You are on page 1of 11

Volume 1, Spring Issue, 1988

B O O K R E V I E W
A N ART I F I CI AL I N T E L L I GE N C E A P P R O A C H
TO LEGAL REAS ONI NG
By Anne vonder Lieth Gardner.
Cambridge, Massachusetts: The MIT Press, 1987,
pp. 193.
Reviewed by Edwi na L. Ri ssl and*
Anne Gardner is both a lawyer and a coruputer scientist. She
obtained her J.D. from Stanford in 1958, ar:d her book is a revision
of her 1984 dissertation submitted to Stanford's Department of
Computer Science. She plays, in part, the role of pioneer; artifi-
cial intelligence ("AI") techniques have not yet been widely ap-
plied to perform legal tasks. Therefore Gardner, and this review,
first describe and define the field, then demonstrate a working
model in the domain of contract offer and acceptance.
I, THE FIELD: AI AND THE LAW
A. Artificial Intelligence
Malwin Minsky, in his preface to the collection Semant i c In-
f ormat i on Processing, defined artificial intelligence as "the
:science of making machines do things t hat would require intel-
ligence if done by [people]. 'u Better known successful applications
have included playing chess, identifying bacterial strains, recog-
nizing and manipulating structures built with childrens' blocks
and formulating concepts in set theory. 2 Typical programming
languages include LISP and PROLOG, while typical specialties
in AI research include case-based reasoning, expert (rule-based)
systems, natural language processing, nonmonotonic reasoning,
and learning from examples. Above all, the field is marked by its
current diversity, rapid growth, and apparent potential2
* Associate Professor of Computer and Information Science, University of Massachusetts
at Amherst; Lecturer on Law, Harvard LaW School. Professor Rissland will serve as Program
Chair for the second International Conference of Artificial Intelligence and the Law, current-
ly scheduled for late Spring, 1989 in London. Asecond, more technical version of this review
will appear in the Fall, 1988 issue ofA/Magazine.
I. S E M A N T I C I N F O R M A T I O N P R O C E S S I N G ( M . M i n s k y ed. 1 9 6 8 ) .
2 . S e e g e n e r a l l y W a l t z , A r t i f i c i a l I n t e l l i g e n c e , S C I . A M . , O c t . 1 9 8 2 . F o r a r e c e n t , r e a d a b l e
o v e r v i e w , s e e L i n d e n , P u t t i , ~ g K n o w l e d g e t o W o r k , T I M E , M a r c h 2 8 , 1 9 8 8 .
3. O p t i m i s m is f o u n d e d i n p a r t u p o n p a r a l l e l d e v e l o p m e n t s i n h a ~ ' d w a r e a n d o p e r a t i n g
systems, such as Intel' s 386 and MicmsoR's OS/2. See, e.g., Gralla, Chips Of f t he Old Block:
A History, PC WEEK, Jan. 19, 1988 at S/13 (noting the AI capabilities of new microcomputers).
224 Harvard Journal of Law and Technology [Vol. 1
B. Attraction
The l aw is an at t r act i ve domai n for AI r esear ch for sever al
reasons. Fi rst , t he l aw has a t r adi t i on of exami ni ng i t s own
r easoni ng process. Second, legal r easoni ng is stylized: one r easons
according t o stare decisis, wi t h cases and by analogy. Thi rd, much
legal knowl edge is r eadi l y accessi bl e and r el at i vel y wel l st ruc-
t ured, codified and indexed. Never t hel ess i t will not surpri se, and
may even pl ease, l awyer s to l ear n t hat t he Res t at ement , t he
Uni f or m Commerci al Code, and case l aw, like t he t heor i es of legal
r easoni ng proposed by Karl Ll ewel l yn, Ronal d Dworki n, Edwar d
Levi, and H. L. A. Har t , ar e of l i mi t ed i mmedi at e use to AI
pr ogr ammer s. Ther e is want ami dst pl ent y because t he cent ral
ques t i ons - what we know and how we know i t - a r e answer ed onl y
par t i al l y by such mat er i al , and even t hese par t i al answer s prove
difficult to har nes s comput at i onal l y.
