This bibliography cites the works which were consultedfor preparing a term paper. The bibliography served to demonstrate the direction which the paper was to take, and sources wereadded as research progressed. (Added sources are not included in this bibliography. Please see the paper for citations.)

History of visual resources libraries

Roberts, H.E. (1978, Winter). The image library. Art Libraries Journal,3 (4), 25-32.

Helene Roberts' paper described the traditional method of organization in imagelibraries for art history, an arranging of images by categories of medium,period, culture or country, school or artist, in that descending order. (26)While this method of organization served art historians well because itmirrored the categorization that is typical of the study paradigm of arthistory students, it did not serve researchers who, at the time, wereincreasingly searching for visual materials by different categories. Sheindicated that the alternative approach would be to additionally organizeimages by sign, symbol and meaning the use of indexes. This alternativeapproach referred to the three levels of meaning long ago defined by ErwinPanofsky. Roberts' article defines the image organization and thereforesearching paradigm that has dominated visual arts image collections to thisday.

Sunderland, J. (1982, Summer). Image collections: Librarians, users andtheir needs. Art Libraries Journal, 7 (2), 41-49.

Sunderland raises the issue of feasibility in indexing an image collection. Hecites examples of slide libraries which have esoteric classification schemeswhich suit their users, but notes that many of these institutions have not comeclose to indexing the entirety of their collections. As Bradfield stated in1976, the main method of access in image collections is classification byphysical arrangements and browsing. Sunderland's survey suggests that thetraditional method of image organization -- recording into a database or printcatalogue empirical aspects of works, such as artist, title, physicaldescription and broad classification -- is adequate for image provision in amuseum. He states that if more information of this nature can be entered, themore valuable the information system will be. In specialized collections suchas museums, he is unsure whether indexing deeply is economically feasible,given that traditional methods of image searching and image use do not relyheavily on indexes.

Theories of image organization and methods of imageprovision

Barnett, P.J. (1988). An art information system: From integration tointerpretation. Library Trends, 37 (2), 194-205.

Barnett discusses the dependence of traditional image retrieval on the mannerin which attributes of images are pre-coordinately recorded. She describes forus the process of creating a record for an art image as the process of creatinga bibliographic description and a knowledge description. Barnett states that tolink the art object to the image and thus the searcher to the art object, onesearches for either the object description or the subject description. InBarnett's terms, "...the object description enumerates and the bibliographiccontent description abstracts". (201) The enumeration, she argues, occurs on anempirical level listing intrinsic elements such as author and title, analogousto the levels of book description employed in cataloging. The knowledgedescription abstracts the meaning of the image that is extrinsic to the object,such as the scholarly opinions of iconographic meaning, analogous to subjectheading classification in book cataloging. Barnett's paper suggests to me thatfor scholarly research, access to visual images is best facilitated bysearching for established and controlled terms pre-coordinately provided byindexers steeped in the standard modes of art object description.

Markey, K. (1988). Access to iconographical research collections.Library Trends, 37 (2), 154-174.

Markey discusses access to visual resources subject matter examiningiconographic research collections which utilize subject headings orclassification schemes. Her paper describes the various levels of subjectmatter access relating to pictorial images and explains why information systemstraditionally have provided subject access only to "secondary subject matter",iconographical and iconological identifications, and not to "primary subjectmatter", pre-iconographical identifications regarding "ofness" descriptions anddescription of attributes such as line and color. Markey observes that thesole provision of secondary subject matter content for image access has placedthe "burden of access" on users and has created barriers to images for userswho are insufficiently steeped in the systems' classification or subjectheading schemes. She believes that "providing access to collections by primarysubject matter can ease the task of users since it requires only one level ofinterpretation, practical experience, and knowledge of the history of style."(168)

O'Connor, B.C. (1985, December). Access to moving image documents:Background concepts and proposals for surrogates for film and video works.Journal of Documentation, 41 (4), 209-220.

O'Connor believes that "there are three major sets of internal representationwith which a user might approach a document collection, 1) Well-articulated,fits system vocabulary, 2) Vague awareness of lack, index vocabulary notadequate, and 3) Monitoring of the information environment, shake up personalknowledge store". (211) Indeed, Kuhlthau (1991) has enumerated severaldescriptions of information seeking behavior proposed by Kelly, Taylor, Belkin,and herself. O'Connor has proposed that a different model for providingbook-like access to moving image documents (MIDs) is necessary to supportbrowsing and to allow users to make personal judgments of relevance based onthe surrogate, which may stand in the full MIDs place in the IR system.O'Connor considers the elusiveness of the so-called "text" of shots andsequences in MID and believes that still images, words, and mathematicalexpressions may be adequate to surrogate a MID. He proposes the use of asecondary surrogate which enumerates the objects recorded and the depictionvariables, that is, describing the physical document and the topic. Thesurrogate model serves to place a distance between user and object (MIDs) inorder to afford scrutiny of the surrogate and consequently to link the user tothe object. O'Connor's proposal mirrors the attempts in art image indexing toprovide secondary access to images because there is no otherwise reliable wayto provide access to images. As such, these methods may serve as singleretrieval tools but their reliability cannot be adequate.

