Berkeley Technology Law Journal

Volume 1
Article 2
Issue 1 Spring

January 1986

Full-Text Databases and Legal Research: Backing

into the Future
Robert C. Berring

Follow this and additional works at: https://scholarship.law.berkeley.edu/btlj

Recommended Citation
Robert C. Berring, Full-Text Databases and Legal Research: Backing into the Future, 1 Berkeley Tech. L.J. 27 (1986).

Link to publisher version (DOI)

https://doi.org/10.15779/Z38ZW98

This Article is brought to you for free and open access by the Law Journals and Related Materials at Berkeley Law Scholarship Repository. It has been
accepted for inclusion in Berkeley Technology Law Journal by an authorized administrator of Berkeley Law Scholarship Repository. For more
information, please contact jcera@law.berkeley.edu.
FULL-TEXT DATABASES AND LEGAL RESEARCH:
BACKING INTO THE FUTURE
BY ROBERT C. BERRINGt

INTRODUCTION
The use of computers in legal research is a topic at the center of
any discussion about modem legal literature. The appearance of com-
puter terminals at each accredited law school and the presence of on-line
systems at every large law firm signal a major change in the way that
lawyers conduct research. As more and more print publishers introduce
on-line versions of their traditional hard copy products, 1 the legal pro-
fession is increasingly receptive to a primary role for computers in legal
research. 2 This article will examine one3 crucial role for the computer-
the full-text on-line computer databases.
© 1986 High Technology Law Journal

t Professor of Law and Law Librarian, Boalt Hall School of Law, University of Cali-
fornia, Berkeley; Dean, School of Library and Information Studies, University of Califor-
nia, Berkeley. I wish to thank my colleagues Daniel Dabney, Bill Maron, and Patrick
Wilson for helpful suggestions, and my research assistant Kathleen Vanden Heuvel for
her invaluable help. Also, I wish to thank High Technology Law Journal editors Mark
Ryland, Cecelia Fusich, and Jeff Love for their contributions to this article.
1. See, e.g., BNA LABORLINE, an on-line product by the Bureau of National Affairs
similar to its looseleaf service. Commerce Clearing House and Prentice Hall are also
preparing on-line products that cover the same territory as the companies' loose leaf
products.
2. See, e.g., Yates, Nearly Everything You Want to Know About Data Bases, A.B.A.J.,
Nov. 1985, at 90 (includes a list of over 70 databases "for virtually any area of law or
any area related to the practice that you can imagine").
3. There are two important and related topics that I will not consider. First, the use
of non-legal databases in legal research is growing. For example, many lawyers are be-
ginning to use NEXIS, a database of Mead Data Central, that includes all on-line infor-
mation services available from Mead except for those provided to the legal community
as LEXIS. The NEXIS database consists of the full text of 142 newspapers, magazines,
wire services and newsletters. NEXIS is available through the LEXIS system. See MEAD
DATA CENTRAL, INC., GUIDE To NEXIS AND RELATED SERVICES (1985).
Second, the use of computers in law office management and litigation support is bur-
geoning. See Legal Times, Sept. 30, 1985 (Special Supp. 1985 Fall Law Office Equipment
and Services Directory), which lists an enormous variety of computer services. Whole
periodicals such as Computer Lawyer and PC Lawyer are devoted to this topic, and even
established periodicals such as the American Bar Association Journal have continuing
columns on law office computing.
28 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

The giant on-line databases produced by Mead Data Central

(LEXIS) and the West Publishing Company (WESTLAW) are the main
arena for the face-off between lawyers and computers. They have
grown from simple repositories of case law into integrated databases of
primary and secondary source materials. 4 The systems are full-text5 and
allow free-text searching 6 for any word or combination of words, with
query structures that incorporate the power of Boolean logic. 7 This com-
bination of sources and search capabilities creates an entirely new genre
of legal literature.
LEXIS and WESTLAW already have become an integral part of the
arsenal of research tools available to the lawyer, but we need to re-
evaluate the role that they play in legal research. Recent studies have
raised questions about both the general efficacy of full-text systems, and
the research skills necessary to use them efficiently. These questions
must be examined in order to assess the usefulness of the legal data-
bases. One purpose of this article is to explore some problems in compu-
4. The legal databases now include various citation services, numerous legal periodi-
cals, libraries for specialty practices, ALR annotations, etc. In this article I will focus on
the use of the databases to search for judicial opinions, but much of what I will say
could be applied to other on-line source materials.
5. A "full-text" database is one that incorporates every word of every document rath-
er than the more usual (and less expensive) method of putting only an index entry or
abstract of the document on-line.
6. "Free-text searching" enables the researcher to search for every occurrence in the
database of any word or combination of words without using a pre-existing index.
7. "Boolean logic" is a syntactical calculus used for the comparison of data items
(words or numbers) and combinations of data items. In Boolean logic, data items can be
related in only one of two ways: true (matched) or false (not matched). For purposes of
searching, one data item can be combined with others using the Boolean operators
"and," "or," and "not." With the use of conjunctions, disjunctions, and negations, a
search can list instances in a database where a given item or a combination of items ex-
ists.
The power of a Boolean search is this ability to match items that have specific rela-
tionship within a document. In a full-text search system, such as LEXIS or WESTLAW,
the use of these conjunctions allows the researcher to create a context - to specify a re-
lationship between the terms for which he is searching. For example, without conjunc-
tions, the searcher would use one term at a time, calling up every instance of the word
in a database. He would then have to examine each of these retrievals in order to dis-
card those items not relevant to his issue. The use of a conjunction, "and," would allow
him to search for instances where two words (or numbers, etc.) are found in a single do-
cument, paragraph, or sentence. The conjunction allows the researcher to specify a rela-
tionship between two terms, and thus formulate a more precise search.
The advantage of this search technique over a prepared index is that the researcher
can find every occurrence of a significant word. This allows the researcher to both nar-
row the search to specifics and to broaden it, as he or she does not rely on the preselec-
tion of certain cases or sections of a document by the individuals who created the
prepared index. This method also has inherent weaknesses. See infra text accompany-
ing notes 48-86.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 29

ter-based research, and to point out certain significant limitations of on-

line full-text legal databases.
But we cannot examine the utility of legal databases in isolation.
The impact of LEXIS and WESTLAW is not simply a matter of a new
technology simplifying or speeding up a preexisting process; it involves a
change in the structure of legal literature. More work needs to be done
on the relation between the structure of legal literature and the substan-
tive development of law,8 but it seems clear that in law, more than any
other discipline, the structure of the literature implies the structure of the
enterprise itself. 9 I will attempt to show this interrelation by first describ-
ing the history of the traditional hard-copy primary sources, and assess-
ing their influence on the lawyering process. Next, I will examine the
emergence and growth of the legal databases. Finally, I will point out
certain practical problems inherent in full-text searching and in making
each lawyer his or her own on-line researcher, and suggest some
theoretical difficulties with this new form of legal literature.

I. THE STRUCTURE OF THE OLD PARADIGM:

THE WEST REPORTER SYSTEM
Before the arrival of computerized legal research in the 1970s,
American legal publishing was a highly integrated and well-developed
system of comprehensive publication and retrieval in hard copy. Most
aspects of this system can be attributed to a few enterprising publishers
who conceived of the intriguing publication formats. A brief sketch of
the hard copy system is necessary to put the advent of the legal data-
bases in context.

A. The Development of the System: Comprehensive

Regional Reporters
The publication of case reports was organized, systematized and
perfected by the West Publishing Company of St. Paul, Minnesota at the
end of the nineteenth century. 1 John B. West, an entrepreneurial office
8. Cf. Childress, The Hazards of Computer-Assisted Research to the Legal Profession, 55
OKLA. B.J. 1531 (1984) (suggesting certain tangible and intangible links between the
structure of legal literature and styles of practice).
9. Recall Langdell's famous aphorism, "The library is to us what the laboratory is to
the chemist or the physicist and what the museum is to the naturalist." HARVARD LAW
SCHOOL ASS'N, THE CENTENNIAL HISTORY OF THE HARVARD LAW SCHOOL 1817-1917, at 97
(1918).
10. See W. MARVIN, WEST PUBLISHING COMPANY: ORIGIN, GROWTH, LEADERSHIP (1969)
for an extensive if over-flattering portrait of the history of the West Publishing Com-
pany. For a more balanced treatment, see Woxland, "Forever Associated with the Practice
of Law": The Early Years of the West Publishing Company, LEGAL REFER. SERV. Q., Spring
1985, at 115-24.
30 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

supply salesman, noted the disorganization of the case reporters that his
lawyer-customers were purchasing. The existing forms of publication
were slow, unorganized, and inaccurate.1 1 In response he began to pub-
lish The Syllabi, a reporter that contained the text of Minnesota Supreme
Court cases and summaries of decisions from surrounding states. The
Syllabi was so successful that West introduced a successor entitled
Northwestern Reporter.12 This series contained the full-text of all deci-
sions from those states that Mr. West considered the "northwestern" re-
gion. Publication was frequent, and the reporter was inexpensive and
reliable. It was a success.
West soon realized that a "regional" reporter, which gathered to-
gether the decisions of a variety of jurisdictions into a single series, was
useful to lawyers and, consequently, easy to sell. One of the motiva-
tions for the regional reporter approach was to gather enough cases to
produce a regional biweekly advance sheet that was marketable to
lawyers in a number of different jurisdictions. These advance sheets,
which might have been too costly to produce for limited circulation in
each individual jurisdiction, delivered judicial opinions quickly into the
hands of the lawyers throughout the region. West made agreements
with various courts to obtain decisions directly and rapidly, and spared
no effort to locate opinions. His deserved reputation for completeness as
well as accuracy rapidly earned him a substantial following.
Northwestern Reporter was only the first step in West's process of
innovation. Mr. West extended his system nationwide by dividing the
entire country into seven regions and by producing a reporter for each
region. 13 Within a few years, West Publishing Company provided
comprehensive coverage of all state cases. The introduction of Federal
Reporter and Supreme Court Reporter in 1886 completed this pattern. 14
Although more established publishers also created regional reporters,
no one emulated West's decision to divide the entire nation into seven
11. Young, A Look at American Law Reporting in the 19th Century, 68 LAW LIBR. J. 294
(1975), provides an overview.
12. The title Northwestern Reporter was actually used to describe two separate publi-
cations. The first Northwestern Reporter, introduced in 1877, was an enlargement of The
Syllabi. It was still more like a newspaper and included the full text of only Wisconsin
and Minnesota decisions. The Northwestern Reporter as we know it first appeared on
April 26, 1879. F. HICKS, MATERIALS AND METHODS OF LEGAL RESEARCH 145-46 (3d ed.
1942).
13. John B. West's division of the country into seven geographic regions demonstrat-
ed his talent as a publisher but it did not show him to be prescient as a geographer.
Ironically, West did not anticipate the development of the West; Oklahoma is not con-
sidered by many a Pacific state.
14. For example, New England Reporter, Central Reporter and Western Reporter by the
Lawyer's Co-operative Publishing Company of Rochester, New York. See W. MARVIN,
supra note 10, at 48.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 31

