context as a factor in personal information management …a personal information management system...

Context as a Factor in Personal InformationManagement Systems

Deborah K. BarreauCollege of Library and Information Services. 4105 Hornbake. University of Maryland,Coliege Park, MD 20742. E-mail: barreau© wam.umd.edu

Personal information management (PIM) systems are in-formation systems developed by individuals for use in awork environment. Seven managers were interviewed toobserve how their electronic documents were organized,stored, and retrieved. The purpose of the study was to in-vestigate classification behavior both to identify the fea-tures of a PIM system and to suggest whether the factorswhich influence classification decisions in an electronicenvironment were consistent with the factors that Kwasnikobserved for physical documents in an office. It is sug-gested that these behaviors may be influenced by thehardware and software environment and may impact per-sonal as well as organizational efficiency.

Introduction

A personal information management system (PIM) isan information system developed by or created for anindividual for personal use in a work environment. It in-cludes a person's methods and rules for acquiring the in-formation which becomes part of the system, the mech-anisms for organizing and storing the information, therules and procedures for maintaining the system, themechanisms for retrieval, and the procedures for produc-ing the various outputs required. Interest in PIM systemshas advanced as a result of the rapid proliferation of mi-crocomputers in the 1980s. A variety of personal infor-mation management software and hardware, such asnotebook computers and pen-based input devices, havebeen developed to support individuals and the way theywork. The industry's efforts have concentrated on mak-ing systems easier to learn and easier to use, respondingto market demand for such features as graphical inter-faces and multitasking environments.

Research into how people work and how they use sys-tems is limited, but the studies which have been done are

Received September 29, 1993: revised June 30, 1994-. accepted June30, 1994.

© 1995 John Wiley & Sons. Inc.

revealing and useful. It has been demonstrated that usersfrequently employ satisficing strategies when using sys-tems, concentrating more on accomplishing job-relatedtasks than on learning to become more proficient withthe system itself (Carroll, 1987). Furthermore, it is gen-erally accepted that users will use what is convenient andreadily accessible to them rather than attempt to findother sources which may be more appropriate but arenot so easily obtained (Marchionini, in press). Work is ahighly situational. context-based process which requiresthe accomplishment of tasks within the constraints of tbeworking environment. The classification of work prod-ucts and processes rarely fit neatly into document-spe-cific categories such as subject and form which are com-mon to many information storage and retrieval systems.Subject and form are important, but other factors suchas frequency and date of last use, relative importance toa task, and the amount of cognitive energy required toassign a category must be considered when classifying,organizing, and retrieving information from an individ-ual's personal environment.

Software to support PIM systems must be flexible andadaptable to the task needs and personal idiosyncraciesof the individuals who use them. To be useful to the in-dividual, the software should help the user to integratethe processes of selecting, organizing, storing, maintain-ing, and retrieving information. It should be flexibleenough to support the way the individual works whileallowing transformation of information into formats andschemas that can be used by others. Although softwareenvironments in many places have far to go to achievethis, systems are far better and more flexible than everbefore. Consequently, individuals develop unique waysof using software and of organizing and managing infor-mation within their workspaces that are highly personal-ized. These "personalized" tasks and strategies within astandardized business computing environment are whatis described here as a PIM system.

The context in which a document is acquired or cre-

JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE. 46(5):327-339, 1995 CCC 0002-6231/95/050327-13

ated may have much to do with how it is classified,stored, and later retrieved. The importance of situationalfactors in human cognition was illustrated in studies ofcues used to retrieve members of natural categories(Walker & Kintsch, 1985) and in studies of skilled mem-ory and expertise (Ericsson & Staszewski. 1989). Thesestudies are relevant because PIM systems are more inti-mate and more in tune with the way people think andwork than information systems designed for general use.The purpose of this article is to examine context as a fac-tor in PIM systems to suggest how it may influence clas-sification decisions and ultimately retrieval. It describesa study that was undertaken to explore the factors thatinfluence the way individuals manage informationwithin their personal space. Interviews were conductedand analyzed to compare the patterns observed by Kwas-nik in her study of the influence of context upon classi-fication of documents in an office (Kwasnik, 1989b) withclassificatory behavior of individuals in an electronic en-vironment, and to explore the feasibility of employingthe methodology in an electronic environment.

Characteristics of Personal InformationManagement Systems

The structure of an information system includes in-puts, the individual with information needs as well as theinformation and entities which are part of the system:the information storage and retrieval (ISAR) system, themechanisms for organizing, storing, maintaining, andretrieving information; and the outputs, the informationor entities obtained from the system (Soergel, 1985). Fig-ure I is a representation of an information system basedon Soergel's model.

Acquisitions

In a PIM system the user whose problems, needs, andquestions are inputs to the system is usually the personwho decides upon the information and the entities to beincluded in that system. This is not exclusively the caseas any environment imposes certain constraints. Com-pany-selected hardware or software, limitations upon ac-cess to electronic information sources, and requirementsto build upon work started by someone else are some ofthe constraints that may impact decisions about what toinclude within the system. Despite constraints, much ofthe information acquired for a PIM system is generatedby the user, or has value added by the user. Informationentities may vary widely by subject and format, includ-ing formal documents that require polish, precision, andaccuracy; informal notes and messages which permit ir-regular formats and spelling; data files; spreadsheets andtables; graphics; and program utilities, among others.

This article is concerned with inputs to informationsystems as they relate to the organization of informationwithin the information space; that is, how the user de-

INPUT

Theus«rwHhneeds and problems

INFORMATION SYSTEM

D«flnltlon or aoop* for aoqulaltlonalanrtB tor oroantzlng •rx* •tortng InfomrwUon

Rul«» «nd proo*<]ur*« for malntalnino aystttmafor InfomMttlon r*trt<»vaJ

Anewwi to quMOonsVahiMKldMl produdi

RaportiSummviM

OttMT

FIG. I. Structure of an information system.

fines, groups, and labels the information he or she in-cludes within the system.

Organization and Storage

Mechanisms for organizing and storing informationare those tools and procedures that classify, name, group,and place information in a location for later retrieval.Within a typical ISAR system, these are the accesspoints, the controlled vocabularies, and the storage me-dia and processes used to facilitate organization of thecollection and retrieval of data relative to anticipatedneeds. Assigning titles, names, and subjects to items, or-ganizing them logically, and storing them in physical lo-cations are typical activities. Attributes such as docu-ment type, date, language, and target audience are alsoused to classify and distinguish information within col-lections.