Shatford, S. (1986, Spring). Analyzing the subject of a picture: Atheoretical approach. Cataloging & Classification Quarterly, 6 (3),39-62.

In the visual resources field Sara Shatford is usually identified as thelibrarian who explained most succinctly what Erwin Panofsky meant by the "threelevels of meaning in an art work". Panofsky's theory is the basis for ourtraditional methods of image description in a library setting. Shatfordrecognized in her paper that there were very few user studies in the area ofpicture use and called for theories which allow image management workers toidentify more clearly the methods for facilitating image access. Herinfluential paper is typically cited as the driving force behind many currentefforts to provide subject access in art historical image collectionstoday.

Image Use studies

Bradfield, V.J. (1976, Autumn). Slides and their users: Thoughts following asurvey of slide collections in Britain. Art Library Journal, 2 (3),4-21.

Bradfield's paper is one of the earliest surveys of slide collection use, whichhas called for different ways of looking at image retrieval systems. Bradfieldbelieved that observation of slide collection use was a valuable method fordiscovering information seeking behavior in slide libraries. She identifiedissues that contributed to success in satisfying users' needs and lesseningfrustration: 1) browsability in the collection, 2) comfort of the facility , 3)an atmosphere conducive to searching and to consulting librarians, and 4)having considerable time available for searching. The most significant reasonsfor failure in retrieval dealt with browsability and time. She believed that"browsing is vital" because browsing presents alternatives in the ideagenerating process, and that "serendipity is valuable." Most users identifiedthat a failure to retrieve materials was often due to a "lack of time tobrowse." (9)

Cawkell, A.E. (1993). Picture-queries and picture databases. Journalof Information Science, 19 (6), 409-203.

Cawkell reviews query by image technology, a new paradigm by which users posequestions about color, shape, and texture, and by which users may even submit aquery by drawing a sketch. Cawkell's research is significant because itillustrates that the addition of pre-iconographic elements of images may now beused for image access. Cawkell suggests that such computerized access methodsare reasonable because there is little published literature regarding the kindsof questions asked and the success of librarians or information systems inresponding to those questions. He also believes that the field has not donesufficient failure analysis on searches of image collections.

Enser, P.G.B. (1993). Query Analysis in a Visual Information RetrievalContext. Journal of Documentation & Text Management, 1 (1),25-52.

Enser looks into question analysis in his work with the Hulton picturecollection. Enser's research supports Bradfield's (1976) early theory thatsearch time and expertise factors relate significantly to success in using apicture system.

Jörgenson, C. (1995). Image Attributes: An investigation. Ph. D. ,Syracuse University.

Jörgenson's dissertation research investigated human pictorial imageperception and description, and developed a template of typical attributeswhich humans use to describe images, such as objects, people, etc.Jörgenson based her attributes on the terms which participants typicallyused to describe images in a variety of tasks.

Jörgenson, C. (1996). Indexing images: Testing an imagedescription template. ASIS 1996 Annual Conference Proceedings, October19-24, 1996 [On-line]. Available:http://www.asis.org/annual-96/ElectronicProceedings/jorgenson.html

Jörgenson's follow-up to the research of her dissertation tests thetemplate of image attribute classes she identified in the dissertation. Sheidentifies that her participants (naive image users) described images in threecategories of description, perceptual, interpretive, and reactive. The studywas concerned with investigating whether a template for image description wouldbe useful to participants in framing their image descriptions. The findings ofthis study indicated that users may have difficulty when using a template inassigning descriptors to higher-level classes based upon conceptual orfunctional relationships. She suggests in reaction to the results that thesolution to this difficulty may lie in user training or guidance.

Jörgenson, C. (1997, November). How people describe images:Continuing research. Paper presented at the meeting of the American Societyof Information Science, Washington, DC.

In her most recent research, Jörgenson asks participants to describeimages within a template of 47 attributes (developed in her 1995 study) whichshe has identified as typically characteristic of an image. Her recent datasuggests that the terms chosen after being introduced to a template aresignificantly different from the terms they may have typically and naturallychosen without use of the template. This suggests to me that users are lesslikely to describe images as they naturally perceive them (using abstractconcepts such as "mysterious", for example) when confronted with an imageretrieval system, because users are aware of the shortcomings of informationsystems in dealing with such descriptions. Jörgenson's research sets aprecedent for considerations in designing or radically changing image accesssystems. It illustrates the need for providing image access in a manner thatsatisfies a variety of user needs.

O'Connor, B.C. (1996). Aboutness and user-generated descriptors.
Explorations in indexing and abstracting: Pointing virtue, and power.Englewood, CO: Libraries Unlimited.

