Chapter 13
The Future of Menu Selection
Some have suggested that menus are merely an interim solution to the problem of human/computer interaction. Menus are regarded as only a stage in the metamorphosis of novice users into expert users. Or perhaps they are a way station on the road between command language and direct manipulation of the "soft machine," a malleable interface that takes on the shape and characteristics of the device that it emulates.
Some may feel that research should wait until the interface settles down. Findings today may not be relevant tomorrow. However, it is argued that although the implementation of the human/computer interface may change dramatically in the future, the theory and principles of cognitive control grounded in the user will not. While the technology driving the computer display and input devices undergoes evolution, the psychological processes driving user behavior is fixed. Although users vary among themselves, learn new ways of doing things, and adapt to the interface over time, they basically come in only one configuration with no revisions.
Furthermore, the concept of menu selection and underlaying cognitive processing is not likely to be supplanted by a new technology but rather further entrenched in new designs. Whatever metaphor one projects, whether the restaurant menu, the desktop, soft machines, or beyond, all involve the organization and activation of objects and require the fundamental processes of attention, comprehension, selection, choice, traversal of functions, and search on the part of the user to establish cognitive control over the interface.
Rather than diminish, it is expected that the impact of menu design will increase in the next generation of systems. Furthermore, the need for well designed menus will increase as more of our work becomes computer mediated and as more of our files become computer accessible. The complexity of the computer environment will approach, if not surpass, that of the physical environment in which we operate. Advanced menu selection, navigation, and direct manipulation are required for the user to operate with control in such an environment.
In this chapter a number of new directions in menu selection will be explored. Projecting the future development of a technology is not without its embarrassing pitfalls. Nevertheless, it is worth taking the shot. It is expected that menu systems will move far beyond the current generation as high performance workstations become the norm. Innovations in menu selection design will continue to reshape the human/computer interface. A number of new types of menus will be discussed. It is expected that menu systems will be provide a number of new capabilities to the user in an effort to enhance their power and adaptability. Systems will continue to grapple with the problem of when to use menu selection as the mode of interaction, but it is expected that systems will provide a better integration among alternative modes of interaction. Finally, one must think about the future of research on menu selection. A number of fundamental research issues are yet outstanding that need to be addressed in the coming years.
To project the future it is instructive to plot the past. Menu selection has come through several generations up to this point in time. Surveying a number of systems, one can date the software by the stereotypic way in which menus are implemented. Listed below are three generations of menus. As is the case with most technologies, a later generation rarely supplants a former generation, but rather it coexists for an extended period of time. Furthermore, one cannot conclude that later generations are necessarily better than former generations. Older tools may get the job done just as effectively as new ones.
First Generation: Menus are characterized as lists of options. They have been displayed one by one on slow terminals or rapidly on a fast workstations and personal computers. Selection is generally by a keyboard response such as a letter or number or function key. The first generation is typified by videotext services, electronic bulletin boards, and programs designed for novices.
Second Generation: Menus are characterized by pull-down and pop-up lists. Selection is generally made by cursor keys, a mouse, or by touch screen. The second generation is typified by applications such as WordPerfect and Lotus 1-2-3. The second generation sought greater speed and ease of access, thus securing a greater market among experienced users.
Third Generation: Menus are characterized by selecting and moving icons representing objects or tools. Selection is made by clicking, double clicking, and dragging objects using a mouse or some other pointing device. The third generation is typified by the desk top metaphor and direct manipulation.
Several trends are discernible in the progression from one generation to the next. The first trend has been increased speed and flexibility. Faster user response time, alternative selection methods, and increased number of items were the driving force of leading to the second generation. The second trend has been the addition of graphics and metaphor to the selection process. The third generation was motivated by an analysis of the user task and the need for a usable mental model to support the task.
Looking to the fourth generation, there are three formidable problems that must overcome. First, it must provide fast access to extremely large numbers of items. Second, it must allow complex formulations of processes and commands. Third, it must provide new methods for user customization. The next section explores some of the innovations that hope to make this possible.
