Tobacco Documents Online

MAi'i~i~l'~ LUEUNti AIND i+Ei ii iNVT/.r yy V Y iiJ Go=- T'I7 0~-ji MERT.= L;.~D FORETIuN CO HD; MCE HKEE OF RERESENTLTiF~~ ~c'Z~.~T~ -r Is.ZST COti GB.ESS STAT=:IT OF TfiEODOR D. STFRZIYG, PROFSSSOR, DEPART`'~iT 0F A.PPLTFD X3TEEX ATICS AYD C0:",1PUTs'... R'...R SCIENCF, WASSLYG- TON U:TIYERSITY, ST. LOUIS, i,10. The Cx_-,rRarA.,r. We are happy to have you with us. We are sorry that we have to run clte=e l~earin:s in the afternoon, bat we h~.re to do it to complete these hrarin,a. Mr. SrEr.LZac. I undErstand. Str. Chairman_ Tlie Cx:u:rarati. Very well. You may proceed. Z-tr. Szmr.z.tic. 'My name is Theodor D. Sterling, 2nd I,tm pro- tessor in the Depnrtment of Applied Hathema•tics and Computer Science at ZWa:hin`*ton L~nivec~itv, St. Louis. I have been professor of biostatistics for 3 years at the College of Medicine. at the lr ni- versity of Cincinnati. I arn the former chairman of the Bioloeical Prccessinz Or,^ini7acinn. Ilu~+•o served tor G crears on the :Lmericun lLarhematics :Lssocic,tion Commictee on Undergraduace Programs in JLathematics. I am at present n, member of the Committee of the April 29, 1969 TIMN 0164768

. , f 931 linerican Xssociation of Physicists iil -lledicine dealing with com- putation matters in medicine. The field in which I teach und in whicll I cclnclluct the nlajori'ty of mv research concerns the collection. prnce=sinz, and interpretation nf rh.r;i. A lnrae number of nlv research etturrs li:il•e coulcentrtret.l specilira,lly on the analysis and iiiterpretation of data concerning environinertal heti.ith problems such as the effect of lead on industrial wrn•kers, tlte effect of low dnses of radiation nit tht~ sul>~c~qiu~nt a(~heat•,tnce I)f neoplasia, and the effect of llollurant., on c•onlulunitv Il~;lltll. And, in- cidentally, the efPect of smnkin: on the health of the ,eneral public. I aul also serving no«. and Ita~•r servetl in the l;,lst, as n cc,n;ultctnt to public agencies on problems related to tlata acquisition and analysis. These include the 1 atlonal .-~.cienwe Foundation, the \'ational Academy of Sciences, the Veternns* A(Inlilllsti;ttion Research Centcrs, as well as a number of universities and inclu,tries. I ant here on nly own be- htllf. and at the invitation of several ci-at•ette manufacturers, not in behalf of Wnshincton i; tlivelsity. It is inevitable thctt com•litsiun:; ntay sometimes be reached hy the Public Health Service or by in~•e5ti;.*.Ltot~ sponsored by it that «tll be in disab eement with those reached by persons attected and by other scientists. It is in the intere~zt of the public that these instances of disat,reement be res-olved on the halsis of inerit of tlte c(at,l, and that relerant analvses be made available to all interested parties, rather than resolved bv the clash of experts in front of cona-ressionn,l cc)cu- mittees. When IwRs asked to review the PHS Publication, "Ci~l,rette $mokin, and Health Characteristics." prepared by Ronald Wilson, of the Division of Health Interview Statistics of the \ational Center for Health Statistics, and b,l~;ed on the results of the Yational Health : tlrvev, it was agreed that whatever conclusions my analysis produced .roula be made available to the Public Health Service, as -«•ell as to the industrv. Indeed, three reports dealing with this study have been submitted to the Public Health Service. My CORCIUSi()11S are based on a detailed stlldv of the information publislled by the National Center for IIeaith Stattsttcs and on adili- tlonal analyses conducted on the vast collection of data on which the report "Cigarette Smoking and Health Characteristics!' is based. These conclusions were detailed in three lengthy reports. All reports and supporting data were forwarded to the director of the Yationa,l Center for Health Statistics. My conclusions were forwarded also to a committee of the tirational Institutes of Health concerned with evaluat- in~ d<tta on smokin_ and heulthy lfy statement. then, covers material of cc•hicli the Public Health Service hfls been made aware some time a~o.j _lIv concern here is with recent claims that cigtrette snlokin, cauc,es awide variety of chronic and acute diseases as well as disabilities. and -with the procedures by «hich the source data were collected and bv which morbiditv and disability rates were produced for purposes of comparing smokers to nonsmokers. There are a number of major $aws in these data and in their analyses that are so serious that clear- cut conclusions should not be drawn. A thorough assessment of the report "Cigarette Smoking and Health Cha.racteristics" must deal in detail with violations of statistical and 3 Coptes of nn relevant correspondence hace beea piaced in the record of this hearing (see p. 941). TIMN 0164769 .: .,

