Reacquisition following extinction in appetitive conditioning

Transcription

1 Animal Learning & Behavir 16,4 (4), Reacquisitin fllwing extinctin in appetitive cnditining SEAN T. RCKER and MARK E. BOUTON University fvermnt, Burlingtn, Vermnt n fur experiments utilizing an appetitive cnditining preparatin, reacquisitin f cnditined respnding was fund t ccur bth rapidly and slwly fllwing extinctin. n Experiment 1, acquisitin frespnding t a tne that had been cnditined and extinguished ccurred mre rapidly than acquisitin in either a grup that received equivalent expsure t the fd uncnditined stimulus r a "rest" cntrl grup that received nly expsure t the apparatus in the first tw phases. Hwever, reacquisitin was impaired relative t acquisitin in a "learning-experienced" grup that had previusly received cnditining and extinctin with a different stimulus. Experiments and 3 prduced similar results, but als fund that high respnding during reacquisitin was cnfined t trials that fllwed reinfrced, rather than nnreinfrced, trials. Experiment 4, in which very few initial cnditining trials were used, prduced reacquisitin that was slw cmpared with bth learning-experiencedand rest cntrls. The results are cnsistentwitha rle fr sequentiallearning: Reacquisitin is rapid when animals have learned that reinfrced trials signal ther reinfrced trials. t is ften assumed that reacquisitin f cnditined respnding fllwing extinctin ccurs mre rapidly than acquisitin with a nvel cnditined stimulus (CS; e.g., McAllister & McAllister, 14). The assumptin is based, in large part, upn a number f early studies (Brgden, Lipman, & Culler, 13; Finch & Culler, 135; Frey & Butler, 177; Frey & Rss, 16; Hilgard & Marquis, 135; Hehler, Kirschenbaum, & Lenard, 173; Knrski & Szwejkwska, 150, 15a, 15b; Smith & Grmezan, 165; Szwejkwska, 150). n these experiments, respnding during recnditining f the CS was cmpared with respnding during the riginal cnditining phase in the same animals. Rapid reappearance f the cnditined respnse after extinctin was taken as evidence that the CS retains sme excitatry strength fllwing extinctin, and that this strength transfers psitively t the reacquisitin phase. The CS was seen as starting with a greater assciative strength during reacquisitin than during cnditining. The idea that the CS survives extinctin with sme excitatin is clearly cnsistent with a variety f evidence (e.g., Butn, 1, 13). Hwever, as Butn (16) pinted ut, numerus difficulties exist in interpreting the literature n reacquisitin. The high level frespnding btained during reacquisitin in studies prir t 16 ften culd have been due merely t spntaneus recvery. That is, in many early experiments, reacquisitin began after a time interval that wuld have permitted This research was supprted by Grant BN-0454 frm the Natinal Science Fundatin. Experiments and were part fa thesis submitted by the first authr in partial fulfillment fthe requirements f the master's degree at the University f Vermnt. Crrespndence shuld be directed t M. E. Butn, Department f Psychlgy, University f Vermnt, Burlingtn, VT spntaneus recvery t be present at the utset. t is als pssible that reinstatement, where uncnditined stimulus (US) presentatins in the same cntext as extinctin restre respnding t an extinguished CS (Butn & Blles, 17b; Butn & Peck, 1; Rescrla & Heth, 175), r even renewal, where testing an extinguished CS in a different cntext than that used in extinctin results in increased respnding (Butn & Blles, 17a; Butn & Ricker, 14), culd have been respnsible fr the high level frespnding bserved. Early studies freacquisitin ften used massed cnditining and recnditining trials. With trials spaced clsely tgether, it is pssible that stimulus aftereffects f the US culd have prvided a cntext fr renewal fcnditined respnding (Butn, Rsengard, Achenbach, Peck, & Brks, 13). Early studies had nt cntrlled fr any f these effects (fr argument, see Butn, 16). n cntrast, clear evidence f slw reacquisitin has been btained in the cnditined emtinal respnse (CER) prcedure (Butn, 16; Butn & Swartzentruber, 1). n these experiments, attempts were made t minimize recvery effects that might cmplicate interpretatin frecnditining. The general prcedure invlved giving ne grup cnditining, extinctin, and recnditining phases with ne CS, and cmparing the rate f acquisitin in the final phase with that f a cntrl grup that received cnditining with a nvel CS at the same time. (These were the first experiments t cmpare acquisitin in an extinguished CS with acquisitin f a nvel CS paired with the US at a similar pint in the experiment.) Butn (16) used a "rest" cntrl, which received expsure t the apparatus but n Pavlvian events during the first tw phases. Butn and Swartzentruber (1) used a "learning-experienced" cntrl, which received cnditining and extinctin fa different CS dur- 43 Cpyright 16 Psychnmic Sciety, nc.

2 44 RCKER AND BOUTON ing the first tw phases. The basic finding was that, given enugh extinctin trials, cnditining in the third phase prceeded mre slwly fr the grup that received cnditining with the extinguished CS than fr grups that received a nvel CS. Butn interpreted these findings as indicating that, fllwing extinctin, a memry f that phase exists which interferes with subsequent cnditining. Slw, rather than fast, reacquisitin has als been btained in taste-aversin learning. Fr example, Danguir and Niclaidis (177) fund extremely slw reacquisitin; their results suggested that an aversin t saccharin culd nt be relearned fllwing extinctin. n cntrast, Revusky and Cmbes (17) fund that reacquisitin.did ccur. t was rapid cmpared t acquisitin in a rest cntrl grup but cmparable in rate t acquisitin in a learning-experienced cntrl grup (Revusky & Cmbes, 17, Experiment 1). Mre recent studies (Hart, Burne, & Schachtman, 15) have fund reacquisitin t ccur mre slwly than acquisitin in a learning-experienced grup. The results btained with CER and taste-aversin learning cntrast sharply with data recently reprted in the rabbit nictitating membrane respnse (NMR) paradigm (Napier, Macrae, & Kehe, 1). n a series f fur experiments, Napier et al. addressed the criticisms ffered by Butn (16) cncerning previus research with this methd. Hwever, even after cntrlling fr reinstatement, renewal, and spntaneus recvery, they cnsistently fund reacquisitin t ccur rapidly when cmpared with a rest cntrl. Rapid reacquisitin ccurred whether extinctin cnsisted fcs-alne presentatins r explicitly unpaired CS and US presentatins (ruling ut renewal by US presentatins as a mechanism in the final phase). Rapid reacquisitin als ccurred when the CS was subjected t prcedures that culd have given it an inhibitry value in the secnd phase, such as feature-negative and differential cnditining. The effect persisted despite a large number f extinctin trials (three times the number required t eliminate spntaneus recvery). Reinstatement tests als prduced n evidence that US-alne presentatins reinstated extinguished respnding in the NMR paradigm (Napier et a., 1, Experiment 1). Napier et a!. interpreted their results with a mdel that frmally captured the idea that residual excitatry strength was available t grups that had been thrugh cnditining and extinctin (Kehe, 1). A recent fear-cnditining study has als fund evidence cnsistent with rapid reacquisitin (McAllister & McAllister, 14). Using the acquired drive methd (acquisitin f a hurdle-jumping respnse t escape fear cnditined t cntextual cues), McAllister and McAllister fund that recnditining ffear fllwing 30 h f extinctin expsure t the cntext ccurred mre rapidly than did cnditining in a grup analgus t a rest cntr!. Hwever, the result was nt btained in three earlier experiments that used less extinctin (McAllister & McAllister, 14, Experiments 1-3). t is nt clear why mre extinctin shuld prduce the result when less extinctin shuld be mre likely t d s (e.g., Kehe, 1). Given that reacquisitin can ccur bth rapidly and slwly, ur verall gal was t determine what cnditins prduce the tw results. One pssibility, suggested by Napier et al. (1), is that different cnditining preparatins have intrinsic differences that yield rapid r slw reacquisitin. Our initial gal, then, was t characterize reacquisitin in a preparatin that had nt been studied befre. We chse an appetitive cnditining methd that is widely used in the cnditining literature (e.g., Balleine & Dickinsn, 11; Butn & Ricker, 14; Delamater, 15; Farwell & Ayres, 17; Hall & Channell, 15; Kaye & Mackintsh, 10; Pearce & Redhead, 15): Fd pellets served as the US, and the rat's entries int the magazine t which the pellets were delivered served as the cnditined respnse. The magazineentry respnse is ften described as a Pavlvian CR, but since btaining the pellet is directly cntingent n the respnse, the respnse presumably als has an instrumental cmpnent. ndeed, when magazine apprach is put. n an missin schedule (s that respnding n a trial leads t the missin fa reinfrcer that is therwise delivered; Hlland, 17), the respnse is acquired and maintained (suggesting sme Pavlvian cntrl) but is als weaker than that in a cntrl grup that receives the same CS-US pairings withut the missin cntingency (suggesting an instrumental cmpnent). As an instrumental respnse, magazine entry is prbably best described as a partially reinfrced discriminated perant; many magazine entries g unreinfrced during bth the CS and the intertrial interval. The cmplexity f the magazine-entry respnse will require sme cautin in interpretatin. But there nw appear t be many similarities between instrumental and Pavlvian learning (e.g., Mackintsh, 13; Rescrla, 17). And reacquisitin after instrumental extinctin, like Pavlvian extinctin, is ften assumed t ccur rapidly (e.g., Bullck & Smith, 153). The present experiments btained results suggesting that bth slw and rapid reacquisitin can ccur in the magazine-entry preparatin. They als identified sequentiallearning (e.g., Capaldi, 14) as a variable that cntributes t which fthese results is btained. During cnditining, reinfrced trials may be assciated with ther reinfrced trials. When reinfrced trials are presented fllwing extinctin, they may renew respnding n subsequent trials.. EXPERMENT 1 We designed Experiment 1 as an initial investigatin f reacquisitin in ur appetitive cnditining preparatin. n Experiment 1A, we examined the rate f reacquisitin in a grup that had received cnditining and extinctin with a tne CS (Grup R) relative t cnditining in several cntrl grups. The first fthese was a learningexperienced cntrl similar t the ne used by Butn and Swartzentruber (1); this grup received an iden-

