8/12/2019 Wasserman 1993

1/18

Psychological Bulletin1993.Vol.113, No. 2 ,2 1 1 - 2 2 8Copyr ight 1993 by the Am erican Psychological Association, Inc .0033-2909/93/S3.00

ComparativeCognition:Beginning the SecondCenturyoftheStudyof Animal Intelligence

E.A. WassermanComparative psychologyh as undergone many changes since itsinception inV ictorian Englandsome 100 years ago.Gone are the amusinganecdotes of pet owners and amateur naturalists,replaced by the detailed observations of behavioral scientists made under carefully controlledconditions. Yet, man yof thepersistent problemsin thecom parat iveanalysisofintelligenceremain:Arethe cogn itiveprocessesof animalslikethose ofhumans?Can researchers construct aphytogenyof intelligence? W hatiscognition w ithou t language? This articlebriefly reviewsth ehistoryof thestudy of comparative cogn ition. It then discusses 2 of the m ost active and im por tant areas ofempirical inquirymemorya nd conceptual izat ion to acquaint readers with contemporaryre -search in the field.Given increased contact withthe related areaso fcognitivescience, behavioralneuroscience,an dbehavioral ecology, comparative cogn ition should continuein its 2ndcenturytomakesignificant con tributio nsto theoverall understandingof theprinciplesofbehavior.

How familiar asceneit is. Alearned an d eloquent scientistgazes up ward toward th e distant heavens, sighswistfully, an dplaintivelyasks Is there life beyond earth? Isthat life intelli-gent? If so, can w ecommunicate with it?What a noble quest: to search out alien life and to interactwith it. Aquest no t only noble bu t alsogenerously funded bygovernmental agencies. Radio telescopes, orbiting satellites,an d both staffed an d unstaffed cosmic expeditions have beenand w ill be constructed and undertaken to find extraterrestriallifeatgreat public expense an dwith much media fanfare.I wish not to demean these important efforts but rathertosuggest thatanequally challengingan dprobably more fruitfulsearch fo ralien intelligence can take placeright here onearth. Thousands of species of animals inhabit this planet,manyofthem exhibiting notable craft,flexibility, andingenu-ity in adapting to the numerous challenges of survival. Thecarefulan dscientificstudyof ourplanetmatesmayshed consid-erable light on the nature of intelligence, on the charactero fcognition w ithou t language, and on the v ery possibility of com-munication with extraterrestrial lifeif it iseverfound. Per-haps most important, comp aring the intelligence of many spe-cieso fanimals may helpusknow better what itmeans to behuman.Althoughthe study of anim al intelligence has been an ongo-ing concer n of scientists forsome100 yearsRom anes's(1883/1977)classicbook,AnimalIntelligence,waspublishedin 1883 m o s tpeople knowverylittle about it. The present p opu larityofmany nature programs on television plus the great publicitythat several research p rojects on anim al beh avior have recentlyreceivedsuggest that now might be an op portu ne time to out-

Ithank G. Burghardt, H. Davis, M.Ril l ing,D.Rumbaugh ,N .Spear,and L. VanH a m m efo rtheir editorialand technical helpinpreparingthis article.Correspondence concerning this art icle shouldbeaddressedto E. A.Wasserman,D epar tmentof Psychology,Un iversityofIowa, IowaCity,Iowa52242.

lineforbothspecialistsandnonspecialists th eessentialsof thescientific field k nown ascomparative cognition.In addition,recent developmentsinhum an cognitive psychology have againmade the study of anim al intelligence central to a traditionalgoalofpsychology:namely, to view complex hum an abilitiesas emergingfrom configurationsofelementary associative pro-cessesthat could be studied in simple organisms (Gluck &Bower,1988,p.227) .This reassessment stands in stark contrast to earlier dis-missals of animal behavior and cognition as irrelevant to theh u ma n species. By way of illustration, on e trio of textbookauthorsjustified their near-total disregard o f nonhumanintel-ligence thusly:

Wheneverh igher mentalprocessesar e involved,weheartily dis-agree that human and animal behavior are necessarily governedby the sam e principles. We regard the h um an as a specializedproductofevolution,as ananim al whose cognit ionisalso special-ized. This means that hu ma ns and animals may share some cogni-tive abilities, but it is not a foregon e conclusion that they do .(Lachman, Lachman, &Butterfield, 1979, p. 42)Although itwouldb e foolhardytoconcludeon the basiso fexisting evidence that there are no substantial differences inhuman an d animal cognition, it is farbetter to examine th ematter with an open mind than from the prejudiced perspec-tiveofanthropocentrism.Thus, thepresentarticleexplores the

questionofanimal intelligencefrom, Ihope, anobjectivea ndcomparative vantage point.In pursuit of this overall objective, thepresent article firstsketches a bit of the interesting history of comparative cogni-tion and then introduces two of its most im portan t and activeresearchareas: memory and conceptualization. Additional dis-cussion concerns num erical competence and language behav-io rand theplaceofcomp arative cognition withinthebiologyofbehavior.Readers shouldbewarned that in areview suchasthis,it isimpossibleto do full justice to all areas of relevant research orto givefull credit to allpast an d present workersin the field.2 1 1

8/12/2019 Wasserman 1993

2/18

E. A. WASSERMANAlthough I make no claims to comprehensiveness an d thor-oughness, I dobelieve that th epresent article effectively intro-duces the field of comparative cognition to newcomers andusefully reviewsan ddiscussesk eytopicsand theoriesforindi-vidualswho already have some kn owledge of the field.Readers should be further warned that clear an d detaileddefinitions of many important termsindeed that ofintelli-genceitselfare no tgiven inth is article;nor are theygiven inmost otherso ncomparative cognition.It may be farsimplertoprovide preciseoperationald efinitionsofelementsoraspectsofintelligence than to givea good definitionof theoverarchingconceptfor animalsorpeople (for more on the definition ofh u ma n intelligence,seeSternberg, 1986; fo rmore on broaderconceptions of hum an and anim al intelligence, see Stenhouse,1973).

HistoricalPerspectiveEvolutionary Theory

In th e middle of the 19th century, in England, HerbertSpenceran dCharles Darwin proposed thatt hemindsaswellasthe bodies of animals had und ergone aprocessof organic evo-lution. Living animal species can thus be comparedbehav-iorally and anatomicallywith the resulting similarities anddifferencesp rovidin g important clues to the ancestry of behav-iorala nd anatomical traits.Sincethen, considerable progress has been m ade in elucidat-ing the structu ral evolution of anim als. Much less progress h asbeen recorded in und erstand ing the evolution of their intelli-gence. Beyondthe obvious problem that most behavior doesnot fossilize, man y actions that one characterizes as intelligentare not those that are common to all members of a species;instead, these intelligent actions are the often idiosyncratic re-sponses that individ ual anim als exhib ittospecificenvironmen-ta l problems or circumstances. This fact means that properdocumentation ofanima l intelligence requiresn otonlycarefulrecording of the relevant behavior but alsoprecise knowledgeand even control of the pertin ent envir onm ental variables. Thelatterpo int is of real historical significance, given the somewhatawkward start that Darwin and his student, George J. Ro-manes ,gave to the field of comparative cognition.1

Methodology and EvidenceBoth Darwin and Romanes had a clear agenda. They be-lieved that support for the theory of the evolution of intelli-gence could come by documenting a continuity of cognitionamonglivinganim al species. Such documentation was initiatedby collecting literally hundreds of tales of animal genius, asrelated by petowners, naturalists, an d zookeepers. Take, fo rinstance, this tale of feline guile by one of Romanes's (1883/1977) friends:

Ourservantshavebeen accustomedduringthelatefrostto throwth e crumbs remaining from th ebreakfast-table to thebirds,and Ihave several timesnoticedthat our catused to waittherein am-bush in theexpectation ofobtaining a hearty meal from one ortw oof theassembledbirds.Now,sofar,this c i r cumstanceinitselfisnot an example ofabstractreasoning. But tocontinue.For thelast few days this practice of feeding thebirdsha s been left off.

The cat , however , with an a lmost incredible amount of fore-though t , was observed by myself , together with two othermembersof thehousehold, toscattercrumbson thegrasswithth eobvious intention of enticing the birds,(p . 418)As might be surmised from this vignette, th e early anec-dotists were no t always impartial observers of behavior no r

werethey n ecessarilycareful recorderso feitherthebehaviorinquestion or the conditions that p romoted the behavior. As inter-esting and suggestive as these anecdotes were, they could notstand the tests of scientific scrutiny, as they wereof dubiousobjectivity and replicability. The anecdotal method simplywouldnot do to establish a science of comparative cogn ition, apoint made forcefully by C.Lloyd Morgan (1894/1896).A countrymanofSpencer, Darwin,an d Romanes, Morgan(1894/1896) is often credited w ith stimu latingthe scientificstudy ofanimal intelligenceby investigatingth e behaviorsofnewlyhatched chicks reared w ithou t their m other. Morgan rea-soned that withveryyoung animals,hecould more effectivelyl imit an d control th e individu al's prior experience; and withmaternal isolation,hecould removetheconfounding influenceofimitationon thebehaviorsh eobserved. Although Morgan'svarious research projects seem quaint an d amateurish by to-day's standards(theyw ere conductedin his ownpoultry yard),theyset thestagefor themorepowerfulan d refinedmethodsofPavlova ndThorndike .Asmost studentsofbehaviorareaware,I. P.Pavlov,inRussia,and E. L.Thorndike , in the United States, developed highlyreliable andobjective metho dsfo rinv estigating associative con-ditioning inanimals.It is afairconclusion that most progressinth e experimental investigationo f comparative cognition ha sbeen theconsequenceof thecreative application orm odifica-tion oftheir two basic methodologies. Later, I reviewa sam-pl ingo fth at em pirica l evidence.Interpretation

The Darwinian agenda not only sought to blur any sharpdivisionsbetween human an d animal intelligence,b ut also itinterpreted animal behavior in terms of human behavior andprivate experience. This anthropomorphic bias is implied inDarwin's(1871/1920)declaration of mental continuity betweenhumans andanimals:The difference in mind between man and the higher animals ,great as it is, certainly is one ofdegreeand not of k ind . We haveseenthatth esensesan d intuitions,th evarious emotionsa ndfacul-ties,such aslove, memory,attention,curiosity, imitation, reason,etc., ofwhich m an boasts,may befound in anincipient , orevensometimes in a well-developed condition, in the lower animals.(p .128)

The question of mental continuity was also of concern toDarwin's student, Romanes. He hypothesized that there might'Most historians have been very harsh in their criticisms of Ro-

manes'scontributions to comparative psychology. However, m ore re -centdiscussionsof hisideashavecastth eworko fRomanesin amuchmore favorable light (see Rilling, 1992,for adiscussion ofRom anes'suncelebrated experimentat ion on counting in chimpanzees, and seeWasserman,1984,for amore generalconsiderationo fRom anes'sinflu-ence on comparative psychology).

