{"title": "Modeling Midazolam's Effect on the Hippocampus and Recognition Memory", "book": "Advances in Neural Information Processing Systems", "page_first": 75, "page_last": 82, "abstract": "", "full_text": "Modeling Midazolam's Effect on the\n\n__H_il!Jlocampus and Recognition Memor!\n\nKenneth J'\" .I\\'lalJrnbeJ~2\n\nDepartn1ent of Psychology\n\nIndiana V'uiversity\n\nBloomington, IN' 47405\n\nRene Le!ele:nD~er2\n\nDepartment of rS'/cnOlCHIV\n\nIndiana University\n\nBloomington, IN 47405\nrzeelenb(~~indiana.edu\n\nDepartm.entsof Cognitive Science and Psychology\n\nRichard 1\\'1.. Sbiffrin\n\nIndiana 'University\n\nBloomington, TN' 47405\nshiffrin@indiaful.etiu\n\nAbstract\n\n'~1idazolam\n\nbenz.odiaze:pine\n\nThe\ncauses dense,but\ntemporary~ anterograde amnesia, similar to that produced\nby- hippocampal damage~Does the action of M'idazola:m on\nthe hippocanlpus cause less storage, or less accurate\nstorage, .of information in episodic. long-term menlory?- \\rVe\nused a sinlple variant of theREJv1. JD.odel [18] to fit data\ncollected. by IIirsbnla.n~Fisher, .IIenthorn,Arndt} and\nPassa.nnante [9] on the effects of Midazola.m, study time~\nand normative \\vQrd...frequenc:y on both yes-no and\nremember-k.novv recognition m.emory. That\nsimple\nstrength. 'model fit well \\\\tas cont.rary to the expectations of\n'flirshman et aLMore important,within the Bayesian based\nR.EM modeling frame\\vork, the data were consistentw'ith\nthe view that Midazolam causes less accurate storage~\nrather\nstorage, of infornlation in episodic\nmcm.ory..\n\nthan less\n\na:\n\n1 Introduction\n\n'Danlage to the hippocampus (and nearby regions), often caused by lesiclns1\nleaves normal cognitive function intact in the short\nincluding long\".tenn\nmemo-ry retrieval, but prevents learning of new1' inJornlat.ion.We have found a ,yay\nto begin to distinguish two alternative accounts for this lea.ming deficit: Does\ndamage cause less storage, or less accurate storagc1 of information in long-term\nepisodic menlQry\u00a3! We addressed this question by using the REM model of\nrecognition 'mC'mQry [18] to fit data collected by Hirshnlan and colleagues [9], vlho\ntested recognition memory in nornlalparticipants given either saline (control group)\nor Midazolam, a benzodiazepine that temporarily- causes anterograde amnesia with\neffects that generally' 'mimic those found after hippoca.mpa1 dan1age.\n\nterm.,\n\n\f2 Empirical findings\n\nThe participants in Hirshman et at [9] studied lists of \\~ords that \u00b7varied in\nnomlative. word frequency (Le., lo\\v-frequency vs. high.-frequency) and the amount of\ntime allocated for study (either not studied, or studied for 500, 1.200, or 2500 ms per\n\u00b7word)+ These variables are known to have a robust effect on rec.ognition memory in\nnornlal populations; Lo\\v-frequency (LF)\n\\vords are better recognized tllan high\u00b7\u00b7\nfrequency (FIT) \\\"rOrd5~ an.d a.n. increase in study tinle inJproves recognition perfbl1:11ance.\nIn. addi.tion~ the probability ofrespon.ding 'oldY to studied words (temJed hit rate~ or FfR) is\nhigher forL\u00b7F \\:vords than forHF '\\\u00a5ords, and t11e probability of responding 'old\u00b7 to\nunstu.died. \\\\lords (~ermed fa.tse alarm. rate, or FAR) is lo\\>ver for l,F \\vords than. tor HF\n'\\Tords. Th.is pattern of data is commonly kno\\vn as a ~l;mirror etIecf' [7].,\n\nIn. Hirshulan et al. [9], participants received either salin.e or l'vfidazolatn a11d\nthen studied a list of \u00b7words. A.fter a delay of about an