regions and to provide coverage of all state court decisions. Indeed, con-
temporary commentators lambasted the idea as being wasteful and
greedy. 15
But it was this national coverage that most historians regard as the
foundation of West's success. 16 Lawyers were entranced by the availabil-
ity of cases from all jurisdictions in a standard, inexpensive format.
Whether by dint of the product's attractiveness, its price, or its market-
ing, the National Reporter System was a resounding success. Other
competitors soon dropped out of sight and1 7 left West with a dominant
position as the unofficial publisher of cases.
Although official reporters sponsored by the various jurisdictions
continued to exist, 1 8 the comprehensive West publication system became
the most prominent feature of American case reporting. During the
same period the number of cases being rendered into written opinions
was rapidly increasing, 19 lending impetus to West's scheme. Moreover,
West's traditionally high standards of speed and accuracy in publishing
enhanced his system's reputation and marketability. The fact that the
West regional reporter structure has thrived until today is testimony to
its quality and usefulness.

B. The Structure of the System: Headnotes and

the American Digest
The American Digest System 2 0 was the key aspect of the new form
of legal literature that Mr. West created. The Digest classified all areas
of law into seven broad categories. These categories were then subdi-
vided into some four hundred and thirty topics. Each topic was then
further subdivided into subsections called "Key Numbers" (a trade-
marked term). These Key Numbers allowed the topic to be broken into
as many subdivisions as were necessary to completely cover that area of
15. The New Reporters, 19 AM. L. REV. 932 (1885).
16. See, e.g., Woxland, supra note 10, at 116.
17. See id. at 122.
18. Today, for example, only 29 states publish official reporters at all, and many only
include cases from the highest court of that jurisdiction. See HARVARD LAW REVIEW
ASS'N, A UNIFORM SYSTEM OF CITATION 136-76 (13th ed. 1981). The only officially pub-
lished federal cases are the Supreme Court cases in United States Reports.
19. The literature bemoaning the volume of published cases is vast. A personal
favorite is High, What Shall Be Done with the Reports, 16 AM. L. REV. 435 (1882).
20. In 1889, West acquired U.S. Digest from the Little Brown Company and editor
Benjamin Vaughn Abbott. U.S. Digest was modified and published by West as its Amer-
ican Digest System. Even more important was the acquisition of Complete Digest be-
cause its editor, John Mallory, also came to West. Mr. Mallory is acknowledged in the
preface of First Decennial Digest as the guiding hand behind the American Digest subject
scheme.
32 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

the law. 21 Eventually, a structure of subject headings was created which

provided for every possible legal issue. A headnote always had a specific
2
location in the Digest System.
West Publishing Company developed an elaborate process for
melding the cases into the Digest. As cases arrived, a lawyer-editor read
each one, editing it first for citation form and other stylistic conventions.
Then the editor prepared a set of headnotes that served as abstracts of
each point of law contained in the decision. Each headnote was as-
signed to a specific topic and Key Number location. 2 3 A headnote could
be assigned to two locations but it had to fit into at least one Key
Number address. The text was then passed to one of West's four senior
editors who verified the accuracy of the topic and Key Number assign-
ments. Although no statistics are kept, according to my conversations
with senior editors, a substantial number of topic and Key Number as-
signments were modified at this point. The importance of the placement
of the headnote into the Digest's subject index cannot be overem-
phasized. This initial placement had a tremendous impact on any subse-
quent manipulations of the data. In recent advertising, West indicated
its internal valuation of the senior editors by calling them "Edi Knights."
The West Digest System exemplified a type of index called a
universal subject thesaurus. 2 4 The concept of a universal subject
thesaurus, while not unusual in information science, reshaped legal
research. For when West Publishing created the Key Number System, it
not only enabled lawyers to research cases by subject, it also allowed
and encouraged lawyers to fit every legal issue into a certain conceptual
framework. At a mechanical level, the West Key Number System
created a comprehensive subject format that allowed for all of the cases
appearing in the National Reporter System to be arranged by subject ac-
cording to their headnotes; the power of the system made it the primary
21. For a classic description of West's American Digest System and a list of the
categories and topics, see F. HICKS, supra note 12, at 233-43.
22. The practical and theoretical implications of this closed-end system are explored
below. See infra text accompanying notes 27-33.
23. The editor writing the headnote and assigning it to a particular Key Number was
engaged in a purposive enterprise - fitting the case into the system. The headnotes
were tailored to this purpose. See infra text accompanying notes 30-32. Thus, even
those merely using the headnotes in a West reporter and not using the Digest proper are
affected by the structure.
24. Notice that the West thesaurus is limited to the legal universe. Some universal
thesauri are truly universal; they cover the entire universe of subjects. One example is
the subject thesaurus of the Library of Congress. SUBJECT CATALOGING DIVISION, LIBRARY
OF CONGRESS, LIBRARY OF CONGRESS SUBJECT HEADINGS (9th ed. 1980).
For an excellent summary of indexing theory, see Dabney, The Curse of Thamus, 78
LAW. LIB. J. 5, 9-17 (1986).
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 33

method of case retrieval. 2 5 But the West system did much more than
that. The Key Number System provided a paradigm for thinking about
the law itself. Lawyers began to think according to the West categories.

C. Strengths and Weaknesses of the System

The Digest System had both enormous strengths and unresolvable
weaknesses. The strengths were the comprehensiveness of the coverage,
the reliability and accuracy of the West editorial staff, and the fact that
an increasingly large number of published cases could be fit into a recog-
nizable and stable subject format. Furthermore, West was able to main-
tain the consistency of its system because it was both the publisher and
the indexer of cases. These strengths were reinforced by the existence of
Shepard's citators, an extremely accurate cross-referencing device.
Perhaps the most important characteristic of the Digest System was
that the West editorial staff acted as a national fixed point in the spin-
ning universes of state common law judges and lawyers. The editors
were trained to "normalize" 2 6 judicial opinions that used strange
language or strange analysis or otherwise appeared to be anomalous, to
25. The West National Reporter and Key Number Systems became an even more
powerful research tool when combined with the citation service provided by the
Shepard's Company. Shepard's developed its series of citators during the early part of
the twentieth century. Frank Shepard was a book salesman who recognized the utility
of comprehensive citations for cases. His idea was straightforward: he would provide a
service that noted every subsequent mention of a particular decision by any other case.
This categorization would allow the researcher ready access to any other decisions
which might modify, expand upon, or even comment on the subject decision.
The miracle of Shepard's was its accuracy and comprehensiveness. It covered every
court in every jurisdiction. Eventually, Shepard's expanded to cover codes, constitutions
and ancillary tools as well. But the heart of the system was always the total,
comprehensive coverage of all cases. The other part of this miracle was its reliability.
Early on Shepard's established an outstanding record for reliability. The literature of the
Shepard's Company on the death of Mr. Shepard demonstrates the company's real pride
in the reliable accuracy of its products, and lays out the Company's philosophy in
delightfully purple prose:
The present management of Shepard's Citations would have to be men of the dramatic Ra-
bot type, with hearts of steel and souls devoid of sentiment, to escape the thrills of satisfac-
tion that come with the realization of the worth while [sic] achievements of their organiza-
tion, and not to realize the enormous debt of gratitude which is owed to the editorial, busi-
ness, and mechanical forces of the Company for their loyal and unselfish service and their
constant devotion to the principles of accuracy which is the one outstanding and indispens-
able feature from which there must be no departure in any Shepard publication.
PUBLISHERS EDITORIAL STAFF, THE FRANK SHEPARD COMPANY, A RECORD OF FIFTY YEARS OF
SPECIALIZING IN A FIELD THAT IS OF FIRST IMPORTANCE TO THE BENCH AND BAR OF THE UNIT-
ED STATES 9 (1923).
26. In other words, opinions should be fit appropriately into the West analytic
scheme. It is an interesting question *whetherWest editors engaged in a kind of com-
mon law decisionmaking, classifying a case by inferring the "proper" holding from the
pattern of facts and the outcome, while downplaying the actual language of the opinion.
34 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

bring them back into the orthodox mainstream, to make them fit past
cases and present expectations. But the centripetal force exerted on the
law by the West staff was also a weakness of the system, as we shall
see.
The major weaknesses of the American Digest System were four in-
terrelated problems. The first two problems were quite practical, while
the second two were more theoretical. First, the West editors could
make mistakes. Second, the tremendous scope of the West universal in-
dex combined with the felt need for precision in the subdivisions created
a deeply layered index. Third, the universal index was inflexible and
resistant to change. Fourth, the editor steeped in the paradigm of the
Digest always interpreted cases in a way that fit the paradigm.

1. Mistakes
One major weakness of the West System was the fact that the very
editor whose job was to "normalize" the judicial language and to
correctly index the decision was also subject to human error. The West
indexer/editor who wrote the headnote or the individual who assigned
the subject location could make a mistake. And if this individual mis-
27
placed the headnote, it might be lost forever.
It's difficult to assess the extent of this problem. My impression is
that it was not severe. West's reputation for accuracy was well
deserved. Still, these kinds of mistakes, even if minor, are eliminated by
free-text searching in computer databases, so now they seem like an un-
necessary weakness.

