The following is a compilation of articles, published between 1995 and 1997, which evaluate and compare screen readers for Windows. Since adaptive technology in this area has been continually developing, I encourage readers to seek more current information before drawing conclusions. This material is intended primarily as a way of introducing concepts of accessibility testing and suggesting questions to consider when selecting a screen reader. (Each article is separated by a line of 10 dashes and a hard page break, as below.) Jamal Mazrui ---------- In July 1995, the National Federation of the Blind released a unique set of consumer reviews it had compiled, covering most screen readers on the market today. This document summarizes the complete text by excerpting the conclusions and contact information sections related to each product, as well as the cross-product conceptual sections. I have divided it into two messages: the first on DOS screen readers and the second on graphical ones. Distributed by Jamal Mazrui National Council on Disability Email: 74444.1076@compuserve.com ---------- SCREEN REVIEW PROGRAMS: COMPREHENSIVE REVIEWS OF SPEECH ACCESS PROGRAMS FOR THE BLIND Produced by the International Braille and Technology Center for the Blind National Federation of the Blind David Andrews, Editor National Center for the Blind 1800 Johnson Street Baltimore, MD 21230 Telephone (410) 659-9314 BBS (410) 752-5011 Copyrightū by the National Federation of the Blind, 1995 INTRODUCTION With over twenty choices on the market, the decision of which screen review program to purchase is an important and difficult one for any blind or visually impaired computer user. Further, the new user isn't going to know which features are important and which features are not; he/she is not even going to know what questions to ask. This document will provide some background information and provide reviews of all of the commonly encountered screen review programs which are currently available for IBM and compatible computers running MS-DOS, Microsoft Windows Version 3.1, IBM's OS/2 and Warp, as well as the Macintosh running Apple's System 7.x. WHAT IS A SCREEN REVIEW PROGRAM? The piece of software we will be discussing throughout this document is generically known as a "screen review program," a "screen reading program," or a "screen reader." We will refer to it as a "screen review program" since we feel that this is the clearest and most unambiguous reference to avoid confusion with a specific product that uses the "Screen Reader" phrase in its name. Essentially, a screen review program is the software that controls the speech synthesizer and allows a blind person to use a computer interactively in real time to run most off-the-shelf commercially available applications. Two components are needed for a blind person to use a computer. The first component is the speech synthesizer (sometimes called the voice synthesizer). This is the actual hardware that produces the voice you hear. It can be either internal or external, in the case of IBM and compatible computers; that is, it is an expansion circuit card that fits into one of the expansion slots inside your computer, or it is an external unit that resides outside your computer. External synthesizers are attached to your computer via either a serial or parallel connection. A serial connection is the most common by far. In the case of the Macintosh, the speech is produced by a chip which is built into the computer. The second component needed in a screen review program is a memory-resident program, also called a TSR (terminate and stay resident) program. That is, it is a program that is run once, establishes itself, and returns control to the computer. It is, however, still present and operating so that it can do its work. A screen review program does a number of things: 1) It allows the user to control the synthesizer: change volume, pitch, and speed of the voice, as well as to shut up the synthesizer. Do not underestimate the value of the shut up or quiet command; you will use it more than any other. How quickly it acts is in part a function of the speech synthesizer itself and in part a function of the screen review program. It is, however, a very important consideration. 2) It gives you a set of commands to read and re-read the screen. You can think of the synthesizer as an audio printer. It prints or says all output from the computer. However, once it says something, the text is gone. This makes it difficult to use an application such as a word processor interactively. Thus, the screen review program gives you a set of commands to read by character, word, line, sentence, paragraph, screen, page, and/or document. Not all programs have all commands, but all screen review programs will allow you to re-read the screen by a variety of text units that you can choose. 