By Marcia J. Bates
School of Librarianship, University of Washington, Seattle, WA 98195

Journal of the American Society for Information Science, v. 30, July 1979, p. 205-214

Received September 12, 1978; revised February 12, 1979. © 1979 by John Wiley & Sons, Inc.

As part of the study of human information search strategy, the concept of the search tactic, or move made to further a search, is introduced. Twenty-nine tactics are named, defined, and discussed in four categories: monitoring, file structure, search formulation, and term. Implications of the search tactics for research in search strategy are considered. The search tactics are intended to be practically useful in information searching. This approach to searching is designed to be general, yet nontrivial; it is applicable to both bibliographic and reference searches and in both manual and on-line systems.

Introduction
For all the developments in automated and semiautomated information retrieval, nothing yet matches the ability of experienced human searchers—whether known as “information specialists” or “reference librarians”—who move skillfully among an enormous range of resources, both manual and on-line, to develop bibliographies or answer questions. We know discouragingly little about just what those skills are and how they develop; we cannot yet define what it is that an experienced searcher knows that a beginner does not.

In this article, the concept of the information search tactic is introduced and various particular tactics are named and described. These tactics are designed to be of assistance to the searcher while in the process of searching, and secondarily in teaching fledgling searchers their work. They may also suggest various lines of research on search strategy, including, ultimately, efforts to describe and distinguish the skills of the experienced searcher.

The plan of this article is first to give a brief literature review of human information search strategy; second, to discuss the concept of the tactic; third, describe the 29 tactics to be used in information searching; and finally, show some implications of tactics for research in search strategy.

The orientation of this article is toward professional, as opposed to lay, information searching. The tactics, intended to be practically useful in information searching, are applicable to both bibliographic and references searches* in both manual and on-line systems, and to all types of questions and subject fields.

Much of the searching done by the professional information specialist requires very little in the way of strategy. Simpler, so-called “ready reference” questions are usually answered by mentally retrieving in a second or two the name of a suitable source, going to the source, and getting the answer. In such cases, the tactics proposed below are seldom needed. The tactics are primarily designed to help in more complex searches—those involving many stages, or those that resist the automatic mental retrieval of satisfactory answering sources.

A few of the tactics described are well known and discussed—though often unnamed—in the library/information science literature; others may be used but are not consciously articulated, let alone named; and still others may be unknown in every respect to most searchers, instructors, or researchers in the field. Even with those tactics that a searcher already uses and recognizes, there may be benefit to be gained by reading about them nonetheless. It is a basic premise of this article that to articulate and name a tactic makes it more available in the searcher's mind; it makes it more easily and readily applied.

A major orientation of the article is to focus on and use the strengths and flexibility of human thinking processes. It is suggested that search theory and practice may be advanced through a greater attention to the specifically human, psychological processes involved in searching, as distinct from the logical, or formal, properties of the process. In addition, we turn from a focus on the machinery, the information technology, to the brain that is running it.

Information Search Strategy Literature

There is relatively little literature dealing with information search strategy—especially by that name. We must look several places to find all that might be relevant to the subject. The phrase “search strategy” is probably found most often in the literature of computer on-line searching. Stevens [1] reviews a number of these articles (also see issues of Online Review, founded in 1977). Typically, in these articles search strategy is viewed as specific to on-line searching. The on-line search is not seen as a part of a broader search process. To put it another way, the unifying factor is the computer on-line system, not the human being running the terminal.

Some literature deals with whether computers can assist the searcher or be used for any of the steps normally performed by humans [see, e.g., refs. 2-5]. Interesting insights about human searching are sometimes provided by these human-machine comparisons, and receive later mention in this article where relevant.

While the phrase “search strategy” is less commonly used in traditional librarianship, the field has had a longstanding interest in the “reference process,” which sometimes includes searching procedures. Though reference work has been recognized as a library function since before the turn of the century, traditional texts in reference work said little or nothing about searching [6-8]. Hutchins [9] was exceptional in her attention to how to answer questions. As for theory of search strategy, there has been essentially none until quite recently; in fact, up to the 1970s, there was ongoing debate about whether there was theory of any kind in ref. 10. Since Wynar’s article noting the lack of theory in the field, a number of articles have appeared positing reference theories [11-15], but these are fairly brief and general and say little about searching.

However, some articles have appeared that deal more specifically with information searching. Neill [16] applies the psychologist Guilford's "Structure of Intellect" model to the reference process, and Benson and Maloney [17], in their analysis of the search process, focus on building a “bibliographic bridge” between system and query.

A number of attempts have been made to analyze the reference process or, more specifically, the search process, in flow chart form. Katz [18], in a modern reference text that deals much more extensively with searching than earlier texts, reviews no less than eight such models. (Incidentally, Katz's chapters on searching in the recent editions of his text [18, 19] constitute excellent reviews of this literature.) These models vary in specificity. Stych's model [18, pp. 134-135], for example, goes into great detail, down to the type of reference book and even specific sources, to try for answers at various decision points. On the other hand, the Rees and Saracevic model [18, p. 137] gives “selection of search strategy” as a single step in a ten-step model of the reference process.