3 REACQUSTON AFTER EXTNCTON 45 tical treatment with a different CS (a light-ff stimulus) during cnditining and extinctin (Grup L). A secnd cntrl received equivalent expsure t the US during the cnditining phase but n pairings f the US with any stimulus (Grup U). The last grup (Grup C) was a rest cntrl that received n stimulus presentatins during the cnditining r extinctin phases (cf. Butn, 16; Napier et a., 1). During a final reacquisitin phase, all grups received tne-fd pairings. n Experiment 1B, we dubled the number fextinctin trials and cmpared a reacquisitin grup with a learning-experienced cntrl. Experiment la Methd Subjects. The subjects were 3 male Wistar rats bred at the University f Yermnt. They were apprximately days ld at the start f the experiment and were individually hused in suspended stainless steel cages in a rm maintained n an 1:6-h light:dark cycle. The experiment was cnducted n cnsecutive days during the light prtin f the cycle. The rats were fd deprived t 0% ftheir free-feeding weights and maintained at that level thrughut the experiment. Apparatus. Tw sets f fur Skinner bxes, hused in sundattenuatin chambers and lcated in tw separate rms, were used. These tw sets f bxes have prvided different cntexts in ther studies, but were nt s used in the present experiment. Each bx in ne set measured 6 x 5 x 1 cm. The frnt, back, and ne side wall were made faluminum; the ceiling and ther side wall were made f clear plastic. The flr cnsisted f tubular steel bars 16 mm in diameter, spaced 3. cm center t center, and munted perpendicular t the frnt wall. On the frnt wall feach bx, cm abve the flr, was a recessed 4 x 4 cm fd cup. A.5 x.5 cm lever prtruded frm the frnt wall 5 cm abve the flr and cm t the right f the fd cup. The subjects were placed in the bxes thrugh a dr in the right wall. Each bx in the secnd set measured 4 x x 1 cm. The frnt and back walls were aluminum, while the ceiling and side walls were clear plastic with vertical black stripes cm wide and.5 cm apart. The flr cnsisted f stainless steel bars 3 mm in diameter, spaced 1.5 cm frm center t center, and munted parallel t the frnt wall. On the frnt wall feach bx, cm abve the flr and centered 3.5 cm frm the right wall, was a recessed 4 X 4 cm fd cup. A 4 x cm lever prtruded frm the frnt wall 5 cm abve the flr and 6 cm t the left fthe fd cup. The subjects were placed in the bxes thrugh the ceiling. n bth sets f bxes, illuminatin was prvided by tw 7.5-W white incandescent bulbs munted n the ceilings f the sundattenuatin chambers, 5 cm abve the flr. Tw CSs were used in this study. One was the 30-sec presentatin fa 3000-Hz tne (0 db re 0,uN/M [AJ) prvided by a generatr wired t identical speakers munted in each chamber 5 cm abve the bx flr. Backgrund nise was 65 db. The ther CS cnsisted f the 30-sec ffset f the huselights, which prduced cmplete darkness. The US cnsistedftw 45-mg fd pellets (Traditinal frmula, P. 1. Nyes, Lancaster, NH) delivered 0. sec apart. Magazine entries were detected by phtcells munted within the magazines, just behind the plane fthe wall fthe Skinner bxes. The apparatus was cntrlled by cmputer equipment lcated in an adjacent rm. Prcedure. Pretraining. Each subject first received 30 min f expsure t the bx it was assigned t fr the duratin fthe experiment. Fd cups were baited with 4 fd pellets prir t the start feach expsure sessin. On the fllwing day, subjects received ne 30-min sessin fmagazine training. Fd cups were again baited with 4 pellets each. During the sessin, the subjects were trained t apprach and eat frm the fd cup upn activatin fthe feeder mechanism. Apprximately 5 pellets were delivered during each sessin. Cnditining. The subjects then received 6 days fcnditining. On each day, there was ne 0-min sessin. Fr Grups Rand L, this sessin cnsisted f eight pairings f the 30-sec CS with the fd US, with a mean intertrial interval (T!) f min (shrtest T! = min). Thrughut, the ffset fthe CS cincided with the nset f the US. Fr Grup R the CS was the tne (T), while fr Grup L the CS was the light ff (L). Grup U received the same eight US presentatins but n CS. Grup C was merely placed in the apparatus fr the entire 0 min; n stimuli were presented. On Day, the first sessin fr each grup started with tw nnreinfrced presentatins each ft and L alne in the sequence TLTL, with a mean T! f min. Extinctin. Fllwing cnditining, the rats received sessins fextinctin. All sessins were 0 min in duratin and ccurred n cnsecutive days. Fr Grups Rand L, the 30-sec CS was presented eight times alne with a mean T! f min during each sessin. The CSs presented were the same as thse used during cnditining. Grups U and C were bth placed in the apparatus fr the 0-min sessin, but n stimuli were presented. N fd was presented during this phase. Reacquisitin. Fllwing extinctin, the rats received fur mre sessins f cnditining. All fur grups received ne 0 min sessin a day in which the 30-sec tne CS was paired with fd eight times, with a mean T f min. Data analysis. Cnditined respnding was measured by means f an elevatin scre f the frm e = c - p, where c represents the number fmagazine entries made during the 30-sec CS and p represents the number fmagazine entries made during the 30 sec immediately preceding the CS. Elevatin scres were calculated fr each trial and then cnverted t fur-trial blck averages prir t analysis. These scres were then analyzed by means fa mixed design analysis f variance (ANOYA), with grup as a between subjects factr and blck as a repeated measure. An identical ANOYA was used t analyze respnding during the pre-cs perid (prescres). A rejectin criterin fp <.05 was used thrughut. Pst hc cmparisns were cnducted using t tests with a pled errr term and a Bnferrni adjustment fr Type errr rate. Results Cnditining prceeded uneventfully, with Grups R and L reaching asymptte by Day 5. The data are summarized n the left side f Figure 1. A grup X blck ANaYA n the cnditining data revealed nly a significant main effect fblck [F(,154) = 7.11]; respnding tended t increase ver blcks. Neither the main effect f grup [F(l, 14) = 1.] nr the grup X blck interactin [F(11,154) = 1.16] apprached significance. Extinctin prceeded uneventfully as well (right prtin ffigure 1). A grup X blck ANaYA carried ut n the data revealed nly a significant main effect fblck [F(l,66) = 15.43]. Neither the main effect fgrup nr the grup X blck interactin were significant (Fs < 1). Grup cmparisns during reacquisitin. The data fr each grup during the reacquisitin phase are presented in the tp panel f Figure. All grups acquired cnditined respnding at this time. Grups Rand L appeared t differ, with Grup R respnding smewhat mre n early trial blcks and smewhat less n later nes. Grups U and C were bth smewhat slwer than