3) It has mechanisms for tailoring their output to speak correctly with specific applications. This takes the form of automatic analysis of the screen to determine what should be spoken and when, and/or setting up a so-called configuration that tells the screen review program how to work with a specific application. Most of the programs fall into this latter category; that is, they require manual configuration by the user, the developer, or a third party. However, configurations for most common programs exist and are distributed with the individual screen review programs and/or are available on bulletin board services operated by the developers or others. If further configuration is necessary, the results can be saved and used again, so configuration is generally a one-time process. How Were the Evaluations Performed? In the following evaluations of screen review programs, we are striving for user-oriented perspectives. All of the reviews were done by blind persons who actually use the screen review programs in their day-to-day work. The evaluators were given a set of guidelines and criteria to use when writing their reviews. The resulting reviews were then edited by David Andrews, who is Director of the International Braille and Technology Center for the Blind. As such, Mr. Andrews has used all of the screen review programs on the market and is able to provide a unifying perspective and set of experiences, through which all of the reviews were passed. Sighted assistance was only used in the evaluation of Microsoft Windows programs. This area is so new that, on occasion, things happen on the screen that are not reflected in speech output. A sighted consultant was used to help identify these problem areas and also to assist with the application software tests. Reviewers used the following guidelines: 1. General description: How does the program work? How are commands laid out? What is the logic of the command structure? 2. What speech synthesizers are supported? Is indexing supported, and if so how well does it work? Does it shut up the synthesizer efficiently and quickly? 3. Requirements: Are there any hardware and/or software requirements? Is a particular version of DOS required, etc.? 4. Installation: How do you install the program? Is installation easy? Do you need sighted assistance? Can you do it if you don't already have some speech up? 5. Memory requirements: How much memory does it take? Can it be moved high? Can it run in expanded or extended memory? Can you reduce the memory footprint? Can you remove it from memory? 6. Navigation: How do you navigate around the screen? Discuss reading units, character, word, sentence, line, paragraph, screen, entire document, etc. Is there a review mode? An audio cursor? Can you route the real cursor or the review/audio cursor? 7. Settings: How do you set things like speech rate, tone, pitch, punctuation level, capitalization alert, etc.? 8. Configuring: Do you need to configure? How do you do it? How difficult is it? How do you define windows, silent areas, monitors, etc.? 9. Monitoring: Can the program monitor different areas of the screen and notify you of changes? How do you set it up? How well does it work? What options are there--beeping, chaining windows, etc.? 10. Alternative cursors: Can the program track an alternate cursor, highlight bar, ASCII character, soft cursor, etc.? How do you set the feature up? How well does it work? 11. Attributes: How do you identify attributes like background and foreground color, extended ASCII characters, highlighted text, bold and/or underlined text, etc.? How can you use this information? 12. Searching: Can you search for things on the screen, character strings, attributes, etc.? If so, how well does this feature work? What can you do with it? 13. Keyboard: How is the key response? Are there keyboard conflicts, and how can you get around them? Is there a pass-through key? Can you tell if the caps lock key is down? Can you find out the status of other toggle keys? 14. Macros: Does the program have built-in macro capability? If so, describe. Is there provision for macros from an external program? 15. Pre-defined configurations: Does the program come with pre-defined configurations for popular software packages? If so, how many and what? Are there other configurations available via the company, dealers, or a BBS? How well do any available configurations work? 16. Autoloading: Does the program autoload configurations? If so, how does it work, and how well does it work? If not, what other provision is there? 17. Pronunciation: Can you change how things are pronounced? Is there an exceptions dictionary? If so, how does it work, and how well does it work? 18. Responsiveness: How responsive is the software? Does it require a fast machine? Does it eat up lots of CPU cycles? Does it adversely affect other applications? Can it run with a multitasker such as Double DOS or DESQVIEW? 19. Support for sighted people: Do the menus appear on the screen or just in speech? How well does it help a sighted assistant, highlighting review cursor, etc.? Can it easily be shut up if a sighted person must use your computer? 20. Documentation: How is the documentation? What formats does it come in? How good is it? 21. On-line help: Is there on-line help? How does it work? How good is it? Is it context-sensitive? 