Designing Menu Selection Systems
`
`Ben Shneiderman
`Department of Computer Science, University of Maryland, College Park,
`MD 20742
`
`Menu selection systems reduce training and memoriza-
`tion, simplify entry of choices, and structure the user’s
`tasks. However, the use of menu selection is no guaran-
`tee that novices or experts will be satisfied or able to
`carry out their work. This article focuses on the multiple
`design issues in creating successful menu selection sys-
`tems. These include the primary issue of semantic orga-
`nization and the host of secondary
`issues such as
`response time and display rates, shortcuts for frequent
`users, titles, phrasing of menu items, graphic layout, and
`selection mechanisms. Novel approaches such as pop
`up menus and embedded menus are covered. Experimen-
`tal results and design guidelines are presented.
`
`1. Introduction
`they can
`Menu selection systems are attractive because
`eliminate
`training
`and memorization
`of complex
`com-
`mand sequences. When
`the menu
`items are written using
`familiar
`terminology,
`users can select an item easily and
`indicate
`their choice either with one or two keypresses or
`through
`the use of a pointing device. This simplified
`inter-
`action
`style reduces
`the possibility
`of keying errors and
`structures
`the
`task
`to guide
`the novice and
`intermittent
`user. With
`careful design
`and high speed
`interaction,
`menu selection can become appealing
`to expert frequent
`users as well.
`is often contrasted with
`by menu selection
`Interaction
`interaction
`by command
`language,
`but
`the distinctions
`are sometimes blurred. Typically menu selection
`requires
`a single keystroke, whereas commands may be lengthy.
`However, how would you classify a menu
`in which the user
`has to type a six or eight letter
`item? Typically, menu se-
`lection presents
`the choices on the display, whereas com-
`mands must be memorized. However,
`how would you
`classify a menu
`that offered 4 numbered
`choices and ac-
`cepted 10 more generic
`choices
`that are not displayed?
`How would you classify a system
`that offers single letter
`
`Received April 10, 1985: revised May 10, 1985; accepted
`
`June 20, 198.5.
`
`0 1986 by John Wiley & Sons, Inc.
`
`two-dimensional menus
`prompts? What about graphical,
`where selection
`is made by pointing with a mouse or on a
`touchscreen?
`Finally, what category
`is voice synthesis/
`recognition menu
`interaction?
`to
`it is more useful
`Rather
`than debate
`terminology,
`the system offers on
`maintain
`an awareness of how much
`the display at the moment
`the selection
`is made,
`the form
`and content of item selection,
`and what problem domain
`knowledge
`is necessary
`for users to succeed. Menu selec-
`tion is especially effective when users have little training,
`are intermittent
`in using
`the system, are unfamiliar with
`the terminology,
`and need help in structuring
`their deci-
`sion-making
`process.
`it does
`However,
`if a designer employs menu selection,
`not guarantee
`that
`the system will be appealing
`and easy
`to use. Effective menu selection systems emerge only after
`careful consideration
`and
`testing of numerous
`design
`is-
`sues such as semantic organization, menu system struc-
`ture,
`the number
`and sequence
`of menu
`items,
`titling,
`prompting
`format, graphic
`layout and design, phrasing of
`menu
`items,
`display
`rates,
`response
`time,
`shortcuts
`through
`the menus
`for knowledgeable
`frequent
`users,
`availability
`of help, and
`the selection mechanism
`(key-
`board, pointing devices,
`touchscreen,
`voice, etc.).
`
`2. Semantic Organization
`is to create a sen-
`The primary
`task for menu designers
`sible, comprehensible,
`memorable,
`and convenient
`se-
`mantic
`organization.
`Some
`lessons can be
`learned
`by
`studying
`the semantic decomposition
`of a book into chap-
`ters, a program
`into modules,
`the animal kingdom
`into
`species, or a Sears catalog
`into sections. Hierarchical
`de-
`compositions,
`natural
`and comprehensible
`to most peo-
`ple, are appealing because every item belongs
`to a single
`category. Unfortunately,
`in some applications
`an
`item
`may be difficult
`to classify as belonging
`to one category,
`and
`the temptation
`to duplicate
`entries or create a net-
`work increases. Despite
`some limitations,
`the elegance of
`tree structures
`should be appreciated.
