1973, vol. 82, no.2, 211-226 decisional analysis of...

Journal of Comparative alid Physiological Psychology1973, Vol. 82, No.2, 211-226

DECISIONAL ANALYSIS OF THE EFFECTS OF LIMBICLESIONS ON LEARNING IN MONKEYS

ABRAHAM A. SPEVACK' AND KARL H. PRIBRAM

Neuropsychological Laboratories, Stanford University

In a 1969 study, K. H. Pribram, R. J. Douglas, and B. J. Pribram came to thehypothesis that behavior during discrimination learning and reversal is undercontrol of two competing variables: the patterned cues to be discriminatedand the noncontingent schedule of reinforcement. The current study using amodified decision theoretic procedure shows that in fact these two variablesare operative and that noncontingent reinforcement produces a strong posi-tion bias against discriminating. This bias is quantitatively more easily over-come b~' normal subjects than by monkeys with hippoc'ampal-amygdala le-sions though the strategy and tactics used are the same for both groups.Thus, hippocampus and amygdala are shown to influence attention throughmechanisms that regulate motivational bias.

."

Recently, Pribram, Douglas, and Pri-bram (1969) reported that monkeys withboth hippocampus and amygdala bilaterallyremoved (hip-am ablations) showed a pecu-liar deficit in the original learning and re-versal of a pattern-discrimination task.During the acquisition phase of the experi-ment, the experimental subjects, when com-pared to their controls, showed a prolongedperiod of chance performance before theybegan to respond differentially to the re-warded cue. Once discrimination com-menced they reached criterion normally.During reversal training the hip-am ablatedmonkeys extinguished their responses to thepreviously rewarded cue as readily as didthe intact subjects. But, as in acquisition,the experimental subjects performed atchance for a long period before respondingto the renrsed contingencies. Again, oncediscrimination commenced, criterion wasachieved as rapidly as by the unoperatedcontrols. Thus, the major difference betweenthe performance of the experimental and in-tact monkeys during both original and re-versal learning was a long period of chanceperformance.

Pribram et al. (1969) suggested that thisperiod of chance performance results fromthe fact that monkeys with hip-am abla-tions were unable to keep their attention

'Requests for reprints should be sent to Abra-ham A. Spevack. who is now at the Department ofPs~'cholog~'. Virginia Polyteehnie Institute andState Uni,'ersity. Blacksb1ll'g. Vir/!inia 24061.

fixed on the relevant stimulus dimensionlong enough during periods of relativelyrandom reinforcement such as occur at thebeginning of original learning and after theextinction phase of reversal training. In-stead, they give up attending (observing)though they continue responding, which ap-parently comes under the control of a non-contingent reinforcement schedule (50%variable ratio) .

The present experiment, using the samesubjects, was undertaken to replicate and toextend the earlier study. The aim was togather more 'data which could be subjectedto a variety of analytic procedures thatwould test the suggested hypothesis. Espe-cially important, a modified response-oper-ator-characteristic technique derived fromsignal-detection theory was instituted in anattempt to determine whether the atten-tional deficit of the hip-am monkeys is dueto a change in selective capacity (discrimi-nating the stimulus pattern from noise) orto a change in the motivational bias thatalters the response to that stimulus pattern.

METHOD

8ubjects and Lesions

The subjects were eight adolescent rhesus mon-kevs individuallv housed with free access to water.Ai'ter behavior;1 testing the~' were fed once perda~' with Purina monke~' pellets and fruit of suffi-cient quantity to maintain normal growth and re-liable responding for the duration of their dail~'test sessions, Fo1ll' monke~'s had been subjected to

211

212 ABRAHAM A. SPEVACK AND KARL H. PRIBRAM

bilateral removal of the amygdala and hippocam-pus through direct visual identification and consti-tuted thc experimental hip-am group (Figllfe 1).Details of the surgical procedure havc been re-ported elsewhere (Douglas & Pribram, 1966). Theremaining four monkeys were unoperated andserved as an intact normal control group. How-

ever, one unoperated monkey died shortly after theinitiation of these experiments, thus reducing thecontrol group to three subjects. This placed con-siderable strain on the statistical procedures thatwere used to determine the reliabilitv of the re-sults obtained. The deceased control 'll1onke~' wasnot replaced, however, since both tIll' hip-lllll ancl

41

25 17

f

413325 17

FIG. 1. Heconstruct ion of hip-am lesions. (8trippling indicates sparpd hi'ppocampus; c)'l)sshalch. Ilwlesion. OIl cro:;s sectiun, Ic:;ion border is indicated by heavy black.)

LIMBIC LESIONS AND DISCRIMINATION LEARNING 213

..

intact monkeys were those used in the earlier study(Pribram et aI., 1969); and thus, it was very diffi-cult to provide a new subject with identical experi-ence.