8/12/2019 Wasserman 1993

3/18

C OMP AR ATIVE C O G N I TI O N A N D A N I M A L I N T EL L I GEN CE 213be acognitive scale rough ly paralleling the phylogenetic scale.Placement along this scale could be determined by objectivebehavioral tests: C an animals learn to behave inaccord withspecific situational demands deemed necessary to evidencesome cognitive capab ility,suchasmemory,associationbycon-tiguity,reason, or commu nication?I fso, then they mu st possessthat mental capacity.Romanes (1884/1969) constructed anelaborate chart tracingthe rep resentation ofnumero us cognitiveprocessesthroughoutth eanimal kingdom .As oneprogresses from more primitivetomore advanced organisms, more advanced form s of cognitionareassumedto beaddedtoless advancedones.2Thus, althoughhumans may sit at the summit of intelligence, humans mayshare with animalsanumbero fcognitive capabilities, that n um -berdecreasing asevolutionary kinshipbecomesmore remote.Romanes (1884/1969) also made an importan t effort to dis-tinguishthe objective and subjective aspects of animal behav-ior. As the following passage illustrates, Romanes saw thatovert action an d subjective experience cannot be equated.Whereas one m ight judge that the performanc e of similar re-sponses by hum ans and an imals implies the operation of com-m on cognitive oremotionalprocesses, an ysubjective experi-ences that may accompany those behaviors are inherently pri-vate; the ir existence can only be hypothesized by analogy withhuman experience.

Now,b ymindwe maymeantwo verydifferent things, accordingas we contemplate it in our own individual selves, or in otherorganisms. For ifwecontemplate ourown mind,wehave animme-diate cognizance of a cer tain flow of though ts orfeelings,whichare the m ost u ltimate things, and indeed the only things, of whichwe arecognizan t .But if wecontemplate m indinother organisms,we havenosuchim mediate cognizanceofthoughtsorfeelings.Insuch cases we can only i n f e r the existence and the nature ofthoughts and feelings from th e activitiesof the organismswhichappear to exhibitthem . (Romanes, 1883/1977,p . 1 )

From its earliest years, comparative psychologists havefounditextremely tan taliz ing to speculate on the nature of consciousexperience inanimals (fora review,see Burghardt, 1985). N oless than Morgan (1894/1896) championed the use of intro-spection in the interpretation of animal behavior:Th ewisean dcautiousstudentnever forgetsthattheinterpretationof thefactsin psych ical terms is based up on the ind uctions he hasreached th roug h introspection.Th e facts ar eobjective phenom -ena;theinterpretationis inte rmsofsubjective experience,(p. 47)

Even today,we find Griffin (1976,1978) insisting that subjec-tiveanim al exper iencefalls within the proper province of thefield of comparative cognition (what he calls cognitive ethol-ogy see upcom ing discussion).However tempting it may be to project onto animals one'sow nthoughtsandfeelings,mo st workersin the field ofcompar-ative psychology have deliberately resisted that temptation.One of the earliest behavioristic students of animal behavior,H. S.Jenn ings (1904/1976), recognized the unverifiability ofconsciousnessas anaccompaniment to action:It isclear th at objective evidence cannot give a demonstrationeither of the existence or of the non-existence of consciousness,forco nsciousness is precisely that w hich cannot be p erceived ob-

jectively. N o statement concerning consciousness in animals isopentorefutationb yobservationand experiment,(pp.335-336)His more celebrated contemporary, J. B. Watson (1913),evenmore vigorously asserted that Ihave virtu ally denied that th isrealm of psychics is open to experim ental investigation (p.175). Inaddition, the late B. F.Skinner (1977)commented onth eregressiveandunproductive character ofmentalistic analy-sis,accordingtowhich

a science ofanimalbehavior m ust be replaced or supplemented bya science ofan imalfeelings.It would be as extensive as the scienceof behaviorbecause therewould presumably be afeelingforeachact. Butfeelingsareharder toidentifyand describe than the behav-io rattributed to the m, and we should have abandoned an objec-tive matter in favor of one of dubious status, accessible onlythrough necessarily defectivechannelsofintrospection,(p. 3)Th epresent stateofresearchan d theoryincomparative cog-nition is stronglyrooted in the experimental investigation ofpreciselycontrolledandrecorded animalbehavior. Interpreta-tions u sually centeron themechanismsan dprocessesofcogni-

tion rather thanon thenatureo rcontentsofsubjective experi-ence. Inaddition,considerableattentionis paid to the biologi-cal substrates of cognition, w ith due regard for the factthata nunderstanding of the biology of intelligent behavior isverymucha long-term goal.Considered next are two key areas of research in com parativecognitionmem ory and conceptualizationsothat one maygainaclearer appreciationof theprogress thathasbeen m adeinevaluating thethesisof mental continuity between humansan d animals. It ismad e evident that although some of the re-search istruly comparative, other work seeks primarily to ex-pand the known cognitive competencies of animals. Learningthe lim its of ani ma l cognition is one of the m ost active avenuesofcontemp orary investigation (Wasserman,1981). Fullydocu-mentingthe k now n competencies of anim al cognition allowsresearchers tobetter ascertain whether similar behavioral re -sults in hum ans m ay be due to the operation of the same biologi-cal processes or mechanisms.Animal Memory

The most fundamental principle of mental operation is thatof memory, for this is the conditio sine qua non of all mentallife (Romanes, 1884/1969, p. 35). These lines by Romanesexpress in the strongest terms the centrality of memory to anunderstandingofcognition inbehavior. Evena cursory surveyof the most prestigious periodical in human cognitiontheJournal of Experimental Psychology: Learning, Memory, an dCognitionwillreveal the continued prom inence of the studyof memoryinh umans s inceth eearly workofEbbinghaus. Yet,despite Romanes's earlypronouncement and the vast body ofresearch on h um an mem ory, the systematic study of memoryha sno t historically been aprimefocus ofworkers investigatingthe behavior of animals under controlled laboratory condi-tions.

2Incidentally, Romanes also assumed that the same accretion ofcognitivecap abilities occurs as a hum an developsfrom bir th to adult-hood.

8/12/2019 Wasserman 1993

4/18

2 1 4 E. A . W A S S E R M A NSeveralfactors probably contributed to thelackofexper imen-ta l attention paid toanimal memory. First, initial interest cen-tered on associativelearning as aresultof thepioneering effortsof Pavlov and T hornd ike. To Watson(1913)and the other earlybehaviorists, little else seemed to be necessary to understandth e intelligenceoforganisms than their learningo f stimulus-

response associations. Second, extreme conservatism governedthe analytical efforts of the early behaviorists. Morgan (1894/1896)had persuasively argued ag ainst the interpretive excessesof attributing gratuitous cognitive processes to animals whensimpler processes seemed to do, and most later researchersabided by his fam ous canon of parsimony. Finally,whereasth ebasic procedures ofPavlovandThorndike servedth eearlyinvestigators of anima l lear ning exceedingly well, the more elab-orate techniques necessary fo r studying memory were no tnearly as well developed. W. S.Hunter (1913) did devise thedelayed-response paradigm, in which, after adelay,an animalmightidentify whic h of several potential foodsites it earlie r hadseen baited. Yet, successful perform ance in this task m ight notbe the result ofsome en dur ing cognitiveor neural process; itmighta swellhave been due to the anim al merely maintain ingits bodilyorientation to thebaitedsiteduring thedelay period(see Fletcher, 1965, for further discussionofsuch behavioralmediation ).Th epast2decades have seenaveritab le explosionofinterestint heexperim ental investigationofmemoryinfamiliar labora-toryanim als like rats, pigeons, and m onkeys. In part, this devel-opm ent is due to a realization tha t learning alone is incapableof explaining intelligent action; other cognitiveprocesses likememory, attention, an d conceptualization must also promoteadaptive behavior in complex and changing environments.Alsocritical to the rise in the study of anim al m emor y has beenthe emergence of new experimental techniques with sufficientpowerand reliabilitytodisclose clearlytheoperation ofmem-oryan d other cognitive processes.Let'sno wconsider th e contemporary studyo fanimal mem-oryparticularlyshort-term memorytogain a fuller under-s tandingof the methods, results,and theories with which mostworkers have been concerned (see Spear, Miller, & Jagielo,1990, for a broader review of the areas of animal memory andlearning).BasicMethods and Findings

Quite independently, J.Konorski(1959),in Poland, and D. S.Blough (1959), in the United States,reported tw o delayed dis-crimination methods thathavecometo bestandardtechniquesin today's investigation of animal memory.In one version of Konorski's(1959) procedure (Wasserman,1976), pigeons view two visual stimuli in succession on anickel-sized pecking key. Each member of a pair of stimulimightbe presented for 5 s, with a variable time interval betweenthem. In the delayed matching-to-sample paradigm, foodwould followthe second (or test) stimulus if it was the same asthe first (or sample) stimulus, but food would not follow thesecond stimulusif it wasdifferent from the first. Bymeasuringth e pigeon's rateofresponse to the test stimulusontrials withmatching or nonmatching sample an d test stimuli, one canassessthedegreetowhich memoryo f the firststimu lus controls