2. Layered Indexing
There is also a large practical problem inherent in the complex
structure of the West Digest. Modem indexing theory criticizes deeply
layered indexes. A layered index is one that creates a series of
subclassifications in order to increase precision. Many of the topics in
the West Digests had such multi-subdivisions, and thus created
significant hazards for the searcher. The depth of the indexing in the
West Digest System, which resulted from a desire for precision, itself
27. Consider the example of certain purposely "lost" cases. The California Supreme
Court occasionally "depublishes" appellate court decisions that have already appeared
in the advance sheets by ordering that they not be included in the official reporters. Be-
cause West publishes their California Reporter and Pacific Reporter so rapidly, subse-
quently depublished cases are sometimes included. However, West does not put the
headnotes from these cases into the Digest System. This effectively depublishes the
case since no one can ever find it. Nevertheless, lawyers want access to these depub-
lished cases. This has led to their inclusion in the on-line databases. See infra note 84.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 35

became a problem. 2 8 A researcher had to figure out not only the first in-
dexing term, but perhaps the second, third and fourth term in order to
find the desired case. The West Publishing Company, while striving to
provide the best possible descriptive indexing of the case, actually made
it harder for the uninitiated to locate items. Only an adept West editor
could maneuver with ease through the variegated latticework of sub-
sub-sub-subclassifications. Thus, the Key Number System represented a
marvelous achievement in its breadth and precision, but its achievement
concealed significant risks for unsophisticated researchers.

3. The Rigidity of the System

Another problem of the Digest was the inherent rigidity of the sub-
ject structure itself. Naturally, the system developed by Mr. West in the
1880s that supposedly provided a discreet subject category for every po-
tential legal issue could hardly have endured for a hundred years
without showing some significant strains. But the size of the system ar-
gued against active adjustments to it. In order to update the system, the
West Company introduced a number of entirely new topics, both at the
issuance of each decennial cumulation and during the publication of the
General Digest volumes. 2 9 The introduction of such topics required a
28. The example chosen by Dabney, supra note 24, at 13, can hardly be improved
upon. West "Securities Regulation" Key Number 327 is layered as follows:
Securities Regulation
II. State Regulation (Blue Sky Laws)
(C) Offenses and Prosecutions
325. Criminal Prosecutions
327. -Evidence in General
That is five levels of subject breakdown between the user and the lead. -And this is just
the index!
29. The NINTH DECENNIAL DIGEST, which covered the years 1976-1981, was the first
compilation issued after only five years, a response to the growing volume of cases to be
processed. This Digest included 24 new or revised topics:
Abandoned and Lost Property Deposits and Escrows
Abortion and Birth Control Dower and Curtesy
Accountants Employers' Liability
Administrative Law and Procedure Extortion and Threats
Bankruptcy Extradition and Detainers
Chemical Dependents Illegitimate Children
Condominium Implied and Constructive Contracts
Consumer Credit Internal Revenue
Consumer Protection Public Utilities
Copyrights and Intellectual Property Urban Railroads
Credit Reporting Agencies Zoning and Planning
Debtor and Creditor
36 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

Herculean purge of the entire system of headnote classification in order

to locate all relevant topics and cases and to rearrange them into the
new subject order. The difficulty and expensiveness of this process
caused West to be fairly reserved in its introduction of topical
modifications.
The effect of the natural rigidity of the West System on the legal
system is unclear. Nevertheless, it is interesting that American legal
literature of the last century was controlled by a paradigm that was na-
turally both conservative and orthodox during a time when many as-
cribed these characteristics to the law itself. The West System was con-
servative in the sense that it resisted change; it was orthodox in the
sense that it self-consciously attempted to maintain internal consistency
and coherence in American law. The instrument of conservatism was
the rigid index. The instrument of orthodoxy was the editorial staff
placing new cases into the national index.

4. The Purposive Role of the Editor

A trained editor could clarify and "normalize" the language of an
opinion. He or she could likewise assess a judge's "real" meaning, read
language in context, correct for idiosyncrasies in style or expression and
then classify the words in the subject structure. But by intervening in
the research process, and by inserting his or her own interpretations, the
editor foreclosed other potential classifications of the subject matter.
Subsequent researchers always felt the mediatingpresence of the editor
in the very location of the case within the Digest.3u
There were two problems that stemmed from the interposition of
editorial judgment. First, the West editors had to choose between alter-
native characterizations of the issues in the case, and to choose between
possible locations for the issues in the West subject thesaurus. Disagree-
ments between editors about proper interpretation and classification
have been shown to be disconcertingly common. 3 1 Even if a certain
choice was not a "mistake," it could be less than optimal.

Since 1981 three new or revised topics have been added to the General Digests. They
are:
Children Out-of-Wedlock (1983) (formerly "Bastards")
Public Utilities (1982)
Commodity Futures Trading Regulations (1984)
30. Of course, each case is actually inserted into the Digest in a number of locations
because the headnotes correspond to the "issues" in the case.
31. See, e.g., Zunde & Dexter, Indexing Consistency and Quality, 20 AM. Doc. 259
(1969). Even one editor will classify the same materials differently at different times.
Id.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 37

In addition to the inevitable indeterminacy of subjective editorial

judgments, the judgment of the West editors was inevitably skewed in a
particular direction, or, more accurately, frozen in a certain shape. Be-
cause of the purposive nature of the editorial process, 3 2 the interpretive
range of the West editors was bounded by the intellectual universe of
the Digest. Subtle shifts and deflections in the attitudes and language of
judges under pressure from new social or legal forces were treated ex-
actly like idiosyncrasy and anomaly. Thus, the greatest strength of the
Digest System - its centripetal force, its "normalizing" will to orthodoxy
- was also its greatest weakness.
Was the editor, then, a friend or foe? For the pre-computer age
lawyer the answer was a resounding "friend." Because no one could ac-
cess the cases in any other efficient fashion, even those who despised
the digests had to use them. The primary significance of the advent of
LEXIS and WESTLAW is that they appeared to eliminate the necessity of
33
a mediating editorial staff, as we shall see.

D. The State of Legal Literature, B.C. (Before Computer)

For all of its flaws, the complete set of case reports with its
comprehensive subject arrangements were powerful tools. By the mid-
dle of the twentieth century, the legal system had available to it in the
publication of its cases a comprehensive system of document production.
This system included total subject availability and a comprehensive and
accurate system of citation. It made legal literature unique among other
disciplines. 3 4 No other discipline had invested the resources and time to
develop these extremely efficient manual systems.
Thus, when on-line databases first appeared, they were not particu-
larly attractive to the legal researcher. The possibility of comprehensive
retrieval and indexing combined with the citation services offered on the
new databases were no nirvana for the legal researcher. She already
had such tools available on her desk. There was no need to wait for the
32. See supra note 23.
33. See infra text accompanying notes 35-47.
34. The relatively high degree of integrity and cohesiveness in legal literature is prob-
ably inevitable, with or without the historical fact of the West System, because law is a
field where the primary source materials have normative force. In most disciplines,
researchers are interested in sources because of the quality the work and the intellectual
power of the authors (or the lack of it), or out of historical interest; but later workers in
such disciplines are in no sense bound by the work of their predecessors. In legal litera-
ture, the primary materials (cases) provide (or, for the legal realist, appear to provide) le-
gal workers with a crucial type of binding social norm - law - that people have to
know in order to structure their relations with others, and to restructure relations that
have broken down.
38 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

conversion of the information or to suffer through the inevitable training

and start up difficulties. As a result, law was slow to turn to on-line
sources for information.
But the change did come.
II. THE NEW PARADIGM: THE ADVENT OF LEXIS
AND WESTLAW
A. The Development of the System
The LEXIS system appeared nationally in the mid-1970s. The enor-
mous cost of creating a full-text database of cases was a significant entry
barrier, but Mead Data Central persevered and, with the active coopera-
tion of state bar associations, eventually expanded its LEXIS database to
include judicial opinions from every state. The history of this expansion
is described elsewhere. 3 5 The end-product was a national system con-
36
taining on-line the full text of every printed case.
As LEXIS made headway, West introduced WESTLAW. In its first
incarnation WESTLAW did not contain the full text of court decisions.
Instead it utilized only the text of the headnotes. WESTLAW used the
power of the computer and free-text searching to enhance its already ex-
isting manual system. This decision proved to be a disaster. West had
failed to grasp the nature of the new research tool, and the real
significance of the new form of legal literature. Why would a lawyer
bother to learn the mechanics of computer research to access the Digest
System which had been designed and perfected as a manual, hard copy
research tool? West soon caught on and began including the full-text of
37
decisions in addition to the headnotes.
When LEXIS initially marketed its system, the most frequently
heard criticism was that free-text Boolean searches were inappropriate
for retrieving judicial opinions. Critics pointed to the variety in judicial
language and to the difficulty of locating desired opinions by attempting
to specify exact common terms. For example, in 1975, Professor J.
Myron Jacobstein of Stanford Law School challenged LEXIS demonstra-
tors at an American Association of Law Libraries convention. He
described the facts and law of a particular case concerning a child and
asked them to locate it. What Professor Jacobstein knew and the
demonstrators did not was that the opinion had uniformly referred to
35. See Harrington, A Brief History of Computer-Assisted Legal Research, 77 L. LIBR. J.
548 (1985); Burson, Report from the Electronic Trenches: An Update on Computer-Assisted
Legal Research, LEGAL REFER. SERV. Q., Summer 1984, at 3.
36. Actually, the full-text databases generally extend their coverage only back to the
1920s or 1930s, although backward expansion continues.
37. This is the "Full-Text Plus" system described in note 68, infra.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 39

the child as an "infant." Because both LEXIS and WESTLAW could

only retrieve the exact terms entered into their databases, 3 8 the
chagrined demonstrators could not find the case.
Both systems struggled to resolve these problems. LEXIS and
WESTLAW added the capability to truncate search terms so that the 39
searcher could retrieve all items by searching with the roots of words.
In addition, both systems altered their search software to automatically
retrieve plurals and convert statutory alphanumerics. 40 As familiarity
with the systems grew, researchers waxed in confidence, and the
relevance of the4 1 kinds of questions posed by the Jacobstein challenge
seemed to fade.