22. Technical support: Is there technical support and help available from the company? Is it toll-free? How good is it? Please note that the above guidelines are for DOS-based screen review programs. Guidelines for those programs that are used with a Graphical User Interface, i.e. Microsoft Windows, OS/2, and the Macintosh will be listed in a later section. The following persons wrote reviews (all or in part) or provided assistance and advice: David Andrews Mike Freeman Brian Buhrow Steve Jacobson Curtis Chong Lloyd Rasmussen Dr. Tim Cranmer Cathy Schroeder Frank DiPalermo Ellen Waechtler Sharon Duffy Jim Willows Dr. Emerson Foulke Tom Wlodkowski For clarity and uniformity in the text which follows, any commands to be typed verbatim are quoted, e.g., "Backup." Specific keys to be pressed or key sequences to be entered are capitalized without quoting, e.g., ENTER or CTRL-SHIFT. Menu titles are capitalized, e.g., "Options." ----- Artic Vision Version 3.08 Conclusions Artic Vision and Artic Business Vision are widely used programs. They were very popular, particularly in the mid- to late 1980's when they were probably the best things around. They provide good performance and a rich set of features at a reasonable price. There are some good configurations around. The ones for WordPerfect are particularly well done and popular, and there are lots of users from which you can get help. On the negative side the program can be difficult to configure. Artic Technologies has been slow to update the programs in the past several years. They promise things early which are a long time coming if at all, such as macros. They have not kept up with DOS screen review features wars very well lately, and seem to be concentrating on their Windows product and other devices. Contact Information Artic Technologies International, 55 Park Street, Suite 2, Troy, Michigan 48083-2753; telephone (810) 588-7370; fax (810) 588-2650; BBS (810) 588-1424 Technical Support--East Coast (810) 588-1425; Technical Support--West Coast (209) 291-3645 Prices: Artic Vision $395, Business Vision $495, SynPhonix 210 and 215 $395, SynPhonix 315 for Microchannel computers $495, TransPort $895, TransPort with Business Vision $1295. ----- ASAP Conclusions ASAP has made a large impact in the four years it has been around. It was the first program to employ extensively automatic techniques for aiding the user in reading the screen. While not perfect, these techniques are often quite helpful, particularly with software that you use infrequently which has a complex user interface and software for which you don't want to take the time to configure. The software provides good performance and a wide variety of features, particularly with the Double Talk family of synthesizers. The author is responsive, willing to help solve problems, and regularly adds features at user request. This software is updated on a regular basis. Some people consider this to be a plus, while others think it is a minus. If you are interested in keeping up with updates you will need to access the Microtalk BBS regularly. Some of the commands and processes can get a little complicated when configuring applications. Nevertheless, there are lots of configurations around as well as people who can help you. Contact Information Microtalk, 917 Clear Creek Drive, Texarkana, Texas 75503; telephone (903) 832-3471; fax (903) 832-3517; BBS (903) 832-3722; technical support (404) 299-6832 ----- CompuSight Screen Reader Version 4.08F Conclusions Overall, CompuSight Screen Reader isn't a bad program. It can do just about anything a user could want. Nevertheless, it would be difficult to recommend it to most users. It isn't widely distributed, and there are few people in the U.S. with experience using it. There are few, if any, configurations available. The Dolphin synthesizers are mediocre in quality, the program takes a good bit of memory, and some of its functions are awkward to use. Contact Information E.V.A.S., 16 David Avenue, P.O. Box 371, Westerly, Rhode Island 02891; telephone (401) 596-3155; toll-free line (800) 872-3827; fax (401) 596-3979 Price: CSR alone $495; Apollo II, internal, or Juno synthesizers $995 each; one of the synthesizers with CSR $1195 Flipper Version 4.06 Conclusions After having played with a few other screen readers for MS-DOS, this writer believes that Flipper is one of the best designed and most versatile screen readers he's come across. In the time he has used Flipper, there have been very few (less than five) text-based applications that he found difficult to use with Flipper. As he has increased in sophistication, Flipper has grown with him. When he is presented with a new application that he must make work with Flipper, he doesn't say, "Will this work with Flipper?" but "How will this work with Flipper?" Flipper itself has never crashed his machines and has always been extremely responsive, even on the slowest computers. (This is starting to change with regard to the later versions on PC and PC/XT machines). The writer has had no trouble using it with networked workstations, even those machines which were extremely slow (a Toshiba T1200 probably falls in this category). With all of that said, there are a couple of minor inconveniences one might encounter when using Flipper: 1. Flipper configurations must be saved from the DOS command line. When you are developing a new configuration, you might find that you need to exit the application for which the configuration is being developed in order to save the configuration to the hard disk. This can significantly increase the development time of a configuration if your technique is to make small changes and save as you go so you can back them out if they go awry. 2. The number of times a repeat character appears on the screen before Flipper switches from saying it multiple times to counting it is hard-coded at 4. It would be nice if this were a user- definable option. 