You are on page 1of 18
The focus is on 38800 ‘of attention in social cognIEION Spodels, which are areas Of in what sense are repre Rie he use of represents ding on factors such as a iit, or ist depen common themes across models and conc! mn themes a aspects of implicitness (6 ‘ional models, which suggest that importan effort) may depend on i Key Words: mental representation, unconscious cognition, implicit cognition Theories of mental representation are always metaphors (Wyer, 2007). We talk of bins, of nodes, and of races 5 ways o transform abstract informa tion into solid topics of study. In some ways, this abstemess can be dangerous, leading to debates about semantic ideas that might be difficult or Iimposible o vei: We belive that such mode Bis nets important case maps : phos mater Beas they can rai ‘ily change Peeeiics understand the ‘any i ‘ r operation history of sciences full of xan i hanging capo changed everything Bon changed the way wc) an ocial scienti , 1999), mPer MeAPhor has en . ache ro ees tore Pe ck, gs well a5 07 © active grow inferences or construals peop! implicit social cognition, implicit repr | entation and how these relate to imp | ond connectionist models, which have 2 odied cognition, situated cognition, Ir Each theory is scrutinized in relationto_ | Jicit or explicit? Second, are the licit? Third, are representations | eention? The chapter discusses | judes with a discussion of attribu- uch as consciousness, intent, and ice about their own representations, | repres -sentations imp! tions imp! -esentation, consciousness, | dominant metaphor for the human mind. This is mainly because computers are easier to unde stand than minds, It is interesting chat with ret advances in understanding neural networks, em puters are pots much more like brains. Mor importantly, models of brains are becoming mF Ue bss Abstract or not, as knowledge about felt grows, the metaphors inevitably chang with Wall metaphors working better than others. Whe iy Harvey discovered the circulation of ‘ lood, the heart bec chan the eater i ; aaa outs Previously understood to be Suddenly vn “tious sounds made alot more sess aes cls of mental representation help ™ chapter ie ings in social cognition? Our aim in! ‘0 overview theories of mental repress tion, We do ae Pestive of implicit social tite or ae, pt Sealuate whether the heoties » But we consider which ones af ame a pump, rather ¢ sentation that this same group is peefectly sociable. “The representation that f ative ata particular tie ic nd a Ee person perceives, interprets, Bacay cy responds in situations involving that group. BB vines anyen hers emphasize that given representation can be Nae tpl pecmahas orcas aperd (Jacoby & Kelley, Ecc, jeg tales teva Jasechiko, 1989). So 2 person might access a stereorype about a specific racial group, and whether that stercotype translates into judgment or behavior will depend on how much awareness and intentional control the person I roosts of : has over its application (Devine, 1989). Our review a ight of as a state versis a will consider h ce ne consider how each representational model 1 physical, bodily state (Barsalo Ye do not incorporate these features into jtion because the theories of mental repre- al al repre- always implicit, as opposed to momentarily implicit pnthat we review differ on these important Your ae as breathi ee Hither than defining any of these Features probably outside of aw Se at tee addresses implicit representations versus implicit tise of re fn presentations ‘A related issue is whether a representation is read this ess, Beak the start, we will address the specific sentence. In this example, the represenration Wis 7 SE esséeiated with each theory in the course of momentarily implicit, but 2 shift in arrendon 0 sion. make it explicit. In contrast, no shift of attention 4 Bs or anything else can make explicit the processes b | This Is Implicic? which your brainstem is controlling your breathing. Ries Begin by adopting a conventional definition Some represenation, 1 be inaccessible under recent es sari in Baglis (999) "our ay Grane ‘Where posible, we conside ll Ofautomaticty: Implicit representations whit implicit about implicic representations from both the momentary and permanent perspectives ness, are efficient, do not require intent A these berween implicit and ‘Why does the distinction cxplicis mental representations matter. We believe fig this distinction matters because it Geape control. Notice that how we fake sense when applied to mental