hour they ,vere sho\\vn studied\n) and unstudied words Cnew1)'1 a.nd asked to give old-new recognition\nwords eoldt\nand. renlenlber/k'11o\\v judgments. The HR and F.AR. ii.ndin.gs are depicted in Figure 1\nas the large circles (tl.I1ed for l,F test\u00a5iords and un.filled for HF test '~i'ords). The\nresults fror.n the saline condition, given in the left panel, replicate tIle standard\neffects in the literature: In the figure; the points labeled with zero study time give\nFAR.s (for ne\u00b7';fl test itelns), and the other poin.ts give HRs (for old test items). Thus\n,ve see that the saline group exhibits better performance forL\u00b7F words al1d a rnirror\neffect: ForLF words~ FA\u00b7Rs are IO\\,l.ler an-dHRs are higheL The Midazolam group of\ncourse gave ]oVi-rer performance (see right pan.el). More critically, the pattern of\nresults differs from that for the sal ine group: The mirror effect was lostL,F \\vords\nproduced both.loweTF~A,..Rsand lower HR.s+\n\n0.8\n0.7\n0.6\n\n0.5\n0.4\n\n0.3\n\n0.2\n\n- , - - - - - - - - - - - - - - - , ....---------------,- 0.8\n\nHRsandFARs\n\nin Saline Condition\n\n....:::::::::::::::: .. ,.@\n\nHRsandFARs\n\nin MidazolamCondition\n\n0.7\n\n0.6 II\n~0.5 0\nc.\n0.4 '-'\n\n~ \"\n\n;\n\n--.- LF Data\n\u00b7\u00b7\u00b70\u00b7\u00b7\u00b7\u00b7 HFData\n--.- LF Fit\n.. \u00b7\u00b70\u00b7 .. \u00b7 HF Fit\n\n0.3\n- ' - - - . , - - - - , . - - - , - - - - - - r - - , - - - - , . . - - - - - t ' - - - - , - - - - . , - - - - - , - - . , - - - - , . - - - - r - - - -+ 0.2\n\no 500 1000 1500 .2000 25003000\n\no\n\nStudy Time (ms.)\n\n50010001500 200025003000\nStudy Time (ms.)\n\nFigure J.!.y\"\u00b7cs-no recognition data. from Hirshman ct at and predictions o:f aR.EMm.odeL\nZerQ U1S study time refers to 'new~ items so the data gives the false-alarnl rate (FAR). Data\nsh(nvn for non-zero study times give hit rates (lIR). Only the REM parameter & varies\nbchvecn the saline andm.idaz:olam conditions. 1~hc fi.ts are hased on 300 JVlonte (~arlo\nsimulations usinggLF ~ .325 t g= ..40\" gnr ~ ,,45; ~:= 16; 10 ~ 4, ~ ~ .8~ !!* ~ ..025, QS.l ,~ .77,\n~M1d = .25, CritQ/N '= .92. \u00b7LF\u00b7= low-frequency words and lIP = high-freque.ncy \\yords.\n\n\fThe participants also indicated 'Athether their \"old'\" judgrnents \\vere\nmade on the basis of '~rememberingHthe study event or on the basis of \"kno\\\u00a5ing\"\nthe v.rord \\vas studied even tb.ougb. tlley could not explicitly rernenlber the study\nevent [5]. Data: are sh.o\\\u00a5n in Figure 2. Of greatest interest (or present purposes,\n~~knowr~\" and \"rernelnber)' responses \\vere differently affected by the \\vord frequen.cy\nand the drug manipulations. In the 1Vlidazol.aul condition~ th.e conditional probability\nof a t'kno\\\\{'~ judgnlent (given an t'o]d:l~ response) was consistently higber tb.an that of\na \"remember'} judg1nent (for both HF and L,F Vi-i'ords). lVl\"oreover, these probabilities\n\u00b7were hardly affected by study timei A different pattern \\vas obtained in the Saline\ncondition.FQT HF words, the cQnditional probability of a '.tknO\\~l~1 judgment vvas\nhigher than that,of a \"rerrleulber\" judgmen.t~ but the difference decreased with study'\ntime..Final1y~ tor LF \\vords, the conditional probability of a \"'kn.o\\v\" Judgment vvas\nhigher than that of a Hrernenlber~' judgrrlent tor nonstudjed foj.ls~ but tor studied\ntargets the conditional probability of a. Hremernber\" judgrnent \\vas 11 igher tha.n that\nof a '~kn.ow\u00b7\" judgrnent The recognition an.d rerrlenlberlk\"u\\\u00a5 re\u00b7sults were interpreted\nby Hirshman et aL [91 to require a dual process account; in particular~ tlle authors\nargued against Hnlenlory strel1gtll~' accounts [4 t\nthe n1uin\nmessage of this tlote~ itvvitl be of som.e inteTest to m.emory theorists to n.ote that our\npresent results. sho\u00b7ws tIllS con.clusion to be in.correct.