Intelligences and brains : an evolutionary bird\'s eye view

First publ, in: Comparative cognition : experimental explorations of animal intelligence / ed. by Edward A. Wasserman and Thomas R. Zentall. - Oxford [u.a.] : Oxford Univ. Press, 2006. - S. 555-579. ISBN 0-19-516765-1

Intelligences and Brains: An Evolutionary Bird's Eye View JUAN D. DELI US AND JULIA A. M. DEL IUS

henever repurrin!; un the cogn itive

W

an ima ls to lar audiences,

olle

~h ili t ie s

u(

is regu larl y COIl -

fronted with rhe question of whether olle is suggesting that animals are, in (acr, intellige nt. The seniur author h:1 S ofre n (u un d himself sidestepping this qu estiun. !lut, havi nJ.\ taken over a se min ar on intelligence in an ill1al s, hum ans, and machines, he could no longer evade ' it. Moreover, the partici pants compl.li ned that the semin ar's readings conveyed a mere mass o( disco nn ected knowledge. They wanted a-succinct te xt that would organize the nlnter ial. This chapter is an updated versio n of that re xt, which presents what rhe a utho rs cunsider to be an interdisciplinary co nsensual overview o( intelligence as a behavioral disposition displayed by so me an im als, humans included. There has never been any doubt about rhe facr that the hum an species is the cleverest on Earth. However, there ha s been much preoccupation with the circumstance that not all human individuals arc equa lly bright. In nea rly all areas of human elldeavor, ir is appreciated that the degree of cleverness or stupidity of an individual is a factor that defines the efficiency with which (s)he can execu te all but the mosr routine tasks. This efficiency is mustly judged by the accuracy, the speed, and the effo rtl essness with which a person ca n solve everyday problems. These problems might involve finding the way in unknown terrain with a map, baking a ca ke when ingredients and implements arc lacking, assembling kit (urniture without th e

in struction s, or serti ng strai!;ht a ll1uddled sc ient ific text. Although there is no undi sput ed definition uf intelligence, there is sO ll1e consCllsus that it has t(l do with varyin g a bilities to successfu ll ), adapt behavior rn novel situations, or more ahstracri)', with var)'ing capacities for a goa l-co nducive processing of cognitive information (Hunt, 1980). A test that meas ured intelligence was first de vised h)' Alfred Binet (185 7- 1911) . Inrelli gencc tests were stead il y improved so th at now severa l o( them provide a consistent, replica hie, and predictive mea sure of indi vidua l intell ectual ability (Kaufman, 2000). Farly rests fended to yield scores that were much influenced hy the cultural hack gro und and the formal edu cation that the subjects had experienced. Modern intelligence tem arc designed to be largely insensitive to these factors. An extreme in this respect arc the so-ca lled Raven sca les which arc thought to give nearly idea l mea sures of the gencral intelligence of individuals independently of their particular expertises and also independently of their more specia l cognitive abilities (Snow, Kyllonen, & Marshalek, 1984 ). Figure 28.1 shows both an easy and a difficl)lt exa mple of Raven sca le items. They must be so lved quickly, as the full test consi sts of many suc h items and has to be completed in a limited timc. These items have been designcd so that their so luti on is unlikely to be influenced by previous knowledge that individu als mighr or might not have. The sca les are constructed so that 'even extended experience with one

555

Konstanzer Online-Publikations-System (KOPS) URN: http://nbn-resolving.de/urn:nbn:de:bsz:352-206120

557

556

~

=#= #~

+0 ~ ~ [D lID [ID [ID ct> [J[) rID ~ ~ [@) ci) cb [i(o ~ Figure 28. 1. An e,ISI' item and a difficult item from the Ral'en intelligence sca les. The gaps mu st he tilll'd with one of the panels offered . (Ba sed on Carpemer, Just, & Shell, 1990.)

version of them has onl), a negligible influence on huw well one performs with the next version of them. The fact· that the intelligence gauged by such tests is an ability that cannot be easily taught and lea rned is undoubtedly the source of widespread un eas in ess. What the test measures is often felt not to correspond with the everyday connotationswhatever they precisely are-of the ter m itttelligel/ce (Lenz, 2000). Many textbooks of cognitive psyc hology avoid the sub ject of individual differences in cognitive competencies perhaps because of political correctness (e.g., Matlin, 2002; Medin, Ross, & Markman, 200 I; but sec Sternberg, 1996). The nution of innate individual differences seems to go against the fact that experience can improve one's performance with tasks such as improving texts or assembling furniture. On the other hand, some of us seem inherently good at drawing (or mathematics), but inherently poor at writing (or singing). We will return to these issues later.