`soups, main
`Restaurant menus
`separate
`appetizers,
`dishes, desserts,
`and drinks
`to help customers organize
`
`JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE. 37(2):57-70; 1986
`
`CCC 0002.8231/86/020057-14$04.00
`
`

`

`fit logically into cate-
`items should
`their selections. Menu
`gories and have
`readily understood meanings.
`Restau-
`rateurs who list dishes with
`idiosyncratic
`names
`such as
`“Veal Monique,”
`generic
`terms
`such
`as
`“house
`dressing,”
`or unfamiliar
`jargon
`such as “Wor Shu Op”
`should expect
`that waiters will spend ample time explain-
`ing the alternatives
`or should anticipate
`that customers
`will be anxious because of their
`insecurity
`in ordering.
`the
`Similarly,
`for computer menu
`selection
`systems,
`so
`categories
`should be comprehensible
`and distinctive
`that
`the users are confident
`in making
`their selections.
`Users should have a clear idea of what will happen when
`they make a choice. Computer menu selection systems are
`more difficult
`to design
`than
`restaurant menus because
`computer
`screens
`typically
`allow
`less information
`to be
`displayed
`than do printed menus. Screen space is a scarce
`resource.
`In addition,
`the number of choices and the com-
`plexity
`is greater
`in many computer
`applications,
`and the
`computer user may not have a helpful waiter to turn
`to for
`an explanation.
`of menu
`organization
`of meaningful
`The
`importance
`items was demonstrated
`in a study with 48 novice users
`[ 11. Simple menu
`trees with 3 levels and 16 target
`items
`were constructed
`in meaningfully
`organized
`and disor-
`ganized
`forms. Error rates were nearly halved,
`and user
`think
`time (time from menu presentation
`to user’s selec-
`tion of an
`item) was reduced
`for the meaningfully
`or-
`ganized
`form.
`In a later menu search study, McDonald
`et
`al.
`[2] found
`that
`semantically meaningful
`categories,
`such as food, animals, minerals,
`and cities,
`led to shorter
`response
`times
`than did random or alphabetic
`organiza-
`tions. This experiment
`tested 109 novice users who worked
`through 10 blocks of 26 trials. The authors conclude
`that
`“these
`results demonstrate
`the superiority of a categorical
`menu organization
`over a pure alphabetical
`organization,
`particularly when
`there
`is some uncertainty
`about
`the
`larger menu
`terms.” With
`structures
`the effect
`is even
`more dramatic,
`as has been demonstrated
`by studies with
`extensive videotext databases
`[3,4].
`[S]
`model
`These
`results
`and
`the syntactic/semantic
`is to first
`suggest
`that
`the key to menu
`structure
`design
`consider
`the semantic organization.
`The number of items
`on the screen becomes a secondary
`issue.
`Menu
`selection
`applications
`range from trivial choices
`between
`two
`items
`to complex
`videotex
`systems with
`300,000 screens. The simplest
`applications
`consist of a
`single menu, but even with this limitation
`there are many
`variations.
`The second group of applications
`includes
`a
`linear sequence
`of menu
`selections;
`the progression
`of
`tree
`menus
`is independent
`of the user’s choice. Strict
`structures make up the
`third group, which
`is the most
`situation. Acyclic
`common
`(these
`are menus
`that
`are
`than one path) and cyclic (there are
`reachable
`by more
`that allow users to repeat menus) net-
`meaningful
`paths
`works constitute
`the fourth group. These groupings
`de-
`scribe
`the semantic
`organization;
`special
`traversal
`com-
`mands may enable users to jump around
`the branches of a
`
`to the previous menu, or to go to the be-
`tree, to go back
`ginning of a linear sequence.
`
`(Y,N)?
`
`A. Single Menus
`to ac-
`is sufficient
`a single menu
`In some situations,
`complish
`a task. Single menus may have
`two or more
`items, may require
`two or more screens, or may allow mul-
`tiple selections. Single menus may pop up on the current
`work area or may be permanently
`available
`(in a separate
`window or on a data
`tablet) while
`the main display
`is
`changed. Different guidelines
`apply for each situation.