032 I scientific procedures. I«ill eiideavor to summ::ri2e briefly the sub- stance of the m.^.jor shortcomin~ which exist in that report. 'There are three Inajor objections to the report and to the use of the data on which it is l,;isect for estimating the incidence of ctisiibility : 1. It is not c!ear wiute the ba: is d,ita actually rerrL~sent. As an index of health or r.listbilitv, the information collected by the \ ational Health 5urrey is beset with errors. These errors are of such demon- strated ma;IUtude that inferences drawn from such material about differences hetween population groups such as betm•een smokers aud nonsmokers, tire proheihlv invalid. Certainly estimates that a particu- lar number of dis:tibiiities--c:r that any disabilitie..-are caused by smok:ng are meaningless. 3. The datea obtained by the 'utional Health Interview Survey are colnplesly affected by a variety of confounding variables. ~'hnt dif- ferences may appear between eiisease• and disZility counts of smok- ers and nonsmokers depend on the manner by which the data are divided into categories accordin~ to health, ses, a~e, employment, and other population ch~zracteristics and by which observed dise~e counts are ~`corrected" according to di8:erences in the distributions of pos- sibly confoundiner factors between smokers and nonsmokets. The com- parison of "raw'pdisease counts does not really show anything about the effect of smokin77 nor would this constitute a proper comparison. On the other hand, it is well known that the multivariate data may be made to show almost anything, dependin~ on how they are divided up for comparison and depending on what 6ases are useZl to ~adj ust" observed numbers. The analyses performed on these complex data faiI to adjust for so many sources and faccors of confounding that their conclusions lack conviction. This weak and unsatisfactory anxlysis of data, lacking already in validity and relia.bility, does not demon- stratu any bealth con5eduences due to smoking or a:sociated with it. 3. Even if the calculated disability rates are accepted on their fnce value, which they should not be, they do not, in fact, show that smok- ers sutfer from morP disabilities than do nonsmokers. In fact, female smokers for whom much more reliable data. are available than for male smokers, have fawer aisabilities than do nonsmokers. Also, the con- sistent pattern w ith which former smokers appear to suffer from the highest disability rates is in contrzdiction not only to the conclusions ot thi, report but also to the insistence that one of the major pieces of evidence for the harInNl effect of smoking is chat stopping rvill be followed by a reversal of ill effects. 1. TliF, ESTF.\'T OF EP2O[tS I:T TIIE Sl>IIr.rE D.tTa When two groups are compared on some mea3urement. the observed (lilTc<rence should ~~e evaluated with due regard for the accuiacs• of the tool which is iised to measure. This is a ver7 commonsense rnle. Let us take an instance in which we are measurin the height of in- dividn;~ls bv asking children under 5 years of ~age the liei~•I:t of °-d.~~ldy." .lssttme tl~at if we compare che height of daddy as estimlterl by a a-yenr-old to daddy's real hei;.*ht, a 5-year-old is wrong on the avera~e by as much as 3 feet. We ~ould hardly pay much attention to a report which claims that the average height of daddies who are firm disciDlinarians, measured by this uncertain method, differs by ? I

I 933 inches from the a1•era_e height of daddies who are soft Clirclplinarians. ''We would insist that the ditference of 2 inclles ill the average reported hei~lit of different tti•l~es of daddies is trivial when c(ralhare(1 to tlle arerasre error in the method by which the height is actually me:l:ure(i. We.~ould be even niore disillclined to talct' tllis rellorted dilkc•re((ce reriously if there is some suspicion that being :1 Il;irsh disciplinarian has solue effect on tlle way a child perceives ille hei~,ltt of his Iutl((.r. The method used br the houaehola-ituer~ ie~~ =1n•%•r%• hro(iucetl (•r- rors of such magnitude in counting tlle incuience of disabilities t!lat our analogy is not ulire:isonable. 'I'u un(let•st,tntl hett er wlly the (intcl are so unsuitable, we might begin by :isl:in~ o(ihel-:e5 i.•hcre exacrl~• did they come from, .rhose iilnesses do thev represent, and to Mhat extent do they measure accurately the incidence of disabilities in the populacion'4 a. Persarts •intera•ietred ivere not a repre•?entati.rc popzthtion and •ntu~t vT the injarm/Itton on iutil('Y t(•ns ohtahterd •~c•conli' rirt•nd•. The report proceeds from the assumption that its tindins are busecl on a representative sample of the U.S. population. In fact, it is stressed in ••Ci-nrette Smobin_ and Health Characteristics." on page 6, that: one of the advantages of this study is that the data on relationships betwf~en cIg(irette smoking and health are ba~ed on responses trmm a probability sanehte which is representative of the cirilian. nunimtitutinnai twh(tl:itiun of the t'uih•d State-4. thus nll(m•in- the (lata to be presvnte(i in rhe Pur(u nf natiunal estinleue-s. The claini that tile data are representative of the I;.5. polnilatiuu forn(s. in fact, the basis on which the "Hea,ltll Consequence of Smok- ill': '?11:i1Ct's its P-Otttlaros on the nllnlller of (liyitl)illtl('s due to slllnl{111~`:. We tincl on page -23: As the primary source of data in the L'nited States on (limability, the survey relHprr. being based on a national hrnbabilit}• ,atnple•. proride., a %oIiei base for estimating the excess overall disability assuciated with ciearette smolcius. Are these statements justitieil? The method of sampling used by the latl()/lal IIt':LITIL Cltri•e.y U•as a mixture of random and nonrandom s:liuht:nep. C(lsi an(i (:(aivenience factors had to prescribe so nlanv comhrornt:es with wh:it are no:•mal tanduui sampling pl•nce(lnlcs that the final resl(lt e•:+(1 be e•:illeel a probnbilitti sample onl~y by some stretch of the itna;*in:>.tion. In fact the desi;ners of the \ati;nal Health Snrvey llointecl out frotn the be-inniri, tltat the (li~•isirnl of t,iE+ land nla,:: ot' a i•mliltl'y into areas for whicli samples are to be drawn was an --art" r:itller tllan ascience.= 'rhe sampling procedure resnlted in the list of acldre-ses of honse- hol(I: whose membeis were to be interviewed and not in a sample of nonin=titntionalized civilians. BUC who is usually found in the ho~u=e- holcl cls•trina normal «-orkina hour-.,? The,v are hon,e~rives, rllildren, unemployed persons. retirea. elderly, and temporarily or perma,- nentlv disabled individuals. The 1arD*e part of the male a.nd the cur- rentli• eniploy ed female polnilation could not he espectecl to be twe5ent in most instances when the interviewer called, and thu5 clid not fnrnish information about themselves. This particnlnr shortconlint hatl been reconnized from the be,innin- . We find in the Concepts :Lnd Detini- tions` in the Health Householc~-Interriew Survez, pu,re .: From the standpoint of reiiabiiity of responses ic would be ideal in hous- hold iucerrieers If every adult couia be interrie.r(kt for him.seif. The cost of z Tht Stattsttcal Deaipn of the flcalth Houaehotddaterriem 3urveg, series e-2. p. 10. TIMN 0164771 •