4 46 RCKER AND BOVTON 6 5 GrupR Grup L 4 ~ 0 u (/) c:: 3 0 ';j :> '" 0 ~ 0+---"-""---"-""--"T-""--"T-..,.---:;:'--.!j'---r-D--rLJ-r-"'--r--;'" Trial Blcks Figure 1. Mean elevatin scres during cnditining and extinctin in Experiment A. Grups Rand L in acquiring respnding, with Grup V being slwest. A grup X blck ANOVA revealed a significant main effect f blck [F(7, 16) = 1.] and a significant grup X blck interactin [F(1,16) =.]. The main effect fgrup was nt significant [F(3,) =.7]. The interactin suggests that grup differences existed nly n certain trial blcks. Therefre, a series f five unplanned cmparisns were cnducted with an adjustment fr errr rate (which set a =.01). These cmparisns cmpared Grup R with each f the cntrl grups separately n the blcks where the largest grup differences appeared, and similarly cmpared Grup L with each f the ther cntrls. Grup R did nt differ significantly frm Grup L n Blck 7 [t(7) = -.11], where the largest difference between these tw grups appeared. Hwever, Grup R respnded significantly mre than Grup V n Blck 4 [t(7) = 3.13] and mre than Grup C n Blck 3 [t(7) =.77]. Grup L als respndedmre than Grup V [t(7) =.] and Grup C [t(7) =.6] n Blck 5. Althugh Grups Rand L failed t differ in the previus analysis, previus CER and taste-aversin experiments that fund slw reacquisitin had cmpared analgus grups by themselves. Therefre, an additinal analysis was cnducted using nly these tw grups. A grup X blck ANOVA revealed a significant main effect fblck [F(7,) = 6.] and a significant grup X blck interactin [F(7,) =.55]. The main effect f grup was nt significant (F < 1). The interactin suggests that acquisitin prceeded differently in Grups R and L. Hwever, the largest difference between the grups (Blck 7) was nt reliable [t(4) = -.00]. The ANOVA carried ut n prescres revealed n significant effects [largest F(7, 16) = 1.7]. Mean prescres fr the reacquisitin phase were 1.71, 1.33, 1.3, and 1. fr Grups R, L, V, and C, respectively. Cnditiningversus recnditining. The upper panel f Figure 3 shws Grup R's perfrmance in bth the riginal cnditining phase (Phase 1) and during reacquisitin (Phase 3). A within-subject cmparisn fperfrmance in these phases can cmplement ur betweensubjects analysis f reacquisitin, althugh it des cnfund cnditining histry fthe CS with ther variables, such as prir experience with the reinfrcer, the apparatus, and the fd-deprivatin schedule. As Figure 3 suggests, respnding was clearly greater during recnditining. A within-subject phase X blck ANOVA revealed a significant main effect f phase [F(l,7) = 16.73]. The ANOVA als revealed a significant main effect fblck [F(7,4) = 4.6] anda significantphase X blck interactin [F(7,4) = 5.41]. A similar ANOVA n prescres revealed a significant main effect fblck [F(7,4) =.53]. Prescres increased slightly ver blcks, regardless fphase. The main effect f phase apprached significance, but was nt reliable [F(,7) = 5.15, p =.06], and the phase X blck interactin was nt significant [F(7,4) = 1.00]. Mean prescres were 1.34 and 1.71 during cnditining and reacquisitin, respectively. Experiment B Methd Subjects and Apparatus. The subjects were 16 male Wistar rats fthe same stck as thse used in Experiment A. They were apprximately 160 days ld at the start f the experiment, and were hused and maintained as in Experiment A. One subject in Grup L became sick during the curse fthe experiment and was remved, leaving 7 subjects in that grup. The apparatus was the same as that used in Experiment A. Prcedure. Pretraining. Bx expsure and magazine training were carried ut as in Experiment A. Apprximately pellets were delivered during each magazine training sessin. Cnditining. Fllwing magazine training, the subjects received eight sessins f cnditining. Each rat received tw

5 REACQUSTON AFTER EXTNCTON 47 Experiment A O+---r--r----,;----,.---r--,---,----, 7 Experiment B e-- ---l:j Trial Blcks GrupR Grup L Grup U Grupe Grup R Grup L ~-r--r--r-.--r--r-!;:=;::=;::=;::=;==;"' Trial Blcks Grup R; the mean elevatin scres fr the phase were 0.63 and 0.33 fr Grups Land R, respectively. A grup X blck ANOVA revealed a significant main effect f grup [F(1,13) = 4.7] and a significant main effect f blck [F(3,507) = 6.4]. The interactin was nt significant (F < 1). Althugh Grup L appeared t extinguish mre slwly, this difference had disappeared by the last fur sessins, where Grups Land R had mean elevatin scres f 0.14 and 0.17, respectively. Fur unplanned cmparisns (at a =.015) between the tw grups n the first blck feach fthe last fur sessins revealed n significant differences [ts(405) < 1]. Grup cmparisn during reacquisitin. The fcal data frm reacquisitin are presented in the lwer panel ffigure. During this phase, bth grups received cnditining with the tne. The grups initially did nt differ, but Grup R shwed less respnding than Grup L ver blcks. This descriptin was cnfirmed by statistical analysis. A grup X blck ANOVA revealed a significant main effect fblck [F(ll,143) = 11.3] and a significant grup X blck interactin [F(11,143) = 3.33]. The main effect f grup was nt significant [F(1,13) =.33]. Cmparisns (at a =.0167) between grups n Blcks 7, 11, and 1, where the largest grup Experiment A Figure. Mean elevatin scres during reacquisitin in Experiments la (upper panel) and 1B (lwer panel). 0-min sessins n each day. Each sessin cntained eight pairings fthe 30-sec CS with the fd US, with a mean T f min. Fr Grup R, the CS was T; fr Grup L, it was L. On Day 1, the first sessin fr each grup started with tw presentatins each f T and L alne in the sequence TLTL, with a mean T f 1 min. Extinctin. Fllwing cnditining, the rats received 0 sessins fextinctin. As in cnditining, there were tw 0-min sessins n each day. n each sessin, the 30-sec CS was presented eight times alne with a mean T f min. A ttal f 160 extinctin trials were given. The CSs presented were the same as thse used during cnditining. N fd was presented during this phase. Reacquisitin. Fllwing extinctin, the rats received six mre sessins f cnditining. Bth grups received tw 0-min sessins a day in which the 30-sec tne CS was paired with fd eight times, with a mean T f min. Data analysis was carried ut as in Experiment. Results Cnditining prceeded uneventfully; during the final fur-trial blck facquisitin, Grups Rand L had mean elevatin scres f5. and 5.14, respectively. A grup X blck ANOVA revealed nly a significant main effect f blck [F(15,15) =.55]. Neither the main effect f grup [F(l,13) =.56] nr the interactin [F(15,15) = 1.5] were significant. During extinctin, respnding appeared t extinguish mre slwly in Grup L than in 0 Phase Phase 3 +--,--.---r----,-=r==;:==::;==;_-' 7 Cl6 u "'5 C.g 4 ~3 iil Experiment B Trial Blcks 1 Phase Phase ,--.--,-r--r-..,...--r'==r=r==;==;==;---' Trial Blcks Figure 3. A cmparisn f mean elevatin scres fr Grup R during cnditining and reacquisitin in Experiments la (upper panel) and 1B (lwer panel). 7

