce, and poplar. The tree category isincluded in the superordinatecategoryplant, as are the categories bush, grass,


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


14/36

442 CASSONand vine, and oak includes the subordinate categories white oak, post oak,scrub oak, red oak, jack oak, pin oak, and burr oak.Ethnobiological taxonomiesare the most thoroughly studied object classi-fication system. Thanks largely to the work of Berlin and Brownand theirassociates (15-19, 26-29), much s nowknown bout their structure, growth,and development.A major finding is that categories in ethnobiologieal taxono-mies are grouped into mutually exclusive hierarchical ranks, whichhave beenlabeled "kingdom," life form," "generic," "specific," and "varietal" (16). Thekingdom ank category is the all-inclusive unique beginner; it is the highestlevel category and delimits the entire domain e.g. plant in English ethnobo-tany). Life-form rank categories are the next most inclusive categories. Theyare few in number nd are labeled by primary lexemes e.g. tree, grass, vine).Categories of genetic rank are next in inclusiveness. By far the most numerouscategories in ethnobiologicalclassification systems, generic categories are alsogenerally labeled by primary lexemes (e.g. oak, maple, pine). Genetic categor-ies that dominate lower level categories (manydo not) immediately includespecific rank categories, which in turn include categories of varietal rank.Specific categories are few in number nd varietal categories are rare. Lexemesfor both are generally binomial secondary lexemes e.g. white oak, post oak,scrub oak).Berlin (16, 17) has shown hat genetic rank categories comptise he coreethnobiological classifications in simpler societies. They re not only the mostnumerous ut also the psychologically most salient or basic categories. Basiclevel categories are those categories in a classification system that are mostfrequently used in everyday interaction, most easily recalled by informants,and earliest in ontogenetic development (58). Dougherty (58) has revisedBerlins account of basic leyel categoties, demonstrating that while genericrank categories are most salient in the ethnobiologicalclassification systemsofpeople who nteract frequently with their biological environment and whosesubsistence depends irectly on it, categories superordinate o the generic level(i.e. life-form categories) are mostsalient in the systemsof people whodo notmaintain a high degree of interaction with their biological environment.Whichlevel is basic in a taxonomicclassification system is not fixed but ratherdetermined elative to the overall salience, or cultural significance, attached bysociety memberso the objects classified in the system.Basic level categories have been described for nonbiological classificationsystems as well. Roschand her associates (168) have established the existenceof a "basic object level" in their studies of classifications of manufacturedobjects such as vehicles, tools, clothing, and furniture. Vehicle, for example, sa superordinate category dominating ar, bus, and truck, which are basic levelcategories dominatingsuch subordinate categories as sportscar, city bus, andpickup truck (168). Basic level categories have long been recognized n studies


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


15/36

SCHEMATAN ANTHROPOLOGY 443of color classifications, e.g. red, green, yellow, blue (20), and kin classifica-tions, e.g. uncle, aunt, nephew, niece (190). Cantor & Mischel (34)described basic level person categories, e.g. extrovert, madman, oker,activist.Another type of classification system that has received attention is theconstituent, or partonomic,classification system (6, 25, 31, 62, 134). Higherlevel and lower level categories in this kind of classification system areconnectedby part-whole relations. In English ethnoanatomical lassification,for instance, fingernail is a part of finger, finger andpalmare parts of hand,handand forearmare parts of arm, and arm s a part of body(6, pp. 347-48).Like taxonomies, partonomies also have a basic level of abstraction: hand,foot, and eye are basic level categories that are dominatedby superordinatecategories arm, leg, and face and in turn dominate subordinate categoriesfinger, toe, and pupil. Other constituent classification systems include mem-bership classifications such as family, which has among ts membersather,son, and brother, and baseball team, which has the membersshortstop,pitcher, and manager 176).Functional classification systems are constructed on the basis of in-strumental, or "used for," relations. Superordinateand subordinate categoriesin this kind of classification systemare related functionally. Oneway n whichshoes, for example, re classified in our culture is by function:a joggingshoeisa shoe used for jogging, and a tennis shoe is a shoe used for playing tennis.Many asic level categories in classification systems hat have been reated, astaxonomically related to superordinate categories are shownactually to befunctionally related (183). Roschs axonomies f manufactured bjects are,fact, not strictly taxonomies t all: a vehicle is any kind of object that can beused as transportation, and car, bus, and truck are categories of objects thatfunction as vehicles. While elations of basic level categories to subordinatecategories are generally taxonomic, elations to superordinate categories arevery often nontaxonomic183, 218). It mayeven be the case that superordinatetaxonomic ategories occur only in ethnobiological classifications (218).This survey of object classification systems supports two mportant observa-tions. First, it establishes that the complex chemataunderlying conceptualsystems are not simply ad hoc collections of simpler schemata, but ratherintegrated, organicwholesoriented arounda basic level of abstraction. Second,it also supports the idea that underlying psychological capacities must beshaped in a way that is consistent with the basic level organization found inobject classification systems.Attributes, as pointedout in discussingprototypes, are properties or qualitiesof objects, typically features of form, suchas color, shape, texture, or size, andfeatures of function, such as use for sitting, for clothing, or for food. Whats acategoryand what s an attribute of a category s determined elative to level of