respondingto thesecond.It so happens that pigeons peck th eresponse key at hig hrateswhen food is likely but at low or zerorates when food isunlikely. With ashort(1-s)interval betweensampleandtest stimuli,th erateo fresponseto thetestst imuluson matching tr ials may exceed that to the test stimulusonnon ma tchi ng trials by a factor of 10 or more. H owever, as thesample-test delay is lengthened, the response rate differentialdeclines in an extremely orderly way. This inverse relation be -tween delay interval and differential test responding docu-ments the forgetting of sample inform ation.In B lough's (1959 ) procedure, two or more simulta neouslypresented test stimul ifollow the sample stimulus.3Only one ofthem is correct and leads to food if chosen; any other selectiondoes not lead to food. In the delayed matching-to-sample para-digm,one of the test stimuliis thesameas thesample and theremainder aredifferent from the sample. With a short intervalbetween th esamplean d test stimuli, pigeons showed astrongtendency to peck the correct (matching) test stimulus. How-ever, as the sample-test delay is lengthened, choice accuracydeclined, again docum enting the forgetting of sample informa-tion.Theessentialdifference between thesemethodsisthatwithKonorski's(1959)procedure the anim al mu st decide whether torespond to the test stim ulus , whereas with Blough's(1959)pro-cedure the animal must decide to which test st imulus to re-spond. This procedural difference notwithstanding, the twomethods produce highly similar results. Thus, no t only doeseachtechniqueyieldadecline inmemory as thesample-testinterval is lengthened u nder the delayed matching-to-sam pleparadigm, but also discriminative performance in each caseisdirectly related to thedurationof thesample stimulusand thetime between trial s (e.g., Nelson & Wasserm an,1978;Roberts &Grant, 1976).An imp ortant varia nt of each general method involves sam-ple and test st imul i drawn from different item pools. Thus,anim als might be showndifferentcolors as sample stimuli anddifferent forms astest stimuli. No physical matches are possi-ble; only arbitraryorsymbolic matches ca nhold.(In aparticu-lar version of Konorski's, 1959, procedure, then, food wouldonly follow red-squarea ndgreen -circle sequencesbut not red-circleandgreen-squa re sequences.)Thisso-called delayed sym-bolic match ing-to-samp le procedure ha safforded researchersspecial opportunities to expand th e investigation of an imalmemory.So, with true matching-to-sample procedures, it has beenshown that pigeonscanremember thecoloro f asample stimu-lus,itsshape,itsorientation,and itsspatiallocation(forareviewsee Spear et al.,1990).However, with the symbolic matc hing-to-sample procedure, it has also been possible to show that tem-poral aspects of ast imuluscan beremembered. Thus, pigeonsgiven Konorski's (1959 ) method remembered different dura-tionsof a redsample s t imulusan d reported that memory dur-ingteststimuli ofdiffering line orientations (Wasserman, D e-Long, & Larew, 1984). They rem embered not only two verydifferentstimulus durations(2 s and16s),bu talsoawhole range

3Actually, his is adelayedvers ion of Skinner 's(1950) matching -to-sampletechnique.

8/12/2019 Wasserman 1993

5/18

C OMP AR ATIVE C OG NI T I ON A N D A N I M A LI N T E L L I G E N C E 215ofmoreorlessdifferent dura t ions (1 ,2 ,3 ,4 ,5 ,6 ,7 ,and 8 s) , andsample-testdurationsaslongas16s.Beyond theattributesofsingle sample stim uli, p igeons givenembellished versionsof the symbolic matching-to-sample para-digm havealso been shown to remem ber the temporal order(e.g.,red-green) of aseriesof twodifferent color sample stimu li(Weisman, Wasserman, Dodd, & Larew, 1980), the spatialorder(e.g.,left-right)of aseriesof two identically colored sam-pl estimu li (Wasserman, Nelson,&Larew,1980),and the rela-tiveduration (e.g.,short-long) of aseriesof two different colorsample stimuli(Dreyfus, Fetterman, Smith, &Stubbs, 1 988).Furthermore, rats havesuccessfullybeen trained to press one oftwolevers, dependingon thenumberofpriorauditory stimuli(two or four) in a simplified version of Blough's (1959) proce-dure (Fernandes & Church, 1982; also see Davis &Albert ,1986).(Alater sec tiono fthis article considers relatedaspectsofnumericaldiscrimination.)Yetanotherway inwhichthe memoryofcomplex informa-tion hasbeen studiedha sinvolved sample s timuli comprisingtw oor more man ipulab le attributes or elements(Riley& Roit-blat,1978).Pigeons were thus shown two -element samples com-posed of color (red or green) and line orientation(horizontalorvertical) elements. Using Blough's (1959) method, tests withjust color comparisons or just line comparisons each yieldedhighlyaccurate performan ce, implying that both the color andtheline orientation informa tion ofthecompound sampleswerediscriminated and remembered by the pigeons. Significantly,accuracyo nthese com pound sample (colorand line) trialswa slowerthanonoth er trials involving on ly element samples (coloror line),suggesting thatthe twosourceso fsample informationon compound trials competed with on e another fo r what inhumans iscalled divided attention (for moreon this dividedattention notion se eBrown& Morrison, 1990,an d Riley&Brown,1991).Flexibility of Memory

Earlyin themodernera ofmem ory research,it wasproposedthat m emo ry might be explained by analogywitha simple elec-troniccomponentthecapacitor.In thesamewaythat electri-ca lcharge growsin thecapacitor when current isappliedto it,the strength of a mem ory trace m ightgrow,the longer the sam-plestimulusisobserved; and in the same waythat electricalcharge dissipates from thecapacitor when current is drawnfrom it, thestrengthof amem ory trace mightfade,thelongeritha s been since th e sample stimulus ha s been observed. Al-though the initial evidence supported this trace theory ofanimalm emo ry (Roberts&Grant,1976;alsoseeGuttenberger&W asserman, 1985), more recent research indicates that ani-mal memoryis farmore intricateand flexiblethan isimpliedby acapacitorlike trace.Cuingth edelayinterval. Suppose that,in thedelayed match-ing-to-sampleparadigm using either the Kon orski(1959)or theBlough (1959) procedure, subjectsare given additional infor-mationdur ingth esample stimulus concerningth elengthof theupcoming delayinterval (Wasserm an, Grosch,& Nevin,1982).Withtwodelay inte rvals (e.g.,2 s and 8 s), twostimuli(e.g.,atonean dwhitenoise)can beusedto cuethose delays. W hen perfor-mancehereiscompared w ithacontrol condition inwhichthe

added stimuli are completely uncorrelated with the upcomingdelay intervals, the correlated condition enhances memory atth e short interval but impairs memory at the long interval.Furtherevidencethatdelay interv al signalsaffect memory per-formancecomesfrom testsinwhichashort delayiscuedbut alongdelayisgivenand inwhichalon g delayiscuedbut ashortdelay is given. Cu ing a short delay but giving a long one im-provedmemory above that obtained at the correctly cued longdelay;cuingalong delaybu tgivingashortoneimpaired mem -or ybelow tha t obtained at the correctly cued short delay.Just wherein thesequenceofevents (sample stimulus, d elayinterval, test) the delaycue isexertingit seffect and how it isdoingso is not yetknown,althoughonepossibilityisthatdif-ferentialattention is paid to the short- and long-cued samples(for more on this issue, seeMacDonald & Grant, 1987;a ndWasserman etal., 1982). What isknownisthat d elay intervalcuingsimilarlyaffects hum an memory performance. Humans,too, show a steeper memoryfunctionwithbetter memory atshort delay intervalsbutpoorer memoryatlong delay intervals on a paired-associate immediate memory task when infor-mativetime tags are provided than when they are not (Hinrichs&Gru n ke ,1975).Cuing th e test stimuli. Imagine that on e wereto combinetrue and symbolic matching-to-sample paradigms withinBlough's(1959)general procedure. H ere, each of the two samp lecolors would requirechoiceof itssame color stimulus whenboth colors were shown as the test stimuli; each of the twosample colors w ould also require choice of a specific line orien-tation when both line orientations were shown as thetest stim-uli.Ifcolor testsan dline tests were equally likelyandunpredic-table, one might suspect that this task would pose no specialchallenge for thepigeon,as itsmemoryfor thecolorof thepriorsample stimulus should afford it the means of choosingcorrectly on each type oftest:color or line. Now, imagine thataccompanying eachsamplestimulus is one of twoformsthatare perfectly correlated withth e upcoming test dimension:Asuperimposed circle signalsa color test and a superimposedtriangle signalsaline o rientationtest.Here, there could be adecided advantage to the bird's discrim inating and u sing theformcues; advance wa rnin g of the test dimension might enableitto preparefo rthe up com ing test and to set itself to respond tojust one particular stimulus. Without correlated test dimensioncues, the animal would have to engage in more elaborate andpossibly moredifficult preparations and to set itself to respondto two particular stimuli depending on the nature of the testitems. Note that test dimension cuing wouldhave an advan-tageous effect only if there were an anticipatory or expecta-tionalaspectto animalmemory.Adecayortracetheorywouldleadone to predict that test dimension cuing would have noinfluence onmemory.To see ifcuingthetestdimensionhas anyeffect onmemory,Stonebraker an d Rilling (1984)gave pigeons initial trainingunder the previously mentioned co nditions and then occasion-ally miscued the test dim ension, giving a color test after th etriangle and a line test after the circle. Memory on miscuedtrials was much poorer than wasmemoryon correctly cuedtrials,therebysuggesting thatsubjects had come to anticipateparticular test stimuliafter particular prior cues.As indelayintervalcuing , researchers have much moretolearn abouth ow

8/12/2019 Wasserman 1993

6/18

2 1 6 E. A . W A S S E R M A Nprecuingaffects memory (see Grant &M acDonald, 1990, andSanti, Musgrave,&Bradford,1988, formoreo nthis issue). Yet,thesean dotherdataindicate that anticipatory orexpectationalprocesses ar e importantly involvedi nanimal mem ory (Honig& Dodd , 1986; Honig &Thompson, 1982;Wasserman, 1986)an d hum an memorya swell(Cohen,1989).

Directed forgetting. Within th e domain of research on hu-man memory, it is customary to posit the operation of controlprocesses:means bywhich memoriesa rem odulated in accor-dance w ith the needs of the individ ual or the dem ands of thetask(Atkinson&Shiffrin, 1968). By their very nature, controlprocesses comprise manyof thoseflexibilities and intricaciesof memory that falloutside thescope of trace theories.One such k ey control processis rehearsal. It is often said thatrehearsalis acovert activity that helpstosustain th ememoryofsome prior event. Whereas engaging rehearsal should aid inretainingearlier informa tion, termin ating rehearsal should im -pair retention. To investigate the role of rehearsal in humanmemory, workers havedevelopedth eso-called directed-forget-t ingparadigm (Bjork,1972). In one versionof the paradigm,shortly after presentation of the trial stimulus, subjects ar egiven one of two cues (e.g., red or green colors) either to re-member or to forget the prior stim ulus. Here, it is generallyfoundth at when mem ory testsa reg iven, retention is farworseon forget-cue trials tha n on remember-cuetrials, imp lying thatthe poststim ulus cues wereaffecting the rehearsal process andthereby modulat ing memory. It has further been found thatpostponing th e forgetcue in adelay intervalof fixeddurationleads to a loss in itseffectiveness; that is , memory fo r earlierinformation improvesth elater intoth edelay intervalthe forgetcue isgiven (Weiner& Reed, 1969). This result suggests thatspontaneous or uncued rehearsal before the forget cue protectsth e memory from th e decrementaleffect of the forget cue.