Within a PIM system, the item attributes are oftenfine-grained and so personal that it is difficult to distin-guish and to represent them. This difficulty is furthercomplicated by the dynamic state of the individual's en-vironment as priorities are rearranged and workloadschanged. Several unique attributes have been suggestedfor PIM systems, and it is generally recognized thatmultiple access points are needed. Malone found thatplacement of documents on a desk in an office environ-ment often served a reminding function to the individual(Malone. 1983). This location was not based on the phys-ical characteristics or attributes of the documents, but onthe use or relevance to the person. Malone also found

328 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE—June 1995

that individuals would group things in a location becauseof the difficulty in classifying them, so items were foundtogether which did not fit easily anywhere else, althoughthey might have no other attribute in common. This isevidence ofa satisficing strategy, ofa person's avoidanceof the cognitive effort required to make a decision.

Kwasnik found that the use to which a document isput or intended is often the first classification rule in-voked by individuals in organizing their personal docu-ments {Kwasnik, 1989a). Her work is described in moredetail later.

Maintenance

Maintaining information systems is an importantfunction, yet one often underestimated in ISAR systemdesign. Regular system backups and routines for updat-ing incorrect or out-of-date information are universallyaccepted requirements, but rules for deletion or migra-tion vary widely from system to system. In an era of in-expensive memory and storage devices and fast proces-sors, there has been a tendency to allow systems to grow.Individuals desiring comprehensive treatment ofa sub-ject have need to keep as much information as is relevantand feasible. Unfortunately, growth may occur with littleassessment and review of the relevance or quality of theinformation due to the amount of time required to makesuch determinations. Growth can continue as long as thetools for organizing and retrieving information keeppace with the storage, and as long as system performanceand personal productivity are not impaired. Decisionsfor what and when to delete or migrate to secondary stor-age are critical to efficiency, yet are frequently made at atime and under conditions dictated by circumstancerather than by rule.

Retrieval

Information retrieval within ISAR systems is based onuser-defined queries and system access features, and theability of the system to match the user's request to itsretrieval language is one measure of p>erfonnance. Aswith any information system, the translation ofthe user'squery in a PIM system is not a trivial problem.

The process of retrieval in a PIM system dependsgreatly upon the situation or event, the informationneed, which initiates the search. Formal ISAR systemsrely upon controlled vocabularies, structured files, andelaborate filing mechanisms to facilitate use. Within aPIM system, each person's vocabulary is itself "con-trolled" since it is unique for that person. It takes disci-pline to settle upon a subset of one's vocabulary for use,particulariy when the effort required to do so interfereswith the accomplishment ofa task. On the other hand,there are retrieval cues available to persons in their ownsystems which formal systems lack. Dates, file size, andcontext clues from the user's situation at the time the

document was created or last used are among the cueswhich can combine to jog a person's memory to locatean item or to retrieve a fact. Even the acts of classifying,naming, and storing a document to disk occur in thecontext of other work and events that may themselvesbecome an aid to retrieval later.

Output

One ofthe most important features in the structure ofinformation systems is output. Users come to the systemto resolve a need or answer a question, and the system'scapability to satisfy the request is greatly dependent uponthe quality of output. For PIM systems, output must becriterion-based, subject to the needs and objectives oftheindividual rather than normalized to generic tasks formultiple users as in general ISAR systems. Rexibilityand variability of output become important criteria inevaluating PIM systems.

In summary, one factor that defines PIM systems andthat sets them apart from general ISAR systems is theunique, situational aspect ofthe working environment.The context in which a person works, including the na-ture ofthe task, the subject matter, the available tools,the intended audience for the work, and the intended usefor the information are factors which are in play at thetime the item is acquired, created, classified, stored, orretrieved. Successful PIM systems, and software used toimplement these systems, should accommodate situa-tiona! factors to facilitate system use.

Context and Classification: The Kwasnik Study

Context is the situation in which an event occurs.Context includes all aspects ofa person's experience, andit has long been recognized as a factor in human behav-ior. We are often unable to function well when we aretired, or if environmental conditions are extreme (theroom is too hot or too cold, for example), or if there aredistractions. In a corporation, a manager with years ofexperience in the organizational culture has a perspec-tive on what information is important for the job thatdiffers from that ofa young management trainee. Everyindividual is unique, possessing experiences which im-pact his or her decisions and reactions to new experi-ences, and these experiences become the basis for indi-vidual growth and learning.

People tend to analyze and to classify new objects orexperiences in life by comparing information aboutthese new objects and experiences with those which arefamiliar. This "reasoning by analogy" is the context the-ory of classification {Kwasnik, 1989b; Medin & Schatfer,1978). Prototype theory is an extension of classificationtheory, implying that exemplar experiences lead to pro-totypes on the basis ofa category's central tendency. Ac-cording to the theory, classifications are based on percep-tions of how well items fit the individual's idea of the

JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE—June 1995 329

prototype. Again, each person's experiences are uniqueand, even when people agree on a prototype, they maynot agree upon the characteristics which define it orwhether a particular object is close to the prototype or faraway.

Malone explored how people organize things in thecontext oftheir offices. He interviewed ten people fromvarious job categories to explore the patterns in their or-ganizing behavior and the implications for office infor-mation systems (Malone, 1983). He found that individ-uals tend to organize information in their offices into"files" and "piles." Files are we 11-organized, often la-beled stacks or folders whereas piles contain miscella-neous documents that have no apparent organization orlabeling. Malone discovered that the location ofthe filesand piles on the subjects' desks was an indication oftheirimportance, serving a reminding function of work whichneeded to be done. He also found that the subjects' de-scriptions ofthe information probes used to retrieve adocument did not correspond to their descriptions oftheir organizational schemes in about two thirds ofthecases. Malone concluded that automated systems couldresolve many ofthe problems in these workspaces bysupporting multidimensional classification and by pro-viding devices which would serve the reminding functionsuch as varying the size or color of icons based upon theimportance ofthe document or having an item surrogateappear on the screen periodically.

Malone's investigation raises questions concerningthe process by which people organize and retrieve infor-mation within their workspace, and Kwasnik raises someofthe same questions in her work. She investigated theinfluence of context upon the processes employed by in-dividuals in organizing and classifying their own docu-ments in their own space. Kwasnik interviewed eight re-searchers from various academic disciplines and askedthem to give her a "guided tour" ofthe documents intheir offices and work areas. In a second visit, subjectswere asked to sort a day's mail and to describe each piecealoud and explain what they would normally do with it.In a third visit to four ofthe subjects, she sorted a day'smail for each based upon the "rules" identified in thesecond visit. These interviews were taped, and interviewsfrom the first and second visits were also transcribed. Sheanalyzed the transcripts to identify the categories of doc-uments identified by users and the dimensions used inmaking these classifications. Head nouns were selectedto determine how individuals name and classify docu-ments they deal with, and to identify the dimensionsused to organize them. (A software program, Qualog, adata analysis tool developed at Syracuse University, wasused to facilitate the analysis.) Two examples follow:

"on the top shelf are books that are very seldom used""correspondence I must deal with immediately goes intomy briefcase" (Kwasnik, !992b, p. 4)

mensions of LOCATION {top shelf, briefcase), FORM(books, correspondence), TIME (very seldom, immedi-ately), and USE (used, deal with). She coded and inven-toried all head nouns mentioned by the subjects in theinterviews. Once this task was completed, she analyzedthe dimensions and organized them into categories andsubdivisions (Fig. 2).