\n\n6~ 11]. Although not\n\n1.0 .,----,---------------,\n\n1.0 . . , - - - - - - - - - - - - - ,\n\n,-.. 0.8 D.\"...\n~ 0.6\n\nHF .. Saline\n\u00b7\u00b7\u00b7\"\"O:::::::::::::::::::::::~::::::::::::::::::::..\n\nI:\n\n0\n\n\\..\n~ 1.0 . , - - - - - - - - - - - - - - - - - ,\nE\n4) 0.8\n\nLF \u2022 Saline\n\n2000\n\n1000\n\n~ 0.4\ng 0.2\n~o 0.0\n\n~\n\n! OJ>. C:t,...\n~D:' 0.4\n0.2\n\n0.0 . . J . . - , - - - - - - - , - - - - - - - . - - - - - - - '\n\n0=\n\n1000\n\n2000\n\nStudy Time\n\n~o:t\n\n0.4\n\n0.2\n\n1I:r-\n(U\n..c\n~ 0.8\nE\nOJr(cid:173)\n~cr\n\n0.8\n\nHF ... Midazolam\n~ 0.6 O-\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7:::::~\u00b7..:::::::;;::\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7\u00b7,,,\u00b7o\u00b7..\u00b7\u00b7~--_\u00b7\u00b7\u00b7\u00b7_-_ ..\u00b7~~_\u00b7\u00b7\u00b7\u00b7\u00b7-O\n[J\n\nDr.------ID\n\n.2~ 0.0 ~------,-------.-------'\no\n\n2000\n\n1000\n\no\n\n1.0 . . , . - - - - - - - - - - - - - - - ,\n\nLF ... Midazolam\n\n0.6 ~::::::::::::::::::::::'a:::::::::::::::::::::?\n-Ii\n0.4 ~\n\n-II- p~Uremember~f\u00b7old)\npJlk.now.~~ I~'Old.)\n-0- Rmember Fit\n\u00b7\u00b7\u00b7_O\u00b7\u00b7\u00b7\u00b7--,.\u00b7\u00b7_Ko_o_w_F_it_---r__---'\n\n0.2\n0.0 ..J..-,-\no\n\n1000\n\n2000\n\nStudy Time\n\nFigu.re2~ .Remember/kn.ow data froluHirshman et aL and predictions ora REM\nfllodeL The paramete:r values are those listed in t.he caption for.Figure I,. plus there\narc two remember...know crite.rion:Fo.r the saline group,. CritR/K;; 1.52; for the\nmidazo.lanl group, (:ritRlK;;;' 1~30~\n3 A REM model for recognition and remember/know judgments\n\nAconlmonway to conceive of recognition. nlenl0ry is to posit that memory.\nis probed \\vith the test item, and the recognition decision is based on a. continuous\nrandom variable that is oft.en conceptuali.zed as the resultant strength, intensity, or\nfam.iliarity [6]\" If the familiarity exceeds a subjective c.riteri.on, then the subject\nresponds'~old\"+Otherwise, a \"n.ew\" response is made [8].\n\nA subclass \u00b7of this type of model accounts for the vvord-frequency mirror\nthere exist four underlying d.istributions of fa.nliHarity\n\neffect by assuming that\n\n\fvalues~ such th.at th.e means of these distributions. are arranged along a familiarity\nscale in the follo\\ving n1annct: p(L\u00b7F-nc\\v) ~:::: jl(HF-nc\\\\r) <~ p(HP-old) < p(LF~old).\nThe left side of Figure 3 displa.ys this relation graphical.ly. l\\.. model of this type can\nthe recognition fmdings of IIirshn1a.n \u00b7ct a1. (in press) if the effect of\npredict\n. Midazolam is to rearrange the underlying distribut.ions on the familiarity scale such\nthat }t(L.F-old) < p(HF-old). The right side of Figure 3 displays this relation.\ngraphically. The R.EM 1110del of the \\~lord-frequency effect described by Shiffrin and\nSteyvers [13, 18, 19] is a member of this class of models, as \\ve describe next.