Apparatus

All testing was done with the DADTA III sys-tem described in detail elsewhere (Pribram, 1969).The animal-testing unit consisted of an enclosurewith one of the sides a 4 X 4 matrix of 16 trans-lucent panels with a food cup below. The discrimi-nanda were lighted numerals projected from theback of each panel. Stimulus pattern and positionwere controlled by a PDP-8 computer which alsorecorded and collated response parameters such asstimulus choice, the position of the panel pressed,the latency of the response and whether 01' not 'theresponse was correct and rewarded.

Procedure

.-\ll subjects received the following trammgregimen: pretraining, pattet'll discrimination, twosuccessive reversals using a strict criterion, andfive subsequent successive re,'ersals to a laxcriterion.

Since the monkeys had already been trained onpattern-discrimination and reversal paradigms, itwas not necessary to put them through an exten-sin shaping and pretraining program. However, thelong interval between the end of Pribram et al.'s(1969) study and the beginning of the present ex-periment made it expedient to rehearse the shapingprocedure. On the first day (50 trials) 4 of the 16panels pseudorandomly displayed the numeral "1."the other 12 panels remaining blank. On the foursubsequent 50-trial blocks, only two instead of fourpanels displayed the 1. During these five pretrain-ing sessions, the depression of an uulit panel wasrecorded, but did not affect the display or advancethe program to the next trial. A press of a paneldisplaying a 1 produced the delivery of a 190-mg.Noyes banana pellet, the initiation of a constant5-sec. intertrial interval, followed by a new stimulusdispla~·. The five pretraining sessions were suffi-cient to ensure that all monkeys were respondingconsistently with short response latencies.

For pattern-discrimination training the numerals"3" and "8" were simul t:meously displayed; theother 14 panels remained blank. Unlike pretraining,however, the stimuli did not appear randomly overall panels. Rather, the display was restrieted to thepanels of the left and right columns of the DADTAmatrix with the middle two panel columns alwaysblank. During a trial the 3 and 8 numerals were dis-pla~·ed randomly on one of the four panels ofeither the left or right column. the remaining nu-meral, on one of the four panels of the oppositecolumn. Thus. stimulus position was determinedrandomly on each trial with the constraint thatonly one of the stimuli appeared in each of the twoend columns and that each stimulus appeared ineach column on 50% of the daily 50-trial sessions.

The daily 50-trial blocks \~~re continued until

the subjects pressed the 3 panel 90% on 3 successivedays. This strict criterion was chosen to assure thatall vestiges of position bias were eliminated beforethe monkeys were considered to have learned thepattern-discrimination task.

The first and second discrimination reversals be-gan on the day following the attainment of criterionarid involved similar procedures. (For the firstreversal t.he 8 was rewarded; during the second re-versal, the 3.)

The subsequent five reversals were identical tothe first two reversals except that a more lax cri-terion was adopted. The monkeys were advanced tothe next reversal after onlv one 50-trial session inwhich they responded !lO%' to the reinforeed stimu-lus.

RESULTS

Pre training

All monkeys retained aspects of theirprevious experience. When introduced intothe apparatus during the first pretrainingsession they immediately began to presspanels with lit-panel depressions predomi-nating. In fact, very few blank-panelpresses occurred during any of the five pre-training sessions. Unlike original shaping,there were no reliable differences in thenumber of blank-panel presses between theexperimental and intact monkeys (Mdn =~, M cln = 6, respectively) during pretrain-mg.

Analysis by Trials and Errors

The results of the discrimination problemare summarized in the first column of Table1. As in original acquisition the lesion pro-duced a significant learning deficit, thehip-am monkeys requiring over four timesas many trials as the intact subjects toreach the lax criterion. The differenceproved to be statistically reliable (p < .03)according to a Mann-Whitney V test. Un-like original acquisition, however, we foundthat rate of blank-panel presses was notdifferent for the two groups. No monkeyexceeded a rate of more than one blankpress per day over the acquisition session.Once monkeys reached the lax criterionduring pattern-discrimination acquisition,they went on to reach the strict. criterionafter two successive sessions. The only ex-ception was one hip-am subject that re-quired 1,400 addition trials to reach the'


TABLE 1J\1!-:AN NUMBER OF TRIALS TO CRITImION AND MI'AN Ih:vl':HSAL

RATIOS FOR THI' HIP-AM ANI> INTACT GROUI'S

Acquisition Reversal

Group 1 I . 2 3 4 5 6 7Lax Strict

I I--- ------ --

Lax Strict LaxI

Stric t Lax Lax Lax Lax Lax

Hip-am 675Intact 150

Hip-amIntact

.Mean no. of trials to criterion

1100 1588

I2138 I 1113

1

1438

I875 575 5(i3 450 51:3 ..,

250 250 350 200 300 183 183 183 200 lU7

Mean' reversal ratios l-

I2.6

I-

I1.8

I1.3

I.8 .9 .7 .7

1.4 - 1.2 .7 .7 .7 .8 .7

Note. Reversal ratios were calculated by expressing the number of trials required to reach the strictcriterion for the first two reversals and the lax criterion for the five remaining reversals, as a ratio ofthe number of trials to reach the strict criterion during acquisition.

strict criterion after having reached the laxcriterion.