cJ:34 .nch a procednrt•, howt•ct•r. is quite hi.~h since more recalls to the bousehold are rt•ttuired. The rttlt•s adohrt'd represent:; :t comitrnati:;e benween the n-etltu<i that is icast etpenaire. i.o., one re:iix>n&nt for a household, and the methtxl that is most r•t•linbir, i.t•.• twt•ry adult re,iKtnding for hiorselt. It, is important to lct'rit in iuinil that l:ir~r parts of the rl:itii are reullv seeonclh:zncl, or proxy, infornlation. The report "Cigarette Smokin, and Health Characteristics79 indicates the extent to which proxy reslx>ir-c~ %vrrt• ns(•ti. lttt'c-rlli:ttintl was r•allet•tecl tlirectlv only fron>_ 40 percent of tlm nl:tlr, but front 5:3 liei•cc•ut of tlie fen:lles.t In ,itltlition, it is tilcrtv tli;rt tlrr-se tnnlr, wtu> were clirrc•tlv interviewed c•atuc nii>sllv frrtnt tirt cli=altlrrl, flie ill, the elderly and i•etired, or the unrtllplt-yrcl. I~:I•,•rrr'S in ~rrr~•ry ":linl>iiit.r:tre icnov~•it to he l,ir,p. Errors in information oLtaineci frein proxy interviews ltiverage about twice as rrr:it as tliilsi'• ul)iuinrtl by cliret t intf'rt-ic~~.= Prn~~• ilrtt'rr•ie~~S are also known to be iilnrlt more e*Wilr atFet-ted by a wide range of social sychologit•nl f;tctuts.' and 1~, Thus, the nt('thoil of s:itnplint nl>r:linrd inforlnation from individ- nal, who are not. reltrr~; nr;rti~ t~ or the [TJ4. il(1llilt::tltlltiOlLilizetl popu- latic/tl. .\lsn, the nre•tlrt>dLrnrlril lill u-itil two levels of reliability in the ;011111e clata. I):It,t :tltt,nt Frnt:ilt•s llrrtl>:tblv llavts less than h;llf the erri>r its data al,ortt iri:ilc•,. The f:u•t that clitferences between female aitnl:ri-.' ;uul rir,nsnu-I:ri~: t`f1r liu,,t rli-ruses are either trivial or favor Sinc>l;t•r•-s, as we slinll sr(+ I:iter on. crn_lit to have rr(•eiti'Pr1 special ilttention. 1>. l'rv n1•x ii! rlrrfrl nh/r/i nrrj hrI %n/P)•,•ir. 7ir :.2irr•t?ll.v C/)t+r; /v-i1tf~ the fnCi- rlr');r•!• rrJ' Ilin~('r/:~r• Nl' !r•r'elirrrl Pr'Pr7iN, I•[-v!/x i11 /1oI•lQY, 7lN.yLf01•1c/LtiQn; l,ir•lli. /•t r•r•/! r•n. r/ir• i:;rrlr. ,l.r i1i ilimr:l-t• or ilis;l'-ilitit'- rrttlrirril iil the rr•Ilort were not btl•erl on medi- r•,rl c'X:tillill7ttitms r-r• on rPt•(>ril., of ille(ll(::1l e\:iminAtintla. Instead, a tilt•tlir:tllv tintr.;ineti iritervie«-rw cltiei•ied an informant or respondent :1ht~ut WIrr't110r or ilnt r hr rrSl,trtltlent, or the iutlivitlllnl ;lhoilt whom he i11i'r>t•nrril, s11H'r-•r(l l•rnrli a ciisease or Gllowed particular si~ns or svmp- ti~ius. The taltl:ttitrrl of (lisrilse depended on inflividutils wl•to cllimed to lt:lvr a tlist•:Isr ril lit•i- bet•:iusf, the tluc•t»r told tI1eT1t So or 1)e^allse they illl:l~.;•inetl it, or rrt,r-rtc•li rert:lill sirn5 or s4•mptnln, n•llich were taken to inrlirate that a pec•itir. clisease was precrllt. The :tsrreelnents between t'nnnts of disease r>btotillril by stu•!i intrc•vieu• ;lirrev, unci by inspection of nrrclir:ll terelcis :11e ~rnalli incleeti. A lutirilx•r of studies have corilirtrecl conditions reported in or in- fr•rrt•ti frotll the \ational Health Survey with those from clinical rernrtlg. Of thcl,;~e diseases that were covered by the household interview in ter•m, that nl:lde misunderstanding on the, part of the respondent t Tn the rrpnrt °ttepnrtinc of Hr-s,ptMlir.ation In the Henith Interview Snrcey" (P.H.S. PnbP.enrion No. 10n0--SMri,•q 2-Nn. ;) wo :Ind that in n+nmple ytutly only it mmrter of the mnles and three qmtrn•rt of rhN !'omah•r ~con• •~eif-recrnndenta and the re!t were proaiev. ('lenrrttr ;tmoktnc and tU•nirh ('har.trteristir.e" inrtir,tex rhnr is mnn}• as RM perrnnr of mai,• +mnkcrr hnt onlp 1? p,•rM•nt of f,•malt• sntuktrry Rern prnsies. Theye Ngures were later veritierl as enrrtvt. j A numher of vniidity studies have compared the rates of errors for proag and self- respnndenta• To nsr'rihe only hcire the rnre of ;•rrurv ro prnxira is most likely an unfler- estimate. For detuile+. <Ne m.r report "an Ecniuatinn an4 Critique of the Report Linldng Cii-ttrette Rmnkinc to r:enernl Nrorhirtity and Disahiltty.•" which is on 91e with this com- mitreP (and has been mndo nt•att•tbtN to the Pnblir. Health Sercice). sThe report •'tteporrinc of lTn.;l+icnliaarion in the Health•Tntersietv Snrvey° (:ro. 1004- Serte:c 2-No. B> compare, nr.•ur.tc>• of ropnrtinr illnearna that are in some way °threnten- tnC" and tinQtt va+t diRr•renr•,•s hr•tween the accurnry of rPpnrtin, 41Rerent rilsabiltties and ec„n larger titKerenera h,•twr+„n self and prnsy rt•epondents. Cnnatdertn, the vast amonnt of pnhltcity concerninL rhr• healrh nltecta of vmnkinL. it Is nnrraannnhle to nuppose that such "pwchnlnydcai•• tactoru more not upnrnting rluring the inten•iew. •