repre Bas may differ depending on your theory of ination. If we think of a representation Being ineridy in memory, then ic makes Bey ithar a person is unaware of the rep For inscance, a person might hold an Facial stereotype about a specific th: ansmitted through exposure © PoP : Bur thinking of representations inert esymbols causes problems for the other cri ‘can a symbol be efficient oF inefficient? finintended? Controlled oF Uno” of the thing, racher than the thing said to —reristics id to have these characteris PAYNE, CAMERON | 22 that explori how we think about the co mental representations as implicit or explicit (ol of behavior. informs Placing helps us to f have desires or belief such as racial stereotypes—th our considered - implici in themselves — J-then that oals and values. IF or counter ¢ representations a Tunconscious of tnintentiond Ihniques for raising mental repre- ss, IF implicie rep ed, on the implicit that iss suggests specific re entation into conscious aware sean implic in how they as reper hand, then cat suEBESS We need to cultivate orerene ofan nce Data influence our ici tap re of mental eco om sever emporrans ca ton). Wea ls whose focus i on dying Be eecandconclied procencs rather slr pros and dual seme mode). We th fore focus on theories ‘of mental representation tha M afucnal and on those that we bei — have bee fave oe influential i impli socal cognition incerp™ ; sheth mre carry @ Address Key Questions About Implicit and Expl sce eo imply Aacthe presentations Are representations impli & Ts efiicoriepresenaion _cemporary oF permanenih? wi swe implicit? mean Rmienieprestingaciva» | Representations and their use | Representations can cua De rercmrcious repre arc ipl unl hey pass explicit when they x For i Mentaiondareimplicicunril the threshold of conscious threshold cious acti that they pas the threshold of awareness vation after acral repre eels swarenets through network a no Sy Seema rcprcuations ar Schemas and . obje Beeeeensbot casio nocimplicic in themselves but ence ae only temporas encebhavior and judgments are used implici inf tren y implicit based on ditetion ol = Dime represcniations Active epresen = - det Betrriion Acne pesaionl pa Representational pats i Re craeti | Ersaeconcousbucncr canbe male conscions crwork isnot seis generally workings of net impli presentational Some representational ses systems, both Fepresentations are permanently implicit and usear implicit fr other others an only ery ewes others are only temporarily _2tems, only use is implicie Bodily repr ;resentations and their use can be cith Saerscia) plc explicit jemporarily implicit, but this might vary by sensory modality Situated repre only temporarily implicit, at ‘his might vary contextual formed i G septs tend to be . The nodes have vatying levels [on prior experience and current links between the nodes are of ‘There is some controversy Breil coder abun CAT vw. SIAMESE CAT), and ks or relationships beeween concepts spropositional meaning (i.c., does the link jnodes “CAT” and “MISCHIEVOUS sition “cats are mischievous”?; Fiske = 1991; Wyer & Carlston, 1994). Bur, fy associative network models agree that eis constructed dynamically and bottom-up finked-combinations of related nodes. fstance, Fazio'’s (1990) MODE model posits n attitude is the association between a node ening an attitude object (e.g., a person) and fepresenting an evaluation of that attitude fg. good or bad). five networke models have been especially for understanding accessibility and prim- Gas. The strength of links is thought to be Bel the frequency and recency of ativa- (Higgins, 1996; Higgins, Bargh, & Lombardi ‘connected nodes, like electricity oF water finding. its way through the contours in a tock face (Anderson, 1983; Collins 8 Loftus, 19753 Wyer 8 Carlston, 1979). This spreading activation is typically treated as unconscious. But when activa- tion crosses a high enough threshold, the concept is said to come to mind, meaning that itis consciously thought about. In what sense, then, are represent tions implicit or explci? Parallel search models sug gest that people are not aware of the nodes themselves or of the spreading of activation along the links. However, they become aware of the concept that results from nodes that have reached the threshold of consciousness. Parallel search models also imply that the process of spreading activation is efficient, in that people do not need to expend cognitive effort; acti- vation automatically spreads along interconnected linkages. Spreading activation needs neither inten- tion to begin, nor can people control the spreading, once it has started. Thus, parallel search versions of associative network theories st tations (ie nodes and the links among them) are o all four criteria of ; are mplicit memory ee ee ee me reading activation Links ang He og euch dul proces SOE rey might belp make clear the lying, nese ideas and the then juments that result. mental representation have 1: in social cognition also bee ; ple 10 the gestalt in_ principle geste Towarch, dl Ee cips7)-anea foe and Bruncrs (1957) fa ation psychologists * 4 the information given? 