Kwasnik's analysis produced a frequency distributionofthe incidence of each dimension described in Figure 2among the categorizations ofthe head nouns and generalnouns identified. She defined general nouns to be thosewhich identify documents without mentioning form;things, stuff, and items are examples. She also createdfrequency distributions ofthe occurrence of dimensionsin pairs, when two dimensions were used in the classifi-

SITUATIONAL ATTRIBUTESAccessCircumstanceNeed/RequirementsOwnership of thedocumentRelated to meRoom/SpaceSourceUse/Purpose

DOCUMENT ATTRIBUTESAuthor ^FormTopicPhysical Attribute

DISPOSITIONChangeDiscardKeepLocatePostpone decision

ORDER/SCHEMEAccumulationArrangementGroupSeparateUnfinished arrangement

TIMEVALUE

ImportantInterestingNeeds improvementNot valuableSecret/confidentialUnspecified valueWorks for me

COGNITIVE STATEDon't knowWant to remember"Just know"

In the above examples, each decision invoked the di- p. 6)

FIG. 2. Fundamental categories with subdivisions of dimensionsalong which people classified personal docunients(see Kwasnik. 1992a.


cation, and in triplets, when three dimensions were used,to observe patterns. Ofthe dimensions identified, the fiveoccurring most frequently were USE, TOPIC, TIME,CIRCUMSTANCE, and FORM. When two dimensionswere used, TOPIC and USE were most frequentlypaired, and when three dimensions were invoked, CIR-CUMSTANCE-TIME-USE were used together most of-ten. Situationai attributes were invoked by her subjectsmore often than any category except form.

Kwasnik's work supports the notion that traditionalsystems which rely too heavily on document attributesare insufficient for an individual's requirements for or-ganizing, managing, and retrieving information for per-sonal use. Furthermore, if context is a key factor in theorganization of materials within a person's physicalspace, it follows that context is just as important, or evenmore important, in how those materials are used in anelectronic environment to do work. Certainly, it wouldbe interesting to investigate classification behaviors of in-dividuals in an electronic environment to see if the di-mensions identified by Kwasnik tend to hold up. Thegoal of such an investigation is to suggest whether peopleorganize information in their electronic space by similardimensions and to explore the implications for peoplewho work in environments where documents must beshared. The next section describes research performedwith these goals in mind.

Methodology

A study was undertaken to investigate the factors thatinfluence classification decisions by individuals in theirelectronic environments. Seven individuals, all manag-ers, were interviewed and were asked to provide a guidedtour of their electronic directories both to identify thedocuments in their workspace and to isolate factorsaffecting their decisions to include the documents withintheir systems. Managers were chosen because ofthe va-riety and quantity of information they must deal with ona daily basis. The interviews were analyzed to identifythe dimensions used in organizing and retrieving infor-mation in an electronic environment.

Subjects

Seven managers were interviewed, including threemales and four females, considered to be experts in theirrespective domains, but only two would classify them-selves as experts with computers. Four ofthe managersare employed on the same project and work in the samecomputing environment, a Novell local area network(LAN). These individuals work in a large informationmanagement company providing services to govern-ment and commercial clients. One is a financial man-ager, one oversees document delivery functions, and twomanage reference and referral services. Three of thesefour have computers without hard drives and are there-

fore restricted to the storage available on 3.5" diskettesand on the public and private directories of the fileserver, but the fourth has a 120-megabyte hard drive aswell. A fifth subject works in the same company in adifferent department and supervises a group of graphicartists. She uses a Macintosh computer on an Appletalknetwork which has access to a Novell LAN via MacLinkPlus. This group supports a variety of projects, includingthe project ofthe four managers already mentioned.

Two of the managers interviewed were located atdifferent sites. One was a project manager in a researchdepartment for a government agency, and many ofhisresponsibilities are computer-related. The computingenvironment included a powerful microcomputer withtwo hard drives, operating OS/2, a multitasking operat-ing system. This was linked to a Banyan network. Thefinal subject works in a networked environment as well,running Microsoft Windows as the interface to his ownhard drive, to a local area network, and to an IBM main-frame. This subject is a project manager/research scien-tist at a major corporation providing services to the gov-ernment. His current project is designing a user interfacefor a system to be used by a federal agency.

Procedures

The managers were told in advance ofthe interviewnot to clean-up or reorganize their directories, but toleave them as usual. The interview followed the loosestructure of a guided interview. Subjects were encour-aged to describe freely the documents and informationwithin their workspace, but questions were asked to elicitinformation about the typical functions of personal in-formation management systems. For example, subjectstypically were asked questions about where documentsin their workspace originated, how they came to bewhere they are, when and under what circumstances theymight be moved or deleted, and how they would goabout finding the document when it was needed. The in-terview was concluded when the subject completed thetour.

The interviews were taped, transcribed, and were an-alyzed within the framework ofa PIM system, that is,the rules and procedures applied by the individuals inacquiring, organizing, maintaining, and retrieving docu-ments within their personal workspace. The intervieweralso made brief notes during the sessions. Although noneof the subjects used formal PIM system software, thedata suggest that rules and procedures for managing in-formation were present. The transcripts were analyzedto identify the documents used by individuals in theirelectronic workspace as well as to identify the dimen-sions and rules applied in classifying them. Identificationof the categories of documents and the general nounsused was accomplished with ZYINDEX, commercialtext retrieval software. Each occurrence of a term wasanalyzed in context to determine which if any ofthe di-


Unique Document Identifiers or "Head Nouns'

Subject 1referralsmanualappraisalspeechletterdecisionrecommendationguidepoliciesagreementsearchbits and pieces

Subject 5costinggoalshotsheetthesaurusdirections

Subject 2windowpunnotesrecordjokedecorationarticleoverviewexecutablesfontstalk

Subject 6mailerlogocompositejobsarchivebrochuremastheadtransformationsmaptrashchapterpageslidesimagesillustrations

by Individual Interviewed

Subject 3correspondencemodificationplantaskbillingexpensesappropriationsdraftnarrativebid

Subject 7labelsrequestsanalysisevaluationattachment

Most Commonly Used General Nouns (used by all subjects)

thing, thingsfile, filesdisk, disketteStuffdirectory, directoriesnetwork or LANreportinformationcopyprojectmail, emailmemomessage

2081368374 (not used by subject 7)744337

• fr- '21 (not used by subject 7)21 (not used by subject 7)20 (not used by subject 1)20 (not used by subject 4)16

Subject 4menuservernewsletterinquiriesnewspaperinsertionsspecsbriefingdemo

FIG. 3. A total of 136 unique nouns or general nouns were used by the subjects to describe the items in their workspace.

mensions identified by Kwasnik were applied by the in-dividuals in their classification descriptions ofthe itemsin their electronic space. The dimensions identified werecoded with the classification descriptions and were writ-ten to a file. The document categories identified that areunique to each ofthe individuals interviewed, and a listofthe general nouns used most frequently by ali subjectsappear in Figure 3.