Beginningwith a study by Mak i and Hegvik(1980),severalworkers in the area of animal memory have endeavored toascertain whether directed forgettingisuniquely hum an.Re -search with both pigeons (Grant, 1984; Maki, 1981;Ril l ing ,Kendrick ,& Stonebraker, 1984) and monkeys (Roberts, Maz-manian ,&K raemer,1984)h asadapted th e delayed matching-to-sample paradigm to this end by adding brief postsamplecues tosignal that atest for sample memory either would orwouldnot be given. As in the caseof human memory, animalmem ory proved to be much lower on forget-cue trials than onremember-cue trials. In addition, m emory was more markedlyreduced ifthepostsamplecue waspresented early than if it waspresented latein thedelay interval (Grant,1981;Stonebraker&Rill ing, 1981).4Serialposition function. Further evidence inconsistent withtrace theory is well established in the area of human verbalmemory.There, it isfrequently found that memoryfor a seriesof items is a bowed functionof input position: memory beinghigh fo rinitial (primacy)a ndterminal (recency) itemsbut lowfor itemsin intermediate input positions (e.g., Crowder, 1976).Whatever else the bowed serial position function means fortheories of memoryprimacy perhaps reflecting long-termmemory and recency reflecting short-term memoryit cer-tainlyviolates the m onotonically increasing functionpredictedby trace theory.Isthisbowed serial position function demonstrable in non-

verbal animals? Yes,say Wright, Santiago, Sands, Kendrick,and Cook(1985).They trained both pigeons and m onkeys on aserial-probe-recognition task. Lists of color slides were pro-jected one at a t imeon the upper of two rectangular screens.Each of the four list itemsall different from one anotherwa sshown for 1 s(monkeys)or 2 s(pigeons) witha1-sintervalbetween items. A probe item was projected on the lower screensome time after th e fourth list item. If the probe item was arepeat of one of thelist items( same tr ial) , thenaresponse toth er ight-hand manipulandumwascorrect and wasreinforcedwith foodor dr ink ; if the probe item had not been presented inth eprior four-item list ( different trial), thenaresponseto theleft-hand man ipulandu m was correct and w as reinforced.When1 s to 2 s forpigeonsor 1 s to 10 s formonk eys separatedthe final list item and the probe test, memory fo rList Items1and 4 exceeded mem ory for List Items 2 and 3, thus repro duc-ing the classic bowed serial position memory function. Atshorter list-testdelays, the serial po sition function rose mono-tonically; at longerlist-testdelays, theserial position functionf e l l monotonically.N otonlydid th e pr imacyan d recency por-tions of the serial position curve greatly depend on how longafter the last list item the probe test was given, but also pre-cisely the same changes in memory performance heldfor hu-mans tested under similarcircumstances.5These and other results thus suggest that animal memory isfar more complex and flexible than was once thought and thatsimilar control processes may modulate both h um an and ani-mal memory. Such empirical parallels are, of course, all themore significant given that th e nonhuman an imals in all ofthese investigations were nonverbal creatures.

Conceptual BehaviorHumans an d otheranimals ar econstantly confronted withan extraordin arily complex arrayofexterna l stimu li. Yet, senseissomehow made of these varied and var yingstimuli.One wayof reducing the demands on an organism's sensory and infor-mation-processing systems is for it to treat similar stimuli asmembers of a single class; by so doing, substantial cognitiveeconomy can beachieved, thus freeing itsadaptive m achine ryto deal with other c omp eting exigencies of survival. In add ition,categorical processing permits an organism to identify novels t imuliasmembers ofaparticu lar classand togeneralize kno wl-edge about that categorytothesenewmemb ers. Thus,a norgan-ism need not bebound to respond toonly those stim uli withwhichit has had prior experience, correspondingly enhancingits ability to cope with a continually changing world.Althoughtheorists have often extolled these adaptive virtue sof categorization an d conceptualization, we remain fa r fromunderstanding exactly how organisms process stimuli so as to4Further discussion o f directed forgettingand alternative accountsof it can b efoundin K endrick andKilling's (1986)book, in which theyalsoconsider a broad rangeo fan imal memory phenomena.5Further evidence on the serial position effectcan be found in Bu-chanan, Gill, and Braggio (1981), in Reed, Chih-Ta, Aggleton, andRawlins(1991),and in Kesner and Jackson-Smith (1992).The last arti-cl ealsoconsiders th eneural mechanismsofhu manandanimal mem-or yprocesses.

8/12/2019 Wasserman 1993

7/18

C OMP AR ATIVE C O G N I T IO N A N D A N I M A L I N T EL L I G EN C E 217partition the world into classes of related objects an d events.Indeed, given early writings on the subject, we should wonderwhether nonhuman animals ar e even capable of conceptualbehavior. Nearly a century ago, Morgan (1894/1896) deniedanimalsthe ability tobehave conceptually.To do so, he said,requires that

weneglect allthatis variable and focus the attention on the u ni-formrelation. [Then]wehave reached aconception, an dthis con-ception is not concrete, particular, an d individual, bu t abstract,general,and ofuniversalapplication, (p.263)Morgan believed that only adult hum ans (not even children) arecapable of conceptualization. Several recent lines of evidenceare radically changing that initial opinion .Object Concepts

O ne familiar instance of conceptual behavior involves thekind of open-ended categorization response one mak es wh enone labels different natural (e.g., cat)and hum an-mad e (e.g.,chair) objects withdifferent nouns. Such verbal behaviors ar eoccasioned by specific instances of wide variab ility and individ -uality.In deed , accurate classification even extends to categori-ca lexem plars never seen before.Is it at all possible fornonhu-man animals lacking languagetoengageinthisformofconcep-tual o r classifactory behavior?To answer this questionwith regard to the familiar labora-tory pigeon, a new technique was devised to train it concur-rently to discriminate stimuli from several human languagecategories.Thespecificmethod (Bhatt, Wasserman, Reynolds,& Knauss, 1988, Experiment 1 ) wa s based on the techniqueparents often use to teach their children to label objects in apicture bookthe name game. Wh enth epageisturned, th echild is asked to look at the object and then he or she is re-quested tonameit.Ifheorshe iscorrect,praiseisthereward.Ifhe or she is incorrect, the result is encourag ement to try again.Finally, ifself-correction fails, he or she isprovided with thecorrect name.Toimplement this method with pigeons,acolorsnapshot was displayed on a 3-in. square frosted plastic screen,and thepigeonwasrequiredtopeckaclear plastickeycoveringthe screen 30 times. Com pleting this observing response re -quirement led to the il luminationof four different color keysjust beyond the corners of the viewingscreen. A single choiceresponse was then permitted. If it was to the correct key forreporting the stimulus on the viewing screen, all of the visualstimuliwere turnedoff and thepigeonw as fedmixed grain;ifthe response was to any of the three incorrect report keys,a llreport keylights were turned off and the trial wasrepeated.Onlythe first choice response of a trial w as scored; correctiontrialswere not considered in analyses of performance. In sev-eral studies,the 40slides seen ineach daily session depicted 10differentexamples each of cats, flowers, cars, and chairs. Thepictures containedone ormore instancesof thecritical stimu -lus object; the objects were indoors or outdoors; near or faraway; centered or offcenter; and in different colors, orienta-tions,an dbackgrounds.In on erepresentative experiment (Bhatte tal.,1988,Experi-ment IB), a group of four pigeons attained a mean level ofdiscriminativeperform ance of 76% correct durin g Days 26 to

30 oftraining,afterbeginningthe investigation nearthechancelevelof25 %correct. Als o noteworthy were the results of 2 laterdaysoftestperform ance w ith the 40 original trainin g slides andwith 40 brand-new slides of cats, flowers, cars, and chairs.Mean accuracy to old slides was81%,and to new slides it was64%. Althoughtestperformance washighlydiscriminative toboth setso fstimuli, accuracyw asreliably high erto oldthan tonew pictures, perhaps because the birds remembered someo rall of the old slides. Finally, there was no evidence for any of thestimulus categories being harder or easier for the pigeons todiscriminate (contraryto the suggestion made by Herrnstein,1985, that pigeons cannot categorize human-made stimuli).Thus, pigeonsar eable concurrentlytocategorize stim uli fromfourclasses of natu ral andartificialobjects and also to extrapo-latethatcategorization tocompletely novelteststimuli, albeitat a somewhat lowerlevelof accuracy.Perhaps even more important were the results of a subse-quent investigation (Bhatt etal., 1988,Exp eriment 3) in w hich alarge pool of2,000 unique snapshots (500 from each of fourcategories) were shownto pigeons on a one-time-onlybasis.Without the benefit of any stimulus repetition, the birds at -tained a me an accu racy level of 70% correct on Days 46 to 50.Eitherth epigeon has anundocumented abilitytorememberarather large num berofstimuliit hasseen only once(cf.Vaughan&Greene,1984,w hoshowed thatpigeonscan remember up to32 0pictures seena tleast2 8times each)or it can abstract somekind ofgenericorprototypicalinformation from varied stim-uli,as implied by Morgan's(1894/1896) earlier quotation andby several more recent modelsofconceptualization (Smith&Medin, 1981).It issurely no small matter to demonstrate that nonverbalanimals like pigeons are so adept at categorizing snapshots ofreal objects (also seeHerrn stein, 1985). Yet,one isbound towonder just how similar thisfeatis to the conceptual beh aviorofhumans. Here,tw oadditionalprojects suggestthatthesimi-laritymay be more than accidental.Ino ne investigation (Wasserman, K iedinger,&B hatt, 1988,Experiment 2), two group s offourpigeons w ere trained to cate-gorize the sam e set of 80 snapshots. The first (or tru e category)grouphad topeckone offourkeystoreport eachof 20stimulifrom fourhuman language categories: cats, flowers, cars, andchairs. The second (or pseudocategory) group had to classifyth everysame slides into rando m assortments, inw hich eachofthe four pseudocategories comprised equal numbers of cat,flower, car, and chair slides. Over Days 37 to 40 of training ,pigeons on the true categorization task averaged 79% correct,whereas pigeons on the pseudocategorization task averagedonly 44% correct (asmall bu t reliable rise from 25%). Thus,learning proceeded farfasterwh en the to-be-trained categoriescoincided with hu m an language classes tha nwhenthey did not.These and other results (Astley & Wasserm an,1992;Edw ards &Honig,1987;H errnstein& deVilliers,1980; Wasserm anet al.,1988, Exp eriment 1 )suggest that topigeons, members of hu-man language categories resemble one another more than the yresemble members ofother language categories. Th e pigeon'scategorization behavior thusconfirms the nonarbitrary natureof hum an language terms, at least for the object categories theyhavethusfarbeengiven.Inanother project (Bhatt,1988),three groups offour pigeons