5 “go beyond wal judgments. Unlike the bottom-up { by associative network models, @ a top-down approach: sche: ns that structure and in making social dynamics posited schema models take broad representation Ihological experience. These large be though of following linked 0 fate & Bower, 19731 onski, 1986). For instance, if p consciously remember something {ys might consider specific cues mnake sense of psye ‘Chunks of general knowledge might " 3, following the y 3 ie sion as different lenses with which co view ne a ees pplicnion Like lenses, schemas arc activated anc SOME fo trek mode, atime ahough the thes schemas . Re ceerectesunconscous, ing lees of accessibility that bring chem closer a tect ficss ofthe ot father from being acvated. Once activa» Ree procsetacls owerce is schema brings online all of its general knowles a eerste) eforfulineentional, and and, asa consequence, it directs attention, memory i ary control. Thus, representations in and judgment. ‘models arc only momentarily implicit Consider some classic examples of schemas in Setietisnifaticacion (ic, action. In-a famous experiment, Chen and Bargh p the search) thar determines what is (1997) subliminally primed participants with either and what is explicit, black or white faces as they were completing 2 vis [and serial search models share the basic alization task. Priming was meant to activate the “OF nodes and Tinks. They differ dia- stereotype schema for black individuals as hostile etivation spreads, and in what : : for the implicitness of representations es of models are not necessarily exclu- cause they work on the same basic ight be said t0 make use of ‘implicitly (Via parallel search) Participants then interacted with another petso ina taped interaction. Compared with participan’s Sompared with pa who were sub tic nally primed with white faces: PX tlpants who were subliminal primed with black a were rated by their interaction partners as dis- playing more hostility. M r, ts who y. Moreover, participan ted been primed with black faces also reported tht My grounded ins the (hegpariners were more hostile Thus, activation of i the black-hostili : and inks. 1 ve gil in shea me act more ma made people be schema, and then incerp®et ‘euflatory hostility as due to the inherent hostility ol the interaction partner. oe schema theories have not been fF loped to address the implicit-exPi* fecheros In the form of penton anda fee carly Lae oryping research. In research on and attentional processing in animals and humans — traits are commonly considered a Schwarts, 8 Pourtols, 2011). Although — Hpummarizing a persoris behaviors tives approaches might add much greater predic: BEES Seercorypes are considered” ba specificity to schema theories, they remain cae of group of po bs norte me stem in sol PY aon Hippel. 1996). Studies of how Pieper Hreauggests that inferences about traits pontancously (ies without inte) Mealy (ues efficient) upon learning aeonsbehaviors (Carlton & Skowronski ei 8 Uleman, 2003; Uleman, 1987). research, They are intended to mimic the network March on dispositional atcibutions, srdcate of human newons. Like asia are likely to ascribe behaviors roan work annette eros a a fe Gs opposed co constraints of che stue processing nodes at Hing Ta pacity is low (Gilbert 8 which are connected by weighted Tinks allowing for the transmission of this activation (Rumelhart, McClelland, & the PDP Research Group, 1987) tal differences between Connectionist Models Connectionist models have been widespread in cognitive psychology for the past two decades, but have only slowly been brought into social cognition. ‘when cognitive ca $1985). Finally, many studies have shown fpeotypes cin be activated automatically (eB 1996; Devine, 1989; Gaertner & Bellin, 1983; Payne, 2001; for reviews, see (2001; Bargh, 1999) and that people are mort fis stereotypes when cognitive capac © thausen, 1990; Govorun & Payne 2006; Milne, & Bodenhausen, 1994) Togethes, an be used effi- se are fundament Unlike associative network However the two types of models. models, most connectionist mod that specific nodes have semantic Me Tepresenttionsaredstbuve as mer patterns eros the entre set of connestd nodes. When he network eventually sertles et of inputs, at satisfies the parallel constraiais ‘nodes and the weighted eaninsss nem Importantly, distributed fps ce because there are anywhere it 1996). Different dels do not assume ning, Rather, of into a parern of the activated tions berween th rations are NOt esentations work (Smithy lidies suggest that schemas ¢ thout deliberate reotypes fthout intent, and wi com the content of the traits ste Bias may be available to conse Beiation and applicacon in shinking (Kunda & ¢ discre' resent distinct TP! ‘stored is presumably unconscious st ne y connection 7 creations accion aay recombinations of ERS le connection weights. Meichsracterize the impliciencss . ire sane bof nodes and connect ee 2 ' embeitol Ae vsentations arestiobSUSS Soe rime precisely because, a8 wand Moreover TePES ee” pr veibuted pattern (OF argued, these models are more, the network a = een ease Imany models of representa? represen 07) oly \arned changes it mith, 1996). Schemas 7597" escribe a perso! ea a oo ; Figure 1.2 Schemaicilasion ist model oa connection! work might CAT Consider how 2 connectonst nenwore tele no aprerfor he eepicenton Bee weet eo tion isa seco inputs dhe nerwork, which cou, Gyles such as "daw “meow and “POT Ove lengthy expen such cues a co her via a set of connection weights ro is fine the disribuced representational patter Fe SCAT? Thus, the system learns to more quickly seit into this patern in the future based on prior Jeaming. However, there is no single distributed rep- resentation for “CAT.” Rather, each time, the nodes and weighted connections will reach a slightly dif ferent approximation based on the combination of prior experiences, the specific features fed into the network, and other contextual influences Whereas most treatments of conncctionism sug. gest that single nodes only have meaning in the ontextof the larger distributed patcern, others have argued that single nodes can have semantic mean ing, asin associative networks (Bowers, 200; 2000; Read, Vanman, & Miller, 1997), Such locals, fonneetionist models retain the parallel constrainy Satsfetion logic described above, but allow for dic ‘rete, static representations. For argued for “grandmother calls’—singe- ney that represent specific things, places, or such as your grandmother. Dr of single-cell electro stance, some h some have OF persons, more similar €0 associative network networks ae Mi ncctionist models because thee models than 10 ¢ iption of distributed pat models abandon the assum ion (Smit oe operating conditions disingish hh, 1996). Table 11.2 lists some nd wey features an jative networks, d connection. ing assoc ine networks, and localist conncctionist nerwotks (cf, Smith, 1996). Localist connectionist modds Jistributed re similar to associative network models in many rays, but they are distinguished by che operating inciple of parallel constraint satisfaction (Read ny & Mille Although the locals versus distributed debate is controversial, it suggest that the distributed representation assumption may not be necessary within connectionist models Research applying connectionist_models has not typically been concerned with questions of consciousness or automaticity. Instead, most stud ies have used connectionist networks to model the learning and use of patterns. However the operating characteristics of such neeworks are SUB: Bestive. Connectionist networks are thought 1 always be active, constantly changing and setling {Rte new patterns (Conrey 8 Smich, 2007). And decauise these patterns are cither recombinations erations of the same set of nodes and connection weights, they will ‘i of stored of stored inform, The distributed Reworks migh ight ‘ind unintentional always include some uninvited bis ation (Carlston & Smith, 1996) Tepresentations of connections. herefore be described as efit I. Pe pattem, but not the workings ® Network itself, and it is not Ae a the final emerge COPE tend to become ws the connectionist ‘Connection wei : . ‘weights change Nodes and links change very very slowly Spreading activation with parallel constraint satisfaction Meaning of nodes dacs not change depending on context, so another mechanism is needed to account for contex ‘tual influences