ANIMAL INTELLIGENCES Even before Darwin (1809- 1882) proposed that one must look for the roots of human intellect in animals, their intelligence was of much practical interest. Shepherds and cowboys were aware of the differences in working intelligence a mong in dividual co llies or horses . Older members of the senior author's laboratory still remember one extraordinary pigeon that could solve comp lex

behavioral ta s ks within hours or days, problell1s that other pigeons would not so lve within 'web or months. Because they arc after universal s rather than particulars, animal behavioral scientists have largely ignored the interindividual vari ations in cognitive capacities of their subjects (bu! sec B. Anderson, 2000; Matzcl ct aI., 2003). But also, anybody watching mountain sheep and macaque monkeys at the zoo is likciy to be sure that the latter arc more intelligent than the for mer. The same applies to more gross phyletic groups, such as birds heing rated more clever than fishes (Nakajima, Arimatsu, & Lattal, 2002 ). Using more scientific methods, sc holars bave been interested in such interspecific differences as, for example, are dolphins more cognitivcly capable than monkeys? But the design of a universa l test of animal intelligence has turned out to be difficult.

Learning Sets Harry Harlow (1905- 1981) a nd colleagues attempted to assess the intelligence of various species by judging how well they would acquire lea rning sets. The learning-to-Iearn procedure they used involved challenging individual animals with ,\ series of discrimination learn ing problems, a ll of whicb were procedura lly identical, but each invnlved new stimuli. For example, macaques first had to Ica rn in repeated trials that lifting a toy car rather than a