8/12/2019 Wasserman 1993

8/18

218 E. A . WASSERMANwere trained to categorize photographic slides. The threegroupsof pigeons were given 48 daily trainin g trials comprising12copieso f 1examplefrom the categories cat,flower,car, andchair (Group1 );3 copies o f 4 examples from the same catego-ries (Group4); or 1copyof 12examplesfrom the same catego-ries (Group 12). The rate of learning to a criterion of 70%correct was an inversefunctionof the number of examples percategory.Of additional importance were the results of a general-ization test with 32 novel stimuli, 8from each category. Here,accuracy was adirectfunctionof thenumber ofexamples givendur ing training. The mean percentages ofcorrectchoices ongeneralization test trials were27% forGroup1 ,45%forGroup4, and 62 % forGroup 12 . Thus, although increasing the d i f f i -cultyof original learnin g, greater numbers of training examplesper category enhance the accuracy of generalization perfor-mance, perhaps because of the increased likelihood that anygiven test stimuluswill resemble one or more of the remem-bered train ing stim uli or because abstracted prototypes ar emore representative of the entire class the greater the n umb er ofexemplarsseen (Smith & Med in,1981).Not only are these dataorderly, bu t they neatly correspond with a large body of re-search on categorization in humans (reviewed by Homa ,Burrel,& Field,1987)and w ith a recent report on two-categorydiscrimination inpigeons with5 vs. 35examplesof black-and-whitebird and mam mal sketches per category (Cook, W right,& Kendrick,1990). Whether such interesting correspondencesinconceptualization byhum a nsan d animals willcontinuet obe found issurelyto b eexplored in future research.Abstract Concepts

One of the empiricalhallmarkso fconceptualization is theindependence of discriminative responding on thespecificde -tails of the p revailin g stimu li. Thus, it was imperative in w orkon object concepts to show that discriminative respondinges -tablished to a set of training stimuli also extended to a set ofuntrained test stimuli. To have conceptualized chairs requiresthat new chairs occasion the same response as old ones. Aneven more abstract levelofconcep tualization may beachievedwhenorg anisms attach the terms sameor d i f f e r e n t to apairo fsimultaneouslyor successively presented stimuli. (With succes-sivelypresented stimu li, the termsnovelorfamiliarwould simi-larly suggestanabstractlevelofconc eptualization.) Hereagain,the critical test comes w hen new pairs ofstimuliare given to seewhetherthe organism appropriately labels untrained stimuli.Matching to sample. One setting in which acquisition of asame-different concept has been studied is the matching-to-sample paradigm discussed earlier. The success of pigeons intheiroriginal mastery of this task m ight suggest that m atchingto sample would readily transfer to n ew stimuli. However, sucha result has not b een obtained (for reviews and critical analysessee DAmato, Salmon, Loukas, & Tomie, 1986; Edwards, Miller,& Zen tall, 1985; D. Premack, 1978). Comp ared with pigeons,bothnew-andold-world monkeys morereadily generalizetheirvisual matching-to-sample performance to novel stimuli (e.g.,DAmato&Salmon, 1984),asdoesthebottlenose dolphinin anauditory matching-to-sample task (Herman & Gordon, 1974).However, the clearest evidence of spontaneous transfer ofmatching-to-sample performance comes from chimpanzees.

Oden, Thompson, an d Prem ack (1988) taught four infantchim panze es to match to sample using a set ofonlytwo objects,a lock and a cup. In the simultaneous matching-to-sample pro-cedure the y used the chimp anzee w as handed a sample ob ject;itwas then required to choose from the set of two test objectsthe one that was the same as the sample object on that trial. Acorrect choice resulted in social and gustatory reinforcement,whereas an incorrect choice did not. After reaching a criterionof 83% correct, the anim als were given a series of tests withnovel objectsan dfabrics. Test accuracy onthese trialsaveraged85%. Thus, th e chimpanzees transferred their matching-to-sample perfo rman ce withou t decrement to brand-new stimu li,suggestingthattheyhad strongly conceptualized the sa me-dif-ferentrelation.It is surely noteworthy that dolphins an d primates morereadily generalize their matching-to-sample behavior than dopigeons. However, does this mean that pigeons are completelyunable to appreciate the abstract relation of sameness-differ-ence?O ther evidence suggests not.Pairedcom parison. Yetanother procedure ha sbeen usedt oassesscontrol over behavior bysameand different stimuli. Inthe paired-comparison procedure, two stimuli are eithersimul-taneouslyor successively exposed. Th en, two response alterna-tivesar eafforded tosubjects:one forreporting that th e stimuliwere thesameand theotherforreporting that they were d i f f e r -ent. Correct choices occasion reinf orcem ent, whereas incorrectchoices do not.In a project bySantiago an d Wright (1984), pigeo ns weretrained on a simultan eous visual paired-comparison proce-dure. Durin g original tra ining,105color slidesoffruit,flowers,animals, people, and other natural and human-made objectswere shown inpairson asplit screen.After makingan observ-ing response to a clear panel covering the split screen, twochoice keys were lighted: the left fo r reporting that the twoslidesweredifferent and the rightf orreporting that th ey werethe same. Half of the trials involved same stimuli and half in-volved different stimuli .T he large numbero f training stimulithat were used guaranteed that no stimu lus occurred on morethan one trialin a session, in the hope that the birds' behaviorwould more likely come under control of thesam e-differentrelation than und er controlof the specific featuresof thestim-uli. After tra iningto over 80 %correct on the original set ofslides,thepigeonswere shown 105brand-new slides.First-ses-sion transfer perfo rma nce averaged7 0%correct. This scorewa sa b it lower than transfer performance to the train ing slides; but,itwa smuch high er thanthe 50%score expectedbychance,an dit compared quite favorably withthe 72% first-sessiontransferscore of rhesus monk eys trained and tested under virtuallyidentical circumstances (Wright, Santiago, &Sands, 1984).Familiarity-novelty discrimination. Further evidence thatevenpigeonsc an appreciate abstract stim ulu s relations comesfrom a recent report by Macphail an d Reilly (1989). In thisproject, pigeons were shown a series of color slides depictingindoor scenes, outdoorscenes, objects, faces,and so on. Eachslidewasshown twiceineach da ily 48-trial session,an dslideswere never reused from one session to another. Pecks to thefirst presentation of a given slide were reinforced with food,whereas pecks to the second presentation were not.Afteronlyfoursessionso fdiscrimin ation training, pigeons pecked much

8/12/2019 Wasserman 1993

9/18

C OMP AR ATIVE C O G N I T IO N A N D A N I M A L I N T EL L IG E N C E 219moreoftenon the firstpresentation of aslide thanon itssecondpresentation. Because of the continually changin g compositionof the slide arrays, these results suggest that the pigeons werereadily able to discriminate familiar from novelstimuli,quiteapart from the specific attributes of each stimulus display. Mac-phail an d Reilly proposed that earlier difficulties in traininghighly general same-different or familiar-novel reports in pi-geons may have been due to procedural factors rather than toany cognitive limitations of the species.Oddity discrimination. Whereas the familiarity-nov elty dis-crimination of Macphail and Reilly(1989)necessitates succes-sively presented stimuli, a related problem, also requiring ab-stract conceptualization, doesnot.In anoddityconcept, choiceismade of the odd stim ulus in an array of three or m ore simulta-neouslypresented novel stimu li. Lom bardi, Fachinelli,and De-lius(1984)first established an oddity discr imina tion in pigeonsgivenproblems arranged froma pool of either 5 or 20 white-on-blackvisual forms. After reaching nearly equivalent levelso fperformance(the group givenfewerstimuli learning faster thanthegro up given more stimu li;cf .Bha tt, 1988),al lanimals weretested with brand-new forms. Highly reliable transferwas ob-served,thattransfe r beingbetterfor pigeonsgiven more train-ingstimuli than those givenfewer training stimuli (cf. Bhatt,1988).Interestingly,ananalogous project w ith rats given visualand olfactory oddity tasksfailedto find evidence of an oddityconcept (Thomas&Noble, 1988).Equivalence Class Concepts

The quest for clear evidence of conceptual behavio r has alsobrought about the development of new analytical ideas andexperimentalprocedures.Fo rinstance,theobject concepts dis-cussed earlier could be and probably w ere based solely on physi-ca lsimilarity.Despite our present inability to isolate and manip -ulate the relevant physical featuresofsuchcomplex stimuli assnapshotsofcats,flowers,cars,andchairs, theorists caneasilyappeal to the w ell-documented principle of prima ry stimulusgeneralization to account fo rtransfer tonovel instances fromthe training categories.Yet, other theorists (e.g., Lea, 1984) have insisted that trueconceptual behavior must be based on something more thanmere physical resemblance. Learning a response to somemembersof aheterogeneousset ofstimuli should ideally prop-agate to all membersof the set, without regard tosimilarity(Herrnstein, 1990,p.150).This requiremen t reducesto thedem-onstration of transfer of control through secondary stimulusgeneralization(H ull, 1943), and sets of such physically dissim i-la ry etfunctionallysu bstitutable stimuli are calledequivalenceclasses (see Sidman, 1986, for one specific rendering of thisgeneralidea).Initial research bySidman etal. (1982) using matching-to-sample procedures suggested that although children couldformeq uivalence classes, neither rhesus monk eys norbaboonscoulddo so. More recent research usingdifferent procedures ismore encouraging to the idea that animals other th an hu man sca n formequivalence classes (Vaughan,1988;Wasserman,De-Volder, & Coppage, 1992; Zentall, Steirn, Sherburne, &Urcuioli, 1991). In Vaughan 's (1988) experim ent, fo rexample,pigeonswere reinforced withfood forpeckingat a set of 20 out

