I9 I 3 EEAEWEWQIL 3589189 If PROP WGIEIBIB 11 9543 TO PULL MUSINGS ABOUT THE Cecil 3 AN OBJECTIVE APPROACH TO SCORING TRANSLATIONS 5 A COMPARISON OF NSA AND ATA LANGUAGE CERTIFICATION STANDARDS 8 LETTERS TO THE EDITOR 12 eclassifiecl and Release by Dn 1-2131 tD ED 135213 CREE 54FT3 DOCID 4009730 TOP SBCRBT Published Monthly by PI Techniques and Standards for the Personnel of Operations VOL III No 3 MARCH 1976 PUBLISHER WILLIAM LUTWINIAK BOARD OF EDITORS Editor in Chief ooo oo Arthur J Salemme 56425 Cryptanalys i ooo o ooo 1 P L 1f8025S Language oo o oo ooo oo o Emery W Tetrault 5236s Machine Support oo oo o Mathematics ooooo oooo ooo 1 13i Reed Dawson 3957s Special Research o o o Vera R Fil y 7ll9s Traffic Analysis o oo oo o Frederic O Mason Jr 41425 For individual subscriptions send name and organizational designator to CRYPTOLOG PI TOP SECRET 86-36 -- - DOCID 4009730 P L 'fOP SEEURRE'f UMBRA TO PULL A PONYAL I t March 76 CRYPTOLOG Page 'fOP SECRgT YMBKA 1 EO 1 4 c P L 86-36 86-36 OOCID 4009730 tfOP SBEURRBtf UMBRA EO 1 4 c P L 86-36 March 3-6 Convention of Teachers of English to Speakers of Other Languages 10th New Y9rk NY March 1 -15 North American Conference on Afro-Asiatic Linguistics 4th and American Oriental Society Philadelphia PA Marth 21-24 Association for Asian Studies Washinqton D C 1 lrch 25-27 Conference on Composition and Communication Philadelphia PA April 1-4 Northeast Conference on the Teaching of Foreign Languages New York NY April 3-5 Annual Meeting of the Linguistics Association of Great Britain Edinburgh Scotland Write J Christie Dept Linguistics 15 Buccleuch PL Edinburgh EH8 9LN Scotland April 5-9 International Conference of Nordic and General Linguistics 3rd Austin TX April 6-9 Acoustical Society of America Washington D C April 9-10 College English Association Ci ncinnati 00 April 15 Deadline for abstracts for LSA Surrmer Meeting April 15-17 Conference on African Linguistics 7th Gainesville FL Write P A Kotey Ctr African Studies 470 LGH Univ Florida Gainesville FL 32611 April 19 23 American Educational Research Association Las Vegas NV April 20-24 Congress of the International Association for the Study of Italian Language and Literature 9th Palermo Italy Write Robert J Cl ements AISLLI Rm 701 Main Bldg New York Univ Washington Sq New York NY 10003 April 22-24 University of Kentucky Foreign Language Conference Lexington KY April 22-24 Southwest Areal Language and Lin9uistics Workshop 5th San Antonio TX Write B L Hoffer Dept English Trinity Univ San Antonio lX 78284 April 22-25 Chicago Linguistic Society Chicago IL Write Chicago Linguistic Society Goodspeed 205 1050 E 59th St Chicago IL 60637 May 6-8 Conference on Perspectives on Language Louisville KY June 21-26 Conference on the Psychology of Language Stirlin9 Scotland June 2B-July 2 International Conference on Computational Linguistics Ottawa Onto Canada July 26-31 Phi 1ippine-American Conmunication Conference 1st Manila The Philippines -- August 17-19 World Congress of the InterLSA Bulletin national Reading Association 6th Linguistic Singapore Society of August 26-31 World Congress of Phoneticians 3rd Tokyo Japan AmeZ'iaa August 28-30 European Linguistics Society Datober 1975 Salzburg Austria Write G Drachman Institut F6r Sprachwissenschaft der Universit t Imbergstrasse 2 111 A-5020 Salzburg Austria September 1-4 International Phonology iiilo oiio o UNCIASSIFIED Meeting 3rd Vienna Austria March 76 TOP CRYPTOLOG S CRIi T Page 2 lJMBRIa DOCID 4009730 CONFIDEN IAL MUSINGS ABOUT THE AG-22 IATS Cecil Phillips C03 The foZZowing artiaZe is reprinted from C-LINERS C Group Maahine Proaessing Information BuZZetin VoL 3 No 7 August Sept mber Oatober 1975 March 76 CRYPTOLOG EO 1 4 c P L 86-36 CONFIDENTIAL Page 3 JI I bB VIA SSllHl'F SIM mEhS SI 'bY DOCID 4009730 CONFIBI3N'fIAL crIME'S C UNNING OUT You'd better hurry i f you 86-36 want to enter 1 4 c 'the CMI ESSAY CONTEST or the CLA ESSAY CONTEST this year The deadline for submitting entries to the CMI ESSAY CONTEST is 2 i March and the deadlIne for theCLA ESSAY CONTEST is 19 March For complete information see The NSA TeahniaaZ JournaZ Vol XX No 4 Fall 1975 Or call the following l -------------- UNCLASSIFIED Puz zle 6an6 ma themtLti cia ru CJtyp ta Yrai y-6 t6 H eJLe ' -6 an oppoll-twUty you won I t LIk1n t to rn U -6 The Cltyptoma the maUC6 Iru ti t ILte I CMI ha 6 a numbVl 06 copiu 06 PENNEY PUZZLES a colle ction 06 68 puzzlu by CMI' e xe c uti ve cUAe ctolt Wa UVl Penney WIUle the -6UPply laMA a 6Jie e c oP 06 t Jr U nlUcilutting un ir6- e d book w iU be -6e nt to e ve Jr yone jo-Uring ' 1'1 4 c To jobt -6end you Jr name -6oUal H y um 6-36 bVl OItgan izaUon te1 e phone numbVl and $3 00 60Jt 1976 duu to r I CMI TIte Mu Jr Vl S03 Comments anyone Mr Phil Ups was Chief of the ADVAGENS Joint Meahanization Group whiah aonduated the first tests of an AG-22 60 nd 1961 UNCf ASSIFIED C8UF I BEfiT lAb IPICC8 P L March 76 CRYPTOLOG 86-36 Page 4 CONFIDE TI I IhldlSbE 'ilA eOMIN'f' elb l IU LS OIIL'l P L 86-36 DOCID 4009730 UNCLASSIFIED An Objective Approach to SCORING TRANSLATIONS 1 --------OrR-e-pr-i-nt-e-d-f-ro-m-Q-R-L-r-Q-u-ar-t-er-ly-R-e-v -'e-'w or Lingu ists November 1973 Author's note The philosophy underlying the translation grading system desaribed in this paper has been developed and applied by Emery Tetrault and myself UJith many valuable suggestions from our aolleagues on Professional Qualifiaation Examination PQE Committees and from other Agenay Unguists My use of the pronoun 'We re fl eats this ao llaboration I personally take full responsibility for presenting our findings here Translation as an intellectual activity has been practiced since antiquity for practical as well as esthetic reasons but even today we have reached no consensus on whether it is even possible to fully bring over from one language to another the contents of a fairly pedestrian piece of prose let alone fiction or poetry We debate the merits of machine translation as a precursor to a truly scientific discipline or argue whether a free translation of a Greek lyric poem does not better catch the flavor of the original than one which adheres rather closely to the syntax of the original and results in an awkward Le literal rendering in the target language In short the domain of translation has not been really determined Our own work never demands fortunately the highly creative product of the artist translating Homer Dante or Push in but it does require something more than a purely mechanical approach good translators have a flair for couching in idiomatic English the thought content of the source-language passage Poor translators on the other hand seem never to master this complex operation regardless of how well they may know the source language or how well they may write English when not translating IP16 P L 86-36 tui tive judgments across language boundaries in source language-to-English situations Over the past 2 or 3 years my colleagues and I have developed a way to score translations which may obviate this problem to a large extent even though our results have admittedly been far from perfect total objectivity in grading any kind of connected text is of course impossible Our first large-scale use of the system which r will describe later was with the Russian PQE We have subsequently tried it in a number of other PQEs involving several languages mainly Indo-European but also from other families The results have been encouraging enough in both instances for us to recommend its use in the PQE Handbook Subjeative Approaahes to Evaluating Translations One approach to scoring translations is in a sense historical in that the evaluator tries to guess the reason for the translator's mistake and in so doing draws unwarranted inferences This attempt may be useful when the time comes to counsel the failing or borderline aspirant and the problems can be talked out Unfortunately though the history of an error is irrelevant to the rendering of a particular text and whether the aspirant should have learned a particular grammar rule in a course is beside the point A second method in wide use is to decide arbitrarily that a certain kind of mistake for example failure to recognize proper nouns and inappropriate tr lslation of them is sufficient cause to fail a paper This syndrome is evident among longtime subject-matter experts or specialists in some area of the world who see examinations solely in terms of their immediate interests One example of each approach should suffice In the first case a would-be translator missed a trickly negative conditional construction in If we had only the two kinds of translators the first sentence of a fairly long paragraph to judge we could sort them into sheep and which contained other negative and conditional goats and have done with it The fact is that elements much less difficul t In the interest of most aspiring translators are neither brilliant coherent discourse the translator overrode the nor hopeless but fall somewhere in between signals which he should have learned and proand when professional certification is at stake duced a sensible but wrong text A second transwe need in the interest of justice to the can- lator decided that the name of a group of isdidate as well as to the system reasonably lands which historically means fishing platobjective criteria for making a final decision form should be so translated Not only did on pM or 6a U he fail to recognize the proper noun but he also Lacking such criteria in our early attempts saw the context of an entire paragraph in terms of the translation and produced a peculiar to evaluate translations we ended too often text in English with intuitive decisions not transferable to similar cases involving the same source Ian The evaluators of the first passage initialguage nor was it even possible to compare in- ly decided that the paper should fail because March 76 CRYPTOLOG Page 5 UNCLASSIFIED DOCID 4009730 UNCLASSIFIED transfer of ideas and impressions will be the aspirant did not know enough of the grammar on the whole fairly simple Some kinds of of his language regardless of his performance texts are dependent on an