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


16/36

~d CASSONanalysis (136). Redandblue, for example, re at one level subject to analysis ascategories, as the large literature on color classification testifies (20, 91,103,164, 220). At a higher level RED nd BLUE re units in the analysis ofcategories. Theyare primitive elements connectingconcepts with objects in theworld that do not decomposento constituent parts.Attributes are of two sorts. Absoluteattributes are inherent properties ofobjects (a ball, for example, may have the properties RED,ROUND,ndSMOOTH).elative attributes, on the other hand, are not properties of objectsthemselves. Theyare properties of relationships betweenobjects and norms orthose objects (a person maybe OLD r RICHelative to the normsfor age andwealth) and between objects and other objects (an owner is a person whoPOSSESSESn object and a father is a person who s the male PARENTF achild). Moredetailed discussion maybe found in Casson (37, pp. 84-86; seealso 35, 49, 63, 122, 197,209, 214).The basic level in classification systems s the most nclusive level at whichperceptual and functional attributes are shared by most members f each of thecategories and at which contrasting categories are maximallydiscontinuous(165, pp. 30-31). The categories chair and car are examples. Members fsuperordinate categories share only a small numberof attributes: furnitureincludes chairs, tables, and beds, andvehicle inch/des cars, buses, and trucks.In both cases, included objects share few properties. Members f subordinatecategories share bundles of commonttributes, but these attributes overlapextensively with the attributes of contrasting categories: kitchen chair, deskchair, easy chair, and city bus, school bus, cross-country bus are sets ofcontrasting categories that have manyoverlapping attributes in common.Orientation SchemataOrientation schemata are complexschemata that represent knowledge aboutspatial orientations (110). Particularly in anthropology,schemataof this typeare often referred to as "cognitive maps." Their overall organization is hierar-chical. A complex chema epresenting general spatial concepts occurs at thetopmost level, and a numberof subschemata epresenting object concepts areembedded t lower levels. Orientation schemata represent knowledgeaboutspatial relations among bjects and their relative positions in the physicalenvironment; hey always include a representation of the self because "Egoandworld are perceptually inseparable" (142, pp. 113-17). Kuipers (110,132-34), in proposing a formal model of spatial knowledge, argues thatcognitix~e mapscontain three classes of representations: representations forknowledge bout particular environments, descriptions of the current positionof the self (the "YouAreHere" pointer), and representations of processes thatmanipulate he other two kinds of knowledge, .e. routes, whichare proceduresfor moving he "YouAre Here" pointer through the environmental representa-tion.


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


17/36

SCHEMATA N ANTHROPOLOGY 445Cognitive maps ange from representations of very small- to very large-scalespaces. Maps f relatively small-scale spaces have been formulated by Neisser(142, pp. 111-13), whodescribes a cognitive mapof an office and its setting,and by Linde & Labov 123), whodescribe schematafor the spatial layouts

apartments. In The Image of the City, Lynch (!29, pp. 46-83) describeslarge-scale mapsof portions of cities, showinghat they generally contain fivetypes of elements: paths (routes through the city), edges (boundaries, suchrivers or railroad tracks), districts (sections of the city), nodes junctions wherepaths meet), and landmarks (outstanding features, such as tall buildingsmountains).Wallaces 211) classic account of "driving to work" llustratesuse of just such a large-scale cognitive map. T. Gladwin81) and Oakley 150)have described the complexorientation schemata used by native navigatorsduring lengthy sea voyages in Oceania.Orientation schemata are used not only in traveling through cities andnavigating at sea. Theyare also used in imagining ity travel and sea voyages.Cognitive mapscan be detached from their original environments and usedabstractly to picture the environments hey represent (142). Detached n thisway from actual environments, they can serve as mnemonic evices. The"Method f Loci" is a well-knowndevice employedwidely in performing featsof memory23). First invented by the ancient Greeks, the methoddependsthe representation in cognitive mapsof a series of locations along a path orroute. Items to be memorized re imagined one at a time in association witheachof the locations along the path. Recall of the items in the original order, inthe reverse order, or individually out of sequence s achieved imply bytakingamental stroll along the path and examining the images associated with thevarious locations. In an example related by Harwood 88), the holes offamiliar golf course were used as the loci where images of 18 items formemorization ould be mentally placed. "A brown Bron-) slavic (-slaw) skier(-ski) lying sick (Mal-)" could, for example, be mentally located near atrap on the third green, so that later, in attempting o recall the third item in aseries, the nameBronislawMalinowskiwouldbe brought to mind by envision-ing the third green and its associated image 88, p. 783).Cognitive maps, like classification systems, are organic wholes. Theyarenot just assemblages of object schemata, but rather integrated conceptualsystems that include specific object concepts in general representations ofspatial knowledge142). Furthermore, ust as basic level organization in objectclassifications suggests the existence of matching sychological actors, so the spatial organizationof object concepts n cognitive maps uggests thd existenceof correspondingly organized underlying psychological mechanisms23,142).