13.2 Innovations in Menu Look and Feel
Menu interfaces convey a sense of look and feel. Users immediately see what they look like and quickly gain a feeling for how they work. But it is more than a superficial issue. The "look" of menu displays must convey choice information in a concise, information rich form that allows the user to efficiently scan and select items with a minimum number of errors. The "feel" of a menu must allow the user access to large numbers of objects with minimal movement and effort on the part of the user.
A number of innovations are being explored which increase the field of view of menu options and extend the power of selection to the user. These innovations are exciting, but it will remain to be seen which approaches accomplish that goal in a way that is compatible principles of cognitive control.
13.2.1 Nonlinear/Spatial Menus. Conventional menus list items in a linear array. A list has the characteristics of having a beginning, an end, and hopefully a meaningful ordering among the items. The beginning and end items are special and serve as anchors. Other items have president,/successor relationships that help the user to locate items and structure the list. Many sets take advantage of linear order, but many other cases suggest nonlinear layouts of items that convey different relationships among items.
Rectangular layouts convey multiple dimensions and classes of related items. In addition to the row and column location (a linear ordering), rectangular menus convey similarity and neighbor relations (nonlinear diagonal orderings). For example, the periodic table of elements (see Figure 13.1) arranges items in rows according to their period (number of shells of electrons) and according to their atomic number (number of protons). Vertical columns of elements are related groups of elements that have similar properties.
Other sets of items are best represented with a cyclical order, for example, hours on a clock face or months of the year. Such orders convey cyclical orders in which, for example, January comes both before and after December. Circular menus can convey other relationships too. The color wheel conveys complementary relationships between colors. The circle of fifths conveys multiple sequences of whole-tone and semi-tone scales. Figure 13.1 displays examples of such nonlinear/spatial layouts of menus.
In other cases, items have a traditional or standard graphic layout that is familiar to the user. Graphic menus faithfully represent these layouts on the screen. Figure 13.3 illustrates the layout of the states and zip codes. Other examples abound in cartography, schematics, illustrations, and diagrams that can be used to organize menu items in a way that uses and strengthens the user's prior visual organization of information.
Nonlinear /spatial menus have two advantages. Visual recognition and spatial memory is a powerful and efficient means that users have for locating of items on the screen. The graphic representation conveys a context for choice and structure for visual search.
A second advantage is that nonlinear /spatial menus may be arranged to facilitate selection time. Research on pie menus demonstrates the reduction in selection time that can be achieved by reducing the mouse or cursor movement distance from home to the desired item (Callahan, Hopkins, Weiser, & Shneiderman, 1988). In this case the starting point for the mouse was in the middle of the circle. Slight movement in any direction served to highlight an item for selection. Rectangular and other spatial layouts may also have an advantage in that they allow for diagonal movement across items. Linear menus, such a pull-down menus, only allow for city block movement in which the user can only move across or down at a time. Finally, nonlinear/spatial menus may be able to make the most efficient use of space. Response time is also a function of the physical size of the target. Important or frequently accessed items can be given a larger selection area.
Nonlinear/spatial menus are beginning to have a great impact in human/computer interaction as graphic interfaces become available. Despite their availability, a potential designer should show some restraint in their use. It is easy to overdo a good thing and confuse the user with unfamiliar and complex graphics. There must be a balance of simplicity and functionality.
13.2.2 Analog Menus. Input of numerical values has been a constant problem in human/computer interaction. Values have had to be input via the keyboard or numeric keypad. Menu selection has not been an acceptable mode since the selection of digits on the screen is no better than key input. However, with pointing devices such as the mouse or touchscreen, analog menus allow the user to input apparent continuous values. This innovation has two forms. One is accomplished by moving the cursor and clicking on a scale. The other is done by using the mouse to click on and drag a marker along a scale. Generally feedback is given to the user. The feedback may be a numeric display or a spatially translated effect such as seeing a geometric form rotate or a manuscript scroll to a relative position. One of the obvious difficulties with analog menus is granularity. One solution has been to shift from macro movement to micro movement of the mouse using a gear shift key to change the ratio of translation.