Nothing was found in this literature on search tactics, or dealing with search strategy from the standpoint of tactics. Josel's article, “Ten Reference Commandments” [20], with hints for reference work, comes closest.

Finally, one other body of literature may be expected to be of use: that on the psychology of problem solving. Nothing was found in that literature directly related to information searching, but Adams [21, p. 66] lists a number of verbs as general problem-solving “strategies,” an approach which resembles that used here for tactics. Where appropriate, other references are made below to the problem-solving literature.

The Concept of the Search Tactic

In order to elucidate the role of the tactic in human information searching, it is first useful to distinguish various types of models that may be developed in studying this area. At least four different sorts of models of search strategy can be distinguished, models for idealizing searching, representing searching, teaching searching, and facilitating searching.** An ideal model specifies ideal search patterns on the basis of mathematical, system analytic, or other formal criteria of optimality. Representation models exist for the scientific purpose of describing, and ultimately predicting and explaining, the human behavior known as information searching; these models represent what people actually do or think in searching. A good teaching model is one that makes it easy for people to learn to search. A model for facilitating searching is one that searchers can use while in the process of searching, one that helps them search more efficiently or effectively.

These different functions impose different requirements for what constitutes good models in each case. Conceivably, a very good single model could be found that would be optimal for all these purposes in one. But it is more likely that different models will have to be found for each function. For example, system analytic or mathematical theories of optimal searches, while good descriptions of ideal searching, may be predicted to be of little interest for the practice of searching to the typical information specialist, whose mathematical background is limited.

Good models that join the teaching and practice functions are more likely to be found than ones that join other pairs of these functions. But even here differences can be found. The best way to learn something when first encountering it may be different from the best way to think about that same thing once familiar with the ideas.

The model presented here is intended primarily as a facilitation model and secondarily as a teaching model. It is assumed at this point that the model will function well in both capacities; later testing will determine whether it is, in fact, useful in neither, one, or both of these capacities. Students of search behavior who are attempting to develop representation models may find this an interesting starting place as well. The concept of the tactic, as well as particular tactics, may be found to be useful and appropriate for describing the behavior of skilled searchers. (Different results may be found for those who have and for those who have not been exposed to the concepts in this article.)

“Strategy” and “tactics” are terms best known for their military uses. Webster's New Collegiate Dictionary [22] defines them as follows:
Strategy: The science and art of employing the armed strength of a belligerent to secure the objects of a war, esp. the large-scale planning and directing of operations in adjustment to combat area, possible enemy action, political alignments, etc.; also, an instance of it.

Tactics: 1. (usually construed as sing.) The science and art of disposing and maneuvering troops or ships in action or in the presence of the enemy. 2. (usually construed as pl.) Hence, any method of procedure; esp., adroit devices for accomplishing an end.

Strategy deals with overall planning; tactics deals with short-term goals and maneuvers.

Let us now adapt these terms for use in information searching. The definitions below are general and simple, and may be refined in later theoretical development.

Search tactic: A move made to further a search.

Search strategy (in searching); A plan for the whole search.

Search strategy (as an area of study): The study of the theory, principles, and practice of making and using search strategies and tactics.

Search behavior: What people do and/or, as far as can be determined, what they think when they search.

Every move a person makes toward the goal of finding desired information is seen as a tactic. Hence, there can be good, or effective, tactics and bad ones. The purpose of this model is to suggest tactics that are thought likely to improve the effectiveness or efficiency of a search. The word “likely” is important, however. These tactics are heuristic; they may help, but not necessarily. Furthermore, a tactic may be good in one situation and not in another. Further development of the model may make it possible to state the circumstances under which certain tactics are most likely to be useful. (See also the section “Implications of Search Tactics for Research in Search Strategy.”)

The tactics described here are concerned with the search proper: threading one's way through the file structure of the information facility to find desired sources, fitting the search as conceptualized to the vocabulary of the system/ resources, and monitoring the search, that is, keeping it on track. Another aspect of searching, getting ideas to overcome a stymied search, will be dealt within a future article, "Idea Tactics." Other areas of the overall search process which may lend themselves to the development of search tactics are the reference interview, the initial analysis of question and search design, consultation with the patron during and after the search (which may be considered a part of the reference interview, broadly conceived), and determination of the relevance of information found.

Four types of tactics are distinguished here:
(1) Monitoring tactics are tactics to keep the search on track and efficient.
(2) File structure tactics are techniques for threading one's way through the file structure of the information facility to desired file, source, or information within source.
(3) Search formulation tactics are tactics to aid in the process of designing or redesigning the search formulation. (These tactics are not restricted to computer search formulations.)
(4) Term tactics are tactics to aid in the selection and revision of specific terms within the search formulation.

