Disentangling the Roles of Approach, Activation and Valence in Instrumental and Pavlovian Responding

Size: px
Start display at page:

Download "Disentangling the Roles of Approach, Activation and Valence in Instrumental and Pavlovian Responding"

Transcription

1 Dsentanglng the Roles of Approach, Actvaton and Valence n Instrumental and Pavlovan Respondng Quentn J. M. Huys 1,2,3 *, Roshan Cools 4, Martn Gölzer 5, Eva Fredel 5, Andreas Henz 5, Raymond J. Dolan 1, Peter Dayan 2 1 Wellcome Trust Centre for Neuromagng, Unversty College London, London, Unted Kngdom, 2 Gatsby Computatonal Neuroscence Unt, Unversty College London, London, Unted Kngdom, 3 Medcal School, Unversty College London, London, Unted Kngdom, 4 Donders Insttute for Bran, Cognton and Behavour, Centre for Cogntve Neuromagng, Radboud Unversty Njmegen, Njmegen, Netherlands, 5 Charté Unverstätsmedzn Berln, Campus Charté Mtte, Berln, Germany Abstract Hard-wred, Pavlovan, responses elcted by predctons of rewards and punshments exert sgnfcant benevolent and malevolent nfluences over nstrumentally-approprate actons. These nfluences come n two man groups, defned along anatomcal, pharmacologcal, behavoural and functonal lnes. Investgatons of the nfluences have so far concentrated on the groups as a whole; here we take the crtcal step of lookng nsde each group, usng a detaled renforcement learnng model to dstngush effects to do wth value, specfc actons, and general actvaton or nhbton. We show a hgh degree of sophstcaton n Pavlovan nfluences, wth appettve Pavlovan stmul specfcally promotng approach and nhbtng wthdrawal, and aversve Pavlovan stmul promotng wthdrawal and nhbtng approach. These nfluences account for dfferences n the nstrumental performance of approach and wthdrawal behavours. Fnally, although losses are as nformatve as gans, we fnd that subjects neglect losses n ther nstrumental learnng. Our fndngs argue for a vew of the Pavlovan system as a constrant or pror, facltatng learnng by allevatng computatonal costs that come wth ncreased flexblty. Ctaton: Huys QJM, Cools R, Gölzer M, Fredel E, Henz A, et al. (2011) Dsentanglng the Roles of Approach, Actvaton and Valence n Instrumental and Pavlovan Respondng. PLoS Comput Bol 7(4): e do: /journal.pcb Edtor: Antono Rangel, Calforna Instute of Technology, Unted States of Amerca Receved November 12, 2010; Accepted February 22, 2011; Publshed Aprl 21, 2011 Copyrght: ß 2011 Huys et al. Ths s an open-access artcle dstrbuted under the terms of the Creatve Commons Attrbuton Lcense, whch permts unrestrcted use, dstrbuton, and reproducton n any medum, provded the orgnal author and source are credted. Fundng: Ths work was supported by a Wellcome Trust Grant and a Max Planck Award to RJD. PD was supported by the Gatsby Chartable Foundaton. Publcaton costs were shared between the Max Planck Award and the Gatsby Chartable Foundaton. The funders had no role n study desgn, data collecton and analyss, decson to publsh, or preparaton of the manuscrpt. Competng Interests: The authors have declared that no competng nterests exst. * E-mal: qhuys@cantab.net Introducton The functonal archtecture of respondng nvolves two fundamental components that are behavourally [1] and computatonally [2] separable: Pavlovan and nstrumental. The nstrumental component respects the stmulus-dependent contngency between responses and ther outcomes (stmulus-response and acton-outcome learnng) [3]. By contrast, preparatory Pavlovan responses, chefly nvolvng approach and wthdrawal, are elcted by the appettve or aversve valence assocated wth predctve stmul n a manner that s not dependent on the consequences of those responses [3 5]. The nteractons between the two systems are most evdent when automatcally-elcted Pavlovan responses nterfere wth contngent nstrumental respondng [1,6 9]. For nstance, pgeons wll strkngly contnue to peck at a lght predctve of food (a preparatory approach elcted by the appettve predcton), even f the food s wthheld every tme they peck the lght (the nstrumental contngency) [10,11]. Pavlovan nterference lkely contrbutes to many qurks of behavour such as mpulsvty [12], framng and [13], endowment effects [14] and many other anomales [15], ncludng neurologcal [16 19] and psychatrc dseases [20 26]. Further, puzzlng facets of seemngly purely nstrumental behavour such as the dffcultes n learnng go responses to avod punshments; or nogo to obtan rewards (unpublshed data) and even the restrctons n assocatons evdent n evolutonarly preparedness [27,28] mght be traced to Pavlovan prncples. However, nstrumental and Pavlovan systems share overlappng neural hardware. Ther bdrectonal nteracton s charactersed by two key trads: rewards are ted to approach and vgour; and punshments to wthdrawal and behavoural nhbton. The neuromodulator dopamne (DA) responds predomnantly to rewards [22,29 31], nduces behavoural actvaton and enhances approach [32 35]. Each aspect of ths trad confounds the role of the phasc DA bursts n the flexble acquston of nstrumental values [36 42]. Serotonn appears to le at the heart of the aversve trad, havng been lnked to punshments [43 45], behavoural nhbton and wthdrawal [25,32,46 52], although dopamne actng va D2 receptors lkely also plays a role n lnkng absence of rewards to nogo [17,53,54]. Sgnatures of both trads are also evdent n neural crcuts nvolved n response and choce. In the dorsal stratum, there are nterdgtated pathways for go and nogo, wth the go pathways agan lnked postvely to rewards va dopamne [16,18,55,56]. The ventral stratum s prmarly organzed along an appettve/aversve axs wth drect lnks to approach and wthdrawal behavours [57,58]. The aversve trad s also tghtly lnked to the dorsal raphé and the peraquaeductal gray [59,60]. The man routes to the scentfc nvestgaton of these nteractons conssts of tasks n whch Pavlovan stmul are presented durng ongong nstrumental tasks. However, these have PLoS Computatonal Bology 1 Aprl 2011 Volume 7 Issue 4 e

2 Author Summary Beautful background musc n a shop may well tempt us to buy somethng we nether need nor want. Valenced stmul have broad and profound nfluences on ongong choce behavour. After replcatng known fndngs whereby approach s enhanced by appettve Pavlovan stmul and nhbted by aversve ones, we extend ths to wthdrawal behavours, but crtcally controllng for the valence of the wthdrawal behavours themselves. We fnd that even when wthdrawal s appettvely motvated, t s stll nhbted by appettve Pavlovan stmul and enhanced by aversve ones. Ths shows, for the frst tme, that the effect of background Pavlovan stmul depends crtcally on the ntrnsc valence of behavours, and dffers between approach and wthdrawal. as yet not explored the full set of nteractons charactersng the overlap between the two systems. Two crtcal confounds reman: The frst confound concerns the precse nature of the effect of Pavlovan stmul on nstrumental behavours. The nstrumental behavours studed have largely been appettvely motvated approach behavours (n Pavlovan-Instrumental Transfer (PIT) and condtoned suppresson tasks, [1,6 8,61 63]), and one nstance of aversvely motvated wthdrawal behavour [64]. The relatve role of the appettve-aversve motvaton axs versus that of the approach-wthdrawal axs s unknown. Ths n turn obscures the nature of the nteracton: whether Pavlovan stmul nteract wth the value of the nstrumental behavour, or by promotng specfc responses [1], or even smply by modulatng behavoural actvaton [5]. Second, the extent to whch the separaton of reward and punshment processng nto opponent motvatonal structures apples to nstrumental as well as Pavlovan learnng s ncompletely explored [1,27,28,65]. All these ssues can smultaneously be addressed n a combned PIT and condtoned suppresson task wth both approach and wthdrawal actons n whch the overall motvatonal component of approach and wthdrawal are matched (Fgure 1 and Table 1). The task separates the contrbutons of approach and wthdrawal by usng two counterbalanced blocks, one nvolvng approach go versus nogo, and the other wthdrawal go versus nogo. The comparson between go and nogo controls for effects of behavoural actvaton or nhbton. In each block, subjects frst underwent bref nstrumental tranng (Fgure 1A), learnng from postve and negatve feedback (monetary gans and losses of J0.20) whether to produce a go or a nogo response assocated wth sortng mushrooms. In the approach block (Fgure 1A, top, all 46 subjects), go responses nvolved movng the cursor onto a mushroom (to collect t), whle nogo nvolved dong nothng, thus not collectng the mushroom. To test for the effect of low-level motor varables, subjects performed one of two types of wthdrawal actons. In throwaway (24 subjects, Fgure 1A, mddle), go nvolved movng the cursor physcally away from the mushroom and clckng nto an empty blue box; nogo nvolved dong nothng, and thus keepng the mushroom. Importantly, both approach to and wthdrawal from the nstrumental stmulus were orthogonal to any approach and wthdrawal that mght be drected at the Pavlovan background stmulus. In release (22 subjects, Fgure 1. Task descrpton. A: Instrumental tranng. To centre the cursor, subjects clcked n a central square. In approach trals (top), subjects chose whether to move the cursor towards the mushroom and clck nsde the blue frame onto the mushroom (go), or not do anythng (nogo). In throwaway wthdrawal trals (mddle), they nstead moved the cursor away from the mushroom and clcked n the empty blue frame (go) or dd nothng (nogo). In release wthdrawal trals (bottom), subjects were nstructed to keep the button pressed after the ntal clck n the central square. The mushroom was then presented centrally, under the cursor. To throw away the mushroom, subjects released the button. Outcomes were presented mmedately after go actons, or after 1.5 seconds. B: Pavlovan tranng. Subjects passvely vewed stmul and heard audtory tones, followed by wns and losses. C: On Pavlovan query trals, subjects chose between two Pavlovan stmul. No outcomes were presented, but they were counted and added to the total presented at the end of the experment. D: Pavlovan-nstrumental transfer. Subjects responded to nstrumental stmul wth Pavlovan stmul tlng the background. No outcomes were presented, but subjects were nstructed that ther choces counted towards the fnal total. No explct nstructons about the contrbuton of Pavlovan stmul towards the fnal total were gven. do: /journal.pcb g001 PLoS Computatonal Bology 2 Aprl 2011 Volume 7 Issue 4 e