Analog menus are rapidly being incorporated in the next generation of menu design and are replacing other modes of input. It is expected that they will fulfill an important role in increasing the cognitive control of users over the interface.
13.2.3 Power Pointing. Current selection devices provide only a limited channel for selection. The keyboard is limited by the number of keys. Cursor/arrow keys are limited by the number of key presses required. The mouse and other input devices are more powerful and versatile for selection by pointing since they provide a direct spatial mapping of x-y hand movement to the selection of items laid out on a 2-dimensional screen. However, the movement capabilities of the user are not limited to two degrees of freedom. Maximally, the user has six degrees of freedom in which to move a pointing device: the 3-dimensional x, y, and z coordinates and the axises of pitch, roll, and yaw rotation. Hand controllers and data gloves have the capability of sensing the full compliment of movement.
To a large extent this movement can be mapped to the screen using three dimensional representations in space. One mouse dimension may be used to zoom through layers and the other two to scan across. The axises of rotation may be used to rotate objects on the screen. Selectable items may graphically displayed on the surface of three dimensional objects (e.g., rotating a lunar globe to select landing sites).
Added degrees of freedom may also allow the user to input gestures. Gestures are stereotyped movements that have specific meaning. For example, a squiggle or twist over an item on the screen may be used to delete the item, a "Q" shaped movement may be used to quit an application, a twist in one direction or another may sling an item in different directions. Exploration of these features may add tremendous power to many applications that support high resolution graphics.
13.2.4 Apparent Menus. Menu systems that involve control of processes and functions generally have a build-in interdependency among items. For example, Function C cannot be selected until either Function A or B have been selected. Hierarchical menus build this interdependency into the structure and present items only when they are appropriate for selection. The inherent interdependency of items, however, is buried in the hierarchy menus and is not always apparent to the user. Menus which display all functions at the same time use some mechanism such as graying out items to indicate that they are not active given the current state. Unfortunately, the interdependency that activates or de-activates them is generally not apparent to the user. The user is left to hunt forever, it would seem, to find the right combination to activate a desired item.
A major innovation is needed in making the inherent interdependency among items apparent to the user. One possibility would be to have the user select de-activated items in order to access context dependent help information about that item. Such information might inform the user as to what is necessary to activate the item. For example, selecting the function "Align Objects" would inform the user that two or more objects must first be selected in order to perform the function. This approach may be sufficient for simple interdependencies but not for complex network relationships.
A second approach would be to graphically display relationships using a schematic diagram. Figure 13.4 shows a hierarchy of functions and how a static diagram (second panel) might convey some of the relationships. Unfortunately, this diagram is complex and insufficient.
A third approach is a combination of the first two. The user would select a de-activated item which would then display a trace back through the items in the network showing possible paths to achieve that function (see the third panel of Figure 13.4). If the user wants to perform several functions in a series, the desired functions could be selected and the system would trace a route through a necessary series of functions to perform the desired subset of functions.
The point is to find ways of using the dynamics and graphics of the interface to reveal the structure inherent in the system to the user. This is a challenge to designers; however, it is essential if the menu interface is to provide true transparency and ease of use.
13.2.5 Simultaneous/Linked Menus. Conventional menus generally incorporate only one menu hierarchy. However, it happens that often one needs to traverse the same menu tree in two or more independent passes or one needs to traverse two or more different menus in linked passes. The first case is desirable if one needs to compare or to link different nodes of the same menu tree. For example, if one wants to compare different citations in bibliographic database or to create cross references. One innovation would be to open multiple windows to a menu system and to provide tools to perform link, copy, compare functions between them.