intimate knowin the rest of the examination Fortunately ledge of a vastly different culture s b- we reconsidered the translation as a whole and system religious political and so on gave the paper a marginal pass The second These texts are never used for PQEs The faux pas made in an early PQE apparently caused student needs a vocabulary of a couple of such anguish to the evaluators that they failed thousand words the main carriers of ideas the paper mainly on this point before he can work efficiently regardless I hasten to add that none of the above is of how good his syntactic control may be meant in a caviling spirit Translations wherObviously many members of PQE Committees ever and by whomever done have ordinarily been have not had much formal linguistic training in assessed by these or other equally subjective methods What we propose here is an attempt to syntactic and semantic analysis excellent bring a measure of uniformity to grading trans- though they may be in the language they are con lations so that intuition and idiosyncrasy play cerned with Just as obviously there is no a minor role and some kind of objective standard time for us to give extensive training in these is established The framework and detail of the subj ects to people who are busy with other things However a certain minimum of linguistic thin - proposed system is described below ing is needed to make our system work and we Linguistias and Evaluation of Translation have tried to impart this informally to committees before they select passattes for translation Before discussing precise error weighting I should say a few words about linguistics dreary and evaluate them Once a committee has a general grasp of the system the members should be as the prospect may be and follow with some better able to select passages containing intercomments on how committees apply linguistic esting syntactic and semantic problems After principles in test evaluation having produced translations of passages against The first point to be made is that the grad- which papers will'be judged the committee can then to some extent predict likely errors When ing system to be described is based insofar as possible on a somewhat artificial distinction the papers are actually graded each evaluator between 4yntax and emantiC6 I will therefore works independently to derive a score based on define these terms insofar as they apply to the the weighting described below then compared his present case results with the others o Syntax is the system of relating linguistic Some differences in weighting are bound to elements to one another in phrase clause occur and may be irreconcilable but generally or sentence a process which is concerned the findings show a narrow spread of 5 to not with selecting appropriate grammatical more than 10 points The care required in such careful grading insures that at the very least forms hereafter called affixes positions for vocabulary items or a combinamajor errors are not overlooked a result not tion of both depending upon the particualways achieved by subjective grading lar language Syntactic analysis leads Our weighting of linguistic elements is oased' from particular texts to general statement on the assumption that aspirants should be exabout the linguist elements of a language pected to control their source-language syntax The traditional parsed sentence is an inA failure to understand relations within and ductive exercise proceeding from for examong sentences which are marked by position ample the sentence The quiak brown fox and or affixes as noted above can seriously jwrrped over the lazy dog to a set of names distort the meaning of an entire paragraph or symbols Definite article adjective ad- hence we take either 8 or 4 points off per erjective noun verb past tense preposiror depending on its severity Ii for examtion definite article adjective noun ple a sentence should have been translated We cannot however recapture this gem by Dog bites man but comes out Man bites dO h simply reversing the procedure because we would subtract 8 points because all the reexcept for the definite article a possibly lationships are wrong If a sentence which infinite number of combinations could apought to have been translated We stopped near pear Nonetheless a command of the synthe train is rendered We stopped the train tax of a foreign language which will inwe would take off 4 points since the subjectevitably differ from that of English at verb relationship is correct even though the many points is the first element of the rest of the sentence is wrong As we have set language the student necessarily learns 70 as a passing score four 8-point errors will fail a PQE o Semantics on the other hand is concerned with particular meanings of linguisticeleCertainly the examples above are oversimpliments in context To the extent that pat- fications of translation problems which can be terns of feeling and thinking between lan- knotty Our experience has shown however that guages are similar as to a large extent in 9 out of 10 cases committees can