Behavior Research Methods, Instruments, & Computers 1999,3/ (3), 479-486

ESPbase: A Microsoft Access tool for selecting symbol and sets for usability

OSCAR DE BRUIJN and SINE McDOUGALL University ojWalesSwansea, Swansea, Wales and MARTIN B. CURRY Sowerby Research Centre, British Aerospace, Bristol, England

The ESPbase provides a tool for storing symbols and icons along with information about their char­ acteristics. Information about a wide range of symbol characteristics is included on the database to facilitate the selection of symbol sets for research and design. The database includes information about the graphical characteristics and functions of symbols. Italso includes ratings of symbol concreteness, complexity, familiarity, and meaningfulness. Symbols and icons can be accessed on the basis of each of these characteristics or any combination of characteristics. This makes it easier to select symbols on the basis of usability and design requirements. It also means that symbols can be easily selected for research while controlling their characteristics on a number of dimensions.

A range ofnormative data are available for words (see, There is now a growing body of research examining e.g., Benjafield, Frommhold, Keenan, Muckenheim, & what characteristics ofsymbols affect their usability, and Mueller, 1993; Friendly, FrankIin, Hoffman, & Rubin, it appears that control over these characteristics may be 1982; Gilhooly & Logie, 1980; Paivio, Yuille, & Madigan, as important to symbol usability research as control over 1968) and pictures (Martein, 1995; Sanfeliu & Fernandez, word and picture characteristics is to psycholinguistic re­ 1996; Snodgrass & Vanderwart, 1980; van Schagen, search. For example, it is often assumed that symbols, and Tamsma, Bruggemann, Jackson, & Michon, 1983) in order particularly computer icons, are more easily processed to control their characteristics when such items are used when they are concrete than when they are abstract. Use as experimental stimuli in psycholinguistic research. Norms of concrete symbols is supposed to be easier because available for words include their concreteness, familiarity, they depict objects, places, and people with which we are and meaningfulness, while those available for pictures in­ already familiar in the real world, making their meanings clude their familiarity, meaningfulness, and name agree­ more obvious than those of abstract symbols. However, ment. These characteristics of word and picture stimuli despite the large number ofstudies that have investigated often need to be controlled in order to avoid confounding the effects of concrete versus abstract symbols on us­ them with the characteristic(s) under investigation. Re­ ability (Arend, Muthig, & Wandmacher, 1987; Green & cently, normative ratings were collected for the concrete­ Barnard, 1990; Rogers, 1986; Rogers & Oborne, 1987; ness, complexity, familiarity, meaningfulness, semantic Rohr & Keppel, 1985; Stammers, George, & Carey, 1989; distance, name agreement, and concept agreement of a Stammers & Hoffman, 1991), the precise circumstances large number ofsymbols I (McDougall, Curry, & de Bruijn, under which concrete symbols are more usable than ab­ 1999). These norms have been included in the database stract symbols remains unclear (but see McDougall, Curry and can be used to control the characteristics ofsymbols & de Bruijn, 1998). Part of the inconsistency concerning used in symbol usability research. Inaddition, ratings of the effects of concreteness on usability found in these the semantic, syntactic, and pragmatic usability ofa large studies may be due to a lack ofcontrol over the character­ number of symbols (American Institute of Graphie Arts istics ofthe symbols used in these studies other than their [AIGA], 1982) have been included in the database and can concreteness. For example, in some ofthis research, the be used when one is considering the usability and design concreteness ofsymbols may have been confounded with of symbol sets. their complexity (Garcia, Badre, & Stasko, 1994). In order to examine the effects of concreteness on usability, it is important to be able to select symbols on the basis oftheir concreteness while controlling for their complexity. In­ deed, it may be generally important to control the char­ This research was part of the Effective Symbology Project, which acteristics of symbols used in research in order to avoid was supported by a grant from British Aerospace pie (Grant SRC/UOS/ 060495). Correspondence should be addressed to S. McDougall, Psy­ confounding their effects on usability. chology Department, University ofWales Swansea, Swansea SA2 8PP, Controlling symbol characteristics is also important Wales (e-mail: [email protected]). whenever there is a need to select symbols to be used in