The second case is desirable if traversing one menu is related to traversing another menu. Traversing a menu driven help system provides information to the user about traversing a menu control system. In control systems, operations in one menu system may conditionalize a second menu system. Furthermore, menu systems may be embedded in other menu systems. This sort of complexity will be required if menu systems are to perform complex jobs in command formulation and programming.
Finally, menu systems may be used to alter or adapt other menus. User adaptable menus are an important innovation as seen in a following section, but to implement adaptability, additional menus may be required to operate on top of the menu being adapted.
13.2.6 Vast and Fast Menus. Research on menu depth and breadth suggests that broad organized menus result in faster performance and fewer errors than deep menus. The problem is that in order to access data bases with large numbers of items, depth has been required. Advances in large, high resolution screens, however, makes it possible to display menus that have 50-200 items per screen. The MacIntosh(TM) Finder program, for example, can display over 50 folders in its hierarchical file system. Pull-down menus can display 150 or more options. Numbers such as these can provide access to 125,000-8,000,000 items with only three menu selections. If screens are graphically organized, selection times per screen may be relatively fast.
The difficulty in achieving vast and fast menus is in generating menu structure and graphic layout of vast menu frames that conveys the inherent structure of the database. This can be an extremely laborious task, particularly at the lower levels of the hierarchy. Furthermore, it may be in vein since many menu frames may be only rarely visited. Some systems may overcome this problem by providing a self organizing algorithm. Others may allow the user to organize and even to prune menu frames based on need. Innovations are clearly needed to help the user organize the growing mass of documents and tools in the electronic media.
13.2.7 User-Constructed/User-Adaptable Menus. Storage and retrieval of information is often handled via user constructed menus. Users may create folders and using direct manipulation or conventional menus store files in the folders. Such hierarchical filing systems are essentially menu driven but are constructed by the user.
Similarly, users may organize tools or functions for access via menus. And to an increasing extent software designers are giving users the ability to customize or build their own menu structures. On the one side, this ability allows the user to create an efficient system tailored to their needs. On the other side, it places a burden on the user to become the designer. The trade-off is an important one and users must gauge the extent to which they want to invest their own time constructing or adapting a menu system versus using a fixed off-the-rack system. To make matters worse, user design may result in a morass of redundancy, over-complexity, and sheer idiosyncrasy. Users may not be good designers. They will require considerable guidance to avoid irrational design.
The challenge is to design a user-designable system that avoids the pitfalls of bad design. Such a system must impose the principles of good design in the same way that a syntax-directed editor imposes correct statement construction on programmers. For example, a menu builder system may assess the discriminability of item labels, the organization of items in the frame, and the structure of the network and alert the user/designer when problems occur.
User adaptable menus provide five primary tools for redesign as listed below:
Rename. Users should be able to change the name or other designating information of menu items so as to add to its meaningfulness.
Delete. Users should be able to delete items. Deletion may be of terminal items or whole branches in a hierarchy. Pruning the menu system of unused items can greatly simplify the interaction.
New Node. Users should be able to create new menu items or frames that will serve as new groupings of items.
Copy. Users should be able to create copies of items. Copies may be of terminal items or whole branches in a hierarchy. Copies can be placed anywhere in the menu system to provide multiple access to frequently needed items at different locations in the system.
Link. Users should be able to create links from any item in the menu system to any other item. Links provide new pathways that can be used to create fast access to frequently used items or to link related items.
Construction of customized menus can also be accomplished by the tools listed above. The difference is that the user would begin with a library of items that could be assembled into a custom menu. The tool kit would first have to provide users with access to the library of items. Second, it must provide tools for organizing, editing, and assembling menus. Figure 13.5 shows a schematic of such a process. The left side of the figure shows the library access of objects available to the user. Tasks will not be performed using the library menu structure. It is only used when the user is selecting a menu or an item for inclusion in the user-designed menu system. Once the menu or item is located via navigation through the library, it is picked and marked for inclusion in the new menu. On the right side of the figure, the user builds a menu structure for a particular task. The user locates the desired point in the new system and puts the selected menu or item at that point. Other editing features are needed to create new menu frames (nodes), new links between items, and clusters of items. Working menus are pieced together and customized from stock software.