Attributes of orientation concepts are properties of physical space, e.g. upand down, boveand below, all and short, left of and right of, in front of and inback of. UP/DOWN, BOVE/BELOW,nd TALL/SHORTre relative attri-butes because they are properties of relationships betweenobjects in physical


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


18/36

446 CASSONspace. Anobject has the attributes UPor DOWN,or instance, dependingon itsposition on a vertical axis (defined by the force of gravity) relative tohorizontal plane (ground level) (43, pp. 241-42). Attributes like LEFT/RIGHTand FRONT/BACKre defined in terms of spatial relationships to a canonical,or prototypical, person (44, 54, 132, 139). This person, or self, is typicallyupright, has significant parts (e.g. head, front, back, right side, and left side),and movesand sees in a forward direction (43, 77).Event SchemataEvent schemata epresent a wide range of activities and interactions, varyingfrom simple actions like giving and taking to complex cenes like ordering ameal in a restaurant. In studying event concepts, investigators have devotedconsiderable attention to determining he elementaryunits out of whichsche-mata for events are constructed. Researchers in linguistics (38, 65, 98, 115,133, 169,217), artificial intelligence (182-184, 188), and psychology 137,174) have devised numeroussystems of primitive elements. The earliest ofthese theories were proposed by linguists working in Generative Semantics,particularly Lakoff (115), and Case Grammar, articularly Fillmore (65).principal aim of this linguistic research, whichpostulated primitive predicatesand case relations as elementaryunits, was to represent the semanticstructuresunderlying lexical items and sentences. Research n artificial intelligence andpsychology temming rom this linguistic workadopted the same aim, plus theadditional goal of developing computermodels of these semantic representa-tions (175, 182-184).These theories, which are very similar despite notational differences, allposit sets of primitive predicates and recognize a basic distinction betweenprimitive actions and primitive states (see especially 65, 115, 182). Primitiveacts are predicates that specify active relationships in propositions. Examplesare DO, GO, PROPEL, GRASP, CAUSE, CHANGE. rimitive states arepredicates that specify nonactive states of affairs. Examples nclude BE,TALL,ALIVE, KNOW, OSSESS.Some investigators (e.g. Schank) main-tain that while here is a large number f primitive states, there is a limited set ofprimitive acts, perhaps as few as 11 (183,188). Others (e.g. Minsky) rguethe collection of primitive concepts s quite large (138).These theories also propose elementary concepts specifying the ways inwhichargumentsare related to primitive predicates in propositions. In theirinfluential original formulations, Fillmore (65) and Chafe (38) posited sets"cases" .or "roles" to account for relationships betweenconcepts in argumentpositions and their predicates. These notions have been clarified and refined inmore ecent research (68, 78, 141,174). A current list of cases would ncludeAGENT, OBJECT, RECIPIENT, EXPERIENCER, INSTRUMENT,LOCA-TION, SOURCE, and GOAL.


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


19/36

SCHEMATA N ANTHROPOLOGY 447To illustrate, the GIVEchema s a relatively simple action schema 78; seealso 174, 175). It is composed f a predicate and three arguments: an AGENTGIVESan OBJECTo a RECIPIENT.GIVE s not, however, a primitiveconcept. Its representation breaks down nto subschemata for DO,CAUSE,and TRANSFER;hat is, it is an event in which an AGENTOESomething toCAUSE an OBJECT to be TRANSFERREDo a RECIPIENT. DO andCAUSEre primitive concepts, but TRANSFERs not. The representation ofTRANSFERncludes further subschemata for CHANGEnd POSSESS,which are primitive elements; it is an event in which POSSESSIONf anOBJECT CHANGES rom AGENT o RECIPIENT.Schemata, s described earlier, are linked into ordered sequencesor chains.This is accomplished y way of causal, or contingency, relations. Actions and

states are conditionally connected:a state is usually or necessarily followedoraccompanied y an action, and an action is the usual or necessary consequenceor concomitant of a state (see 37, p. 82). Schank (183, 188) very usefullydistinguishes five kinds of causal relations that are needed n accounting forlinkages between schemata: RESULTausation (action RESULTSN statechange), ENABLEausation (state ENABLESction), DISABLEausation(state DISABLESction), INITIATION ausation (state or act INITIATESmental state), and REASONausation (mental state is REASONOR ction).Causal inking maybe illustrated by specifying the relations that connectstatesand actions in a simple event: John, being thirsty, opensa can of beer and takesa drink (188, p. 28-30).