+ 935 le:t3t 1ike1~-such as diabetes, astluna. high blood pressure, or heart ' trouble-olllv 44 percent were lepttt•ted rttrrecrl~. _\;.rrtielnent lx~t~~•~~rtl less clearlv defined diseases sank too as low as ?0 herccnt. Iielatit•el,v dramatic inedical events, such as a ctlntact Nt•ith ph}•~ician durilt- ? weeks precedin~ interrie.r, u•ere nttt relltlrtt•tl aG Percent of the time. Only •1J percent of the population reported their len,rth of hospital stilr correctly. The fact that hosPitnlizat•irnl occurred -a5 inct>rrectlr reported 13 percent of the time. Even an event «hit:ll mig-ht be con- sidered to be the most dramatic inetiic•ail experience in a f<lnliis•, tile birth of a child, was reported with a.3) percent error by honsellold surre~•s. As inaccurilte as ilitervic« dat;la area, ttlev bec•clnle Ilollblv so when they use proxt resllcindellts.t One studr fcliultl rh;tt tirrrn•, frow prox.y rcs11rnu1rnts were over live times higher thlin erl•ol•s from self-rcilttrrili,;; indiviclli;tls. Ht~~~-- el•er. in most ca.e: errors from proxies ;11•t• rtlag1114 t«•ice that of errol•s from self-reportinI ,-. One study fulind thtl.t for hospital sta,~~,~•s for 2 to 7 days. proxy respont.lellt:: ni;Idc~ el•rt~l•s approximately :4 pe>icenr of the tinle ti~•herP ,tilf-respontlents made such errors only 10 11crcPnt of the time. When the weeks hc•1ween hosllit;tl di`charr•e wcre between ?1 ancl 411, self-reshondents ulatle a n-pcrcPnt error: prow re- shttlidents made a 1_-lx•rcenr error n•lieii rlie.• rclxtrtt•tl !°nr tlielr sllurl~ze ;Inc1 ;1 a_'-percent• error wllen tile}• reported for cltlier relati%•eG. With re-E;ect to delivery o F;t t•llilcl. sel f-resluontlent, t;t ilt4l to rl•lxit•t clelivery :1 hercrnt Of the tinle, prosY respouclent, fitilecl to report t}leln 8 ller- cent n t` tile time.= Errors are not limitecl to tlle ctmnr, tof tli=t•al~.ec. 'I'lie (•nmit of wlitY smoked and wllo clitl not is also sul/.ject to ctlnsi<lt•t•altir tlttilhts. T(iorc. ,irP two large-scale surve\•s that lult•e flirniSiu,tl Il, ll:itrern of tlle clis- tt•ibtlt•ion of smokers and nonsnhlcers in tlh- T'.S. llttlnll:iticm. Tllese slll•ver3 ;lggTee NCitil each other 1>nt Cllsill-trr('Y 1A•irll IttllilEtel• of qInltla'I•s e~tinuited'bv the lationttl ITeultll Survey by :ts ntuc•11 :i, ;; lerc.ent fnt• _rnne tu;ile. ;lnd ?:; percent for ~o1nP ft'nltllt•.:l.gr _*tt,llll.: In t•it+«- of the l;tr,;e et•rols and thr nnct•i•t;lint~• ttf tlle~ir hase. it would be speculation to estiluttte tlit' n111111,t,r tlt' ciise:t-cs tllat: t,t: ilr I11 the L•..1~.. population to say 11Uthllit" of 2IsC1•il)lilir a rl\•Ptl Illlnll)Cr Ot the?P to a particular noplllition " rolln. Certainly we t';lilllot justify a ct~lripil.rison bet•.~ een smokers alld nt>nsmoket~. ally comparison between ,rrollps has to be cl<~na «•it11 clne re:,r:lyd for tlle accur;tcy with ~~•hit;h disease r;ue> are estinutted. Zlthen a =ani111P is a trulv random sample, and when the method of inellsure- 111r'Iit is ••reliable,•' thell the accur.icti- of estilntite>; can be inferred. Tlle measure we use to describe the ;lccur;ic}• is tlit• "st:intl;lrcl error." When dealing with I.uldUln samples and reliable ulethods of ineRsni•e- nlent. the estun;ttes of standard errors, or of tile Iltlssihle; %•;tri:il>ility of tht, .~~rilu;Lte, ignore the coefiicieut of reliiihiiit}•. Ih~rerer. «•lletl the rt•ii:tl,ilir%• of tile nll`;tFttt•t, is kntlu•n to be ltla•, it i5 nect•ss;tl•v to Cot•rc•ct tlle u:mal estimate of the standard error by the size of the reliability t For details and references. .-we mc report '•.tin Ecnltuation and Critiqnr of the Report Linktnc Clcatrette Smukin_ to Grnernl arurbidity ana Dixnhitltc.•• nn iile with tltN cummitttw. = j'l9tr-r urN : Ilutumuud. E. C.. ; utnkiati in lielutiun t,• rlte It••uth [L•tte ut One JIilllon Men and Women. National Cancer institute Monograph. No. 1r+. 1968. •and Iiaensyel. W.. Pt al.. Toh:tcco Sutoking PatternK in the t.•nited States. Publir lretith Munogrnph, \u. 45, 1/15a. 1{rlrr;tut tahlrs hat'e b.wtt .•strncted tu ttt,r tlctailed ropurt tu 1 ht• 1•11S.

: ,, I 3 936 r~~~l1'i~ ieut. If, as in the present instance, the size of this reliability co- efficient is not known, although it is known to be small, and if, in addition, there are considerable doubts about the validity of the meas- nre and the randomness of the sample, then the usual estimate of the :~tandard error is sinii>lti- not valid. This is even more true for male smokers and for workin~ females for whom data was obtained mostly through proxy inrervie~cs. The reliability and validity of their data is l;nn %Vn to he sniaillcr b,v :ti Iua~r.nitude thzin for directly obtained data. Thus, it is siniply not possible to evaluate differences between popu- l;ition groups, especially among male and currently employed females aIainst the e5timate 4;f the standard errors as ;tven in the report ••C.'i':tretto Siuuiciu_• cuul IhM(th Ctutiracteristics." =. flOF.3 'r1IE DATE RR_\LLY BF..1R OUT TSE PUBLICIT.Y:D CO-ICLIISIOY57 Besides the uncertuintie5 ahout the data themselves, there is a real question if such data that were collected really support the conclu- sions which were disseminated. a. Femn.leQ: It was pointed out before that data on females are more reliable than those for males. While most of the information about females came from sel f-resronses, most of the in fnrmation about males came frocn proxies or frnui m;iles who were found in the home during working hours and who included, therefore, any unemployed, elderly, sick. and disabled individuals. The comparison between female smokers and nonsmokers would appear to be of special importance nnder the circumstances. Yet, the vomPat•ison of feniale sniokei:; to nonsmokers does not yield a. picture that is at all in line; with the stated conclusions of the rei)ort. In almost all instances where Rge-specitic•rates are compared, tho female nonsmokers show a hi'her rate of disease and disability than to fonuala sniokers. «'hen a c•rnnlmrlson was made not between the more reliable a:•e-specific rates but bet.veen z`e-adjusted rates, these com- parisons changed sli,ryhtly. I will discuss this peculiar result of age ncijustment more fully in a moment. But, even if we accept the a~e- adjusted rates, then disease rates for females are still 'hI_her for nonsmokers for some of those diseases for which the report Ciaarette Snioking and Health Characteristics claims the opposite. This is true especially for heart conditions. 