User-constructed and user-adaptable menus allow users complete freedom to do what they want. But freedom can be time consuming and full of pitfalls. User-constructed and user-adaptable menus require a great amount of time, effort, and knowledge on the part of users. It may unreasonable to expect users to bear such a great burden of design. Ultimately, it will again be the designer and the human/computer interaction specialist who will be the one to use the menu tool kit.
The number and variety of interaction styles available to the designer for human/computer interaction is steadily increasing. Command languages have long dominated human/computer interaction and have provided a powerful but error prone environment to users willing to spend the time to learn the syntax and semantics of the language. Form fill as an interaction style has been implemented for low level data entry tasks. Although it minimizes learning time, it lacks power and in most cases flexibility. Menu selection has an advantage over command language in that it greatly reduces the amount of syntactic and semantic knowledge required on the part of the user. It has an advantage over form fill when data entry is limited to a relatively small set of values. However, menu selection can't do it all. There are times when command language is clearly required and times when keyboard and form fill are more efficient.
The question is, "What is the optimal interaction style?" The answer depends on the type of user, the degree of cognitive control required, the point in the dialogue, and the type of task being performed. Time critical situations may call for a high degree of user-directed control. Passive learning situations may call for a low degree of user-directed control and active learning situations may require a constant shifting of control between the user and the system. As experience, mental workload, stress, and time pressures change one expects corresponding changes in the way the user interacts with the system. Consequently, interaction is not static. There are ever changing demands on the user and the system.
If interaction is not static, then it stands to reason that the mode and form of that interaction should not be fixed. The interface should be constructed so as to facilitate the changing needs of the user and the task. The appropriateness of menus as the mode of interaction will vary.
13.3.1 Early in the Learning Process. It is often argued that menus are most appropriate for novice users, but that experienced users prefer command languages. Certainly, menus are an aid to the novice in that they do not need to recall commands. However, it must be remembered that users do not remain novices for long and there may be a point at which the user may wish to switch to command input. The question is whether users starting with a menu selection system are hampered later on in learning a command language. If there is negative transfer from menus to command language, it would suggest that users should not start with menu selection. On the other hand, there may be a graceful transition from one mode to the other.
Streitz (1987) provides empirical evidence that menus do provide positive transfer to command language learning. In an experiment one group gained experience using a menu selection system for editing text. A second group gained experience using a command language with a help window for editing text. Two additional groups were given no prior experience. All four groups then participated in the learning phase in which texts were edited using the command language only, with menus, or the help window as a help device. Learning curves were assessed for a limited set of commands and subjects were tested on their knowledge over the all of the commands available. No difference was found in the slope of the acquisition curves for the two groups with prior experience on menus versus commands. However, the menu group fared better in their overall knowledge of the functionality of the system. Subjects in the command language condition learned only a limited number task specific functions. It would appear that menu selection promoted a greater breadth of system knowledge.
Menus are not only appropriate for novice users who are expected to remain novices, they are also appropriate for novice users who are expected to become experienced users. This suggests a radical rethinking of many user training programs. Rather than starting training on the command language, users may begin working productively using menus. This allows them to actively gain an understanding of the functionality of the system through its use. Commands may be introduced at later stages as users master the concepts. This approach encourages active exploration on the part of the user and may promote a deeper understanding of system functionality and a broader range of use.
13.3.2 At Transition Points. Human/computer interaction is often characterized by intense periods of focused interaction (e.g., designing in a graphics application or entering data in a text editor) and transitions from one task to another (e.g., terminating one application and moving to another). Transition points involve important decisions and actions (e.g., what to name a file, what to do next, where to locate the next file). Such points are characterized by changes in the type and degree of cognitive control.