`Binary Menus
`The simplest menu
`is a binary menu
`with yes/no or true/false
`choices such as is found
`in many
`home computer games:
`DO YOU WANT
`INSTRUCTIONS
`or in a medical-history
`taking
`interview
`YOU HAVE HAD SURGERY TO REMOVE YOUR
`APPENDIX
`1 - TRUE
`2 - FALSE
`MAKE YOUR SELECTION NOW:
`In the
`Even
`these simple examples
`can be improved.
`first case, a novice user might not understand
`the (Y,N)
`prompt-really
`an abbreviated
`form of the menu
`of
`choices. Secondly
`this common query leaves the user with-
`out a clear sense of what is going to happen next. Typing
`“Y” might produce many pages of instructions
`and
`the
`user might not know how to stop a lengthy output. Typing
`“N” is also anxiety producing because
`the user has no idea
`of what
`the program will do. Even
`in writing
`simple
`menus, clear and specific choices should be offered which
`give the user the sense of control:
`Your choices are:
`1 - Get 12 lines of brief instructions
`2 - Get 89 lines of complete
`instructions
`3 - Go on to playing
`the game
`Type a number
`and press RETURN:
`Since this version has three
`items,
`it is no longer a binary
`menu.
`It offers more specific items so the user knows what
`to expect, but it still has the problem
`that users must take
`instructions
`now or never. Another
`strategy might be:
`At any time you may type
`?-
`to get 12 lines of brief instructions
`?? -
`to get 89 lines of complete
`instructions
`Be sure to press RETURN
`after every command
`Ready for game playing commands:
`This example calls attention
`to the sometimes narrow dis-
`tinction
`between
`commands
`and menu
`selection:
`the
`menu choices have become more command-like
`since the
`user must now recall the ? or ?? syntax.
`The following examples
`illustrate
`additional
`
`issues in
`
`58
`
`JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE--March
`
`1986
`
`

`

`items can be identified by single let-
`binary menus. Menu
`ter mnemonics,
`as in this photo
`library
`retrieval system:
`Photos are indexed by film type
`B Black and white
`C Color
`Type the letter of your choice
`and press RETURN:
`service:
`or in a shop-by-computer
`PLEASE SELECT THE DESIRED METHOD
`OF BILLING:
`MASTERCARD.............................1
`VISA.......................................2
`KEY IN YOUR SELECTION AND PRESS ENTER:
`The mnemonic
`letters in the photo menu are often pre-
`ferred
`to the numbered
`choices
`in the shop-by-computer
`menu (see Section 7). The long line of periods can be a dis-
`traction,
`and the uppercase-only
`lettering may slow read-
`ing. The mnemonic
`letter approach
`requires
`additional
`caution
`in avoiding
`collision and
`increases
`the effort of
`translation
`to foreign
`languages,
`but
`their clarity and
`memorability
`are an advantage
`in many
`applications.
`These simple examples
`demonstrate
`alternative ways to
`identify menu
`items and convey
`instructions
`to the user.
`No optimal
`format
`for menus has emerged,
`but consis-
`tency across menus
`in a system is extremely
`important.
`Multiple Item Menus Single menus may have more
`than
`two items. Examples
`include online quizzes:
`Who invented
`the telephone?
`1 Thomas Edison
`2 Alexander Graham Bell
`3 Lee De Forest
`4 George Westinghouse
`Type the number
`and press RETURN:
`or the list of options
`in a document
`processing
`EXAMINE, PRINT, DROP, OR HOLD?
`The quiz
`example
`has distinctive,
`comprehensible
`items, but the document
`processing example shows an im-
`plied menu
`selection
`that could be confusing
`to novice
`users. There are no explicit
`instructions,
`and it is not ap-
`parent
`that
`single
`letter
`abbreviations
`are acceptable.
`Knowledgeable
`and frequent
`users may prefer
`this short
`form of a menu selection, usually called a prompt,
`for its
`speed and simplicity.