\n\nRE.M [1.8] assumes that memory traces consist of vectors Y, of length ~, of\nnonnegativ\u00b7e integer feature values v Zero represents no infomlation about a feature +\n()thenvise the values for a given feature\u00b7 are assum.ed to tbllo\\\\l the geometric. probability\ndistribution given as Equation 1: P(V = j) = (l_gy-lg, for j= 1 and higber~ Thus higber\ninteger values represent feature values less likely to be encountered in the environment\nR.EM adopts a \"feature-frequen.cy'\" assumption [13]: the lexicalJsemantic traces of lU\\\\ler\nfrequency \u00b7words are generated \\vith a low\u00b7er value of g (Le. gLP < gllr). These\nlexical/semantic traces represent general knovvledge (ekg~, the orthographic, pl1onological,\nsenlantic, and contextual characteristics of a \\vord) and bave very many non-zero feature\nvalues~ most of'\\vl1icb. are e.ncoded correctly. Episodic traces represent the occurrence of\nstinluli in a certain environmental context; they are built of tlle same feature types as\nlexical/senlantic traces, but tend to be in,cOlnplete (bavemany zero values) and inaccurate\n(the values do not nec.essarily represent correctly the \u00b7v\u00b7alues ofth,e presented event).\n\nWhen a \\vord is studied, an incomplete and error prone representation of the\n'~lord's lexical/semantic trace is stored in a. separate episodic image. The probability\nthat a feature ',eVill be stored in the episodic inlage after! time units of study is given\nas Equation 2: 1 - (1 - 11*)1, where !!* is the probability of storing a feature in an\narbitrary un-it of time~ The number of attempts, 1j, at storin.g a con.tent featur-e for an\nitenl studied for j units of time is co.mputed from Equation 3: 11 == 11.=.1(1 +' ~-JAJ), \\vh.-ere\n\nSaline\n\nnew\n\nLF\n\n(}ld\n\nHF\n\nHF\n\nLF\n\nless.\n\nFmniliarity\n\ntnQ~\n\nMidazolam\n\nlleW\n\ncld\n\nLF\n\nHF I.F HFJI\n\nFami1:L.\"U'ity\n\nless\n\nmore\n\nFigure 3. l\\rrallgomcnt of Inoans of the theoretica.l distributions of strcngth...bascd\nmodels that may give risc to 'Hirshman ct at ~s findings. HF and LF = high or LF\nfreq~ucncy ,vol'ds:; respectively..\n:\u00a7; i.s a rat.e parameter:- and t, is the number of atten1pts at storing a. feature in the' first\n1 s. of study. \"rhus, increased study time increases the storage of features, but the\ngain in the amount of information stored diminishes as the itctn is studied longer.\nFeatures that arc not copied frotn the lexical/semantic trace arc represented by a\nvalu.c of O. If storage of a feature docs occur, the feature value is correctly copied from\nthe ,vord~s lexical/semantic tI'aCC '\\vith probability Q. With probability l ..~ the value is\nincorrectly copied and sall1plcd randolnly from the long-run base-rate gco111ctric\ndistribution:, a distribution defin.ed. by g such that gHF ~> g > gLF.\n\n\fAt test, a probe made with context features only is assumed to activate the\nepisodic traces~ Ij, of the !l list itenlS and no otllefS [24]. Then the content features of the\nprobe cue are tnatched in parallel to the activated traces..For each episodic trace, Ii, the\nsystem notes the values. of features ofIi that rnatch the corresponding feature of the cue\n(njjm stands for the number of matching values in tl1e j-th image that have value i)} and the\nntnnber of nlisulatcbing featq.res (njq stands tor the number of mismatching values in the\nfhimage). Next~ a likelihood ratio, ~j~ is cOlnputed for each Ii:\n\n.\n\n/;1\n\nn\u00b0C.\",\"\"'>. [.. c + (l-,{.~)g(l- g)r-l ].,.\u2022 l1.\n\nfi\n\nm\n\ni-I\n\n(.\n' .