479 Copyright 1999 Psychonomic Society, Inc. 480 DE BRUlJN, McDOUGALL, AND CURRY new applications (e.g., Barnard & Marcel, 1984; Dyson, linguistic Database (Quinlan, 1992) provides a computer­ 1992; Gittens, 1986). As symbol usability research pro­ ized way of selecting word sets, taking into account the vides links between an ever-growing number of symbol syntactic, semantic, phonological, and orthographic infor­ characteristics and their usability, it may be possible to mation ofthe words. Available in this database are prop­ predict the usability of symbols relative to a number of erties of words such as their number of syllabies, their alternatives on the basis of their characteristics. Thus, regularity or irregularity ofspelling, and their frequency whenever there is reason to believe that symbols with of occurrence in printed English. The Oxford Psycho­ certain characteristics can increase the usability ofan ap­ linguistic Database also includes the norms for the con­ plication relative to other symbols, it may be desirable to creteness, familiarity, and meaningfulness, which are select symbols possessing these characteristics. For ex­ available for a large number of words. Each of these ample, it has been argued that usability ofsymbols with characteristics can be used to select words in the database similar meanings is enhanced by the use of common simply by specifying criteria to which these words need graphical elements such as color (e.g., Christ, 1984) and to comply. For example, the inclusion of information shape (e.g., Gittens, 1986). On the other hand, there is about word type, frequency ofoccurrence, and concrete­ some evidence to suggest that symbol usability is en­ ness ofwords stored in the database allows all concrete hanced by the use ofunique graphical elements that make nouns with afrequency ot occurrence 0/500per million symbols more discriminable (e.g., Blackwell & Cuomo, or more to be selected from the list of words for which 1991; Geiselman, Landee, & Christen, 1982). Ifthe cir­ this information is specified. cumstances under which usability is enhanced by com­ The need to select symbols with certain characteristics mon or unique graphical elements are known, this knowl­ for both research and design purposes, together with the edge could be used to select symbol sets on the basis of recent development ofnorrns for a large number ofsym­ their graphical characteristics to best fit the circumstances bols, prompted the development of the Effective Sym­ in which they will be used. bology Project Symbol Database (ESPbase), which offers Given the large number ofcharacteristics that may need flexibility in the construction of symbol sets similar to to be considered, the task ofselecting symbols for sym­ that offered by the Oxford Psycholinguistic Database for bol usability research or for designing new applications the construction ofword sets. However, unlike the Oxford would be greatly facilitated by a computerized database Psycholinguistic Database, which allows only data re­ in which symbols are stored together with information trieval, the ESPbase provides facilities for storing new about a large number oftheir characteristics. Symbol sets symbols together with a large number oftheir character­ can then be accessed by simply specifying their desired istics. Information about each ofthese characteristics can characteristics. However, despite the availability ofa large be used to access these symbols at a later date. Thus, in ad­ number of collections of symbols accessible by com­ dition to providing easy access to symbols already stored puter, particularly on the World-Wide Web (e.g., Crystal­ on the database, the ESPbase provides a way ofstructur­ Vision, 1994; Graphie Artists Guild Foundation, 1995; ing new symbol data in such a way that this information Gray, 1990; Horton, 1994; The Icon Factory, 1997), none can be used in the construction of symbol sets for research ofthese provide facilities for the retrieval ofsymbols on and design purposes. the basis oftheir characteristics. Even if some structure The ESPbase is implemented as a Microsoft Access is apparent in the way symbols are stored in these col­ application running in a Microsoft Windows environ­ lections, this structure is usually based on just a single ment.2 This provides several advantages in terms ofease symbol characteristic, such as, for example, symbol size of interacting with the database and flexibility in cus­ (e.g., 16 X 6 pixels), file type (e.g., tiff,jpeg, or bitmap), tomizing the database to suit the users' needs. Microsoft function (e.g., "delete," "save"), or application (e.g., Mi­ Windows provides an easy way of interacting with the crosoft Windows 95, Apple Macintosh). A collection in database for anyone with a basic knowledge ofthe Micro­ which symbols are stored according to a single symbol soft Windows . It utilizes differ­ characteristic allows symbols to be retrieved in only one ent windows to display symbol information, to specify way, whereas for symbol usability research and design search criteria, and to display the results of symbol purposes, it is important that symbols can be retrieved searches. Navigation between these windows occurs us­ depending on many different characteristics. Thus, any ing bars, , and buttons that are inherent to computerized database that allows symbols to be accessed the Microsoft Windows graphical interface. Using Micro­ on the basis of many different characteristics will in­ soft Access as a platform for the ESPbase also has the evitably be more useful for research and design purposes advantage that all its functionality can be easily custom­ than one that allows symbols to be selected on the basis ized by anyone with a knowledge ofbuilding Microsoft ofonly a single characteristic. Access applications. Computerized retrieval ofwords on the basis ofa large In the next section, a description ofthe different sym­ number of their characteristics has been available for bol characteristics incorporated in the database is given psycholinguistic research for