B. The Structure of the System: Full-Text Databases and

Free-Text Searching
LEXIS and WESTLAW's use of a full-text format was a big step.
The computerized research systems that were coming into use in other
disciplines generally did not contain the full text of documents. Instead,
they consisted of abstracts or index entries. 4 2 These systems were highly
38. If, for example, the searcher wanted to find all cases that analyzed the rights of
unwed fathers concerning adoption of their children, she could frame the search as:
father & child w/15 adoption
This search strategy would retrieve all cases that contained both the word "father" and
the word "child," where the word child appeared within fifteen words of "adoption."
The Jacobstein challenge demonstrated that if a judge had referred to the child
throughout the opinion as "infant" or "son" or "baby" or "minor" the case would not
be retrieved.
As we shall see, the searcher can attempt to resolve this problem by including
synonyms in the search request, but this search strategy results in the retrieval of an
unwieldy number of cases. Many of the cases found by such a search will be irrelevant.
As the number of search terms are increased, the retrieval of unwanted items escalates.
The searcher is therefore placed in a dilemma. To ensure the retrieval of desired items
she must expand the search. Such action, however, increases the search's cost while it
likewise increases the inefficiency of the retrieval. This is the problem of the inverse re-
lation between Recall and Precision. See infra notes 51-56 and accompanying text.
39. For example, the search described in note 38 could be changed to:
father & child w/15 adopt!
This search would retrieve cases that used words like "adopts" or "adopting" in addi-
tion to "adoption."
40. There is speculation that the search software will be modified by so-called
"artificial intelligence" techniques to include synonym retrieval, but that does not seem
likely in the near future.
41. However, several recent papers have raised it again in sharp relief. See infra text
accompanying notes 48-72.
42. Examples are MEDLINE, a database of medical information that indexes and pro-
vides bibliographic citations to articles in over 3,000 journals and chapters from selected
monographs, and SOCIAL SCISEARCH, which indexes and provides bibliographic cita-
tions to articles in 4,500 social science and scientific journals. Neither offer the full text
of the indexed documents.
40 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

efficient search tools, but they left the researcher with the task of docu-
ment location. In addition, they retained all of the problems associated
with the mediating role of the indexer. LEXIS and WESTLAW, on the
other hand, were based on a different concept - the idea that the
researcher needed a totally integrated system that freed him from any
index-imposed restraints and that allowed him to examine the full docu-
ment on-line. The full-text feature made the systems more expensive
than on-line index systems, 4 3 but they also provided the innovation that
attracted users.
43. Also, the expense of the systems was added to that of traditional research tools.
The current cost of the systems is hard to assess and compare because WESTLAW
charges a regressive flat rate for on-line time (starting at $150 per hour for the first three
hours per month) while LEXIS charges a fixed amount per file access, search, search
modification, plus on-line time.
The law librarian at Control Data Corporation recently published some comparisons
that are useful. Griffith, Dual-System Research: The Best of Both Worlds, Legal Times,
March 17, 1986, at 9, col. 1. Griffith divided the searching universe into four categories
that roughly corresponded with search habits in his office, and ran identical searches on
each system: (1) five "search and browse" searches - researching more than 10 minutes
on-line; (2) 12 "search and cite" searches - quick-answer research (less than 10 minutes
on-line); (3) eight "cite check or retrieve" searches - using Shepard's or retrieving a
specific case; and (4) five "case retrieval" searches - finding and retrieving a case of
unknown citation. His findings were:
Research Request WESTLAW LEXIS
Type and No. Retrievals/Cost Retrievals/Cost
(1) "browse"/ 5 98 / $ 127.19 108 / $115.31
(2) "quick" / 12 1202 / $ 57.89 1616 / $207.31
(3) "cite" / 8 - /$ 17.01 - /$ 13.60
(4) "find" / 5 - / $ 9.66 - /$ 62.58
These search costs are significant, but notice that both systems give users access to (at a
minimum) all cases in the West National Reporter System. Bound volumes of all these
cases are not cheap. The base prices for West reporter sets are:
Reporter Set Volumes Price
Atlantic Reporter 2d 1-499 $ 12,081.75
Federal Reporter 2d 1-776 $ 14,537.50
Federal Supplement 1-617 $ 11,876.00
Northeastern Reporter 2d 1-484 $ 11,746.75
Northwestern Reporter 2d 1-375 $ 8,502.50
Pacific Reporter 2d 1-706 $ 16,732.75
Southeastern Reporter 2d 1-335 $ 7,357.50
Southern Reporter 1-477 $ 11,744.50
Southwestern Reporter 1-697 $ 14,026.25
Supreme Court Reporter 1-106 $ 3,308.50
Total $ 111,914.00
(Additional volumes cost about $35.00, and each additional term of the United States
Supreme Court costs $137.50.).
Prices obtained from Donald Blockhus, Sales Rep., West Publishing Co. (Mar. 1986)
(available at High Technology Law Journal). Obviously, a set of reporters would pay for a
lot of computer time. For a law office that mainly uses cases from a few jurisdictions, it
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 41

With LEXIS and WESTLAW the researcher could locate material by

using, not a predetermined subject thesaurus or index, 4 4 but the free-text
searching method. Using search commands that incorporate Boolean
logic, the researcher retrieved documents by requesting cases that con-
tained a specific term or terms. Boolean logic allowed the terms to be
linked by occurrence, proximity, section of a document and various com-
binations thereof.
The real breakthrough with LEXIS and WESTLAW, however, is
that they eliminate the intervention of any editorial judgment. It is now
possible to research efficiently without the mediating presence of the
West editors. No editor or index stands between the language of the
opinion and the researcher as he or she frames the search request. This
simple fact vaporizes the full range of complaints that had accumulated
against the old West system. Editors and antiquated subject structures
no longer burden the research process.

C. Strengths and Weaknesses of the System

The full-text, free-text searching of the on-line literature frees
researchers from many of the serious flaws of the old paradigm. First,
an editor can no longer "misplace" a case by misinterpreting a decision
and placing it in the index in a way that forecloses access by research-

may be cheaper to use a computer system to do research in other jurisdictions than to

buy a set of seldom used reporters. Of course from a client's point of view, these price
differences may not be significant because the cost of a lawyer's time dwarfs the costs of
any legal research method.
44. To increase retrieval speed, full-text databases actually have an index, but the in-
dex contains every word and word root in the database, along with a description of
every location of that word in the database. (Words like "a" or "the" are not included
in this list.). This kind of index is called a "concordance." For example, the word root
"adopt" is indexed along with each of its locations by document, paragraph, sentence
and position in the document. The words "father" and "child" are indexed with the
same location information. If the search request is:
father & child w/15 adopt!
the computer will find "child," "father," and "adopt" in the concordance and compare
all of their locations. Whenever "child" and "adopt" occur within 15 words of each
other in a document that also contains "father" the computer will retrieve that docu-
ment.
The concordance scheme is essential. If the computer had to search every document
in the enormous database one by one, the searches would be interminable. Also, users
of the systems searching for some word "x" may have noticed that they receive the
message "The word 'x' is not in the database" extremely rapidly, far faster than a search
that actually retrieves documents. This is puzzling until the user understands that the
first part of the search is a search of an index that contains every significant word in the
database.
42 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

ers. 4 5 Every case is equally available to every researcher, limited only by

the researcher's training and ingenuity.
Second, there are a number of functions that researchers with full-
text database capability can perform that could have been performed
only inefficiently or not at all by a person employing manual research
techniques. Intelligent use of the "segment searchers" on both LEXIS
and WESTLAW systems is one good example; cross referencing search
terms by a particular judge, a particular court, a particular date, or even
by such indicia as a name of a particular party can yield helpful and
practical information. Or, when legal research problems concern a
specific object, one that has a unique name or phrase that describes it -
for example, a product or trademark - the computer searches can yield
every use of the unique term in the entire corpus of cases. Also, a
researcher can find every case mentioning a certain code section or a cer-
tain previous case. This kind of research was simply impossible in the
old system, and it can be extremely useful.
Third, the new form of legal literature eliminates the rigidity in-
herent in the West paradigm. In the old system, the information in each
of the cases was parsed into a preexisting framework that inevitably
tended to suppress subtle changes and to enforce judicial and profes-
sional conformity and conservatism. In the new legal literature, the in-
formation is strewn into a free-form database without differentiation.
The legal databases provide no guidance and place no restrictions on the
way that judges and lawyers think about cases. The use of specific
words and the presence of specific facts become more important to the
researcher than the "holding" of a case or any other abstract generaliza-
tion about the law. The new paradigm is not merely a more flexible
structure than the old. The new paradigm has no structure at all. 4 6
Fourth, and perhaps most important, the absence of an index
means the absence of indexers. There is no "normalizing" editorial
force, no will to consistency, coherency or orthodoxy. To the extent that
these characteristics are seen as desirable, the responsibility for maintain-
ing them is placed squarely on the shoulders of judges and lawyers, and
not on an anonymous functionary in the bowels of West Publishing.

45. Sometimes, however, typographical errors in the databases can have a similar
effect. See infra note 65.
46. The databases have no subject-matter structure. The division of the legal data-
bases into "files" or "libraries" provides a sort of structure, but this generally amounts
to classification by jurisdiction, precisely the arbitrary system of classification abjured by
the West National Reporter System and American Digest. However, information per-
taining to certain specialized areas of practice, e.g., trade regulation and bankruptcy, are
increasingly gathered together in special files and libraries in both systems.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 43

The legal databases make available the raw materials of legal research as
never before: raw.
But, inevitably, the new system has created its own problems, prob-
lems inherent in the new research process and the new form of legal
literature. 4 7 The two most significant practical problems are, first, the
questionable efficacy of free-text computer searching, particularly in
enormously large databases, and, second, the tendency of those using
and promoting the new paradigm to see every lawyer as an appropriate
end-user of the systems. The third problem is more theoretical: what
kind of legal practice will cohere with a form of legal literature that
makes judicial opinions available according to practical search skills and
that interposes no mediating and integrating editorial judgment between
the raw legal materials and the practitioner?
1. The Mechanical Limits of the System: the Efficacy of
Free-text Searching
The Jacobstein challenge demonstrates that lawyers have shown
some concern about the efficiency and accuracy of free-text searching
since the inception of the legal databases. These concerns have focused
on the ability of free-text searching to deal with the vagaries and variety
of thought in language, and the concern with the effect of even small er-
rors in the huge legal databases. I will examine these criticisms by re-
viewing two excellent articles that have brought them back into the
limelight.

a. The Reemergence of the Issue: Blair and Maron's Study

In March, 1985, David Blair and M.E. Maron published an article
that caused a flurry of interest among legal researchers. 4 8 The two
researchers had a marvelous opportunity. They worked with a large,
operational, full-text document-retrieval system that was set up to serve
as a litigation support system in an actual case. The system contained
approximately 40,000 documents (roughly 350,000 pages of text) that
were thought pertinent to the defense of the lawsuit. With complete ac-
cess to a large full-text database, with search software similar to LEXIS'