3 Table 1. Expermental layout. Approach Block A1 Instrumental tranng (60 trals) Probablstc renforcements 1 : J s I 1,2,3?approach p(rewjgo,s I 1,2,3 )~0:7, p(punjgo,si 1,2,3 )~0:31 s I 4,5,6?nogo p(rewjnogo,s I 4,5,6 )~0:7, p(punjnogo,si 4,5,6 )~0:31 A2 Pavlovan tranng (60 trals) Determnstc renforcements s P zz?reward s P z?reward s P 0? 1 J 0.10 J 0 s P {?punshment s P {{?punshment J 21 J A3 PIT (100 trals) No Renforcements Wthdrawal Block s P s I 1{6?? W1 Instrumental tranng (60 trals) Probablstc renforcements 1 : +0:20 J s I 7,8,9?wthdraw p(rewjgo,s I 7,8,9 )~0:7, p(punjgo,si 7,8,9 )~0:31 s I 10,11,12?nogo p(rewjnogo,s I 10,11,12 )~0:7, p(punjnogo,si 10,11,12 )~0:31 W2 Pavlovan tranng (60 trals) Determnstc renforcements s P zz?reward s P z?reward s P 0? 1 J 0.10 J 0 s P {?punshment s P {{?punshment J 21 J W3 PIT (100 trals) No Renforcements s P s I 7{12?? Note the numercal subscrpts on the nstrumental stmul s I here refer to ther denttes, not to the tme of presentaton. 1 For subject wth determnstc nstrumental renforcements, the outcome probabltes were 1 and 0 nstead of 0.7 and 0.3, respectvely. do: /journal.pcb t001 Fgure 1A, bottom), the subjects had to start by pressng the mouse button. Go nvolved releasng the button to avod collectng the mushroom; nogo nvolved contnung to press the button and thereby recevng the mushroom. In order to orthogonalse the approach-wthdrawal and appettve-aversve axes, the learned nstrumental values n approach and wthdrawal blocks needed to be matched. To acheve ths, both go and nogo responses were, f correct, rewarded. Addtonally, to avod the confound of actvaton, n each block (.e. n both approach and wthdrawal blocks) the go acton was desgnated as the correct response to half the nstrumental stmul, and the nogo acton to the other half (see Table 1). Incorrect responses had opposte outcome contngences to correct responses, yeldng more punshments than rewards. Ths ensured that go, nogo, approach and wthdrawal overall had the same learned assocaton wth rewards and punshments. We tested both determnstc and probablstc outcomes but found no dfferences. In the second part of each block, subjects passvely vewed unrelated, fractal, stmul pared wth separate audtory tones (Fgure 1B). Each compound Pavlovan stmulus s P was determnstcally assocated wth a monetary gan or loss,.e. ts Pavlovan value V(s P ) was equal to that monetary outcome. Every ffth tral n the Pavlovan block was a query tral (Fgure 1C), n whch subjects chose the better of two fractal vsual stmul wthout beng nformed about the outcome. Fnally, n the PIT stage, the nstrumental stmul were presented on a background of fractal Pavlovan stmul together wth the audtory tones, and agan wthout outcome nformaton. Our task addressed the key confounds descrbed above. Wth respect to the trads, we found that the Pavlovan nfluence s acton specfc: appettve Pavlovan cues boosted go approach responses and suppressed wthdrawal go responses; aversve Pavlovan cues dd the opposte. Addtonally, subjects were substantally based aganst wthdrawal, but we found no evdence that the nstrumental learnng component tself dffered between the approach and wthdrawal condton. Results The key results n ths paper concern the nteracton of valued Pavlovan stmul on nstrumental choces. We frst present a drect analyss of the choce data and reacton tmes. We then provde a detaled modellng analyss of the data, employng a strngent form of group-level model selecton that assesses each model s parsmony by weghng ts ablty to ft the data aganst ts complexty. The models quantfy Pavlovan values V(s P ), whch are the expectatons of a gan or loss gven Pavlovan stmulus s P, and nstrumental choce values Q t (a,s I ), whch are the tmevaryng expectatons of a reward gven a response a to an nstrumental stmulus s I. The structure of the most parsmonous model mples the nfluences and nteractons that were sgnfcant (for nstance rulng n a bas aganst actve wthdrawal, but rulng out any dfference between the nstrumental learnng rates assocated wth approach and wthdrawal); the values of the parameters n ths model ndcate the nature of those nfluences and nteractons. PLoS Computatonal Bology 3 Aprl 2011 Volume 7 Issue 4 e

4 Model-free analyses There was no dfference between the results for probablstc and determnstc feedback, and we therefore present the combned data. Analyss of the components of the experment ndcate robust, yet moderate, nstrumental condtonng that was stable durng the PIT perod, combned wth hghly robust Pavlovan condtonng. Fgure 2A shows the nstrumental probablty of choosng the more rewarded ( correct ) stmulus over tme. Subjects rapdly came to prefer the more rewarded acton. Preference was weaker for go wthdrawal, aganst whch there was a consstent bas. We ntended the nstrumental preference to be weak to avod celng effects when assessng PIT. Subjects also exhbted predctable varablty on a shorter tmescale: Fgure 2B shows the mmedate consequences of rewards and punshments on subsequent behavour. It s notable that punshments dd not reduce the repeat probablty below chance level (mean p(swtch t jpun t{1 ) s not v0:5, one-taled t-test pw:2). The same was found when analysng go and nogo choces separately: n both cases, p(swtch t jpun t{1 ) was not sgnfcantly dfferent from 0.5 (both pw:3, two-taled t-test), and was sgnfcantly smaller than p(stay t jrew t{1 ) (both pv4 10 {6, pared t-test). Whether ths really does represent an nsenstvty to punshments depends, however, on the average stay probablty, and on how ths average stay probablty s related to past renforcements. Subjects were nstructed that the outcomes of responses n the PIT block would be counted as n the nstrumental block. Fgure 2C shows that ths led to stable mantenance of the nstrumental response tendences throughout the PIT block. Fgure 2D shows that all but one (excluded) subject showed extremely good performance on the Pavlovan query trals nterleaved wth the Pavlovan tranng (mean correct w95%). Gven the success of nstrumental and Pavlovan tranng, we next analysed the raw effect of Pavlovan stmul on approach and wthdrawal choces. Fgure 2E shows a hghly sgnfcant nteracton between block and Pavlovan stmulus valence. Relatve to neutral stmul, postve Pavlovan stmul enhanced approach and nhbted wthdrawal go over nogo. Conversely, negatve Pavlovan stmul enhanced wthdrawal and nhbted approach go over nogo. A smlar analyss lookng at the probablty of respondng ncorrectly (outsde the blue box) showed no effect of the Pavlovan stmul n ether approach or wthdrawal condton and no nteracton (p~0:26,0:22,0:88 respectvely, ANOVA), suggestng that these results were not due to response competton. Note that the wthdrawal go probabltes were lower than the approach ones, agan reflectng the overall bas aganst go wthdrawal. Average reacton tmes for go approach and go wthdrawal actons dd not dffer (p~0:097, 2-taled t-test). Aganst our expectatons, Pavlovan stmul of both postve and negatve valence shortened reacton tmes n a parametrc manner relatve to neutral Pavlovan stmul (Fgure 2F, p = , ANOVA), although ths effect was not present n ether block separately (p = and p = respectvely, ANOVA). Model-based analyses The sze of the PIT effect may have been affected by the extent of nstrumental learnng (and thus the actual learned acton values), by response bases, and by generalzaton from the nstrumental to the PIT stage. In addton, there may have been dfferences n the nstrumental learnng of approach and wthdrawal actons (Fgure 2A). We decomposed and analysed all such factors usng a detaled renforcement learnng model. Ths contaned explct parameters capturng all the nstrumental and Pavlovan effects n Fgure 2. Raw choce probabltes. A&C: Average probablty (+1 standard error) of choosng the more rewarded ( correct ) acton n the nstrumental (A) and PIT (C) parts. Average performance was above chance n all cases, but worse when wthdrawal go was the more rewarded acton (red). There was no extncton durng the PIT block. Each pont s the average across subjects and across four trals. B: The bars show mean overall probablty of repeatng an acton n the nstrumental part gven that t was last rewarded n the presence of the current stmulus, or the probablty of swtchng gven a prevous punshment. Punshments do not lead to relable swtchng. D: Choce probabltes n the Pavlovan forced choce query trals. Most subjects were close to perfect. The grey bars show the probabltes of left: choosng a very good stmulus (++) over a good (+) or neutral (0) stmulus; mddle: choosng a bad (2) or neutral (0) stmulus over a very bad (--) stmulus; rght: choosng a postve (++ or +) stmulus over a negatve one (-- or -). Subjects that performed submaxmally n the appettve Pavlovan doman dd not necessarly have lower reward senstvtes n the nstrumental task, and vce versa for aversve Pavlovan stmul and punshment senstvty. E: PIT effects. The left part shows the approach PIT block, the rght part the wthdrawal PIT block. Each bar shows the log rato of the choce probablty (go/nogo) n the presence of one of the fve Pavlovan stmul. There was a sgnfcant effect of Pavlovan stmulus valence n each block. In addton, there was a sgnfcant block Pavlovan stmulus valence nteracton. Grey bars are means +1 standard error (red) and +95% confdence ntervals (green). F: Reacton tmes, pooled data for both PIT blocks. The bgger the absolute valence of the Pavlovan stmulus, the shorter the reacton tme. do: /journal.pcb g002 PLoS Computatonal Bology 4 Aprl 2011 Volume 7 Issue 4 e

5 Fgure 3. Model comparson. Each bar shows the dfferental BIC nt score relatve to the model wth the lowest BIC nt score (log e scale). Note that these BIC nt scores are for the group as a whole. Top: Models 1 7 were ftted to the nstrumental data only. Model 1 was a standard Rescorla- Wagner type model whch forced rewards and punshments to be equally nformatve. It assumed equally fast learnng about rewards and punshments, and no bases. Incluson of ether separate reward and punshment senstvtes (2r, Model 2) or separate bases n the approach and wthdrawal blocks (Model 4) mproved the ft. Separate learnng rates for rewards and punshments (Model 3) dd not mprove the ft as much as separate reward and punshment senstvtes (Model 2). The best model (5) ncluded a separate go bas n the approach and wthdrawal blocks, and separate reward and punshment learnng rates. Models that addtonally allowed separate renforcement senstvtes (Model 6), or separate learnng rates (Model 7) n the approach and wthdrawal blocks faled to mprove the ft. Bottom: Comparson of models on both nstrumental and PIT choce data jontly. Models 8 10 used the nstrumental component of Model 5. Models 8 10 ncluded ten Pavlovan factors, capturng the effect of each of the fve Pavlovan stmul n each of the two blocks. Model 9 allowed for extncton by ncludng an exponental decay of the nstrumental values durng the PIT part of the task. Model 10 ncluded random generalsaton nose and provded the best ft. do: /journal.pcb g003 the task, and was ft to the choce data of all subjects. We used group-level Bayesan model comparson [66] to choose amongst a varety of model formulatons (reportng DBIC nt scores relatve to the fnal model), and ensured that nference yelded correct parameter estmates when run on surrogate data generated from the assumed underlyng decson process. Instrumental learnng The fnal model ncluded 5 parameters assocated drectly wth the nstrumental requrements of the task. These comprse one learnng rate ; two parameters bas app and bas wth representng the bas towards go n the approach and wthdrawal blocks; and two separate free parameters r rew and r pun, representng the effectve strengths of rewards and punshments. At a group level, subjects were based aganst actve wthdrawal, but showed no bas for or aganst approach (p~8 10 {8 and p~0:70 respectvely, two-taled t-test), the dfference beng sgnfcant (p~5 10 {5, ANOVA, Fgure 4A). Wthdrawal bases n the release and throw away expermental subgroups dd not dffer (p~0:62, ANOVA), controllng for motor effects. The wthdrawal bas accounts for the lower performance on go wthdrawal n Fgure 2A. One concern s that dfferences n the bases mght have masked dfferences n learnng (.e. the reward senstvtes) n the approach and wthdrawal condtons. We tested ths by allowng for separate reward and punshment senstvtes n the two condtons (Model 6) or separate learnng rates (Model 7). The use of these extra parameters was structurally rejected by the model selecton process (DBIC nt ~12:6; 19:7 respectvely for the purely nstrumental trals); and the freedom to choose dfferent parameter values n these condtons was duly not used (Fgure 5). The absence of any dfference n the learnng parameters for approach and wthdrawal suggests that the nstrumental system treated approach and wthdrawal entrely equally. We wll see below that ths was not true for the Pavlovan system. Although, by desgn, rewards and punshments were equally nformatve, subjects chose to rely more on rewards than punshments (Fgure 4B). Rewards had a stronger effect than punshments both at a group level and for all ndvdual subjects, the dfference beng sgnfcant (pv1 10 {15, ANOVA). Indeed, the average punshment senstvty was not dstngushable from zero (p~0:37, two-taled t-test). Ths remaned true when we separately tested subjects who were gven determnstc (p~0:34, two-taled t-test) and probablstc (p~0:0627, two-taled t-test) feedback. Supplementary analyses (Text S1) excluded two further explanatons for the punshment nsenstvty: frst, that t s due to choce perseverance (Fgure S1 Text S1); and second that t s due to an emergng maxmsaton behavour (Fgure S2 n Text S1). Thus, t appears that the pattern seen n Fgure 2B s ndeed due to a dfferental senstvty to rewards and punshments. Generalzaton: Extncton versus nose We next analysed the generalzaton of nstrumental Q(s,a) values from the nstrumental to the PIT blocks. Generalzaton could be mperfect n two ways - the startng Q(s,a) values n the PIT block could dffer from the endng Q(s,a) values n the precedng nstrumental block, and the Q(s,a) values could then decay over tme or trals durng the PIT block gven the lack of nformaton about the outcomes. We constructed models ncludng such effects, and tested whether ther excess complexty was outweghed by ther ft to the data. PLoS Computatonal Bology 5 Aprl 2011 Volume 7 Issue 4 e