John is THIRSTYstate)INITIATESJohn DESIRES eer (mental state)REASON FORJohn DO omething (unspecified action)RESULTS INbeer can OPENstate)ENABLESJohn INGEST eer (action).

Persons and objects in events stand to states and actions in the various caserelationships: John is AGENT f DOand INGEST nd EXPERIENCERfTHIRST and DESIRE.Linked schemata of this sort that represent recurrent, conventionalizedactivities and interactions are known s "scripts." Schankand Abelson 1, 2,183-188),whooriginated the script notion, have generally defined a script as"predetermined, stereotyped sequence of actions that defines a well-known


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


20/36

448 CASSONsituation" (188, p. 41). Recently, however,Schank as revised this definition,stating that a script is not a data structure that is available in one piece inmemory, ut rather a structure that is reconstructed from memorys it is neededto interpret experience(185, p. 264; see also 24). Agars(3) accountof gettingoff, or shooting up(injecting heroin), a central event in the culture of streetjunkies (urban heroin addicts), can be rephrased as an exampleof a script. Theorganization of some of the sequenced actions comprising this event is asfollows:

junkie COOKSheats) heroin and water mixture (action)RESULTS INheroin DISSOLVEDstate change)ENABLESjunkie DRAWtransfer) heroin into works (action)RESULTS INworks PRIMEDstate)ENABLESjunkie HIT insert) works nto vein

Case relations in this event include: junkie is AGENTf COOK, RAW,ndHIT; works is INSTRUMENTf HIT and RECIPIENTof DRAW.The GETTINGFFscript is an example of what Schank & Abelson (188)call an "instrumental script." It is an invariant sequenceof actions that isemployed, enerally by one participant, to accomplish particular task (188, p.65). Dougherty & Kellers (59) "taskonomies" can be regarded as furtherexamplesof instrumental scripts. The two other types of scripts identified bySchankand Abelsonare "personal scripts" and "situational scripts." Personalscripts are idiosyncratic sequencesof actions that single actors use to achievepersonalgoals(e.g. making date with a waitress in a restaurant), and as suchare not of much ultural interest (see 199). Situational scripts have been the

principal focus of script research. Theypertain to specific situations---charac,teristically institutionalized public situations---in whichseveral participantsassume interconnected roles and, on the basis of shared understandings,cooperate to achieve certain well-defined goals (188, p. 61). ORDERINGnrestaurant is a situational script, comprisedof a chain of actions and stateslinked by causal relations, e.g. EXAMININGhe menu, CHOOSINGooditems, SUMMONINGhe waitress, etc (188, p. 42-43).The GETtINGOFF and ORDERINGcripts may be viewed from a widerperspectiveas constituents of still larger conceptual tructures, as subscripts orscenes in wider scripts. GETTINGFF s causally linked in a cyclical chain ofevents with two other events, COPPINGbuying heroin) and HUSTLING(getting money)4, pp. ~.~. ~.6):


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


21/36

SCHEMATA N ANTHROPOLOGY 449junkie POSSESSread (state)ENABLESjunkie COP eroin (action)RESULTS INjunkie POSSESSeroin (state)ENABLESjunkie GET OFF/SHOOT P heroinRESULTS INjunkie BE STRAIGHTi.e. not sick) (state)ENABLESjunki~ HUSTLEread (action)RESULTS INjunkie POSSESSread (state)

ORDERINGn a restaurant, as stated earlier in discussing linking, is one offour scenes that comprise the full restaurant script: ENTERING,RDERING,EATING,nd EXITING188). Several of Frakes important articles (71-73)are accounts of aspects of Subanun nd Yakan ulture that could be rephrasedas scripts.Afinalpoint to note about event schematas that events represented n scriptsare very likely basic level categories. Rosch065, pp. 43-44) has tentatively

concluded on the basis of pilot research that events like "makingcoffee,""taking a shower,"and "going o statistics class" are basic level event categor-ies, as contrasted with such superordinateevent categories as "getting out of thehouse in the morning"or "going to afternoon classes" and such subordinatecategories as "picking up the toothpaste," "squeezing he toothpaste," and soon. More esearch is neededhere, but this seems o be another case in which heorganization of systems of cultural concepts suggests underlying psychologicalmechanisms.MetaphorRecently, the nature of metaphorhas become topic of considerable interest inanthropology, linguistics, and psychology. A vast topic, metaphor has beenapproachedfrom a large numberof theoretical perspectives (9, 99, 152-156,177, 181, 206). The strategy here will be to focus on metaphor researchemploying schema theory.One line of research has been concerned with showing howschema heorycan account for the comprehension f metaphors (152-154, 172). In generalterms, a metaphor states an equivalence between two concepts from separatedomains.The metaphor,"George s a lion," for example,states an equivalencebetweena human eing, George, and an animal, a lion. Georges the "tenor" ofthe metaphor, the concept that is continuous with the topic of discourse,