3. Flipper version 4.06 is noticeably slower on the Toshiba T1200 than was version 2.74. There are a lot of additional features which make it extremely useful; however, it does have a noticeable response lag on very slow machines. This lag does not make it unusable, and we are sure that new Flipper initiates wouldn't notice the drag at all. 4. The 100K or so Flipper requires on the disk from which it is started can be an inconvenience on low density floppy disks. On the Toshibas, which have non-volatile RAM which can be used as a boot disk, it is downright annoying. The fact that Flipper allows you to edit the messages and menus it voices to the user is nice, welcome, and much appreciated. The fact that it requires those files for operation and is unable to work with compiled defaults in the absence of those message files is extremely unfortunate. All in all, Flipper is well worth its asking price of $495 and should provide fast, reliable service for as long as you use text-based applications under MS-DOS. Contact Information Flipper, Omnichron, 1438 Oxford Street, Berkeley, California 04709; telephone (510) 540-6455 ----- IBM Screen Reader for DOS Version 1.2 Contact Information IBM Corporation, Special Needs Systems, 1000 NW 51st St., Boca Raton, Florida 33432; telephone 800-426-4832, technical support 800-426-8637. ----- JAWS Version 2.31 Conclusions JAWS is a powerful and flexible screen review program that usually makes the information displayed on the screen of the monitor easy to acquire when a speech synthesizer is used. It handles troublesome screen displays fairly well when its default settings are used, and it can be customized to provide precise solutions for the problems posed by such displays. Configuration files that prepare JAWS for use with several of the most popular application programs, such as WordPerfect, Lotus 1-2-3, Professional Write, Procomm Plus, First Choice, FormMate, and Arkenstone Easy Scan, are supplied with the JAWS installation disks, and other sets of configuration files may be purchased for little more than the cost of the disks on which they are recorded, or obtained without charge from the electronic bulletin board maintained by Henter-Joyce. Some people have criticized JAWS because they believe that when "Smart Screen" is used with its default settings, it does not always read troublesome screens as effectively as it should. The default settings for "Smart Screen" appear to be a consequence of the principles that have guided Henter-Joyce's approach to the design of the program. For instance, default settings could have been chosen that would cause "Smart Screen" to be more selective in determining what should be spoken, but to do so, it would be necessary to make some additional assumptions concerning what designers of programs do to draw the sighted operator's attention to the significant information on the screen. The problem is that not all software designers take the same approach in designing screen displays. If, in a particular case, the "Smart Screen" assumptions are warranted, the screen will be read more effectively. If they are not warranted, reading of the screen will not be improved and may be less effective. The course taken by Henter-Joyce appears to make "Smart Screen" a little less precise, but more generally useful in its default mode by employing less stringent criteria for determining what should be spoken, and to provide "Smart Screen" options that are easy to learn and use in order to make adjustments on the fly that solve the problems offered by whatever program happens to be running at the time. As has already been mentioned, the JAWS user who needs to make frequent use of a program for which specific solutions to screen reading problems have not been worked out, and who is willing to learn and use the tools provided by JAWS for the solution of such problems, can create configuration files that implement very precise solutions to the screen-reading problems presented by the program in question. Many good screen review programs are now available. In the final analysis, deciding which screen review program to buy may be determined more by personal preference than by objective measures of functionality. Some users have no interest in learning how to find their way around in the screen review program's menu system, or in learning how to construct frames, macros, configuration files, and the like. They are willing to settle for less than perfect functioning, and just want a screen review program that works well enough to meet their needs without fine tuning. Others enjoy learning how to modify the performance of the screen review program and enjoy the challenge of fine tuning it so that it will meet their needs more precisely. Both orientations are defensible, but those who just want to use JAWS would doubtless be pleased if Henter-Joyce made a more conscientious effort to develop and distribute sets of configuration files that tame more popular application programs than the application programs for which sets of configuration files have so far been developed. Self-reliance is unquestionably a virtue, but the reality is that many JAWS users have little interest in developing the kind of self-reliance that would allow them to design precise solutions to screen-reading problems. It seems to us that traditionally, JAWS has had two main groups of users. The first group includes tinkerers with some technical acumen and willingness to