6 Fgure 4. Instrumental model parameters. A: Go bases for the approach and wthdrawal condton n the full experment. Subjects were only based aganst go, compared to nogo, n the wthdrawal block. B: Reward and punshment senstvty. Subjects were sgnfcantly more senstve to rewards than punshments. C: Generalzaton nose. Effectve Q value dfferences between go and nogo actons for all stmul and subjects, at the end of nstrumental learnng and durng the PIT block. Generalzaton seemed noser when acton preferences were weaker. D: Mean Q values of correct (.e. more frequently rewarded) actons. There was no dfference, and all correct actons had postve expectatons on average. E: PIT parameter estmates, correctng for nstrumental learnng, response bases and generalzaton nose. Postve Pavlovan stmul enhanced approach go actons and nhbted wthdrawal go, whle negatve Pavlovan stmul nhbted approach go actons and enhanced wthdrawal go actons. The nteracton was hghly sgnfcant, as were the two lnear man effects. F: There was no dfference between the effect of Pavlovan stmul on throwaway versus release go actons (all p values n E and F are ANOVA). Throughout, grey bars are pror means wth estmates of standard error (red) and 95% confdence nterval (green). Black dots show ndvdual data ponts, and ndvdual subjects parameters are connected by a dashed grey lne n A and B. do: /journal.pcb g004 As expected from the stable raw probabltes of choosng the correct (.e., more rewarded) opton (Fgure 2C), a model n whch the nstrumental Q(s,a) values decayed exponentally over tme durng the PIT block (mmckng extncton) dd not provde a good account of the data (Model 9, compared to Model 10 DBIC nt ~865). Fgure 5. Reward senstvtes and learnng rates n nstrumental approach and wthdrawal blocks do not dffer. A: The dark bars show the reward (left) and punshment (rght) senstvtes n Model 5, whch collapses across approach and wthdrawal condtons. The grey and lght grey bars show the senstvtes when ft separately for approach and wthdrawal blocks (Model 6). There s no dfference between blocks; and the jont parameter dffers from nether (all parwse comparsons pw:19). B: Dark bar shows learnng rate collapsed across both condtons n Model 5. Grey and lght grey bars show learnng rates when ft separately for approach and wthdrawal condton. Agan, no parwse dfference s sgnfcant (all pw:2). Throughout, black dots show ndvdual data; bars show pror means and red and green error bars 1 estmated standard error and 95% confdence nterval, respectvely. do: /journal.pcb g005 PLoS Computatonal Bology 6 Aprl 2011 Volume 7 Issue 4 e

7 Rather, the fnal model allowed for the addton of random generalzaton nose to each Q(s,a). These factors were drawn ndependently from the same normal dstrbuton for all stmulusacton pars, and the mean and varance of ths dstrbuton were both nferred wthout constrants (see Methods). Fgure 4C vsualzes the resultng changes; each dot represents the preference for the go acton (Q(s,go){Q(s,nogo)) for all subjects and all stmul. The abscssa shows ths at the end of the nstrumental stage, the ordnate after addton of the nose for the PIT stage. Importantly, there was no systematc dfference n mean correct acton values ether n the nstrumental or PIT stage (Fgure 4D). Pavlovan-Instrumental transfer We were manly nterested n the effect of the Pavlovan values on nstrumental performance. We therefore ftted 10 unconstraned parameters to separately capture the nfluence of each of the fve Pavlovan stmul on nstrumental go actons n both the approach and wthdrawal condton. All models accounted for performance n the PIT part by addng up nstrumental and Pavlovan nfluences pror to takng a softmax [67,68]. Ths amounts to treatng nstrumental and the Pavlovan controllers as separate experts, each of whch voted for ts preferred acton. The model captured n detal, and thereby controlled for, varablty n nstrumental learnng and generalzaton. The fnal model predcted the choces of every ndvdual subject better than chance (bnomal probablty, pv:0001 for every subject, overall predctve probablty ). The maxmum a posteror (MAP) estmates of ths model s parameters panted a pcture very smlar to that seen n the raw data. Fgure 4E shows the parameters of the model related to the nfluence of each Pavlovan stmulus. The pattern mrrored that seen n the raw data: there are hghly sgnfcant, and opposte, effects n the approach and wthdrawal blocks, wth appettve stmul (++ and +) promotng approach but nhbtng wthdrawal; and aversve stmul (-- and -) promotng wthdrawal but nhbtng approach. At a sngle subject level, the effect n the approach block was seen n 45/46 subjects (98%), whle t was seen n 30 subjects (65%) n the wthdrawal block. Snce there was no dfference n the learned value of go or nogo actons n ether approach or wthdrawal blocks, and n ether the nstrumental learnng or the PIT stages (Fgure 4D), any PIT effects are unlkely to be due to a preferental assocaton of a Pavlovan stmulus wth the learned value of an acton. Rather, they reflect the approach or a wthdrawal nature of the acton. We ncluded two separate groups of subjects who ether performed a throwaway wthdrawal acton, or a release wthdrawal acton. Ths was both to test the contrbuton of an approach/wthdrawal component amed at the Pavlovan stmul tlng the background, and n recognton of the sophstcaton of defensve reactons [27]. Fgure 4F shows that Pavlovan stmulus value had a sgnfcant, lnear effect on both wthdrawal acton types, and that ths overall lnear effect dd not dffer between the two acton types. At an ndvdual level, lnear correlatons were postve for 16 (72%) and 14 (58%) subject n the release and throwaway condton, respectvely. Psychometrc measures No psychometrc measure of anxety or depresson correlated wth any of the parameters n the man model. Dscusson Our task was desgned to look nsde the trads of valence, behavoural actvaton and nhbton, and specfc actons assocated wth Pavlovan nfluences. Ths ssue has been ncompletely explored n the past. Ether these trads as a whole have been nvestgated: aversve actons allowed avodance of, or escape from, a negatve renforcer; appettve actons, the acquston of a reward [6,8,64], or, as n negatve automantenance [10], the relevant Pavlovan contngences have been tghtly embedded n the nstrumental task. Here, we found that Pavlovan nfluences dstngushed approach from wthdrawal when carefully controllng for actvaton, for appettve versus aversve nstrumental motvaton, and for detals of the motor executon. Thus, for nstance, a Pavlovan stmulus predctng reward had opposte effects on two dfferent nstrumental actons (approach and wthdrawal) even though both those actons were themselves equally motvated by the acquston of reward. Approach and avodance were defned n two parallel ways: by the cogntve label for the acton ( throw away, collect ) and by the relaton to the stmulus (movng the mouse/fnger towards or away from the stmulus). Our task dd not set out to dstngush these two contrbutons (cogntve and motor), and we also dd not attempt to quantfy subjects explct nsght nto ther strateges. However, both possbltes are mportant. At a cogntve level, subjects should neglect the Pavlovan stmul: by desgn, they are not nformatve about the nstrumental task. Upon enterng the PIT stage, subjects were also explctly nstructed to contnue dong the nstrumental task as before. If despte these facts subjects were cogntvely swayed to nclude the rrelevant backgrounds n ther goal-drected decson process, then our fndng show that Pavlovan contngences extend even nto cogntve choces. Ths s of course consonant wth a large number of behavoural rregulartes n human decson makng [12 15]. The motor aspects are equally nterestng snce they suggest a fne level of detal n the archtecture of Pavlovan nfluences. There s qute some evdence for ths; for nstance, Pavlovan CRs are known to be hghly adaptve to the detals of the CS (for nstance evokng a groomng condtoned response to a rat whch functons as a food CS, rather than a gnawng CR [69]) and to the nature of the US [70]. In humans, a plexglass postoned between subjects and an appettve US abolshes an ncreased wllngness to pay [71]. The performance on the purely nstrumental porton of the task was also revealng. We observed a dfference n the nstrumental performance of approach and wthdrawal acton; and ths came (unlke n prevous tasks) after controllng for the motvatonal dfference between approach and avodance. Our model-based analyss revealed that the dfference was not due to a dfference n learnng (.e. a dfference n the nstrumental parameters relatng renforcements to performance), but due to a statc bas aganst performng a wthdrawal go acton. Of course, lke all other tasks, our nstrumental task also had embedded Pavlovan contngences, and, ndeed, a Pavlovan suppresson of actve wthdrawal by the overall appettve framng of the task (subjects on average chose the correct, rewarded, acton more often) could mrror what we saw n the PIT stage of the task. Alternatvely, ths could be the result of subjects experences upon enterng an expermental stuaton n whch they are gven a computer mouse. We have nterpreted such as bas n terms of evolutonary preparedness or programmng [2,9,24,50,72]. That s, the flexblty of the arbtrary outcome-contngent mappngs of nstrumental control comes at the prce of the experence necessary for t to be specfed. Pavlovan prors substtute nflexble hard-wred choces that are mmedately avalable for ths flexble nstrumental adaptatvty wth ts potentally substantal sample complexty (.e. the potental need for extended experence). Related bases are wdely known: dogs wll happly learn to run, but not to yawn, for food; teachng a PLoS Computatonal Bology 7 Aprl 2011 Volume 7 Issue 4 e