A note on “file structure,” since the phrase is used here in a somewhat unconventional way—All the information in a typical information facility can be seen as organized into a structure. Typically, there are a number of different types of file, each ordered according to different principles, and these files are interrelated through common access in a central file such as a catalog. (The concept of “file” as used here includes not only those things conventionally called by that name, but also any ordered set of information individuals. The “book” is a typical such individual; thus, for example, the ordered set of books contained in the A/Z call number range of a library's main stack collection is also a file.) The closest traditional librarianship comes to a name for this structure of files is “bibliographic control,” a phrase used in so many different ways as to be almost meaningless. In information science, “file organization” is generally restricted to computer usage. So, “file structure” is used here to refer to the overall pattern, or structure, of organization of information in an information facility.

This concept is not to be confused with particular classification or indexing systems. The point here is that there is a more general way of viewing information organization than through particular systems. All indexing and classification systems provide a structure; the interest here is in the fact of the structure, not in the specific character of that structure. It is seen that the tactics, though they deal with threading one's way through the file structure, do so independently of particular systems of information organization.

With regard to the whole set of tactics, some overlapping will be noticed, as well as some hierarchical relationships. VARY, for example, will be seen to be the general case, against several specific forms of variation that follow it in the term tactics. These relationships have been retained for two reasons.

First, if we view the identification and understanding of tactics as our problem to solve as researchers, then the following point on research strategy needs to be made: While our goal over the long term may be a parsimonious few, highly effective tactics, our goal in the short term should be to uncover as many as we can, as being of potential assistance. Then we can test the tactics and select the good ones. If we go for closure too soon, i.e., seek that parsimonious few prematurely, then we may miss some valuable tactics.

Second, it may prove that a larger set of tactics, including even overlapping ones, makes a better facilitation model anyway, for psychological reasons. I suggest that this is likely the case. The use of these tactics is a form of creative problem solving. In such cases the mind may not work in logical, regular patterns. It may come at a problem from many different levels and angles. It may use one tactic on a particular type of problem one time and a different one on the same sort of problem the next time. In other words, we may find that the requirements of a good facilitation model of searching include some redundancy, and that an across-the-board application of parsimony (economy of logical formulation and expression) is more appropriate to ideal and representation models. (Even if this is so, it nonetheless does not preclude the possibility that parsimony is desirable in some respects in a facilitation model. For example, while using a large, nonparsimonious set of tactics, it may still be desirable to have definitions of the tactics that are as concise as possible.)

Search Tactics

In Table I are listed and defined all of the (presumably useful) tactics for the search proper that I have been able to discover to date. The 29 tactics are adapted from my own experience and thinking, from the literature, and from the comments of colleagues and students.

Since tactics are moves, they are presented in verb forms, one may think of them either as commands (CUT!) or as infinitives (TO CUT). Names of tactics have been selected to be brief and striking.

Discussion of Tactics

Definitions of tactics are given below in the context of a broader discussion and explanation. Where no discussion is deemed necessary, definitions are stated alone.

MONITORING TACTICS: Tactics to keep the search on track and efficient.
MI. CHECK. To review the original request and compare it to the current search topic to see that it is the same.

M2. WEIGH. To make a cost-benefit assessment, at one or more points of the search, of current or anticipated actions. Among other things, the searcher might consider whether any other approach would be more productive for the effort.

M3. PATTERN. Frequent experience with a type of question may lead to an habitual pattern of search [see also 19, p. 87ff] . If, for example, a common request in an academic library is for addresses of researchers, then the librarian may soon develop a sequence of sources to search, arranged by their likely productivity. To PATTERN is to make oneself aware of a search pattern, examine it, and redesign it if not maximally efficient or if out of date.

M4. CORRECT. To watch for and correct spelling and factual errors in one's search topic. These may exist in the topic as presented originally by the user (cf. Josel's first “reference commandment” [20, p. 146]), or may slip into the searcher's thinking in translating a verbal request, or in remembering (without having in hand) a written request. In observing bibliographic searching done by several librarians, Carlson noted that the searchers would allow inaccuracies, particularly spelling errors, to slip into their search formulation. One librarian, for example, had a request on “neuroglia,” and searched instead on “neuralgia,” a very different concept. He noted several cases where a difficult technical term was not written down and the librarians “would search for the remembered spelling, usually not find it, and then stop the search for that term” [5, p. 29].

A clue to errors in the request as stated may be provided by suspicious coincidences. Josel's fourth “reference commandment” is “Coincidence is no coincidence.” As he says: “When a patron wants to have a biography of Saint Edmund Hall, born 1226, and you find the same name listed as a college of Oxford University, and 1226 as its date of construction, do not doubt the patron needs further talking to” [20, p. 146].