\n\ngel-g)\n\n)12,\nA ~ l-c )Il\nj\n~ is the likelihood ratio for the fh itnage~ It can be thought of as a\u00b7 runtch(cid:173)\nstrength bet\\veen the retrieval cue and.Ii. It gives tlle probability ofthe data (the olutcl1es\nand misn1atches) given that tlle retrieval cue and the inlage represent the san1e word (in\nwhich case features are expected to luatch, except for errors in storage) divided by the\nprobability' of the data given t11at the retrieval cue and tIle irnage represent different \"fords\n(in \\vhich case features matell only by chance).\n\n(4)\n\nTI,e recognition decision is based on the odds1 <1>, giving the probability that the\ntest item is old divided by the probability the test itetn is ne,\" [18]. This is just the\naverage ofthe likelihood ratios:\n\n1 n\n=-LA.\u00bb\n\n11 j=4 J\n\n(5)\n\nIf the odds exceed a criterion~ then an Uoldj~ response is 1nade, The default criterion is '1.0\n(wllich maximizes probability correct) although subjects could of course deviate from\nthis setting.\n\nThus an Hold\" response is given 'Vvnen there is more evidence that the test ,vord\nis old. !\\1atching features contribute evidence that an item is old (contribute factors to the\nproduct in ,Eq. 3 greater than 1~O) and n1ismatching features contribute evidence that an\nitem is ne\\\\' (contribute factors less than l .O)~ RE!vlpredicts an effect of study time\nbecause storage of Olore non-zero features increases the number of matching target-trace\nfeatures; this factor outweighs the general increase in variance produced by'\" greater\nnunlbers of non-zero features in an vectors. 'RENt predicts a L\u00b7F HR advantage because\nthe matching ofthe more uncon1mon features associated 'W'\"ith LF words produces greater\nevidence that the item is old than the matching of the more COOlmon features associated\nwith H.F words..For foils~ however~ every teature match is due to chance; such matching\noccurs n10re frequently for HF tl1an LF \\vords because HF features are ,nore common\n[12]. TIlis factor outweighs the higher diagnosticity of matches tl1f theLF words, andHF\nvV'otds are predicted to have higher FARs than L\u00b7F '\\vords~\n\nMuch evidence points to the critical role of the hippocampal region in\nl5, ]6l 20]. Interestingly, f\\.1idazolam has\nstoring episodic memory' traces ['I, 14,\nbeen sho\\vn to affect the storage, but not the retrieval of memory traces [22]. As\ndescribed above, there are tw'o parameters in R.EM that affect the storage offeatures\nin tnemory: 11* detennines the nuolber of features that get stored, ~nd \u00a3. deternlines\nthe accuracy with which features get stored. In order to lower performance, it could\nbe assun1ed that Midazolanl reduces the values of eitl1er or both oftl1eseparameters.\nHo\\vever, Hirshulan et at '8 data constrain wl11ch of these possibil ities is viable.\n\nfe\\ver features, relative to the saline condition (i.e.\n\nLet us assutne that MidazoJam only causes the hippocampal region to store\nll* is reduced). In REM~ this\n\n\fcauses te\\\\>Ter terms in the product given by .Eq. 4~ and a lO\\\\>Tervalue for tlle result~\non the average. Het1ce~ if Midazolam causes fe\\ver features to be stored~ subjects\nshould approach chance-le,\\tel performance for both HF and .LF \\-'lords: LF{FA.R) ~\nH.F(F..A.R) .....,/ L-F(HR) ....~ HF(HR). However, Hirshnlan et a1 found that the difference\nin the LF and H'F FA.Rs \\\u00a5as not affected b:y 1\\1idazolam. In RETvl this difference\nwould n.ot be much affected; if at al1~ 'by changes in criterion, or c.hanges in & that\none 1111ght assume Midazolam induces. Thus \\vithin the fratnework of R.ENf, the\nmain effect of l\\1idazolam on the functioning of the hippocampal region is not to\nreduce the n.umber of features \"that get stored.