of 40 slides, all 40 of which depicted trees. Pigeons were shownthe 40 slides indifferent random orders each session. The slideswere divided into two arbitra ry assortments: one positive (rein-forcement in theirpresence, 1+) and the second negative (noreinforcement intheir presence, 2 ) .Different birds w ere givendifferent 1+, 2-slide assortments toreducethepossibility thatsome unknown physical feature of the slides was associatedwith category membership. After the pigeons learned to dis-criminateth eslides (respondingto 1+ but not to 2-stimuli),thecontingencies w ere reversed 1 , 2+) and then reversed again,repeatedly. After dozens of reversals, the pigeons were able todiscrim inate the positivefromthe negative collectionsafterpre-sentation of only the first fewstimuli in each. Evidently, thebirds had form ed a com mo n stimu lus class for each of the twoslide collections;bysamplingafewslides,thepig eons could tellwhichequivalence classhad apositiveor anegative valenceinan yparticular session.In the familiar terminologyo fRosch and Mervis (1975),itmightbesaid that V aughan's(1988)birdshadlearnedasubordi-nate concept, in which pa rticula r instances of a basic-level con-cept, liketrees, were segregated insubgroups. In higherlevelgroupings, isthereanyevidence that animalscanlearn tocol-lect stimu lifromdifferentbasic-level concepts into sup erord in-ate concepts? Yes. Wasserman et al. (1992) used a three-stepprocedure with pigeons, in which collections of perceptuallydissimilar stimuli, like chairs an dcars, werefirstassociatedwithacommo n response. Then,a newresponsewaslearnedtojust one of those classes ofstimuli,like chairs. Finally, subjectsweretested for thei r tendency to m ake the new response to theclass of stim uli (cars) notgivendur ing the second step. Pigeonsshowed a strong propensity to m ake the new response to stim-uliwithwhichthat responsew asnever before associated. Thus,merelyby being associated with a com mon response in the firststep, classes of perceptually dissimilar stimuli appear to amal-gamate intoa newsuper ordinate categoryoffunctionally equiv-alentstimuli.Althoughnot the end of the story, itdoesappear that diverseconcepts are learnable by common laboratory animals, likepigeons, with different experimental methods being more orlessconducive to disclosing those conceptual a bilities.

Other Active Research AreasIn the previous discussions of animal m emory and concep-tual behavior, I tried to providereadersw ith a broad overviewof two of the most active and systematic areas of research in

comparative cognition.O fcourse,thereis no way todiscussthef u l l rangeo f research in the field in one review article. Inter-ested read ers sho uld consu lt several edited volu mes (e.g., Hulse,Fowler, & Honig, 1978;Kendrick , Rill ing, &Denny, 1986;Roitblat, Bever, & Terrace, 1984; Weiskrantz, 1985) and text-books (Pearce, 1987; Roitblat, 1987)form ore comprehensivecoverage of comparative co gnition.Beforeleavingspecificresearch dom ains with in comparativecognition, h owever, two additional topics are considered, w ithaspecial eyetoward matters ofexperim ental strategy an devi-dential interpretation. These two areasnumerical compe-tence an d language behaviorhave fostered rather more de -

8/12/2019 Wasserman 1993

10/18

220 E. A . W A S S E R M A Nbate,butless systematicdata,than th eareaso fan ima l memoryand conceptual behavior.Numerical Competence

Does number control the behavior of animals? If so, thendoes such discriminative control resemble counting an d othermathem atical behaviorsinhuma ns? Recent research suggestsaclear yes to the first question and a guarded perhaps to thesecond (fora review of the literature an d additional criticalcomm entary, see Davis & Perusse, 1988).Asisoftenthecaseincomparative research, matterso fdefini-tion can be critical to deciding if the behaviors of differentorganisms ar e alike. So, the seemingly simple question D oanim als count? requires opera tional definitio n of counting.Unfortunately, considerable debate e xists conc erning th at verydefinition (see, for example, D avis & Perusse, 19 88, and Gallis-tel,1989, fordifferent viewpoints). Rather than tryin g to solvethisdefinitionaldisputedirectly,someinvestigatorshaveunder-taken behavioral analysestoidentify more basic cog nitive skill sthatappearto benecessaryfor anyorganism toshow countingor higher mathematical abilities. Washburn and Rumbaugh(1991)thu s sought evidenceon themoredefinitequestion Cananima ls learn that differentsymbols areassociated with differ-ent quantities of food? An affirmative answer to this morel imitedquery wo uld indicate that the an imal under study has aprerequisite skilltoengage in more complicated mathematicalperformances, l ikecount ingan dadding.6In one set of exp erim ents (Exp erim ents 1, 2, and 3), Wash-b u r n an d Rumbaugh (1991) gave tw o rhesus monkeys th echoice of responding to two arabic numerals; whichever nu -meral was chosen determ ined the nu mb er of food pellets given.Fo rexample,in a 2-9pairing, choosingthe 2resulted inreceiv-ing 2 pellets, whereas choosing the 9 resulted in receiving 9pellets. Th e monkeys showed a clear tendency to choose th elarger numeral, that tendency increasing as the difference be-tween th e numerals increased. Also, one of the monkeysshowed clear generalization topairsofnumerals thatha dneverbefore been presented during training. In a finalexperiment(Experiment 4), Washburn and R umbaugh further examinedthe monkeys'symboliccapabilitiesby presenting fivedifferentnumerals on each trial. Each selection delivered th e appro-pria te num beroffoodpelletsandeliminated that numeral fromth e next choice opportunity. Here, both animals tended tochoose the largest numeral available, with their performancevaryingsomewhat as afunction ofboth th e numbero f numer-als available and thedifferencesamong the numera ls. As a sam-ple of performance, one monkey first chose the 8, then the 6,then the 5, and finally the 4 when given successive choices of8-6-5-4-1,6-5-4-1,5-4-1, and 4-1,respectively. These results ledWashburn an d Rumbaugh to conclude that rhesus monk eyslearned. . .that symbols (i.e.,arabic numerals 0-9) wereasso-ciated with different quantities of pellets and ordinally se-quenced the symbols that indexed those quantities (p .192).A fur therclaim by the authors imp lies that this ability maybelimitedtoprimates: These rhesus monkeys displayeda pro-ficiency at discriminating, representing, an d ordering quanti-ties beyond thaty etdemonstrated withanynonhumanspecies(Washburn & Rumbaugh, 1991 p. 192). However, Bhatt and

Wasserman(1987)had earlierfoundthat pigeons could learn toassociate fourdifferentcolor keys withfourdifferent quantitiesof food reinforcementa ndthat theirchoiceo f areference sched-uleo ffood reinforcementwas aclear functio nof thealternativequant i tyof r einforcement the y were given, choice of the refer-ence qua ntity declining as the alternative quantity was in-creased. In addition, Hulseand O'Leary (1982)had even earlierfound that rats given access to a four-arm radial maze withdifferent quantitiesof food at the endso feach ar m learned toorder th eir with in-trial selections in accord with those q uanti-ties. After 5 weeks of training, at least 75% of therats'choiceswereordered18-6-1-0(0 = nochoiceto the 0-pellet arm).A wide range of animal species can thus associate differentarbi t rary stimuli (e.g., numerals, colors, or spatial locations)with different quantities of food. Furthermore, their choiceamong those arbitrary stimu li clearly preserves th eordinal rela-tions those st imul i have with different food quantities. Now,researchersc an logically pursue questions concerned withn u-mericaloperations,suchascounting an d adding, which mightbe performed on those symbols (see Boysen & Capaldi, 1992,for much more workon this question).LanguageBehavior

The popular an d scientific attention paid to recent researchin language behavior in animals is unprecedented. Many of theapes, dolphins,andavia ns that have been studied inthis workare as well or better k now n by name th an their h um an investi-gators. In addition, th e rivalry among those investigators ha soccasionally sparked rancorous debates, which some wouldclaim havedone rather little toadvance th e understanding oflanguage behavior in nonhuman animals.As most readers are already aware, a focal question of thisgeneral l ineofworkis Islanguageauniquely hum an phenome-non? Some of the most excitinga nd controversial researchaddressing this question involves teaching animals various hu-man-made comm unication systems, includ ing vocal (Pepper-berg, 1981), gestural (Gardner & Gardner, 1984; Herman,Mo rrel-Samuels, & Pack, 19 90), token-based (A. J. Premac k &Premack, 1972), and computer-based (Rumb augh, 1977)schemes.N oeffort ismade hereto ruleonwhetherth ebehav-ioral feats heretofore performed by the many animals in thesediverse projects proves them to becapableo fhumanl ike lan-guage. Rather,as in thecaseofnumerical competence, theques-tion is whether animalspossess any of the component skillsnecessary for language behavior. This more limited approachyields some clearan d unequivocal conclusions.

Central to languageis the idea of reference: To the extentthat th ecom municat ion of information depends on the arbi-t rary pai r ing of terms with conceptual categories, then thatbiological function of a natural language depends on the rotelearning of paired associates (Gard ner & Gardner, 1984, p .401). Thus, in Project Washoe and in follow-up studies, theGardners successfully trained chimpanzees to make one of6Sucha componential analysisofcountingcan beseentoaccordwiththe general approach of Gluck and Bower (1988) to complex cog-nitiveperformance mentionedin the introduction.

8/12/2019 Wasserman 1993

11/18

C O M P A R A T I V E C O G N I T IO N A N D A N I M A L I N T EL L I G EN C E 221many manual gestures in American sign language to refer tomembers of such human language categories ascats,flowers,balls,and shoes.