several years (see Coltheart, together with the reasons why these particular character­ 1981; Quinlan, 1992; Wilson, 1988). The Oxford Psycho- istics were included. This is followed by a description of SELECTING SYMBOL SETS 481 the different facilities incorporated in the database that than their graphical characteristics, which are described allow sets of symbols conforming to a number ofcrite­ below. With this information, symbols can be retrieved ria to be selected. from the database by specifying one or more keywords that describe the semantic contents of the symbol. For NATURE OF TUE INFORMATION example, a symbol that is described as "a silhouette ofa IN TUE ESPbase man viewed from the front" can be retrieved by using the keyword man. Information in the ESPbase is displayed in the Data Image characteristics. Given a large collection of Form, shown in Figure I. Because symbols are different graphic images, it is often desirable to retrieve these im­ from words, characteristics that allow effective symbol ages on the basis oftheir graphical characteristics. This search are not necessarily the same as the word charac­ could be achieved by using, for example, mathematical teristics included in the Oxford Psycholinguistic Data­ image mapping routines (e.g., Liu & Sun, 1997; Mehtre, base. Information in the Data Form is divided into three Kankanhalli, & Lee, 1997)or neural networks(e.g.,Mazza, sections: (I) symbol information (left section in Figure I), 1997), but we opted for providing a list ofgraphical fea­ (2) function information (upper-right section), and (3) tures (Figure 2) that can be used to characterize symbols ratings information (lower-right). graphically. Dyson (1992) showed that most symbols can Precise details regarding the information contained in be categorized in accordance with a list ofgraphical fea­ each section are given below. tures that the symbols either do or do not have. This means that it is possible to characterize a symbol by simply Symbol Information checking its features off on a list and using this charac­ The Symbol Information section ofthe Data Form con­ terization to retrieve symbols from the database. Thus tains a graphical image ofthe symbol along with all the symbols can be selected on the basis of individual features information directly concerning this image and its visual or sets offeatures. Furthermore, it is possible to avoid cre­ characteristics. ating symbol sets that share too many similar features Symbol ID number. Each symbol is indexed by a and therefore could become confusable. Symbol ID number, which provides a unique reference Achecklist with graphical features is available for each to each ofthe symbols. All other information in the data­ symbol in the Image Characteristics Box (Figure 2), which base is linked to the Symbol ID numbers. Thus, the Sym­ pops up when its is clicked in the Data Form. bol ID number works as a key that links together all the Graphical features are divided into four types: (1) colors, separate pieces ofinformation concerning one particular which are separated into foreground and background symbol. colors; (2) global features, which are those characteristics Source. Symbols are drawn from a variety of different concerning the entire image, such as, for example, sym­ sources. These include standards organizations (e.g., the metry and line drawing; (3) local features, which are the International Standards Organisation [ISO], the Interna­ graphical elements from which an image is constructed, tional Electrotechnical Committee [lEe], and the British such as, for example, lines, circles, and boxes; (4) special Standards Institute {BSI]), the AIGA, computer soft­ features, each of which can be considered as carrying a ware packages (e.g., Microsoft Access, Microsoft Word, certain meaning that is derived from the conventions that PaintShop Pro, etc.), Horton (1994), Dreyfuss (1971 ), and exist for their use, such as, for example, arrows, which a variety ofwebsites (e.g., Graphie Artists Guild Founda­ are often used to indicate direction ofmovement. tion, 1995; The Icon Factory, 1997). Source information allows symbols to be selected from particular sources. Function Information This may be useful when symbols are needed for use in Unlike words, symbols are used to represent functions applications for which symbols have already been devel­ that are not necessarily directly related to whatever is de­ oped or if the symbols have to be part of (international) picted in the symbol. Indeed, in the evaluation ofthe fac­ standards. tors underlying symbol usability, a distinction is often Symbol image. Symbols are central to the database made between a symbol's primary referent (what is de­ and are stored as graphical images. Symbols are saved as picted in the symbol), its secondary referent (its func­ bitmap image files that are then embedded in the symbol tion), and the closeness of the relationship between the database, from which they can be subsequently retrieved.3 symbol and its function (e.g., Familant & Detweiler, 1993). This method of storing symbol images in the database Thus, a characterization ofsymbols is not complete with­ ensures that images can be edited (using graphics soft­ out reference to their functions. ware such as Microsoft Paintbrush or PaintShop Pro), cop­ Information related to functions includes a unique ref­ ied to other applications (using Windows Clipboard func­ erence number for each function (the Function ID num­ tion), or saved in separate files using different file formats. ber), the name ofeach function, and a Ionger description Image contents. The symbol database contains a ver­ of the function if required (Figure I). This information bal description ofeach ofthe symbols' images present in allows symbols to be selected on the basis ofthe function the database. This field contains descriptions ofthe sym­ or type of function they have been used to represent. bols in terms of what is depicted in the symbol rather Functions can be represented by more than one symbol 482 OE BRUIJN, McDOUGALL, AND CURRY