`Extended Menus Sometimes
`items
`list of menu
`the
`may require more
`than one screen but allow only one
`meaningful
`item to be chosen. One resolution
`is to create a
`tree-structured
`menu, but sometimes
`the desire
`to keep
`is very appealing. The
`the system to one conceptual menu
`first portion of the menu
`is displayed with an additional
`menu
`item that
`leads to the next screen
`in the extended
`menu sequence. A typical application
`is in word process-
`ing systems, where common choices are displayed
`first but
`
`system:
`
`infrequent
`screen:
`
`or advanced
`
`features
`
`are kept on the second
`
`MAIN MENU
`
`SUPERDUPERWRITER
`PAGE 1
`Edit a file
`1
`Copy a file
`2
`Create a file
`3
`Erase a file
`4
`Print a file
`5
`View the directory
`6
`P2 Go to PAGE 2
`Type the number of your choice
`and Press RETURN
`
`MAIN MENU
`
`SUPERDUPERWRITER
`PAGE 2
`7 Alter line width
`set
`8 Change character
`recovery of damaged file
`9 Attempt
`10 Reconstruct
`erased file
`Set cursor blink
`rate
`11
`12
`Set beep volume
`13 Run diagnostics
`to PAGE 1
`Pl Go back
`Type the number of your choice
`and Press RETURN
`the extended
`screen menu will go on for
`Sometimes
`many screens of command
`items or data items. More elab-
`orate scrolling capabilities may be needed.
`Pop-up Menus The
`or “pull down”
`term “pop-up”
`menus refers to the process of forcing a menu to appear on
`the screen
`in response
`to a click with a pointing
`device
`such as a mouse. The Xerox Star, Apple Lisa, and Apple
`Macintosh
`(Figure
`1) made
`these possibilities widely
`available. There
`is a great satisfaction on the part of most
`users in making
`the menu appear
`rapidly. Selection can
`be made by moving
`the pointing
`device over the menu
`items, which respond by highlighting
`(inverse video, a box
`surrounding
`the item, or color have been used).
`The contents of the pop-up menu may depend on the
`position of the cursor when the pointing device is clicked.
`Since
`the pop-up menu covers a portion of the screen,
`there
`is a strong motivation
`to keep the menu
`text small.
`Hierarchical
`sequences
`are also used in pop-up menus.
`Permanent Menus Since menus can be used for per-
`manently
`available
`commands
`that can be applied
`to a
`displayed object. For example,
`the Bank Street Writer, a
`word processor
`designed
`for children,
`always shows a
`fragment of the text and this menu:
`
`ERASE
`UNERASE
`
`MOVE
`MOVEBACK
`
`TRANSFER
`FIND
`REPLACE MENU
`
`JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE-March
`
`1986
`
`59
`
`

`

`f6
`
`File
`
`Font
`
`Fonts
`
`iu 1 FatBits
`
`Show Page
`Edit Pattern
`Brush Shape
`Brush Mirrors
`Introduction
`Short Cuts
`
`FIG.
`
`I.
`
`Pulldown menu on the Apple Macintosh
`
`appears when the user clicks down on the mouse button.
`
`the left and right arrow keys causes items to be
`Moving
`highlighted
`in reverse video. When
`the de-
`sequentially
`sired command
`is highlighted,
`pressing
`the RETURN key
`initiates
`the action.
`include Apple
`of permanent menus
`Other applications
`MacPaint,
`computer-assisted
`design
`(CAD)
`systems, or
`other graphics
`systems
`that display an elaborate menu of
`commands
`to the side of the object being manipulated.
`Price [6] describes
`a CAD system with 120 choices
`in an
`on-screen menu. Lightpen
`touches or other cursor-action
`devices allow the user to make selections without using the
`keyboard.
`Multiple Selection Menus A further variation on sin-
`gle menus
`is the capacity
`to make multiple
`selections
`from
`the choices offered. For example,
`this menu from a politi-
`cal interest
`survey allows multiple
`choices on one touch
`screen:
`
`ISSUES
`
`POLITICAL
`HIGH UNEMPLOYMENT
`AID TO ELDERLY
`NUCLEAR FREEZE
`HIGH DEFENSE BUDGET
`GOVERNMENT
`REGULATION
`FOREIGN AID
`PERSONAL TAXES
`CIVIL DEFENSE
`RIGHT TO ABORTION
`CRIME CONTROL
`DONE
`MINORITY RIGHTS
`ISSUES THAT YOU
`TOUCH UP TO THREE
`IMPORTANT,
`FEEL ARE THE MOST
`AND THEN TOUCH DONE:
`
`is nicely handled with a multiple selection
`This situation
`single menu;
`it would have been cumbersome
`to ask 11
`binary choices when the user could not scan the full list of
`issues. The system might highlight
`already selected
`items
`with a check mark or bold face.