6 4 RCKER AND BOUTON differences appeared, revealed less respnding in Grup R than in Grup L n Blck 7 [t(35) = -.6] and Blck [t(35) = -.53] but nt n Blck 1 [t(35) = -.50]. The ANOVA n prescres revealed nly a significant main effect f blck [F(11,143) = 3.67]; respnding during the pre-cs perid increased smewhat and then decreased ver blcks. All ther effects were nt significant (Fs < 1). Mean prescres fr the reacquisitin phase were 1.7 and 0.0 fr Grups Rand L, respectively. Cnditining versus recnditining. The lwer panel f Figure 3 presents a cmparisn f Grup R's perfrmance during reacquisitin and the riginal cnditining phase. A phase X blck ANOVA revealed a significant main effect fphase [F(l,7) = 1.0], a significant main effect f blck [F(11,77) = 7.0], and a significant phase X blck interactin [F(11,77) =.46]. Again, cnditining ccurred mre rapidly in the reacquisitin phase than in the riginal cnditining phase. A similar ANOVA n prescres revealed n significant effects [largest F(l,77) = 1.1]. Discussin The results f Experiments la and B indicate that reacquisitin in the appetitive methd can be rapid when cmpared with "learning-naive" cntrls (Grups U and C in Experiment la) and with the riginal cnditining phase. Hwever, in Experiment 1B, perfrmance during reacquisitin was impaired relative t that in Grup L, a learning-experienced cntrl (a related grup X sessin interactin was btained in Experiment la). Grup L itselfappeared t shw a "learning-t-earn" effect analgus t that previusly bserved in NMR cnditining (e.g., Kehe & Hlt, 14; Kehe, Mrrw, & Hlt, 14). That is, previus cnditining (and extinctin) with the light-ff CS facilitated cnditining with the tne relative t learning-naive cntrls. Smewhat paradxically, a learning-t-learn effect was less evident in a reacquisitin grup that had received all three phases with the same CS. That result might implicate an interference mechanism in additin t the learning-t-earn mechanism. Althugh the fact that Grup R was lwer in respnding than Grup L later in the phase indicates sme interference by prir cnditining and extinctin with the same CS, it is wrth nting that, in CER, the interference effect is bserved starting early in reacquisitin. nstead facquiring respnding mre slwly, Grup R appeared t apprach a lwer asymptte (see especially Experiment B). Theries that predict slw reacquisitin wuld actually expect grup differences t ccur early, rather than late, in the phase (Butn, 13; Pearce & Hall, 10; Wagner, 11). EXPERMENT n Experiment, we cnducted reacquisitin with a partial reinfrcement prcedure that intermixed nnreinfrced trials amng the reinfrced trials. This srt fprcedure was used by Butn (16) and Butn and Swartzentruber (1), wh btained slw reacquisitin. Accrding t Butn's (13) retrieval view, presentatin f nnreinfrced trials during reacquisitin might help retrieve a memry fextinctin fr a reacquisitin grup and thus cause reacquisitin t ccur slwly. This analysis resembles Capaldi's (e.g., 14) welldevelped sequential thery. t implies that the utcme fa given trial can serve as a signal (r retrieval cue) fr the utcme fthe next trial. Thus, during cnditining, reinfrced trials might cme t signal that the next trial will be reinfrced, while during extinctin nnreinfrced trials might cme t signal that the next trial will be nnreinfrced. This view implies a particular pattern f respnding during reacquisitin with partial reinfrcement. Fllwing cnditining and extinctin, respnding n a trial after a reinfrced trial shuld be high, while respnding fllwing a nnreinfrced trial shuld be lw. Since reacquisitin was cnducted with a partial reinfrcement prcedure in this experiment, we culd test this predictin. A retrieval view emphasizing the rle fprevius trials predicts mre respnding after reinfrced than after nnreinfrced trials. We used a learning-naive cntrl similar t Grup U in Experiment la (as ppsed t Grup C) s that the cntrl wuld have had equal expsure t the US. This grup als received cnditining in the third phase with a partial reinfrcement prcedure. Here ne might als expect higher respnding n trials fllwing reinfrced trials, ifnly because assciative strength might increase after each CS-US pairing. Clearly, any sequential learning effect in a reacquisitin grup wuld have t be greater than that bserved in Grup U. We als perfrmed a cntext switch fllwing US expsure in rder t reduce pssible blcking by cntext cnditining (e.g., Randich, 11). There is sme evidence t suggest that this ccurred in Experiment 1A, because in the final phase Grup U appeared t acquire cnditined respnding mre slwly than Grup C, the cntrl that had nt received US expsure in Phase. Methd Subjects The subjects were 16 male Wistar rats frm the same stck as in the previus experiments, and were maintained and hused as befre. They were apprximately 0 days ld at the start f the experiment. Apparatus The apparatus was the same as in Experiments A and B, except that a dish cntaining ml f'heinz distilled white vinegar (H. 1. Heinz C., Pittsburgh) was placed in each sund-attenuatin chamber f the first set f bxes in rder t prvide a distinctive scent cue; a dish cntaining g f Vick's Vaprub (Richardsn Vicks, nc., Sheltn, CT) was similarly placed in each sundattenuatin chamber fthe secnd set fbxes. n the present experiment, the tw sets f bxes, which prvided the different cntexts, were fully cunterbalanced. Prcedure Pretraining. Bx expsure and magazine training were carried ut as in Experiment, with the exceptin that each subject received ne sessin in Cntext A (the cnditining cntext) and ne in Cntext B (the US expsure cntext). Sessins in different

7 REACQUSTON AFTER EXTNCTON 4 cntexts were cnducted n the same day, althugh bx expsure and magazine training were cnducted n separate days, as befre. Apprximately 0 pellets were delivered during each fthe magazine training sessins. Cnditining. The subjects then received days fcnditining. On each day, there were tw 0-min sessins, ne in Cntext A and ne in Cntext B. Fr Grup R, sessins in Cntext A cnsisted f eight pairings f the 30-sec tne CS with the fd US, with a mean T f min. Sessins in Cntext B fr this grup cnsisted nly f expsure t the apparatus. Fr Grup U, sessins in Cntext A cnsisted f expsure t the apparatus, while sessins in Cntext B cnsisted feight presentatins fthe US alne, with a mean T! f min. As in the preceding experiments, n Day all sessins started with tw presentatins fthe tne alne, with a mean T! f min, s that cnditining began with a nnnvel CS in all grups. Grup R received sessins in the rder ABBA, while Grup U received them in the rder BAAB. Extinctin. Fllwing cnditining, the rats received 0 sessins fextinctin in Cntext A. There were tw 0-min sessins in Cntext A each day. Fr Grup R, each sessin cnsisted f eight presentatins f the tne alne, with a mean T! f min. Fr Grup U, these sessins cnsisted nly fexpsure t the apparatus. N fd was presented during this phase. Reacquisitin. Fllwing extinctin, the rats received 3 mre days (six sessins) fcnditining in Cntext A. Bth grups received tw 0-min sessins a day in which the 30-sec tne CS was paired with fd fur times and presented alne fur times, with a mean T f min. Tw trial sequences were used ver the six sessins. n ne sequence, reinfrced (R) and nnreinfrced (N) trials alternated (i.e., RNRNRNRN); in the ther sequence, trials semialternated (i.e., RNRRNNRN). The rder f presentatin f the tw sequences was ASSAAS, where A refers t alternating and S refers t semialternating. Results By the end f cnditining, rats in Grup R were respnding well t the CS; the mean elevatin scre in the last fur-trial blck was 7.7. A ne-way repeated measures ANOVA was cnducted fr Grup R, and it revealed a significant main effect f fur-trial blck [F(l5,5) = 4.1]. Extinctin als prceeded withut incident, with rats in Grup R shwing little r n spntaneus recvery by Day 4. By the last sessin f extinctin, Grups Rand U had mean elevatin scres f 0.00 and -1., respectively. A trial-blck ANOVA revealed a significant main effect f blck [F(3,73) = 7.71]. Reacquisitin data are presented in Figure 4. During reacquisitin, Grups Rand U initially did nt differ, but ver trials, R respnded mre than U. A grup X blck ANOVA carried ut n elevatin scres revealed a significant main effect fgrup [F(l, 14) = 7.0], a significant main effect fblck [F(l,154) =.], and a significant grup X blck interactin [F(,154) = 3.45]. Althugh respnding verall was higher in Grup R, the grups did nt differ initially and tward the end freacquisitin. This was cnfirmed with fur pst hc cmparisns (at ex =.015). On Blck 1, Grups Rand U did nt differ [t(44) = 0.53]. Hwever, n Blcks 6 and, they did [ts(44) ~ 3.44]. By Blck 1, the grups n lnger differed [t(44) = 1.6]. The ANOVA n prescres revealed nly a significant main effect f blck [F(l, 154) = 3.13]. Neither the main effect f grup [F(1,14) < 1] nr the interactin [F(l,154) = 1.] was significant. Prescres increased smewhat and then decreased ver blcks, but did nt differ between grups. Mean prescres were 1.64 and 1.55 fr Grups Rand U, respectively. Figure 5 presents the respnding by Grups Rand U n trials that fllwed reinfrced versus nnreinfrced trials. A grup X trial type X sessin ANOVA was cnducted n the elevatin scres. The ANOVA revealed a significant main effect fgrup [F(l,14) = 7.0], a significant main effect fsessin [F(5,70) = 1.5], and a significant grup X sessin interactin [F(5,70) = 3.3] (these effects crrespnd t the preceding analysis f reacquisitin). n additin, there was a significant main effect ftrial type [F(l,14) = 5.4] and, mst imprtantly, a significant trial type X grup interactin [F(,14) = 17.6]. While respnding verall was higher fllwing reinfrced than fllwing nnreinfrced trials, this effect was cnfined t Grup R. The sessin X trial type interactin was significant [F(5,70) =.70], but mre imprtantly, the sessin X trial type X grup interactin was als significant [F(5,70) =.37]. nitially, fr Grup R nly, respnding was higher fllwing reinfrced trials than fllwing nnreinfrced trials, but, ver sessins, this difference disappeared. This was cnfirmed with simple cmparisns (at ex =.015). n Grup R there were significant differences between trial types during the first tw sessins [ts() > 4.44], but in Grup U there were n such differences (ts < 1). A similar analysis was cnducted n prescres. The ANOVA revealed significant main effects nly f sessin [F(5,70) = 5.63J and trial type [F(,14) = 5.77J. Respnding during the preperid was smewhat higher fllwing reinfrced trials than fllwing nnreinfrced trials, where the respective means were 1.74 and 1.54 fr Grup Rand 1.1 and 1. fr Grup U. The fact that this effect did nt vary with grup [grup X trial type interactin, F(l,14) = 1.15] is interesting, given that the <U 0 u '" C 5.g ~ 4 > <U ~ Grup R Grup U Trial Blcks 11 1 Figure 4. Mean elevatin scres during reacquisitin in Experiment. 6, "... " ", -. 'e,"'--.,,