;It:hicve sllch an imlllediate switching strat('~y simultaneously presented toy airpl'lne would reveal (Ileliu.s, Amciing, L.ea, & StaJdon, I ~95; Diebmp, a hidden peanut reward. Next, the)' had to learn Prior, & Giintiirkiin, 1999; Staddon & Frank, that a cup. but not a call, signaled rcw.Hd, and so 1974). So me other hird spec ies pcrformed hetter on with a number of "junk-object" pairs. The anion this ta sk, achiel'ing Iel'els comparable to those mals initially rook a considerable number of trials of monkeys (Gossene &. Gossette, 1967; Kamil, )vith eac h I;air of new objects before consistentll' picking the rewarded one, but, by about the lOath .iones, Pit;cwicz, & \-Ianldin _ 1977). Of a sample of 20 universitv students, mal1\' achiel'ed nearl y pair, th e macaq ues had learned to choose correctly rule-like behal'i;'" aiter lil'e reversals on a ta sk in bl' the seco nd trial with each nOl'el pair (Warren, 1973 1. It \\'as found that some species, foremo st vo ll'ing the concurrent discrimination of two pairs chimpanzees, showed ne:H-pc r(cct dis("riminatioll uf irregular polygons. However, a fell' students did not ex hibit any significant performance improve on the secolld pair after experience with on'" about mems . Remarkahly, among the successful st udents , a dozen stimulus pairs (Schustcrrnan , 19641 . Other spec ies, slich as the rat, oilly show minor improve - severa l could not verbally ex pl 'li n at all what rule the)' had fin'lily used (Siemann, I'on Selzam , &: ments in this learning-to -icarn task. The laner Borchert, 2004; see later). species were accordingl), considered to be less illtel According to an carly study, fats show CO Ill ligent thall the former. ivlost 6- to 7-year-old schoolchildren arc quicker than chimpanzees at ac- parativciy little learning-to -Iearn ability (Warren, 1973). The rats had been trained to jump toward quiring this optimal str,ltegy, needing onl), pre a pair of doors bearing different visual patterns experience with two or three pairs of stimuli, but apparently, whether the individu'll perform'lnce of such as a !I'ian~le ,,,HI a circle. The door di sp lav ing the triangic' would, for example, be designated children on this kind of task correlates with their test intelligence has not bee n examined (5 . iluschio, co rrect" :lnd open to giv(.' access to il food reward. personal cOllllllunication). Learning-set tasks can The other displaying the circle lVould be the in correct one and would be locked , the rats falling be viewed as the learning of a special expertise that helps to solve a particular kind of problem but it is onto a net as a penalty. However, r,lts arc nightactive animals and arc not well ,1(iapted for the not quite what human intelligence tests attempt to rccognition of visual panerns in daylight. When Illea sure (sec later). Somc work ha s bec n done on the learning-set task with Ilolllllamrnalian species, Siornik and Katz (19 74; sec also Slotnick, Han ford, & Hodos, 2(00) used different odors as and some hird species have done quite well (Kamil, stimuli and arranged a more suitable procedure, Lougee, & Schulman, 1973; Plornik & Tallarico, thc)' found that rats lVere considerably more ca 1%6). There arc dara on a somewhat wider rang(.' of pable than originall), judged. LJn surp risingly, rats were hener with stimuli that they arc better species for another ta sk of a similar, bur simpler nature: serial reversal learning (Warren, 197.1). equipped to recognize . One can expect that hu mans, heing relativel), microsmatic, would con Subjects learn to discriminate a pair of stimuli as versely do worse on an olfactory test than on a before, but once the)' have acquired the discrimina visual test, even if the odors were chosen to be tion, instead of being confronted with a new pair, disc riminable for them (d. Danthiir, Roberts, Pal they continue with the same pair, now with the contingencies of reinforcement reversed. That is, if lier, & Stankow, 200 I). The point is that it is hard to des ign a testing situation that is sufficiently choices of A were rewarded and choices of II were uniform to )'ield comparable meas ures while be not rewarded previously, choices of B arc now rewardcd and choices of A arc not. When the su h- ing equall)' appropriate to the perceptual, motiva tional, and mororic dispositiuns of species that jects have learned to respond adequately to this arc adapted to quite diverse environments. The reversal, the reinforcement allocation is reversed difficulties arc not unlike those arising when hli again, and so on. t\ cognitively gifted species can man ps),chologists attempt to design intelligence be expected to lea rn to behave according to the tests, which are also fair to motorically ·di sabled rul e, "if a stimulus ceases to yield reward, switch 01' blind persons. The upshot of this con undrum is to the other st imulus on the next trial." Indecd, that behaviorists have large l), given up t rying to most macaques manage to adopt such a strateg), within a few reversals. Pigeons show improvements desi gn a universa l animal intelligence test. The)' ill learning the sliccess ive reversals, but never have rather turned to exa mining the perforn1R.

0

()

91:

0

~

35 u 0

::r

0

o· «) (j)