`Summary Even
`the case of single menus provides a
`rich domain
`for designers and human
`factors researchers.
`Questions of wording,
`screen
`layout, and selection mech-
`anism all emerge even in the simple case of choosing from
`one set of items. Still more challenging
`questions
`emerge
`from designing
`sequences
`and trees of menus.
`
`B. Linear Sequence of Menus
`menus can be used to
`Often a series of interdependent
`guide the user through a series of choices in which the user
`sees the same sequence of menus no matter what choices
`are made. A document
`printing package might have this
`linear sequence of menus:
`
`printed at
`
`Do you want the document
`1 - your terminal
`line printer
`center
`2 -
`the computer
`laser printer
`center
`3 -
`the computer
`Type the number of your choice and press RETURN:
`
`Do you want
`1 -
`single spacing
`2 - double spacing
`Type the number
`of your choice and press RETURN:
`
`60
`
`JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE-March
`
`1986
`
`

`

`Do you want
`1 - no page numbering
`2 - page numbering
`on the top, right justified
`3 - page numbering
`at the bottom,
`centered
`Type the number of your choice and press RETURN:
`Another example would be an online examination
`that
`had a sequence of multiple
`choice
`test items, each made
`up in the form of a menu.
`Movement Through the Menus Linear
`sequences
`guide
`the user through
`a complex decision-making
`pro-
`cess by presenting
`one decision at a time. The document-
`printing
`example could be improved by offering
`the user a
`mechanism
`for going back to previous menus
`to review or
`change
`choices made
`earlier. A second
`improvement
`would be to display the results of previous choices, so users
`could see what decisions had been made. A third improve-
`ment might be to let the users know how many and which
`menus are yet to be seen.
`traversal,
`allowing backward
`The
`first
`improvement,
`could be handled
`easily by changing
`the instructions
`to:
`Type
`the number
`of your choice and press RETURN,
`or type “B” and press RETURN
`to go back to the pre-
`vious menu:
`the record of previ-
`showing
`improvement,
`The second
`ous menus, could be handled by displaying
`the choices al-
`ready made. The third
`improvement,
`showing upcoming
`choices, could be handled by displaying
`a descriptive
`term
`about
`the menus
`to follow, or simply an indication
`that
`this is the third of six menus. Unfortunately,
`as more im-
`provements
`are made
`there
`is a greater possibility of cre-
`ating cluttered
`displays.
`Judgments
`based on experience
`can resolve many decisions, but experimental
`tests with
`alternative
`formats
`and several classes of users may be
`useful to guide designers.
`Summary Linear
`are a simple
`of menus
`sequences
`the user through
`a deci-
`and effective means
`for guiding
`sion-making
`process. The user should be given a clear
`sense of progress or position within
`the sequence and the
`means
`for going backwards
`to earlier choices (and possi-
`bly to terminating
`or restarting
`the sequence).
`is of-
`Choosing
`the order of menus
`in a linear sequence
`ten straightforward,
`but care must be taken
`to match user
`expectations. One strategy
`is to place
`the easy decisions
`first to relieve users of some concerns,
`enabling
`them
`to
`concentrate
`on more difficult choices.
`
`C. Tree-Structured Menus
`When a collection of items grows and becomes difficult
`to maintain
`under
`intellectual
`control, people form cate-
`gories of similar
`items,
`creating
`a tree structure
`[7,8].
`Some collections
`can be easily partitioned
`into mutually
`exclusive groups with distinctive
`identifiers.
`Familiar
`ex-
`amples
`include:
`Male, female
`Animal, vegetable, mineral
`
`a
`
`is
`for
`
`autumn, winter
`Spring, summer,
`Sunday, Monday, Tuesday, Wednesday, Thursday,
`Friday, Saturday
`Less than 10, between 10 and 25, greater
`Percussion,
`string, woodwind, brass
`lead to confu-
`Even
`these groupings may occasionally
`indexing
`is a
`sion or disagreement.
`Classification
`and
`is no perfect
`complex
`task, and
`in many situations
`there
`solution acceptable
`to everyone. The initial design can be
`improved as a function of feedback
`from users. Over time,
`as the structure
`is improved
`and as users gain familiarity
`with it, success rates will improve.