Lawyer s' i nt er est i n AI t echni ques and sof t war e is nascent but
concisely expr essed in mar ket i ng t er ms. Lawyer s use tools t ha t
gat her, sift, and/ or st r uct ur e legal ar gument s i n a cost-effective
manner . LEXIS and WESTLAW, whi ch are essent i al l y keywor d-
based pr ogr ams si f t i ngl ar ge ful l -t ext dat abases, oper at e at a level
of sophi st i cat i on f ar bel ow t hat cont empl at ed by AI pr ogr ammer s
such as Al ane Gar dner , Mar ek Sergot , 4 my col l eague Kevi n Ash-
ley, 5 or ma n y ot he r s . 6 Yet p r o d u c t s s uc h as LEXI S a nd
WESTLAW do i ndi cat e t he pot ent i al ubi qui t y of legal anal ysi s
sof t war e and t he symbi osi s bet ween comput er science and legal
research.
C. The Hurdles
Several specific, fami l i ar aspect s of legal anal ysi s t hat chal-
l enge AI may be not ed.
Fi rst , logical deduct i on al one cannot resol ve all legal i ssues. In
Gar dner ' s t er ms, legal r easoni ng is a "rul e-gui ded" r at her t han
rul e-governed act i vi t y (p. 3). Legal r ul es have t he s t at us mor e of
heuri st i cs t han of t heor ems, in t ha t all rul es have except i ons, and
rul es ma y cont radi ct .
Second, t he t er ms empl oyed i n legal di scourse ar e "open-tex-
t ured. " That is, t he meani ngs or defi ni t i ons of ma ny legal t er ms
4. Sergot & Sadri, The British Nationality Act as a Logic Program, 29 COMMUNICATlONS
OF THE ACM 370 ( 1987l.
5. Ri s s l a nd & Ashley, A Case.Based System for 7}'ade Secrets Law, PROC. OF THE ANN.
CONF. OF THEAM. A. FOR ARTIFICIAL INTELLIGENCE 60 (1984); Ashl ey, ModellingL~galAr-
gument: Reasoning with Cases and Hypotheticals ( unpubl i s hed manus cr i pt ) .
6. See, e.g,, PROC. OF THE FIRST INT'L CONF. ON ARTIFICIAL ][N~rELLIGENCE AND LAW
(ACM 1987}. Ga r dne r he r s e l f offers a n excel l ent bi bl i ogr aphy (pp. 201-14), a nd pr e s e nt s i n
Cha pt e r 4 a s ur ve y of t he fi el d compl et e t hr ough t he s pr i ng of 1986.
Spring, 1988] Artificial Intelligence 225
and pr edi cat es are i nher ent l y i ndet er mi nat e i n t he phi l osophi cal
sense of natural kind classes.
Thi rd, legal quest i ons commonl y i nvi t e more t han one answer ,
yet legal ar gument is bot h t i me const r ai ned and resource limited.
For whi l e conflict, di sagr eement , and ar gument are par t and par-
cel of t he law, 7 t he l aw mus t provi de a t i mel y ans wer for one si de
or t he ot her, r eached at a socially accept abl e cost.
Fi nal l y, in law, t he answer s change. Whet her t he change is in-
crement al , as in t he mode of Kuhn' s nor mal science, or abr upt , as
i n a Kuhni an par adi gm shift, a har d l ear ni ng i ssues l ur k close t o
t he surface of t he legal i ssues t hat AI pr ogr ams at t empt to resolve.
To accomodat e even gr adual change, t he al gori t hms mus t adapt
to a dynami c knowl edge base: a shi ft i ng foundat i on of cases,
st at ut es, and t he indices, rul es, and nor ms whi ch mani pul at e
t hem. Event ual l y one mus t confront t he change in pr edi cat es and
in t he r epr esent at i on itself, ei t her t hr ough t he emergence of new
legal concept s or t hr ough t he subst ant i al modi fi cat i on of old ones. 9
In s umma r y, l egal r eas oni ng r equi r es cer t ai n mi ni mum
capabi l i t i es:
1. t he abi l i t y to r eason wi t h cases and exampl es, part i cul ar-
ly t hr ough anal ogy;
2. t he abi l i t y t o handl e ill-defined, open- t ext ur ed predi -
cat es;
3. t he abi l i t y to handl e except i ons;
4. t he abi l i t y t o handl e f undament al conflicts bet ween
rul es; and
5. t he abi l i t y to handl e change and nonmonotonicity.10
D. Several Possible Approaches
Each AI speci al t y ment i oned i n Sect i on A above is descri bed
bri efl y bel ow. The di st i nct i ons dr awn ar e noZ absol ut e; t hes e
speci al t i es ar e compl ement ar y and oft en overl ap.