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


22/36

450 CASSONand lion is the metaphors "vehicle," its discontinuous concept (180, p. 7).Placing the two concepts in juxtaposition, the metaphor orms a concept thatsubsumesboth tenor and vehicle (9, pp. 96-97). The WesternApaches, o citeone of Bassos (9) examples, convey the concept "living earth dweller whowastes food" by juxtaposing carrion beetle (vehicle) and whiteraan(tenor)their "Wise Words" metaphor, "Carrion beetle is a whiteman." In schematheory terms, the comprehensionof metaphors like "George is a lion" and"Carrion beetle is a whiteman"depends on similarities and dissimilaritiesamong he values bound to variables in the schemata underlying tenor andvehicle (154, pp. 359-60). General variables that are similar are maintainedvariables in the metaphorical concept (George and lions are ANIMATEBEINGSand carrion beetles and whitemen are LIVING EARTHDWEL-LERS);specific variables that are dissimilar are omitted from the metaphor(George does not have PAWSr TAILand whitemenare not INSECTS); ndleast one specific variable (or subschemataembeddedn a variable) thatshared is recognized as a crucial variable in the metaphoricalconcept (Georgeand lions have COURAGE, TRENGTH,nd AGRESSIVENESSnd white-men and carrion beetles WASTEOOD) 9, 180).A second line of metaphor research employing schema theory deals withlarger-scale metaphorical processes. This is the research of Lakoff & Johnson(118, 119), which is concerned with demonstrating that muchof everydayexperience is structured by metaphorical concepts. In essence, Lakoff andJohnsonsarguments that basic abstract concepts that are not clearly deline-ated in experience, such as ARGUMENT,IME, LOVE, and IDEAS, aremetaphorically structured in terms of other basic concepts that are moreconcrete in experience, such as WAR,MONEY, RAVEL,and FOOD.Metaphorical structuring is not simply a matter of individual concepts butrather of "experiential gestalts" multidimensional structured wholes (i.e.schemata) that coherently organize experience in terms of natural dimensionsof experience, e.g. participants, parts, stages, causes, purposes (119, p. 81).

The metaphorical concept ARGUMENTS WARs an example that Lakoff& Johnson 119) treat at some ength. Conversation hat is seen as argumentunderstood as such on the basis of the ARGUMENTS WARmetaphor.ARGUMENT,n abstract, not clearly delineated concept, is "partially struc-tured, understood, performed, and talked about in terms of WAR," hich is amore concrete concept that emergesnaturally from the experience of physicalcombat 119, p. 5). The WARchemaspecifies an event in which participantsare adversaries whoattack and defend positions, plan strategies, maneuver,advance, retreat, counterattack, declare truces, surrender, and triumph. TheARGUMENTchema, derived systematically from the WARchema, repre-sents an event in whichparticipants are adversaries embattled in a conflict ofopinions (not actual combat), who truggle over positions, gain ground, winlose, and so on.


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


23/36

SCHEMATA N ANTHROPOLOGY 451Lakoff & Johnson (119, p. 7) argue that the systematicity of metaphoricalconcepts llustrated in this examples reflected in the languageused in talking

about these concepts and that, as a consequence, inguistic expressions are asource of insight into and evidence for the nature of the humanconceptualsystem. Again, he research strategy, as stated in the passtige quoted earlierfrom DAndrade56, p. 3), is to look for regularities in knowledgetructuresthat are built into language to discover basic properties of thought. Someeveryday speech formulas or fixed-form expressions reflecting the ARGU-MENTS WAR etaphor are evident in the following sentences (119, p. 4):

Your claims are indefensible. He attacked every weak point in myargument.His criticisms wereright on target. I demolishedhis argu-ment. He shot downall my arguments.

Lakoff & Johnson 118, 119) distinguish three types of metaphors: ontolo-gical, orientational, and structural. Ontologicalmetaphors re used in compre-hending events, actions, activities, and states. Events and actions aremetaphorically onceptualized s objects, activities as substances, and states ascontainers. The metaphor IDEASAREFOOD,which entailsIDEASAREOBJECTS,s apparent in a great many expressions.