`menu systems
`Despite
`their problems,
`tree-structured
`have the power to make
`large collections of data available
`to novice or intermittent
`users. If each menu has 8 items,
`then a menu
`tree with 4 levels has the capacity
`to lead an
`untrained
`user
`to the right frame out of a collection of
`4096 frames.
`and compre-
`at each level are natural
`If the groupings
`hensible
`to the user, and if the user knows what he/she
`is
`looking for, then the menu
`traversal can be accomplished
`in a few seconds-more
`quickly
`than
`flipping
`through
`book. On the other hand,
`if the groupings
`are unfamiliar
`and
`the user has only a vague notion of what he/she
`looking for, it is possible
`to get lost in the tree menus
`hours [9].
`Depth Versus Breadth The depth
`(number of levels)
`of a menu
`tree depends,
`in part, on the breadth
`(number
`of items per level).
`If more
`items are put
`into the main
`menu,
`then the tree spreads out and has fewer levels. This
`is advantageous,
`but not if clarity is substantially
`compro-
`mised or if a slow display
`rate consumes
`the user’s pa-
`tience. Several authors have urged four to eight items per
`menu, but at the same time they urge no more than three
`to four levels. With
`large menu applications,
`oue or both
`of these guidelines must be compromised.
`in retriev-
`D. P. Miller [lo] studied user performance
`ing items from 4 versions of a tree-structured menu system
`containing
`64 target
`items. Menus had 2,4, 8, or 64 items
`in each screen, with corresponding
`depths of 6, 3, 2, and
`1. The 64 items were carefully
`chosen
`that “they
`form
`valid semantic
`hierarchies”
`in each of the 4 versions.
`Speed of performance was fastest with 4 or 8 items per
`menu, and the lowest error rate occurred with 8 items per
`menu. These results are useful, but there were two special
`conditions
`that may limit
`the applicability
`of this study:
`subjects became very familiar with the menus during
`the
`training
`and 128 trials, and
`the 64 items were chosen so
`that there were meaningful
`groupings
`in all 4 versions.
`Kiger [ 1 l] grouped 64 items in 5 menu
`tree forms:
`8-2: 8 items on each of 2 levels
`4-3: 4 items on each of 3 levels
`2-6: 2 items on each of 6 levels
`4-l + 16-1: A 4 item menu followed by a 16 item
`menu
`16-1 + 4-l : A 16 item menu
`menu
`
`than 25
`
`followed by a 4 item
`
`JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE-March
`
`1986
`
`61
`
`

`

`least
`the slowest,
`tree, 2-6, produced
`The deep narrow
`accurate,
`and least preferred version, whereas
`the 8-2 was
`among
`the best for speed, accuracy,
`and preference. The
`22 subjects performed
`16 searches on each of the 5 ver-
`sions.
`a one-level menu having 23
`Dray et al. [12] compared
`one-word
`target
`items arranged on 6 lines with a two-level
`menu having 6 items in the main menu. Selection was by
`cursor control arrow keys and an ENTER key. Subjects
`had 138 trials
`in each condition
`in this counterbalanced
`within subjects design. Although neither version emerged
`as superior,
`there was a significant
`order effect
`that
`the
`authors
`interpreted
`as evidence
`that
`the one-level menu
`was easier
`to learn.
`Informal
`reports
`from subjects
`sup-
`ported
`the conclusion
`that continuously
`seeing
`the full
`picture aided decision making.
`l-4096,
`the numbers
`When
`the menu
`tree contains
`time to locate a target number was found
`to increase with
`the breadth
`of the
`tree
`[13]. Search
`times were almost
`twice as long in a 12 level tree having 2 choices at each level
`(2-12) as opposed
`to a 3 level tree with 16 choices at each
`level (16-3). Although
`the 6 subjects made choices more
`rapidly
`in the shorter menus,
`the effort to work through
`the more numerous menus did slow them down substan-
`tially. Each subject did 12 trials with each of 4 widths.
`While
`the semantic
`structure of the items cannot be ig-
`nored,
`these studies suggest
`that fewer levels aid decision
`making. Of course,
`display
`rates,
`response
`time,
`and
`screen clutter must be considered
`in addition
`to the se-
`mantic organization.
