Scribd Logo
Upload

Welcome to Scribd!

  • 2001 UrcuioliDeMarseDeNolf AssessingContributions PDF

    Uploaded by

    sdfoiuhjwer
    11 views13 pages

    Original Title

    2001-UrcuioliDeMarseDeNolf-AssessingContributions.pdf

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    11 views13 pages

    2001 UrcuioliDeMarseDeNolf AssessingContributions PDF

    Uploaded by

    sdfoiuhjwer

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 13
    ex of nnn Fy er: Cre Aaetstae cox oloow a T3 Assessing the Contributions of SO and R-O Associations to Differential-Outcome Matching Through Outcome Reversals Peter J. Urcuioli, Thomas DeMarse, and Karen M. Lionello-DeNolf Purdue University Pigeoos were wained on symbolic mwching with 2 samples, 2 pairs of comparisons, and diferent ‘outcomes forthe cect respons within eath comparison pai, For one group, the 2 samples were also associated with diferent outcomes, whereas for another group they were aot When the responso~ ‘outcome (R-O) relations for one pair were subsequeally revered, the group sined with differenti ‘ample_outcome (8-0) sieciatons was sigaiGicanlly disrupted ints perfonance on bth revere = and ‘onreversed-outoome trials. By contrat, the group tained with just differential R-O associations was RL + OD S25 R2 + (02) S2=R2 +0) SIRS + (01) 31-3 +00 S2—+Ra + (0), S29 Ra + OD) Incongruest $-O st-R1+@D Si RI+ 02) S27 R2 + (02) S25 R2+ (0) 81+ R3 + (023 SI > R34 02) S2=R4+ 0} S2>R6 +O) (Note, Sh and SZ represent red aad green sample stimuli; RI and R2 represent vertical and horiodtal comparison choice respontes; RS and RA represent dot and white choice responses. Comectrespontes following each faple ate ndicated ‘by Whe phi sign (+) Ol and O2 represen the ‘uloome for comect choice. Incorrect choice responses have been omitted for clarity for each tial type is not shown) The groups differed in the relations between the sample stimuli and the outcomes and were Jabeled ina manner similer that of Peterson and Trapold (1982) Group Congruent $-O received waining with differential S-O associations soch that cach sample reliably signaled which out- come was available for a correct comparison response, These birds, then, could anticipate the available outcome on secing the. sample. By contrast, the $-O associations for Group Incongrvent 5-0 were nondiffeential: Each outcome occurred equally often ‘with both samples. This was accomplished by reversing the $-O relations acress the two comparison pairs. Thus, birds in this group, could not determine the available outcome for corect choice during sample presentation Following acquisition, the R-0 relations for one pair of com parison responses were reversed, as shown in the right-mast col tumn of Table 1. The tials involved in outcome reversal are shown, above the space separating the four rain til types for each group. ‘This patil reversal (f, Peterson & TrapoK, 1982) provided atest of the relative contributions of S-O versus R-O associations to performance. Specifically. if comparison response-outcome (R-O} relations exert the primary influence on performance indepen- dently of the nature of the $-O relations, then the disruption of matching accuracy on reversed-obtcome tials should be similar in the two groups. Moreover, the drop in matching seceracy on reversed-ootcome trials for both groups should be much greater than on tials in which the R-O associations remain the same (ines below the space). (On the other hang, if differential S-0 relations, wien present, exert a greater influence on performance, then matching accuracy in Group Congruent $-O should be cistupied on both reversed ‘outcome and same-outcome matching tials, andthe size of that same curcome = reversed outcome (Rodger, 19758, Equation 23). For the incongruent $-O group data, post hoe contrasts showed that the ‘baseline accuracies did not significantly differ, F(3, 9) = 0.01, neither did they differ from that observed on the same-outcome test tials, F13, 9) = 0.06. Accuracy on the reversed-outcome test tials, however, was significanly lower than on the remaining tial types. F3, 9 = 67,06. These statistical decisions imply che following ordering of accuracies in the incongruent S~O group: baseline-same = baseline-reversed = same outcome > reversed 2a LURCUIOL!, DeMARSE, AND LIONELLO DeNOLF 100 80 E & 6 E ‘Groep Congreen 50 & © Same 40 i Reversed, op Tnmguen 0 © Sane 20 (G Reversed. 123 4 5 6 7 8 9 10 Session Fgwe 2 ecenage of cone chie respons by cmpuson fr each group over theft 10 outcome reversal setsions ia Experi | 5 = sample; 0 = outcome Finally, we used one-way ANOVAS to compare matching 8c- curacy across groups on same-oulcome and reversd-outcome ial averaged over the first five test sessions. On reversed outcome tral, the congruent §-O and incongruent 5-0 groups performed st comparably low levels of accuracy, Fl, 6) = 0:00, whereas on same-outcome als, he incongruent $-O group was smoge accurate in its choices than the congruent S-O group, (1, 6) = B02 Discussion [Experiment 1 was virually identical in design co that reported by Peterson and Trapoid (1982), although our pattern of results differed considerably from theirs. First, we fond that a pamtial outcome reversal had an equally