achieve conthey will be in the Indo-European langu-ages sensus on the kind and degree of syntactic error March 76 CRYPTOLOG Page 6 UNCLASSIFIED DOCID 4009730 UNCLASSIFIED not only for Russian examinees but also for aspirants in Indo-European and Semitic languages generally Where we have had problems it has been in languages such as Chinese or Vietnamese where the intersection of syntax and selDantics has not begun to be explored let alone taught to unversity or DLI students If aspirant translations from these languages go far afield and reverse the meaning of the original or even worse do not even fall in the correct subjectmatter area the point system above 'is still applied wrong is wrong It would of ourse be possible to set as passing a score lower than 70 for languages structurally removed from English but a decision to do so would be made on grounds other than linguistics Semantic errors are charged at a lower rate 2 points per error not because we regard vocabulary as unimportant -- far from it As I noted above a knowledge of syntax alone won't do for the idea structure mainly depends upon lexical units However our experience has shoWIl' that students of foreign languages pick up vocabulary unevenly probably because no systematic attempt to teach it has ever been devised and successfully applied Nor for that matter are students properly instructed in the use of either monolingual or bilingual dictionaries To the extent that vocabulary instruction is given at all it is given on the job and takes the form of ad hoc corrections and suggestions on how to use one or another dictionary to better effect This process may be tedious but provided the subject matter translated is not abstruse is easy in comparison with mastering a new syntax For these reasons it seemed best for us to accommodate to the existing situation and apply the lighter penalty EVen so we still encounter situations in which it is almost impossible to decide whether an error is in fact best viewed as semantic or syntactic If a source-language pronominal reference is translated by the wrong pronoun is this a simple error at a given point 1 h erea f ter punct Inear erro II or cou ld 1 t open the way fo mIsunderstandIng seve al sen ences syntactIC error Contextual Ju ent IS the only method for makIng such a deCISIon On the other hand some apparently pur 7 grammar markers such s tense fo s ca be mlstra slated WIthout l any way dlstort n g the ba lc noun-verb relatIonshIps Thus the dog b1 tes d th do b t the h e the same the man an e g 1 man av noun-verb-noun arrangement the dIfference lIes in the time of the event identified at a given point in the text A tense error is hence punctilinear and costs 2 points As I have suggested it is often difficult for PQE Committees to reach a consensus at every point of semantic and syntactic error in translation from the source language Thetask of evaluating the translation from the point of view of English usage and violation of convention is even harder but extremely awkward English does have an overall effect of blurring meaning or at the very least causing the reader unnecessary strain Violation of acceptable English we call poor cooccurrence When the members of a committee are unanimous in judging an English phrase to reflect poor cooccurrence one point is assessed Here are examples of the kind of poor usage we have met an oblivious scholar to practical matters instead of a scholar oblivious to awkward splitting of adjective and accompanying preposition the car collided against the bus instead of o with the bus a seemingly unlikely mistake for a native speaker of English to make but this type is of high frequency nonetheless Foreign Minister Rogers instead of Secretary of State Rogers a violation of conventional usage Hundreds if not thousands of similar peccadilloes have been and are being committed under the present system and are the object of heated remonstrances from the failing aspirant Why do they have to nitpick everything It has been our personal experience that committees on the whole err on the side of generosity in cases of one-point errors and overlook altogether misspellings and wrong punctuation where these do not cause major problems in understandin the English text Even those evaluators who take violent exception to misspelling will not fail a paper for this alone To sum up it has been our intention to establish a system of grading which will be as objective as possible Using a base of 100 and with 70 as passing we have specified 3 kinds of errors The first 2 syntactic and semantic concern command of the source language the third bears upon English usage Syntactic errors are of two orders of severity reflected in the 8- and 4-point penalties while semantic mistakes are charged 2 points and English misuse or poor cooccurrence costs one There are to be sure many doubting Thomas' es who question the efficiency of the system either because they feel it is too complex to master or that the time spent in mastering it might better be devoted to something else and that the intuitive