M5. RECORD. To keep track of trails one has followed and of desirable trails not followed up or not completed. In complex searches it is sometimes necessary to return to the source of information or citations recorded earlier in the search. For example, after recording a number of citations from a periodical index, the searcher may then attempt to retrieve the articles cited and find a blind lead. The citation needs to be checked again in the original source. But unless the source, volume date, and subject term searched under were recorded, the searcher may have to go through the entries under a dozen terms or in several volumes to locate the desired citation. Similarly, if productive on-line and manual bibliographic search formulations are retained, later repeat effort may be saved.

As for trails not followed up or not completed, Carlson noted the following in his observations of librarians:

They are not consistent in recording what they find or what they intend to check later. In many cases, the human will find cross-references and state that they will check these cross-references later. Unless they make some written now, they never seem to check them. Each librarian studied, at some time during the search, noticed some discrepancy either in an item being scanned, or in an item recorded as acceptable, and made the verbal comment that he would check this later. Once again they almost never made a written note about this and when they did not write a note, they never did check the item. The discrepancies which arose during the search were thus not clarified. [5, pp. 29-30]

The searcher may choose not to follow up some side trail or problem, but it appears that such choices are often made by default, rather than deliberately.

FILE STRUCTURE TACTICS: Techniques for threading one's way through the file structure of the information facility to the desired file, source, or information within source.

Fl. BIBBLE. One way to cope with the file structure is to find a way to do without it altogether. The only neologism among the set of tactic names, BIBBLE is based on the abbreviation “bibl.” for “bibliography.” To BIBBLE is to look for a bibliography already prepared, before launching oneself into the effort of preparing one. More generally, to BIBBLE is to check to see if the search work one plans has already been done in a usable form by someone else.

F2. SELECT. To break complex search queries down into subproblems and work on one problem at a time. This tactic is a well-established and productive technique in general problem solving. As each subproblem is solved, the parts can then be knit into a solution to the whole, larger problem.

F3. SURVEY. To review at each decision point of the search the available options before selecting. In Carlson's description of human searching behavior, he noted the following problem: “There is almost no look-ahead in the human search procedures. All of the librarians studied exhibited to some extent this lack of look-ahead. They would often scan each entry as they came to it and then encounter a heading which would alter the search procedure.” He concludes: “Here the lesson is very clear: humans should scan over a reference document before making any detailed searches through it” [5, p. 35]. Psychologically, this is a problem of “going for closure” too soon, that is, settling on a source or approach prematurely. In employing SURVEY, one resists that temptation and presumably achieves a more effective search.

For example, in a bibliographic search, instead of selecting the first index that comes to mind, one thinks of all the major indexes in the subject and then selects the one best suited to the particular query. Then, instead of moving immediately to a subject entry term within the index, one first scans through the thesaurus to find the best term or terms for the subject.

F4. CUT. When selecting among several ways to search a given query, to CUT is to choose the option that cuts out, eliminates, the largest part of the search domain at once. In my opinion, this tactic is of fundamental significance in our field, and is relatively little known or discussed. Here are some examples: When looking up a book written by Smith and Brzustowicz, the search will be much briefer if one looks under Brzustowicz (assuming the file has entries under co-authors). In most files, there will be far fewer entries to scan under the latter name. Thus, in choosing to search under the latter name, with its few entries, one has cut out a larger part of the search domain than would be the case when searching under Smith, and has shortened the search accordingly.

Similarly, in a subject search, other things being equal, one should look up the most specific elements of the topic first. For example, in using a KWIC (rotated title term) index, the searcher will find desired material on the topic “Research in Retinopathy” much faster by looking under the more specific term “retinopathy,” because there will be fewer entries.

The concept of CUT has received the most explicit use in information science in manual coordinate indexing searching. If one pulls three subject-term cards for an ANDed search on terms A, B, and C, then makes comparisons among term cards to find the documents indexed under all three terms, the smart tactic is to start with the card with the fewest document numbers posted on it [cf. 23, p. 232]. Since all acceptable documents must have all three terms assigned to them, the card with the fewest documents posted exercises the most control and eliminates the largest part of the search domain.

F5. STRETCH. Naturally enough, one tends to think about information resources in terms of the uses for which they are intended. However, almost all reference sources can be used productively for some other purpose than intended. The internal organization of a file or reference book is designed around certain uses. Thus, access via certain record elements is provided, and access via other elements is not. But even though formal access is not provided, that other information is there in the source nonetheless. Introductions, which are outside the formal internal file organization of an information source, may also be informative in unexpected ways.

In general, it may be assumed that the most efficient searching involves using sources for their intended purposes. But when such approaches fail, answers may still be found by putting in the harder work to ferret out information incidentally provided. Thus, to STRETCH is to use a source for other than its intended purposes. However, it should be kept in mind that to STRETCH effectively the searcher must first think differently, he/she must think about all the information that is in a source, not just about the ordinary uses of it.