Case-based reasoning seeks to emul at e t he case-based reason-
i ng of legal pract i t i oners. Cur r ent r esear ch is focused i n par t on
case memor y and indexing, as s es s ment of si mi l ari t y and rel evan-
cy, cont ext - dependent case compari son, and ar gument gener at i on
and eval uat i on. Ashl ey' s wor k wi t h t r ade secret s u is a fai rl y in-
cl usi ve cur r ent effort.
Logic and expert systems, adapt ed to t he doct ri ne of precedent ,
7. As Ga r dne r poi nt s out , not onl y ar e act or s i n our l egal s ys t e m free t o ar gue, on h a r d
ques t i ons t hey ar e expect ed t o do so (p. 3).
8. T. KUHN, THE STRUCTURE OF SClENTWIC REVOLUTIONS (1970}.
9. Legal r e a s oni ng t hus i ne vi t a bl y i nvol ves t he h a r d e s t pr obl ems i n l ear ni ng, s uch as bi as
( t he pr obl em t h a t a ny knowl edge r e pr e s e nt a t i on l a ngua ge di sposes one t o e mpha s i z e or mi s s
cer t ai n concept s/ and t he new t er m pr obl em ( concer ni ng t he pr ogr am' s own cr eat i on of new
concept s a s needed).
10. N n m n t n i c r e a s n i n g i nv I ves t he r ej ect i n a nd/ r r e f r m f pr i r deci si ns"
11. Ri s s l and & Ashl ey, supra not e 5.
226 Harvard Journal of Law and Technology [Vol. 1
empl oy "backchai ni ng" to resol ve legal predi cat es. The pr ogr ams
proceed t owar d conclusion by execut i ng act i ons necessar y to
achi eve i dent i fi ed subgoal s. The danger court ed here, as Gar dner
poi nt s out , is t ha t t he "rul es will r un out" before pr edi cat es have
been resol ved. Conflicting resol ut i ons al so ma y be out put .
Natural language processing, i ncl udi ng t echni ques for st or y
under st andi ng, is most readi l y appl i ed wher e st ereot ypi cal fact
pat t er ns and par t y roles ar e recognized. TMScri pt -based under-
s t a ndi ng t e c hni que s mi ght pr ove es peci al l y f r ui t f ul whe n
r est r i ct ed to shor t case summar i es, such as headnot es. I n general ,
however, nat ur al l anguage processi ng i n t he l aw confront s t he
same daunt i ng chal l enges of l anguage i n ot her domai ns. The law,
as a mi crocosm of huma n experi ence, pr esent s in aggr egat e t he
wi dest r ange of expressi on and i nt er pr et at i on, whi l e t he sense of
even common wor ds is probl emat i c, '3 and shi ft s wi t h context.
Fi nal l y, nonmonotonic reasoning syst ems ma y be adapt ed to
handl e l aw' s pr opensi t y to l i mi t or over t ur n pri or r esul t s. Ex-
ample-based learning may emul at e any syst em, i ncl udi ng t he
legal process, whi ch devel ops pr i mar i l y i n r esponse to exampl es.
II. THE BOOK AND THE PROGRAM: GARDNER' S AP-
PROACH
A. Gardner's Theory of Hard and Easy Cases
Gar dner ' s AI model refl ect s t he j ur i s pr udence of har d/ easy
quest i ons, par t i cul ar l y as di scussed by H.L.A. Har t , Lon Ful l er,
and Ronal d Dwor ki n (pp. 38-39). In Gar dner ' s model, har d ques-
t i ons can ari se i n t hr ee ways:
1. Ther e exi st compet i ng rul es.
2. Ther e exi st unr esol ved predi cat es.
3. Ther e exi st compet i ng cases.
Ha r d cases ar e det ect ed by t hr ee heur i st i cs whi ch resol ve
predi cat es:
1. "If an ans wer can be deri ved usi ng [t he pr ogr am' s Com-
mon Sense Knowl edge (CSK)] rul es and i f no obj ect i ons
(i.e., opposi t el y-deci ded cases) to usi ng t hi s ans wer can
be found, as s ume t he quest i on of pr edi cat e sat i sfact i on
is easy and t hat i t s ans wer is t he ans wer j us t deri ved. "
(p. 45).
12. Gardner herselfadopted this approech in the simpler context ofsentences and phrases
rather than in a continuous story.