'yVhere differences in prevalence rates are reversed by the zg-e adjustment used and female smokers have & higher 1gge-adjusted disease rate than nonsmokers, the ditferences are trivi;til, llmost exactly the same results are found in comparing typps of disability days. for restricted activity, bed days, and work-loss days. In jeneral, female nonsmokers show a higher rate of disability •.lavs than do female smokers. There is one peculiar manifestation among female smokers with respect to work-loss da,vs which is especially worth mentioning. Fe- male smokers are restricted to home or bed for fewer days than non- Smokers. Yet, currently employ ed female smokers show a higher rate of work-loss da-rs than cio nonsmokers. This reversal is rather curious since it would inRke little sense to conclude that employment causes a larger increase in disease among smokers than among nonsmokers. Another, and more obvious, explanation is that females currently em- ployed had data concerning their health and well-being supplied

" 937 through proaies while females who were not employed supplied that infor.mntlon themselves in the majority of instances. b. rl.djustment3 for confounding•sa?`icable4: 'I'here are a number of sources of differences in the slmple population l.,rtween dispcise and clisabilitv rates of slnolcel's and nonslnolcc'rs. 'I'hc five obvious m:!ieer sout•ces t;f difference are: 1. ses: ?, a<_e; :>. proxy reshondin~,: -1. t`:n- ployment status, and 5, characteristics of int'ornt;tnrs for proxy re- shondents. Adjustments were made in the analysis only for differences in the set and a,:te distributions between smokers lndnonsnlokere. The other three major confoundind factors were i(*norecl. But proay responding, employment status, and char;icterist.ic~ o E rc-.ypondents all are rela,ted to tlle frequencies with which disenses are reported., Xow, e.ll these effects on reported disease freqnenc`ies due to sex. zti,e, employment, proxy reportin,, and dtte to other factors interact in a very conlples pattern and lnay rl'sult• in large clitF('r('nt-t's in tlte re- portea disease freqnencies, derendin* on whether or ltr~t hrnc-isie~n~ for ttdjustin; for their etrects'are incltuierl in the <ltutly-es of tlte (l:,rct and on the kinds of adjustments used. In the pnt'sent. in::te,nre tlte-e adjnstments were of p,u:unount iluport:utre. Slnokrr5 (liFfhred very much from nonsmokers in the 1,itional IIeulth til:rt'e}- sample. TI1Pr differed in their o,vaile,bilitv for direct interview. in their emplot-menc status, in the characteristics of individuals who infen•lnc`cl about tlit•at. and not only in their age and sex di~trilmtions. But manv of the etFert: of these conftinncline_ vilrial>le• art, Snc`li tlt:(r I11(tlr:tCtiOllS Sblllo!le t•llelll l'(1(ll(t t'tUlti(' tlitl't`1'(`nt'('S ill fl'l`t(ti('t~t'it~ of disabilities to aPpear for ~rottps ucciclellhtll_c ul~e) related to >!ntt!:- in- and nonsmoking. T'rol+er adjustments for these cnnfnnnclin4r f;utrn•~- t•:ln lie m;icle. '1'liF' luo,t reasonable ;tdjilstlllent would be t(> 1)rovitlc' St'l,;irnte ;lae ati.jttst- ments for each "directness of inter%-ien-. enll>loE•nic'nt =t;ltus. infot'nler.=' characteristic" constellation and Conlh.ire tllese, ,rrotllt 1)~• ~,rrnllf. I'a•c`!t better would be to compare .t`,e-:pectfic rates ~t•uhut t'atcll t:t tL•e-c constellations. However, neither tlle~%e nrn• ;tnY utitE'r ;itil-uhri:ttt' :u,jnamc•nt- %%t`rr done. adjustments for sItlelklnes-11P;:1[ll The analysis only used indirect age category frequencies. Indirect age adjustment is the weakest of tl(ree commonly used age adjustment procedures. Tlti; proc•edure ic ntost easily ltftectecl by cvnPounclin~ t-artahlc`, snt-h ;ts l)r()x~• r(•l~orti;t~s. (`tu- hlc~rntent ;tatu~. or tile ch;iracteristics of informant. It is Iuost curious that thw National Ilcttlth Slu•vey does not seem to c,ue either for indirect a_e adjustment nletluxls. The report on ci_;irette smoking and lleul*.h clturacteristics is the only one in a series, i.lty report. "Esnluation of the Analysis Procudnres nf the \iIS Interview Data." ?z- amincs ln detail the diRereat d!stributions nf the satm(ile ie/mulntion among ngwsr<- eraployment atatus-directness of lnteroiea categorio+. Profound diRrrences emerged betireea thr diatribution of smokers and nonsmokers. :\lso. euch of these cntcanriea had an npprrr•i- nle eRect on rettortetl disabUity trequencies. To tbese rRerr:e must be ndded the Was dur to smoking and other personal charncteri+tic-s ot the• informant (!or the• mnny pros,r incer- rieers). As a resuit. the composition of the smo:ing and the nonsmoking groups nre so diRerent that no good procedure actually esists by which an unconfounded eRect due to .mokiaR can be Isolated. The report referred to was sent to the National Center for Health Statistics and has been placed on Hle with this committee.