8 430 RCKER AND BOUTON Grup R(r) Grup R(n) ---e-- Grup D(r) Grup D(n) this idea, it des appear that the rapid reacquisitin bserved in the present methd is initially restricted t trials that fllw reinfrced trials, as a retrieval view r sequential thery predicts. EXPERMENT 3 -t----r---,---,, , Sessin Figure 5. Mean elevatin scres during reacquisitin in Experiment. Here data fr Grups Rand U are brken dwn int trials fuwing reinfrced trials (r) and trials fuwing nnreinfrced trials (n). elevatin scres did vary with grup n these trials. This suggests that the aftereffects f a reinfrced trial were experienced in bth grups (higher prescres), but nly in Grup R did a reinfrced trial signal reinfrcement n the next trial. Discussin Reacquisitin again ccurred rapidly cmpared with a learning-naive cntrl, even when a partial, rather than cntinuus, reinfrcement prcedure was used. Hwever, clser analysis revealed that high respnding in Grup R was initially cnfined t trials fllwing reinfrced, rather than nnreinfrced, trials. This result is cnsistent with a retrieval view, especially ne emphasizing the retrieval rle findividual trials (e.g., Capaldi, 14). The fact that these differences in respnding did nt last beynd the first tw sessins may be due t the change in cntingencies between reinfrced and nnreinfrced trials and the trial types that fllwed them. That is, during reacquisitin, nnreinfrced trials nw tended t signal reinfrced trials and reinfrced trials tended t signal nnreinfrced trials, reversing what might have been learned during extinctin and cnditining. Respnding fllwing nnreinfrced trials was nt lwer in Grup R than in Grup U. This culd present a prblem fr the retrieval view. f, fr Grup R, nnreinfrced trials were signaling nnreinfrcement n the next trial, then respnding fllwing nnreinfrced trials shuld have been lwer in this grup than in Grup U, fr whm n such signaling shuld ccur. One pssibility is that ifretrieval fan extinctin memry is mre susceptible t retrieval failure (Butn, 13), it may be mre difficult t retrieve the memry f nnreinfrcement than freinfrcement in Grup R. Whatever the merit f 6 One purpse fexperiment 3 was t replicate the sequential learning effect bserved in Experiment. Anther purpse was t further explre the learning-tlearn effect that was bserved in Experiment 1A. n that experiment, cnditining and extinctin with the lightffcs facilitated cnditining fthe tne CS in the final phase. This srt f result is captured by the mdel presented by Kehe (1), which states that cnditining and extinctin with any stimulus results in residual excitatin which may transfer t recnditining r cnditining f anther stimulus. An alternative pssibility, hwever, is that the sequential learning effect identified in Experiment may cntribute t this result as well. Fllwing cnditining, the ccurrence fthe US might ptentially serve as a general signal freinfrcement, regardless f the identity f the CS. f s, we might likewise bserve enhanced respnding fllwing reinfrced trials in a learning-experienced cntrl grup. n Experiment 3, we therefre cmpared respnding n trials after reinfrced and nnreinfrced trials in Phase 3 in a learning-experienced cntrl (Grup L), a reacquisitin grup, and a rest cntrl. Methd Subjects The subjects were 4 female Wistar rats btained frm Charles River, Canada. (The rats in Experiments A-B and were derived frm this stck.) They were apprximately 0 days ld at the start fthe experiment and were hused and maintained as befre. During the curse f the experiment, 3 fthe animals became sick ( frm each grup) and were remved, leaving 7 animals in each grup. Apparatus The apparatus was the same as that in Experiments and, with the exceptin that the tw Nyes fd pellets that served as the US were delivered 0.4 sec apart. Prcedure Pretraining. Bx expsure and magazine training were carried ut as in Experiments and. Apprximately 1 pellets were delivered during each fthe magazine training sessins. Cnditining. The subjects received eight sessins f cnditining, ne per day. Fr Grup R, the sessins cnsisted feight pairings fthe 30-sec tne with the (d US, with a mean T f min. Fr Grup L, the sessins cnsisted feight pairings f the 30-sec light ff with the US, with a mean T f min. Fr Grup C, these sessins cnsisted nly fexpsure t the apparatus; n stimuli were presented. Fr all grups, the first sessin started with tw nnreinfrced presentatins each fthe tne and light ff, in the rder TLTL. Extinctin. Fllwing cnditining, the rats received 0 sessins f extinctin, per day. Fr Grup R, these sessins cnsisted f eight presentatins f the tne alne with a mean T f min. Fr Grup L, the sessins cnsisted feight presentatins

9 REACQUSTON AFTER EXTNCTON 431 fthe lightffalne with a mean T! f min. Grup C received nly expsure t the apparatus during these sessins. N fd was presented during this phase. Reacquisitin. Fllwing extinctin, the rats received three mre sessins fcnditining, again ne per day. Fr all grups, the tne was paired with fd fur times and was presented alne fur times with a mean T! f min during these sessins. n each sessin, reinfrced and nnreinfrced trials alternated (i.e., RNRNRNRN). Results Cnditining prceeded uneventfully, with Grups R and L reaching mean elevatin scres f 5.3 and 5.75, respectively, during the final fur-trial blck. A grup X fur-trial blck ANOVA revealed a significant main effect nly f blck [F(15,) ==.0]. Neither the main effect f grup nr the interactin were significant [F(1,14) == 1.7 and F(15,1O) == 0.3, respectively]. Extinctin prceeded uneventfully as well, with rats shwing little r n respnding by Sessin 7. A grup X blck ANOVA revealed a significant main effect nly f blck [F(3,46) == 1.5]. Neither the main effect f grup nr the interactin were significant (Fs < ). The data frm reacquisitin are presented in Figure 6. During reacquisitin, Grup R nce again acquired respnding mre rapidly than Grup C but mre slwly than Grup L. A grup X blck ANOVA revealed a main effect f grup [F(, 1) == 5.17] and a main effect f blck [F(5,0) ==.57]. The interactin was nt significant [F(,0) == 1.15]. The main effect fgrup was explred with unplanned cmparisns (at a ==.017). These cmparisns revealed that Grup R shwed less verall respnding than Grup L [t(1) == -3.1], that Grup R shwed mre respnding than Grup C [t(1) == 4.64], and that Grup L shwed mre respnding than Grup C [t(1) == 7.3]. Althugh respnding in Grup L seems unusually high at the utset, the first trial elevatin scre in this grup was nly -0.. The ANOVA n prescres revealed a significant main effect nly f blck [F(5,0) == 3.]. Prescres increased slightly and then decreased ver blcks. Neither the main effect fgrup nr the interactin were significant[f(,1) == 1.4 andf(,0) == 0.6, respectively]. Mean prescres fr Grups R, L, and C were 1.6,.1, and.15, respectively. Figure 7 presents respnding f each grup n the trials fllwing reinfrced versus nnreinfrced trials. A grup X trial type X sessin ANOVA revealed significant main effects fgrup and sessin as in the previus analysis [F(,1) == 5.17 and F(,36) == 11.06, respectively]. The interactin f these factrs was nt significant [F(4,36) == 1.3]. There was a significant main effect f trial type [F(l,1) == 11.46], indicating that respnding was higher verall n trials fllwing reinfrced trials. Nne fthe interactins invlving trial type were significant (Fs ~ 1.). Because we had an a priri interest in which f the grups shwed differentiatin between trial types, trial type X sessin ANOVAs were cnducted fr each grup 7 ~ 6 '" 5.~ i) 4 G3 3 0,,, ' ", /).-- Grup R Grup L Grupe,..(]-----', " ~~... -i---,---,---,---..,.---..,.----, Trial Blcks Figure 6. Mean elevatin scres during reacquisitin in Experiment3. separately, using the errr terms frm the verall analysis. Fr Grup R, this analysis revealed bth main effects f sessin [F(,36) ==.60] and trial type [F(1,1) == 7.14]. Fr Grups Land C, hwever, there were n signi ficant effects in these analyses [largest F(1,1) == 3.]. Further cntrasts in Grup L n Sessins and revealed n differences in respnding fllwing reinfrced versus nnreinfrced trials [ts(4) ~ 1.36]. Thus, Grup R, but neither Grup L nr Grup C, gave evidence fthe sequential learning effect. The idea that facilitated respnding in Grup L was mre than a result f sequential learning was further supprted by ANOVAs cmparing the grups' respnding after reinfrced and nnreinfrced trials. Respnding was greater in Grup L than in Grup C fllwing bthreinfrced [t(1) == 3.15] and nnreinfrced [t(1) >.74] trials; facilitatin was nt dependent n trial type. n cntrast, there was mre respnding in Grup R than in Grup C fllwing reinfrced trials [t(1) ==.14,.05 <p <.05] butnt fllwing nnreinfrced trials [t( 1) == 1.34]. The facilitatin effect fr Grup R, but nt Grup L, was thus demnstrably cnfined t trials fllwing reinfrced trials. A grup X trial type X sessin ANOVA n the prescres revealed a significant main effect f sessin [F(,36) == 4.75]. Hwever, the three-way interactin was als significant [F(4,36) ==.0]. N ther effects were significant in this analysis [largest F(4,36) == 1.56]. Fllwing reinfrced trials, mean prescres fr Grups R, L, and C were 1.64,.63, and 1.75, respectively; fllwing nnreinfrced trials, they were.07, 3.1, and'.15, respectively. The three-way interactin suggests that differences in trial types existed fr sme grups nly n sme sessins. Therefre, separate trial type X sessin ANOVAs were carried ut fr each grup separately, as in the analysis felevatin scres. The nly significant effects were a main effect fsessin fr Grup L