8 rat to escape s easer than teachng t to avod the shock [3,27,28]; humans perform actve go responses slower f nstructons are n terms of aversve feedback [51] or f they are followed by aversve nformaton [73]. Fnally, n humans, an nstructed joystck approach response to a happy face s qucker than a wthdrawal response, dependng on the cogntve/affectve label n a manner smlar to our own fndngs here [74]. Alternatve nterpretatons of the response bas nclude endowment effects [14], whereby an over-valuaton of tems notonally n one s possesson makes one reluctant to gve them up. Ths s unlkely because such a bas should be present across all nstrumental stmul,.e. across both stmul for whch a go and a no-go s the more rewarded acton (Fgure 4). Another possblty s a frame dependence [13] snce we compared go wth nogo rather than two alternatve go actons aganst each other. The negatve frame assocated wth sortng to remove bad mushrooms could have nhbted go actons. Neurobology One of the central motvatons for our nvestgaton was the observaton that the neural substrate does not respect the logcal ndependence of reward/punshment and approach/wthdrawal. Rather, as we have dscussed, these are ted together, va the structure of the stratum and also specfc neuromodulators. Whle the neural bass for the promoton of approach responses by appettve stmul s known to nvolve both amygdala and stratum [62,63,75], the neural bases for the effects of aversve Pavlovan stmul are less clear. There are no data on wthdrawal responses per se,.e. wth postve expectatons. Nevertheless, anmal models, genetc studes and pharmacologcal manpulatons suggest that serotonn plays a crucal role n the nhbton of actve behavours by aversve expectatons [25,47,48,50,73,76 78]. In humans, there s evdence for the serotonergc medaton of the nhbton of actve approach by aversve predctons [51], and of approach responses to stmul that are predctve of negatve renforcement [73]. It should be noted, though, that, actng va the ndrect path and D2 receptors, dopamne tself has also been suggested to be mportant n medatng nogo behavour due to punshments [18,53,79]. Aversve Pavlovan stmul can also potentate behavour [1,64,80,81], wth both serotonn and dopamne nvolved. Dopamne may have a domnant nfluence n ths: t s both known to be released, and nfluental, n some aversve settngs [82 85] and has a more evdent relatonshp to vgour [33,34]. Ths observaton has led to a re-nterpretaton of prevous notons [43] of the opponency between dopamne and serotonn, puttng an axs spannng nvgoraton and nhbton together wth spannng reward and punshment [52]. Thus, the lterature suggests three predctons for genetc correlates of the Pavlovan nfluences we observe. When consderng these, the caveats concernng the nteracton of genetc varaton wth psychopathology (e.g. anxety or depresson), and wth development need to be kept n mnd. Nevertheless, the condtoned suppresson effect of aversve Pavlovan stmul on approach should be enhanced by D2 receptors, and hence be postvely related to D2 stratal receptor densty thought to be modulated by C975T (rs6277; [17]). Second, condtoned suppresson should be ncreased n subjects wth hgher serotonn levels,.e. as mght be the case wth the less effcent (s) allelc varaton of the serotonn reuptake transporter (5HTTLPR SLC6A4 [86]). Thrd, gven dopamne s establshed postve correlaton wth approach and PIT [87,88], we expect genetc polymorphsms that boost DA levels, such as the SLC6A3 polymorphsm of the dopamne transporter [89], to ncrease the mpact of appettve Pavlovan stmul on approach. A smlar effect may be expected from DARPP-32, although ts closer relatonshp to synaptc plastcty would also suggest effects on nstrumental learnng [90 92]. Instrumental punshment nsenstvty Although the learnng parameters assocated wth nstrumental approach and wthdrawal dd not dffer, the mpact of rewards and punshments on the acquston of respondng was hghly asymmetrc. In general, subjects neglected punshments, whlst mantanng a fxed senstvty to reward. Ths was gratutous as, n our settng, rewards and punshments were equally nformatve. It s, however, the case that the optmal strategy can be arrved at by concentratng on ether. Subjects were not globally nsenstve to punshments, as ther choce behavour n the Pavlovan learnng was hghly accurate both for rewards and punshments. Furthermore, t should be emphaszed that ascrbng punshments a value of zero outcome would stll effectvely behave as a punshment because a zero outcome s well below the average expectaton of correct actons (Fgure 4D) and as such would reduce the tendency to emt the acton that caused t. The asymmetry has been noted before. Others have ftted models wth separate learnng rates for rewards and punshments and reported sgnfcantly slower learnng rates for punshments than rewards [93,94]. In some restrcted regmes, learnng rates and nverse temperature parameters can trade off, and we explctly tested both types of models to address ths. One potental confound s the emergence of determnsm. Subject were nstructed to perform choces relatve to mushrooms. Real world mushrooms are ether edble or posonous, and ths dchotomy may have predsposed subjects towards a determnstc, rather than a matchng, strategy. (For nstance, subjects may have chosen responses based on a classfcaton of the mushrooms nto good and bad ones, rather than on the partcular value of a response for a mushroom.) Indeed, n RL settngs t s typcally optmal to start wth a low, exploratory, senstvty to outcomes, but to ncrease ths over tme to encourage explotaton, culmnatng n a determnstc strategy [2]. However, subjects dd not behave determnstcally at any pont (Fgure 2A) and supplementary analyses showed that the tme-varyng pattern of renforcement senstvtes ths would predct s not observed n the data (Text S1). A further potental confound s the average stay probablty. If ths were precsely half-way between the stay probabltes after rewards and punshments n Fgure 2B, then rewards and punshments would have the same effect relatve to the baselne, and hence arguably be equally nformatve. However, ths argument would neglect the fact that the mean stay probablty tself must be a functon of the renforcement hstory; and that ths must be ncluded n makng nferences about the renforcement senstvty. We have prevously made the argument on theoretcal grounds that part of the asymmetry observed n appettve and aversve systems mght be due to the nherent dfference n how nformatve rewards and punshments are processed, enshrned agan n the archtecture of the stratum and neuromodulaton [50]. Rewards tell us what to do; punshments tell us what not to do. The former s more nformatve n naturalstc settngs where many optons are avalable but only few are good. The fact that subjects gratutously rely on rewards rather than on punshments n the present settng may reflect an mplct apprecaton of ths fact, although our fndngs are certanly n no way conclusve evdence. Interestngly, t s known that stronger optmalty results can be shown for a stochastc learnng automata rule called lnear reward-nacton, whch does not change propenstes n the lght of punshments but PLoS Computatonal Bology 8 Aprl 2011 Volume 7 Issue 4 e

9 only rewards ([95,96]; also known as a benevolent automaton [97]), than for a rule that changes propenstes for both. Modellng The computatonal model served several central roles. Frst, t encapsulated the manfold aspects of behavour and learnng jontly, thereby controllng for them: the bas aganst wthdrawals s not a due to a dfference n learnng; and varatons n learnng or generalzaton do not account for the PIT effects we saw. Secondly, ts close ft to the behavour argues that the PIT effects can be accounted for by a smple superposton of an nstrumental and a Pavlovan controller: the acton propenstes due to both controllers were smply multpled (as addtve factors n an exponental), rather than beng allowed to nteract n more complex ways. The smplcty of ths nteracton eschews questons about perpheral versus central response competton, whether appettve and aversve systems compete centrally [7], and whether Pavlovan learnng s nvolved n nstrumental learnng [1]. It takes the vew of multple, separate controllers contrbutng n parallel [98], and weghtng the ultmate choce by the reward expected from that choce. One alternatve would be to wegh contrbutons by dfferent controllers accordng to ther certanty [99], although t s unclear how to compute the Pavlovan controller s certanty. Lmtatons There are varous pressng drectons for future studes. Frst, despte the role the archtecture of decson-makng has played n the argument, our work does not drectly address the neural mechansms concerned. These could be examned usng magng and pharmacologcal manpulatons. Second, our task was not desgned to dstngush between outcome-specfc and general mechansms [63,75] as we reled on one, monetary, outcome throughout. Studyng dfferent outcomes s mportant, gven evdence for partly parallel pathways through dfferent nucle of the amygdala and dfferent targets n the nucleus accumbens [100,101]. Thrd, we are mssng one crucal further orthogonalzaton to do wth the overall framng of the nstrumental task. It s mportant to consder the case n whch subjects can at best avod losng money by dong the correct acton [51]. We would expect punshment to mantan ts nstrumental force n ths case; but there could also be a systematc dfference n the nature of the Pavlovan nfluences. Concluson Pavlovan responses are beleved to be hard-wred to reflect evolutonarly approprate atttudes to predctons, beng hghly adaptve and senstve to envronmental structures [102]. Here, we showed that Pavlovan nfluences on nstrumental behavour depend on the ntrnsc affectve label of an acton, ndependent of ts learned reward expectaton. It has long been known that prepared or compatble [27,69] behavours are easer targets for nstrumental condtonng. These ntrnsc bases, or prors, may serve a crucal functon both by reducng the need for collectng data (.e. sample complexty) about the effects of actons, and by reducng the need for executng complex processng necessary to work out optmal actons (.e. computatonal complexty). Both of these can be expensve or dangerous, partcularly n an aversve context. Our fndngs sharpen the understandng of the relatve contrbuton of Pavlovan and nstrumental contngences n general tasks. We showed clearly that the nteracton of Pavlovan and nstrumental behavours s organzed along the lnes of appettve and aversve motvatonal systems, and that a crtcal contrbutor to ths s the affectve nature of actons. Methods Subjects and procedure 54 healthy subjects of central European orgn were recruted from the Berln area. Subjects were screened for a personal hstory of neurologcal, endocrne, cardac and psychatrc dsorders (SCID- I screenng questonnare), and for use of drugs and psychotropc medcaton n the past 6 months. Subjects receved performancedependent compensaton (5 32 Euro) for partcpaton. Three subjects dd not meet ncluson crtera and one subject dd not complete the task; the data for three further subjects were lost due to a programmng error. One further subject was excluded from the analyss because the nstrumental task was not satsfactorly performed. The 46 remanng subjects were 25:3+4:7 years old. 59% were female (n~27). The study was approved by the local Ethcs Commttee and was n accord wth the Declaraton of Helsnk Subjects were gven detaled nformaton and gave wrtten consent. They were seated comfortably at a table n front of a laptop wth headphones and used a mouse wth ther domnant hand to ndcate ther choces. The amount earned was ndcated by the computer, and the sum pad n cash at the end of the sesson. The computer task was followed by completon of self-ratng scales. Task descrpton The task was wrtten usng Matlab and Psychtoolbox ( psychtoolbox.org). It conssted of one approach and one wthdrawal block separated by a 2 mnute break. Each block was n turn dvded nto a nstrumental tranng, a Pavlovan tranng and a PIT part. Table 1 llustrates ths. Instrumental tranng. The nstrumental task was framed n terms of a mushroom collectng and sortng task. Instrumental stmul were generc, coloured mushroom shapes. Trals started when subjects clcked n a central square (Fgure 1A). In the approach block, nstrumental stmul s I 1,2,3 and si 4,5,6 (wth subscrpts ndcatng the dentty of stmul, not the tme of presentaton) were then presented to one sde, surrounded by a blue frame (Fgure 1A, mddle column, top). Subjects ndcated that they wanted to collect the mushroom by movng the cursor onto the mushroom and clckng on t (approach go). They could also decde not to collect the mushroom by dong nothng for 1.5 seconds (approach nogo). At the end of each tral (after a clck for go trals or after 1.5 s for nogo trals respectvely), the stmulus dsappeared and the outcome was shown n the mddle of the screen (Fgure 1A). In the wthdrawal blocks, nstrumental stmul s I 7,8,9 and si 10,11,12 were presented. Subjects chose whether to throw away mushrooms (wthdrawal go) or do nothng (wthdrawal nogo). Two dfferent wthdrawal go actons were tested. The throwaway group (n~24) had to clck n a blue frame located on the opposte sde of the stmulus (see Fgure 1A, mddle column, mddle). The release (n~22) group was nstructed to press and hold the mouse button after clckng n the central square to begn the tral. The mushroom was then presented underneath the cursor (Fgure 1A, mddle column, bottom), and they could throw away a mushroom by releasng the button (wthdrawal go) or not throw away the mushroom by not releasng (wthdrawal nogo) untl 1.5 seconds had elapsed. Each block contaned three good (s I 1,2,3 and si 7,8,9 ) and three bad (si 4,5,6 and s I 10,11,12 ) mushrooms, randomly selected from the pool of 12 stmul. Subjects were gven explct renforcng feedback after every choce ( Correct, +20 cents or Wrong. 220 cents ), ether determnstcally (n~19) or probablstcally (n~27), but were not told whch mushrooms were good or bad. Correct trals were those on whch subjects threw away a bad or kept a good mushroom, and those on whch they collected a good or refraned from collectng a bad mushroom. Importantly, ths means that correct go actons of both types (approach ( collect ) and wthdraw ( throw away )) were followed PLoS Computatonal Bology 9 Aprl 2011 Volume 7 Issue 4 e