Table 1 also summarizes the results ofthe seven successive reversals. It can beseen that the hip-am subjects required sig-nificantly more trials to reach the lax cri-terion than the intact monkeys over all thereversals (p < .03; Mann-Whitney U test).Moreover, in the first two reversals, thehip-am subjects showed a marked difficultyin reaching the strict criterion even afterthe lax criterion was acquired. Whereas allthe intact monkeys required only the mini-mum 100 trials to reach the strict criterionafter the lax, only one hip-am in the firstand a different experimental monkey in thesecond reversal were able to accomplishthis. This difference was reliable at the .06level according to a Mann-Whitney U test.

It is possible that the significant differ-ences in the reversal performance of the ex-perimental and intact monkeys could be at-tributed to their initial differential abilitiesto acquire the pattern-discrimination task.Thus, the animal's reversal performancewas adjusted to take account of their abil-ity to perform the pattern discriminationby expressing trials to criterion on succes-sive reversals as a ratio of the number oftrials to reach the strict criterion during ac-quisition. The results of. this analysis arealso summarized in Table 1. It can be seen

that higher ratios than those shown by theintact monkeys were again found for thehip-am subjects during the initial t\vo re-versals. However, the ratio differences be-tween normal and experimental monkeysrapidly decrease during the initial two re-versals, with a slower rate of decrease ap-parent during the later reversals. There wasonly a single overlap in group ratios duringthe first reversal, but over half the hip-amanimals showed lower ratios than the intactsubjects by the seventh reversal. Overallgroup differences reached the .06 level ofsignificance according to a :Mann-WhitneyU test.

To demonstrate whether the acquisitionand reversal data of the present study werecomparable to those of Pribram et al.(1969), we used the same method of analy-sis as they did, i.e., the method of successivecriteria. This analysis (as shown in Figure2) assessed the number of trials required byhip-am and intact monkeys to approachthe strict criterion in successive 10% incre-ments in performance. This figure indicatesthat, below the 50% lewl of performanceduring acquisition and reversal, both experi-mental and intact monkeys improved atcomparable rates. The major difference,.. inperformance between hip-am and intactmonkeys thus became apparent only afterthe 50% level of performance had been


o

18 20

R~v~r5al 1

0-0 HIPAM5

00----0 Intacte

R~vcrsal '2

Acquisition

4 6 () 10 12 14 16Trials to crit~rion x fOO

2


improvement by the hip-am group wasslower than that of the intact Ss at allpoints beyond the 50% level of performanceas indicated by a decreased curve slopewhich is maintained up to the acquisition ofthe strict criterion. By the second reversal,however, the hip-am and intact curvesshowed striking parallels to the curve gener-ated by the experimental and intact sub-jects during pattern-discrimination learn-ing. As in acquisition, the intact and hip-am curves separate at the 50% level of per-formance with the experimental monkeysshowing major difficulties reaching the 60%level of performance but no difficulty inachieving the lax criterion. Unlike acquisi-tion, however, the experimental monkeysalso showed deficits in reaching the strictcriterion after the lax criterion had beenacquired.

Analysis by Latency of Responses

Table 2 shows median response latenciesand mean number of nonreinforced respon-ses for the first 50 trials of the second andseventh reversal divided into five lO-trial

TABLE 2J\


.'

viously rewarded or previously nonre-warded trials during the course of the sec-ond reversal. To assure that latency of re-sponding during similar phases of the rever-sal were compared for all monkeys, thetrials were organized into five Vincentizedblocks of trials. Thus, for different mon-keys, blocks of trials might be comprised ofdifferent numbers of trials but accountedfor equal proportions of trials required toreach the strict criterion. It can be seen inTable 3 that during the first Vincentizedblock of trials the hip-am monkeys showedlonger response latencies than the intactmonkeys regardless of whether the previoustrial had been rewarded or not rewarded (p< .06; Mann-Whitney V test). In addition,an interesting interaction was observed.The hip-am group response latencies after arewarded trial were longer than those fol-lowing nonreward; the opposite relation be-tween previous trial choice and response la-tency holds true for the intact monkeys (p< .05; Fisher Exact test). It can also beseen in Table 3 that the substantial groupdifference in response latencies apparentduring the first Vincentized trial block wasreduced progressively during the remainingtrials of the reversal.