10 43 RCKER AND BOUTON '" 5.~ > 4.,!:l. >Ll 3 Grup R (r) --i:}- Grup R (n) ---e-- Grup L (r) Grup L (n)... Grup C (r)..._...f:s...- Grup C (n) O+----""T""----, , 3 Sessin Figure 7. Mean elevatin scres during reacquisitin in Experiment 3. Here data are brken dwn int trials fllwing reinfrced trials (r) and trials fllwing nnreinfrced trials (n). [F(,36) = 4.64] and a trial type X sessin interactin fr Grup C [F(,36) = 4.5]. The interactin in Grup C was further explred using simple cmparisns (at a =.017). These revealed nly a small nnsignificant effect ftrial type n Sessin 3 [t(31) = -.05,p =.05]. Therefre, it can be cncluded that there were n prblematic differences in prescres. Discussin As in the previus experiments, reacquisitin again ccurred rapidly as cmpared with the learning-naive cntrl (Grup C) but was slw as cmpared with the learning-experienced cntrl (Grup L). n this experiment, the interference effect appeared early in the third phase, as wuld be expected given the results fprevius studies (Butn, 16; Butn & Swartzentruber, 1). This result is again cnsistent with the idea that tw mechanisms influence reacquisitin rate: a savings mechanism that allws grups that have been thrugh cnditining and extinctin with any CS t respnd mre in Phase 3 than grups that have nt, and an interference mechanism that prduces less fthis effect in a reacquisitin grup than in a learning-experienced cntrl grup. While respnding was greater n trials fllwing reinfrced trials than n trials fllwing nnreinfrced trials, further analyses determined that this effect was cnfined t Grup R. This result replicates the main result f Experiment, and cntinues t suggest that sequential learning plays a rle in prducing rapid reacquisitin. n cntrast, Grup L did nt shw a reliable differentiatin. t is pssible that the present experiment was nt pwerful enugh t detect an effect that might have existed in Grup L; there was a clear numerical trend in the data. Hwever, the fact that Grup :s respnding was significantly facilitated relative t Grup C's fllwing bth nnreinfrced and reinfrced trials clearly indicates that the learning-t-earn effect was nt restricted t trials fllwing reinfrced trials. Therefre, the learningt-learn effect evident in this preparatin cannt be cmpletely attributed t sequential learning. EXPERMENT 4 The purpse fthe final experiment was t take a last lk at whether reacquisitin culd ccur slwly relative t a learning-naive cntrl in this cnditining prepara.. tin. We decided t take advantage fan implicatin f the sequential learning view suggested by Experiments and 3. Specifically, ifrapid reacquisitin was due in part t reinfrced trials' signaling further reinfrced trials, then ne way t reduce the facilitative effect, and thus create an pprtunity t view the interference effect, might be t drastically reduce the number f cnditining trials. With fewer cnditining trials, there wuld be fewer ccasins fr the animals t assciate reinfrced trials with ther reinfrced trials, and thus reacquisitin might actually ccur slwly as cmpared with a learning-naive cntrl. t is interesting t nte that previus experiments shwing slw reacquisitin in CER and taste aversin have invlved very few cnditining trials (eight in CER, and ne-tw in taste aversin). n Experiment 4 we therefre cnducted cnditining with nly eight trials. T btain cnditining with this methd in s few trials, we increased the magnitude f the fd US frm tw t five 45-mg fd pellets and increased the Ts frm t 0 min. Trial-spacing parameters were identical t thse f Butn and Swartzentruber (1) in the CER preparatin. We hped that by using nly eight cnditining trials, acquisitin f sequential learning wuld be minimized fr Grup R and reacquisitin wuld

11 REACQUSTON AFTER EXTNCTON 433 ccur slwly cmpared with bth a learning-experienced cntrl and, fr the first time, a rest cntrl. Methd Subjects The subjects were 4 female Wistar rats btained directly frm Charles River, Canada. The rats were apprximately 0 days ld at the start fthe experiment and were hused and maintained as befre. Apparatus The apparatus was identical t that in Experiments and 3, except that new, quieter feeders (Med Assciates, Gergia, VT) replaced the riginal nes and Frmula All pellets (45 mg, P. 1. Nyes) were used. n this experiment, the fd US cnsisted f five pellets delivered 0.4 sec apart. Prcedure Pretraining. Bx expsure and magazine training were carried ut as in Experiments and 3. Apprximately 0 pellets were delivered during each magazine training sessin. Thrughutthe experiment, there was ne sessin a day. Cnditining. The subjects then received tw sessins fcnditining. Fr Grup R, the sessins cnsisted ffur pairings f the 30-sec tne with the five-pellet US. The mean T! was 0 min. Grup L received the light-ffcs instead fthe tne. Grup C received equal expsure t the apparatus but n stimuli. Extinctin. Fllwing cnditining, the rats received sessins fextinctin. Fr Grup R, these sessins cnsisted feight presentatins f the tne alne with a mean T f.5 min. Grup L received the light-ff instead fthe tne. On the last trial f the last extinctin sessin, Grup L received a nnreinfrced preexpsure f the tne instead f a light-ff presentatin (t eliminate the tne's nvelty at the start freacquisitin; see Butn & Swartzentruber, 1). Grup C received nly expsure t the apparatus during these sessins. N fd was presented t any grup during this phase. Reacquisitin. Fllwing extinctin, all rats received fur sessins f tne cnditining. Fr all grups, the tne was paired with fd tw times and presented alne tw times with a mean T! f 0 min during these sessins. Reinfrced and nnreinfrced trials alternated (i.e., RNRN). Results Cnditining was successfully btained in eight trials; Grups Rand L had mean elevatin scres f and 0.1 n the first tw-trial blck and 4.44 and 3.50 n the final tw-trial blck f the phase. A grup X tw-trial blck ANOVA carried ut n all elevatin scres frm the phase revealed a significant main effect f blck [F(3,4) = 7.5]. Neither the main effect f grup [F(1, 14) = 1.6] nr the grup X blck interactin [F(3,4) =.0] were significant. A t test cmparing the mean f the final tw-trial blck with a hypthetical ppulatin mean f zer indicated significant respnding in Grup R [t(7) = 4.64] and in Grup L [t(7) = 5.4]. Extinctin ccurred fairly rapidly, with bth grups shwing little r n respnding by Sessin 4. A grup X blck ANOVA revealed nly a significant main effect f blck [F(43,60) =.44]. Neither the main effect f grup [F(1,14) = 1.71] nr the grup X blck interactin [F(43,60) = 0.6] were significant. The data frm the reacquisitin phase are presented in Figure. During reacquisitin, Grup R acquired respnding mre slwly than Grups Land C. A grup X blck ANOVA revealed a marginal but nnsignificant effect f grup [F(,1) = 3.17,p =.06], a significant effect f blck [F(7,147) = 4.], and a significant grup X blck interactin [F(14,147) = 1.]. Respnding tended t increase ver trials fr all grups, but there were grup differences n sme trials. Three unplanned cmparisns were cnducted n the largest differences between Grup R and each fthe cntrl grups and als between Grup L and Grup C (at a. =.017). These revealed that Grup R respnded significantly less than Grup C n Blck [t(16) = -.47] and significantly less than Grup L n Blck 5 [t(16) = ]. Grup L failed t differ frm Grup C n Blck 5 [t(16) =.0]. A similar analysis carried ut n prescres revealed n significant effects [largest F(7,147) = 1.74]. Mean prescres fr Grups R, L, and C were 1.1, 1.56, and 1.5, respectively. A grup X trial type X sessin ANOVA was als perfrmed n the data t assess pssible sequential learning. This analysis revealed a marginal but nnsignificant effectfgrup[f(,1) = 3.17,p =.06] and a significant effect f sessin [F(3,63) = 6.73], as in the previus analysis. Hwever, the grup X sessin interactin was nt significant [F(6,63) = 1.57]. There was a significant main effect f trial type [F(1,1) = 1.6], althugh nne f the interactins invlving trial type apprached significance [largest F(6,63) = 1.1]. The pattern indicates that the trial-type effect was evident regardless f grup and sessin. The verall mean elevatin scres fllwing reinfrced and nnreinfrced trials, respectively, were 1.53 and0.3 fr Grup R,.6 and 1.6 fr Grup L, and 3.1 and.1 fr Grup C. As in Experiment 3, separate sessin X trial type ANOVAs were carried ut fr each grup t determine Grup R Grup L Grupe ,...---r--, Trial Blcks Figure. Mean elevatin scres during reacquisitin in Experiment4.