. e 4 r ; 93S of 4` reports on the same or similar data that uses this indirect method. All others compare age-specific rates or rates that were age-adjusted bv direct methods. -c. SPh.at, hophens to forme•r,e»colLens? The report on cigarette smok- ing and health characteristics contrasted all former male smokers. those who quit within 1 year and within 5 years or more, to present smokers and to nonsmokers for one or more chronic conditions, for acute conditions, and for restricted activity days per person per year. .1 consistent pattern is quite oov iouslv e.pparent. For all categories present smokers have a much lower incidence than do former smokers. _1lthou_h no detailed comparisons were made for females, former smokers show also a consistentl~• higher incidence of disease and d.is- ability than do present smokers. ' The fact that the incidence of disabilitv did not decline, but rather increased, for those who stopped smokin, is of utmost importance. These results stand in direct contradiction to other findings and to the espect.ations. that generally morbidity would decrease if it was caused by smoking. One of the major clRims that smoking is a hazard to he.llth is based on the ars.*iinietlt that stopping results in an ip~;l,eent illcrease in life expectancy. In fact, -'The Health Consequence of Smok- inb"goes so far as to state that : I'revious findings on the lower death rates among those who have di3con- tinned cigarette smoking are confirmed and strengthened by the additional data reviewed ... The Sndings ... support the view that epidemiological data show- inx lower death rates among former smoker than among continuing smokers can not be dismissed as due to selective bias and that cbe benetits ot giving up smok- ing have probably been understated Yet, the one finding consistent for all sexes, categories of diseases, and t,ypes of disabilities of the National Health Surrey rs as that to stop smokin ; was associated with an increasing incidence of illness. jVhat findings are we to credit ? CONGZZ'isIONS The report on cigarette smokir., and health characteristics is based -on most uncertain and inaccurate dara, it is based on very inadequate analytic procedures, and it lacks conviction that it really demonstra,tes any difference between smokers and nonsmokers. Yet the same dlita from the same report have been used to estimate a number of different t,ylms and categories oE disabilities stihposedly caused by smoking in 'The Health Consec~uence of Smohin;,'' which is a 196? follow-up pub. liration bv the Public Health Service to the 1964 Report, Smoking and I-Tealth. 2`llis rE•lort maiccs sl,ecitic claims about the total days Ameri- c•ai15 a,re Sulilu>~ed to llaVe lost from work or spent at home or in bed due to.sninking. TlleFe tia-lll•es are lar.,e: TT million days lost from work. .14S iuillion slx•nr• in tx~d..:11it nlillioil claSs of restricted activity, all '•clue to" sniokin~. \nne of rhese fi~ires even appear in the original studv- t.h~~v are e5tiluares sa ici to l,e b,~:ed on much smaller fia-ures contained in tllat,tudy.'I'hese same ti-mires aiao ;ire used oontinuously in pamphlets circrulated by the \ational Clearinghouse for Smoking and Health. If '1nv lesson is to be lecirnecl froeu this issue, it is not that smolcin~ c~alus+s ulitold diseases an~l disabilities and loss from work, but that clsiims about such controversial topics need to be very carefallr and intelligently reviewed even if they are made by public agencies or by TIMN 0164776

J . 939 ~ .. t,rl0e'r inlpece;tbll• srnu•res. ~ntlt claittt= le'ntl rrt atrtietn; tlt:tr tn:ty all'r•e•r tltr lu'nlrltd rtntl wt'll-l,e•irtt: or 1:trct' ntnulx•r:s of ritizt'rm It wottfrl•:tl>1>rar thnr a tliorou`*1>h reriem. 'r)f rlle ~nzlity of dcua, the quality of anal~•~es nt rhe5e d,>.ta. aild the ;otineltu'.: of ee,itl•lll.iotl, have to hr )ne,=t , i•ttll\• l'sial>lislteti !x't:ore rht'}• arl' nseti in stil)lwrt of lmbiic policy. (Attachments to Mr. Sterling's statement follow:) ('vittttct-t.t'M 1'rrAa:-'1'tiN:otx,t: P. A•rF:r,r.txe: Eelnc:ttir>n : :1.B, tceitb lhnnoral• Unie't'r:rit:r of t'hir:t_o. 1:1•ld). M.A.. Cniversity of Chica,,^o,1S):i2. i'h. D.. Tulane tniversity, 10:>Z. Prttfe::.4ionotl .\Bitiations : I'rnfesgnr of Applied .1GlthemntiCs anel (`nrttpnter Scir'nce. ZYa:rhin_ton 1'ni• e•ersity. St. Lrrttis. Jlis.,nnri: 1'isitin_ I'rnt'e'ssor in ('nmpttte'rv and iinnurni- ties. Hebrew i'nion Colle,e. (.'inrinnnti. Ohio: L'remiQent. .IiP:PC'r)M1' Re-earclt Corporation. Cincinmtti. Ohio: Senior Consulting I:elitor in ('ont- putt'rScienc•e• llae'ntill:tn l'umiutny. \e',e Ynrl:. I•'ornte•rlc Prnfesaor ut' Bir>~taticric~ anei PireK•tnr of the Cnmpntine Crntt'r. College of Medicine. Cttirervity of t'incinnati. C inoinnati. (lltio: .1ssist:tnt Protessnr of Staristicv. Pei+nrtment of fitotti,4tic~:• Jiichizan ~tute Cni- vt•rsity. East Laming, 'Michiran. i'rr>f>'+cinnnt actiritle-: 1'te,3-ti7: Member of t•he Panel for P.inlng-y. Niau:e_e'tnont and Srrcial Ccie'ncP:r of tlte 3lathematics .\swN•icttion of .\nu•rica. 1rJ('~3: Chttirman. Committe'e nn .\ctirities of Plind I'rnfes,,innal:: of the .lsse>• ciation for Cnmpntinc Ji:tchir.er}•. 1!1ic.1--R:r: PresiQettt. Iliolr!,icrtl Infnrntation i'rnce•sins (?r_nnizntir>n. 1966: JlNttthPr of l'ntttmittet- on Ra4iarir>n Itnsittte•try. .\uu'rie:tn .lssar•iatinn r>f Phcsir•i:rr, in Jlrrlirine. liltiR-fiT: Chairman. l'ointuitter on :\ce•roelitntiun. .1scociation fur Computing lt:u•hiner}• • Consultanr r>n rampnte'r Q:ttat m:ttta_emrnt nnd st:tti.,tir•:ti prr~h- lents to the National Scic'nce Foundation, the Veterans Atiminisrration. and other health a,-(,nci",. Prnfrssionnl ancl hIonorar. Cor•ietiea. ITnnnr...\a•:rrd:r: .\itter. Math. 1L•trh. Srx•.. Instit. Math. stat.. .\tue'r. St:tt. .\s:roe.. iilo- nterric Smtthern 1'hii.. Assoc Coutpntin, NIac•hiua•n•. :\mer .\ssrx•. Lnie. Prof.. N.Y. :1cnd. Science. _\ruer. .