For example, after searching unsuccessfully through many directories for the address of an engineer, the searcher may recall that patents contain the name of the inventor and also the business affiliation, since the patent is usually owned by the company she/he works for. If the engineer has patented anything, then the address should be available in the nearest patent file.

F6. SCAFFOLD. Hodnett discusses the use of what he calls “auxiliaries” [24, p. 94ff.] which are aids in problem solving which may or may not themselves be a part of the solution, but which make the solution possible. The technique of using auxiliaries is often employed in mathematics, where a seemingly irrelevant theorem is introduced, a theorem with little intrinsic interest, but one that enables the main theorem to be proved.

The use of scaffolding in construction is another such example. When the building is finished, the scaffolding is torn down, but the building could not have been built without it. In information searching, it is sometimes the case that the shortest route through the file structure is a dead end. In that case one may build a roundabout path to the answer by going through files or sources that themselves may seem to have nothing to do with the question. One may acquire an additional piece of information that in no way contributes directly to the answer but which makes it possible to search for the answer in some other source. Thus, to SCAFFOLD is to design an auxiliary, indirect route through the information files and resources to reach the desired information. For example, after unsuccessfully seeking information on an obscure poet, the searcher may find out who the poet's contemporaries were and research them in hopes of finding mention of the poet.

F7. CLEAVE. To employ binary searching in locating an item in an ordered file. (For those unfamiliar with this principle: In binary searching one first looks at a record in the middle of an ordered, e.g., alphabetized, file. One then determines the half of the file in which the desired record must lie. Then the middle record in that half of the file is looked at, and the quarter of the file in which the record must lie is determined. Then one looks at the middle record in the quarter section of the file, and so on until the desired record is discovered. In each case, the file is split in two, hence the term “binary” [see also ref. 25].)

Formally, binary searching is a more efficient approach than serial or random searching. Yet a rigid adherence to this principle would probably be wasteful, since human beings have additional contextual knowledge about many files. For example, a searcher looking for the telephone number of the Ajax Corporation will not start the search in the middle of the white pages. On the other hand, a general awareness of binary searching may enable searchers to improve efficiency, particularly when confronting large and unfamiliar files. The use of CLEAVE, as well as means of testing its usefulness, are discussed in more detail in ref. 26.

SEARCH FORMULATION TACTICS: Tactics to aid in the process of designing or redesigning the search formulation.

SI. SPECIFY. To search on terms that are as specific as the information desired. Specificity is one of the crucial concepts in systems of information access. Almost all systems of classification and indexing require that descriptions assigned to materials be as specific as the content of the materials and as the indexing system itself allows. Sears and Library of Congress subject headings use the “rule of specific entry” which requires entry of materials under the most specific terms that still encompass the content of the item; coordinate indexing, with its focus on “concept” indexing, brings about highly specific description, and so on.

Thus, specificity at the time of indexing requires specificity at the time of retrieval. An indexing system may or may not allow entry under broader terms as well, but it will almost always require specific entry. Thus, it is probably the case that starting with specific terms in all kinds of searches (including both bibliographic and reference) will be the most productive approach [for further discussion see ref. 26] .

S2. EXHAUST. To include most or all elements of the query in the initial search formulation, or to add one or more of the query elements to an already-prepared search formulation. Both this and the next tactic, REDUCE, are related to Lancaster's use of "exhaustivity" [27, p. 71ff.]. In searching, the more exhaustive a search is, the more of the elements of a complex request have been included in the search formulation. For example, the searcher interested in the “training of teachers of mathematics for the elementary grades” has a four-element problem. An exhaustive search would include all four elements in its formulation. Both EXHAUST and REDUCE deal implicitly with the number of elements in the query that are to be ANDed together in the search formulation. The more exhaustive the search statement, the more stringent the requirements, and thus the fewer the documents likely to be returned on a search.

While this tactic is probably most useful for Boolean searching, it is also meaningful for other kinds of searches. For example, in a catalog using Library of Congress subject headings, one can decide between searching under the main heading only or more exhaustively under the main heading plus geographical, bibliographical form, or other nontopical subdivisions.

S3. REDUCE. To minimize the number of elements of the query in the initial search formulation, or to subtract one or more of the query elements from an already-prepared search formulation. REDUCE is the opposite of EXHAUST. This tactic reduces the number of ANDed elements in the search formulation, making the search specification less stringent, and thus increases the number of documents likely to be returned on a search.

S4. PARALLEL. To make the search formulation broad (or broader) by including synonyms or otherwise conceptually parallel terms. PARALLEL and PINPOINT deal implicitly with elements in a query that are to be ORed together. Though these tactics are most readily applied in on-line Boolean searching, they may also be used in manual searching. For example, in the process of manually compiling a bibliography, one may look over catalog subject headings and terms in periodical indexes and expand the number of similar terms searched under (PARALLEL), either at the beginning of the search or after getting some experience with the type and quantity of materials under each term.