13. E.g., Alien & Saxon, Computer.Aided Normalizing and Unpacking: Some Interesting
Machine-Processible Transformations of Legal Rules, in COMPUTER POWER AND LEGAL
REASONING (West 1985).
Spri ng, 1988] Art i f i ci al Intelligence 227
2. "I f no ans wer about t he sat i sfact i on of a legal pr edi cat e
can be der i ved usi ng t he CSK rul es, t hen look at cases. "
(p. 46).
3. ' ~/ ha t e ve r t ent at i ve ans wer has been deri ved usi ng non-
legal knowl edge, look for cases calling for t he opposi t e
answer . " (p. 46).
The har d/ easy anal ysi s proceeds i n t he manner of gener at e-
and-t est . As Gar dner says, "[T]he gener al i dea is, first, to allow
ever y undef i ned pr edi cat e i n a legal rul e t he pot ent i al for r ai si ng
a har d quest i on and, second, to provi de means for concl udi ng fair-
ly quickly, in any par t i cul ar case, t hat most quest i ons of predi -
cat e appl i cat i on ar e easy. " (p. 43) I f t he quest i on of predi cat e
sat i sfact i on can be det er mi ned by appl yi ng t he CSK rul es to t he
fact s, t he pr ogr am has r eached a tentative answer. Tent at i ve
answer s carl be overri dden by opposi t e cases t hat object. Overri d-
i ng al l ows t he pr ogr am to conclude t hat what appear ed at fi rst
gl ance to be a har d quest i on (because meani ngs coul dn' t be
resol ved) is in fact eas y (because t he l aw has shown how to resol ve
t hem), or t o spot har d quest i ons (where l egal r ul es or cases con-
flict) to whi ch t he CSK r ul es wer e appl i ed wi t hout difficulty.
Whi l e t he easy mi sl abal ed as har d l eads t o a was t e of resources,
t he har d mas quer adi ng as easy ma y l ead to f at al l y fl awed argu-
ment s and l ost cases.
B. The Program
1. Domai n and Task
Gar dner ' s pr ogr am wor ks i n t he subfi el d of offer and accep-
t ance from cont r act law. Cont r act l aw is r el at i vel y st abl e, what
Edwar d Levi woul d call a "second st age" domai n, !4 r oughl y cor-
r espondi ng to Kuhn' s nor mal science stage.15 Cont r act l aw is fair-
ly r el i abl y s t r uct ur ed by t he Rest at ement (Second) of Contracts,
abundant case l aw, and t he Uni f or m Commerci al Code. Gar dner
di scusses in gr eat es t det ai l her program' s t r eat ment of Ad a ms v.
Li ndsel l , t he cl assi c cl assr oom crossed-offers case. I nput is
- pr esent ed i n t he t ext i n t he gui se of "[a] t ypi cal exami nat i on
quest i on" (p. 4). The f ami l i ar t ask, i n brief, is to spot t he i ssues,
a qui nt essent i al l y "i nt el l i gent " process.
The i ssue- spot t i ng t as k ma y be vi ewed as a sear ch t hr ough a
space of possi bl e al t er nat i ve i nt er pr et at i ons of t he facts. Since not
14. E. LEVI, AN INTRODUCTION TO LEGAL REASONING 9 (1949) ('rhe second stage is the
period when the concept is more or less fixed, although reasoning by example continues to
classify items inside and out of the concept.").
15. T. KUHN, supra note 8.
228 Harvard Journal of Law and Technology [Vol. 1
all issues are wort h argui ng about, and in the context of an exam
or litigation t here are time and space constraints, t he nat ur e of
the search must of necessity be heuristic. The t ask is how to select
issues which raise defensible argument s, or, in the language of
search, how to distinguish t he plausible from t he merel y pos-
sible. 16
2. Input
Gardner' s program-whi ch is never given a persona with a
name- st ar t s with a set of facts which have been ent ered by hand
into a form acceptable to her program.
3. Structure
Legal knowledge in Gardner' s program is contained in two
sources: net work st ruct ures and legal rules. An Augment ed Tran-
sition Net work ("ATN") represent s the st andard states in a con-
t ract situation with i nt erpret at i ons of events as t he links, or
"arcs, ''17 bet ween them. Legal rules represent certain prototypi-
cal legal fact pat t erns and are drawn upon to resolve whet her an
ATN arc may be t raversed. Legal ant ecedent s necessary to con-
clude whet her a contract exists are formalized (e.g., reasonable-
certainty, exchange, may-be-willing-to-enter).