10 by both rewards and punshments. Thus, the renforcement expectances of correct approach and wthdrawal actons were equal and postve on average. Smlarly, ncorrect actons of both types were also followed by rewards and punshments, but more by the latter than the former. To ensure replcablty across expermental desgns, four expermental confguratons were ncluded, crossng determnstc/ probablstc nstrumental feedback and the two wthdrawal acton types ( throw away or release ). These manpulatons are beyond the mathematcal model descrbed below, and thus should not affect our fndngs. We present both data for all subjects and, testng nternal consstency, across the four groups. 10 subjects were n the determnstc throwaway group, 9 n the determnstc release, 14 n the probablstc throwaway and 13 n the probablstc release group. One-way ANOVA comparsons of MAP parameter estmates from the most parsmonous model (Model 10; see below) for determnstc and probablstc feedback dd not reveal any sgnfcant dfferences. Pavlovan tranng. Fve compound Pavlovan stmul consstng of a fractal vsual stmulus (Fgure 1B) and a tone were classcally condtoned. Each stmulus was presented 20 tmes and determnstcally followed, 1 second later, by the assocated outcome. Outcome presentaton lasted 1.5 seconds. Outcomes for the best (s P zz ), good (sp z ), neutral (sp 0 ), bad (sp { ) and worst (sp {{ ) stmul were, respectvely, gans of 100 cents, 10 cents, zero, and losses of 10 and 100 cents. To ensure that subjects pad attenton, every ffth tral was a query tral n whch subjects had to choose between two Pavlovan stmul (Fgure 1C). No feedback was gven n these trals, but subjects were nstructed that the choces would contrbute to ther compensaton. Pavlovan-Instrumental transfer. In the fnal part of each block, the nstrumental task was presented n extncton and on the background of Pavlovan stmul (Fgure 1D). Subjects were nstructed to contnue dong the nstrumental task; that choces were stll earnng them the same outcomes and were beng counted, but that they would not be told about the outcomes. Note, mportantly, that the Pavlovan stmulus was presented over the entre background, and as such could not by tself modulate the drectonalty of actons. Psychometrc measurements. After completng the tasks, subjects completed self-ratng scales (Beck Depresson Inventory II (BDI), Beck Anxety Inventory (BAI), State-Trat Anxety Inventory STAI [ ]), followed by the admnstraton of clncan rated scales (Montgomery-Ashberg Depresson Ratng Scale (MADRS), Hamlton Depresson Scale (HamD), Structured Intervew for the Hamlton Anxety Scale (SIGHA) and Clncal Global Impresson (CGI) [ ]). Models We modfed a standard renforcement learnng model to capture the behavoural choces n the experment. We frst descrbe the man model, and then the alternatve control models. Consderng frst the nstrumental part, let s I t be the nstrumental stmulus (out of up to 12;.e. the subscrpt t now desgnates tme rather than dentty as n Table 1) presented at tral t, and a t the acton (choce) on that tral. An acton can be one of four types: go wthdrawal and nogo wthdrawal n the wthdrawal block, and go approach and nogo approach n the approach block. Let also r t [ f{1,1g be the renforcement obtaned, ether {1 for a punshment, or z1 for a reward. We wrte the probablty of acton a t n the presence of stmulus s I t as a standard probablstc functon of ) the renforcement expectatons Q t (s I t,a t) assocated wth that par on that tral, and ) a tme-nvarant, fxed, response bas b(a t ): W I (s I t,a t)~q t (s I t,a t)zb(a t ) ð1þ p(a t js I t )~ exp WI (s I t,a X t) exp a WI (s I t,a ) ð2þ where W I s the nstrumental weght of acton a t, and where the varable b(a t ) can take on value bas wth for wthdrawal go actons, or bas app for the approach go actons. It s always zero for the nogo acton. There was no delayed outcome n the nstrumental task, and the expectatons were thus constructed by a Rescorla-Wagner-lke rule wth a fxed learnng rate. The mmedate, ntrnsc, value of the renforcements delvered n the experment may have dfferent meanng for dfferent subjects. To measure ths effect, we added two further parameters: the reward senstvty r rew and the punshment senstvty r pun, yeldng an update equaton for the expectatons: Q tz1 (s I t,a t)~q t (s I t,a t)z R t {Q t (s I t,a t) ( R t ~ r rew f r t w0 r pun f r t v0 Ths s model 5 n Table 2, whch has the lowest BIC nt score (see below). Alternatve models tested on the nstrumental data only are as follows: Model 1 assumes that {r pun ~r rew ~b, and that bas wth ~bas app ~0. Model 2 allows only for separate reward and punshment senstvtes and model 4 for separate bases. Model 3 agan assumes {r pun ~r rew ~b,andthatbas wth ~bas app ~0,but allows for two separate learnng rates,.e. n Equaton 3 s replaced by rew on trals where r t ~1, andby pun on trals where r t ~{1. Model 6 and 7 are expansons of the fnal model, allowng for separate reward and punshment senstvtes (model 6) and for separate learnng rates (model 7) n the approach and wthdrawal condtons. Our man measure of nterest s the effect of Pavlovan stmul on the approach and wthdrawal actons. Let addtonally s P t be the Pavlovan stmulus on tral t. We can then wrte an equaton smlar to equaton 2 for the trals where both nstrumental and Pavlovan stmul were present, but ncludng a term f (a,s P t ) that quantfes the effect of the partcular Pavlovan stmulus s P t on the acton a. Ths means that the acton weghts due to the nstrumental and Pavlovan controllers are added nsde the exponent of equaton 2, and that thus the probabltes each controller attaches to a partcular acton are multpled and renormalzed. The two controllers are therefore treated as two dstnct enttes, each separately votng for a partcular acton to be emtted. The nfluence of each system on acton choce s relatve to the strength wth whch the other enhances one partcular acton. We wrte the PIT weght of acton a as: W PIT (a,s I,s P )~W I (s I,a)zf (a,s P ) Here we force f (nogo,s P )~0 at all tmes. The go values f (go,s P ) can take on 10 separate, nferred, values, meanng that there s one separate parameter for each of the fve Pavlovan stmul s P n each of the two blocks. Each of these parameters captures how much s P boosts the go over the nogo acton (f f (go,s P )w0) orthenverse(f f (go,s P )v0). Note that because these are separately nferred, ndependent, parameters, ths formulaton does not mpose any assumptons about the effect of the value of the stmulus s P, or about the relatve effect of dfferent stmul s P wth dfferent values. Hence, ths controls for varaton n learnng durng the Pavlovan tranng block (though the query trals ndcate that learnng was very robust). Equaton 3 (Model 8 n Table 2) assumes that the stmulus-acton values Q(s I,a) at the end of the nstrumental block are perfectly and exactly generalzed to the PIT block. We frst tested an alternatve model (Model 9 n Table 2) that ncluded an exponental extncton ð3þ PLoS Computatonal Bology 10 Aprl 2011 Volume 7 Issue 4 e

Bonsai Trees in Your Head: How the Pavlovian System Sculpts Goal-Directed Choices by Pruning Decision Trees

Bonsai Trees in Your Head: How the Pavlovian System Sculpts Goal-Directed Choices by Pruning Decision Trees Bonsa Trees n Your Head: How the Pavlovan System Sculpts Goal-Drected Choces by Prunng Decson Trees Quentn J. M. Huys 1,2,3. *, Ner Eshel 4., Elzabeth O Nons 4, Luke Sherdan 4, Peter Dayan 1, Jonathan

More information

Appendix for. Institutions and Behavior: Experimental Evidence on the Effects of Democracy

Appendix for. Institutions and Behavior: Experimental Evidence on the Effects of Democracy Appendx for Insttutons and Behavor: Expermental Evdence on the Effects of Democrac 1. Instructons 1.1 Orgnal sessons Welcome You are about to partcpate n a stud on decson-makng, and ou wll be pad for our

More information

Incorrect Beliefs. Overconfidence. Types of Overconfidence. Outline. Overprecision 4/22/2015. Econ 1820: Behavioral Economics Mark Dean Spring 2015

Incorrect Beliefs. Overconfidence. Types of Overconfidence. Outline. Overprecision 4/22/2015. Econ 1820: Behavioral Economics Mark Dean Spring 2015 Incorrect Belefs Overconfdence Econ 1820: Behavoral Economcs Mark Dean Sprng 2015 In objectve EU we assumed that everyone agreed on what the probabltes of dfferent events were In subjectve expected utlty

More information

N-back Training Task Performance: Analysis and Model

N-back Training Task Performance: Analysis and Model N-back Tranng Task Performance: Analyss and Model J. Isaah Harbson (jharb@umd.edu) Center for Advanced Study of Language and Department of Psychology, Unversty of Maryland 7005 52 nd Avenue, College Park,

More information

Using the Perpendicular Distance to the Nearest Fracture as a Proxy for Conventional Fracture Spacing Measures

Using the Perpendicular Distance to the Nearest Fracture as a Proxy for Conventional Fracture Spacing Measures Usng the Perpendcular Dstance to the Nearest Fracture as a Proxy for Conventonal Fracture Spacng Measures Erc B. Nven and Clayton V. Deutsch Dscrete fracture network smulaton ams to reproduce dstrbutons

More information

Parameter Estimates of a Random Regression Test Day Model for First Three Lactation Somatic Cell Scores

Parameter Estimates of a Random Regression Test Day Model for First Three Lactation Somatic Cell Scores Parameter Estmates of a Random Regresson Test Day Model for Frst Three actaton Somatc Cell Scores Z. u, F. Renhardt and R. Reents Unted Datasystems for Anmal Producton (VIT), Hedeweg 1, D-27280 Verden,

More information

Modeling Multi Layer Feed-forward Neural. Network Model on the Influence of Hypertension. and Diabetes Mellitus on Family History of

Modeling Multi Layer Feed-forward Neural. Network Model on the Influence of Hypertension. and Diabetes Mellitus on Family History of Appled Mathematcal Scences, Vol. 7, 2013, no. 41, 2047-2053 HIKARI Ltd, www.m-hkar.com Modelng Mult Layer Feed-forward Neural Network Model on the Influence of Hypertenson and Dabetes Melltus on Famly

More information

Copy Number Variation Methods and Data

Copy Number Variation Methods and Data Copy Number Varaton Methods and Data Copy number varaton (CNV) Reference Sequence ACCTGCAATGAT TAAGCCCGGG TTGCAACGTTAGGCA Populaton ACCTGCAATGAT TAAGCCCGGG TTGCAACGTTAGGCA ACCTGCAATGAT TTGCAACGTTAGGCA

More information

Encoding processes, in memory scanning tasks

Encoding processes, in memory scanning tasks vlemory & Cognton 1976,4 (5), 501 506 Encodng processes, n memory scannng tasks JEFFREY O. MILLER and ROBERT G. PACHELLA Unversty of Mchgan, Ann Arbor, Mchgan 48101, Three experments are presented that

More information

Project title: Mathematical Models of Fish Populations in Marine Reserves

Project title: Mathematical Models of Fish Populations in Marine Reserves Applcaton for Fundng (Malaspna Research Fund) Date: November 0, 2005 Project ttle: Mathematcal Models of Fsh Populatons n Marne Reserves Dr. Lev V. Idels Unversty College Professor Mathematcs Department

More information

ARTICLE IN PRESS Neuropsychologia xxx (2010) xxx xxx

ARTICLE IN PRESS Neuropsychologia xxx (2010) xxx xxx Neuropsychologa xxx (200) xxx xxx Contents lsts avalable at ScenceDrect Neuropsychologa journal homepage: www.elsever.com/locate/neuropsychologa Storage and bndng of object features n vsual workng memory

More information

International Journal of Emerging Technologies in Computational and Applied Sciences (IJETCAS)

International Journal of Emerging Technologies in Computational and Applied Sciences (IJETCAS) Internatonal Assocaton of Scentfc Innovaton and Research (IASIR (An Assocaton Unfyng the Scences, Engneerng, and Appled Research Internatonal Journal of Emergng Technologes n Computatonal and Appled Scences

More information

The Importance of Being Marginal: Gender Differences in Generosity 1

The Importance of Being Marginal: Gender Differences in Generosity 1 The Importance of Beng Margnal: Gender Dfferences n Generosty 1 Stefano DellaVgna, John A. Lst, Ulrke Malmender, and Gautam Rao Forthcomng, Amercan Economc Revew Papers and Proceedngs, May 2013 Abstract

More information

Physical Model for the Evolution of the Genetic Code

Physical Model for the Evolution of the Genetic Code Physcal Model for the Evoluton of the Genetc Code Tatsuro Yamashta Osamu Narkyo Department of Physcs, Kyushu Unversty, Fukuoka 8-856, Japan Abstract We propose a physcal model to descrbe the mechansms

More information

Price linkages in value chains: methodology

Price linkages in value chains: methodology Prce lnkages n value chans: methodology Prof. Trond Bjorndal, CEMARE. Unversty of Portsmouth, UK. and Prof. José Fernández-Polanco Unversty of Cantabra, Span. FAO INFOSAMAK Tangers, Morocco 14 March 2012

More information

Joint Modelling Approaches in diabetes research. Francisco Gude Clinical Epidemiology Unit, Hospital Clínico Universitario de Santiago

Joint Modelling Approaches in diabetes research. Francisco Gude Clinical Epidemiology Unit, Hospital Clínico Universitario de Santiago Jont Modellng Approaches n dabetes research Clncal Epdemology Unt, Hosptal Clínco Unverstaro de Santago Outlne 1 Dabetes 2 Our research 3 Some applcatons Dabetes melltus Is a serous lfe-long health condton