or seat-of-the-pants method will do as well While conceding the time required to be considerable the final result we f 1 t h stifies the effort and in the ee more an Ju absence of a better system ought to prevail R - O Editor's note The system described in the reprinted article has been in use at the Agency since July 1972 is currently UJI'iting a paper inl-w h-1 -c h h ewill sumnarize the experience that the PQE Corrrnittees have had in using the system and will suggest theoretical and practical il1TprGvements I March 76 CRYPTOLOG Page 7 UNCLASSIFIED - - - - - ------------------ P L 86-36 DOCID 4009730 FOR OFFICIAL HSE eNLY ' ' P L How does an NSA language professional stack up against language professionals elsewhere That question cannot really be answered inasmuch as apples and pineapples are not directly comparable An insight at the very lowest level' of professionalization can be gained however by comparing the criteria for NSA' open-source translation Part IA of the Professional Qualification Examination and those for accredi tation by the American Translators Association ATA NSA's test The NSA language test is of such length in the foreign language as to generate a total of 450 to 600 words of English translation Time is not a factor therefore examinees can take as long as 4 hours for this section of the PQE Texts do not pose any problems of content substance that is they do not require an overly specialized knowledge of subject matter They are usually articles by professional journalists on a topic of interest to the general public -- at the difficulty level of Reader's Digest or a weekly news periodical Any lexical item in the text which is not in the standard bilingual dictionary is glossed in a footnote NSA graders use a points-off system in grading the test They give more relative weight to errors involving the ZogiaaZ relationships among major sentence parts and to identifying specific items in the source text that is markers for number case gender tense aspect mood voice etc Each sentence is graded by itself no more than 8 points is taken off for anyone sentence If the logical subject-verbobject complement relationship in the translation is not substantially the same as that of the source-language sentence no fewer than 4 and no more than 8 points are taken off Experience in 'arious test conunittees has shown that examinees who get the maximum of 8 points taken off on each of 3 sentences seldom pass the test For each translation error involving a singZe word content words such as nouns adjectives verb stems function words such as prepositions and conjunctions and grammatical affixes 2 points are deducted -up to a maximum of 8 points in anyone sentence For errors involVing poor selection of dictionary meanings of a lexical item or poor word aoZZoaations words that just don't fit together right in English only one point is deducted From the foregoing explanation it can be seen that a sentence in the English translation might contain one major error involving logical relationships say 8 points 3 single-word errors 6 points and perhaps 2 collocation or word-choice errors 2 points but the NSA grader would not deduct 16 points for that sentence -- instead he would deduct only the maximum of 8 points per sentence If the translation of Part IA of the Agency's PQE has fewer than 30 points deducted it is considered to be a pass But a translation that just passes is far from good It is not as simple as saying that a translation with 30 points deducted is 30% inaccurate Instead such a translation may be all wrong because it will almost certainly contain several errors of fact anyone of which could be critical in a real-life translation situation So if such a poor examinee were to be certified as a linguist and were to produce on an everyday basis operational translations containing such critical errors of fact one can only imagine the tremendous amount of work and responsibility that would be placed on the language checker who would have to correct the translations before issuing them Or one could imagine the inherent dangers if such translations in a stress situation were issued to the customer without being corrected March 76 CRYPTOLOG Page 8 Q Q I IAb HSE ONhY 86-36 DOCID 4009730 F8R 8FFICIAh MSE 8NhY i r If a particular grader deducts between 30 These selections do not reflect the diffiand 40 points the translation must be debated culty level of the foreign-language texts that in committee if the majority of the committee usually confront professional translators But members feel that enough of the minor errors the grading criteria do reflect the requirement would have been caught and corrected in the of the customers who let contracts for transchecking proofreading and analysis phases lators The ATA examinees are told to be prea translation with as many as 40 points might cise and fairly literal and to be extremely actually be allowed to pass PQE committee carefuly in paraphrasing so as not to lose one members sometimes lean over backwards in atiota of the sense of the clause or