47. In many respects, the problems of the new paradigm appear to be the flip-side of
the problems of the old paradigm. This fact, along with conventional prudence, seems
to suggest that currently the optimal research tool is the two systems used together. See
infra text accompanying notes 87-89.
48. Blair & Maron, An Evaluation of Retrieval Effectiveness for a Full-Text Document Re-
trieval System, 28 COM. ACM 289 (1985).
44 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

and WESTLAW's, 4 9 and with sufficient funding to back them, 5 0 Blair

and Maron were in an unusual position to attempt a test of the efficiency
of full-text search systems.
Blair and Maron were primarily interested in two measures of re-
trieval effectiveness: Recall and Precision. "Recall measures how well a
system retrieves all the relevant documents; and Precision, how well the
system retrieves only the relevant documents." 5 1 If Recall is low the sys-
tem is retrieving only a small percentage of the total number of relevant
documents. If Precision is low the system is retrieving too many useless
documents. In full-text searching systems, Recall is inversely related to
Precision. 5 2 Most lawyers probably would be more immediately con-
cerned with Recall. They would want all the relevant materials, even if
they have to weed out a lot of irrelevant stuff. But in reality, Precision
is just as important. Low Precision in a large database produces what
researchers call "output overload." A high Recall/low Precision search
in a large database might retrieve 1000 documents of which 700 or 800
are irrelevant. Most organizations don't have the necessary time or
resources to cull that much information.
The study utilized a database searching team made up of two legal
assistants and two attorneys, all of whom were intimately familiar with
the case and the content of the computerized litigation file. When an at-
torney wanted to see certain information from the file, he or she would
give a written description of the research to one of the assistants. The
legal assistant would frame an inquiry and run a computer search. The
results of the search were evaluated by the requesting attorney. 5 3 If an
attorney was not satisfied that 75% of the relevant documents in the da-
tabase had been retrieved, he or she would ask to have the query refor-
mulated and run again. The research was considered complete (usually
after a number of searches) only when the attorney was satisfied that
the search had produced 75% of the desired documents. When the at-
torney was satisfied, Blair and Maron's team would compare the number
49. The search software was IBM's STAIRS, an acronym for STorage And Informa-
tion Retrieval System. Id. at 289.
50. Their project cost almost half a million dollars in direct and indirect expenses. Id.
at 298.
51. Id. at 290. Recall is the ratio of the relevant documents retrieved by the search to
the total number of relevant documents in the database. For example, if a database con-
sisted of 1000 documents, 100 of which were relevant, then a search that retrieved 50 of
the relevant documents would have 50% Recall. Precision, on the other hand, is the ra-
tio of relevant documents retrieved to total documents retrieved. For example, if a
search retrieved a total of 75 documents, 50 of which were relevant, then the Precision
of the search would be 50 - 75 = 66%.
52. Id. at 293.
53. Id. at 291.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 45

of relevant documents retrieved by the search (or searches) with the total
number of documents retrieved to determine the Precision of the
searches. 5 4 The computation of Recall was much more complex, 5 5 but it
amounted to a very conservative estimate of Recall.
The results were surprising and dismaying. The full-text retrieval
litigation support system proved to be a fairly inefficient search mechan-
ism. On the average it retrieved about 20% of the desired documents,
i.e., Recall was about 20%. On the other hand, Precision was relatively
high at about 79%. The study also confirmed the observation of earlier
studies that Recall and Precision are inversely related. 5 6 Even more in-
teresting was the fact that the lawyers working with the research team
had estimated the Recall efficiency of the system at a minimum of
75%.57
The most crucial fact about the Blair and Maron study is that it was
the first time a file of this size was used to study full-text searching with
Boolean operators. Seminal studies that "demonstrated" the desirability
of full-text searching had used smaller databases. 5 8 It was only because
54. Id.
55. Because it was impossible to have the two attorneys (who were making all
relevancy determinations) read the entire 350,000 pages of text in order to find all of the
relevant items, the researchers had to find another way to calculate Recall. However, it
is not clear from the article how Recall was estimated. Their explanation is contained in
one paragraph:
To find the unretrieved relevant documents, we developed sample frames consisting of sub-
sets of the unretrieved database that we believed to be rich in relevant documents (and from
which duplicates of retrieved relevant documents had been excluded). Random samples
were taken from these subsets, and the samples were examined by the lawyers in a blind
evaluation; the lawyers were not aware they were evaluating sample sets rather than re-
trieved sets they had personally generated. The total number of relevant documents that ex-
isted in these subsets could then be estimated. We sampled from subsets of the database
rather than the entire database because, for most queries, the percentage of relevant docu-
ments in the database was less than 2 percent, making it almost impossible to have both
manageable sample sizes and a high level of confidence in the resulting Recall estimates. Of
course, no extrapolation to the entire database could be made from these Recall calculations.
Nonetheless, the estimation of the number of relevant unretrieved documents in the subsets
did give us a maximum value for Recall for each request.
Id. at 291-92 (emphasis in original). It's hard to see how the last sentence - claiming a
maximum value for Recall - follows from the explanation.
In a telephone conversation, author M.E. Maron explained the Recall calculation as
follows. The authors found rich subsets by using very broad search techniques. (For a
more complete description of this process, see Dabney, supra note 24, at 28-29.) They
took random samples from these rich subsets and had the lawyers evaluate them think-
ing that they were search results. From the number of relevant documents in the ran-
dom samples, they extrapolated Recall for the rich subsets. Then, they postulated that
there were no other relevant documents in the database. This technique, although still
an estimate, is quite conservative and does approximate a maximum value for Recall be-
cause there were bound to be relevant documents outside the rich subsets.
56. Id. at 293.
57. Id. at 295.
58. E.g., Salton, Automatic Text Analysis, 168 SCIENCE 335 (1970); Swanson, Searching
Natural Language Text by Computer, 132 SCIENCE 1099 (1960).
46 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

of an unusual opportunity that Blair and Maron could afford the ex-
tended time and effort needed to evaluate a large system. Their evalua-
59
tion raises some serious questions.

b. The Curse of Thamus: Finding Words but not Wisdom

One explanation for the very low recall rate described by the Blair
and Maron study is that the human use of language is inexact. Full-text
searching is premised on the assumption that "it is a simple matter for
users to foresee the exact words and phrases that will be used in the
documents they will find useful . The problems of imprecise
"...-60
usage, synonyms, jargon and even misspellings challenge this assump-
tion. To quote Blair and Maron, "it is impossibly difficult for users to
predict the exact words, word combinations, and phrases that were used
by all (or most) relevant documents . Daniel Dabney provides an
"...,61
interesting analysis of these kinds of problems in his recent article The
62
Curse of Thamus.
Dabney divides the problem of matching words into three
categories: synonymous words, ambiguous words, and complex expres-
sions. The first two categories - synonymous words and ambiguous
words 6 3 - involve the problem of linguistic imprecision. Because judges
can refer to a person or a thing in many different ways, it is difficult to
be certain that any search term or terms will retrieve the relevant cases.
Blair and Maron argue that if the earlier studies had utilized large databases they
would have reached less sanguine conclusions. Unlike the litigation database investigat-
ed by Blair and Maron, the small databases could be searched with high Recall and low
Precision techniques without "output overload." Blair & Maron, supra note 48, at 298.
59. There is at least one potential limitation on the applicability of the Blair and
Maron study to the LEXIS and WESTLAW systems. A litigation support file contains a
heterogeneous mix of documents that includes, among other items, reports, memos,
letters, invoices, transcripts of meetings, conversations, etc. The LEXIS and WESTLAW
databases are primarily composed of judicial opinions, a relatively homogeneous form of
discourse. To use Dabney's example, a judge might call a child a "minor" or an "in-
fant," but it is unlikely that he will call a child a "punk" or "rug rat." Still, even if
free-text searching in an on-line legal database were twice as efficient as the litigation
support database studied by Blair and Maron, a 40% recall rate still would be uncom-
fortably low.
60. Blair & Maron, supra note 48, at 295.
61. Id. at 295. This is, in substance, the same concern raised by critics at the advent
of the legal databases. See supra text accompanying notes 37-40.
62. Dabney, supra note 24. The title of the article comes from a legend in the
Phaedrus of Plato. According to Plato, the Egyptian King Thamus disapproved the in-
vention of writing by the god Theuth. Thamus thought that because the mere posses-
sion of writing could not give wisdom, writing would cause far more harm than good.
Dabney notes that we possess an almost unimaginable amount of writings, but asks
"how are we to extract from this almost incomprehensibly large collection of written
records the knowledge that we need?" Id. at 5-6.
63. Id. at 18-19.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 47

Dabney illustrates the problem of synonyms by giving the example of a

search for a case concerning a ten-year-old boy. The court might refer
to the boy as "boy," "minor," "child,".
64
"juvenile," "youth," "ten-year-
old," "infant," or "young man."
Ambiguous words create the converse problem. The searcher may
use an apparently specific word that has few or no synonyms and that
should isolate the relevant cases, only to find that the word has an en-
tirely different meaning. Dabney's example is a researcher looking for
cases involving the drug DES (diethylstilbestrol), and retrieving Tinker v.
Des Moines Independent Community School District.65 This problem is aug-
mented by the ability to search for word roots. For example, in a search
for cases involving the adoption of a child, the searcher might attempt to
retrieve cases that use the noun "adoption," the verbs "adopted" and
"adopts," and the adjective "adopted" by using the following search:
father & child w/15 adopt!
This search would also retrieve cases involving a father and child were
the opinion "adopts" a rule of law or a particular version of a disputed
factual finding.
64. Id. at 18. Blair and Maron provide an amazing example of the problem of
synonyms.
Sometimes we followed a trail of linguistic creativity through the database. In searching for
documents discussing "trap correction" (one of the key phrases), we discovered that
relevant, unretrieved documents had discussed the same issue but referred to it as the "wire
warp." Continuing our search, we found that in still other documents trap correction was
referred to in a third and novel way: the "shunt correction system." Finally, we discovered
the inventor of this system was a man named "Coxwell" which directed us to some docu-
ments he had authored, only he referred to the system as the "Roman circle method." Us-
ing the Roman circle method in a query directed us to still more relevant documents, but this
was not the end either. Further searching revealed that the system had been tested in
another city, and all documents germane to those tests referred to the system as the "air
truck." At this point the search ended, having consumed over an entire 40-hour week of
on-line searching, but there is no reason to believe that we had reached the end of the trail;
we simply ran out of time.
Blair & Maron, supra note 48, at 295. Of course, this example comes from a heterogene-
ous litigation support file, not from a set of judicial opinions. Still, anyone with a little
imagination can think of a similar trail through the cases.
65. 393 U.S. 503 (1969). This example is somewhat deceptive. A court that used the
abbreviation DES is likely (though not certain) to have used the full term at some point.
For example, I searched for the term "des" in the LEXIS States/Omni database on
March 2, 1986, and retrieved 6888 cases. I browsed through the first thirty cases and
found that nearly all involved either the city Des Moines or an Alaska criminal case, Des
Jardins v. State, 551 P.2d 181 (Alaska 1976). I then searched for the term "diethylstil-
bestrol" and retrieved 62 cases and 33 ALR annotations, a much more manageable
search. Of course, the second search may have been incomplete.
My "des" search also brought home the thorny problem of typographical errors in the
databases. Three of the first 30 cases were "hits" because of typographical errors, in-
cluding two misspellings of "does," and one misspelling of "describe." This rate may
not be representative, but it is nevertheless disconcerting.
48 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