... Microsoft Access • [Data Entr}' Form) 1!!IrJ' EI []I Eie Edit fonnat go To Aboul ..d4I2fJ IE _ IShowFunctiom~ l!I ElliI.....IilII1II.....liiIiI..fjjJ ...... iiiiiiiiiiIW • s...... Fwaction inforaalion SymbolID: 1346 FlM'lCtion ID 1169 Souree: l":":AI:'::'GA':""---- Function

T0 indicate thepresence 01 currency exchange facilities

Ratinga inforllUltioft ESP Ratings AIGA Ratings A wad01 banknotesfealuring lhe Concreteness: r:r.:r- Semantics: r number 100 CompleKi!y: 12.08 NameAgree: J'7ö Syntactics: rr­ Familiarity: 13.87 Concept Agree: r;o Pragmatics: r Meaninglulness: 13·92 N rciiCi"iö,"imag«tJ ame: L...._..J.!;.!!gr~lL._...! Semantic Distance: IJ07 I~m:'=o:':':ne;:;;y-----

01 253

Figure 1. The ESPbase Data Form used to enter and view symbol, function, and ratings information. so that when a function is used to select symbols in the ratings),and the American InstituteofGraphic Arts (AIGA database, all the symbols used to represent that particu­ ratings). Details ofeach ofthese ratings are given below. lar function will be retrieved. ESP ratings. ESP ratings are available for a set of239 Function ID. Each function in the database has a symbols included in the database (see McDougall et al., Function ID number. All information concerning a par­ 1998, for further details). Ratings have been obtaincd for ticular function is linked together by means ofits Func­ symbol concreteness, familiarity, meaningfulness, visual tion ID number. complexity, and semantic distance. The ratings provide Function names. Functions are identified by a func­ a measure ofthe degree to which a certain quality is pre­ tion name. A function name consists of a verbal label sent or absent in each symbol. All ofthe ratings were ob­ (e.g., "save document" or "save" for writing documents tained using a 1-5 scale (I = largely absent, 5 = strongly to disk). These function names are the most common present). Information has also been obtained about the way ofreferring to functions. extent to which users are likely to agree on the meaning Function descriptions. In addition to a function of a symbol when they first encounter it (see name and name, some functions are accompanied by a somewhat concept agreement, below). The database can be used to fuller description ofthe function. For example, the func­ select symbols on the basis of the relative strengths of tion labeled "tape recorder" is described in more detail particular characteristics required either for experimen­ as, "On amplifying and measuring equipment, to identify tal or for design purposes. For example, once the precise the terminals, switches, and controls by means ofwhich circumstances under which these dimensions affect sym­ a tape recorder is to be connected and operated." bol usability are known, it may be possible to predict, to a certain extent, the usability ofa set ofsymbols relative Ratings information to alternative sets on the basis ofthe relative strength of The database also includes information about a num­ their characteristics. ber ofvisual and semantic dimensions of symbols (Fig­ ESP ratings are easy to obtain and it is conceivable that ure I) in the form of subjective ratings. Ratings on the data­ users ofthe database may want to obtain ESP ratings for base come from the Effective Symbology Project (ESP different sets ofsymbols and include these symbols and SELECTING SYMBOL SETS 483

Elmage chafactefistics ..... Colou,s.· Foreground colour: IBlack [!J Background colour: IWhile [!J