`Semantic Grouping in Tree Structures Rules for se-
`mantic validity are hard
`to state, and there
`is always the
`danger
`that some users may not grasp
`the designer’s or-
`ganizational
`framework. Young and Hull [14] examined
`“cognitive mismatches”
`in the British Prestel viewdata
`system
`[ 151. Problems
`included
`overlapping
`categories,
`extraneous
`items, conflicting
`classifications
`in the same
`menu,
`unfamiliar
`jargon,
`and generic
`terms. Based on
`this set of problems,
`the rules
`for forming menu
`trees
`might be:
`
`1. Create groups of logically similar items. For exam-
`ple, a comprehensible menu would list countries at
`level one, states or provinces at level two, and cities
`at level three.
`that cover all possibilities. For exam-
`2. Form groups
`ple, a menu with age ranges O-9, 10-19, 20-29,
`and older
`than 30 makes
`it easy for the user
`to
`select an item.
`3. Make sure that items are non-overlapping.
`Lower-
`level items should be naturally
`associated with a
`single higher-level
`item. Young and Hull offered
`an example
`of a poorly designed
`screen with
`“Places
`in Britain”
`and “Regions of England”
`as
`overlapping
`items on the same menu.
`items
`that
`4. Use familiar
`terminology,
`but ensure
`from each other. Choosing
`are dtitinctive
`the right
`
`from sam-
`task; feedback
`is a difficult
`terminology
`ple users will be helpful during design and testing.
`Menu Maps As the depth of a menu
`tree grows, it be-
`comes
`increasingly
`difficult
`for the user
`to maintain
`a
`sense of position
`in the tree, and a sense of disorientation,
`or of “getting
`lost,” grows. To overcome
`this sense of dis-
`orientation,
`some menu systems come with a printed
`in-
`dex of terms
`that
`is easier to scan than a series of screen
`displays. The British Prestel system offers a detailed cross
`referenced
`index
`that
`in 1982 was 34 pages long and con-
`tained
`thousands
`of entries. The CompuServe
`Informa-
`tion Service’s November
`1984
`index contained
`almost
`1000 subjects.
`It included
`a diagram, or map, of the first 3
`levels of the tree structure, which contains 26 menus.
`The relative merits of a map and an index were studied
`in a small menu structure with 18 animals as target
`items
`[ 161. In this case, users who had the chance
`to study an in-
`dex did somewhat better
`than a control group that had no
`special navigation
`aids. The group with an overall map
`did substantially
`better
`than both
`the index and control
`groups:
`
`Number of subjects
`Mean time per search
`Mean choices per search
`
`Control
`
`10
`35.3
`12.3
`
`Index
`
`8
`30.7
`8.4
`
`Map
`
`8
`19.2
`4.7
`
`w as
`
`(3-3) menu
`
`the menus
`
`three-item
`for a three-level
`learning
`Menu
`studied with four forms of training
`[17]:
`Online exploration. Subjects could explore
`online.
`sequences.
`Command
`the 27 paths
`studied
`Subjects
`typed on paper, e.g., Plans Division, Concepts, Sys-
`tems Analyst
`Frames. Subjects
`Plans Division
`Concepts
`Designs
`Proposals
`Menu map. Subjects
`of the 13 frames.
`train-
`subjects had a 12 minute
`The 65 undergraduate
`ing period,
`followed by a 10 minute work period. The re-
`sults indicate
`a strong advantage
`for those who had the
`menu map:
`
`studied
`
`the 13 menu frames, such as
`
`studied
`
`a tree-structured
`
`layout
`
`Online
`Explora-
`tion
`
`8.2
`
`10.6
`10.1
`
`3.6
`
`Command
`Sequences
`Study
`4.7
`20.4
`
`8.4
`
`3.1
`
`Frames
`Study
`6.5
`19.6
`9.8
`
`2.8
`
`MCIlU
`Map
`Study
`
`8.5
`9.4
`16.7
`
`4.8
`
`Sig-
`nifi-
`cance
`
`N.S.
`p <
`.lO
`p < .05
`
`p <
`
`.Ol
`
`Targets found
`Average number of
`menus visited
`Recall of tree
`(max = 27)
`Satisfaction
`(best = 5)
`
`As the tree structure grows, users have greater difficulty
`in maintaining
`an overall understanding
`of the semantic
`
`62
`
`JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE-March
`
`1986
`
`

`

`organization. Viewing the structure one menu at a time is
`like seeing the world through
`a cardboard
`tube:
`it’s hard
`to grasp
`the overall pattern
`and see relationships
`among
`categories. Offering a spatial map can help overcome
`this
`difficulty.