Analysis by Signal-Detection Technique

A modified signal-detection procedure al-lowed us to partial out the part of the sub-jects' performance based on the detection ofstimulus pattern from that due to positionbias and is shown in Figures 3-9. Since werestricted the presentation of the reinforcedand nonreinforced cues to the extreme leftand right columns of panels of the DADTAwe made only four stimulus/position con-tingencies available for the monkeys' po-tential responses; i.e., the monkeys couldpress the reinforced or nonreinforced cue ineither the left or right column of the stimu-lus array. Within a given block of trials thenumber of occurrences of each stimulus/po-sition contingency was set according to apredetermined schedule. Thus, the numberof responses to either cue presented in eithercolumn could be expressed as a relative fre-quency of the number of times each of thestimulus/position contingencies actually oc-

TABLE 3HI I'-AM AND INTACT 1\1 ED!.\N Ib;sPONSI'; LATI·;I\CI ES

(IN :;I·;C.) A~'TEH RF:WARDED AND NONHE-

WAHDED RESPONSES FOH FIVI';

VINCENTIZED BLOCKS OF

TRIALS REQUIRF:D TO

.' ACHI EV I': TH E STHICT

CHITEHION DURING

THE SECOND

I{ IW E Hi3c\ L- ..

Trial blockCondition and group -

I 2 3 4 I 5---~_._---------- --.-After rewarded

Hip-am 4.210 5.674 2.554 2.314 2.095Intact 1. 893 1. 913 2.483 2.185 1.469

After nonrewardedHip-am 3.309 2.884 2.449 2.537 1.780Intact 2.488 2.120 2.485 2.359 2.171

-

curred within the block of trials. However,the magnitudes of these four relative fre-quencies were not independent. Rather, bycalculating the probability of the subjects'responding to the reinforced and nonrein-forced cue in one column we simultaneouslydetermined the probability of responses tothe reinforced and nonreinforced cues pre-sented in the other column. Therefore, wemerely calculated the relative frequency ofresponses to the reinforced and nonrein-forced cue presented in one column whichwas sufficient to summarize the animals' de-tection of pattern and position bias. Ani-mals that responded predominantly on thebasis of position preference produced highrelative frequencies of responding to boththe reinforced and nonreinforced cues pre-sented in the preferred column of panels.Moreover, it became apparent that themonkeys detected the difference betweenpatterns when they produced high relativefrequencies of responding to either the rein-forced or nonreinforced cue presented inthis position. The value of the relative fre-quencies of responses to both cues presentedin one column were then used to define thelocus of a point within an appropriatelyconstructed square to represent graphicallyan individual subject's performance over ablock of test trials. Figures 3-9 furnish ex-amples of this procedure. The ordinate ofthe square represents the relative frequen-


t

cies of the monkeys' responses to the rein-forced cue; the abscissa of the square repre-sents the relative frequencies of respondingto the nonreinforced cue in the same col-umn. Those points which approach the up-per-left and lower-right corners of thesquare represent good detection; points inthe upper left indicate that responses werecontrolled by the reinforced pattern and oc-curred in the latter stages of pattern dis-crimination and reversal learning; pointsnear the lower right indicated that the sub-jects' responses were controlled by the non-reinforced pattern and occurred during theinitial stages of reversal training. Thatmonkeys responded more on the basis ofposition bias than stimulus pattern was in-dicated by points located at either the low-er-left or upper-right corners of the square.

Figures 3-9 summarize the results of suchan analysis for an individual monkey's per-formance during pattern-discrimination ac-quisition and the first two reversals, each ofthese phases being broken into 50-trialblocks. Responses to the reinforced (the 3during initial discrimination and the secondreversal; the 8 during the first reversal) andnonreinforced cues presented in one columnwere expressed as relative frequencies of alloccasions. These cues appeared in this col-umn for each block of 50 trials. Thus twoparameters were available to assess eachanimal's performance during acquisitionand reversal: the number of points in thesquare representing the number of trials re-quired by the monkeys to reach criterionand the position of each point in the squareindicating in what manner individual sub-jects' performances 1110\'e through thesquare in achieving criterion.

It is apparent from Figures 3-9 that thehip-am monkeys required more trials toreach the lax criterion during pattern dis-crimination and both the lax criterion andthe strict criterion during the first two re-versals. It is also apparent that the major

struetcd by joining in temporal sequcncc point:,summarizing paUern-rliseriminat ion and po:,it.ionpreference of the "lesion" and "intact" g:roup:, ofmonke~'s for 50-1 rial tcst sc,.:,ion:, riming patt crn-discrimination and rc\"ersal training:. Thi" figmcshows pcrformnncc of control :,ubject. 294.)

Acqul~ition

OL.-L-L-L-L-L-'---'''''''-L-J.....--J

o .20 .40 .60 .50 100P hitting


Acquieition

Rev12rsol 1

~ 1.004-~

C .50ol')

c60~

L~t') .400"c~ .20L

CL O'--'--..l-....l...-....L--L.--L..-J---L----'-----'02040 .60 .50 1.00P hitting 8 wh~n .3 on right

4 iOOti

c .50ocoC .60~

L3

co .400"C~ .20L

0.. 0 '-.L.-.L..-...L.-......l...---L--L.----L----l---L---.Jo 20 .40 .60 .50 1.00P hitting .3 wh12n 8 on right

-+- 1.00~ p--------C .80oI')

R12v12rsol '2.