\.w>e. Phys. Ile'Q.• Sictn:t \ i. T'i Jlu Epsiion. Morrison Cressy -\«;trti in \atural Scienc'N. 1'rr.vonal: I:orn: 1923, atarriecl: 1948. TTwo ('hilQrt'n: 1tr.~:. 1:):.:. BtnGIOGtt.10IIT OF TREOOOR I). ST(:r:I.I\G. PRUFP:-;sr)K nC>•rr:S r•nmprrter.• and Nrc Lifc Scic'necQ. Cnlumhi:t Pre?s. Ilecemher. 1:N:7-. lntrurlueti4rt to Vctti4irrel Data 1'rnerxxirrg. Prt'ntic:e-Ilttll., .lnne. 1tuN. .-1 t::uirlc to PL; t. Holt. Rinehart and Winston. I'ebrttzr.r• 1969. f'nutiputirr!7 and Computer Scir'ncc. NfacmiIlan. 1!)74 (in press). Cu,nputation in Parliologlr. (T. Sterlins, ed.) Thomas Publishers. 1971 (in press). A nT1(: r.F:S Air i'nllutinn-The Tndnstri:tl Viewpoint :A cnneme•nr. •1F) tl. i in prrs:c'>. Robot Data Screenin,-Att L-i>ietuitnn:: Anrntu:tric Se:trcit Tecltniqne. Prnrrcdir,gm urt ("r>rl/crrwcc on Stati.-itir•n1 r•nntlrrrtrrlinrr, T6r !'rrirr•r:•ril/r of 11'irrorrxirc, .ilurli- .<..n. .I pril lJtifl• > in prrv•..>. \I,•cr.,ttrin, tite I:IYe•rt rlC Air 1'nllntiun on il•batt \Tnrbltiit.. .trrd. I:'nrirn>r. ITr•rrfth, r in 1>re-.r. Dia>l:ec ik'rices fr>rCumiratc•rTr,tttslateri Rr;tillt'. I'rorrrdirr!IY r n,r,irrrvrrr •,n 1'r rr^ P/'I.P(XBPY for IPrnrllr .11'rrnrcirrr•hrrr. !:)ltS. t•evrt(-r for Son.nr}• .\idk h:v:ritutfiuu aud I)evelui+ntent. \iatssachase•tt+ In:rtitute of Te'c•lenuler,y• pl>. t-I-_'t). ['e'I+. 1S)Ii.y. :\ttte'nuttion of R:tdistion Tre'ntuteur l'lannitt, \'I :.\ elrn>•r:ri C'ir•iel 1•:qtt•rtirm m l'airulttte I'>>rcent I)epth Pn.<• in rlte Irr:uli:ttr•d L"nituue• uC a t'ui>alr cr> L•r•uut, Brit. .1. b'adiol.. 40: -163•-!(i5: i!lei?• TIMN 0164777

, i 940 Urban Hospital Morbidity and Air Pollution. _1 Second Report. .dreli. Etrz-iron. ITcaltl6, 1:i: .^,li.-3i-1:. 19ti7. 3\etv I)irotaion in Rt•h:ebilitatian Throur;h Adcanced Instrutnentntion and Computation, Ja3la. LY)O: 6:'r-t;211, May 1.5. 1i)t1T. Final Narrativc ICcport• Training of tlee Blind for Prore.T•tional Compater iPorla, Ihqtt. of ILE.N.. P1>'-1$Y:T-.~-rii-~'?. 1!11;7. Robot Data Screenin, :A Solution to 1[nltivariate Type Problems in the Biologi- cctl anti Sa:ial Sc•iences• r;osem, ACII, 9: :i2J-432, 1J6R. .l Riolneicails-Oriecued Computer Lan);uuge, 1nn. J.Y. dcad. of Sci., 133:7T,:r TG:;• 1906. Cancer Therapp-C.omputation and Visualization of Dose Distributions ia Ec- tornnl Iteum Theraley,.T. Ckrorr• Div.,1J:;r?-•;-:,3J, 1J66. Use nf the Computer to Teaeh Introeluctory Statistics, Comm. dCJl, J:•374-.76, 1960. Auroneation of Radiation Trc•:ument Planninz V. Calculation and Visualizarion of tlur Total Treatment Volume, Brit. .T• rrerlto.. .'.5::1O:-!i2:3, 1:)Gr`>. Conepaters: No Lons;er 1 Bip Iiarxain for Uneducated User:e. Coniputrrs anrl d«to- mrelian, /:.:1.1!2(if:. (?vith .>`. F'i rilac•I: a:csenirir author). St•!cr•tinn. Tra•ininr7 anrl Plart•ment nf Rlind C•nnclrtetcr Programmers, ReI>,)rt of the .>,sco;•iettion fur Computing.lIarhinery. 1!Hlll• Is Medical Diagnosis a Uent-ral Computer Problem. •T. American Jfcd• 198: 241 -2Sti• 17Fi(i• Urban Morbidity and Air Pnllation, A First Report. Arch. Enriron. IIr alth• 13: The Plind ns Computer Prn;r.ttnme•ra, R(7habilitatinn. rrnnrd, 7-10• i!lRfl. Role of tit:etistics in the World of Computers, Ann. 1'.F. Acad. Sci.,12S:11OS--I11G 1:1(iF• Canror Thorapy-•f.'nmpntatinn and Visualizatinn of Dose Distributions in Ez- ternal Beam Therapy. .1. l'leron. ni.v.,19:-52•:-.i.;!). I!li;R. Prncrrrlinpc on the Rraillt• Rcse•arr.h and Derelonnecnt rrynfrrnnce, Senu)r:>vid4. Ec•aeGtation and Development, liassachusetts Institute of Technolog.'r, \ot•en:- lir•r. 11166. Robot Data Processinr Tc•clrninurv fnr )[uiticaricur' ia,itien3iulugic•al Preclictinus. prnr'• .l'.Z'. acrtrl.: i-ir•nr•r, 1°ti: i7R-7!1 f.1J6i. Cunt[mtatiun of I1.'tdi:ttiun leosar;es, Stacy, It. ancl Wacman. B. (I?I)), Computcrs i,t l:4•mrdical Rc:vrarch• .\cariemic E'res•?• \ew York. 10f;,. (Chapter 1i )• Tvtvnrci an Undergraduate Itathematics Pro.-ram for Future Researchers In the 1'ir lds of Biologc anti 'meriicine, Fed. T'roc., 10F>.5• Careers= for the Plintl in illectronic Data Prucessiei_, 0reetp. Outlook Qetart., 9: 1-4. 1Uti.i• The 1.'li•ted in EDP. ACM Committee on Professional Activities of the Blind. 1!)6.i. Design and Operation of :t .ltrriical Cnntpntin-- Center, .Inn. Y.Y. Arr,tt, fici., 11:?:MM-3:10, 1:H.}, (With E. Sm•ugr-r as senior eutthor). Planning Radiation Tre:ctmeut on the Cnmpnter. :lteee. Y.Y. Arad. Scicu.••.e• IIi: '•.17li-!It)i • I!N:1. Prnfes+inn:eL (.'ntn[rntor?Vnrk fur the Blind. Gnnene.:1 cJT• 1964. Ccrntprrtc•r Work f6r the i;lind. .1. lerlettb.• aB:r::^_0-_1. 1fJf.}. Epidemiology of Disease :lrsociuted with L ead, Arch. Ene-irrrn. Health, S:•3.'.3-•;45, 1064. Mathematical .\nalrsis of Lefld Burdens, Arch. Enrirore. FTcaith, 3:-14-11. 1064. Autriu:atinu rlP Radiation Treatment Plauninn. IV. Derivation nf a 3[atLi•maricai Expression for the Irercent Dept I)ose *Surface of Cohetlt 11.e) Bennra and Viauali- zttion of .1[nltiple Field I)oce Distribntions. Rrit. .1. Radio., a7:54-b-3:*A. 1t)6•}. Locating Placenta Praevia. Proe. of the ltoche:rtrr Conf. on Data acqeei.titinn and• Proc•eaxinp iu Bioirrdy a»rl .ilcdieine. Per,euaon Press. [ip. 10:3 =11. 1^(1-E. Do ll:tli,.tnancies Result frane Lius-tin.ric and Therapentic Radiation?. CFenctica and the Epidemiolopy nf C.hrrinic Di: cave, U.S. Dept. of Health. Education and R'ei fa rrn• pp. ?.53-.i71, 1063. ]fl•:Dc:n,111'• Part 1. Stati.Vtirrtl : pstc•ite.e• J[c'ciic•al (_ompatin.; Center, e"olie;e of Medicine, University uf't;incinnati• Ohio. 1()t«;• Boron. Cadmium, Chroneiuue• an:t \irkr•1 in the Blntxi anci Urine. Arch. Envirnn. Health, ri :281i-2S .,-i. 1063. (IT• Iurbus. as senior author). Potential Fiazarel of Baixi,rrr• to [.e•ati, arch. F.'~rrirnn. ITrrtltle, 1!)F'i,°.. (R. Kehoe as senior authur). Carcinogenic h1lCta•1s of t-I:tt l'umparerl with t-Irraciiatiuu-:\ Review. Ilcrrith Phpaics, 'J :1371-138•1•, 1(lFh4.