.- 6100111felllu,es.· r-r- l oClllfelllu,es.· Bold ...... ------_ ..... P' [f:~:~~:~:~~~::~:~~~::l .. _....-_.. --_. r Line drawing ...... r Enclosure ...... r Angles Monolinear ...... ---.-.. r ...... _---_ .. r nlerseclions 3D --_ ...... --.... r I ...... --.----_ .. r Curves P' Silhouelle ...... -----_ .. r ...... ----_ .... Symmelry ... --_ ...... r S-shaped curves ...... P' Circles -... -----_ ...... P' Semicircles r- SpecilllFelllu,es.- ._-----_ ...... r Text -----_ ...... r Squares -...... r olher characlers P' Boxes r _ .. ------_...... Triangles Arrows ...... -.-.------r .__.__...... r

Figure 2. The Image Characteristics Box used to enter and view information about a symboI's graphical features. their ratings in the database. The way in which each of symbol's concept agreement. A further measure, name the ratings was obtained is explained briefly below. agreement, was calculated for those symbols for which the Concreteness. Experimental participants were asked most commonly given function differed from the given to rate symbols as concrete ifthey depicted real objects, function. Where concept agreement and name agree­ materials, or people. If these features did not appear in ment differ, this indicates that there is not a good fit be­ symbols, the syrnbolswere to be regarded as abstract. A tween the symbol and the designated function and that 5-point scale was used to rate concreteness (I = definitely other, better possibilities for a symbol 's function exist. abstract, 5 = definitely concrete). Participants were en­ Semantic distance. Symbols vary in their closeness of couraged to use all points on the scale inc1uding these the relationship between the image and the function two extreme values. being represented. In some cases this relationship is Visual complexity. Symbols were regarded as complex fairly direct (e.g., a printer used as a symbol for printing ifthey contained a lot ofdetail or were intricate, and as documents from word processors), whereas in other cases simple if they contained few elements or little detail. this function is much less direct (e.g., a triangle used to (I = very simple, 5 = very complex). indicate "hazard ahead"). In the first case, the semantic Familiarity. Symbols were rated as "familiar" if they distance between function and symbol might be regarded depicted a concept that participants had encountered as quite smalI, and in the second case, much larger (I = often before, and as "unfamiliar" ifthey depicted a con­ not closely related, 5 = very strongly related). cept that participants had never, or rarely, encountered AIGA ratings. AIGA ratings were obtained as indi­ before (I = very unfamiliar, 5 = very familiar). cations regarding the usability of a set of symbols ac­ Meaningfulness, concept agreement, and name agree­ cording to a number of criteria as judged by a panel of ment. Ratings of the perceived meaningfulness of sym­ professional designers (see AIGA, 1982). These usabil­ bols were also obtained (I = completely meaningless, ity criteria are semantic, syntactic, and pragmatic. Se­ 5 = very meaningful indeed). mantic usability refers to the ease with which users can de­ After rating the meaningfulness of symbols, partici­ rive a symbol's meaning; syntactic usability refers to the pants were asked to state briefly what they feit the mean­ relationship between symbols; and pragmatic usability ing ofthe symbol to be. The percentage of participants refers to a symbol's visibility. Thus, AIGA ratings are di­ who were able to ascertain a symbol 's designated func­ rectly related to forms ofusability, whereas the ESP rat­ tion, or meaning, was calculated as a measure of each ings are only indirectly related to usability. The advan- 484 OE BRUIJN, McDOUGALL, AND CURRY tage ofESP ratings is, however, that they can be used in this symbol consistent with that ofother symbols? (4) Are any context in which it is known how ESP ratings relate the most important elements recognized first? to usability, whereas AIGA ratings can be used only in Pragmatic. The pragmatic rating refers to the rela­ the context in which these ratings were obtained. tionship ofa symbol to auser. When determining this rat­ AIGA ratings are available for only a small number of ing,judges were asked to take into consideration questions the symbols presently stored in the database. However, a like (1) Can a person see the symbol? (2) Is perceptibility large number of symbols for which AIGA ratings are avail­ ofthis symbol seriously affected by poor viewing condi­ able can be obtained from AIGA and put into the database. tions? (3) Is this symbol vulnerable to vandalism? AIGA In addition, these ratings mayaiso be obtained for sets of ratings vary between 1 and 5, where I means that a sym­ symbols for which these ratings are not available yet (see bol has few of the characteristics on which the rating is AIGA, 1982, for more information about AIGA ratings). based and 5 means that a symbol has many ofthe char­ The three types of AIGA ratings are described below. acteristics on which the rating is based. Semantic. The semantic rating refers to the relationship ofa symbol to its meaning. When determining this rat­ RETRIEVING SYMBOLS ing, judges were asked to take into consideration ques­ tions like (1) How weIl does this symbol represent its Specifying the criteria that symbols have to satisfy in function? (2) Do people from various cultures and ages order to be retrieved is done using the Search Form (Fig­ misunderstand this symbol? (3) Is it difficult to learn this ure 3). Each ofthe symbol characteristics included in the symbol's meaning? database can be used in the Search Form. The Search Syntactic. The syntactic rating refers to the relationship Form is divided into six sections, described below. Each between one symbol and another. When determining this search field can be used either in isolation or in combi­ rating,judges were asked to take into consideration ques­ nation with other search fields in a single search. How­ tions like: (1) How does this symbol look? (2) How weIl ever, if more than one search field is used to retrieve does it relate to other symbols? (3) Is the construction of symbols, search criteria are combined by the "and" op-