sentence tempting to pass certain marginal papers They They are warned that omissions of words except claim that if only a few examinees pass manfor such low-information words as schon doch agement tends to say that the test is too hard una zwar en effet znachit can be extremely never that the applicants are just not certifi- costly and can even result in failing a selecable as professional linguists tion A candidate must pass two out of the three passages selected Only one serious erOne thing must be borne in mind however A person who passes the PQE even with distinc- ror in a passage is permitted two serious ertion is only certified as a linguist it still rors constitute a fail for that selection of might take him many years of on-the-job experi- course a large number of nonserious errors could also constitute a fail ence to become a senior linguist something Following is a passage from a statistical that a person becomes without having a piece text Calcul des probabilit s that was used in of paper to prove it Dr Jaffe used to compare this situation with that of a medical stu- the French test administered at the ATA workshop in California in the summer of 1975 The dent being awarded a diploma from medical passage is followed in turn by three of the school and the right to apply for a license to practice medicine he would still need years of translations done by examinees for the convenience of the grading committee the translaexperience to become an expert in his field If tions are typed prior to grading The underthen a person who passes a PQE without debate lined words and the notations in the left-hand has demonstrated the potential to become a margin were made by one of the ATA graders a senior linguist the person who is allowed capital E is the symbol for a major error a after debate to squeak through a PQE has demonstrated if anything the ability to function lower-case e is the symbol for a inor error at only the very lowest level of linguistic 'mistr stands for mistranslation competency Cemparison of ATA and NSA grading systems ATA test The American Translators Association is a To compare the ATA and NSA grading systems different animal Its members include both the typed translations including the ATA gradliterary translators and technical translators er's underlines and notations were shown to an Literary translation is a creative art and is NSA linguist with experience in grading French not tested But since all translators make PQEs He made his own underlines most of their money from translating technical most of which coincided with errors noted by scientific and legal material technical transthe ATA grader and indicated in the right-hand lation is used for ATA accreditation The margin the number of points that would be delength of the ATA test is' approximately 750 ducted in accordance with the Agency's PQE words Time is a factor precisely 3 hours is criteria For example in translation #1 allotted for the test As with the NSA text the ATA grader considered the error for both the texts do not pose any problems of content mistranslation of pour les dieux for the substance lexical items not in the standard gods misread as pour les deux to be a major bilingual dictionary are glossed in footnotes error whereas the NSA grader deducted 2 points Selections tend to reflect the subject matter for it of professional translators international meetings symposia national projects medical Since the Agency's PQEs are usually twice as research technology science etc but in a long as the ATA French selection let us assume popularized version at the newspaper or Reader's that the translators would have had an equal Digest difficulty level error rate for an additional 300 words In that event the results would have been as follows Backgrounds and specializations of the ATA members vary greatly The examinees are acNSA grade cordingly offered five s elections in five someATA RCI1JJ TJ'ansExtrapolated Pass what different subject fields from which they lator grade score score fail choose three to translate The topics in the test that I recently took were the Apollo-Soyuz 1 Fail 84 68 Marginal flight the Helsinki Conference blood immunolo2 Fail 72 44 Fail gy a legal case involving patents and the 3 Pass 100 100 Pass prospective Canadian Health Service System March 76 CRYPTOLOG Page 9 F8R OFFICIAh MSE ONhY DOCID 4009730 fQR Qff E Ab HSE 8HbY French passage A -- from a statistical text Calcul des probabilites Translator #1 I LE HA5ARD I CHANCE Cormnent oser parler des lois du hasard Le hasard n'est-il pas l'antithese de toute loi Ainsi s'exprilne Bertrand au d but de son Calcul des probabilites La probabi lit est oppos e la certitude c'est donc ce qu'on ignore et par consequent semblet-il ce qu'on ne saurait calculer II y a la une contradiction au mains apparente et sur laquelle on a d j beaucoup crit Et d'abord qu'est-ce que Ie hasard Les anciens distinguaient les ph nomenes qui semblaient ob ir des lois harmonieuses etablies une fois pour toutes et ceux qu' ils attribuaient au hasard c' taient ceux qu'on ne