Once the document categories and rules were identi-fied and the dimensions coded, the rules were further an-alyzed for their application in acquisition, organization,maintenance, and retrieval of information for PIM sys-

tems. These are discussed briefly in the sections whichfollow.

Acquisitions. The following categories of docu-ments were specified by the subjects when asked to de-scribe the information included in their workspace. Themanagers employed descriptions consistent with the di-mensions identified by Kwasnik and. similar to Kwas-nik's findings, situational attributes were used most oftento describe documents acquired.

(I) Things from a public directory, or in a public direc-


tory (SOURCE, USE. CIRCUMSTANCE. FORM.ACCESS, OWNERSHIP). All subjects includeditems from shared space within their collections.Subjects distinguished between items that werecopied from a public directory and those that wereused and maintained in the public directory. In theformer case, the public directory is the source for adocument the subject has acquired. In the latter, thepublic directory is a resource for shared files such asa referral database and an accessions list used forretrieval of factual information.

(2) Electronic mail messages (IMPORTANT. CIR-CUMSTANCE. LOCATE. FORM. TOPIC, AC-CESS. COGNITIVE STATE, NEED, TIME). Elec-tronic-mail messages were kept for easy accesswhen they contained information that the subjectmight need in the future, and most respondents in-dicated messages were kept as reminders of some-thing that needs to be done. One subject suggestedthat "sometimes I tind something that's really in-teresting so I just keep inhere for awhile."

(3) Software (FORM. TIME, USE. CIRCUM-STANCE). All subjects interviewed referred to thesoftware that they use when asked to describe thedocuments in Iheir electronic environment. Appli-cations mentioned included word processing,spreadsheets, graphics, demonstration programs,utilities and tools (commercial and self-generated),database management systems, bulletin boards,and electronic-mail. Although subjects referred tothe software as a too! to accomplish work, such as"XYWrite, the word processor I use," it is clear thatsoftware is used as a dimension in classification aswell for the files created with the software; that is,"Lotus files," "SAS logs," and "Harvard Graphicsdiskettes."

(4) Things created by the subject which are used regu-larly or currently (LOCATE, USE, CIRCUM-STANCE, TIME. FORM, RELATED TO ME,TOPIC). One olthe most frequently mentioned de-scriptors of documents within the individual's elec-tronic environment was frequency of use. Subjectsdescribed files that were used "most days" andthings they had been working on "for a couple ofweeks."

(5) Things on a relevant topic (TOPIC, FORM). Onesubject described his collection by saying that 75%ofhis files are budget-related. Others described in-formation relevant to their projects. In the elec-tronic environment, topic was rarely mentionedwithout reference to form.

(6) Something that someone else provided (OWNER-SHIP OF THE DOCUMENT, SOURCE, RE-LATED TO ME). Documents sent to or given tothe subjects included reports for review, informa-tion to be converted into some other formal or tobe used in some other product, and files left by apredecessor. "George's disk." the "budget fromMike." and "someone has sent me bis progress re-pon" are examples.

(7) A copy of something the subject created for some-one else (CIRCUMSTANCE, USE, OWNERSHIP

OF THE DOCUMENT. USE, VALUE). A data-base search that was done for someone else was keptbecause the subject thought it might be useful.Something "special" created for and at the requestofthe project monitor was kept for reference shouldadditional work be required.

(8) Something created for a special reason (CIRCUM-STANCE. USE). Policies and procedures were cre-ated by one subject before she went on maternityleave to help others to keep track of responsibilitiesIn her absence. When she returned, she found ituseful to maintain the information.

(9) Things that may prevent more work later on, orthat might be needed again (TIME, CIRCUM-STANCE. ACCESS, DISPOSITION). Subjects of-ten referred to things which were kept just in casethey might be needed again. There were two exam-ples of individuals describing a situation where adocument was kept to avoid work in the future. Thefirst was a case where multiple copies ofa documentwere kept as a precaution against loss, beeause itwas something that is issued periodically and "hasto go out quickly, be accurate." Another examplewas a logo, created by scanning, which required alot of time to clean-up for use in a publication. Thiswas saved to avoid having to redo the work shouldit ever be needed again.

(10) Personal stuff (USE. FORM, CIRCUMSTANCE.RELATED TO ME). Only one subject admittedthat there were personal documents in her collec-tion. She mentioned specifically instructions forhow to feed her cats, and her fiance's resume. Othersubjects mentioned receiving personal messages inelectronic mail.

Classification and Organization. The rules for clas-sifying and organizing information in the subject's work-space include the dimensions described below. Often,multiple dimensions were described in a classificationevent, and the following list is inclusive of all dimensionsapplied, whether used singly or in conjunction withsomething else. FORM. LOCATE. TOPIC, CIRCUM-STANCE, and USE/PURPOSE were mentioned mostoften in classification decisions.

(1) Things used regularly, or used currently (USE, LO-CATE, CIRCUMSTANCE, TIME). All the sub-jects classified some items based upon their patternsof use. Examples include "my working stuff.""things we use all the time," "the main things I'minvolved in," and rarely used documents which arestored "in there, out ofthe way."

(2) Topical information (TOPIC. FORM), Every sub-ject mentioned topic as a dimension in classifica-tion. Examples include "construction stuff," "OS/2 stuff," and things that "relate to AIDS," Topicwas usually mentioned with other dimensions suchas location and form. Kwasnik considered titles, thenames of documents, as a feature of the TOPIC di-mension, but in an electronic environment, the file


name is just as often an indicator of its form, itspurpose, or some other characteristic.

(3) Things that belong to somebody (OWNERSHIPOF THE DOCUMENT). Ownership is somewhatproblematic in electronic environments. In anoffice, ownership might refer to "books that belongto me" or "a newspaper that belongs to her." In anelectronic environment, once a copy of somethingis in our space, we own it in a sense, and may mod-ify it to suit our purposes. Additionally, once wecreate a document SF)ecifically for someone else, itmay become associated with that person. Itemswere classified by the subjects based upon owner-ship, but it was not always clear whether they were"from" someone, "for" them, or "about" them.