Exiting and entering applications require considerable structured control. The initiation and shut down of a system may be a demanding task involving check lists and sequencing of events. Menus are likely to aid in structuring the task and providing user guidance.
13.3.3 At Different Mental Workloads. Menu selection is often viewed as a more casual, relaxed mode of interaction; whereas, command language is a formal, fast paced mode used under stress. When the all the alarms are going off, users may not be in the appropriate mental state to wait for a menu display, scan the options, and make selections. Instead they want to blurt out commands as fast as possible to control the situation. Menus that are several layers deep are a particular problem. Broad menus may display the needed items, but the under stress the perceptual scanning and selection process may be impaired. Menus may accommodate stress by providing large, dominant alternatives that can serve as panic buttons or they can allow the user to input direct commands.
Mental workload, stress, and other situational factors vary greatly during interactive sessions using a computer. How should mode vary during such changes? A study by Eberleh, Korfmoacher, and Streitz (1987) and reported by Streitz (1987) investigated the use of direct manipulation versus command modes under different levels of mental workload. The task was to work with a graphics program. Sixty subjects learned both a complex but fast method of interaction using a command code and a simple but slow method using direct manipulation. After all of the subjects met the learning requirement, one group of 15 subjects had to use only the command code, another group of 15 had to use direct manipulation, and a third group of 30 subjects had the choice of either method. The situation was varied in the same way for all three groups. They were subjected to a time pressure condition, an office noise condition, and a neutral control condition. Mental workload was measured using a dual task paradigm in which the user performs an primary task but has to respond to a secondary task when it is presented. Reaction time to the secondary task and the number of omissions gives a measure of mental workload on the primary task (Wickens, 1984). It was found that mental workload was not closely tied to any one mode but depended on the compatibility of the interaction mode and the situation. Overall, there were more omissions with command language than with direct manipulation; however, under time pressure, command language led to fewer omissions that direct manipulation. It would appear that when working quickly under time pressure, command language requires less mental effort than direct manipulation. Results from the third group indicated that under time pressure subjects chose to use the command mode (68%) significantly more often than direct manipulation (32%). The implication is that under time pressure, menu selection may not be as appropriate as command codes. This result must be taken with caution; however, as much depends on the way in which direct manipulation is implemented.
Perhaps the best solution in terms of cognitive control is to provide users with both modes of interaction and let them decide which is appropriate. A further analysis of the third group indicated that 60% of the subjects switched between the command and direct manipulation modes depending on the situation while 20% primarily used the command mode and 20% primarily used direct manipulation.
13.3.4 Flexibility. It may be that one cannot direct design toward the "best" mode of interaction. Instead, the interface might allow users to individually arrange their interaction style, sequence of subtasks, and screen layouts so to best suite their needs and preferences. Ulich (1987) extends the "principle of differential work design" applied in organizational psychology to human/computer interaction. The point is that different people may for different reasons prefer different tasks or work structures. The "principle of differential work design" is that there should be an optimal development of personality in interaction with the work activity in the context of individual differences. Consequently, he suggests that dialogue sequences in software should leave various modes of procedure open. The program sequence should have as few preset stages as possible allowing users to work in a way best suited them.
Menu systems have varied tremendously in terms of how much flexibility has been afforded to the user. At one extreme menus may present a rigid, fixed flow of interaction. All work proceeds in a predetermined sequence of menus and the user cannot vary that order. In some cases, such structure may be required by the logic of the task. In others it may be capricious, based on an arbitrary selection by the designer of one of many possible ways of doing the task. Moreover, the interaction may be unforgiving, in that if one component is incorrect, it is difficult if not impossible, for the user to correct the error.
At the other extreme, menus may provide extreme flexibility. Pull-down event driven menus may allow the user to perform a set of procedures in any order. Errors may be corrected at any point by retracing back to the point of error. Menus themselves may be open to modification and restructuring by the user. Such a system allows users to set up the organization and sequence of menus according their preferences and work styles.