What e said left a bad aste in mymouth.All this paper has in it are rawfacts, half-baked ideas, and warmed-overheories. I cant swallow hatclaim. Thats food for thought (119, pp. 46-47).VITALITY S A SUBSTANCEs evident inShes brimmingwith vim and vigor. Shes overflowing with vitality.Hesdevoid of energy. That took a lot out of me. Im drained(119, p.51).and LIFE IS A CONTAINERs reflected inIve had a full life. Life is empty for him. Her life is crammed ithactivities. Get the most out of life (119, p. 51).

Reddy 161) has analyzed a particularly intriguing and subtle example, the"conduit metaphor,"whichstructures both lay and many inguistic conceptionsof language.Orientational metaphors re used to structure abstract concepts that are notwell grounded n experience n terms of concrete conceptsarising from experi-ence with spatial relationships. For example, UP-DOWNetaphors include,amongmany others,HAPPY IS UP/SAD IS DOWNIm feeling up. My pirits rose. Youre n high spirits. Im feeling down.Im depressed. Hes really low these days.


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPontificiaUniversidadCatolicadeChileon09/12/07.

Forpersonaluseonly.


24/36

452 CASSONHEALTHIS UP/SICKNESS IS DOWNHes at the peak of health. Hes in top shape. He fell ill. Hecamedown

with flu. His health is declining,CONTROLIS UP/BEING CONTROLLED S DOWNImon top of the situation.. Hes n a superiorposition. Hes n the upperechelon. Hes under my control. He fell from power. Hes low manonthe totem pole (119, p. 15).

Structural metaphors are used in comprehending omplex, highly structuredconcepts. Whileontological and orientational metaphors re basic in providingthe meansof referring to concepts, quantifying them, orienting them, and soon, structural metaphors provide for the use of "one highly structured andclearly delineated concept to structure another" (199, p. 61). In structuralmetaphors, he constituent structures of complex bject, orientation, and eventschemata serve as the means of structuring other complex schemata. TheARGUMENTS WARmetaphor is an example of a metaphor based on acomplex event schema. LIFE IS A GAMBLINGAMEs another structuralmetaphor involving a complex event schema. LOVE S A JOURNEYs astructural metaphor based on an orientation schema:

Lookhow far weve come. Wereat a crossroads. Well just have to goour separateways. I dont think this relationship is going anywhere199,p. 44).

Examples of metaphors based .on complex object schemata have also beenpublished: Basso (8) has described how he Western Apache lassificationhuman ody parts is used metaphorically in classifying parts of pickup trucksand automobiles, and Casson (36) has shownhowa Turkish kin address systemis extended metaphorically in addressing nonkin.NarrativeNarrative is another burgeoning rea of research in anthropology, inguistics,and psychology.Like metaphor,narrative is a huge topic that has been studiedfrom a variety of perspectives (see 30, 42, 60, 75, 76, 79, 87, 114, 124, 194,202, 208). Again, the strategy here will be to review only research employingthe schemaconcept.The three types of complex chematadescribed in this review have all beenshown o provide the underlying organization for narrative discourse. Narra-tives based on event schematahave been by far the most extensively studied,undoubtedly because they are the most widely occurring, and will, for thisreason, receive the most attention here. But interesting accounts of the use of


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


25/36

SCHEMATA N ANTHROPOLOGY 453object and orientation schemata in organizing narrative have also been pub-lished. In describing "the way of the hot pepper," a Cunacuring chant usedagainst high fever, Sherzer (193) showshow he Cuna"hot pepper" .t.axonomyis projected onto a parallelistic verse pattern. The chanter inserts names ortaxonomic ategories into. the verse pattern, beginning t the top of the hierar-chy and movingdown hrough a series of included subcategories, then return-ing to the top and movingdown hrough another series of included subcategor-ies, repeating this process through as manyas 53 verses (193, pp.. 283-84).Linde & Labov 123) describe the use by apartment dwellers Of cognitive mapsof the layouts of their apartments n producing imaginary our" narratives, andHarwood88) discusses the use by Trobriand narrators of a spatial orientationschema epresenting locations in the TrobriandIslands as a mnemonicevice inrecounting particular mythsand relating them o other myths n the totality ofTrobriand mythology.Eventschemata hat are used in producing arratives are generally referred toas story grammarsor story schemata. They are global structures that breakdown nto two majorconstituents: a setting subschema,which s comprisedof aseries of states specifying the time and place of the story, and an episodesubschema, which is comprised of an external event and the protagonistsreaction to it (175). Episodes generally involve a problem-solving motif:something occurs that makes he protagonist set up a goal, which he or sheattempts, successfully or unsuccessfully, to attain (171, 175). There have beenmany ormulations of story schemata, which; although they vary in detail,generally incorporate this problem-solving motif (107, 130, 131, 162, 170,171, 198,204, 210). The EPISODEch,e, maproposed by Rumelhart(170, 171,175) is representative. This schema pecifies relationships among everalvariables: an initiating EVENT, GOAL, nd an attempt, or TRY, o accom-plish the goal. The TRY ariable is a subschematahat specifies the internalstructure of the attempt(or attempts) to attain the goal. It takes much he sameform as the General Problem Solver proposed inNewell & Simonrs (143)theory of problem-solving.Rumelharts 171, 175) story about Mary nd the icecream ruck illustratesthis EPISODEchema.