(4) Things in a particular format (FORM, TOPIC).Form was used by every subject as a dimension inclassification. Form in this case has as much to dowith the type of software used to produce a docu-ment as it does with the final output format. Exam-ples include spreadsheets, graphs, logs, statistics,proposals, and performance appraisals. Form wassometimes the only dimension used in describingdocuments, yet it appeared often with topic, loca-tion, and use.

(5) Age or currency ofthe documents (TIME). Age anddate ofa document were often cited as dimensionsin classification. One subject organized her memosto an individual by month. Another subject had adiskette of "historical files."

(6) Things organized by circumstance, habit, or due toprocedures (CIRCUMSTANCE). An example ofcircumstance as a dimension in classification issomething shared, such as "a disk I share withPSG," and "because there are so many peopleworking on the project" all items related to the proj-ect were stored on a single diskette.

(7) Things organized by document characteristics(PHYSICAL ATTRIBUTE). Rarely mentioned inthe interviews was the classification of informationbased upon physical features. In an office, physicalfeatures such as the size ofa publication may welldetennine whether it is organized with other itemsof similar size on a shelf, or whether it is filed in adrawer with items on the same topic. In an elec-tronic environment, size ofthe document is rarelya factor unless it is so large that it creates a problemof storage or manipulation. Additionally, datacompression algorithms and other features make itpossible to minimize these differences. However,one example ofthe attribute of size determining aclassification decision occurred when one of thesubjects made a decision to put two "huge, huge"spreadsheets on a diskette. In the environmentswhere graphical interfaces are used, physical attri-butes are more varied. In one subject's environ-ment, a variety of colorful, creative icons ("bau-bles") were used to classify items that perform thefunction of opening up windows. This subject re-ferred lo other icons as "executables," a type ofphysical attribute as it distinguishes files that dothings.

(8) Things which may be needed quickly, or whichmay be needed sometime (ACCESS. NEED). Sub-jects frequently mentioned organizing informationbecause ofthe need to retrieve it quickly, or to haveit "where I can always find it."

(9) Things which are special, interesting, or important,or not important at all (VALUE). The value or im-portance ofa particular item was sometimes a fac-tor in document classification. An example is thesubject who described a set of documents as the"major decisions that deal with the agency."

(10) Things organized according to factors unknown orforgotten (COGNITIVE STATE). It was not un-common for subjects to state that they were not re-ally sure why certain files were found where theywere.

System Maintenance. System maintenance is de-fined here as modifying, moving, or deleting documentsfrom the PIM system. More than one dimension was in-volved in most maintenance decisions, and Kwasnik'sdimension of DISPOSITION (to keep, discard- change,locate or postpone decisions) is integral to all mainte-nance tasks and should be understood as applicable ineach ofthe following rules.

(1) Documents are removed once they exist in anotherformat (NEED, USE. FORM). Some documentswere removed once they were converted to anothersoftware format. In other cases, documents were cre-ated by the subject for use in some larger publicationand the pieces remained in the subject's environ-ment until the publication was done. Sometimesdocuments were kept only until a hard copy was ob-tained.

(2) Documents are moved or deleted because they arefinished or because they will not be needed again(NEED, USE. ACCESS, TIME. CIRCUM-STANCE). Closely related to the above rule, theseare documents that were created for a specific pur-pose, and once that purpose was fulfilled there wasno need for the document. Sometimes items weresimply moved to less prominent locations "whenI'm finished with it here and I may need it for some-thing else."

(3) Documents are moved or deleted based upon thetime available to do this (TIME, CIRCUM-STANCE. USE, NEED). Most subjects admitted toirregular system maintenance. Most relied upon net-work managers to back-up their data, leaving harddrives and diskettes vulnerable. Maintenance was re-ported to occur "when I'm not doing anything." "I'lljust be looking at them," or "in a random sort ofway." During the seven interviews, each subject de-leted at least one file or message. Three ofthe subjectsstated that they removed files when they were told toby their LAN administrators.

(4) Documents are deleted or moved based upon the ageofthe items (TIME. DOCUMENT ATTRIBUTE).During the interviews, two of the subjects deleteditems because ofthe age ofthe documents. In one


case, the subject said "that's old." and in anothercase, "that's an old business plan."

(5) Documents are deleted or moved based upon sizefactors such as the amount of available space, thenumber of files in a director), and file size (ROOM/SPACE. PHYSICAL ATTRIBUTE, CIRCUM-STANCE). In some cases a rule of thumb was ap-plied, such as the manager who tried to limit his Lo-tus directory to "30 to 35 files." In other instances itwas more an act of desperation "in order to get someroom on my hard drive." As might be expected, sub-jects with the least amount of space available per-formed document deletion or transfer more fre-quently than those who had additional media at theirdisposal. At least two subjects indicated they weregetting additional hard drives to combat the spaceproblem.

(6) Documents are deleted or moved based upon the im-portance ofthe information (VALUE. CIRCUM-STANCE. FORM, TOPIC). One subject had veryelaborate back-up schemes for his most importantdocuments. Although one manager said that shekeeps everything, it was clear that moving copies tostorage as a safeguard against losing it was left to net-work administrators. Several subjects indicated thatthey backed-up important information, but not oneperson interviewed performed regular back-ups of alltheir files. At least one person was aware that therewere no back-ups of certain important informationbeyond the LAN tapes, and that it would be diflicultto retrieve anything from the tape back-ups that wasmore than a few weeks old.

(7) Documents are maintained or updated when new in-formation is received (CIRCUMSTANCE, USE).Most subjects used some documents which requiredregular modification. These items were always orga-nized so the user could access them easily, but thelocation wasdiRerent for each subject. Spreadsheets,databases, referral lists, and reports are among thecategories of items routinely maintained.

(8) Documents are deleted or removed because thecontents are unknown to the user, or because theuser does not remember the document (COGNI-TIVE STATE). Almost every subject had items in adirectory which he or she could not identify. In someinstances, these were macros or software-generatedset-up files. In other instances, they were files thatwere left by a predecessor or files given to the subjectby someone else, and due to lack of time or interestnothing had been done to identify and classify theinformation in a way that might be meaningful.Some subjects found items they had created in placesthey did nol expject to find them, and could not ex-plain or remember their reasons for putting themthere.

Information Retrieval. One question asked of eachsubject in the interview was how he or she would goabout finding a document. Each subject employed strat-egies for organizing information with the ease of futureretrieval in mind, but many ofthe strategies employed in

retrieving information not only took advantage of thatorganization, but compensated for its limitations. In aphysical space, there are many simultaneous physicalcues present to assist in retrieval, but in a PIM system,the individual must compensate by developing informa-tion seeking strategies which maximize the situationalfactors and system-defined features available. ORDER/SCHEME and FORM were the dimensions cited mostoften in retrieval as illustrated below.