\n\nAlternatively let us assunle that Midazolam causes tIle hippocalTIpal region\nto store '~nQisier\" episodic traces, as o.pposed to traces with feV~ter :non-zero features~\ninstantiated in RE\u00b7Tvf by d.ecreasing the valu.e of th.e ~ parameter (that governs coo-ect\ncopying of a feature value). Decreasing Q only slightly affects the false alann rates~\nbecause these FARs are based on chance matches 14\n.HO\\\\feVer, decreasing ~ causes\nthe LF an.d .HF old-itenl distributions (see Figure 3) to ~pproacb. the L~F and HF ne\\\\L..\nitem distrihutions; \\vhen. the decrease is large en.oug:h.~ this factor tnust cause the LF\nand .HF old-item distributions to reverse position. The reversal occ.urs bee-ause the\nH,F retrieval cues used to prope melTIOry have more comnlon features (on average)\nthan the LF retrieval cues, a factor that cornes to dominate \\vhen the true 'signar\n(mate-hing features in the target trace) begins to disintegrate into noise (due to\nl.o\\vering of~).\n\nFigure 1. shows predictions of a REM nlodel incorporating the\u00b7 assumption\nthat only ~ \\taries benveen the saline a.nd IVIidazolalTI groups~ a.nd only at storage,\n.For retrieval the same ~ value \\vas used iri both the saline and Midazolanl conditions\nto calculate the likelihoods in E,q~ation 4 (an assumptioll consistent with retrieval\ntuned to the partiei.pant's lifetime learning, and consistent\n,vith prior findings\nsh{)~ring that Midazolam affec.ts the storage of traces and not their retrieval [17].\nThe criterion for an. oldlnc\\v judgment '--va.s set to ;,92~ rather than. tlle nornlatively\noptimal value of I ~O!lin order to obtain a good qua.ntitative fit, but the criterion did\nnot vary betw~een. the 1v1idazolarn and saline gro~ps, and therefore is 110t of\nconsequence tor the present article x \\Vithin the RE,M framework;\nthen; the main\neffect of Midazolan1 is to ca\u00b7use the hippocampal region to store more noisy episodic\ntraces. These conclusions are based on the recognition data.\n'h7e tum next to the\nremenlber/kno\\v judgments.\n\n\\Ve chose to model renlenlber...kno\\v judgments in \"vhat is probably the\nshnplest way. The approach is based on the olodels described by Donaldson [41 and\n.Hirsbrnan and Master [10, II]. As described above~ an totd t decision is given when\nthe familiarity (Le~ a,ctivation~ or in RE1vf tenns the odds) a.ssociated '\\vith a test\n'word exceeds tb.e yes-no criterion.\nth.en. it is aSSUllled th.at a\nhigher remember/kl10\\V criterion is set. \\llords ,,,bose familiarity exceeds the higher\nrenlenlherllo,O\\v\" criterion a.re given the \u00b7'renlenlber\"\nresponse, and a \"knowH\nresponse is givenw'hen the remember/kno\\\u00a5 criterion is 110t exceeded. Figure 2\nthis lnodel predicts the effects of MidazQlam and saline both\nshows\nqualitatively and qua,ntitatively\u00b7. TIllS fit was obtained by' using slightly different\nrenlenlber~know criteria in the saline and 'Midazolam conditions (1.40 and 1.26 in\nthe saline and Midazolam conditions, respectively), but aJl the qualitative effects are\npredicted correctly even\\vhen the same criterion is adopted for remembetlknow.\n\n\\\\7Jlen this happens,\n\nthat\n\n1 Slight din-'erences are predicted depending on the interrelations of ,g~ gl1f~ and\ngLf\n\n\fThese predictions pro'lide a.n existence proof that Hirshman et aL [9] were a bit\nhasty in. usin.g tlleir data to reject single...process tnodels of the present type [4, 11]:t\nan.d sho\\v that single- versus dual-process models \\\\lQuld hav\u00b7e to be distinguished on\nthe basis of other sorts of studies. There is already a large literature devoted to this\nas-yet-unresolved issue [10], and spa.ce prevents discussion here.