But the problem is larger than the mere "imprecision" of language

- for example, whether a child will be called an "infant" or a "minor."
The fact is that law involves ideas, and ideas are not directly correlated
with particular words. 6 6 Dabney describes this 67
as the problem of com-
plex expressions, his third analytical category.
The difficulty of matching words with ideas is in some ways more
insurmountable than the problems of matching words with persons or
things. On the one hand, research problems that involve specific factual
questions or specific statutes or administrative rules are quite amenable
to straightforward computerized searches. Also, the skillful researcher
can develop strategies for searching with words that have specific deno-
tations but synonyms or multiple meanings. But for searches involving
legal concepts - or any ideas that can be expressed without using a
particular word or phrase - the computers are not very effective. Con-
ceptual questions are difficult to frame in the Boolean search strategy be-
cause judges are not likely to use exactly the same words to describe the
68
same ideas or concepts.
66. See Childress, supra note 8, at 1533:
Time cannot correct the inherent limitations of the word-search method, however, and con-
cordance logic may produce its own inefficiencies. LEXIS' dependence on words, for exam-
ple, grounds search capabilities in the opinion's language rather than its content. An unusu-
al or incomplete description of the facts or issue may "lose" a very relevant case from a rea-
sonable search.
67. Dabney, supra note 24, at 19.
68. West Publishing attempts to solve this problem with its "Full-Text Plus" system.
Full-Text Plus refers to the fact that the WESTLAW database contains the full text of
cases plus the same text of headnotes and Digest summaries printed in the National Re-
porter System. West claims that this addition introduces "normalized" language be-
cause the trained editor has again entered the picture. The uniform language in the
headnote and syllabus are supposed to compensate for the imprecision of the judicial
author. Thus, the searcher can formulate a search strategy knowing that his search
phrase will be matched up both with the text of the judicial opinion and with the "nor-
malized" language introduced by West editors in the headnotes and case synopsis.
A recent study by Professor Al Coco lends some credence to this claim. Coco, Full-
Text vs. Full-Text Plus Editorial Additions: Comparative Retrieval Effectiveness of the Lexis
and Westlaw Systems, LEGAL REFER. SERV. Q., Summer 1984, at 27. The study indicates a
substantial difference in retrieval produced by running the same search on both LEXIS
and WESTLAW, with the latter consistently retrieving more cases.
Dabney has questioned this result, noting that no relevancy verification of the cases
was made. He has also questioned the basic theory that Full-Text Plus' addition of
headnote and synopsis language is a major amelioration of the problem. Dabney's point
is threefold: (1) headnote language invariably tracks the text of the case, thus adding lit-
tle in the way of "normalized" language; (2) while subject headings accompany the
headnote in the database, only two levels (the highest and lowest) of West's deeply-
layered subject structure are included, and therefore, most of the relevant headings are
dropped; and (3) because the synopsis paragraph is so general, it is of marginal assis-
tance to the searcher. Dabney, supra note 24, at 31-34. Also, my impression is that
WESTLAW edits its inputted material more carefully than LEXIS, so that the additional
"hits" found by Coco may have resulted from correct spelling as well as from Full-Text
Plus. See supra note 65.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 49

Dabney summarizes this point with an excellent example. He pos-

tulates a search for the question:
"If a person waives his or her right to trial by jury in one trial, can a
jury trial still be demanded in a subsequent new trial of the same
matter?" The key words for this question, "trial," "jury," "waiver,"
and "retrial" are common in judicial opinions, but discussions of the
specific point of law of the question are relatively rare. A computer
cannot reliably find cases that are on point because too much of the
meaning of the desired cases is tied up in the syntactical relation-
ships between the words, which are not "understood" by the com-
69
puter.
In other words, unless a particular legal concept can be reliably mapped
to a relatively unique word or set of words, the concept will be invisible
70
to the researcher on a free-text system.
The Blair and Maron study indicates that the problems outlined by
Dabney remain. Indeed, as the size of the databases expands, so does
the magnitude of the problem. The body of case law increases dramati-
cally each year. The West Publishing Company calculates that it adds
65,000 full opinions to the corpus annually. 7 1 These numbers give an
idea of the truly monstrous scope of the legal databases. And this be-
comes a problem in itself.

c. Error Rates and the Staggering Size of the Databases

The sheer size of the databases is a primary source of inefficiency.
Retrieving 10% of a database of 100 documents presents few problems.
The ten documents that contain search terms used in the research query
will yield a manageable file that can be scanned easily to assess
relevance. If the file contained 40,000 documents, however, a 10% re-
trieval rate would produce 4,000 documents. A researcher cannot
thoroughly evaluate such a large number of documents. In fact, 30 do-
cuments may be too many. The researcher needs to construct high Pre-
cision search strategies that recover mostly relevant documents; unfor-

69. Dabney, supra note 24, at 19-20.

70. An excellent example of this difficulty is LEXIS and WESTLAW's failure to mark-
et effectively on-line databases of state statutes. Because the content of statutory materi-
als is highly conceptual and uses language that is either repetitive or sui generis, it is re-
latively inefficient to research with free-text searching. Also, these materials are costly
to load on-line. Although WESTLAW is experimentally loading Illinois statutory materi-
al, neither LEXIS nor WESTLAW plan in the future to market state statutes, and, in part,
their decision originates from these problems.
71. West supplied this figure as a part of a packet of information distributed during a
Summer 1985 tour. The estimate was confirmed by Bill Lindberg, a West administrator,
in a telephone conversation.
50 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

tunately, this strategy is bound to exclude many relevant documents as

well.
Dabney analyzes this problem in detail. He explains the inevitable
dilemma in which the researcher is caught. As the searcher expands the
search to retrieve all relevant cases, he or she pulls in irrelevant ones as
well. In order to screen out irrelevant materials, the searcher will add
more detail to the search request. This strategy does reduce the number
of cases retrieved, but it also contributes to the exclusion of relevant ma-
terials. 72 This is the problem of the inverse relation between Recall and
Precision described by Blair and Maron. Their study demonstrates that
when the researcher conjoins additional search terms to reduce the size
of the search output, more and more relevant documents are excluded.
As the LEXIS and WESTLAW databases continue to expand in size these
difficulties will only be exacerbated.

2. The Limits of the User

The second basic problem with computer-based free-text searching
is the limitations of the individuals who use the computer. Given the
limitations of free-text searching, who should be expected to search an
on-line, full-text database effectively and to evaluate the quality of his or
her search? In the language of information science, who is the proper
end-user?
Well-trained and experienced computer searching experts are more
effective full-text computer searchers than subject-matter experts. 7 3 But
in the legal profession, most LEXIS and WESTLAW searching is con-
ducted by lawyers. Both legal database sellers push the model of a law
office with a database terminal on each lawyer's desk. Are lawyers the
proper end-users of the full-text databases?

a. Training Incompetents ... or Worse

The first issue is the adequacy of lawyers' training. Although both
database vendors make their own training systems available to law firms
who subscribe, most attorneys are first exposed to and receive their basic
training on the systems during law school. Every accredited law school
in the United States now has either a LEXIS or WESTLAW terminal, and
an increasing number have both. The task of training students in the
use of these on-line systems has become their responsibility.
72. Dabney, supra note 24, at 21-26.
73. Cf. Curry, The Value of the Search Request Form in the Negotiation Process Between
Requester and Librarian, 20 AM. Soc'Y INFO. SCI. PROC. 115 (1983); Obermeier, Expert
Systems - Enhancement of Productivity?, 20 AM. Soc'Y INFO. SC. PROC. 9 (1983).
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 51

Unfortunately, most law schools have spotty records for any kind of
research training. The discussion in the literature on the failure of
manual research training programs is vast. 74 Most law schools have
made little headway in solving the age old problem of how to train their
students in traditional research methods. With this unstable foundation
it is unlikely that law schools will successfully handle their new respon-
sibility for training efficient on-line researchers.
In many law schools computer training is the responsibility of the
law library staff. Very rarely does the staff at these law libraries receive
enough money to develop a truly successful training program. Other
law schools hire students who are already familiar with computers
and/or LEXIS and WESTLAW to train students. At a conference held
during the summer of 1985, a group of people who were brought to-
gether because of their expertise in providing LEXIS and WESTLAW
training admitted that their own programs did not adequately train po-
tential users. This group concluded that the most that could be asked of
a law school training program is that it acquaint the computer user with
the capacities of the system. 75 Due to the skewed ratio of trainers to stu-
dents and number of machines to students currently involved in LEXIS
and WESTLAW training, it is not possible to train each student to be an
76
effective and efficient searcher.
74. A personal favorite is Brock, The Legal Research Problem, 24 DE PAUL L. REV. 827
(1975). See also Mills, Legal Research Instruction in Law Schools, the State of the Art, or,
Why Law School Graduates Do Not Know How to Find the Law, 70 L. LIBR. J.343 (1977);
Achtenberg, Legal Writing and Research: The Neglected Orphan of the First Year, 29 U.
MIAMI L. REV. 218 (1975).
75. These observations are based on a discussion at a West Publishing Conference
held in August 1985, at St. Paul, Minnesota. Both West and Mead Data have instituted
regional workshops for law librarians to discuss the problems training law students in
the use of their systems and possible solutions.
76. Both LEXIS and WESTLAW are now concentrating on assisting law schools in in-
troducing their students to the on-line systems. During the summer of 1985, both sys-
tems sponsored special workshops to talk to legal educators about such training pro-
grams. Moreover, in order to train students one-on-one, both systems have made avail-
able to large law schools a number of terminals on a temporary basis. LEXIS and
WESTLAW experimented with these Temporary Learning Centers during the 1985-1986
school year. In some locations LEXIS and WESTLAW are setting up Permanent Learn-
ing Centers (PLCs) in law schools. PLCs allow LEXIS and WESTLAW to train law firm
subscribers at a local law school library. When the database vendors are not using these
terminals for their own professional training programs, the law schools are free to use
them.
Recently, both LEXIS and WESTLAW have developed another training program.
During the summer of 1985, each announced that it would allow law school subscribers
to use the schools' terminal subscriptions as a free route of entry for up to three person-
al computer users of the same system. This will permit the Deluxe terminal and three
personal computers to be in use at the same time as a part of the same subscription.
The only limitation is that the usage be confined to off-peak hours. But despite these
52 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