experiment, write macros, etc. For these people, JAWS offers lots of grist for their mills. The second group of users is either people who don't care about such things and use applications in the best way they know how, or people who have someone such as a rehab technologist who sets up their system and applications and trains them. JAWS is probably weakest at serving a middle group of users, those people who want and are willing to learn and make some modifications, but who aren't willing to take the time necessary to learn the macro system and the other powerful features of JAWS. The "Smart Screen" and "Smart Focus" with their ability to automatically provide good output are minimizing these problems to a certain extent though. Contact Information Henter-Joyce, Inc., 2100 62nd Avenue North, St. Petersburg, Florida 33701; telephone (813) 528-8900, toll-free (800) 336-5658; fax (813) 528-8901; BBS (813) 528-8903; CompuServe "go JAWS" Price: $495 ----- Master Touch Version 1.32 Conclusions All in all we would have to say that the Keynote Gold PC, Master Touch, and Keysoft products from HumanWare are a mixed bag. All of them have some nice features and some drawbacks. At this time, they are probably best suited for a beginner who wants to write with Keyword and use DOS programs a little. HumanWare promises improvements so the program bears watching. The Touch Tablet is a unique approach and may appeal to some. Master Touch also is easy to use because of its extensive use of menus. Its automatic features, speech sluggishness, and lack of configurations could use some attention. Contact Information HumanWare, Inc., 6245 King Road, Loomis, California 95650; telephone (916) 652-7253; fax (916) 652-7296; toll-free (800) 722-3393; BBS (916) 652-8637. Price: Master Touch software $495, Keynote Gold PC and PCMCIA synthesizers $950, Keynote SA external synthesizer $1245, Touch Tablet $395, and Keysoft software $595. ----- PCMASTER Conclusions PCMASTER is an interesting and unique screen review alternative. It allows a user to employ his/her Blazie Engineering note taking device both as a speech synthesizer and as a computer input device. PCMASTER is straightforward and easy to use. On the negative side, it does not come with any configurations for commonly used software. Some commands involve a number of keystrokes when using Braille entry, and others are impossible to enter, and the program does not have the breadth of features and customization that some screen review programs possess. Further, the program is not updated that often, and some users have reported bugs and other problems with the software. Contact Information Blazie Engineering, Inc., 109 East Jarrettsville Road, Forest Hill, Maryland 21050; telephone (410) 893-9333; fax (410) 836-5040; e-mail: Telnet to blazie.com or write to (first name)@blazie.com where (first name) is the name of the person you wish to reach. Price: $395 ----- PROVOX Version 4.0.05 Conclusions While PROVOX has some interesting features, it isn't for everyone. It offers a relatively low price, a small memory footprint, and efficient resource utilization. If you are running a BBS or other dedicated application, to which you need speech access, PROVOX may be the program for you. When you purchase PROVOX, you can also obtain a copy of the assembly language source code, which may be of interest to new programmers. On the negative side, it does not have all the bells and whistles that other screen review programs have, and it may not have the tools to be configured adequately in complex environments. We had an opportunity to test briefly a beta version of an updated PROVOX, Version 6.1. It used somewhat less memory, 28K, although this is likely to increase as features are added. It had the ability to read pop-up boxes automatically as they appeared, and it used some mnemonic commands such as "Prefix-s" to read the screen, "Prefix-l" to read the current line, etc. The program is now loaded as a device driver from the "CONFIG.SYS" file and activated, deactivated, and changed with a program called "PV." Finally, the ability to mark and define sub-screens from "Direct Mode" has been restored. Contact Information KANSYS, Inc., P.O. Box 1070, Lawrence, Kansas 66044-8070; telephone (913) 843-0351; E-mail 70431.3476@compuserve.com. Price: $295.00 ----- The Speaqualizer Version 2.2 Conclusions The Speaqualizer is the most unusual and atypical system that we have reviewed. We wouldn't recommend it for most people who are word processing, using a modem, and the like, on a regular basis. However, it does have its niche, and if you need it, nothing else will do. If you have a very tight memory situation, are involved in complex terminal emulation situations, and/or need access to your system prior to the operating system being up, nothing else will do. Contact Information American Printing House for the Blind, 1839 Frankfort Avenue, P.O. Box 6085, Louisville, Kentucky 40206; telephone (502) 895-2405; fax (502) 895-1509. Price: $810.00 ----- SCREEN POWER SPEECH Version 1.0A Conclusions Screen Power Speech is a good and straightforward product. Its lack of speech synthesizer support and configurations are disadvantages. Its menuing system will be familiar to users of other Telesensory products, and it is easy to use and full-featured. This is likely to be the last major DOS screen review program release by a major