More information

Hierarchical Prediction Errors in Midbrain and Basal Forebrain during Sensory Learning

Hierarchical Prediction Errors in Midbrain and Basal Forebrain during Sensory Learning Artcle Herarchcal Predcton Errors n Mdbran and Basal Forebran durng Sensory Learnng Sandra Iglesas, 1,2, * Chrstoph Mathys, 1,2 Kay H. Brodersen, 1,2 Lars Kasper, 1,2 Marco Pccrell, 2 Hanneke E.M. den

More information

310 Int'l Conf. Par. and Dist. Proc. Tech. and Appl. PDPTA'16

310 Int'l Conf. Par. and Dist. Proc. Tech. and Appl. PDPTA'16 310 Int'l Conf. Par. and Dst. Proc. Tech. and Appl. PDPTA'16 Akra Sasatan and Hrosh Ish Graduate School of Informaton and Telecommuncaton Engneerng, Toka Unversty, Mnato, Tokyo, Japan Abstract The end-to-end

More information

Modeling the Survival of Retrospective Clinical Data from Prostate Cancer Patients in Komfo Anokye Teaching Hospital, Ghana

Modeling the Survival of Retrospective Clinical Data from Prostate Cancer Patients in Komfo Anokye Teaching Hospital, Ghana Internatonal Journal of Appled Scence and Technology Vol. 5, No. 6; December 2015 Modelng the Survval of Retrospectve Clncal Data from Prostate Cancer Patents n Komfo Anokye Teachng Hosptal, Ghana Asedu-Addo,

More information

Integration of sensory information within touch and across modalities

Integration of sensory information within touch and across modalities Integraton of sensory nformaton wthn touch and across modaltes Marc O. Ernst, Jean-Perre Brescan, Knut Drewng & Henrch H. Bülthoff Max Planck Insttute for Bologcal Cybernetcs 72076 Tübngen, Germany marc.ernst@tuebngen.mpg.de

More information

Appendix F: The Grant Impact for SBIR Mills

Appendix F: The Grant Impact for SBIR Mills Appendx F: The Grant Impact for SBIR Mlls Asmallsubsetofthefrmsnmydataapplymorethanonce.Ofthe7,436applcant frms, 71% appled only once, and a further 14% appled twce. Wthn my data, seven companes each submtted

More information

A Mathematical Model of the Cerebellar-Olivary System II: Motor Adaptation Through Systematic Disruption of Climbing Fiber Equilibrium

A Mathematical Model of the Cerebellar-Olivary System II: Motor Adaptation Through Systematic Disruption of Climbing Fiber Equilibrium Journal of Computatonal Neuroscence 5, 71 90 (1998) c 1998 Kluwer Academc Publshers. Manufactured n The Netherlands. A Mathematcal Model of the Cerebellar-Olvary System II: Motor Adaptaton Through Systematc

More information

THE NATURAL HISTORY AND THE EFFECT OF PIVMECILLINAM IN LOWER URINARY TRACT INFECTION.

THE NATURAL HISTORY AND THE EFFECT OF PIVMECILLINAM IN LOWER URINARY TRACT INFECTION. MET9401 SE 10May 2000 Page 13 of 154 2 SYNOPSS MET9401 SE THE NATURAL HSTORY AND THE EFFECT OF PVMECLLNAM N LOWER URNARY TRACT NFECTON. L A study of the natural hstory and the treatment effect wth pvmecllnam

More information

The High way code. the guide to safer, more enjoyable drug use. (lsd / magic mushrooms)

The High way code. the guide to safer, more enjoyable drug use. (lsd / magic mushrooms) The Hgh way code the gude to safer, more enjoyable drug use (lsd / magc mushrooms) ntroducng the GDS Hgh Way Code GDS knows pleasure drves drug use, not the avodance of harm. As far as we know no gude

More information

The Limits of Individual Identification from Sample Allele Frequencies: Theory and Statistical Analysis

The Limits of Individual Identification from Sample Allele Frequencies: Theory and Statistical Analysis The Lmts of Indvdual Identfcaton from Sample Allele Frequences: Theory and Statstcal Analyss Peter M. Vsscher 1 *, Wllam G. Hll 2 1 Queensland Insttute of Medcal Research, Brsbane, Australa, 2 Insttute

More information

Non-linear Multiple-Cue Judgment Tasks

Non-linear Multiple-Cue Judgment Tasks Non-lnear Multple-Cue Tasks Anna-Carn Olsson (anna-carn.olsson@psy.umu.se) Department of Psychology, Umeå Unversty SE-09 87, Umeå, Sweden Tommy Enqvst (tommy.enqvst@psyk.uu.se) Department of Psychology,

More information

II. Key stimuli in avoidance learning

II. Key stimuli in avoidance learning Anmal Learnng & Behavor 1986, 14 (/), 101-109 Ethologcal analyss of predator avodance by the paradse fsh (Macropodus operculars L.): II. Key stmul n avodance learnng V. CSANYI L. Eotvos Unversty of Budapest.

More information

Richard Williams Notre Dame Sociology Meetings of the European Survey Research Association Ljubljana,

Richard Williams Notre Dame Sociology   Meetings of the European Survey Research Association Ljubljana, Rchard Wllams Notre Dame Socology rwllam@nd.edu http://www.nd.edu/~rwllam Meetngs of the European Survey Research Assocaton Ljubljana, Slovena July 19, 2013 Comparng Logt and Probt Coeffcents across groups

More information

Clinging to Beliefs: A Constraint-satisfaction Model

Clinging to Beliefs: A Constraint-satisfaction Model Clngng to Belefs: A Constrant-satsfacton Model Thomas R. Shultz (shultz@psych.mcgll.ca) Department of Psychology; McGll Unversty Montreal, QC H3C 1B1 Canada Jacques A. Katz (jakatz@cnbc.cmu.edu) Department

More information

The High way code. the guide to safer, more enjoyable drug use. (alcohol)

The High way code. the guide to safer, more enjoyable drug use. (alcohol) The Hgh way code the gude to safer, more enjoyable drug use (alcohol) ntroducng the GDS Hgh Way Code GDS knows pleasure drves drug use, not the avodance of harm. As far as we know no gude has ever outlned

More information

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi HIV/AIDS-related Expectatons and Rsky Sexual Behavor n Malaw Adelne Delavande Unversty of Essex and RAND Corporaton Hans-Peter Kohler Unversty of Pennsylvanna January 202 Abstract We use probablstc expectatons

More information

Prototypes in the Mist: The Early Epochs of Category Learning

Prototypes in the Mist: The Early Epochs of Category Learning Journal of Expermental Psychology: Learnng, Memory, and Cognton 1998, Vol. 24, No. 6, 1411-1436 Copyrght 1998 by the Amercan Psychologcal Assocaton, Inc. 0278-7393/98/S3.00 Prototypes n the Mst: The Early

More information

J. H. Rohrer, S. H. Baron, E. L. Hoffman, D. V. Swander

J. H. Rohrer, S. H. Baron, E. L. Hoffman, D. V. Swander 2?Hr a! A Report of Research on o ^^ -^~" r" THE STABILITY OF AUTOKINETIC JUDGMENTS J. H. Rohrer, S. H. Baron, E. L. Hoffman, D. V. Swander A techncal report made under ONR Contract Nonr-475(01) between

More information

Alma Mater Studiorum Università di Bologna DOTTORATO DI RICERCA IN METODOLOGIA STATISTICA PER LA RICERCA SCIENTIFICA

Alma Mater Studiorum Università di Bologna DOTTORATO DI RICERCA IN METODOLOGIA STATISTICA PER LA RICERCA SCIENTIFICA Alma Mater Studorum Unverstà d Bologna DOTTORATO DI RICERCA IN METODOLOGIA STATISTICA PER LA RICERCA SCIENTIFICA Cclo XXVII Settore Concorsuale d afferenza: 13/D1 Settore Scentfco dscplnare: SECS-S/02

More information

Using Past Queries for Resource Selection in Distributed Information Retrieval

Using Past Queries for Resource Selection in Distributed Information Retrieval Purdue Unversty Purdue e-pubs Department of Computer Scence Techncal Reports Department of Computer Scence 2011 Usng Past Queres for Resource Selecton n Dstrbuted Informaton Retreval Sulleyman Cetntas

More information

ALMALAUREA WORKING PAPERS no. 9

ALMALAUREA WORKING PAPERS no. 9 Snce 1994 Inter-Unversty Consortum Connectng Unverstes, the Labour Market and Professonals AlmaLaurea Workng Papers ISSN 2239-9453 ALMALAUREA WORKING PAPERS no. 9 September 211 Propensty Score Methods

More information

The High way code. the guide to safer, more enjoyable drug use [GHB] Who developed it?

The High way code. the guide to safer, more enjoyable drug use [GHB] Who developed it? The Hgh way code the gude to safer, more enjoyable drug use [] Who developed t? What s t? The frst gude to safer drug use voted for by people who take drugs. How was t was developed? GDS asked loads of

More information

An Introduction to Modern Measurement Theory

An Introduction to Modern Measurement Theory An Introducton to Modern Measurement Theory Ths tutoral was wrtten as an ntroducton to the bascs of tem response theory (IRT) modelng and ts applcatons to health outcomes measurement for the Natonal Cancer

More information

A Linear Regression Model to Detect User Emotion for Touch Input Interactive Systems

A Linear Regression Model to Detect User Emotion for Touch Input Interactive Systems 2015 Internatonal Conference on Affectve Computng and Intellgent Interacton (ACII) A Lnear Regresson Model to Detect User Emoton for Touch Input Interactve Systems Samt Bhattacharya Dept of Computer Scence

More information

Unobserved Heterogeneity and the Statistical Analysis of Highway Accident Data

Unobserved Heterogeneity and the Statistical Analysis of Highway Accident Data Unobserved Heterogenety and the Statstcal Analyss of Hghway Accdent Data Fred L. Mannerng Professor of Cvl and Envronmental Engneerng Courtesy Department of Economcs Unversty of South Florda 4202 E. Fowler

More information

Disconnection of the Amygdala from Visual Association Cortex Impairs Visual Reward-Association Learning in Monkeys

Disconnection of the Amygdala from Visual Association Cortex Impairs Visual Reward-Association Learning in Monkeys The Journal of Neuroscence, September 1988, 8(9): 31443150 Dsconnecton of the Amygdala from Vsual Assocaton Cortex mpars Vsual Reward-Assocaton Learnng n Monkeys E. A. Gaffan, Davd Gaffan, and Susan Harrson

More information

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi Unversty of Pennsylvana ScholarlyCommons PSC Workng Paper Seres 7-29-20 HIV/AIDS-related Expectatons and Rsky Sexual Behavor n Malaw Adelne Delavande RAND Corporaton, Nova School of Busness and Economcs

More information

Optimal Planning of Charging Station for Phased Electric Vehicle *

Optimal Planning of Charging Station for Phased Electric Vehicle * Energy and Power Engneerng, 2013, 5, 1393-1397 do:10.4236/epe.2013.54b264 Publshed Onlne July 2013 (http://www.scrp.org/ournal/epe) Optmal Plannng of Chargng Staton for Phased Electrc Vehcle * Yang Gao,

More information

VALIDATION TOOL THE SETTING OF THE COMMUNITY PHARMACY

VALIDATION TOOL THE SETTING OF THE COMMUNITY PHARMACY #VT01-1 VALIDATION TOOL THE SETTING OF THE COMMUNITY PHARMACY The pharmacy settng can alter the qualty of patent care and may nfluence patent satsfacton. An approprate settng may ncrease the probablty

More information

Drug Prescription Behavior and Decision Support Systems

Drug Prescription Behavior and Decision Support Systems Drug Prescrpton Behavor and Decson Support Systems ABSTRACT Adverse drug events plague the outcomes of health care servces. In ths research, we propose a clncal learnng model that ncorporates the use of

More information

The High way code. the guide to safer, more enjoyable drug use. [cannabis] Who developed it?