Even if students originally were trained in the efficient use of LEXIS

and WESTLAW, including requisite skepticism about the usefulness of
free-text searching, the frequent changes in the databases and the con-
stant stream of enhancements call for continuous retraining. Few
lawyers can commit the time or energy to maintain their skills. This
means that, at best, law schools are graduating students who think that
they have been trained in the use of LEXIS and WESTLAW, but who
approach the systems with little or dated sophistication. The ease of use
of the legal databases may actually compound these problems by giving
lawyers a false sense of competency.

b. User-Friendly or User-Seductive?: The Moron Cadillac

When the Mead Data Central Company first marketed LEXIS, it
bundled access to LEXIS with a dedicated terminal. 7 7 The LEXIS
"Deluxe" dedicated terminal was large and ugly, but its operation was a
model of simplicity. The distinctive labeling of the keys allowed even
the most unsophisticated user to quickly master the mechanical aspects
of terminal operation and interaction with the search software. For ex-
ample, if a lawyer wanted to see the next case, she simply pushed the
button labeled "next case."
For years I have described the LEXIS Deluxe terminal as a "moron
Cadillac," designed so that it could easily be operated by even the most
machine-resistant lawyer. The premise of the design (a correct premise I
might add) was that the average practicing lawyer would not read ac-
companying "How to Use" manuals, nor would he or she attend train-
ing sessions. Lawyers intimidated by computer jargon and worried
about interacting with computer trainers found that they could almost
"train" themselves. After a false start, West eventually introduced a
78
similar terminal with dedicated keys.
efforts by the vendors, the burden of training still lies with law schools, and it is not
clear that they can shoulder it effectively.
77. A dedicated terminal is one that is designed to be used in a particular application;
generally, it is not compatible with other systems.
78. The original terminal marketed by the West Publishing Company as part of its
WESTLAW on-line system was a more or less standard dumb terminal, not nearly as
user-friendly as the LEXIS Deluxe. It required a higher degree of computer sophistica-
tion because it required that the operator learn and understand command codes. On the
other hand, because the terminal was not dedicated, it could be used to access other da-
tabases and computer systems. West thought that this innovation would be a significant
economic advantage to the user. The West terminal also cost much less than the moron
Cadillac. Because of the premium placed on price and flexibility in most parts of the in-
formation industry, West's marketing decision appeared sound. The LEXIS Deluxe ter-
minal was expensive and could not be adapted for other uses. But the legal community
as consumer made its own judgment, and the LEXIS Deluxe terminal was a great suc-
cess.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 53

The simplicity of the user friendly terminals is more problematic

than it might first appear. By encouraging the lawyer to believe that he
has the requisite sophistication to use the system, these terminals may
delude the researcher into overestimating his or her abilities to search
effectively. The ability to operate the terminal and to sort through li-
braries and files does not guarantee adequate searching skills. The sim-
plicity of the terminal's operation permits a lawyer to attend a training
session and then to allow his skills to atrophy because it is months be-
fore he plops down in front of the terminal again. At that point he will
be able to puzzle out the mechanics of operation, but that is no guaran-
tee of effective searching. This is one significant source of inefficient and
expensive searches.
As both LEXIS and WESTLAW make their systems available for use
with personal computers, the problem of inefficient searching both
deepens and widens. The problem deepens because individuals who
operate their own personal computers to search the LEXIS or WESTLAW
databases may be used to other on-line databases, which generally use
traditional subject thesaurus style searching. As a result, they may be
even further deceived about the efficacy of free-text searching in the
LEXIS and WESTLAW databases. The problem widens because any
lawyer with her own personal computer and modem 79 can now dial in
to either system to use the database. Because the personal computer will
lack a "Deluxe" dedicated keyboard, using a PC would seem to force
lawyers to confront their lack of computer literacy and warn them that
their use of the systems may be inefficient. However, software develop-
ments are running apace, and soon it will be possible to buy a reason-
ably priced software package that allows the lawyer to interact with a
"shell" program that is quite simple to use. 8 0 Thus, the same problems
will occur on a wider scale.

The success of the moron Cadillac did not go unnoticed, and West eventually
changed its strategy and marketed its own user-friendly terminal "WALT." (West Pub-
lishing Company held a nationwide contest to name the new user-friendly terminal.
They sought a warm, avuncular name. See Woxland, Anthropomorphism and the WEST-
LAW Custom Terminal OR "Hi Margie, This is Tom. It's About WALT... " LEGAL REFER.
SERV. Q., Winter 1983, at 89.) Although still not as simple to operate as the LEXIS
Deluxe terminal, the WALT terminal was a step towards easing the need for mechanical
skills. Of course, some degree of compatibility with other information systems and low
price was lost in this trade.
79. A modem is a device used for communication between computers over standard
telephone lines.
80. Such software will allow the user to chose commands from "menu" screens that
briefly explain the effect or result of each command. The software will then translate
these commands so that LEXIS or WESTLAW understands them. Lacking such
software, only the intrepid can figure out how to proceed.
54 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

Because the purveyors of the on-line databases generate income by

charging per search and per unit of time that the system is in use, it is in
their interest to encourage wide-spread terminal operation. It should
come as no surprise that the marketing strategies of LEXIS and WEST-
LAW have centered on every lawyer having his own terminal. This has
only exacerbated the problem of inefficient searching by non-expert
end-users.

3. Some Theoretical Implications of the New Paradigm

The full-text on-line legal databases are a new form of legal litera-
ture. The new literature is more or less identical in content to the old
West system, but it is accessible in an entirely new way. If we concen-
trate on the notion of access to the case law, we can begin to understand
how radically the legal databases break with the literature of the past.
The Digest was the internal, mediating structure within the old
mode of discourse. The West editors were, in effect, the Platonic Guar-
dians 81 of legal language and legal meanings. The discourse, in turn,
was the ground of integration and coherence in substantive law. The
very notion that it was appropriate to place cases arising in state jurisdic-
tions into a national index and national categories betrayed an underly-
ing jurisprudence, a non-positivist view of the nature of law.
The location of issues and cases in the old paradigm was part of
their meaning. Because the cases were only accessible through the Dig-
est, they were always presented to the practitioner as situated. The si-
tuation was a substantive context, a setting that told the searcher the
meaning of the case as much as did the opinion itself.
Free-text searching in legal databases, however, deprives the
researcher of context. The materials are presented in a mechanical and
(given the deficiencies of searching outlined above) an almost arbitrary
fashion. Found cases that are relevant are like prizes in a computer
game, rather than instantiations of the legally and socially appropriate
categories of the West Digest. For example, in the legal databases the
notion of making or deciding law by analogy is no longer a part of the
primary source material itself, but must be added onto the raw data by
the practitioner. Analogy has been a primary mode of legal discourse,
and a primary instrumental technique for those advocating changes in
the law. 8 2 The Digest categories were themselves suggestive of analo-
gues, but the simultaneous occurrence of search terms is not.
81. See L. HAND, THE BILL OF RIGHTS 73 (1958).
82. See Childress, supra note 8, at 1534 (arguing that on-line databases will
discourage reasoning by analogy, and focus litigation practice on arguing with only "on
point" cases, thus stifling growth and development in the law).
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 55

One way of thinking about the structural differences of the old and
new paradigms is to think about legal research as a sort of economy.
The goods exchanged in the legal research marketplace are the contents
of the cases. I don't mean to suggest the literal research marketplace
where information is available to everyone in proportion to the amount
of money they have to pay for talented researchers, to pay for experts to
weed out irrelevant retrieved information, and to pay for the necessary
computer time, although that's certainly an important issue. 8 3 What I
mean, rather, is an economy based on the "exchange" of information
from the corpus of law (the sellers) to practitioners (the buyers).
The West Digest System was like a centrally planned economy.
The practitioner could not obtain information directly from the cases, but
was forced to go through the regulating mechanism of the Digest. This
system was "efficient" because there were no alternatives; the buyer
(practitioner) could not find the seller (sources of information) in the ab-
sence of the Digest. Also, the system was relatively leveling and egali-
tarian; it held fewer rewards for pure searching skill than does free-text
searching. Reasonably competent searchers were able to find most
relevant information, and only somewhat less relevant information than
a very good searcher.
The West and LEXIS computer systems substitute a kind of market-
place for the planned economy of the Digest. The practitioner can ob-
tain information directly from the cases by means of Boolean search
techniques without reference to a central authority. 84 The overall
efficiency of the system is questionable, although there are certain kinds
of exchanges that it facilitates far better than the old system (e.g., finding
all cases referring to a certain statute), and there are other kinds of ex-
changes that were impossible in the old system but now are quite simple
(e.g., finding cases by judicial author).
There are several implications of this new form of exchange. First,
the researcher with more skill can obtain a lot more information than
83. See Childress, supra note 8, at 1532 (pointing out that the financial costs of com-
puter research make it available only to wealthy participants in the legal system, and ar-
guing that this research advantage, e.g., more up-to-date Shepardizing, can raise ethical
problems).
84. The full-text databases are subversive of authority in a much more direct and less
metaphorical sense. The California Supreme Court "depublishes" opinions of lower
courts that it disapproves without actually overruling the case or vacating the judgment.
See supra note 27. Depublished opinions are not included in California's official report-
ers, and West does not insert headnotes from depublished cases in the American Digest
System. They may not be cited and have no precedential value - they exist only in a
kind of legal limbo. But Mead Data does not remove these opinions from their data-
base. And, for some reason, enough lawyers have clamored for access to these cases
that West has put them on WESTLAW!
56 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

those less skillful. Thus, the new system differentiates researchers based
on merit; it rewards skill far more than did the old system. Second, the
new system encourages "legal realist" practice, because it enables the
practitioner to acquire and analyze cases by judge, by opposing counsel
and by opposing party. Third, the new system can result in pluralistic
legal discourse. The old system almost guaranteed that opposing coun-
sel would be using the same source materials. The inefficiency of the
new system in finding the relevant cases available through the West Dig-
est 85 and its ability to pull in arguably relevant cases from anywhere in
the corpus could result in different source materials for opposing coun-
sel. 8 6 Counsel may end up talking past instead of arguing against each
other, and a judge may be forced to choose the cases she prefers rather
than the arguments she prefers.
Of course, interpretation of the meaning of cases, assessment of
relevance, and analogy from rules and facts should result primarily from
the professional abilities of the practitioner rather than the structure of
the legal sources. This is true whether the raw materials of research (the
cases) are obtained from the Digest or from a database. Also, the pru-
dent practitioner will use both research sources, so that many if not most
important cases will be discovered in the context provided by the old
paradigm. 8 7 Still, the new paradigm is bound to influence the practice of
law.