(1) Location (ORDER/SCHEME. ACCESS. USE. CIR-CUMSTANCE. LOCATE). One ofthe conventionsused frequently in retrieving documents was by loca-tion. Software applications used most often were ac-cessible from the root directory of hard drives or inprominent places on the menus of networks. Sub-jects kept "all my Harvard Graphics files on a disk"orturned to shared directories on the network wherethe referral lists, macros, or logos were always found.

(2) Names or titles (OWNERSHIP, RELATED TOME. AUTHOR, SOURCE, ORDER/SCHEME.VALUE, DOCUMENT ATTRIBUTE. TOPIC,CIRCUMSTANCE). Document names may benames that serve as a surrogate title or alternate titleto fit the constraints ofthe 8 to 11 characters avail-able in the MS-DOS environment. One subject saidhe tried to create meaningful labels in naming files sothat they wouid trigger his memory when he went tolook for something.

A typical naming convention used was to attachmeaningful suffixes or file extensions to make filenames more distinguishable. Examples includesuffixes to group items together topically, suffixes togroup them by use, and suffixes to make a distinctionamong documents with a similar prefix.

Mnemonic devices such as file names that hint atwhat the item does or is were also used. An exampleofthe former is MOVESOUP, a batch file devised byone person to move electronic-mail messages re-ceived by someone named Campbell to a subdirec-tory. Icons were used creatively in the OS/2 andMacintosh environments as decoration or some-times as puns for the directory contents or forthe action performed by an executable file. In theMacintosh environment which allows naming con-ventions to exceed the 8 to 11 character maximumofthe DOS world, it was common to find full andmore detailed titles.

Naming files after people was a convention em-ployed by all of those interviewed, and included sit-uations where the item was created for the person,produced by the person, was about the person, orconcerned an area which fell under the person's re-sponsibilities. One subject named his contract filesafter the last name of the contracts officer so theywould be grouped in alphabetic proximity of eachother for retrieval.

(3) By reordering things (DATE. USE, DOCUMENTATTRIBUTES, ORDER/SCHEME). Subjects useda variety of methods of grouping and arranging in-formation lo expedite retrieval. In some cases, they


used "arrange" functions within the software or op-erating system to order items by date or by title andthen refer to dimensions such as file si e or extensionto locate the desired item. The natural order orscheme provided by Microsoft Word is "last in, firstout" and the subject who worked in this environ-ment retrieved the things he used most frequentlyand most recently without using any other strategy.One subject made use ofthe WordPerfect "find" fea-ture to retrieve files of like extensions. Yet anothersubject prepared his own utilities and tools to helphim navigate through his extensive workspace.

Subjects were not observed using utilities such as"whereis" or the Macintosh "find" facility whichwere available within their environments. It is possi-ble that they were unaware these utilities exist, al-though they may have overlooked them at the time,or may have bypassed them intentionally, preferringthe ordering schemes described here to command-driven find features.

(4) Date (TIME, USE). Date was a dimension used byseveral subjects in retrieval. Usually, it served as adiscriminating factor when several candidate fileshad been identified as the possible target. However,there were instances where the files were displayed indate order to make it easy lor the subjects to findwhat they had been working on most recently. TheOS/2 environment offered the greatest flexibility, al-lowing the user to see the date the file was last usedas well as the date created.

(5) Browsing (USE. ORDER/SCHEME. DOCUMENTATTRIBUTES. ACCESS. CIRCUMSTANCE).Browsing was a strategy used frequently to retrievefiles. It was the observed method of choice amongthesubjectswhenlookingforold tiles, but it was usedjust as frequently to retrieve documents from a cur-rent list.

Browsing strategies varied and the activity servedseveral purposes. Browsing was used to reduce cog-nitive load by perusing file lists to facilitate docu-ment recognition. This required less effort than re-membering the file name and commanding the sys-tem to retrieve it. One subject entered her wordprocessor during the interview and immediately pro-duced a listing ofthe files to find a document she wasworking on that morning. sa\ ing aloud. "Now whatdid 1 call that file?" Browsing usually involvedmultiple dimensions, with date and file size often asimportant as the name in locating the desired item.

Most often the subjects used some kind of probeto enter the system for browsing. This probe was usu-ally a partial filename (prefix or suffix), but some-times it was a date or a location (a subdirectory ordiskette). Subjects also demonstrated instances ofbrowsing without a probe, broadly searching for theitem needed, perhaps knowing or not knowing anypan ofthe name, or even if what they were lookingfor would be there. This is the most difficult type ofbrowsing for specific data retrieval and it requires ahigh level ofconcentration.

(6) Hard copy aids (ACCESS, ORDER/SCHEME, CIR-CUMSTANCE). One subject worked in a depart-

ment which used a hard copy directory and an elab-orate logging system to facilitate retrieval from a vastarchive of information stored on diskette.

(7) Use (USE, PHYSICAL ATTRIBUTES. EORM).Use is a powerful dimension in retrieval from PIMsystems. Subjects in this study were observed oftensearching for files by accessing the software used tocreate them. These subjects relied upon the softwareto display the files from wherever the software storesthem by default.

Discussion

There are parallels between the results ofthe Kwasnikstudy and this investigation into how individuals orga-nize their electronic files. Many ofthe same types of doc-uments were identified, and many ofthe same dimen-sions were used in making classification decisions. Thereare some interesting differences, however. Kwasnik at-tempted to identify and count incidents of dimensionsused most often in making classificatory decisions, butthis study focuses upon the dimensions used in the rulesfor organizing electronic documents in PIM systems todetermine if similar rules apply.

The methodology used to analyze the data was helpfulin discerning patterns in organizing and retrieving infor-mation in an electronic environment. The anaiysis posedsome problems for the researcher, however. Often it wasdifficult to isolate the particular dimensions applied inclassification of a document, determining whether thedimension applied was an ACCUMULATION or anARRANGEMENT, for example. In these cases, thesummary category (ORDER/SCHEME, etc.) was usedinstead of a more specific one. It was not always easyor even possible to determine with confidence that thecategory assigned was the appropriate one, especiallywhen multiple dimensions were applied, as was fre-quently the case. Kwasnik reported similar problems,and computed a validity measure for ber results by hav-ing the data double-coded. A validity measure was com-puted for these interviews as well, based upon a sampleset of 35 observations. The coders agreed in identifica-tion of the dimensions 82% ofthe time. All ofthe obser-vations in the sample set except one involved more thanone dimension, and there was agreement on at least oneofthe dimensions in every case.