large disruptive effect on reversed- and same-outcome matching trials when pigeons had been trained with differential (congraent) $-O relations. By con- trast, Peterson and Trapold (1982) reported greater dismuption on reversed- than on same-outcome trials after congruent taining. Second, we found that a reversal following training with nondif- ferential Gincongrvent) S-O relations produced a drop in accuracy ‘only on the trials in which the R-O relations were reversed. By contrast, Peterson and Trapold (1982) reported that a partial out- ‘come reversal following incongruent training hed litle or no effect fon matching performances on either reversed- or same-outcome sials ‘AS indicated earlier, however, Peterson and Trapold’s (1982) findings may have been compromised by a number of factors. First, they did not counterbalance across comparison sets. Thus, it is unclear if the differences they observed on same- versus reversed-outcome trials following congruent S-O training may Ihave arisen solely for that reason. Second, matching accuracy in Peterson and Trapold’s (1982) incongruent group prior wo it reversal was 20%-30% lower than in the congruent group. Con- sequently, the ineffectiveness of the partial reversal on matching performance in their incongruent group may simply reflect the fect ‘that those binds had not completely leamed the baseline relations ioe to testing (ie, the associations involved in the reversal had ‘ot fully formed). By contrast, n the present experimental birds ‘were trained fo the same high level of basline accuracy, and the ‘utcome reversal in each group was balanced over the two pairs of comparisons. Our results, then, provide a clearer, more interpret able picture of the effect of partial outcome reversals on pigeons’ iffecential-outcome matching Although the comparably sized drop in accuracy in the two groups on reversed-outcome trials suggests that RO relations ‘were dominant, the ational finding of an equally large drop in matching accuracy in Group Congruent S-O on the same-outcome tials indicates that theie differential S-O relations were, in fact, the more influential of the two. On same-outeome tials, the ‘outcomes for correct choice were unchanged relative to baseline Despite this, and despite the fact that the sample-comparison response relations themselves were unchanged, the ability of the same outcome = reversed cateore (Rodger, 19756, Equation 23). By convas, matching ‘couracics onthe sareoutcome test als forthe incongruent §-O sroap were comprable to these of the two baseline Wal pes hich di not differ from each oer, FC, 9) = O14 and 0.53, respectively. Accuracy on the reversd-outco ts trials, tow: ever, was signiticamly lower than an the former thes tril type, FIG.) ~ $05, These sitisicl decisions imply thatthe ordting of matching accurces in the incongnieat'S-O group was a8 follows: buscline-same = basclinereversed = sane outcome > reversed outcome. One-way ANOVAs comparing accuracies scros groupe om sume-ostsome and on evered-outome tras showed no sig cant beweea-roupdifernce on the later FU, 6) = 0.78, but did show signiiandly more accurate same-oteome performance in ‘the incongruent group than in the congruent group, F{1, 6) = 7.42. "To suaistically evaluate whether the patil reversal following exch type of training generted the sae between differences in this experiment a in Experiment 1, we called an accuracy difference forsame- vers seversed-outome tal foreach bird ‘over ts at and second five tet sessions in each experiment. We then compared diffrence scores seross experiments fo the con- sven training condtion and forte incongruent trsning cond tion. Although these comparisons are confounded wit amount ad ‘type of prior matching training, they do provide a rough sense of how silar the pail reversal ofoct were. Ym all eases there $-0 AND R-O ASSOCIATIONS 247 Percent Correct e-Sane C-Rev ISane ev Group-Condition Fire 6 Percentage of cmrect choice responses averaged over the fe five outcome reversal sessions in Expeioteot 2 by each greup on same- ‘outcome and revered-outcoes easy tals (tipped bars). Basline sccaraces on each il type averaged over the las five sessions preceding the oucome reversal ate shown by the voli bas, The asterisk by each group's Tabet (C= congraent sample-outcome (S-O}. | = incongruent 5-0) indioutes the nate of tbe S-O relations created by dhe partial several in Experiment 1 ‘were no significant differences across experiments in same- misias ‘eversed-trial accuracies following either congruent or incongruent ‘waining, all Fs(L, 6) < 401. So, for example, the large perfor- mance difference between the two tiah types during the initial test sessions following incongruent taining was equally large shether Dirds received that type of training initially (in Experiment 1) or after they had undergone a previous partial reversal that had switched their S-O relations from coagruent to incongruent (i.e. their incongruent training was the baseline for Experiment 2). ‘The only possible exception to this was en indication ofa larger accuracy difference during the second five test sessions following ‘congruent raining in Experiment 2 than in Experiment 1, as seen by the greater seperation between the flled-symbel functions over ‘Test Sessions 6-10 in Figure $ than in Figure 2. Aldhough he difference.score analysis described earlier returned a nonsignifi- cant result, F(J, 6) = 4.10, every congraent $-O group bind in ‘Experiment 2 matched more securately on same: than on reversed- ‘outcome trials over those later test sessions, F(. 