S5. PINPOINT. To make the search formulation precise by minimizing (or reducing) the number of parallel terms, retaining the more perfectly descriptive terms. PINPOINT is the opposite of PARALLEL.

S6. BLOCK. To reject, in the search formulation, items containing or indexed by certain term(s), even if it means losing some document sections of relevance. This tactic deals implicitly with the Boolean AND NOT. The term NOT was not used, however, because the concept extends beyond the usual applications of Boolean searching. For example, in doing a manual literature search, one may choose to reject all items containing a certain word in the title. BLOCK was selected as the name of this tactic to draw attention to the tricky side of NOT—to the fact that in eliminating a document that contains an undesired term, one may also block out desirable material that happens to be found in the same document.

TERM TACTICS: Tactics to aid in the selection and revision of specific terms within the search formulation.

Tl. SUPER. To move upward hierarchically to a broader (superordinate) term. Searchers may be assisted by pointers in a thesaurus or may have to rely on their own knowledge to devise the term.

T2. SUB. To move downward hierarchically to a more specific (subordinate) term.

T3. RELATE. To move sideways hierarchically to a coordinate term.

T4. NEIGHBOR. To seek additional search terms by looking at neighboring terms, whether proximate alphabetically, by subject similarity, or otherwise. Coates pointed out many years ago that all manual (and we should add today, most automated) information organization systems do two fundamental things: locate and collocate [28, Chap. 3]. The primary function of such systems is, of course, to enable the searcher to find, or locate, desired materials. However, such systems also necessarily collocate entries. In any ordered file everything must be next to something else. Many of the historical arguments over the relative merits of classification and indexing systems were as much about collocation as location. Consider, for example, the old debate over whether to have classified or alphabetico-specific subject catalog access. A classified catalog collocates entries by their conceptual relationship; an alphabetico-specific catalog collocates entries only by their alphabetical order. These two approaches have different strengths and weaknesses and different consequences for search strategy [28, Chap. 3].

To use this tactic is to expand the search by examining the proximate entries, whatever they are. In on-line searching, one examines whatever proximate entries are made available by the on-line program one is using. (NEIGHBOR happens to be the current term for the appropriate command in the SDC ORBIT® search language.)

Incidentally, the use of NEIGHBOR may be extended beyond term selection to resource selection as well. Since classification systems collocate books, it is easy to extend a search by examining related sources collocated on the shelves of the reference stacks.

T5. TRACE. To examine information already found in the search in order to find additional terms to be used in furthering the search. Two of the most common ways of doing this are to scan descriptor term lists in citations retrieved in on-line searching, and to scan on a catalog card the list of other headings that have been given to the document in question. These other headings on the catalog card are called the "tracings," hence the name for this tactic (cf. Josel's seventh “reference commandment” [20, p. 147]).

T6. VARY. To alter or substitute one's search terms in any of several ways. See remaining term tactics for some specific variations.

T7. FIX. To try alternative affixes, whether prefixes, suffixes, or infixes. Several may be done at once through truncation routines.

T8. REARRANGE. In any system where terms may contain more than one word, word order may make a difference in retrieval success. To REARRANGE is to reverse or rearrange the words in search terms in any or all reasonable orders.

T9. CONTRARY. To search for the term logically opposite from that describing the desired information. For example, one may want information on “cooperation” and, after an unsuccessful search, change the term to “competition.”

T10. RESPELL. To search under a different spelling. CORRECT dealt with maintaining correct spelling, among other things. But with RESPELL the concern is not with correctness, but with effectiveness. Particularly in current on-line search systems, there are a great many spelling variations that show up in the citations. One must expand the spelling variations to insure good recall. RESPELL is occasionally needed in manual systems too, where, for example, one needs to change from U.S. to British spelling to search successfully in a source.

T11. RESPACE. Spacing, particularly in hyphenated words, or words that appear with various spacings, can be critical in search success. To RESPACE is to try spacing variants. While spacing problems are most glaring in some automated search files, such problems can also be serious with manual files. The two fundamental variants in filing rules—word-by-word filing and letter-by-letter filing—differ on how the blank space is to be treated in filing [29; 30, p. 339]. Both of these rules are in wide use. The searcher who is thinking in terms of one filing rule and enters a source that uses the other may miss the desired material.

Implications of Search Tactics for Research in Search Strategy

As experience is gained with these tactics, leads for the development of ideas and research in other areas of human information search strategy should emerge. Some possible directions that have already come to mind follow.

(1) Various tactics form clusters as responses to situations where a search produces too many or too few documents. For example, where too few documents are produced, the searcher might try SUPER, RELATE, REDUCE, PARALLEL, NEIGHBOR, TRACE, and VARY, among others. Too many documents, on the other hand, might lead to the use of SUB, EXHAUST, PINPOINT, and BLOCK.