12 434 RCKER AND BOUTON if there was differential respnding in each. These revealed that the main effect f trial type was significant nly fr Grups Land C [F(1,l) = 5.37 andf(1,l) = 4.4, respectively]. The trial type effect was nt significant fr Grup R [F(1,l) =.65]. There were n significant interactins between sessin and trial type [largest F(3,l) = 1.7J. A similar analysis carried ut n prescres als revealedamaineffectftrial type [F(1,l) = 5.1]. Hwever, it was in the ppsite directin fthat fr elevatin scres; respnding was greater fllwing nnreinfrced trials (mean f 1.61) than fllwing reinfrced trials (mean f1.7). Such a difference makes the previus analysis a cnservative ne, and is therefre nt prblematic. Discussin n this experiment, reacquisitin ccurred slwly as cmpared with the perfrmance f bth the learningexperienced and rest cntrl grups. These results are cnsistent with the CER findings f Butn (16) and Butn and Swartzentruber (1), and they establish that reacquisitin can ccur mre slwly in the appetitive methd than in bth learning-experienced and learningnaive cntrls. With the prcedure used in this experiment, reacquisitin was clearly slw, nt fast. nterestingly, the learning-t-learn effect bserved in Experiments la and 3 was nt bserved in the present experiment, inasmuch as the learning-experienced cntrl (Grup L) failed t differ frm the learning-naive cntrl (Grup C). This result may be cnsistent with ther evidence suggesting that learning t learn may require training beynd a minimum number fcnditining trials (Schreurs & Kehe, 17). Slw reacquisitin in this experiment is cnsistent with ur predictin based n sequential thery: The smaller number f cnditining trials used here might have been insufficient t allw the rat t learn that reinfrced trials predicted future reinfrced trials. Therefre, perfrmance during reacquisitin was dminated by extinctin perfrmance retrieved by nnreinfrced trials. t is wrth nting, hwever, that the smaller number fcnditining trials used here als meant that the rats had had less experience retrieving pellets frm the fd magazine by the start fthe reacquisitin phase. That inexperience might have intrduced a lnger delay between the tne and the US (r the respnse and the reinfrcer), althugh ur infrmal bservatins suggested that, as in earlier experiments, the rats retrieved the pellets immediately upn their delivery during reacquisitin. Anther pssibility is that the larger US used in this experiment was capable f generating mre frustratin n nnreinfrced trials (see, e.g., Amsel, 15; Wagner, 16). Thus, anticipatin fa large appetitive US might be especially effective at eliciting frustratin behavirs that might cmpete with the magazine-entry respnse. And since Grup R's tne had previusly predicted fd, Grup R might be especially frustrated by trials n which it was nnreinfrced. This pssibility cannt be excluded by the present data. Hwever, frustratin thery cannt explain analgus results that have been reprted with aversive, rather than appetitive, USs (Butn, 16; Butn & Swartzentruber, 1; Hart et a., 15). Thseresults, fcurse, can be handledby sequential thery. The sequential learning effect evident in Experiments and 3 was nt evident here. Respnding was generally higher n trials fllwing reinfrced than fllwing nnreinfrced trials, but further analysis revealed that this effect was nt reliable in Grup R. The greater respnding fllwing reinfrced than fllwing nnreinfrced trials, particularly in Grup C, was presumably due t the larger US used in this experiment, which culd have caused large increments in assciative strength fllwing each reinfrced trial. n fact, the increments must have been relatively large, given this US, because it prduced significant cnditining in Phase 1 with nly eight cnditining trials. The fact that we did nt detect reliably greater respnding fllwing reinfrced trials in Grup R is cnsistent with the pssibility that whatever mechanism causes slw reacquisitin might be sufficiently strng t diminish thse gains. GENERAL DSCUSSON n Experiments 1,, and 3, reacquisitin was mre rapid than acquisitin in a rest cntrl r a grup that received an equal number f US presentatins but n CS presentatins prir t the third phase. n Experiments 1A and lb, reacquisitin was als mre rapid than riginal acquisitin in the riginal cnditining phase. These results clearly suggest that rapid reacquisitin can ccur in the appetitive magazine-entry cnditining preparatin. n cntrast, in Experiments 1 and 3, reacquisitin was impaired relative t acquisitin in a learning-experienced cntrl. Mrever, it was impaired relative t bth a learning-experienced cntrl and a rest cntrl when cnditining was cnducted with nly eight trials and a partial reinfrcement prcedure was used during reacquisitin (Experiment 4). These results suggest that bth slw and rapid reacquisitin can ccur in the appetitive cnditining preparatin. Overall, then, the entire scpe fprevius results was btained using a single cnditining methd. This cnclusin has at least tw cnsequences. First, it strngly suggests that the different results btained previusly with different preparatins were nt necessarily due t inherent differences between the preparatins themselves (cf. Napier et a., 1). (Napier et al. had suggested that fast and slw reacquisitin might be linked t cnsummatry and preparatry cnditining preparatins, respectively.) Secnd, it is a difficult pattern fr many existing theries t explain. Fr example, Kehe's (1) mdel des nt predict slw reacquisitin as cmpared with the perfrmance f learning-experienced r rest cntrls underany circumstances. Cnversely, althugh ther mdels predict slw reacquisitin after a large number f extinctin trials, neither Pearce and Hall's (10) mdel nr Wagner's (11) SOP mdel predict

13 REACQUSTON AFTER EXTNCTON 435 the rapid reacquisitin bserved relative t the perfrmance f learning-naive cntrls. Finally, withut further refinement, the Butn (13) mdel cannt explain why a reacquisitin grup and a learning-experienced cntrl bth acquired cnditined respnding faster than did learning-naive cntrls in Experiments 1 and 3. One pssibility implicated by the present results is that sequential learning is a critical determinant freacquisitin rate. The results f Experiments and 3 indicated that respnding during reacquisitin was higher n trials fllwing reinfrcement than it was n nnreinfrced trials in a reacquisitin grup; the same pattern was nt evident in cntrls. Thus, reinfrced trials appeared t signal subsequent reinfrced trials in the reacquisitin grup. During cnditining, reinfrced trials were presumably assciated with ther reinfrced trials; similarly, during extinctin, nnreinfrced trials might have been assciated with ther nnreinfrced trials. This srt f sequential learning culd be implicit in many cnditining arrangements (e.g., Capaldi, 14). A sequential learning view actually predicts a cmplex set futcmes in reacquisitin experiments. fthe acquisitin f sequential learning depends n the number f acquisitin trials, then the rate f reacquisitin will be a functin fthe relative amunt ftraining with each trial type. Rapid reacquisitin wuld result frm prcedures invlving a large number finitial reinfrced trials, while slw reacquisitin wuld result frm prcedures invlving a large number fextinctin trials. Slw reacquisitin wuld als be mre likely t ccur when a partial reinfrcement prcedure is used during reacquisitin, because the presence f nnreinfrced trials wuld signal additinal nnreinfrced, as ppsed t reinfrced, trials. These pssibilities are generally cnsistent with the knwn facts f reacquisitin. First, slw reacquisitin des depend n extensive extinctin training (Butn, 16). Secnd, rapid reacquisitin appears t be limited t prcedures invlving extensive initial cnditining. n Experiments 1-3 f the present study, where the reacquisitin bserved was rapid relative t that flearningnaive cntrls, the animals had received a large number (~4) fcnditining trials. n cntrast, in previus CER and taste-aversin experiments, where slw reacquisitin had been bserved, there had been very few initial cnditining trials (::; ). Furthermre, when appetitive cnditining was carried ut in Experiment 4 with nly eight trials, slw reacquisitin was bserved. Prcedures that use many cnditining trials, then, will, due t strng assciatins amng reinfrced trials, be mre likely t prduce rapid reacquisitin. This wuld be especially likely in the NMR paradigm (e.g., Napier et a., 1), which requires hundreds ftrials fr cnditining with typical parameters. This view predicts that if numerus cnditining trials were delivered in CER r taste aversin, rapid reacquisitin might be bserved in these paradigms. t als predicts that iffewer cnditining trials were used in NMR cnditining, slw reacquisitin wuld be bserved. One limitatin f the sequential learning interpretatin at present is that it des nt explain why learningexperienced cntrls acquire cnditined respnding mre quickly than learning-naive cntrls-the learningt-learn effect (e.g., Kehe & Hlt, 14). There is a pssible sequential learning explanatin: Assciatins amng reinfrced trials learned with ne CS might generalize t reinfrced trials with a secnd CS. Thus, a learning-experienced grup might learn that reinfrced trials generally fllw ther reinfrced trials, regardless f the CS. Hwever, the results f Experment 3 did nt supprt this view; a learning-experienced grup respnded well in Phase 3, but did nt respnd reliably mre after reinfrced than after nnreinfrced trials. On the ther hand, it is pssible that the reinfrced-trial signal generalized in such a grup but that a nnreinfrced trial signal did nt. Such a pssibility wuld explain why respnding was high, but relatively undifferentiated, in the learning-experienced grup. Thugh speculative, this pssibility is cnsistent with ther evidence suggesting that retrieval f extinctin is relatively easy t disrupt with stimulus change (e.g., Butn, 13). REFERENCES AMSEL, A. (15). The rle ffrustrative nnreward in nncntinuus reward situatins. Psychlgical Bulletin, 55,-11. BALLENE,., & DCKNSON, A. (11). nstrumental perfrmance fllwing reinfrcer devaluatin depends upn incentive learning. Quarterly Jurnal fexperimental Psychlgy, 43, 7-6. BOUTON, M. E. (16). Slw reacquisitin fllwing extinctin f cnditined suppressin. Learning & Mtivatin, 17, BOUTON, M. E. (11). Cntext and retrieval in extinctin and in ther examples f interference in simple assciative learning. n L. Dachwski & C. F. Flaherty (Eds.), Current tpics in animallearning: Brain. emtin, and cgnitin (pp. 5-53). Hillsdale, N: Erlbaum. BOUTON, M. E. (13). Cntext, time, and memry retrieval in the interference paradigms fpavlvian learning. Psychlgical Bulletin, 114,0-. BOUTON, M. E., & BOLLES, R. C. (17a). Cntextual cntrl fthe extinctin f cnditined fear. Learning & Mtivatin,, BOUTON, M. E., & BOLLES, R. C. (l7b). Rle f cnditined cntextual stimuli in reinstatement fextinguished fear. Jurnal fexperimental Psychlgy: Animal Behavir Prcesses,S, BOUTON, M. E., & PECK, C. A. (1). Cntext effects n cnditining, extinctin, and reinstatement in an appetitive cnditining preparatin. Animal Learning & Behavir, 17, 1-1. BOUTON, M. E., & RCKER, S. T. (14). Renewal f extinguished respnding in a secnd cntext. Animal Learning & Behavir,, BOUTON, M. E., ROSENGARD, C., ACHENBACH, G. G., PECK, C. A., & BROOKS, D. C. (13). Effects f cntextual cnditining and uncnditinal stimulus presentatin n perfrmance in appetitive cnditining. Quarterly Jurnal fexperimental Psychlgy, 46, BOUTON, M. E., & SWARTZENTRUBER, D. (1). Slw reacquisitin fllwing extinctin: Cntext, encding, and retrieval mechanisms. Jurnal fexperimental Psychlgy: Animal Behavir Prcesses, 15, BROGDEN, W.., LPMAN, E. A., & CULLER, E. (13). The rle f incentive in cnditining and extinctin. American Jurnal fpsychlgy, 51, -ii7. BULLOCK, D. H., & SMTH, W. C. (153). An effect f repeated cnditining-extinctin upn perant strength. Jurnal fexperimental Psychlgy, 46, CAPALD, E. J. (14). The sequential view: Frm rapidly fading stim-