Another problem in analyzing the data was thedifficulty in determining how to define a document in theelectronic environment. Many ofthe files in the user'spersonal workspace were not documents in the tradi-tional sense, but were complex database systems or toolscreated by or used by the individual in accomplishingwork. These tools and utilities function sometimes as ref-erence tools, but often function as mechanical devicesto move files from one place to another, to open a newwindow, or to strip special characters from a file. Sincethese items occupy space in the environment and since


they can often be changed and enhanced by the user, thedecision was made to include them in the scope of thisanalysis, but a case can be made against this. In an office,a dictionary is a reference tool as well as a document,but a ruler and calculator are only tools. In an electronicenvironment, the distinction is less clear. A macro and abatch file may be tools that perform functions, but theyare also documents in the sense that they can be edited,stored, and retrieved in much the same way.

The interviewer's influence on the responses alsoraises questions. Although the interviewer attempted touse general terms (stuff, things, items) in referencing doc-uments to allow the subject the opportunity to define hisor her documents more specifically, this did not alwayswork. Analysis of the interviews reveals two instanceswhere the interviewer first used a more specific docu-ment label that was later picked up by the interviewee.Some of the managers interviewed were known person-ally to the interviewer, and as they were familiar with theinterviewer's interests and areas of responsibility, theremay have been some degree of posturing. The in-terviewer oversees LAN activities, and these people werecareful to state they performed maintenance on their di-rectories when asked to do so.

Despite these obvious problems, it was useful to ana-lyze the dimensions individuals employ in describing thedocuments in their electronic space. It is useful in under-standing how individuals work and in understandinghow PIM systems differ from general systems. Some ofthese distinctions are described below.

Acquisitions

The personal computing environment offers manyopportunities for individuals to expand both their per-sonal document collections and their access to externalsources. Unlike the physical offices in the Kwasnik study,the workspace of these subjects spilled over to sources farbeyond the desktop, but few were maximizing the capa-bilities of these systems. Additionally, there were files ineach individual's workspace that were unidentified bythem, or that were known but never used. Much of thisis evidence of satisficing behavior, of the individual's at-tention to traditional work-related tasks at the expenseof cognitive effort to gain proficiency with software {Car-roll, 1987; Marchionini, in press). Only one person withlinks to the Internet and to his local electronic bulletinboard regularly sought information there or saved itemsfound that looked interesting. Systems which invitebrowsing and active user exploration are needed to helpreduce the cognitive load required to seek new informa-tion and to acquire or access sources which will contrib-ute to effectiveness on the job. It should be emphasizedthat none of these users expressed a need for comprehen-sive collections as their motives for acquiring informa-tion was based upon need and relevance to their work.

Classification

Document attributes alone are insufficient to describeand organize items within the dynamic environment ofa PIM system. The conditions in which documents arecreated and used change, and these conditions impactthe way that individuals classify and organize informa-tion. Rules that are applied fora periodof time to reflectthe priorities of the moment may soon be abandoned orforgotten; however, this study shows some consistencyamong subjects in the dimensions applied to classifica-tion events.

In an electronic environment, the quantity and com-plexity of the work influence conventions used in classi-fying and storing information. Some subjects inter-viewed here used completely different rules for classify-ing documents based upon the level of granularityrequired to support the workload. An example of thisis the financial manager who organized budget data forcontracts by project name. One project was too large andtoo complex to be supported by this scheme. This projectwas divided into tasks which were organized either by thename of the manager responsible for the task or by thename of the task. The rule employed was, if the projectis too large to file by project name, then break the projectinto tasks, and if the task has only one manager, file bythe manager's name, otherwise, file by the task name.A similar case was observed for individuals who addedsuffixes to further subdivide topical information whentheir naming schemes left them without the flexibility toassign similar prefixes to all items relevant to a topic.This finer granularity isrequiredby the individual to dis-tinguish among the large volumes of data on the subject,and he or she may extend the classification, or reclassify,by such dimensions as a person's name, a date, or sub-topic. While this is essential to the individual, it presentsproblems for colleagues or successors who must interpretthe scheme to retrieve something from the space. Addi-tionally, the individual must rely upon personal knowl-edge of the growth and development of a project or taskto trigger hisor her memory of former schemes.

In a physical office, folders can be tagged with the proj-ect name and with subheadings, and items can be storedin physical proximity. Similar features are available inan electronic environment, but the subjects in this studydid not typically store things this way. The subjects inthis study consistently used satisficing strategies in organ-izing the information within their workspace. Few hadever created subdirectories. Only the two users whosework was computer-related and the Macintosh usershowed knowledge of directory structures to organizedata. One subject continued to put files on diskette forstorage and for quick access, although he had ample freespace on a hard drive (100 of 120 megabytes unused).Most subjects used software defaults to organize and dis-play information and to browse data for retrieval. Thisbehavior is consistent with the high volume, varied, and


temporal nature of work performed by managers. Manydocuments are created quickly and with little concernfor naming or organizing the information, includingmemos to staff, on-demand status reports, special re-quests, and other small projects that are the purview ofmanagement.

Maintenance

The subjects in this study performed maintenanceupon the system when they were required to by theirLAN managers, when they faced storage limitations, orwhen moving or removing files was a conscious effort tofacilitate retrieval of the information later. Most of thesubjects had some awareness of their options for storinginformation, but most had only limited knowledge ofhow to maintain their files or even considered that regu-lar maintenance should be part of their routine. PIM sys-tem maintenance seems to be guided by a philosophyof "if it ain't broke, don't fix it," rather than regularlyscheduled maintenance to avoid crises.

There was evidence that at least one user experienceddifficuhy in modeling the system, or difficuhy in accept-ing that things stored in a place that was beyond the per-son's reach and grasp were safe. She felt more comfort-able having files stored on diskettes and put in a box onher desk than trusting security to a LAN server that wasdistant and unseen.

One thing apparent from this study is that, as is truewith hard copy document collections and most informa-tion systems, collection size is not static and the needfor additional storage grows. The two individuals whoalready had substantial hard drives were expecting addi-tional drives to be installed soon. Individuals who storeditems on diskette sometimes reorganized files by formatwhen the need for storage exceeded the one disk and asecond was required. These collections can be expectedto grow at least until budgets can no longer support pur-chase of more diskettes, hard drives, file servers, andother storage media, or until the users can no longermanage the information.

Retrieval

The people in this study retrieved information basedupon a variety of dimensions. The strategy observedmost often was browsing, usually initiated with a probe.When the software made this easy, individuals were ob-served doing this even when the document was some-thing they had used recently or often. Although browsinginvolved additional steps, it required less effort than re-membering the file name and the command to retrieveit. The use of browsing techniques suggests that multiple,situational dimensions may be used to retrieve informa-tion as well as to organize and store it, as browsing pro-vides clues to age, context, format, topic, and other di-mensions. Browsing provides the secondary benefit of al-

lowing the individual to get a general overview of what isin the system. This is particularly important in a PIMsystem since it is an overview of their personal work-space.