\n\n'Vvork.\n\nthe reverse assumption cannot\n\nThus far we detnonstrated tlle sufticien.cy of a model assulning that\nlVHdazolanl reduc.es storage acc\u00b7uracy rathe-r than storage quantity, an.d have argued\n\\Vhat degree of Inixture of tllese\nthat\nassumptions tnight be conlpatible with the data'? A.l1 ans\"ver \"~lould require an\nexhaust.ive ex:ploration. of the paralnet.er s.pace\" but \\\u00a5e found that tD.e use of a 50~/Q\nreduced value of y* for the Midazola.m group (11* suI == .02; Y*rrti*i == .01) predicted an\nLF-Fi\\R. advantage that deviated from the data by bein.g noticeably snlaller in. the\nMidazolanl than saline condition. Within. the RE.1\\1 fratnework this result suggests\nthe maill effect of l\\1idazolalu (possibly all tIle effect) is on ~ (accuracy of storage)\nrather than Otll1* (quantity of storage).\n\nAJtern.atively~ i.t is possible to conceive of a much more complex RE\u00b7M\nmodel that assurnes that the effect of IVIidazolatll is to reduce the aOlount of storage.\nAccordillg1.y~ one might assunle th.at relatively little in.f1)rnlation is stored. in. m.emory\nin the Mid.azo]am. cOl1dition.~ an.d that the retrieval cue is Inatch.ed primarily aga.inst\ntraces stored prior to tl1e experiment Such a modeL Inightpredict Hirshman et at \"5\ntin.din.gs bec.ause\u00b7 once again. targets will only be randonlly similar to contents of\nm.emory.. Ho\\vever, suell a lTIodel\nis far tnore com:plex. than. the InQdel described\nabove. Perhaps, future research will provide data that requires a Olore complex\nm.odel~ but for n.O\\V the simple m.odel presented here is sufficien.t+\n\n4 Neurosc.ientific. Speculations\n\nThe }lippocatnpus (proper) consists of approximately 'I O~/~ C] ABAergic\nintern.euron.s, and these intern.eurons are th.ought\ntbe firing of the\nremaining 909/~ of the hippocan1pal principle neurons [21]. Some of the principle\nneur011S are gra.nule neurons and SOlne are pyramidal neurons~ The granule cells are\nassociated ,vitb. a rhythmic pattern of neuronal activity k~llown as theta,,vaves [1]~\nTl1eta \\\\laves are associated \",tith exploratory activities in both animals [1.6] and\nhUlnans [2]~ activities in \\vhic.h infortnation about novel situations is being acquired.\nMidazolam is\ntiring of\n(]ABAergic interneurons in the hippocampus [3]. Hence, if tv1idazolan) inhibits the\nregulate the orderly firing of the vac;t majority of\ntiring of those cells that\nhippocampal cel1s!l then it is a reasonable to speculate that the result is a \"noisier\"\nepisodic memory trace~\n\nand benzodiazepines\n\na. benzodiazepine~\n\nto control\n\ninhibit\n\nthe\n\nThe a.rgUlnent\n\nthat?vt idazolaln causes noisier storage rather\n\nthan less\nstorage raises tb.e question whether a sitnilar process produces th.e silnilar effects\nlesions or other sorts of datnage (e.g.. Korsakoff's\ncaused by hippocampal\nsyndrolne). l'hi8 question could be explored in future research.\n\nRefel~ences\n\n[1 ]Bazsaki~ Gy (1989), T\\vo..stagc mode1 of memory trace formation: A role for HnoisyH\nbrain stales. Neu70sciencelj 31 j 55l-510.\n\n[2] Caplan, J. B.1 R.aghavachari, S. and Madscn~ J. R..,Kahana, M. 1 (2001). Distinct patterns of\nbrain Qscillat1(lns underlie two b~qjc parameters of hum.an n13ze learning. .l ajNeurophys,,> 86} 368(cid:173)\n3g0~\n\n\f[3] Dcad\\\\rytcr, S. t.