A few steps may be taken in moving from a facilitation model of tactics to a facilitation model of strategy (i.e., to a model that suggests helpful search strategies or techniques for developing strategies) by distinguishing typical stages of searches and then looking for useful patterns in tactics use at those stages. If the searcher is aware that a small cluster of tactics is most likely to be useful at a given stage, then she/he can concentrate on just those few at that stage.

(2) It was stated earlier that the tactics presented in this article are restricted to those dealing with the search proper: monitoring the search, threading through the file structure, and fitting search query to system/resource vocabulary. With experience, other tactics in the aforementioned categories should emerge. For example, in a system as complex as an information facility, there are surely more useful file structure tactics than have been noted here. Tactics can also be developed for the other elements of the reference process that have been left out here. (As noted earlier, I will present “idea tactics,” ways of getting ideas to help with stymied searches, in a future article.) Tactics can also be developed for moves in the reference interview. The complexity of the interview process has been recognized in recent years, and a considerable literature is developing in that area [see, e.g., ref. 31]. In many cases, the reference interview, in effect, continues into the search process itself, as the searcher returns with partial material to show the requester. User feedback during the search adds another dimension of complexity to the search, and feedback-related tactics should ultimately be included in any comprehensive view of search tactics.

Another aspect of the reference process with tactical potential is the stage in which the search query is analyzed. It is at this point that search strategy is developed. It might be possible to enrich or advance upon the flow-chart models of search strategy already developed by devising specific tactics suitable for this initial analysis phase. For example, Jahoda recommends as one step in the reference process: “select sequence of specific titles to search” [32, p. 155]. Thus, “SORT,” to mean, for example, sorting the sources responsive to a query from most to least likely to be of help, might be seen as one tactic to be used in this phase. Finally, tactics to aid in the evaluation of relevance of retrieved materials would represent yet another element of the reference process.

After these various sets of tactics have been developed, an ultimate goal of creating a single, comprehensive set of tactics can be envisioned. The set would incorporate all elements of the reference process, from initial interview with patron all the way to final determination of relevance and final negotiation with patron. It would provide a unifying mode for viewing the reference process and could constitute the core of a course in reference/information searching.

WEIGH suggests a whole small subarea of investigation. WEIGH is a capsule name for an on-the-spot cost-benefit assessment. Cost-benefit analysis in library/information science generally involves extensive studies and mathematical models. WEIGH, on the other hand, deals with what people can evaluate in their minds in a few seconds. What is needed is a class of rules-of-thumb that the searcher can use while searching. Such rules would be concerned with, among other things, how to choose among several sequences of actions or among several sources. For the latter, the rules might take into account likely productivity of source, likely effort in use, and characteristics of query requirements, such as exhaustivity desired or “importance” of query.

These rules might look simple but they would be based on sophisticated testing to discover them in the first place. The point here is that while there is a well-developed science of cost-benefit analysis for systems researchers to use, there is no such science for information searchers to draw on while they are in the process. What is needed, in short, are searching decisions rules that minimize cognitive strain [cf. 33, p. 82ff.] Since staff costs are usually the largest part of an information facility's budget, anything that can be done to enable searchers to work faster should be a valuable improvement in information facility system performance.

(4) One of the fundamental issues in search strategy is when to stop. Two example stopping questions: How does one judge when enough information or citations have been gathered? How does one decide to give up an unsuccessful search? At least two of the tactics in the set proposed here suggest testable areas for stopping. SURVEY was recom¬mended as a way of making sure one has found a good source, or best term, for searching. Use of SURVEY can not only aid in effectiveness, but also in efficiency. If the searcher does not use this tactic and searches under the first term that comes to mind, time may be wasted with that term in reading and recording (or having printed out) citations, before the realization that there are better terms to use. Thus, SURVEY may be presumed to aid both efficiency and effectiveness of the search. But there must be a limit. In surveying, one could review every source in a library or every term in a thesaurus. Such thoroughness would waste time. Usefulness of this tactic probably follows a curve in which additional SURVEYing beyond a certain point produces diminishing returns. Testing may demonstrate where that point is.

WEIGH also includes some stopping questions. If, for example, one has mentally sorted relevant sources from most to least likely to produce information on the query, then when is the optimal time to stop searching in one source and move to the next one? That point may well come before one has exhausted every conceivable possibility in a given source.

Summary

After reviewing briefly the literature of human information search strategy, four types of models of information search strategy were defined: models for idealizing searching, representing searching, teaching searching, and facilitating searching. The model in this article was presented as being primarily a facilitation model and secondarily a teaching model.

The concept of the search tactic was defined, as well as four categories of tactics for the search proper: monitoring tactics, file structure tactics, search formulation tactics, and term tactics.

Twenty-nine tactics were named, defined, and discussed, and various implications of search tactics for research in search strategy were discussed.