The ch imp anzees transferred the signs they had learned for a fewballs, shoes, flowers, or cats to the f u l l range of the categorieswhenever they found them and howeverrepresented,as if theydivided th eworld into conceptual categories justa shumansd o.(Gardner &Gardner, 1984,p .400)Thus, thechimpanzeesh adpassed thed efinition al criteriono freferencewh erein the anima l has a lexicon of arbitrary signalsthat sy mb olically stand for objects, events, concepts, and fea-tures (Roitblat, 1987, p. 278).Interestingly,the Gardn ers (Gardner & Gardner, 1984)enti-tled their article describingthiswork A Vocabulary Test forChimp anzees. Note that Bhatt etal . (1988)found essentiallythesame resu lt with pigeons, albeit w ith onlyfourmotor behav-iors being used for their reporting four categories ofobjects.Although use of the word vocabularyf or either pigeons orchimpanzees may be controversial, the findingsfromeach spe-cies clearly demonstrate the operation ofreferenceinnonhu-m anbehavior,atleastforsome kindsofconceptualcategories.7Alsocriticalto languageis the notion ofgrammar: a finitelist of rules that can be used to produce an infinitely largenumb er of expressions (Roitblat, 1987, p. 278). Allied to thenotion of grammar is that of symbol sequence or syntax: whereint hesame symbolsindifferent orders. . . canexpressdifferent meanings (Roitblat, 1987,p.278). Manyof the lan-guage-learning projectsinan ima ls have been criticized fo rfail-ing to show grammatical competence. For instance, Terrace,Petitto, Sanders,andBever(1979)concluded tha t although apescan learn ma ny isolated symbols, these projects yielded noevidence of anape'sabilityto use agrammar (p .891). Lawfulregularities in symbol order were observed in those projects;however, such ordered res pon ding can, in each case, be ex-plained by reference to simpler non lingu istic processes (p.900). It is no small matter to show that even pigeons can betaughtto p roduce an ordered string of up to fivedifferent re -sponses in the absence of external d iscriminative feedback(Terrace,1991).Such sensitivitytoresponse orderisobviouslyanecessaryskillforgramm atical performance. However,itdoesno t easily allowfor the production ofnovel grammatical se-quences entailingthesameor new responses.The issue of symbol order has been also investigated in ani-mal behavior with regard to the question ofdifferent ordersconveyingdifferent meanings. Here,it hasbeenfoundthat dif-ferent sequences ofcolors ca n effectively signal different con-tingencies of reinforcement to pigeons (e.g., Weisman et al.,1980, Experim ent 3). Such early results suggest that there isgood reasontobelieve thatatleastthecom prehen sive aspecto fsyntax is operative in animal behavior. No clear separation ofthe comprehensiveand productive aspects of syntaxhas yetbeen achieved in monkeys (Devine, Burke,& Rohack, 1979)and pige ons (Parker,1984),each havin g succeeded in reproduc-ingtwo-item sequences previously shown tothem.8Obviously,there are m any component cognitive skills that areconcatenated in hum an language. Wh ich individu al and com-bined skills appear in the behaviorofanimals mayprovideimpo rtant clues in evaluating the uniq ueness of human lan-

guage. R eal insights may still resultfrom directefforts to teachhuman communication systems to animals, or they mayemerge from more oblique inq uiries into the limitsofanimalcognition. In either event, comparative cognition should con-tinuetocontributeto researchers' understanding th ebiologicalorigins of this most notable formo f huma n behavior. The re-cent remarksofGisinerand Schusterman(1992)providea fit-ting finale for this discussion of language behavior: It seemshighly implausible that the linguistic abilities of humans havearisen incomplete ontogenetican dphylogenetic isolation fromnonlinguistic learning abilities (p .90).

OverviewAfter a long,fallow period , researchers are again explor ingthe cognitive processes ofanim als with renewed vigor. Muchha s been learned from th e experimen tal investigationo fani-mal be havio r since Spencer and D arw in first ignited interest inanimal intelligenceand itsrelationtohum an cognition.U se ofricher andmore refined methodso finqui ryare nowbeginning

to pay real divide nds in disclosing th at such cognitive processesasmemory, attention,a ndconceptualization importantly par-ticipate in anim al behavior. In additio n, as the previou s reviewindicates, therea re good reasons tobelieve that at least somecognitive processes may be common to animals an d hum a nbeings (also see Was serma n, 1990).This rather o ptimistic evaluation notw ithstanding, a few re-marks are in order considering the factthat, just 20 years ago,manyhadproclaimed com parative psychologyto beeither ter-minally ill or dead (see Hodos & Campbell, 1969; Lockard,1971;Tobach,Adler, & Adler,1973,forpresentation an d dis-cussion of this thesis). Thus, th e concluding sections of thisarticle raise and, I hope, answer many questions that havearisen conc erning the place of comparative cognition in thebiology of beh avior.fs Comparat iveC ognit ion Comparative?

Evena cu rsory examination of the research reviewed in th isarticlewillreveal that only a smallfractionof it entails explicitcomparisons ofdifferent species w ithi n the same study. Howthencan oneclaimth at comp arativecognitionistrulycompara-tive?The answer to this query is that most comparisons amongspecies are conducted acrossdifferent studies. With a few nota-ble exceptions (see Bitterma n,1975),mo st researchers of cogni-tion in an ima ls have concentrated on one species. Such concen-7Others would also insist that reference is contingent on the effec-tivecom mun ication between speakers about thingsthatare notnec-essarily present; this additional requirement appears to have beenpassed bychimpanzees(Savage-Rumbaugh,M urphy, Sevcik,&R u m -baugh, in press).8Recent research b y Greenfield and Savage-Rumbaugh(1990,1991)has reconsidered the issues of gramm ar and sy ntax by examining theself-generated sequences of lexigramuse by achimpanzee. They con-cluded that thisape had thepo tential to inventarud imentary gram-mar. Gisinera ndSchusterman(1992),studyingalanguag e-traineds eal ion, have also made interesting observations on syntactical control.They concluded that this anima lca nlearnanumberofsyntactic rela-tionsfrom alimitedset ofstandard com binatorial sequences.

8/12/2019 Wasserman 1993

12/18

222 E. A. WASSERMANtration enhances the chances of that research incisively an dcomprehensivelyelucidatingthecognitiv e processesofthat spe-cies. Of course, the risk of overspecialization is that rathermuch may be learned about rather fewspecies. Beyon d rats,pigeons, monkeys, and apes, researchers know rather littleabout cognition in n on hum an animals. The field welcomes thesystematic stu dy of un derrepresented species.All ofth is discussion,ofcourse, ignoresth e most salientofal lcomparisons, namely, the comparison of anim al and hu ma ncognition.Here, the richempirical literature on cognitive pro-cesses in hum a nsaffords countless comparative o ppo rtunitiesto investigatorsofan ima l behavior. Also,as thepresent reviewclearlydiscloses, researchers havefrequentlyevaluatedthe per-formanceof their a nim al subjects in the context of hu ma n cog-nition.This most naturalof allcomparisonscan beexpectedtocommand a good measureof future research in comparativecognition.Is Comparat iveC ognit ion Cognitive?

Today, the term cognitive psychology has come to denote aparticular dom ain of hum an behavior and the termcog nitivisma specificapproach to investigatin g cognitive processes in hu-man behav ior. Most wh o advocate this cognitive approach havecuriously found it easier to accept functionalan d structuralparallels between human beings and digital computers thanbetween h u m a n an d nonhuman animals (Haugeland, 1978;Lachmaneta l . ,1979).Indeed, the information-processing anal-ogy of hum ans to computers has spawned thecomputationalviewof though t, which sees th ink ing as the m anipulation of aninternalrepresentation ('men tal model') of an extern al dom ain(Hunt ,1989,p.604).Most w orkers in comparative cognition do not ascribe to thiscomputational viewfo r fear ofsubstituting what theysee as aform ofmentalism forbehavioral ana lysis. Thus,as to the roleof internal representations in behavior, Sch naitter (1987) fol-lowsSkinner's(1985)lead inobserving that

there is no independentan d direct way to determine ho w thisinternal environment works. . . . The behaviorally relevantaspects of the inner environm ent are neither independe ntly assess-able nor are they directly controllable. C onsequently it is impossi-bl e to state any constraining generalizations about internal con-text, an dwithout knowledge ofconstraininggeneralizationsit isimpossibletoestablish their relationshipt op erformance general-izations.Th ewhole projectfor ananalysisofbehavior fails, (p .10)Segal(1978)added the following:

Therei s agreat dangero f are tu rnto theworst fallaciesof mental-ism and dualismin thecurrent riseo fcognitiveconcepts.Aware-nessor ahom unculus wi thin th e brain that controls informa-t ion processingis not a necessary p a r tof cognition, (p.214)Such concerns with cognitivism have not, however, pre-ventedinvestigators of animal behaviorfrom researching manyof the most complex an d challenging aspects ofcognition.Within a behavioristic framework (Wasserman, 1981, 1982,1983), researchers have studied such benchmarks of cognitionas m emory, attention, conceptualization, an dlanguage.One ofthe initiators of modern research on comparative cognition,WK. Honig(1978),has assessed this approach thusly:

The analysis is plausible because it places cognitive process andcognitivebehaviorwithinth eframeworkof a functional and exper-imental analysiso f behavior . . . . There is nothing magical ormysterious about the relevant experimental or criterion behav-iors, and thus processes remain within the realm ofbehavioralidentificationand analysis. We do not need a new kin d of psycho l-ogy to deal with cognitiveevents,(p . 1 1 )To study cognitive processes in b ehavior, then , in no way forcesone tofollowthe theor etical lead of cognitivists, althou gh someresearchers of animal behavior have explored theheuristicvalueof one of its most controversial notionsrepresentation(e.g.,Roitblat, 1982).Is Comparat iveCognit ionRelevant to EvolutionaryBiology?

Comparative psychologists have often been criticized fo rtheir choice of species to study. For instance, Ratner (1970)accused many of engaging in capricious comparison, an ideamore f u l l y developed byHodosand Campbell (1969):Muchof the curre nt research in comp arative psycho logy seems tobe based on comparisons between animals that have been se -lected fo rstudy according to rather arbitrary considerations an dappearsto be witho ut any goal other tha n the comparison of ani-mals for the sake of comparison. T his rather tenuous approach toresearchha sapparently been brought aboutby theabsenceo fanybroadtheoretical foun dat ion for the field, (p. 337)

Furthercontributingtoconfusionoverth enatureofcompari-son incomparative psychologyis theclaimbyKingandNich-ols(1960) that the concept of evolution, w hich is basic to thezoological system ofclassification, is probably not the conceptmost usefulfor a classification ofbehavior (p .22) .How canresearchers inaugurate the field of comparative cognition withDarwin'stheory of evolution and n ot adopt phylogenesis as theframework forclassifyingbehaviorand forselecting animal stostudy? One answ er isgivenin a recent article by Gottlieb (1984;also see Campbell & Hodos, 1991;and Yarczower & H azlett,1977).After reviewing past an dpresent practices inevolution-ar ybiology, Gottlieb concluded th atthe discernment ofevolutionary trends has always gone on bothwithin and without strict phylogenetic lineages since the incep-tionof the theoryofevolutionin theearly 19thcentury and . . .thispracticecontinueswithfruitful intellectual results to the pres-en tday.(p.448)

For Gottlieb, it is not a matter of comparative psychology lack-ingaunifying theory, as some have claimed (Hodos & Cam p-bell,196 9); rather,it is the natureo fthat theory.There isa theory of comp arative psychology and th at theory is,and always has been, foun ded on a psychological concep t o f ana-genesis; the progressive evolution of adaptive behavior, learningability,or intelligence (pp.448-449).