Acquieition'

~ 1.004-~

c .50ol')

c .60~

L~t') .40

0"c~ .20L

CL

c .50 R~v~r501 10coc 60~

L3

co 400"'c

.20~L

O'--'---~-'--....l.--'----'-----'---'---'----'020 .40 .60 .50 1.00P hitting 8 wh~n .3 on right

4 iOO~

CL 0 L.-L-..-'--..l-....l...-....L---l...--L..-J---L---.Jo .20 .40 .60 .50 1.00P hitting .3 wh~n 8 on right

-+- 1.00'-+-~

C .80oI')

• I

C .60 C .60~ ~

L L,~ ~

I') .40 I') .40

0"' 0"' R~ver6ol '2.' c .20 c .20

~ ~L L

0.. 0 0.. 00 .20 .40 .60 .50 100 0 .20 .40 .60 .80 100

P hitting 8 wh12n .3 on right p hitting 8 wh12n .3 on rightFIG. 4. Performance of control Subject 293 (see

Figure 3 caption).FIG. 5. Performance of control Subject 243 (see

Figure 3 caption).


l·

O~l--~ '--'--J..--J

o .20 .40 .60 .50 1.00P hitting 8 when J on right

+- 100'+~

c .80 Acquisition()

I')

c .60~

L3

J') .40

0"c~ .20.s:::.Q.. a '---'--'--'--'--'--'---'--'--J..__J

02040 .60 .80 1.00P hitting [; when:; on right

C 1.00~

Q. 0 '---'--'--'--'--'--'---'---'-----.Jo .20 .40 .60 .50 1.00P hitting J when [; on right

+- 1.004-~

C .80()

t')

C .60~

.s:::.3t') .40

0-c~ .20.s:::.Q.

c .50 Revereal i c .50 Reversal 1() ()


Reversal 2

Acquisition

~ 100 .--,----,----,----,----,----,----,----,----,---,4-~

c .50oI')

c .60c:ll

£3I') .400'c~ .20..r:.Q... 0 '-'-'-'-'-L.-.L.-.L.-.L.-.L-.J

Q... 0 L.---'----'---'----'----'----'--'-----L----L.---lo 20 .40 .60 .50 1.00P hitting J when 8 on right

~ 1.004-c:ll

c50oI')

o .2040 .60 .50 1.00P hitting 8 wh~n ::; on right

t: 1.00~

c .80ococ60c:ll

£3

co .400'c~20..r:.

c .60~

.s::3I') .40

0--cE .'20

.s::Q.. 0 =--'--'--'--'--'--'-'-'-L---l

o .20 .40 .60 .50 LOOP hitting


difficulties encountered by the experimentalmonkeys during acquisition and reversaloccurred after above-chance responding tothe reinforced cue had been initiated, themajority of points produced by the hip-ammonkey being positioned above thechance-performance line joining the upper-right corner to the lower-left one. In gen-eral, it can be seen that the acquisition ofcriterion by the experimental animals waspreceded by a long period of responsesbased on position bias, most hip-am mon-keys showing a cluster of points at eitherthe lower-left or upper-right corners of thesquare. Even as their performance ap-proached criterion levels, the experimentalmonkeys still retained vestiges of their po-sition bias, committing a majority of theirerrors when the nonreinforced cue appearedin the preferred position in the stimulusarray. In marked contrast, the intact sub-jects showed much less position bias thanthe experimental monkeys. Few if any oftheir points appear at either the lower-leftor upper-right corners of the squares.

It is noteworthy that the intact monkeys'performance was characterized by smoothand progressive increments until criterionwas achieved. Rarely was a point represent-ing a particular day's performance moredistal from the upper-left corner of thesquare than the point summarizing the pre-vious day's performance. However, it wasnot unusual for the experimental animals toshow drastic and sudden decrements (ormore rarely increments in performance).Thus, even after the lax criterion in reversalhad been reached, hip-am monkeys went forlong periods before achieving the strict cri-terion.

In this analysis we had divided the datafrom each problem into 50-trial blocks.However, performance in 50-trial blocks isnot an appropriate method to define phasesin the solution of pattern-discriminationand reversal problems. Intact monkeys wereso efficient in the solution of some phases ofthe experiment that they completed them inless than 50 trials. This made it difficult tocompare the fine grain of the performancestrategies of the two groups. We thereforedivided each subject's data into 10 blocks

containing equal numbers of trials (Vincen-tization). Thus, trial blocks consisted ofdifferent numbers of trials for different ani-mals but accounted for equal proportions,e.g., 10% of the trials required to completethe pattern-discrimination and reversaltasks. The median value of each of the 10points was determined for the hip-am andintact subjects permitting the hip-am andintact monkey data to be represented bytwo curves, each curve composed of 10points. These curves are shown in Figure 10.Note that the hip-am curves almost coin-cide with their appropriate counterpartsfrom the intact group during acquisitionand the first two reversals indicating thatboth groups show essentially identical strat-egies in the solution of these problems.