\\.~ Wcst, M., &, Lynch!! G. (1979). I\\ctivity of dentate granule cells during\nlearning; difJerentiation of peri'orant path input. Brain Res~, \"] 69~ 29-43.\n\n[4] Donaldson, Vl. (1996). The role oJ decision processes in l~emembering and kno\\ving,\n.J.\\1emory & Cognition\" 24, 523-533.\n\n[5] Gardincr~ J. tvL (1988). Functional aspects or rccollective experience. A-fenu)ry &\n(To&rnition, 1671 309-313.\n\n[6] Gillund\" G., & Shiffrin~ R__ !Vt (1984). A retrieval model for both recognition and rcc\u00b7alL\n.P,r.;yeh. Rev., 91, 1-67.\n\n[7] Glanzer, 1v1., & Adams~l :K.. (1985). The mirror effect in rc\u00b7c\u00b7ognition. lllcmory. .l\\lenl0ry &\nCognition., 12, 8-20.\n\n[8] Green7 D\u00b7, tVL~ & S\\veisJ J. A. (1966). Signal detection theo~y and psycho]Jh...vsics. Ne\u00b7\\\\l\nYork: Wiley.\n\n[9] H\u00b7h\u00b7shman, E\", 'Fisheric J.~ H'cnthorn:1 T.~Arndt~ J-, &. Passannantc~ l\\. (in press) :Mjdazolam\namnesia and dual-process models of the ,Yord frcquc\u00b7ncy mirror effect..I. qll\\{enUJJ:V and\nL\u00b7anguage~ 47~ 499-516.\n\n[lOJ Hirshman~ E. & Henzler, A_ (1998) The role of decision processes in conscious memory,\nPsych. ScL, 9, 61-64+\n\n[11J IIiTshman~ E~ & JvIaster., S. (1997) I\\1odeling the conscious c(ltrelates ofrecQgnitioll\nmemory: Reflections on the 'Rctnctnbcr-Know paradigm. .Atemory & G\"()gniti(J11~ 25~ 345--352.\n\n[12] Malmberg, K. J. & t\\,furnane:t\u00b7K. (2002)_ List compos111on and the word-frequency\neffecL for recognition memory. J. oj\u00b7Exp. P~ych.: LeanlingJ\n\nl\\;lemory~ and c..of:,rnition:t 28~ 616-630.\n\n[13] l\\lfa.hnbcrg, K. J~, StcY\\lCrsf. IVL, Stephens, I D., &. Shiffrint\nfrequency efiects in recognition men1Qry. A/emory & G'o&:TTlition.\n\nR~ 1.f~ (in prcss)~ Feature\n\n[14] tv1a.rr~D~ (1971).. Simple lUCl110ry: a theory tor the atchicortcx. Proceedin.gs o,.(the Royal\nSociety, L,ondon B 84l, 262:23-81.\n\n[15] lVIc\u00b7CLcUand., J. L,.\u2022 rv1.c\u00b7Naughtou, .B~ L..\u2022 & 07~Rcil.ly,R. C. (1995). \\Vhy there arc\nCoUtplCnlcntary learning systems in the hippocampus and ncocortex: Insights fronl the\nsucce\u00b7gses and failures or connecti()uist n1{)d~ls of learning and memory.. Psych. Rev\"\n419-457.\n\n\"J 02,\n\n[1.6] O~.Kccfc, J. &,N'adcl~ .L. (1978). 17:e hippoclunpus as a cOJJnilive 11't~p. Oxford:\nClarendon IJnivcrsity Press.\n[17] Polster,MA, ~1cCarthy, RA, O;Sullivan, G., Gray,P., & Park, G. (1993). lVlidazolam.(cid:173)\nInduced amn.csia~ Implications for the hnpUcit/cxplicit tncrnory distinction. Br(tit1 &\nCognition.., 22, 244-265.\n\n[18J Shiffrin~ R.M.'t & Stcyvcrs. M. (1997). A tllodcl for-recognition memory: REM ~\nretrieving effectively frotu lllcmory. Psy'cho/1(Jlnic Bulletin & Review~ 4; 145-166.\n\n[19] Sh.iffrin,R...M.. & Steyvcrs~ .'1\\4:. (1998). The effectiven.ess ofretrieval fronl m.crnoty. In\n.1Vt Oak.sforrl &N. Chater (Eds.), .Rational fttodels o.l'cogt1Uio!2. (pp\" 73...95)~ London: Oxford\nUniversity Press~\n\n[20] Squirc~ L. 'R.. (1987)~ }~lenlory and the .Brain. 'Nc\\v York: Oxfor.d~\n\n1:21] 'lizi~ E. S. & Kiss:t K. P. (1998). Neurocnemistry and pharmacology of the nlajor\nhippocatnpaJ. tranSfilittcr systcnls: Synaptic an.d .No.nsynaptic interactions. 'H~ppoCall~p1JS, 8,\n566-607.\n\n\f", "award": [], "sourceid": 2283, "authors": [{"given_name": "Kenneth", "family_name": "Malmberg", "institution": null}, {"given_name": "Ren\u00e9", "family_name": "Zeelenberg", "institution": null}, {"given_name": "Richard", "family_name": "Shiffrin", "institution": null}]}