Mfhis machinery could be made fully conscious. control does seem possible: ycalist some cases ppests We are capable of inputting new cues into the | may Fetionist network to modulate the pattern that ly emerges (Conrey & Smith, 2007). has feast some of the time, these new cues might 1s of Blitatively unique. Some theorists have posited tional explicie processor overlaid on top of Ge connectionist processor (Sloman, 1996: 1996; Smith & DeCoster, 2000; Smolensky processor is said sug B). In such cases, the explicit Slate via conscious manipulation of lingu s. By contrast, the esitional representation: fist processor governs the unconsci associations. The snd. associative yy thae activa sed ng of es between connectionist models suggest the possibilit fitern above some threshold could be explicitly represented concep’ However, fir oflany research that has taken this short, the currently availble search I teasiccasonist models ae Po” of implicit representa’ One way M leaning of nodes changes depending on context, resulting in now distributed pattern auielly Spreading activation with parallel constraint satisfaction Both ways Meaning of nodes does not change depending on context, so another mechanism is needed to account for contex tual influe Connectionist models are able to account for many of the same effects as associative network and schema models. Instead of appealing to spreading activation to explain evaluativeand semanticpriming effects, connectionist models can utilize the notion of pattern completion and. overlapping. patterns (Carlston, 2010; Smith, 1996). Ti take it easier to complete s tern may ilar patterns representing other concepts. This can also account for recency effects because recently instantiated pat= terns make the connect to fall into similar patterns fi lity effects can be Jearned changes in weights ore likely jonist network more likely jor a certain period of time. Chronic accessib accounted) for by fen nodes, making some parterns m ‘e network's ourputs than others appeal to slowly betw to emerge on th Finally, ennnectionist models can account For sehes varie processing: tiggeing one portion of a Biss al parcern makes it easier for the rest to complete. C ifstinction between process and con fin of connection. weights becween as the process of representing representation of the pattern donot makea tent. ‘The pat nodes acts both onnectionist networks PAYNE, CAMERON \ 27 ‘accrue it is sill unclear ‘supposed ro accourtt ic-and explicit repre- {mith & DeCoste P1988), others accoune for explicit reP bby appealing to contol a : ‘alter pattern completion ( FS 007, Garon 8 Bodenbausn, 2000 Format Models he aims of the foregoing models are broad: they f Intendsoacenunforhowinformarionsrpresenc However, some recent models suggest j ~ remenigaatt Hinds of representation that re processed in qualitatively distinct ways ) Garlstonis associated systems theory (AST Girkton, 1994) posits four distinct process ‘systems: verbal, visual, behavior " Giticaly representations in cach ofthe four systems ate qualitatively distinct. The individual systems arc -amranged hierarchically, from lower level representa ‘that correspond 10 neural and physiological to higher level representations that are consciously accessible, and amenable idescription. These higher level representa ‘of propositions in the verbal system, "visual system, goals in the behavioral ‘verbal representations of emotion in Higher level representations lower level representations within They can also intermingle ete ACTOS systems, lead systems. For and verbal and affectiv fently f eel eT ee ‘adh which this happen eoeput the ‘on irrelevan PEOGNITION AND seunra, evel “ble to. conscious —e al fem. For instance, afc I ee are more Sel Tocused n verbal and visual re difficult scious: from syste funincentionally drawing (Carlston, 2010). Indeed, formato in ommonly seen in g errors are Co Payne, Hall, Cameron, & of affect (Paynes oo y (Wegner, 2002). For 010) and agency (Wegs aeribu Bishara these reasons, affec be more difficult to mal Shad bal representations. sry systems model (MSM ke conscious, and may be tional and uncontrollable, somewhat than visual and vert Like AST, the memory systems moc Amodio, 2008; Amodio & Ratner, 2011 The MSM is critical of traditional associa ralise snist_ models, tive network, schema, and connectionist ‘which tend to assume that all representations are cut from the same cloth and operate in the same way Whereas such models might effectively capture ce tain kinds of