Maryheard the icecream truck comingdown he street. She rememberedher birthday money nd she rushed into the house.The major constitutents of this EPISODE re an EVENTMary HEARSheicecream truck), which initiates a GOALMary WANTScecream), whichthe reason for a TRYMaryTRIESo get icecream). The principal constituentsof the TRY chemaare an ACTIONselecting a~problem-solving method, hereBUYING)nd a GOALthe icecream). The BUYchema, it Will be recalled,


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPo


Chileon09/12/07.

Forpersonal

useonly.


26/36

454 CASSONrequires the BUYERo have MONEYMaryrushes into the house and gets laerbirthday money) 171, pp. 271-73; 175, pp. 113-15).Thebulk of the story schemaesearch is culture bound n that it is designed oaccount only for stories in the general European radition (106). Thoughthardly news that story-telling traditions are quite different in non-Westernsocieties, only a handful of studies have systematically examinedculturaldifferences in story schemata (46~48, 106, 162). This work has, however,revealed specific differences in the waystory schemataaffect the comprehen-sion and recall of stories. Particularly impressive s Rices (162) demonstrationthat American ollege students recall American tories more accurately thanEskimostories because the former fit their Americanstory schemaand thelatter, whichhave heir owndistinctive structure, do not.

Story schemata, as described to this point, are hierarchical structures inwhich processing is top-down. A second approach to narrative comprehensionconcentrates on bottom-up processing. Schank (183, 184, 187, 188), theleading proponent f this approach,argues that stories, are interpreted not onlyin terms of high level schemata like SETTING nd EPISODE,but moreimportantly in terms of the stereotyped sequences of actions that comprisescripts. The restaurant script, for example, is invoked in interpreting thefollowing simple story;

John went o a restaurant. He ordered chicken. He left a large tip. (188,pp. 47--48)Mention n the story of a restaurantand the actions of orderingandtipping issufficient environmentalnput to instantiate the restaurant script. On he basisof this instantiation, the story is understoodmore ully than it is represented nthe three sentences. Actions not mentioned n the story are supplied as defaultvalues of script variables, so the story is interpreted as

John went to a restaurant. He sat down. He read the menu. He orderedchicken. Heate the chicken. He eft a large tip. Hepaid the check. He eftthe restaurant (188, p. 48).Gaps n the causal chain linking actions in the event are filled in by inferencesmade n the basis of knowledgeepresented n the script. It is explicitly statedin the story that John ENTEREDhe restaurant and ORDEREDhicken, so itmay be inferred that in between these actions he SATdownand READhemenu; it isalso specified that John, having ORDEREDhicken, TIPPEDhewaitress, so it maybe inferred that he ATEhe chicken; and because TIPPINGprecedes PAYINGnd LEAVING,t may be inferred that John EXITEDherestaurant (188, p. 48). "The waitress," "the menu," and "the check," which


Annu.

Rev.

Anthropol.1983.1




Chileon09/12/07.

Forpersonal

useonly.


27/36

SCHEMATAN ANTHROPOLOGY 455are not mentioned n the story, are included in the inferred account becauseknowledge f the roles and props (i.e. agents, recipients, instruments, etc)involved n restaurant events is also contained n the script.