Gardner' s program has an ATN with t went y-t hree states,
t went y legal rules (of which two pairs are conflicting and t hree
pairs are complementary), and approxi mat el y 100 generalized
fact pat t erns. Her simplified t ransi t i on net work is reproduced
below: 18
4. Al gori t hm
To produce t he analysis of legal choices, Gardner' s program
employs several sources of legal, linguistic, and common sense
knowledge. Common sense knowledge is represent ed in two ways:
(1) a slot-filler l anguage is used to describe fact situations; and (2)
a hi erarchy of such things as events, states, and objects, is imple-
ment ed t hrough a mechani sm of Common Sense Knowledge
16. Gar dner relies upon Llewellyn to make this point (p. 9): "[W]hile it is possible to build
a number of divergent legical l adders up out of the same cases and down agai n to t he same
dispute, t here are not so many t hat can be built defensibly. And of t hese few t here are some,
or t here is one, toward which the prior cases pret t y definitely press." (K. LLEWELLYN, THE
BRAMBLE BUSH: ON OUR LAW AND ITS STUDY 73 (Oceana Publications 1960)).
17. "~rhe arcs of the net work correspond to event s t hat can change the legal relations be-
tween t he partiosZ(p. 124) There is no arc to represent the lapse of time.
18. ~State 0: No legal relations exist.
St at e 1: One or more offers are pendi ng the offeree has t he power to accept.
St at e 2: A contract ex-ists.
St at e 12: A contract exists, and a proposal to modify is pending." {pp. 123-241
Spring, 1988] Art i f i ci al Intelligence 229
rejection,
revoking acceptance
oounteroffer.
modification / [
o f o t h e r 4 " . p t a o c e /
acceptance,
rejection,
revocation,
death
rejection, counteroffer,
revoking acceptance acceptance plus
proposal to modify
Transi t i on Network for Offer and Acceptance Problems (p.124)
rules. The rules themselves are highly structured objects with
"ext r a" component s (such as "el i mi nat e- on- f ai l ur e" and
"eliminate-on-success," which prune ATN arcs from considera-
tion), and "secondary" antecedents, which provide some look-
ahead for information useful in resolving hard questions. 19
These ideas are embodied in an algorithm which forms the
backbone of Gardner' s program. This algorithm is outlined below.
The mechanisms of applying CSK and legal rules to work one's
way around the ATN are in fact quite involved and are discussed
at length in Chapters 5 and 6.
If (there is a t ent at i ve answer)
t hen if (there are no "opposite" cases)
t hen (the case is easy);
but if (there ar e opposite cases)
t hen if (there are also similarly aligned cases)
t hen (the law is unset t l ed and t he question is hard);
otherwise (let the technical legal rule override the
confilicting CSK rule; t he answer is easy).
If (there is no t ent at i ve answer)
t hen if(no cases match)
or if (two or more cases match but conflict)
t hen (the cases fail to resolve the issue and the question
is hard);
otherwise (the question is easy and the answer is t hat
indicated by the case(s) t hat match.
19. These CSK .'-ales are encoded in a standard fashion using Michael Genesereth's MRS
language (e.g., *one carload of salt" is "{carloads C1) (number C1 1)"), described in The MRS
Dictionary, Memo HPP-80-24, Stanford Heuristic Programming Project, Stanford University.
230 Har var d Journal of La w and Technology [Vol. 1
5. Out put
The out put of Gar dner ' s pr ogr am is a t wo-l evel anal ysi s rep-
r es ent ed i n t wo gr aphs (Fi gures 7.4 and 7.5, pp. 173-76). The
upper level of Fi gur e 7.5 summar i zes i nt er pr et at i ons of t he event s
i n t he fact si t uat i on. Br anch poi nt s r epr es ent har d ques t i ons -
poi nt s wher e al t er nat i ve, compet i ng i nt er pr et at i ons mus t be
recognized. I n effect, ~he upper level is a decision t r ee whose
br anchi ng nodes r epr es ent har d quest i ons and whose l eaf nodes
correspond to s epar at e sequences of i nt er pr et at i ons of t he event s.