The High way code. the guide to safer, more enjoyable drug use. [cannabis] Who developed it? The Hgh way code the gude to safer, more enjoyable drug use [cannabs] Who developed t? What s t? The frst gude to safer drug use voted for by people who take drugs. How was t was developed? GDS asked loads

More information

Study and Comparison of Various Techniques of Image Edge Detection

Study and Comparison of Various Techniques of Image Edge Detection Gureet Sngh et al Int. Journal of Engneerng Research Applcatons RESEARCH ARTICLE OPEN ACCESS Study Comparson of Varous Technques of Image Edge Detecton Gureet Sngh*, Er. Harnder sngh** *(Department of

More information

Fast Algorithm for Vectorcardiogram and Interbeat Intervals Analysis: Application for Premature Ventricular Contractions Classification

Fast Algorithm for Vectorcardiogram and Interbeat Intervals Analysis: Application for Premature Ventricular Contractions Classification Fast Algorthm for Vectorcardogram and Interbeat Intervals Analyss: Applcaton for Premature Ventrcular Contractons Classfcaton Irena Jekova, Vessela Krasteva Centre of Bomedcal Engneerng Prof. Ivan Daskalov

More information

Economic crisis and follow-up of the conditions that define metabolic syndrome in a cohort of Catalonia,

Economic crisis and follow-up of the conditions that define metabolic syndrome in a cohort of Catalonia, Economc crss and follow-up of the condtons that defne metabolc syndrome n a cohort of Catalona, 2005-2012 Laa Maynou 1,2,3, Joan Gl 4, Gabrel Coll-de-Tuero 5,2, Ton Mora 6, Carme Saurna 1,2, Anton Scras

More information

RENAL FUNCTION AND ACE INHIBITORS IN RENAL ARTERY STENOSISA/adbon et al. 651

RENAL FUNCTION AND ACE INHIBITORS IN RENAL ARTERY STENOSISA/adbon et al. 651 Downloaded from http://ahajournals.org by on January, 209 RENAL FUNCTION AND INHIBITORS IN RENAL ARTERY STENOSISA/adbon et al. 65 Downloaded from http://ahajournals.org by on January, 209 Patents and Methods

More information

Reconciling Simplicity and Likelihood Principles in Perceptual Organization

Reconciling Simplicity and Likelihood Principles in Perceptual Organization Psychologcal Revew Copyrght 1996 by the Amercan Psychologcal Assocaton, Inc. 1996. Vol. 103, No. 3, 566-581 0033-295X/96/$3.00 Reconclng Smplcty and Lkelhood Prncples n Perceptual Organzaton Nck Chater

More information

What Determines Attitude Improvements? Does Religiosity Help?

What Determines Attitude Improvements? Does Religiosity Help? Internatonal Journal of Busness and Socal Scence Vol. 4 No. 9; August 2013 What Determnes Atttude Improvements? Does Relgosty Help? Madhu S. Mohanty Calforna State Unversty-Los Angeles Los Angeles, 5151

More information

Active Affective State Detection and User Assistance with Dynamic Bayesian Networks. Xiangyang Li, Qiang Ji

Active Affective State Detection and User Assistance with Dynamic Bayesian Networks. Xiangyang Li, Qiang Ji Actve Affectve State Detecton and User Assstance wth Dynamc Bayesan Networks Xangyang L, Qang J Electrcal, Computer, and Systems Engneerng Department Rensselaer Polytechnc Insttute, 110 8th Street, Troy,

More information

Balanced Query Methods for Improving OCR-Based Retrieval

Balanced Query Methods for Improving OCR-Based Retrieval Balanced Query Methods for Improvng OCR-Based Retreval Kareem Darwsh Electrcal and Computer Engneerng Dept. Unversty of Maryland, College Park College Park, MD 20742 kareem@glue.umd.edu Douglas W. Oard

More information

The Influence of the Isomerization Reactions on the Soybean Oil Hydrogenation Process

The Influence of the Isomerization Reactions on the Soybean Oil Hydrogenation Process Unversty of Belgrade From the SelectedWorks of Zeljko D Cupc 2000 The Influence of the Isomerzaton Reactons on the Soybean Ol Hydrogenaton Process Zeljko D Cupc, Insttute of Chemstry, Technology and Metallurgy

More information

Sparse Representation of HCP Grayordinate Data Reveals. Novel Functional Architecture of Cerebral Cortex

Sparse Representation of HCP Grayordinate Data Reveals. Novel Functional Architecture of Cerebral Cortex 1 Sparse Representaton of HCP Grayordnate Data Reveals Novel Functonal Archtecture of Cerebral Cortex X Jang 1, Xang L 1, Jngle Lv 2,1, Tuo Zhang 2,1, Shu Zhang 1, Le Guo 2, Tanmng Lu 1* 1 Cortcal Archtecture

More information

(From the Gastroenterology Division, Cornell University Medical College, New York 10021)

(From the Gastroenterology Division, Cornell University Medical College, New York 10021) ROLE OF HEPATIC ANION-BINDING PROTEIN IN BROMSULPHTHALEIN CONJUGATION* BY N. KAPLOWITZ, I. W. PERC -ROBB,~ ANn N. B. JAVITT (From the Gastroenterology Dvson, Cornell Unversty Medcal College, New York 10021)

More information

Delving Beneath the Covers: Examining Children s Literature

Delving Beneath the Covers: Examining Children s Literature Chmamanda Ngoz Adche: The danger of a sngle story Personal Bases Delvng Beneath the Covers: Examnng Chldren s Lterature Hdden Messages of Gender, Ablty, Dversty, Body Image Commercalsm, Power & Prvlege

More information

Bimodal Bidding in Experimental All-Pay Auctions

Bimodal Bidding in Experimental All-Pay Auctions Bmodal Bddng n Expermental All-Pay Auctons Chrstane Ernst and Chrstan Thön August 2009 Dscusson Paper no. 2009-25 Department of Economcs Unversty of St. Gallen Edtor: Publsher: Electronc Publcaton: Martna

More information

Single-Case Designs and Clinical Biofeedback Experimentation

Single-Case Designs and Clinical Biofeedback Experimentation Bofeedback and Self-Regulaton, VoL 2, No. 3, 1977 Sngle-Case Desgns and Clncal Bofeedback Expermentaton Davd H. Barow: Brown Unversty and Butler Hosptal Edward B. Blanchard Unversty of Tennessee Medcal

More information

National Polyp Study data: evidence for regression of adenomas

National Polyp Study data: evidence for regression of adenomas 5 Natonal Polyp Study data: evdence for regresson of adenomas 78 Chapter 5 Abstract Objectves The data of the Natonal Polyp Study, a large longtudnal study on survellance of adenoma patents, s used for

More information

A comparison of statistical methods in interrupted time series analysis to estimate an intervention effect

A comparison of statistical methods in interrupted time series analysis to estimate an intervention effect Peer revew stream A comparson of statstcal methods n nterrupted tme seres analyss to estmate an nterventon effect a,b, J.J.J., Walter c, S., Grzebeta a, R. & Olver b, J. a Transport and Road Safety, Unversty

More information

Investigation of zinc oxide thin film by spectroscopic ellipsometry

Investigation of zinc oxide thin film by spectroscopic ellipsometry VNU Journal of Scence, Mathematcs - Physcs 24 (2008) 16-23 Investgaton of znc oxde thn flm by spectroscopc ellpsometry Nguyen Nang Dnh 1, Tran Quang Trung 2, Le Khac Bnh 2, Nguyen Dang Khoa 2, Vo Th Ma

More information

Prediction of Total Pressure Drop in Stenotic Coronary Arteries with Their Geometric Parameters

Prediction of Total Pressure Drop in Stenotic Coronary Arteries with Their Geometric Parameters Tenth Internatonal Conference on Computatonal Flud Dynamcs (ICCFD10), Barcelona, Span, July 9-13, 2018 ICCFD10-227 Predcton of Total Pressure Drop n Stenotc Coronary Arteres wth Ther Geometrc Parameters

More information

DS May 31,2012 Commissioner, Development. Services Department SPA June 7,2012

DS May 31,2012 Commissioner, Development. Services Department SPA June 7,2012 . h,oshawa o Report To: From: Subject: Development Servces Commttee Item: Date of Report: DS-12-189 May 31,2012 Commssoner, Development Fle: Date of Meetng: Servces Department SPA-2010-09 June 7,2012 Applcaton

More information

NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 TIANHONG ZHOU

NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 TIANHONG ZHOU NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 by TIANHONG ZHOU B.S., Chna Agrcultural Unversty, 2003 M.S., Chna Agrcultural Unversty, 2006 A THESIS submtted n partal fulfllment of the requrements

More information

THIS IS AN OFFICIAL NH DHHS HEALTH ALERT

THIS IS AN OFFICIAL NH DHHS HEALTH ALERT THIS IS AN OFFICIAL NH DHHS HEALTH ALERT Dstrbuted by the NH Health Alert Network Health.Alert@dhhs.nh.gov August 26, 2016 1430 EDT (2:30 PM EDT) NH-HAN 20160826 Recommendatons for Accurate Dagnoss of

More information

EFFECTS OF FEEDBACK CONTROL ON SLOW CORTICAL POTENTIALS AND RANDOM EVENTS

EFFECTS OF FEEDBACK CONTROL ON SLOW CORTICAL POTENTIALS AND RANDOM EVENTS Hnterberger, Houtkooper, & Kotchoubey EFFECTS OF FEEDBACK CONTROL ON SLOW CORTICAL POTENTIALS AND RANDOM EVENTS Thlo Hnterberger 1, Joop M. Houtkooper 2, & Bors Kotchoubey 1 1 Insttute of Medcal Psychology

More information

Mathematical model of fish schooling behaviour in a set-net

Mathematical model of fish schooling behaviour in a set-net ICES Journal of Marne Scence, 61: 114e13 (004) do:10.1016/j.cesjms.004.07.009 Mathematcal model of fsh schoolng behavour n a set-net Tsutomu Takag, Yutaka Mortom, Jyun Iwata, Hrosh Nakamne, and Nobuo Sannomya

More information

Ependymal cells Cilia on one surface Movement of material or fluid over surface of the cell

Ependymal cells Cilia on one surface Movement of material or fluid over surface of the cell 2004 Bology GA 1: Wrtten examnaton 1 SPECIFIC INFMATION Secton A Multple-choce Ths table ndcates the approxmate percentage of students choosng each dstractor. The correct answer s the shaded alternatve.