`Semantic Versus Alphabetic Organization Since the
`creation of a universally
`acceptable
`semantic decomposi-
`tion with several tree levels is a challenge,
`some designers
`have attempted
`an
`index
`strategy
`that provides
`a tree
`structure based on simple alphabetic
`organization
`of the
`target
`items.
`in the use of a
`tested
`There were 30 student volunteers
`tree-structured
`database of 470 index pages and an alpha-
`betic
`index of 453 terms
`[18]. A typical
`search of both
`forms proceeded
`as follows:
`
`User sees:
`
`User selects 2 and sees:
`
`Tree Search for Baseball Scores
`1. News
`2. sports
`3. Entertainment
`SPORTS
`1. Hockey
`2. Baseball
`3.
`. . .
`BASEBALL
`1. Scores
`2. Standings
`3.
`. . .
`BASEBALL SCORES
`
`User selects 2 and sees:
`
`User selects 1 and sees:
`
`User selects 1 and sees:
`
`Alphabetic Search for Baseball Scores
`1. A-B-C
`User sees:
`2. D-E
`3.
`. . .
`1. A
`2. B
`3. c
`DIRECTORY
`. . .
`. . . . . . . .22114
`Banks
`Baseball.
`. . . . .221313
`Books.
`. . . . . . . . .2516
`
`User selects 2 and sees:
`
`FOR B
`
`User keys 221313 and sees:
`
`User selects 1 and sees:
`
`BASEBALL
`1. Scores
`2. Standings
`3.
`. . .
`BASEBALL SCORES
`
`half
`design,
`this counterbalanced-within-subjects
`In
`the subjects began with one method and then tried the sec-
`ond method, while the other half of the subjects worked in
`the opposite order. Subjects performed
`20 searches, and
`no significant differences were found
`in mean search time,
`number of keypresses, or number of menus accessed. Un-
`der both conditions,
`subjects
`“required
`about
`twice the
`
`to find the informa-
`number of pages necessary
`minimum
`tion” and “they made one or more errors on 40% of the
`questions.” Users eventually
`succeeded
`in 98.7% of the
`questions,
`so that, although performance was far from op-
`timal,
`successful
`searching was possible with both meth-
`ods. Subjective evaluations
`did not favor one method over
`the other, but when one method
`required more pages for a
`specific question,
`the preference was for
`the
`shorter
`method.
`that offering
`[ 181 conjectured
`and McEwen
`Tombaugh
`subjects
`can
`both methods may be the best resolution:
`for each ques-
`choose the method
`that
`is most appealing
`leads to difficulty,
`then the users can
`tion.
`If one method
`try the other.
`Summary There
`that
`structure
`is no perfect menu
`do-
`matches every person’s knowledge of the application
`initial
`main. Designers must use good judgment
`for the
`implementation
`but
`then be receptive
`to suggested
`im-
`provements
`and empirical data. Users will gradually gain
`familiarity,
`even with extremely
`complex
`tree structures,
`and will be increasingly
`successful
`in locating
`required
`items.
`
`D. Acyclic and Cyclic Menu Networks
`some-
`Although
`tree structures
`are very appealing,
`For ex-
`times network
`structures
`are more appropriate.
`ample,
`it might make sense to provide access to banking
`information
`from both
`the financial
`and consumer parts
`of a tree structure. A second motivation
`for networks
`is
`that it may be desirable
`to permit paths between disparate
`sections of a tree rather
`than
`requiring
`users to begin a
`new search from the main menu. These and other condi-
`tions
`lead to network
`structures
`in the form of acyclic or
`even cyclic graphs. As users move from trees to acyclic net-
`works
`to cyclic networks,
`the potential
`for getting
`lost
`increases.
`the user can form a mental model
`With a tree structure,
`of the structure
`and
`the relationship
`among
`the menus.
`Developing
`this mental model may be more difficult with
`a network. With a tree structure,
`there
`is a single parent
`menu,
`so backward
`traversals
`toward
`the main menu are
`straightforward.
`In networ