Inference is a major topic in discourse analysis, to which investigatorsworking n several traditions have contributed (5, 86, 87, 113, 114, 157,176,178,191,192, 196,200). In addition to Schanks (183-188) pioneering work,schema theory research on inferences and event chains has been done byWarren, et al (213), whohave developed an "inference taxonomy,"a classi-fication of mutually exclusive categories of inferences, including three majortypes of inferences and a number f subtypes: logical inferences (motivation,psychological cause, physical cause, and enablement), informational infer-ences (pronominal,referential, spatiotemporal, worldframe, and elaborative),and value inferences (evaluative). In discussing a particularly detailed ethno-graphic example,Trobriand itigation discourse, Hutchins(95, 96) describesgeneral schema or possession and transfer of land tenure rights, showinghowactions and logical connectives that are only implicit in discourse about landtenure cases are supplied by inferences madeon the basis of cultural knowledgerepresented in the general schema. Quinn(160) provides a detailed criticalreview of Hutchins contribution to the continuing developmentof schematheory in cognitive anthropology.CONCLUSIONThe preceding sections have described schema heory and its wide range ofapplicability in cognitive anthropological research. Research nd results frommany eemingly disparate fields of cognitive study have been drawn ogetherand discussed in terms of a single, comprehensive nd coherent, explanatoryframework.Newwork in cognitive anthropology specifically guided by sche-ma heory has been considered and familiar workhas been reconsidered n lightof this perspective. One conclusion suggested by this review is that schematheory offers a broad, unified theoretical frameworkhat has the potential tointegrate research in cognitive anthropologyand cognitive science generally.Although already evincing someof this enormouspromise, schema heory isstill in the early stages of its development. t present, it is undoubtedlyverlypowerfuland too general. It is invoked n providing accounts of a great manydifferent kinds of data, often in a way hat does not take into consideration heparticularity of specific aspects of cognitive organization and processing. Theworkof the next fewyears hopefullywill fill in the details of a unified schematheory and specify the factors that constrain the structure and function ofspecific varieties of schemata.A secondconclusion, then, is that the develop-mentof schema heory offers a very exciting prospect for future research incognitive anthropology.


Annu.

Rev

.Anthropol.1983.1

2:429-462.D

ownloadedfroma

rjournals.annu

alreviews.org

byPontificiaUniversidadCatolicadeChileon09/12/07.

Forpersonaluseonly.


28/36

456 CASSONACKNOWLEDGMENTSA number of people were kind enough to read a draft of this review and offerresponses, reactions, criticisms, suggestions, and bibliographic information. Iwant particularly to thank MichaelAgar, Keith Basso, Roy DAndrade,Charles rake,Naomi uinn, ndStephen yler. This reviewhas benefited nmany ays rom dviceofferedby these individuals,but theyare in no way obe blamedor anyerrorsor infelicities of thought r expressionommittednthe review.These re my ault entirely.Literature Cited

1. Abelson, R. P. 1975. Concepts for repre-senting mundane eality in plans. SeeRef. 21, pp. 273-3092. Abelson, R.P. 1976. Script processing inattitude formation and decision-making.In Cognition and Social Behavior, ed. J.S. Carrel, J. W. Payne, pp. 33-45. Hills-dale, NJ: Edbanm3. Agar, M. 1974. Talking about doing:Lexicon and event. Lang. Soc. 3:83-89.See Ref. 37, pp. 114--204. Agar, M. 1975. Cognition and events.See Ref. 178, pp. 41-565. Agar, M. 1980. Stories, backgroundknowledge and themes: Problems in theanalysis of life history narrative. Am.Ethnol. 7:223-396. Andersen, E. S. 1978. Lexical universalsof body-part terminology. In Universalsof Human anguage.Vol. 3: WordStruc-ture, ed. J. H. Greenberg, C. A. Fergu-son, E. A. Moravcsik, pp. 335-68. Stan-ford, Calif: Stanford Univ. Press7. Bartlett, F. C. 1932. Remembering:AStudy in Experimental and SocialPsychology. Cambridge: CambridgeUniv. Press8. Basso, K. H. 1967. Semantic aspects oflinguistic accnimration. Am. Anthropol.69:471-779. Basso, K. H. 1976. Wise Wordsof theWestern Apache: Metaphor and semantictheory. See Ref. 11, pp. 93-121, Ref.37, pp. 244-6710. Basso, K. H. 1979. Portraits of "theWhiteman" Linguistic Play andCulturalSymbols among the Western Apache.Cambridge: Cambridge Univ. Press11. Basso, K. H., Selby, H. A., eds. 1976.Meaning n Anthropology. Albuquerque:

Univ. NewMexico Press12. Bateson, G. 1972. A heory of play andfantasy. In Steps to an Ecologyof Mind,ed. G. Bateson, pp. 177-93. NewYork:Ballantine13. Bauman, R., Sherzer, J. 1975. The

ethnography of speaking. Ann. Rev.Anthropol. 4:95-11914. Bean, S. S. 1975. Referential and indexi-cal meanings of ammain Kannada:Mother, woman, oddess, pox, and help!Southwest J. Anthropol. 31:31 3-30. SeeRef. 37, pp. 188-20215. Berlin, B. 1972. Speculations on thegrowth of ethnobotanical nomenclature.Lang. Soc. 1:51-8616. Berlin, B. 1976. The concept of rank inethnobiological classification: Some vi-dence from Aguaruna folk botany. Am.Ethnol. 3:381-99. See Ref. 37, pp. 92-11317. Berlin, B. 1978. Ethnobiological classi-fication. See Ref. 166, pp. 9-2618. Berlin, B., Breedlove, D. E., Raven, P.H. 1974. PrinciplesofTzeltalPlantClas-sification: An Int

casson - schemata in cognitive anthropology

Documents