Manyreadersmay befamiliarwiththeevolutionary conceptofcladogenesis,thesplittingof organisms intodistinctand re-productively isolated populations over time. Whereas someevolutionary psychologists have insisted that only the latter isthe appropriate subject matter of comparative psychology,Gottlieb (1984)argued th at the form er is also a legitimate con-cern forcomparative psychology.

8/12/2019 Wasserman 1993

13/18

CO MP ARAT I VE CO GN I T I O N AN D A N I MA L I N T EL L I GEN CE 223Wha t distinguishesthe twoclassifications and causes uneasei nsome zoologistsa ndpsychologistsisthat th egradeo rlevelsno -tion[o fanagenesis]dealswithth erankingo fbehavioral organiza-tion (capacity),notkindsofanimals . . .;that is, it ismostoftennot aphyletic[o rcladistic]ordering, (p.453)

Thu s, behavioral analysismaysuggestaparticular hierarchi-cal organization of cognitive processes, some b eing relativelybasic an d others being rather more advanced (Razran, 1971;Romanes, 1884/1969).Thosedifferentgradesorlevelsofcog-nitiveprocesses may represent a mea ning ful evolutionary p ro-gression, one that might be recapitulated by species in differentand even distant evolutionary lineages. Tryingto reconstructsuchbehavioral hierarchies, ifthey exist,isunlikelyto succeedby s tud yin g a few closely related species, although this phylo-genetic strategy isuseful forotherevolutionary purposes:

Wheno nereads th ebiological literature insearch ofexamplesofbiological anagenesis, one finds that th e instances recited ar eusually th e readily discerned ones that come from comparinglarger(supraspecific)taxo nom ic units, as, for example, in the evo-lution of homoiotherms (birds and mammals) f rom poikilo-ther ms (reptiles), th eevolutionof the three-cone retina from th esingle-coneretina,and the like(Gottlieb,1984, p.451)

Why?Here,on e might suggest that most phylogeneticor cladisticreconstructions have involved behavioral traits of remarkablenarrowness and inflexibility, such as the reproductive behav-iorsofavians (Tinbergen, 1 959)androdents (Dewsbury, 1975).This suggestion is supp orted by a wealth of material showingthatinterspeciesdifferences in motor patterns of the kind m ostcommonly used in comparative [evolutionary] studiesare al-most invariably innate (Hinde & Tinbergen, 1958, p. 255).Also, because most of the studies made hitherto have dealtwithrelativelysmall behavior elements within groupsofcloselyrelatedspecies. . .[the]conclusionsdrawn. . .referat most tomicroevolution (Hin de & Tinbergen, 1958, p. 253).Different issuesan dstrategiesar eboundtoarise whenone istryingtocompare not thebehaviors themselvesbut the adap-tive processes manifested by those behaviors, such asintelli-genceand cognitionthe very processes that concerned Dar-win an d Spencer an d that launchedthe field of comparativepsychology. This contrast between behavior an d processw asnot made byself-servingcomparative p sychologists but b y twoleading experts in ethology, R. A . Hinde and N. Tinbergen(1958):

Ultimately itwil l be desirable to make comparative studies notonlyof overtbehaviorbutalsoof the causal mechanismsunderly-ingit.H owever, sincethemotor patternsa redirectly observable,itisthese which have been studied most often,(p. 253)Whereas H inde and Tinbergen (1958) surmised that itsgreater ease might make cladistic reconstruction a forerunnerof anagenic reconstruction, thereis no reason not toconsidertheseas twoparallel an dcomp lementary linesof evolutionaryinquiry.In a ddition, although they are strong critics of manyinterpretations of anagenesis, even Cam pbell and Hodos(1991)conceded that

ifanagenesis isconsidered to be a temporal sequence of grades(stages in the improvement of an organic design) exhibited in anumber ofdifferent lineages. . . ,then this functional anagene-

sis iscompatible with whatwereferredto as theanalysisofadapta-tion, whichdoesno tconfine comparisons towithin-lineage com-parisons,. . . suchstudies[being]bothvalidandvaluable, (pp.21 6 , 220)This discussion thus places comparative cognition squarelywithinthe realm of the evolutionary biology of behavior.

What Is theRelationBetween Comparative Cognitionand Cogn itive Ethology?

From the foregoing,on e mightdefine comparativecognitionas thecomp arative analysisofcognition inhumanand animalbehavior. Researchers inthisfieldgenerally adopt experimen -tal methods of investigation, thereby allowing careful controlan d manipulation of relevant variables, precise an d unbiasedmeasu remen t of behavior, and replicability of experim ental re-sults. They endeavor to abide by M organ's(1894/1896)canon inexplaining the processes of cognition with a small number ofoperationally defined theoretical notions.Se tapartfrom comparative cognition is cognitive ethology, aself-proclaimedantibehavioristic approach toproblemso fcog-nition inbehavior. Initiatedby D . R.Griffin (1976),the princi-pal reason for creating this field of inq uiry is to learn as muchas possible about the likelihood that nonhuman animals havemental experiences, and insofar as these do occur, what theyentail and howtheyaffect the animals' behavior,welfare, andbiologicalfitness (Griffin, 1978, p. 528).Recalling previous discussion of interpretation in the earlyhistory of comparative psychology,Griffin's (1978) call for acognitiveethology appearsto be athrowback to a prescientificanalysis of behavio r in term s of conscious experience. Writingin 1928, Warden (W arden, 1928) described the influence thattheemergence ofbehaviorism had oncomp arative psychology:Since that time, comparative psychologyhasbeen attemptingto readjust itself to a strictly natural science position as thelogical outcom eof theD arwinian conception ofpsychologyasa biological science (p. 508). It is eviden t that the adjustm entprocess is farfromcomp lete if researchers are still debating theusefulnessofsuch notionsa smin d, mental experience, aware-ness, consciousness, and other ideas Griffin believes can beproper subjectsforscientificstud y (see Ristau ,1991,fo rfurtherdiscussion ofcognitiveethology).9Thatone canim agine anima l behaviorto beaccompaniedbysubjectiveexperiences canno tbe the issue,for it isequally p ossi-ble to imagine those behaviors to be performed withoutconscious accompaniments, a point recognized nearlyacen-tury ago by Jennings (1904/1976; see his earlier quotation).There is simply no clear or necessary role for subjective experi-ence toplayin behav ior, as has been observed by Segal(1978):

9Concern with mental experience, althoughitmightnot be a fittingtopic forscientificinquiry, mayn everthelessbe an important source ofinvestigablehypotheses.Thus, w i thinth eexperimental analysisofani-mal psychop hysics, num ero us studies have beenconductedtoconfirmthe existence of various perceptual illusions in humans and animals;for anexcellent recentexample,seeFujita, Blough,an d Blough(1991),wh o reported that pigeonssee thePonzo illusion.

8/12/2019 Wasserman 1993

14/18

4 E. A. WASSERMANAnimalcognition doesno tim ply awareness. To saythatana n i m a llearns about environ mental relationships or relationships be-tween it s behavior and its consequences an d then acts on thatknowledgeis not to saythatt heanimal knowsit hasknowledge,o rknows what it is doing. It may simply be that organ isms wit hcom-plexbrainsreactto stimul us inpu t in complex ways. At no time isit necessary that th eorganism be an active, conscious participantin its information-processing functions; biological matter may besufficient to do the job as physical matter is sufficient to docomparable problem-solving tasks in the digital computer ,(pp.213-214)

One possible reason for the resurrection of mental istic termsatthis particular timeis thekeen publicandscientific interestthat surrounds many of the animal language projects men-tionedearlier. Because those projects havesucceededin estab-l i shingtwo-way communication betweenahumanand anani-mal, Griffin(1978) saw apossibilityforusing suchacommuni-cation system as a window on the minds of animals, throughwhichthey might themselves make known their thoughts, feel-ings,and emotions.

This ideaisindeed enchanting;but can itachieveitsobjec-tive?How canresearchersbe surethatthisandother so-calledwindowson the minds of animals are not in reality mirrors,reflecting backthethoughts, feel ings,andemotionsof the hu-mans? In addition, if researchers must resort to anthropomor-phism to interpret the behaviors of animals (see Burghardt,1985) ,must not researchers also now solve the persistent prob-lemsof introspectionism, whose intractability spurred the riseofbehaviorism?

Perhapsit isbesttoconclude this discussionofcomparativecognition and cognitive ethology with these sobering reflec-tions ofMason (1976) :

Whathasbeen achieved[ bybehavioristic investigationsofanimalcognition] mayseem pale incomparison withth evisiono fsittingdown like Dr.D oolittlefor an infor mal and revealing chat w ith ananimal friend. But even if this were possible, how much wouldnecessarily remain unsaid? If we have learned one thing fromyearso feffort devoted to the problem, it isthat there is no win-dow that wil l allow us to gaze directly on another mind, eventhato fanother hum an being,and to see itsworking s clearlyand tose e them whole. Mind, afterall,lacks thing qua lity ; it is but aconstruct, hardly more than alabel, really,forcomplex processesand functions that we are still farshort of understanding in anycreature, inclu ding ourselves. W e have learned what isperhapsth ehardest lessonofall: There is noroyalroadtomind;w e areforcedto approach along th eonlypathsthat ar eopen to us,through th etortuous byways of analysis, inference, hypothesis, an d recon-struction, (p.931)

ProspectusFrom the preceding reviewand discussion, it isclear thatresearch on the comparative psychology of cognition is enteringa period of real growth and accomplishment. As research in thearea of comparative cognition continues, there is likely to bemuchgreater contact withtheareasofcognitive science,on theone hand,andbehavioral neuroscience, on the other. Fullerelucidationof the similaritiesand differences between humanandanimal cognition plus greater appreciationof thebiologicalmechanisms of cognition wi l lsurely come fromthesecontacts.Alsounderway is an important effortto investigate theroleofecologicalfactors in the evolution of adaptive behavior and cog-

nition (e.g.,Kamil &Roitblat,1985) ,thus connectingthestudyof naturalisticcontingencies with complexandmodifiable pat-terns of action.10These increased interactions and interdisci-p l i n a ry effortsshould also reducetheintellectual isolation thathasheretofore characterizedthe field.

Thus, thesecond centuryofworkincomparative cognitionisof fto a faststart.Itsstrong commitmenttocareful lycontrolledexperimentalmethodsand itsclear focus onobjectively verifi-able processes ofcognitionshould help the field avoid the meth-odologicalandinterpretive trapsthat hampered progress dur-in gits firstcenturyofinqui ry . Thedebtto theearly evolution-ists can be no better paid than by advancing the field ofcomparative cognition in accordance with the best methods ofbehavioralscience.

10With in th i sareao