w ' 941 intimt.^.tion of Radiation Treatntent Phtnninr. tIT. _\ Sirniriitirvl R.sterrt of izin, L3ndoses and Dirt.•ca Yrint-Anc •rt' D.rse 3?fatribntiutt. tirit. •I. Nu+lin[.• Intrarenous Jiethocrex;)te (:\merhoptrrin) Therapp in the Trnntntent of :\rnte Lc•ubetuia, Pccliatrics• 3l:ia.l-Sii). 10uS. (R•ittt J. Perin, senior author). :\ttt.:utation of Radiation Trt•at:nent I'lanniuge ii : l'alrul,'ttiun uC Sun-t'uuc•r•r,t•ttt Fie1Q Dose Distributiurts. 13rit. .1. Ruttini., .iri :t4:-E31. it„i:i. Rnbot Troatnienc Yl:tnuitt,_. ProC. of tita Rnr'hrstrr l'r,rr.i. nrr nula .(r'ryrrixilir.rr rrud 1'rnrrxxing in hinlnJU und .1lydirinr, New Yortc, iN'r,:ttuutt i'rr'.., i•t•. :",);~-.31J. 10t33. Robnc Data Screet•.ing. Proc. of the h'nr'kr•atrr nn 1)nta Ae•rltti.,itiorr nnr7 Procryain7 in Piolnpf and .llcdicine. \ew luric. 1:luL', I'E•r,:tmwn Press. 1't). _31-9•t'2 10t3.,. Q1;cuLF:uionsi Eslwsure to Ur_anic Lead Cuutiiouueis. 3rch. !s',rrirorr. IIc•rrith• J::*302-a:6, 19B_'. (R. DeTrirille• senior author). New Derelolunents ln t'Srnnic .Dise:tse I?iridrntinirr_,r• (•nmpeting Hi.ky and I?;i- cibilitS, 3rner. Inrl• F3_uyivne dsyoc. •1., 2.3 :-I:13-1-16. I:hrd. A I'r::cticsl Procedure Por .\ntuul:ttin, Radiation Trr•:tttur•ut I'IanuinL, /trit. •J. rurt'iol.. 3-) :?_U-T3:3. 1at;1. Epiur:mioiogical Methods antl Communitc Air Pnilutinn. arrh. h.'rrriron. Ilr,rlth, n:'3t5T :;:,. 1thii. (T. Phair. senior attthor). Competing Causes of Death In Cofll Tar «'ortc.•rs. I:etterin, Report. 1:>ta. Cesium-1::7 Retentiun and Distribution in S-Irrndiateel Rats. ('.S. artnp 'ltedicai Research Laboratory. Fort Iinos, Kentttcl:y. Itrport ._:iD +• lt)60. ((I. Kt•rinices• ?rnIor author). ":'e•:t`•uuai Variutiotts in the• Birth of the Jleut:tity Detic•irttt. Jnrr:n .1. Pub. llcnttie. «'l:ar is so 1'eeutiar About 3ecelrting the \ull Hclwthesis?, 1'sycltol. Itcpor•ts, :3t,:.-V~rl, 101'0. \eol•lua:t Fnllowiu>; Ther2pentic Irrsdiation for Benign Conelitiona in ChilQlta+d, R•nliniopU, 7.):55t)-fit}I. 1.1t;0. (E. S:ten:;er. sPnior:)nthnr). I'nl•iir'atinnc Ptrisinn, ctnti Their T'n..'hh• Rt?'r•vt• mn tttforrnr•r•: T)r:twn fr:nn 1',',t ur, aisttitivttttct•-ur 1'ivt-1"cr::t• .1. .IA;rv'. StUt. .l.'arnr. ;, f^it t•1• f:1 7,a. I'ttlt'nouary 1••unrtiou in Chilsren I and 11. .1. <lllerg?r, .3O :; t4-::K:. Zll~+:). (L. ltern- ~n•in, senior author). Tal•i~- of rurmnl I'ttln:onary Function Vaines in t'hiirirrn .\,e-s ic-i t, I:rt•;or Pr:utit:, Cu.• (.'ittciututti. Ultio, 1t1M), (I.. I'.rrnstr'in• sr•uinr::u;h41r). Tiu• .\;turcxi,;, cuie -\ctiun of llextru-Jutpitrt:uui::e tiulf:tte n1Ntu B'eerlirt:; Fl"t.onses of Differing Strength. J. Cont p. and 1'Ir psioJ. P. Ychol., SJ :Ili)-1S'., (P. Sie.-el. sHnior atuhor). ~nnz•ee3 vP Contextual Cnust: tinc upon R nrtis in Sentt•ur'e'+, .J. R.rp. Pxrrrlenl., :171-14). 1:15:1. t II. :lburn. .rniur autlurri. 1•YM4•t sit' Irrolcv:utt llrirv on Extinction. I':r,lr•knl.. I2t•lwrt 1!1:r7. A (:,utratl Z")urreillance 1-y-tc•tn .licasnrc•taenc .llorlei, Part 1. II, I'roj. lilehigsn 7.*~(:~. Z:11•t• L'nirersitc of Miclti,:tn. 1t)S;, .1 tirnrCai : urrec oE Statistical Declsion Theory. Prnj. ltichignan 7M15 . L: nirersits c,' JL'chi;;an, 1057s. TIMN 0164779