More information

EVALUATION OF BULK MODULUS AND RING DIAMETER OF SOME TELLURITE GLASS SYSTEMS

EVALUATION OF BULK MODULUS AND RING DIAMETER OF SOME TELLURITE GLASS SYSTEMS Chalcogende Letters Vol. 12, No. 2, February 2015, p. 67-74 EVALUATION OF BULK MODULUS AND RING DIAMETER OF SOME TELLURITE GLASS SYSTEMS R. EL-MALLAWANY a*, M.S. GAAFAR b, N. VEERAIAH c a Physcs Dept.,

More information

Introduction ORIGINAL RESEARCH

Introduction ORIGINAL RESEARCH ORIGINAL RESEARCH Assessng the Statstcal Sgnfcance of the Acheved Classfcaton Error of Classfers Constructed usng Serum Peptde Profles, and a Prescrpton for Random Samplng Repeated Studes for Massve Hgh-Throughput

More information

The High way code. the guide to safer, more enjoyable drug use. (ketamine)

The High way code. the guide to safer, more enjoyable drug use. (ketamine) The Hgh way code the gude to safer, more enjoyable drug use (ketamne) ntroducng the GDS Hgh Way Code GDS knows pleasure drves drug use, not the avodance of harm. As far as we know no gude has ever outlned

More information

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and Ths artcle appeared n a journal publshed by Elsever. The attached copy s furnshed to the author for nternal non-commercal research and educaton use, ncludng for nstructon at the authors nsttuton and sharng

More information

The Effect of Fish Farmers Association on Technical Efficiency: An Application of Propensity Score Matching Analysis

The Effect of Fish Farmers Association on Technical Efficiency: An Application of Propensity Score Matching Analysis The Effect of Fsh Farmers Assocaton on Techncal Effcency: An Applcaton of Propensty Score Matchng Analyss Onumah E. E, Esslfe F. L, and Asumng-Brempong, S 15 th July, 2016 Background and Motvaton Outlne

More information

Concentration of teicoplanin in the serum of adults with end stage chronic renal failure undergoing treatment for infection

Concentration of teicoplanin in the serum of adults with end stage chronic renal failure undergoing treatment for infection Journal of Antmcrobal Chemotherapy (1996) 37, 117-121 Concentraton of tecoplann n the serum of adults wth end stage chronc renal falure undergong treatment for nfecton A. MercateUo'*, K. Jaber*, D. Hfflare-Buys*,

More information

THE NORMAL DISTRIBUTION AND Z-SCORES COMMON CORE ALGEBRA II

THE NORMAL DISTRIBUTION AND Z-SCORES COMMON CORE ALGEBRA II Name: Date: THE NORMAL DISTRIBUTION AND Z-SCORES COMMON CORE ALGEBRA II The normal dstrbuton can be used n ncrements other than half-standard devatons. In fact, we can use ether our calculators or tables

More information

Inverted-U and Inverted-J Effects in Self-Referenced Decisions

Inverted-U and Inverted-J Effects in Self-Referenced Decisions Inverted-U and Inverted-J Effects n Self-Referenced Decsons Kenpe SHIINA (shnaatwaseda.jp) Department of Educatonal Psychology, Waseda Unversty, Tokyo, Japan Abstract Ratng one s own personalty trats s

More information

NeuroImage. Decoded fmri neurofeedback can induce bidirectional confidence changes within single participants

NeuroImage. Decoded fmri neurofeedback can induce bidirectional confidence changes within single participants NeuroImage 149 (2017) 323 337 Contents lsts avalable at ScenceDrect NeuroImage journal homepage: www.elsever.com/locate/neuromage Decoded fmri neurofeedback can nduce bdrectonal confdence changes wthn

More information

A Geometric Approach To Fully Automatic Chromosome Segmentation

A Geometric Approach To Fully Automatic Chromosome Segmentation A Geometrc Approach To Fully Automatc Chromosome Segmentaton Shervn Mnaee ECE Department New York Unversty Brooklyn, New York, USA shervn.mnaee@nyu.edu Mehran Fotouh Computer Engneerng Department Sharf

More information

Biased Perceptions of Income Distribution and Preferences for Redistribution: Evidence from a Survey Experiment

Biased Perceptions of Income Distribution and Preferences for Redistribution: Evidence from a Survey Experiment DISCUSSION PAPER SERIES IZA DP No. 5699 Based Perceptons of Income Dstrbuton and Preferences for Redstrbuton: Evdence from a Survey Experment Gullermo Cruces Rcardo Pérez Trugla Martn Tetaz May 2011 Forschungsnsttut

More information

HYPEIIGLTCAEMIA AS A MENDELIAN P~ECESSIVE CHAI~ACTEP~ IN MICE.

HYPEIIGLTCAEMIA AS A MENDELIAN P~ECESSIVE CHAI~ACTEP~ IN MICE. HYPEGLTCAEMA AS A MENDELAN P~ECESSVE CHA~ACTEP~ N MCE. BY P. J. CAM~CDGE, M.D. (LEND.), 32 Nottngham Place, Ma~'y~ebone, London, W, 1, AND H. A. H. {OWAZD, B.So. (Lol, m.). h'~ the course of an nvestgaton

More information

Reconstruction of gene regulatory network of colon cancer using information theoretic approach

Reconstruction of gene regulatory network of colon cancer using information theoretic approach Reconstructon of gene regulatory network of colon cancer usng nformaton theoretc approach Khald Raza #1, Rafat Parveen * # Department of Computer Scence Jama Mlla Islama (Central Unverst, New Delh-11005,

More information

NHS Outcomes Framework

NHS Outcomes Framework NHS Outcomes Framework Doman 1 Preventng people from dyng prematurely Indcator Specfcatons Verson: 1.21 Date: May 2018 Author: Clncal Indcators Team NHS Outcomes Framework: Doman 1 Preventng people from

More information

Lateral Transfer Data Report. Principal Investigator: Andrea Baptiste, MA, OT, CIE Co-Investigator: Kay Steadman, MA, OTR, CHSP. Executive Summary:

Lateral Transfer Data Report. Principal Investigator: Andrea Baptiste, MA, OT, CIE Co-Investigator: Kay Steadman, MA, OTR, CHSP. Executive Summary: Samar tmed c ali ndus t r esi nc 55Fl em ngdr ve, Un t#9 Cambr dge, ON. N1T2A9 T el. 18886582206 Ema l. nf o@s amar t r ol l boar d. c om www. s amar t r ol l boar d. c om Lateral Transfer Data Report

More information

THE ROLE OF FRONTAL AND PARIETAL CORTEX IN COGNITIVE PROCESSING

THE ROLE OF FRONTAL AND PARIETAL CORTEX IN COGNITIVE PROCESSING .,......._._,,,,-_.._---_..._-_..._.,, ---;, ----,..,.-,..,.-. - _---_....!)-)oo " Bran (1978), 101,607-633 THE ROLE OF FRONTAL AND PARETAL CORTEX N COGNTVE PROCESSNG TESTS OF SPATAL AND SEQUENCE FUNCTONS,

More information

Experiment. shows the materials used in the study and, for each item, the percentage of choices for the matching cause.

Experiment. shows the materials used in the study and, for each item, the percentage of choices for the matching cause. J ameson,j.,& Gent ner,d.( 2008).Causalst at usandexpl anat or ygoodness ncat egor zat on.i nb.c.love,k.mcrae,& V.M.Sl out sky( Eds. ), Pr oceed ngsoft he30t hannualconf er enceoft hecogn t vesc encesoc

More information

WHO S ASSESSMENT OF HEALTH CARE INDUSTRY PERFORMANCE: RATING THE RANKINGS

WHO S ASSESSMENT OF HEALTH CARE INDUSTRY PERFORMANCE: RATING THE RANKINGS WHO S ASSESSMENT OF HEALTH CARE INDUSTRY PERFORMANCE: RATING THE RANKINGS ELLIOTT PARKER and JEANNE WENDEL * Department of Economcs, Unversty of Nevada, Reno, NV, USA SUMMARY Ths paper examnes the econometrc

More information

AN ENHANCED GAGS BASED MTSVSL LEARNING TECHNIQUE FOR CANCER MOLECULAR PATTERN PREDICTION OF CANCER CLASSIFICATION

AN ENHANCED GAGS BASED MTSVSL LEARNING TECHNIQUE FOR CANCER MOLECULAR PATTERN PREDICTION OF CANCER CLASSIFICATION www.arpapress.com/volumes/vol8issue2/ijrras_8_2_02.pdf AN ENHANCED GAGS BASED MTSVSL LEARNING TECHNIQUE FOR CANCER MOLECULAR PATTERN PREDICTION OF CANCER CLASSIFICATION I. Jule 1 & E. Krubakaran 2 1 Department

More information

DECREASING SYMPTOMS IN INTERSTITIAL CYSTITIS PATIENTS: PENTOSAN POLYSULFATE VS. SACRAL NEUROMODULATION. A Research Project by. Katy D.

DECREASING SYMPTOMS IN INTERSTITIAL CYSTITIS PATIENTS: PENTOSAN POLYSULFATE VS. SACRAL NEUROMODULATION. A Research Project by. Katy D. DECREASING SYMPTOMS IN INTERSTITIAL CYSTITIS PATIENTS: PENTOSAN POLYSULFATE VS. SACRAL NEUROMODULATION. A Research Project by Katy D. Prce Bachelor of General Studes, Unversty of Kansas, 2005 Submtted

More information

RHEUMATOID ARTHRITIS PATIENTS CANNOT ACCURATELY REPORT SIGNS OF INFLAMMATORY ACTIVITY

RHEUMATOID ARTHRITIS PATIENTS CANNOT ACCURATELY REPORT SIGNS OF INFLAMMATORY ACTIVITY Brtsh Journal of Rheumatology 995;4:547-55 RHEUMATOID ARTHRITIS PATIENTS CANNOT ACCURATELY REPORT SIGNS OF INFLAMMATORY ACTIVITY S. E. HEWLETT, J. HAYNES, L. SHEPSTONE and J. R. KIRWAN Unversty of Brstol

More information

Perceptual image quality: Effects of tone characteristics

Perceptual image quality: Effects of tone characteristics Journal of Electronc Imagng 14(2), 023003 (Apr Jun 2005) Perceptual mage qualty: Effects of tone characterstcs Peter B. Delahunt Exponent Inc. 149 Commonwealth Drve Menlo Park, Calforna 94025 Xueme Zhang

More information

Toward a Unified Model of Attention in Associative Learning

Toward a Unified Model of Attention in Associative Learning Journal of Mathematcal Psychology 45, 812863 (2001) do:10.1006jmps.2000.1354, avalable onlne at http:www.dealbrary.com on Toward a Unfed Model of Attenton n Assocatve Learnng John K. Kruschke Indana Unversty

More information

Cutaneous and Kinaesthetic Perception of Traversed Distance

Cutaneous and Kinaesthetic Perception of Traversed Distance Cutaneous and Knaesthetc Percepton of Traversed Dstance Wouter M. Bergmann Test L. Martjn A. van der Hoff Astrd M. L. Kappers Helmholtz Insttute, Utrecht Unversty, The Netherlands ABSTRACT Dscrmnaton thresholds

More information

Length of Hospital Stay After Acute Myocardial Infarction in the Myocardial Infarction Triage and Intervention (MITI) Project Registry

Length of Hospital Stay After Acute Myocardial Infarction in the Myocardial Infarction Triage and Intervention (MITI) Project Registry JACC Vol. 28, No. 2 287 CLINICAL STUDIES MYOCARDIAL INFARCTION Length of Hosptal Stay After Acute Myocardal Infarcton n the Myocardal Infarcton Trage and Interventon (MITI) Project Regstry NATHAN R. EVERY,

More information

Lymphoma Cancer Classification Using Genetic Programming with SNR Features

Lymphoma Cancer Classification Using Genetic Programming with SNR Features Lymphoma Cancer Classfcaton Usng Genetc Programmng wth SNR Features Jn-Hyuk Hong and Sung-Bae Cho Dept. of Computer Scence, Yonse Unversty, 134 Shnchon-dong, Sudaemoon-ku, Seoul 120-749, Korea hjnh@candy.yonse.ac.kr,

More information

Does Context Matter More for Hypothetical Than for Actual Contributions?

Does Context Matter More for Hypothetical Than for Actual Contributions? Dscusson Paper Seres March 2008 EfD DP 08-02 Does Context Matter More for Hypothetcal Than for Actual Contrbutons? Evdence from a Natural Feld Experment Francsco Alpzar, Fredrk Carlsson, and Olof Johansson-Stenman

More information

Jurnal Teknologi USING ASSOCIATION RULES TO STUDY PATTERNS OF MEDICINE USE IN THAI ADULT DEPRESSED PATIENTS. Full Paper

Jurnal Teknologi USING ASSOCIATION RULES TO STUDY PATTERNS OF MEDICINE USE IN THAI ADULT DEPRESSED PATIENTS. Full Paper Jurnal Teknolog USING ASSOCIATION RULES TO STUDY PATTERNS OF MEDICINE USE IN THAI ADULT DEPRESSED PATIENTS Chumpoonuch Sukontavaree, Verayuth Lertnattee * Faculty of Pharmacy, Slpakorn Unversty, Nakhon

More information

A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA

A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA Journal of Theoretcal and Appled Informaton Technology 2005 ongong JATIT & LLS ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA 1 SUNGMIN

More information

An expressive three-mode principal components model for gender recognition

An expressive three-mode principal components model for gender recognition Journal of Vson (4) 4, 36-377 http://journalofvson.org/4/5// 36 An expressve three-mode prncpal components model for gender recognton James W. Davs Hu Gao Department of Computer and Informaton Scence,

More information