III. CONCLUSION
Does all this mean that computers should have no place in the
research process? The answer quite clearly is no. To use on-line search-
ing efficiently in the short term, lawyers have to develop strategies for
dealing with its limitations. In the long term, the old and new para-
digms will merge in the technologies of the future.

85. See supra text accompanying notes 48-72.

86. But see Childress, supra note 8. Childress suggests that use of the computer data-
bases will tend to narrow the focus of practitioners to the "on point" cases rather than
expand it in unpredictable ways, as I have suggested. Id. at 1534. This seems to be
based on his belief that computer searchers will only find cases with matched facts
and/or matched holdings. My guess is that, due to the shortcomings of free-text search-
ing outlined above, even "on point" cases would sometimes not be retrieved. On the
other hand, much that would be retrieved would appear relevant to the practitioner be-
cause it was retrieved, rather than because it was relevant. The result would be pluralis-
tic discourse.
87. See supra note 47.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH

A. The Short Term: the Use of Computers as an Adjunct to

Traditional Research
The strengths and shortcomings of the old and the new forms of le-
gal literature are complementary. Prudent lawyers will continue to use
both manual hard copy research and on-line free-text searching side-
by-side. As I have suggested, the special difficulties and limitations of
computer research mean the average lawyer is not the optimal end-user
of the system.
Because of law school training programs, law firms increasingly will
be composed of lawyers who know that LEXIS and WESTLAW are
powerful tools. But lawyers will not maintain and expand the search
skills necessary to use the databases efficiently. Attorneys will do com-
puter research as they have done hard copy research - senior members
of firms will refer problems to the newest lawyers on the staff who, be-
cause of their recent graduation from law school, will be more familiar
with the computer systems. In my view, however, this practice is an
inadequate response to the need for special skills and constant updating.
Law firms have to recognize that the average attorney has no time
to maintain his or her on-line research skills. The attorney who is a true
computer "jock" is an exception. As the marketing struggle between
LEXIS and WESTLAW continues, they will offer more databases with
ever-expanding search capabilities. As the systems become more com-
plex, lawyers will need to continually develop more sophisticated com-
puter skills. Law firms will reach a point where they must decide to
create a new professional position: an expert in computer research.
Attorneys must recognize the need for an expert who straddles the
law librarian's function and associate/researcher's function. This new
position may require individuals trained both in law or legal research
and in librarianship and computer technology. These intermediaries
must be able to fully understand a lawyer as he or she describes a prob-
lem, and then be able to employ their in-depth understanding of the da-
tabases to formulate effective searches and to retrieve relevant informa-
tion. Some large law firms are beginning to recognize that computer-
literate law librarians are well-suited to fill the role of the
LEXIS/WESTLAW expert.
The Berkeley law library can serve as a model for this kind of sys-
tem. One member of our reference staff, who is both a lawyer and a li-
brarian, is responsible for maintaining current knowledge of the data-
bases. Even those of us who think of ourselves as computer literate, and
who use the databases with some frequency, cannot hope to keep up
with day to day changes and evolutions in the systems. Instead, we rely
on this individual to keep us posted on the steady stream of updates.
Law firms will have to recognize the need for this type of specialist and
58 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

to compensate them accordingly. Until then, the enormous potential of

the on-line systems will be diluted or lost, and standard of care ques-
tions centering on ineffective and inept use of the systems may arise in
the near future.

B. The Long Term: Replacing Traditional Research with

Enhanced Computerized Research
The perceptive reader will have noticed that all of my criticisms of
computer searching hinge on the use of free-text searching. This is a
significant limitation. Computers can accomplish many traditional
research tasks more efficiently than hard copy products. For example,
the on-line version of Shepard's Citations is always much more up to
date than the shelf versions, it contains all the relevant citations in one
place (rather than across several volumes), and the on-line version al-
lows the researcher to jump quickly back and forth between the citator
and the cases. 88 Since all of the references in Shepard's need to be
checked (although most are irrelevant), the on-line researcher does not
need to spend most of her time running around the library and physi-
cally locating the cases. The on-line citations systems being imple-
mented by Mead Data (Auto-Cite) and West (Insta-Cite) are powerful
new tools in the research world as well.
One quick and dirty way of improving the computer systems would
be to load the entire West Digest System on-line. 89 I have already ac-
knowledged the importance of the ability of the on-line systems to
search for unique words or terms with free-text searching. But for
efficient research in the realm of legal concepts, the intervention of a
highly trained and dependable human indexer may be irreplaceable.
The Digest could be used both as a subject for free-text searching and
for browsing like a traditional subject thesaurus. All the cases could be
cross-referenced with the index so that the researcher scanning the all-
inclusive Digest (no more searching from decennial to decennial) could
jump immediately to the indexed cases and back, much like the current
Shepard's citator. The experience, vocabulary and expertise of the West
editors would be an extremely useful addition to on-line research.
Over time, computers may begin to replace hard copy as the
medium for traditional index-based research through the convergence of
emerging technologies. For example, consider the recent advent of CD-

88. See Dabney, supra note 24, at 38-39.

89. The West "Full-Text Plus" system is supposed to provide many of these benefits,
but it provides them only in the context of free-text searching, and there are other
significant limitations. See supra note 68.
1986 FULL-TEXT DATABASES AND LEGAL RESEARCH 59

ROMs. 9 0 CD-ROM technology allows small computers to use the laser

disks used by home stereo compact disk players as optical storage dev- 91
ices. A single CD-ROM will hold 600 megabytes of information
(about 300,000 typescript pages), the equivalent of about 150 volumes of
United States Reports. The last fifty years of the entire West National Re-
porter System would fit on fifty or sixty CDs, which could be easily
stored in a small desk drawer. Never before has so much information
been available in such a small and physically stable format. Also,
manufacturing costs of CD-ROMs are low; each disk costs less than ten
dollars to produce in quantity. 9 2 Although many users don't realize it,
LegalTrac (a computer-based legal periodical index marketed by Infor-
mation Access Company) is a CD-ROM device that has already arrived
in many law libraries. In the law office of the future, the firm computer
will provide reporters and digests on-line, and lawyers will be able to
use both powerful indexes and free-text searching to access the data-
93
base.
90. Compact Disk - Read Only Memory. A CD-ROM is a small (4.72-inch-diameter)
plastic-coated metal disk with binary information etched onto the metal. The informa-
tion is read by a "player" disk drive that bounces a small laser beam off the surface of
the disk and reads the modulations. Information can be read from the disk but not writ-
ten to it.
91. Crabb, CD-ROM Arrives; It's Fast but Limited, InfoWorld, Mar. 31, 1986, at 49, col.
1 (evaluating performance of Phillips CD-ROM device and the few software packages
currently available).
92. Compare these manufacturing costs to the retail prices of the hard copy products,
supra note 43, and you will see that replacing hard-copy products with CD-ROMs could
mean both enormous profit margins for West and significantly lower prices to legal con-
sumers. The CD-ROM disk drives are also quite inexpensive. Single disk drives now
cost as little as $845, and should soon be available for under $500. Welch, Manufactur-
ers to Propose CD-ROM File Standard, InfoWorld, Feb. 3, 1986, at 1, col. 1. For more on
CD-ROMs and the attempt to develop industry standards in order to facilitate market
expansion, see id.
93. Of course the paperless office is a myth. People will always want to interact with
words on paper. But a computer-based library is not incompatible with the desire for
printed products. First of all, the quality of computer visuals is skyrocketing. Personal
computers and terminals of the near future will have very large and very high-
resolution screens. Legal materials will appear on-screen in a life-size black on white
replication of actual typeset pages. The user will be able to instantaneously "thumb"
through pages as easily as through a book. This will make scanning search results far
faster, easier, and more natural than current dumb terminal technology. Second, law
offices will see the emergence of a technology called "on-demand publishing." High-
speed laser printers in the library will typeset cases and other materials onto book-
quality paper in a matter of seconds. The paper output can be bound or inserted into
looseleaf binders.
Thus, the entire contents of a legal library will be on-line, with most of the collection
on a local computer (e.g., on CD-ROMs updated every month or two), and very recent
materials available from remote sources such as LEXIS and WESTLAW. Portions of the
collection that are used frequently will be "published" in hard copy and bound by the
library staff. The printed collection can be instantly updated in order to make informa-
tion access easier and to give legal workers the printed page that they will long desire.
60 HIGH TECHNOLOGY LAW JOURNAL Vol. 1:27

For now, LEXIS and WESTLAW will remain part of the armory of
legal research - useful, practical, and particularly helpful for questions
that call for the location of a specific word or name. And we should not
underestimate the role of computers and free-text searching in the law
library of the future. But unless and until subject thesauri implemented
by professional indexers are added to the databases, these systems will
not be ultimate research tools. Until that time, lawyers must be careful
that the shortcomings of the on-line legal literature do not distort or di-
minish the quality of their practice.