___________________

* A major orientation of the article is to focus on and use the strengths and flexibility of human thinking processes. It is suggested that search theory and practice may be advanced through a greater attention to the specifically human, psychological processes involved in searching, as distinct from the logical, or formal, properties of the process. In addition, we turn from a focus on machinery, the information technology, to the brain that is running it.

**These four types of models may also be useful in conceptualizing research on human behavior in other areas of information science, e.g., classification/indexing and relevance assessment

TABLE I. Summary of information search tactics and definitions

Acknowledgements

This work was supported in part by a grant from the University of Washington Graduate School Research Fund.

References

1. Stevens, M.E. "Strategies for Organizing and Searching." In: Fenichel, C., Ed. Changing Patterns in Information Retrieval. Tenth Annual National Information Retrieval Colloquium. Washington, DC: American Society for Information Science; 1974: 47-79.

2. Paisley, W.; Butler, M. Computer Assistance in Information Work. A report to the Division of Science Information, National Science Foundation, Contract #C76-05489. Palo Alto, CA: Applied Communication Research, Inc.; 1977.

3. Meredith, J.C. "Machine-Assisted Approach to General Reference Materials." Journal of the American Society for Information Science. 22(3):176-186; 1971.

4. Jahoda, G. "Reference Question Analysis and Search Strategy Development by Man and Machine." Journal of the American Society for Information Science. 25 (3):139-144 ; 1974.

5. Carlson, G. Search Strategy by Reference Librarians. Part 3 of the Final Report on the Organization of Large Files. NSF Contract C-280. Sherman Oaks, CA: Advance Information Systems Division, Hughes Dynamics; 1964.

6. Wyer, J.I. Reference Work: A Textbook for Students of Library Work and Librarians. Chicago: American Library Association; 1930.

7. Shores, L. Basic Reference Sources; An Introduction to Materials and Methods. Chicago: American Library Association; 1954.

8. Cheney, F.N. Fundamental Reference Sources. Chicago: American Library Association; 1971.

9. Hutchins, M. Introduction to Reference Work. Chicago: American Library Association; 1944.

10. Wynar, B.S. "Reference Theory: Situation Hopeless But Not Impossible." College and Research Libraries. 28(5):337-342; 1967.

11. Vavrek, B.F. "A Theory of Reference Service." College and Research Libraries. 29(6):508-510; 1968.

12. Emery, R. "Steps in Reference Theory." Library Association Record. 72(3):88-90; 96; 1970.

13. Rugh, A.G. "Toward a Science of Reference Work: Basic Concepts." RQ. 14(4):293-299 ;1975.

14. Holler, F. "Toward a Reference Theory." RQ. 14(4):301-309; 1975.

15. Whittaker, K. "Towards a Theory for Reference and Information Service." Journal of Librarianship. 9(1):49-63; 1977.

16. Neill, S.D. "Problem Solving and the Reference Process." RQ. 14(4):310-315; 1975.

17. Benson, J.; Maloney, R.K. "Principles of Searching." RQ. 14(4):316-320;1975.

18. Katz, W.A. Introduction to Reference Work. Volume II. Reference Services and Reference Processes. New York: McGraw-Hill; 1974: 2nd ed.

19. Katz, W.A. Introduction to Reference Work. Volume II. Reference Services and Reference Processes. New York: McGraw-Hill ; 1978: 3rd ed.

20. Josel, N.A. "Ten Reference Commandments." RQ. 11(2):146- 147; 1971.

21. Adams, J.L. Conceptual Blockbusting. San Francisco: San Francisco Book; 1976.

22. Webster's New Collegiate Dictionary. Springfield, MA: G & C. Merriam; 1956.

23. Foskett, A.C. The Subject Approach to Information. London: Clive Bingley; 1969.

24. Hodnett, E. The Art of Problem Solving. New York: Harper; 1955.

25. Meadow, C.T. The Analysis of Information Systems. Los Angeles: Melville; 1973: 2nd ed.

26. Bates, M.J. "The Testing of Information Search Tactics." Proceedings of the American Society for Information Science Annual Meeting. 15:25-27; 1978.

27. Lancaster, F.W. Information Retrieval Systems. New York: Wiley; 1968.

28. Coates, E.J. Subject Catalogues: Headings and Structure. London: The Library Association; 1960.

29. ALA Rules for Filing Catalog Cards. Chicago: American Library Association; 1968: 2nd ed. abridged.

30. Needham, C.D. Organizing Knowledge in Libraries. Seminar Press; 1971: 2nd rev. ed.

31. Lynch, M.J. "Reference Interviews in Public Libraries." Library Quarterly. 48(2):1 19-142; 1978.

32. Jahoda, G.; Olson, P.E. "Analyzing the Reference Process." RQ. 12(2):148-156; 1972.

33. Bruner, J.S.; Goodnow, J.J.; Austin, G.A. A Study of Thinking. New York: Science Editions; 1956.