A study by Aschwanden and Zimmermann (1984) cited by Ulich (1987) supports the contention that such flexibility is desirable. Two groups of 15 female subjects were compared using two different types of dialogue to perform the task of handling a customer order. One type was highly rigid and the other highly flexible. The flexible dialogue allowed the user to control screen layout, choice of procedures and modify keyboard functionality. No significance difference was found in the time to perform the task; however, the rigid dialogue resulted in a greater number of gross errors than the flexible dialogue. Furthermore, it was found that (a) users did in fact make use of the greater flexibility afforded them by the dialogue; (b) the increased use of the flexibility did not result in decreased performance or additional strain; and perhaps most importantly, (c) the individualized work procedures stimulated the user to become innovative in that they offered numerous suggestions for improvement of the system.
It is obvious that interest in menu selection has generated considerable amount of research. The content and direction of this research reveals the major issues and emphases. Although some lines of research have dealt with fundamental questions, much of the research has had a tendency to dwell on detail. Detail is important since seemingly minor design issues can have large effects on performance and subjective impressions. There is no end of additional design details that need to be studied. Moreover, whole new cases of details will need to be addressed as new innovations in menu design are explored.
The results of these studies can be divided into two classes: enduring results and technology dependent results. The former are more or less independent of technological advances. They deal with the general principles that are cut across all designs (e.g., complexity, functionality) or they involve psychological processes in the user that will always be the human part of the interface (e.g., mental models, visual scanning).
On the other hand, technology dependent results will be obsolete when the technology changes. This is not to say that this type of research lacks importance, only that its value is time dependent. Such results often help to drive technological improvement. For example, the work on screen readability of older VDUs has helped to motivate the development of high resolution workstations.
It is anticipated that great changes are in the wings for menu selection. Research is needed more than ever to support this development.
What is particularly exciting is that research on human/computer interaction contributes to the development of cognitive psychology in an unprecedented way. The human/computer interface taxes all of the cognitive aspects of the human and requires a greater understanding human processing capacities than ever before. Research in this area will consequently be doubly fruitful, adding both to the technology and to the basic understanding of human cognitive processing. Menu selection in particular will contribute to theories of attention, problem solving, decision making, memory, and intelligence.
Menus have come a long way. But there is still a long way to go before menu selection as a mode of interaction reaches its maximum potential. The thrust of current and foreseen innovation is to provide a control surface to the user that (a) is spatially organized, (b) accommodates selection in both discrete and continuous sets, (c) conveys visually apparent relationships, (d) provides maximum degrees of freedom in selection, (e) is vast and fast, (f) allows complex forms, and (g) is user adaptable.
Advancements in these areas has already placed menu selection as prime candidate for control of the human/computer interface. Furthermore, trends in the development of different modes of interaction suggest that menu selection is headed in the right direction. The development of command language as a mode of interaction came first from within the computer community and has only recently been extended to the user in market place by simplification, training, and sometimes use of natural language. It has moved from a versatile and powerful mode to limited application. In contrast, menu selection was developed first for the novice user and later extended to the more experienced and demanding computer literate community. Its versatility and power are constantly being enhanced.
Moreover, interfaces have in the past been characterized as being of one mode or another. Currently, applications find it necessary to shift modes as needed. The intelligent construction of the human/computer interface makes the best use of each mode as tasks shift between sequencing and control on the part of the user to computer directed requests for data input. Menu selection is perhaps the most versatile of modes for this purpose, accommodating novice to experienced users, running at low to high workloads, and providing transitions between tasks and environments. Consequently, designers are using menu selection as the principle base of operations; and only when needed, do they digress to other modes.
In conclusion, it must be realized that the human/computer interface is a rapidly changing surface. It is not clear what it will ultimately be. But it is becoming clear that it is extremely important. Indeed, the balance of power of the computer rests at the human/computer interface. In the years to come research and development of the interface will be a major thrust in bringing about the next wave in computer revolution of society. Menu selection is on the crest of that wave.