14 436 RCKER AND BOUTON ulus traces t the rganizatin fmemry and the abstract cncept fnumber. Psychnmic Bulletin & Review, 1, DANGUR, J., & NCOLADS, S. (177). Lack freacquisitin in learned taste aversins. Animal Learning & Behavir, 5, DELAMATER, A. R. (15). Outcme-selective effects f intertrial reinfrcement in a Pavlvian appetitive cnditining paradigm with rats. Animal Learning & Behavir, 3, FARWELL, B. J., & AYRES, J. J. B. (17). Stimulus-reinfrcer and respnse-reinfrcer relatins in the cntrl f cnditined appetitive headpking (gal tracking) in rats. Learning & Mtivatin,, FNCH, G., & CULLER, E. (135). Relatin ffrgetting t experimental extinctin. American Jurnal fpsychlgy, 47, FREY, P. W., & BUTLER, C. S. (177). Extinctin after aversive cnditining: An assciative r nnassciative prcess? Learning & Mtivatin,, FREY, P. w., & Rss, L. E. (16). Classical cnditining fthe rabbit eyelid respnse as a functin f interstimulus interval. Jurnal f Cmparative & Physilgical Psychlgy, 65, HALL, G., & CHANNELL, S. (15). Differential effects f cntextual change n latent inhibitin and n the habituatin f an rienting respnse. Jurnal fexperimental Psychlgy: Animal Behavir Prcesses, 11, HART, J. A., BOURNE, M. J., & SCHACHTMAN, T. R (15). Slw reacquisitin f a cnditined taste aversin. Animal Learning & Behavir, 3, HLGARD, E. R, & MARQUS, D. G. (135). Acquisitin, extinctin, and retentin f cnditined lid respnses t light in dgs. Jurnal f Cmparative Psychlgy, 1, -5. HOEHLER, F. K., KiRSCHENBAUM, D. S., & LEONARD, D. W. (173). The effects f vertraining and successive extinctins upn nictitating membrane cnditining in the rabbit. Learning & Mtivatin, 4, 1-1. HOLLAND, P. C. (17). Differential effects fmissin cntingencies n varius cmpnents f Pavlvian appetitive cnditined respnding in rats. Jurnal fexperimental Psychlgy: Animal Behavir Prcesses, 5, KAYE, H., & MACKNTOSH, N. J. (10). A change f cntext can enhance perfrmance f an aversive but nt f an appetitive cnditined respnse. Quarterly Jurnal fexperimental Psychlgy, 4, KEHOE, E. J. (1). A layered netwrk mdel fassciative learning: Learning t learn and cnfiguratin. Psychlgical Review, 5, KEHOE, E. J., & HOLT, P. E. (14). Transfer acrss CS-US intervals and sensry mdalities in classical cnditining fthe rabbit. Animal Learning & Behavir, 1,1-1. KEHOE, E. J., MORROW, L. D., & HOLT, P. E. (14). General transfer acrss sensry mdalities survives reductins in the riginal cnditined reflex in the rabbit. Animal Learning & Behavir, 1, KONORSK, J., & SZWEJKOWSKA, G. (150). Chrnic extinctin and restratin fcnditined reflexes:. Extinctin against the excitatry backgrund. Acta Bilgiae Experimentalis, 15, KONORSK, J., & SZWEJKOWSKA, G. (15a). Chrnic extinctin and restratin f cnditined reflexes: 111. Defensive mtr reflexes. Acta Bilgiae Experimentalis, 16, 1-4. KONORSK, J., & SZWEJKOWSKA, G. (15b). Chrnic extinctin and restratin f cnditined reflexes: V The dependence f the curse f extinctin and restratin f cnditined reflexes n the "histry" fthe cnditined stimulus (the principle fthe primacy ffirst training). Acta Bilgiae Experimentalis, 16, MACKNTOSH, N. J. (13). Cnditining and assciative learning. Oxfrd: Oxfrd University Press. McALLSTER, D. E., & McALLSTER, W. R. (14). Extinctin and recnditining fclassically cnditined fear befre and after instrumental learning: Effects f depth f fear extinctin. Learning & Mtivatin, 5, NAPER, R. M., MACRAE, M., & KEHOE, E. J. (1). Rapid reacquisitin in cnditining fthe rabbit's nictitating membrane respnse. Jurnal fexperimental Psychlgy: Animal Behavir Prcesses, 1, 1-1. PEARCE, J. M., &HALL, G. (10). A mdel fr Pavlvian learning: Variatins in the effectiveness f cnditined but nt f uncnditined stimuli. Psychlgical Review, 7, PEARCE, J. M., & REDHEAD, E. S. (15). Supernrmal cnditining. Jurnal fexperimental Psychlgy: Animal Behavir Prcesses, 1, RANDCH, A. (11). The US preexpsure phenmenn in the cnditined suppressin paradigm: A rle fr cnditined situatinal stimuli. Learning & Mtivatin, 1, RESCORLA, R A. (17). A Pavlvian analysis f gal-directed behavir. American Psychlgist, 4, RESCORLA, R. A., & HETH, C. D. (175). Reinstatement f fear t an extinguished cnditined stimulus. Jurnal fexperimental Psychlgy: Animal Behavir Prcesses, 1, -6. REVUSKY, S., & COOMBES, S. (17). Reacquisitin f learned taste aversins. Animal Learning & Behavir, 7, SCHREURS, B. G., & KEHOE, E. J. (17). Crss-mdal transfer as a functin f initial training level in classical cnditining with the rabbit. Animal Learning & Behavir, 15, SMTH, M., & GORMEZANO,. (165). Effects f alternating classical cnditining and extinctin sessins n the cnditined nictitating membrane respnse f the rabbit. Psychnmic Science, 3, 1-. SZWEJKOWSKA, G. (150). The chrnic extinctin and restratin f cnditined reflexes: 11. The extinctin against an inhibitry backgrund. Acta Bilgiae Experimentalis, 15, WAGNER, A. R. (16). Frustrative nnreward: A variety f punishment. n B. A. Campbell & R. M. Church (Eds.), Punishment and aversive behavir (pp ). New Yrk: Appletn-Century Crfts. WAGNER, A. R. (11). SOP: A mdel fautmatic memry prcessing in animal behavir. n N. E. Spear & R. R. Miller (Eds.), nfrmatin prcessing in animals: Memry mechanisms (pp. 5-47). Hillsdale, NJ: Erlbaum. (Manuscript received June, 15; revisin accepted fr publicatin December 1, 15.)