pouvait pr voir pa ce qu'ils taient rebelles toute loi Dans caaque domaine les lois pr cises ne decidaient pas de tout elles tragaient seulement les limites entre lesquelles il etait permis au hasard de se mouvoir Dans cette conception Ie mot hasard avait un sens pr cis objectif ce 'qui tait hasard pour l'un tait aussi hasard pour l'autre et meme pour les dieux Mais cette conception n'est plus la notre nous sormnes devenus des d terministes absolus et ceux memes qui veulent r server les droits du libre arbitre humain laissent du moins Ie determinisme r gner sans partage dans Ie monde inorganique Tout ph nomene si minime qu'il soit a une cause et un esprit infiniment puissant infiniment bien informe des lois de la nature aurait pu Ie pr voir des Ie cormnencement des s1ecles 5i un pareil esprit existait on ne pourrait jouer avec lui aucun jeu de hasard on perdrait toujours Pour lui en effet Ie mot de hasard n'aurait pas de sens ou plutot il n'y aurait pas de hasard C'est a cause de notre faiblesse et de notre ignorance qu'il y en aurait un pour nous Et meme sans sortir de notre faible humanit ce qui est hasard pour I 'ignorant n'est plus hasard pour Ie savant Le hasard n'est que Ie mesure de notre ignorance Les phenomenes fortuits sont par d finition ceux dont nous ignorons les lois How dare we speak of the laws of chan Ysnot cnancethe antithesis every law Thus Bertrand expresses himself at the beginning of his Calculation of Probabilities Probability is op ' posed to certainty thus it is what is un- t t il ' known what we would not know how to A e 1II1 r calculate There is at least an apparent contradiction here one on which a great deal has already been written 'r' e First off what is chance The ancients distinguished between phenomena O which seemed to obey harmonious laws established once for all and those which they attributed to chance they were the ones' that couldn't be anticipated because they were unamenable to all law In each domain no precise laws were determined they tra ed-only-th bo d rI s- g which chance was allowed to operate In this concept the word chance had a precise objective meaning what was chance for one was also chance for the other and t ' even for both 'i @ 'b ' -'#1' - However this is o urwa YOflOOking_O ' o at it we have beCOllle absolute determinists - d even the S QJe obes who want to save a plac f oift iril iii mankin O nevertheleslf letdeterm inism hold untr ammeled s Jti in the inorganic world Every phenomenon nO ma ter how minute it may be a cause nd an infinitely powerful mind infinitely well informed of the laws of nature would have been able to foretell it since the'beginning of time If such a mind existed one would not be able to play f r any games of chance with iit ' -- one would always lose bI has e For that mind in fact the word chance would have no meaning or rather there would be no chance It is because of our weakness and our ignorance that there would be chance ti is -r wi _ s Also even wi thout-goi -goiitsId o -ourYeeble humanity what is chance to the ignorant is no longer chance to the man Chance is only the measure of our ig- I norance c hg n c e Phenomena by defi- I nition are those whose laws we do not know March 76 CRYPTOLOG Page 10 PSH SPPlE HSE S IH DOCID 4009730 eK e IeIAL HSE ON Translator #2 Translator #3 I CHANCE I CHANCE How does one dare to speak of the laws of chance Isn I t chance the anti thesis of all law Bertrand thus expresses himself at the beginning of his Calculation of Probabilities Probability is opposed to certainty it is thus what one ll does not know and consequently it seems ' what one does not k ow how to calculate @ There is at-least an apparent con1 tradictio one on which much has already been written tf f r PJ J first what is chance The an- CD $ I I1'ie ingle out phe omena that seeme y to obey harmonious firmly established laws and those they attributed to chance Y there were those that one couldn 't forsee@ because-they were g g to all law In each domain precise laws did not decide everything they only traced the limits between which chance was permitted to operate In this conception the word chance had a precise objective meaning that which was chance for one was also chance for the other and even for the gods But this idea is no longer ours we have become absolute determinists and even those who wish to reserve the laws of human free-Wil1 tM determin- j ism to rule without division in thein-@ elll 40rgam c v wor ld All - phenomenlJItl Evf IN o Oflit f d 'nI as small as t s has a cause an an e e iI infinitely powerful infinitely well formed to the laws of nature would have Been iffire to forsee it from the - fo't g ex sted no game of chance could E ' rr r t1 ginn ng T e en uries If CD @ Q @ '1t le bepTa-yed with it one would always lose For it indeed the word Q chance f not have any sense or-s-oon there f @ rr G' #0' would not be any chance It is because of our weakness and M our ignorance that %1T' there would be one f6F us And even j ' '01 e t'f'would 1 ' D without 2 our weak humanity that which is chance for the