Figure 4 also indicates that 80% of thepoints which form the first hip-am reversalcurve and 70% of the points of the secondhip-am reversal curve arc above the chancediscrimination diagonal, indicating that forthe experimental monkeys between 70%and 80% of the trials required to achievethe strict reversal criterion occur after sub-jects begin above-chance pattern discrimi-nation. In marked contrast the intact mon-keys required only 50% of trials to reachstrict criterion in Reversal 1 and 2 afterthey began to discriminate above chance. Arelated difference between the two groups isthat the intact monkeys required 50% oftheir reversal trials to extinguish their re-sponses to the previously rewarded cue; thehip-am monkeys, only 20%.

Correlation bettceen Extent of Lesion andBehavioml Effect

As can be seen in Figure 1, all hip-amanimals did not suffer an equal amount ofremoval of the hippocampus althoughamygdaleetomy was total in each. Rankorder of the monkeys on the basis ofamount of sparing of hippocampal tissuecame out to be 290 < 278 < 292 < 277, thelast showing the most sparing. Rank order-ing of the severity of deficit in all of theanalyses matched rrmarkably with this or-dering.

Four additional monkeys were tested onall procedures, two with amygdalectomy

J.


-

-

Reversal 2

o '--'---'-'--J---'_...J.--...l.--...l.--L-.Jo .2040 .60 .30 I no

p Hitting 8 when .3is on the non-preferred siae

AcquisitionI-

I-

100 " , , ,

r')~ 80 ~--J \" ~,lC1j . - ... - .. ~~

~~ /3 ~ 60 ",r')'t - --HIPAM5 '"cr- '- 0 -----Intact \c: 0..4~1r+- .201-o..§

.~ o , , ,o .20 406050 100

p Hitting 8 when ~is on the non-preferred side

100 r-r-r-----,----r-,--r,--;-,-,---, 100 r-_""_=-__'_'.r--c----r.--r----r---r~--,tl tl

~~ ~~.50 no5] c:~u..c: '- " ~ ~3,~60 ~ 3 ~60~ I ~

~~ : ~o0" Q. \ 0"' L.c: .:,0 - '---, c: 0..40

:€ ~ Reversal 1 \:€1::J:- , I~0..6 .20 - I 0..5 20

.~ ~O'-__~~~_--'----'----.J.0 .20 .40 00 .50 LOO

P Hitting .3 when 8is on the non-preferred side

_.~

FIG. 10. Group curves (Vincentized) for the hip-am and intact monkeys durin" original learningand during the first two reversals constructed by joining the 10 points for each group summarizingdiscrimination and position-preference response. (Each point was the median value of discriminationand position-preference responses produced by individual monkeys in each group during one-tenth ofthe trials required to achievc the strict criterion during original learning and Heversal 1 and 2.)

"

and two with hippocampectomy. The amyg-dalectomized subjects behaved as those re-ported in the study by Barrett (1969). Onehippocampal subject performed just as themonkeys in the current study; the other be-haved more like the normal unoperatedgroup and was found to have an incompleteremoval of the hippocampus much likeMonkey 277 but with somewhat more dam-age to the inferotemporal cortex.

DISCUSSION

Analysis by Trials and Errors

In the current study the hip-am monkeysrequired significantly more trials than nor-mal controls to acquire pattern-discrimina-tion and reversal habits. However, this defi-cit was not equally apparent during allphases of these tasks. Rather, at the begin-ning of acquisition the experimental mon-

keys required more trials than the intactsubjects to move from the 50% to the 60%level of performance, but achieved both thelax and strict criterion as readily as theintact controls once they reached the 60%performance level. Further, at the beginningof reversal the hip-am animals extinguishedtheir responses to the nonreinforced cue asrapidly as the controls and thus reached achance level of responding in about thesame number of trials as the intact mon-keys. After the extinction phase of reversal,however, the experimental subjects requiredmore trials than the intact monkeys to im-prove their performance (as measured bysuccessive 10% increments) and to reachthe strict criterion after tile lax' criterionhad been achiend.

These results replicate in general thosereported by Pribram et al. (1969), There isone difference, however, between the hvo


bodies of data. In the earlier study the defi-cits occurred o-nly at the chance level ofperformance, whereas in the current studydeficits were also obtained in achieving andmaintaining criterion performance afterabove-chance discrimination was apparent.