D

Figure 28.2. Concept forlllalion

0

lead ing

acco rdin g to funct io nal equi va lencies

in pigeon s. (:Vllldified fro m Siemann , \'o n Ferse n, & Dciius, 1998.1

tra iling pair

Ilolhuis, 200 I). Indeed, thi s facu lt), appea rs to be dea lin g wi th aud ito r\' Ihan with visua l st imuli associated with il relative enlargement of the hi p(Roitblar & vo n Fersen, 1992 ). Siema nn et a l. (2004) found lila r, although uni - pocampus, il hra in area rhat appropria tciy, among versitv srudents were natura llv more eflicie nt tha n ot her things , is kn ow lI to be illvo lved ill ellwd ing pigeo;ls or do lphins at lea rnil;g the basi c disaimi - the spalial layout of the environment. Howeve r, it nati ons constituring the ser ial reversa l t'lsk, the)" is not altoget her ce rta in whet hcr thc latter en large still varied individuall)' in th eir abi lity wi th tesrs in - ment is not a secondary ontogelletic adaptation (d: volving equivalence cla ss tran sfe r. The students Ek stro m et al. ,)OOJ; M aguire et aI., 20nO). also diffe red co nsiderabl y in their awareness of the task's logica l stru cture, hut rh e relati on berwecn the stml ents' rra nsfer pe rfor mance and their test in- Tra nsitive Respond in g telligence was not exa mined . Experiments using a different mer hodolog)' hav e shown that mentall y It is not immedi atel y obvious tha t if x < d, III > d, th en it mll st follow tha t «III . The more handicapped sc hoo lchildren have cons idera bl y and lIl ore difficulty forming equiv alences rhan sa me· terms that arc involved and the less well ordered age co ntro ls (Sidman, 1992) . In any case, tbese th e premi ses rh at arc presented, rh e harder it is to studies suggest th ar some behaviora l tas ks with draw the correct con clu sions. Transi ti ve inference which at least some birds and mammals ca n cope probl ems of a si mila r nature have been designed so also cha llenge the cogn iti ve competencies of th at they ca n he given to very yo ung children or, inyou nger or lesser gi fted humans, eve n if it is not deed, to anim als . In the simples t case , t he)' lea rn clear whether th ese competenci es arc those mea- ove r many tria ls to discriminate co ncurrently four overlapping pairs of five st imuli, A+B- , B+C- , sured by inte lligence tests. Whar is obvious, howeve r, is th at the uniform C+D- , ,md O+E- , ra ndo mly o rdered, umil they assessment o f th e intelligence of animal spec ies is ac hieve a criterion level of choice accuracy. Then, made difficult b), the different spec ia lizatio ns th at anima ls and child re n arc presen ted with a test pair the various spec ies have evolved. It is reaso na ble to of stimu li not prev ious ly presented together, 13000, speculate that the olfactory intelligence of rars is witho ut reinforcement (0) for their choice. If the lin ked to th eir nocturnal way of life and th at the subjects imerpret the task in terms of a transitive auditory inte llige nce of dolphins is rel ated to their inference problem (1\ >13 , I.I>C, C>D, O > E), th en sonar Ill ode o f or ientation (Shettieworth , 1998). the)' should prefer 1.1 to O. In fact, that is what Sim ilar eco logica l arguments ha ve been proposed subjects acru a lly do (children, squ irrcl monkeys: in co nnecti on with the presence of an enhanced Cha lmers & McGonigle, 1984; pigeons: von memory compete ncy in bird species that cache sur- rersen, Wynne, Delills, & Staddon, 1991; d. figure plus food in fllilny different locations for later, 28.3). Note that all th ree spcc ies did aboll! equa ll y well wi th test pair 13000. However, the pigeon s lea ner times (K rebs, 1990; but see Macphai l &

«x,

560

premises A+8-

100 % U

Ul

_

A~

75

monkeys

u

Fersen, &

C •

pigeons, squirrel mo nkeys. and human s.

50

lOok longer to learn the premise pairs than the monke),s, who took longer thall the children. Siemann and Delius (1998; sec a lso Dclius & Siema nn , 1998) comluctcd analogous expcrimellls with studellls .. li s reinforcements for correct and illcorren stimll iu s choices, srudents gai ned or losr symbo lic coi ns. The srudenrs learned the premises far fasrer rhan pigeons or indeed children, but rhey were nor Illuch better on the co nclusion pairs. When faced with rhe test pairs, two rhirds of rhe stud enrs did quire well, but about one third of the srudenrs chose randomly. Interesting ly, these failures were not linked with lower test intelligence. Rather, it appeared that these individual s had lea rned to cOlTectly respond to the va;'ious stimulus pairs (liB, BC, etc.) as compound patterns without atte nding to the fact that the component elemellls (B for examp le) recurred ill differenr pairs. When later faccd with BoDo con clusion pairs, these students did not have any rcady response to these new co mpoulld stimu li . It is likely that a n en hanced capacity for configura I perception is a parr of human intelligence (Matsumoto O higas hi , Fujimori, & Mori, 2000), but here it was counterproductive because it cou ld not provide the information needed to solve the problem. Questionn aires after the experiment revealed thM abour half of the studellls who solved the test pairs were aware of the logica l structure of the inference task but that the ot her ha lf were not. Nevertheless, the students with an explicit understanding did not perform better than did the ones without. We assume that the form er-des pite