,I. MicfOsoft Access . (S,mbol Search] 1!Ir;} EI

Figure 3. The ESPbase Search Form used to enter search criteria for retrieving symbols from the database. SELECTING SYMBOL SETS 485 erator. For a full explanation of how to use the symbol search, a number of improvements can be made. Some database in order to construct symbol sets, see the "ESP improvements may be prompted by the experimental Symbol Database Users' Manual," available from the au­ findings obtained within the Effective Symbology Pro­ thors on request. ject at the University ofWales Swansea and other future findings relating symbol characteristics to usability. For Search Form example, ratings of symbol distinctiveness (i.e., the Keyword search. Keyword searches can be used to amount by which a symbol "stands out" from others) have retrieve symbols that contain certain words in their text already been obtained (McDougall et al., 1998), and these fields, such as image contents and function fields. For ratings could easily be added to the characteristics ofsym­ example, typing the word temperature in the function bols already in the database. field will retrieve symbols with functions like tempera­ In addition to adding more symbol characteristics, it ture control and rising temperature. may be possible to incorporate some ofthe research find­ Source. Using the source field to retrieve symbols al­ ings relating symbol characteristics to symbol usability. lows one to either retrieve all symbols from a single Inclusion of these findings could potentially lead to the source or to limit other types ofsearches to retrieve only development ofa tool for selecting symbol sets that are symbols from the specified source. optimized for usability given the circumstances in which Ratings. Symbols can be retrieved on the basis of the symbols will be used. For example, if information their ESP and AIGA ratings. This facility has had enor­ about the effects ofsymbol concreteness, complexity, and mous value to our symbol usability research because it distinctiveness on symbol usability is incorporated into has allowed us to construct, almost at the flick ofa switch, the database, it may be possible to select symbols that are symbol sets that vary orthogonally on their concreteness maximally distinctive and as concrete and complex as and complexity rating scores (see McDougall et al., 1998). the context in which they are to be used dictates. This was achieved simply by specifying ranges for the Notwithstanding possible extensions of the database, concreteness and complexity ofsymbols that fall into the however, the abundance of symbol characteristics al­ extremes on these scores. A set ofsymbols that were ab­ ready included in the database ensures that the ESPbase stract and simple was obtained by searching for symbols allows the construction of symbol sets conforming to a with concreteness and complexity ratings less than 2.5 host ofdifferent criteria. Given the need for research into out of5. the effects of a growing number of symbol characteris­ Image characteristics. Symbols can be selected on tics on usability, as well as the importance ofconsider­ the basis of either the presence or absence of certain ing many symbol characteristics in applying symbols in image features. This information can be specified in the various contexts, the ESPbase may provide useful con­ Feature Specification Box, which pops up ifits button in tributions to both research and design. the Search Form is clicked. The Feature Specification Box is similar to the Image Characteristics Box except Availability that a yes/no option is available for each of the local, The database is available from the directly from the global, and special features, depending on whether the authors at the University of Wales Swansea. There is a search is for symbols with these features present or ab­ small cost involved to cover postage and packaging. Run­ sent. Conceivably, such a search could be used to con­ ning the database requires Microsoft Windows 3.1 or struct symbol sets in which the symbols are maximally higher and Microsoft Access 2.0 or higher. distinctive in terms of shared image features. For exam­ pie, a set in which all symbols do have curves but do not REFERENCES have angles would be quite distinctive from a set in AMERICAN INSTITUTE OF GRAPHIC ARTS (1982). Symbol signs: A system which all symbols do have angles but do not have curves. ofpassenger/pedestrian oriented symbols based on an evaluation of Operators. Arrhythmie operators allow one to spec­ existing symbology. New York:Hastings House. ify ranges ofrating scores for the symbols to be retrieved. AREND, U., MUTHlG, K. P., & WANDMACHER, J. (1987). Evidence for For example, symbols with a familiarity rating greater global feature superiority in menu selection by icons. Behaviour & Information Technology, 6, 4 I 1-426. than 3 can be selected by using the "greater than" opera­ BARNARD, P.,& MARCEL, T. (1984). Representation and understanding tor in the familiarity field. Logical operators allow more in the use of symbols and pictograms. In R. Easterby & H. Zwaga complicated searches to be performed by combining (Eds.), Information design (pp. 37-76). Chichester, U.K.: Wiley. search terms within search fields. For example, combin­ BENJAFIELD, J., FROMMHOLD, K., KEENAN, T., MUCKENHEIM, R., & ing keywords in the function field with the "and" opera­ MUELLER, D. (1993). Imagery, concreteness, goodness, and familiar­ ity ratings for 500 proverbs sampled from the Oxford Dictionary of tor allows one to narrow down the number ofsymbols re­ English Proverbs. Behavior Research Methods, Instruments, & Com­ trieved relative to using just a single search term. puters, 25, 27-40. BLACKWELL, J. S., & CUOMO, D. L. (1991). Evaluation of a proposed FINAL COMMENTS space and missile warning symbology standard for graphical displays. In Proceedings ofthe Human Factors Society Thirty-fifth Annual Meeting (pp. 102-106). Santa Monica, CA: Human Factors Society. The ESP symbol database is still under development. CHRIST, R. E. (1984). Research for evaluating visual display codes: An Although it has already been an invaluable aid in our re- emphasis on colour coding. In R. Easterby & H. Zwaga (Eds.), In- 486 DE BRUIJN, McDOUGALL, AND CURRY