The differences in results between our ex-periment and the earlier one can be ac-counted for by the different procedures usedin the two studies. We used the numerals 3and 8 as discriminanda, whereas the earlierstudy used the numerals 2 and 4. From pre-vious experience we know that the cues usedin the current experiment are more difficultto discriminate than those used earlier. Ouruse of difficult discriminanda appear tohave had the effect of prolonging the hip-am reversal deficit into periods of above-chance levels of responding. The curves forthe first and second reversal support thissupposition: By the second reversal thecurve for the hip-am monkeys more closelyapproximates those hip-am reversal curvespreviously published by Pribram et a!.(1969). With increased experience with thediscriminanda the subjects in the presentstudy progressively were able to overcomethe difficulties presented by the less discrim-inable patterns.

Another difference between our proce-dures and those of the earlier study wasthat for the strict criterion we required ourmonkeys to maintain a 90% level of correctperformance for 3 successive days, whereasthe earlier study terminated discriminationof reversal after only 1 day of 90% correctperformance. Thus, our experiment clearlydemonstrated that experimental subjects, inmarked contrast to the unoperated mon-keys, showed deficits in maintaining criter-ion performance.

One of the purposes of the present studywas to determine whether the reversal defi-cit produced by bilateral amygdalectomyand hippocampectomy is alleviated bypractice with the reversal paradigm. Thus,monkeys were required to complete sevensuccessive reversals after pattern-discrimi-nation training. It is apparent that the ex-perimental animals required progressivelyfewer trials to complete each successive re-versal. The intact subjects also showed im-

provement albeit (since their initial per-formance was better) at a slower rate thanthe hip-am monkeys. Thus, deficits in theability of hip-am animals to reverse inde-pendently of their ability to perform pat-tern discrimination was substantially alle-viated by practice with the reversal para-digm. Nevertheless, the hip-am subjectsstill took more trials to complete the sev-enth reversal than did the intact monkeys.

Our results are in only partial agreementwith those obtained when the amygdalaalone is removed bilaterally. Barrett (19691showed that the development of reversallearning sets is retarded by amygdalec-tomy. Our adoption of the lax criterion forlater reversals may have mitigated againstthe development of the rather larger differ-ence in learning-set performance betweenoperated and control groups found in theearlier study. This interpretation is bortleout by the sudden improvement in reversalperformance by the experimental subjectson the third reversal when the lax criterionwas adopted as the signal for the initiationof reversal.

Analysis by Latenc.ll of Responses

In view of the ambiguity of interpreta-tion allowed when analysis is restricted totrials and errors, other methods of recordingand analysis were instituted in the currentexperiment.

The recording of response latencies wasone such addition. The results showed thathip-am monkeys were slower to respond ingeneral than their controls. :iVIore interest-ing, we obtained a marked difference be-tween groups when the analysis was madein terms of 'whether the previous trial hadbeen rewarded or not (correct or incorrect) .Under the assumption that a longel' re-sponse latency indicates some disruption ofhabitual response, the conclusion can bereached that nonreward is more disruptivefor normal subjects while rewards interferemore with facile performance of hip-ammonkeys. These results can be interpretedin the light of previous findings (Douglas &-Pribram, 1966) to indicate that, in the cur-rent study, error sensitivity ascribed to hip-pocampal function was more severely im-

I'~ .

-!"


.--"••

".

paired than registering reward ascribed tothe amygdala.

Signal-Detection Analysis

Pribram et al. (1969) suggest that thedeficits in discrimination and reversallearning produced by bilateral hippocam-pectomy and amygdalectomy are due totwo separate, but related, factors. First, thelesion reduces the frequency with which thesalient dimensions (pattern) of a stimulusarray is observed, thus prolonging the peri-ods of chance performance apparent at thebeginning of acquisition and after the ex-tinction phase of reversal. Second, a VR2(50% variable ratio) schedule of reinforce-ment is sufficient to maintain the experi-mental subjects' chance performance longafter the intact monkeys have switchedtheir attention to the salient stimulus di-mensIOn.

Clearly, any experiment which proposesto test the validity of the Pribram (1967)hypothesis, as does the present study, mustdevelop means to specify the variableswhich effectively exert control over themonkeys' performance throughout allphases of discrimination and reversal learn-ing. A modification of the analytical tech-niques developed by the theory of signaldetection offered one such possibility. Thisanalysis permitted us to determine the rela-tive amount of control exerted by the di-mensions of the stimuli and by the positionbias on the subjects' discrimination and re-versal performance. Thus, when equal sizeblocks of 50 trials were each represented bya point in an appropriately constructed de-cisional square, it was possible to observethe daily changes in responding to the stim-ulus pattern and position bias as the mon-keys performed discrimination and reversaltasks.

In this analysis the position bias wastaken as the major alternative to stimuluspattern in controlling an animal's discrimi-nation and reversal behavior. 'Most subjectswhile acquiring discrimination and reversalshowed, during at least some periods, analmost exclusive position preference. This isdemonstrated in the analysis by the factthat the points representing their daily per-

formance are positioned at either the up-per-right or lower-left corners of the deci-sional square. The periods of maximum po-sition bias were most apparent at the begin-ning of discrimination learning and afterthe extinction phase of reversal learning.This result, therefore, strongly supports thePribram et al. (1969) assumption that dur-ing periods of chance performance the be-havior of the monkeys comes under the con-trol of noncontingent reinforcement. Thedecision-theoretic analysis allows this to bestated in terms of discrimination and bias:Monkeys tend to revert to position prefer-ences when their behavior is not under thecontrol of a discriminable stimulus pattern.