ignorant is no longer chance for the well-informed Chance is only the measure of our ignoe rance Fortuitous phenomena are by 11 t T C definition those of which we do not 0 know the laws f How can we dare to speak of the laws of h i not chance the antithesis of all law Thus speaks Bertrand at the beginning of his Calculation of Probabilities Probability is opposed to certitude thus it is what we do not know and so apparently what we cannot calculate There is here an at least apparent contradiction about which much has already been written First of all what is chance The ancients distinguished between phenomena which seemed to obey harmonious definitely established laws and those which they attributed to chance The latter were phenomena that could not be foreseen because they were contrary to all law In each field the precise laws were not all-determining they merely defined the limits within which chance was permitted to operate In this conception the word chance had a precise objective meaning what was chance for one man was chance for another and even for the godb But this conception is no longer ours we have become absolute determinists and even those who wish to reserve the rights of human free will at least permit determinism to reign unchallenged in the inorganic world Every phenomenon however small has a cause and an infinitely powerful mind infinitely well informed of the laws of nature could have foreseen it from the beginning of time If such a mind existed we could 'not play any game of chance with it we should always lose For this mind in fact the word chance would have no meaning or rather there would be no chance It is because of our weakness and our ignorance that one may exist for us Even without departing from our weak humanity what is chance for the ignorant man is no longer chance for the scientist Chance is merely the measure of our ignorance Fortuitous phenomena are by definition those whose laws we do not know J 3 - March 76 CR pTOLOG Page 11 FOR OFFICIA ijgB QN FOR OFFICIAb HSE OHhY tteJt 10 tlte dilo t Chief W ro 1 24 December 1975 I SUBJECT Cryptolog-J rticle Oct 75 In one of the gaps_1fihich occurred during the holiday season I got to read your ooo Proud and Bitter Memories-' __ article Your message is clear as are your fEi-elings about the people you worked with and f' il I think you've done a service for those of us here who didn't have much to do with the --_ problem by writing that article It wasn't enjoyable reading but I'm recommending it One final remark about NSA PQEs and ATA's accreditation examinations they are both designed to set the lowest threshold level of professionalization no more At NSA the emphasis is placed on accuracy of content In the ATA accent is placed on precise literal accuracy and on speed A person who can complete the exam in exactly 3 hours will be able to make a bare living just above the poverty level in translating material of that difficulty using conventional paper-and-pencil techniques and typing up the finished copy himself Translators therefore specialize in specific technical fields where they can get at least double or triple the basic translation pay rate Then too professional translators often dictate their translations into recording equipment and pay a typist to transcribe the translation into finished copy despite this additional expense this method is beneficial to the translator since it enables him to at least double his translation output Consequently a free-lance professional translator after gaining experience and one or more specialties should be able to earn at least $36 000 a year Even a part-time free-lance moonlighter translator can earn as much as $5000 a year These figures indicate that ATA accreditation is worth quite a bit in dollars and cents They also indicate that according to ATA standards the majority of certified NSA linguists are overpaid while the few truly professional certified linguists at NSA are underpaid ------------- FeR eFFIEHth H6E Q lb j to all in W I cc Ch B Ch PI Editor Cryptolog PIG Communications Security Establishment I very much enjoyed the October i sue of Cryptolog especially I I- proud and Bitter Memories and the extract from his Uncertain Origins which you put across as a splendid Double-Crostic Noting that you are obviously very enthusiastic about this kind of puzzle I am enclosing a copy of a related type which I constructed for fun but on completion realized was not economically viable since it is much more difficult and time-consuming to create than to solve I don't think I shall ever construct another one so this is probably an authentic O'Neill hapax l egomenon Kevin O'Neill Readers ho woul d l ike a copy of get one 56428 Mr O'Neill 's puzzl e TTt Zy by calling the Editor on UNCLASSIFIED March 76 CRYPTOLOG Page 12 PI-F b 76-S3-24 l67 L 86-36 DOCID _4_0_0_sg_7_30 This document is from the holdings of The National Security Archive Suite 701 Gelman Library The George Washington University 2130 H Street NW Washington D C 20037 Phone 202 994-7000 Fax 202 994-7005 nsarchiv@gwu edu