Fr om such a r epr esent at i on of i ssues, i t is a shor t i nferent i al hop
to ans wer t he bi g quest i ons of whet her a cont r act exi st s and what
ar e t he ar guabl e i ssues. For each upper level br anch poi nt t her e
is a l ower level det ai l ed anal ysi s (such as i n Fi gur e 7.4) support -
i ng t he di vergi ng i nt er pr et at i ons.
Gar dner ' s model t hus is abl e t o anal yze i ssue spot t er quest i ons
such as t hat pr esent ed by a st yl i zed Ad a ms v. Li ndsel l . As she
di scusses, her pr ogr am' s ei ght maj or t wo- way br anch poi nt s and
ni ne concomi t ant anal yses ar e per haps mor e t han a huma n
l awyer mi ght consider, but t hes e br anch poi nt s compar e wel l to
her es t i mat e of a possi bl e sear ch space of five t o t he ni nt h power.
The pr ogr am r epor t edl y per f or med cr edi t abl y on an as s or t ment
of pr obl ems from Gilbert' s La w Summar i e s (pp. 183-88).
III. COMMENT AND CONCLUSI ON
Anne Gar dner is ent i r el y correct i n st r essi ng such par adi gms
as "open t ext ur e" and ' ~hard/easy." These ar e cent r al t o legal
r easoni ng and open up val uabl e l i nkages to ot her AI di sci pl i nes
as wel l as legal phi l osophy. Moreover, Gar dner ' s wor k gi ves some
real comput at i onal fl esh to t he phi l osophi cal skel et on of t he
har d/ eas y di st i nct i on, provi di ng a model for cri t i que by bot h legal
and AI pract i t i oners.
However , wi t h specific r egar d t o Gar dner ' s manner of di st i n-
gui shi ng har d and eas y quest i ons, I suggest t ha t any real i st i c case
base will i nvar i abl y provi de cont rary, opposi t e cases. Consequent -
ly, her heur i st i cs will char act er i ze near l y all quest i ons as har d.
One ans wer is to recognize t ha t not all opposi t e cases ar e equal ,
and t ha t some cont r ar y ar gument s ar e mor e r obus t t han ot hers.
Of course, t hi s r equi r es one t o r eason wi t h cases and, mor e dif-
ficultly, about a r gume nt s - a t r emendous t ask, but pr ef er abl e t o
overrel i ance upon r ul es and general i zed f act pat t er ns. I n pa r -
t i cul ar, I see cases as pl ayi ng a much mor e cent r al r01e t han t hey
do i n Gar dner ' s model. Therefore, I woul d addr ess wi t h some care
t he det ai l s of i ndi vi dual cases whi ch ar e critical for i ndexi ng, anal -
Spring, 1988] Artificial Intelligence 231
ogy, and other aspects of case:based reasoning. However, using
cases in a more central role and not j ust as annotations or exis-
tential embodiments of concepts would require deep changes in
Gardner' s program.
Nevertheless, Anne Gardner has done l andmark research in
t he field of AI and legal reasoni ng. Not wi t hst andi ng our
"jurisprudential" differences, I feel t hat Gardner has made a sub-
stantial contribution to AI and legal reasoning which will have
major impact upon both disciplines. Not only has she performed
ground-breaking work on such topics as issue spotting, open tex-
t ured predicates, and the hard/easy paradigm, she has also
pointed the way to furt her work on case-based reasoning and ar-
gumentation. The book is well written and has served my classes
at Harvard and Amherst well. In particular, her introduction and
explication of legal philosophy from the perspective of AI (Chap-
ters 2 and 3) are unrivalled in the current literature. An Artifi-
cial Intelligence Approach to Legal Reasoning is essential to
anyone working in the field, and invaluable to those who wish to
work in the field. And t hat is an easy question.
Spring, 1988] Books Received 233
B OOKS R E C E I V E D
IF WE CAN KEEP A SEVERED HEAD ALIVE..., by Chet Fleming.
Polinym Press, P.O. Box 22140, St. Louis, Missouri 63116. pp. 461
(1987).
AN INTRODUCTION TO KOREAN INTELLECTUAL PROPERTY LAW, by
Byong Ho Lee. Central International Law Firm, Korea Rein-
surance Building, 80 Soosong-dong, Chogno-ku, Seoul, Korea. pp.
453 (1987).
ENGINEERS AND THE LAW / AN OVERVIEW, by Bruce Schoumacher.
Van Nostrand Reinhold, 115 Fifth Avenue, New York, New York,
10003. pp. 337 (1986).