Retrieval tools made available by the software or cre-ated by individuals were frequently employed to retrieveinformation. Examples of such tools include batch filesto navigate through subdirectories, the find feature inWordPerfect, the arrange function available within soft-ware or operating environments which supports sortingby various characteristics, and software defaults that or-der the documents in some way. Few individuals indi-cated having trouble finding things, but the interviewerobserved one subject who was unable to find a particularfile he was seeking during the session.

Items were sometimes organized in prominent posi-tions in the personal information space to serve as re-minders of things that need to be done. This was foundto be true among all the subjects who used electronic-mail.

Conclusion

Classification behavior in a PIM system is a dynamicprocess characterized by satisficing strategies in selectingand organizing information, by browsing strategies forretrieval, and by the personal and situational factors sur-rounding the classification event. Classification behaviorin PIM systems raises questions about individual pro-ductivity, organizational efficiency, and the quality ofsoftware and hardware used to support work.

Individuals have as much flexibility and autonomywith PIM systems as the hardware/software environ-ment supports and their organizations allow. Individualscan organize their information as they choose, selectnaming conventions for their files, and design menusand desktops to meet their needs. Personal productivity,creativity, and job satisfaction are enhanced by these fea-tures, and organizations benefit when workers are pro-ductive and successful. However, the individualismwhich characterizes PIM systems may also pose prob-lems for organizations. Organizations require some con-formity in the electronic environment to support collab-orative work, to reduce duplication of effort, to facilitatetransfer of tasks from one person to another, and to en-able persons to retrieve needed information in the ab-sence of the individual responsible for a task.

This investigation suggests that electronic documentsare not well ordered in a PIM system. There is evidencethat items are stored often by system default on the avail-able media with reliance upon software features for sort-ing and reordering information "on the fly." Addition-ally, the observed behaviors suggest that some individu-als employ organizational techniques similar to thoseused with paper documents and they have not adjustedtheir thinking to take advantage of features unique to theelectronic environment. Marchionini found similar ten-


dencies among persons who were learning to use an elec-tronic encyclopedia based upon the printed model (Mar-chionini. 1989). The physical clues which aid in storingand retrieving items in an office are replaced with verydifferent personal and situational clues in an electronicsystem.

Individuals are inventive, creative, and flexible, andthere is evidence that they are using software successfullyto perform the work they need to do. What they are notdoing well is exploring the full range of capabilitieswithin their environments or maximizing the availablestorage. The challenge for software developers is formore powerful tools to aid the user in organizing, main-taining, and retrieving information, and for intuitive in-terfaces which improve the learnability and usability ofthese tools. It has been demonstrated often that expertusers prefer to bypass menus for the power of commandlanguages, but evidence here suggests that browsing strat-egies, even among experts, are preferred for retrievaltasks. This is consistent with the dynamic and situationalnature of the work environment.

Future research should look into the similarities anddifferences between the PIM systems within an organiza-tion to determine the how these systems impact organi-zational efficiency. Additionally, a better understandingof the relationship between classification behavior andinformation-seeking behavior within a PIM environ-ment over time is needed to understand the strategiesthat are most effective and to design systems that cansupport the way individuals work in these environments.Finally, it would be helpful to know if there are limits tothe amount of information that can be managedeffectively, whether limits may vary by individual, bysystem, and by situation, and if there are limits, how ac-cess to virtually unlimited electronic informationthrough national and international networks will impactpersonal performance and satisfaction in a work envi-ronment.

Acknowledgments

The author thanks Dr. Gary Marchionini for his en-couragement and guidance in developing the topic andfor his advice on improving the manuscript. The authoris indebted to Danuta Nitecki and Emily Marsh who as-sisted in the coding of data, and to the managers who

were generous enough to give their time and thoughtfulresponses to questions.

ReferencesBruner. J. S. (1958). Social psychology and perception. In E. E. Mac-

coby. T. M. Newcomb. & E. L, Hartley (Eds.). Readings in sodafpsychohgyOTd ed.) (pp. 85-94). New York: Henry Holt and Co,

Burton. H. D. (1987). FAMULUS revisited: Ten years of personal in-formation systems. Journal of the American StKiclyfor InfonnaliouScience, S2. 440-443.

Carroll. J. M. (1982). Creative names for personal files in an interactivecomputing environment. Inlernalional Journal of Man-MachineStudies. 16. 405-438.

Carroll. J. M. (1987). The Paradox of the active user. In J. M. Carroll(Ed.). InU'rfacin^lh<mf>hl {pp. 80-111). Cambridge. MA: MIT Press.

Englebart, D. C. (1961). Special consideration of the user as a user,generator, and retriever of information. Anwriean DiKumeniation.121-125.

Ericcson. K. A,. & Staszewski. J. J, (1989). Skilled memory and ex-pertise: mechanisms of exceptional performance. In D. KJahr & K.Kotovsky (Eds.). Complex information processing: The impact ofHerbert A. Simon. Hillsdaie. NJ: Erlbaum.

Kwasnik, B. H. (1989a, June). How a personal document's intendeduse or purpose affects its classihcation in an office. In Proceedings ofthe .KWf-SlGIR I2tli .Annual International Conference on Researchand Development in Information Retrieval. Cambridge. MA. NewYork: Association for Computing Machinery.

Kwasnik. B. H. (1991), The importance of factors that are not docu-ment attributes in the organization of personal documents. Journalof Documentation. 47. 389-398.

Kwasnik. B. H. (1989b). The injUience of context on classifactory be-havior. Doctoral thesis. Rutgers University.

Kwasnik. B. H. (1992a). The naming of personal documents: The vo-cabulary problem revisited, (unpublished paper.)

Kwasnik. B. H. (1992b). The role of classification structures in reflect-ing and building theory. In Proceedings of the 3rd .ISIS SIG/CRClas.silnation Research Workshop (.'i5th ASIS annual meeting).Pittsbui^h, PA. Silver Spring. MD: Printed and distributed by Amer-ican Society for Information Science Special Interest Group/Classi-fication Research.

Malone. T. W. (1983). How do people organize their desks? Implica-tions for the design of office information systems. ACM Transactionson Office Information Systems. 10. 99-1 12.

Marchionini, G. (in press). Information seeking in electronic environ-ments. New York: Cambridge University Press.

Marchionini. G. (1989). Information-seeking strategies of novices us-ing a full-text electronic encyclopedia. Journal of the American StK'i-etyfor Information Science. 40. 54-66.

Medin, D. L.. & Schaffer. M. M. (1978). Context theory of classificationlearning. Psychological Review. 85. 207-238.

Nitecki, D- Developing positive impact personal information systems.Unpublished manuscript.

Soerget, D. (1985). Organizing information: Principles of data base andretrieval systems. Orlando, FL: Academic Press.

Walker. W. H..& Kintsch. W. (1985). Automatic and strategic aspectsof knowledge retrieval. Cognitive Science, 9, 261-283.