semantic processes, they might fail t do justice to affective processes and behavioral out comes (Amodio & Ratner, 2011). To address these concerns, the MSM appeals co distinct memory systems with unique operating characteristics and neural underpinnings (Squire & Zola, 1996). The model distinguishes between a semantic associative system, an affective memory system, and a proce dural memory system. The affective memory system learns through mechanisms, including fear condi tioning, that are implemented by subcortical path ways in which the amygdala is central. The semantic memory system learns through more concepttal means that are implemented by neocortical regions And the vena : ce nemory system learns through @ network connecting th eeting the striatum and basal gan to prefrontal cortex ay : In some c: cognition m; the distinee nd Motor regions. ses, the kinds of tasks used in social wy be used to dissociate the effects of ‘ystems. For instance, Amodio and 1 REPRESENTATION ‘mirror neurons might PC orms of empty and =P goals, and mental sears ition models handle mn? Ke seems that this Imay be orthogonal to the modal vers format of representations. Representations different cmbodied modalities might be ‘or kcss conscious, intentional, controllable, ‘our discussion of multiple format TProdals, The multimodal representations used in “fimulations scem to be only temporarily implicit, jJnasmuch as conscious attention can shift focus to “information resulting from specific modalities. ‘And fess resides as much in embodied represen: ‘tations themselves as in how they arc used. However, there has been more debate over the “implicituseofembodied representations. Theembod- {ged simulations underwritten by mirror neurons are to be efficient, uncontrollable, and “unintentional, although their results are amenable to ‘conscious awareness with proper attentional direc~ ‘tion (Gallese, 2007). On the other hand, PSS theory “suggests that simulation using embodied informa. tion is an effortiul process, contingent on cognitive “capacity and motivation. A person can intentionally “draw upon embodied, modality-specific informa- for instance by trying 10 use his or her taste its history) to understand the ingredients ‘Under less ideal circumstances, people are ‘se “shallow conceptual” processing, or that do not re-enact modalit 2007). embodied cogti explicit distinction prime- (2010) exam influence aed squee fying, wheth were faster 0.96 were facing awa coGNin 3 ON AND MENTAL REPRESENT, rom side co side (an embod, od their heads up ang ‘validation). Only peg. dition showed priming 1 implicit aggression, moderate classic jor efeets- ee Cale “dy, Witt and colleagues motes howe embodied cues For (ool we recognition of tools. People vd while quickly identi Is or animals. They tools whose handles a ball in one han er objects were 00 ccurately identify ay from the hand grasping the ball appeat to have blocked 1 ball, people yr simulations By grasping, he automatic motor hat would oth- an ke have been activated when seeing images of handles faced the grasping hand. rools whose ition models can Finally, embodied social co help us better understand implicit : ‘or instance, embodiment perspectives attitude mea surement. F can illustrate how different implicit measures might rely on different types (i.c., modal ws. amodal mental representations. In one study, people were asked to hold a pen in their mouth—thus block ing the use of the zygomatic facial muscle—while they completed an evaluative priming task and an Implicit Association Test (IAT; Foroni 8 Semit 2011). Whereas this manipu on IAT performance, it disrupted affective priming presumably because it blocked the facial muscle movements that occur after affect-inducing primes and that would facilitate speeded responding (© affectively congruent targets. Findings such as these tion had no influence est that embodied social cognit ectives social cognition perspe might help to resolve long-standing debates about ced by. different the mental representations implicit measures, In summary, embodi on mod Satan dino odied social cognition about implicit and ex and how th drawback a in revising how we think plicit mental representations ey ate measured. Pehaps the bigse* this point is lack of th 1a " s lack of theoretical it Bravion across the fast-growing number of studies “ATION ico iio ations | bra she ‘3009: ¢ Alchoug embod theorist (Smith cal ass embod maint ment tion ¢ more that all. 4 have pute this rese cor int net

You might also like