formation design. The design and evaluation ofsigns andprintedma­ measures for content-based image retrieval: A comparison.lnforma­ terial (pp. 209-228). tion Processing & Management, 33, 319-337. COLTHEART, M. ( 1981). The MRC psycholinguistic database. Quarterly PAIVIO, A., YUlLLE, 1.C, & MADIGAN, S. A. (1968). Concreteness, im­ Journal ofExperimental Psychology, 33A, 497-505. agery and meaningfulness values for 925 words. Journal ofExperi­ CRYSTAL VISION (1994). 4000 lcons with lcon Wizard [Computer soft­ mental Psychology Monograph Supplement, 76, (3, Part 2). ware). Carlsbad, CA: CrystalVision Software. QUINLAN, P. T. (1992). The Oxford Psycholinguistic Database [Com­ DREYFUSS, H. (1971). The symbol source book. New York: McGraw­ puter software]. Oxford: Oxford University Press. HilI. ROGERS, Y. (1986). Evaluating the meaningfulness oficon sets to rep­ DYSON, M. C. (1992). How do you describe a symbol? The problems in­ resent command operations. In M. D. Harrison & A F.Monk (Eds.), volved in retrieving symbols from a database. Information Services People and computers: Designingfor usability (pp. 586-603). Cam­ & Use, 12, 65-76. bridge: Cambridge University Press. FAMILANT, E. M., & DETWEILER, M. C. (1993).lconic reference: Evolv­ ROGERS, Y., & OBORNE, D. 1. (1987). Pictorial communication of ab­ ing perspective and an organising framework. International Journal stract verbs in relation to human-computer interaction. British Jour­ ofMan-Machine Studies, 39, 705-728. nal ofPsychology, 78, 99-/12. FRIENDLY, M., FRANKLlN, P. E., HOFFMAN, D., & RUBIN, D. C. (1982). ROHR, G., & KEPPEL, E. (1985). Iconic interfaces: Where to use and The Toronto Word Pool: Norms for imagery, concreteness, ortho­ how to construct. In H. W. Hendrick & O. Brown (Eds.), Human fac­ graphie variables, and grammatical usage for 1,080 words. Behavior tors in organisation design and management (pp. 269-275). Amster­ Research Methods & Instrumentation, 14,375-399. dam: Elsevier. GARCIA, M., BADRE, A. N., & STASKO, J. T. (1994). Development and SANFELlU, M. C; & FERNANDEZ, A. (1996). A set of 254 Snodgrass­ validation of icons varying in their abstractness. Interacting with Vanderwart pictures standardized for Spanish: Norms for name Computers,6,191-211. agreement, image agreement, familiarity, and visual complexity. Be­ GEISELMAN, R. E., LANDEE, B. M., & CHRISTEN, F. G. (1982). Percep­ havior Research Methods, Instruments, & Computers, 28, 537-555. tual discriminability as a basis for selecting graphie symbols. Human SNODGRASS,1. G., & VANDERWART, M. (1980). A standardized set of Factors, 24,329-337. 