3) = 32.52. This slower recovery of performance on reversed-outcome trials was ‘ot seen in the congruent S~O group in Experiment 1. is impli- cations are addressed in the General Discussion, Finally, the rightmost pars of bars in Figure 4 show how sample responding changed in reaction to the second outcome reversal Here, the group labels correspond to cach bird's original group assignment in Experiment 1 and the nature of the SO associations resulting ftom the second outcome reversal. When the second ‘reversal re-created differetial S-O associations (left pane), birds pecked at very high rates to the sample (S2) that was now consis- tently followed by food and at very low rates co the sample ($1) that was now consistently followed by no food. By contrast, when the seversal re-created nondifferential S-O associations (right panel), the differential pauem of sample responding evident atthe end of the first reversal (Le., at the end of Experiment 1) dissi- pated, Again, the higher overall rates observed to S2 at the end of the second reversal in this group primasily reflected the reappear ance of the sample-response bias seen in acquisition. I i clear, however, thatthe shift back to nondifferential S-O associations in Group Incongroont $-O produced a large corresponding shift in their sample peck rates. General Discussion ‘Rescorla (1992) and Rescorla and Colwill (1989) have argued ‘hat in differential-outcome discriminations, R-O sssociations have a greater influence on performance thaa S-O associations. By contrast, the conclusion we draw from the results reported here is {quite different: When pigeons’ conditional discriminations involve ‘both differential $-O and differential R-O associations, the former ‘exe stronger control over performance than the Isiter, The main ‘nding supporting this conclusion is that, after waining ofthis sort, reversing the R-O associations on some matching trials adversely affects performance on other matching trials on which the R-O associations are unchanged, Just 5 important, this generalized disruption occurred only ‘when (raining involved differential S-O associations (Group Con- gruent S-O). By contrast, when pigeons learned one-to-many ‘matching with nondifferential $-O associations (Group Incongru- ent $-O), their reversa-test performances were disrupted only on tals directly invoived inthe outcome reversal. Although this latter effect confirms pigeons” sensitivity to R-O associations (Ureviel, & DeMarse. 1997; Ureuioli et al, 1998), the specificity of the effect inthe incongruent SO group clearly shows that the gener. alized dissuption observed in the congruent S-O group cannot be explained simply’ by this or by procedural changes per se that were experienced during the reverse Moreover, these between-group differences were rephiceted with the sarue pigeons that had eatier shown the contrasting pattem of results. In other words, when the pigeons trained with differential '$-O associations in Experiment 1 (Group Congruent $-O) were “retrained” on the same task with nondifferential S-O associations (ia the first outcome reversal), their same-outcome matching performances during the partial eversal of Experiment ? remained act, while their performances om ceversed-outcome trials col- lapsed. By contrast, when pigeons intially waned with nondiffer- ceatial S-O associations (Group Incongruent S-O) were “retrained” ‘with differential $-O associations in Experiment 2, their second arial reversal now produced as much disruption on same- ‘cutcome as on reversed-outcome trials, atleast initially 1s importamy, then, thatthe pattern of test results consistently ‘coincided with, and is understandable interme of, the changes in the prevailing S-O associations caused by the paral reversal, For instance, a shift from differential to nondifferental SO associa tions did, and would be expected to, have an impsct on all ‘matching trials if baseline performances were atleast partly cucd by the anticipated outcomes arising from the differential $-O. assoctations. Conversely, the opposite shift did nor, and would not, be expected to, have an effect on all tits because the nondiffer- ‘ental $-O associations in baseline would not engender a sample- specific amicipatory cue. 248 URCUIOLI, DeMARSE, AND LIONELLO-DENOLE Other sesearchers have reported effects, and offered interpeeta- ‘ions, similar 10 ours. For example, Peterson and Trapold (1982) {found that following one-to-many matching with difercatial S-O associations, reversing R-O associations with ene comparison set disrupted pigeons’ choice accuracy with the other, unchanged set, Likewise, Honig, Matheson, and Dodd (1984) found that reversing the outcomes for one of two concurrent matching tasks decreased E SIE 2 EL Nat Testing RL ve SI and $2 ate diferent discriminative stimuli: Re RI, and R2 are diferent responses: O1 and O2 are

    You might also like