This analysis further indicated that mon-keys did not necessarily completely 0"81'-come the position bias even after theyachieved above-chance levels of responding.Rather, the fact that the subjects' perform-ance points approach either the upper-rightor the lower-left side of the square as theynear criterion suggests that throughout theacquisition of discrimination and reversalthe position bias vies with the stimulus pat-tern for control of the monkey's behavior.Thus, errors committed throughout the ac-quisition of these tasks occur because theappropriate stimulus dimension fails to holdthe attention of the subjects. This is clearlydemonstrated by the fact that selection ofthe incorrect pattern during periods ofabove-chance performance most frequentlyoccurs at the animal's preferred position.This result also suggests why the hip-ammonkeys, especially during reversal learn-ing, show such sudden and large shifts inperformance. It is difficult to account forthese performance shifts by recourse to anypossible reductions in the subjects' abilityto discriminate between the cues. It is morelikely that these large swings in perform-ance levels occur when animals fail to at-tend to the stimulus dimensions, fall backon their bias, and settle for the VR2 sched-ule of reinforcement with the concomitantreduction in effort required.

This hypothesis suggests that the majoreffect produced by bilateral amygdalectomyand hippocampectomy may be to alter theintensive aspects of attention (Berlyne.


1969). If hip-am subjects expend less effortthan the intact monkeys for obtaining re-ward they will show position preferencesduring a larger proportion of discriminationand reversal trials than do intact controls.Whenever there is a reduction in the incen-tive value accruing to the stimulus dimen-sion-as during the chance reward periodsof the reversal-the experimental monkeysresort to a period of position preference(chance levels of responding) more rapidlythan do the intact controls. The resultingperiod of chance performance is more pro-longed for the experimental animals be-cause they show less incentive than the un-operated monkeys to attend to the stimulusdimension in order to gain access to thehigher density of reinforcement potentiallyprovided by this dimension. Even after thehip-am subjects do eventually begin to re-spond on the basis of the stimulus dimen-sion, presumably because the intensity oftheir attentional state has been sufficientlyaltered, they still retain their increasedtendencies to give up attending and resortto the position bias.

This hypothesis also suggests an explana-tion for the difficulties of the hip-am mon-keys to maintain a criterion level of per-formance once they have already achievedthe lax criterion. The same bias for positionpreference which caused the experimentalanimals to remain at a chance level of re-sponding at the beginning of pattern-dis-crimination learning and at the end of theextinction period of reversal, and to retardtheir achieving criterion, is still apparenteven after criterion has been reached. Thus,the experimental subjects, because they in-consistently attend to the stimulus dimen-

sion, have difficulties in maintaining criter-ion performance during 3 successive testdays.

Signal-detection analyses, therefore, haveclearly demonstrated that the major differ-ence between the hip-am and the intactsubjects is in the intensive rather than theselective dimension of attention-or, to putit more baldly, in the monkeys' motivation.Further, the analysis has shown that thismotivational difference between the experi-mental and unoperated monkeys during dis-crimination and reversal learning is a quan-titative and not a qualitative one; The Vin-centized discrimination and reversal re-sponse-operator-characteristic curves showidentically shaped learning curves for thetwo groups.

REFERENCES

BARRETT, T. W. Studies of the function of theamygdaloid complex in Macaca mulot/a. Neu-ropsychologia, 1969, 7, 1-12.

BERLYNE, D. E. The development of the concept ofattention in psychology. In C. R. E"ans & T.B. Mulholland (Eds.), Attention in neurophys-iology. New York: Appleton-Century-Crofts,1969.

DOUGLAS, R. J., & PRIBRA)!, K. H. Learning andlimbic lesions. Neurop8ychologia, 1966. 4, 197-220.

PRIBRAM, K. H. The limbic systems, efferent con-trol of neural inhibition and beha\'ior. In W.R. Adey & T. Tokizane (Eds.) , Prog re88 inbrain research. Vol. 27. Amsterdam: Elsevier,1967.

PRIBRAM, K. H. DADTA III: An on-line computer-ized system for the experimental analysis ofbehavior. Perceptual and l"Iotor Skills, 1969,29, 599-608.

PRIBRAM, K. H., DOUGLAS, R. J., & PRIBR.\)I. B. J.The nature of nonlimbic learning. Joul'llal ofComparative and Physiological Psychulogy,1969,69, 765-772.

(Recei"ed February 18, 1972)

J .-( .

t1

I;,

1973, vol. 82, no.2, 211-226 decisional analysis of...

Documents