260 pietures: Norms for name agreement, image agreement, famil­ GILHOOLY, K. J., & LOGIE, R. H. (1980). Age-of-acquisition, imagery, iarity and visual complexity. Journal ofExperimental Psychology: concreteness, familiarity, and ambiguity measures for 1,944 words. Human Learning & Memory, 6, 174-215. Behavior Research Methods & Instrumentation, 12,395-427. STAMMERS, R. B., GEORGE, D. A, & CAREY, M. S. (1989). An evalua­ GITTENS, D. (1986). Icon-based human-computer interaction. Interna­ tion of abstract and concrete icons for a CAD package. In E. D. tional Journal ofMan-Machine Studies, 24, 519-543. Megaw (Ed.), Contemporary ergonomics 1989 (pp. 416-421). Lon­ GRAPHIC ARTISTS GUILD FOUNDATION (1995). Disabilityaccess symbols don: Taylor & Francis. [On-line). Available HTTP: www.gag.org/das STAMMERS, R. B., & HOFFMAN, J. (1991). Transfer between icon sets GRAY, L. A (1990). lcon Manager [Computer software]. Chino, CA: and ratings oficon concreteness and appropriateness. In Proceedings Impact Software. ofthe Human Factors Society Thirty-Fifth Annual Meeting (pp. 354­ GREEN, A. J. K., & BARNARD, P.J. (l990).lconic interfacing: The role 358). Santa Monica, CA: Human Factors Society. of icon distinctiveness and fixed or variable screen locations. In VAN SCHAGEN, 1., TAMSMA, N., BRUGGEMANN, F., JACKSON, 1. L., & MI­ D. Diaper et al. (Eds.), Human-eomputer interaction-Interact '90 CHON, J. A. (1983). Namen en normen voor plaatjes [Names and (pp. 457-462). Amsterdam: Elsevier. norms for pictures). Nederlands Tijdschrift voor de Psychologie, 38, HORTON, W. (1994). The icon book: Visual symbols for computer sys­ 236-241. tems and documentation. New York:Wiley. WILSON, M. (1988). MRC Psycholinguistic Database: Machine-usable THE IcON FACTORY (1997). PixelPlace [On-line]. Available HTTP: dictionary, version 2.00. Behavior Research Methods, Instruments, pixelplace.com & Computers, 20, 6-10. LIU,Z. Q., & SUN, J. P. (1997). Structured image retrieval. Journal of Visual Languages & Computing, 8, 333-357. NOTES MARTEIN, R. (1995). Norms for name and concept agreement, famil­ iarity, visual complexity and image agreement on a set of 216 pic­ I. Unless otherwise specified, the term symbol refers to icons, picto­ tures. Psychologica Belgica, 35, 205-225. grams, symbols, and signs. MAZZA, C. (1997). Neural-net inference and content addressable mem­ 2, Microsoft Access is protected by copyright held by the Microsoft ory.IEEE Transactions on Neural Networks, 8,133-140. Corporation. Microsoft Windows is a registered trademark ofthe Micro­ McDoUGALL, S. J. P., CURRY, M. B., & DE BRUIJN, O. (1998). Under­ soft Corporation. standing what makes icons effective: How subjective ratings can in­ 3. In order to save disk space, the original image files were RLE en­ form design. In M. A. Hanson (Ed.), Contemporary ergonomics 1988 coded before being embedded into the symbol database. Encoding re­ (pp. 285-289). London: Taylor & Francis. duces their file down to approximately 10% ofthe non-encoded image McDoUGALL, S., CURRY, M. B., & DE BRUIJN, O. (1999). Measuring files. The number of different colors used in the symbol images was symbol and icon characteristics: Norms for concreteness, complexity, limited to 16 for the same reason. meaningfulness,familiarity,and semanticdistancefor 239 symbols.Be­ havior Research Methods, Instruments. & Computers, 31, 487-519. (Manuscript received December 5, 1997; MEHTRE, B. M., KANKANHALLI, M. S., & LEE, W. F. (1997). Shape revision accepted for publication April 23, 1998.)