A

8

c

o

image of the sample stimulus. Choices of the sa mple -identical lateral stimulus were penalized with a period of darkness. It was import'lIlt that the birds learned to perform this !ask wi.th several differe nt visual shapes to ensure that thc)' had acquired a genera l preference for the mirror stimuli. Pigeons were ea rlier thought to be unable to learn such nn "a lwa)'s choosc odd stim ulu s" rule, but it is now dear that rhe)' ca n do so under propitious conditions (Cook, 2002; Deiiu s, 1994; Young & Wasse rman, 200 I). The test stage involved presenring both comparison sr imuli rotated hy rhe "l llle angle relative to the sa mple stimulus. Aparr frolll rhe 0° training di sparit)" rhe pigeons were tested with 45°, 90°, 135°, and I SOO sample-comparison di s parir~' triplets. The tests showed that the pigeons made about the sa me number of errors, about 10% , and took about the same amount of time (iust under a second) to choose, regard less of the varying orientation disparities (figure 2H.4; Hollard & Deli us, 1982; Lombardi, 19S 7). It /leed s to bc srressed that rhis resu lt docs not mean thar pigeon s arc invari'lbly in scnsitive to the orientation of visual palterns, as some critics seem to have ass lIllled (c.g., Halnm & Matheson, 1997) . In differently de signed experi Illents, we ourselves showed that, under differcnr

competencies like learn in!', sets, concept learnin!'" or transitive responding arc borderline or absent (but see l\sh'H~', Wickler, & Frickl', 2002). This picture appears to be true despite the facr that Macphai l (198 7) has proposed rhar all I'ertebrates except human s arc of about eq ual intelligence.

~

PigeOn~

'0 .c

VOII

8000 ---'

~- ~ . ---........::>- - 10'1) . Chichesrer, UK : Wil e)'. Anderson , ,vI. (I '.>\1 81. ivlcnr;tl retard arion, gm eral intellige ncl', a nd Ill odularity. /. (" 'l'II illg 2,. Ill'/iI'idllil l Oi/(en·" a's. 10, 15'.>- 1n. ,\rhas, 1'. A., Ml' inerrzha gl'n, I. A., & Shaw, S. R. (1 99 1J. Evo lurioll in Il CrnHl S S\'S rCIll S, AIIIIII(,1 I':'3H. Asrl C)', S. I.. , & Wasse rlllan , E. 1\ . 11 999 ). Supcro rdinate calcgorv fonn ari o n in pi geons: Associ a' rion with a CO llllll On dela )' or pro bability of food rcinfOl'cclIl cllt Illake s perceplually di ssilllil ar stillluli funerion a ll y equi v;tien!. JOIlI'll(/1 IJ( 1':xllel'il/l l' lIl{/ll's),chl/ log),: Allill/al /l chaui" r I'm cesses, 1.1, 4'1 5-- 1'12). Calli bridge: Call1bridge Univc rs itl' Press. Ccci, S. .I ., N ightingale, N. N., & 'Baker, .I . C. (1 993 ). T he eco logies o f intelli gencc: Clwl · Icnges to tradirional views. In D. K. Detrerman (Ed.), Is Ihe /IIilld /IIodll/ar or IlIIil"n? Cllrrel/I l opics

0/

hlllllllll illtel/igell ce

1,;1"

6 1- 82 ). Norwood, N.I : Ablex. Chalmcrs, M., & McGo nigle, B. (1984). Arc dtil · dren any more logica l than monk eys 011 rh e fi ve tcr m serics prohlem? .1011 mal o( EX/leri· IIl elllll l Ch i ld l'sYc/lll log)" 37, 355- 377.

Chevc rud, J. lVI. , Fal k, D., Yalllli er, M., Konigsberg, 1.., HcllIlk '"11 p, R. C , & H idebolt , C (1 99 0). Heritability of brain si ze and surface features in rh esus macaques (MacllCil 11/11I,l//a l . j Olll'llal o( H aec/il)'. 8 1, 5 1- 57. Cia'rroc hi , .I. V., Chan, A. Y. C, & Ca puti, 1'. (2000) . :\ crirical evalu arion of rhe emotional intelli l:\c ncc consrnrc!. l'