O'Connor's chapter on aboutness and user-generated descriptors looks at twocase studies involving uses of photographs, one based on lantern-slide imagesand the other based on PhotoCD digital images. He deals with the question ofaboutness using pictures because it is difficult to describe what an image isabout. The studies O'Connor cites in the article illustrate the inconsistencythat occurs when humans describe what images are about. While therepresentation of image content is historically handled by external agencies,O'Connor believes that placing the power of representation in the user's handswill serve to adequately improve image retrieval. User-generated descriptors,he believes will cluster around certain images, and retrieval of images withindescriptor clusters may be retrieved in ranked order through computermanipulation. This proposal serves to increase our "aboutness" access to imagesand represents a radical solution that has unfortunately not been considered asyet in the contemporary literature.

Ornager, S. (1997, November). Image retrieval: Theoretical analysis anduser studies on accessing information in images. Proceedings of the 60thAnnual Meeting of the American Society for Information Science, 34,202-211.

Ornager's research empirically observed the use of digital image collectionsin newspaper archives to determine the kind of questions users ask of thearchive, and to group into categories the kinds of users of the archive.Typical observation activities included word association testing as a methodfor observing image descriptions. The intention of Ornager's research is todefine an operational subject indexing strategy for images. She approached theidea of subject content from the perspectives of Erwin Panofsky and SaraShatford, who both discuss the levels of "of" and "about" meaning related toart objects, and from the perspective of Roland Barthes who approaches text asa process of identifying signs and feelings associated with such things astextual documents and images. Ornager believes, following this analysis of userdescription of images, that indexing must encompass factual description(ofness), "expressional" content description (aboutness), and indication of thecontext in which the image can be used. The most significant conclusion shemakes is that user queries follow a pattern that correlates to their placementwithin a user typology. This sort of idea has been suggested by Robert Taylor's(1991) research related to information use environments, specifically dealingwith information seeking in a hospital. The theory suggests that sense-makingwithin certain contexts may directly relate to the behaviors associated withseeking information. While Ornager is concerned with enhancing user interfaceto deal with aspects of the information querying system, I believe the moresignificant implication of her research deals with the observation of patternsof use and the possibilities of enhancing systems using vocabulary control andclustering.

Turner, J.M. (1994). Indexing "ordinary" pictures for storage andretrieval. Visual Resources, X, 265-273.

Turner's user study gathered empirical data on spontaneous image description tosuggest directions for providing image access in information retrieval systems.His study sought to confirm whether participants distinguished betweenpre-iconographical (ofness) and iconographical (aboutness) content in images.As the data relates to "ordinary pictures" Turner found that pre-iconographicalindexing is necessary, but iconographical indexing of ordinary pictures isquestionable. He believes that in the case of ordinary pictures it isreasonable to place the onus of aboutness description on users. In addition hebelieves that a general classification of ordinary objects is necessary in animage retrieval system, an ofness classification which describes what istypically spontaneously observable by a human subject.

Turner, J.M. (1995). Comparing user-assigned terms withindexer-assigned terms for storage and retrieval of moving images: Researchresults. [On-line]. Available: http://www.asis.org/asis-95/papers/turner.html

Turner's study compares the most popular terms assigned to moving imagedocuments by users to the terms assigned by indexers using the National FilmBoard of Canada's stockshot collection. His study shows that there was a highdegree of agreement between the two groups. Turner believed that the findingssuggested that pre-iconographic (ofness) level indexing in addition toiconographic (aboutness) indexing would help improve retrieval rates in thiscollection.

Turner, J.M. (1997, November). Explorations in using audiodescription as a tool for indexing moving image documents. Paper presentedat the meeting of the American Society of Information Science, Washington,DC.

Turner's research observes natural language descriptions of moving imagedocuments in shot by shot analysis. His latest study expands this type ofobservation by observing and indexing voice-over audio descriptions (the audioanalog for sight-impaired to closed-captioning for the hearing impaired) ofmoving image documents. These words are transcribed into textual documents thatare indexed in the same manner as print documents. Turner's previous researchexplored issues of "ofness" and "aboutness" of images, making recommendationsthat were closely dependent on the type of image being accessed and the typesof image use. The present study attempts to continue observing imagedescription primarily at the ofness level, and it remains to be seen how hewill interpret the data, and whether he will find any value in the use ofofness description in this study.

Future research

Holt, B. & Weiss, K. (1997, November). The QBIC project in the Departmentof Art and Art History at UC Davis. Proceedings of the 60th Annual Meetingof the American Society for Information Science, 34, 189-195.

Holt's and Weiss' paper on the QBIC project illustrate the usefulness ofproviding pre-iconographical access to images, which Markey identified in 1988as the key to removing barriers to access in traditional image access systems.Holt and Weiss discuss in this paper their observations opinions of this newmethod of image access, which allows users to query image indexing systems byselecting attributes of color, shape and texture, or by electronic sketching ofoutlines. Users may then iteratively process the search by searching for imageswhich are similar to an image selected from the result set. The authors notedthat QBIC-style searching cannot replace text systems for thematic searches,but they do offer added means for image access, which helps to possibly reducethe failure rate of image retrieval. The project began after a user survey wasperformed; the user survey is not publicly available, however.

15 December, 1997