company since everybody's efforts are now going into Microsoft Windows and other GUI systems. Contact Information Telesensory Corporation, 455 North Bernardo Avenue, Post Office Box 7455, Mountain View, California 94039-7455; telephone (415) 960-0920; fax (415) 969-9064; Customer Service (800) 227-8418; technical support (800) 537-3961. Price: $495 ----- TinyTalk Personal Version 1.67 Conclusions We would judge that Tinytalk Personal has met its goals. It provides more than the necessary minimum feature set to use most modern text-based DOS applications. It is relatively easy to use and it provides exceptional value for the dollar. Finally, it is a good way for people to experiment, to get their feet wet with screen review programs and/or to try out a sound card that they already have. Contact Information OMS Development, 610-B Forest Avenue, Wilmette, Illinois 60091; toll-free (800) 831-0272; telephone (708)251-5787; e-mail ebohlman@netcom.com Price: $150.00 ----- Vocal-Eyes Version 2.2 Conclusions We think that Vocal-Eyes is an all-around good program. It isn't as automatic as some, but it has a good and complete set of features, clear and complete documentation, good technical support, and a relatively straightforward approach. It is widely used, and we in the International Braille and Technology Center for the Blind find ourselves recommending it more often then probably any other screen review program. While it is not perfect, it is a very good general, all-around choice. Finally, when GW Micro releases new versions, they tend to work well without bugs or problems. While this isn't always true, they do have a better track record than most in this area. Contact Information GW Micro, 310 Racquet Drive, Fort Wayne, Indiana 46825; telephone (219) 483-3625; fax (219) 482-2492; BBS (219) 484-0210; E-mail support@gwmicro.com Price: $450.00 ---------- THE GRAPHICAL USER INTERFACE Introduction In 1985, a new system of computer interaction became widely available although experiments and early efforts had preceded it by several years. It is commonly known as the Graphical User interface or GUI, commonly pronounced as "gooey," as in "The chewing gum stuck to the bottom of my shoe is a gooey mess." Apple's Macintosh was the first widely available GUI system. Other examples that blind persons are now able to use include IBM's OS/2 Presentation Manager and Microsoft's Windows 3.1 and Windows for Workgroups. The final major GUI, which we are not yet able to access, is X-Windows which is commonly used with Unix systems. The Graphical user Interface is characterized by the use of a mouse or other pointing device and icons and other graphical elements. To better understand the differences between a GUI and a more traditional character-based system, let us go back and explain the character or text-based system. These systems include the old Apple II line, the Commodore 64, Ataris, and all IBM and compatible computers running MS-DOS and other character-based operating systems. The standard IBM or compatible computer normally displays text in a layout of 25 horizontal rows by up to 80 vertical columns. There is a portion of the computer's memory where this information is stored. Thus, there is a memory location that represents each possible position on the computer monitor. In that location, information on the character, its attributes, and location are stored. The screen review program only has to look at the proper memory addresses to see what is being displayed by the computer. This one-to-one memory-to-video display correspondence is what makes screen review programs possible and relatively simple, technically speaking. A GUI, on the other hand, does not have this one-to-one correspondence. Individual letters are drawn onto the screen by turning on and off little dots called pixels (picture elements). A GUI display does not necessarily have 25 rows or 80 columns of text. The size of the letters can be changed by redrawing them, using different dots, and the space between letters and/or lines can be infinitely varied. Further, the GUI uses pictures called icons. These pictures represent objects, such as documents or programs, and actions such as formatting a disk or erasing a file. Further, text and icons are moved around the screen by a "pointer" which is controlled with a mouse or other pointing device. A mouse is a small box-like object, about the size of a pack of cigarettes, which is attached to your computer with a wire. The mouse is moved around on a flat surface, and it moves the mouse pointer correspondingly on the computer's monitor. This pointer is used to move objects, choose objects, highlight text, re-size text and windows, and other actions. Why would a blind person want a GUI? Well, left to our own devices, we wouldn't. The GUI's available today use a variety of visual metaphors which are difficult to translate into linear audio terms. Further, some programs employ true graphics, that is, drawings or pictures of objects. It is difficult, if not impossible, for software to know what a picture is and to convey this information to a blind person. It would be even more difficult to create, manipulate, and/or change these drawn objects. Nevertheless, the general computing world is adopting GUI-based systems almost universally and we are being forced to them ourselves. Many employers now only use Microsoft Windows on their computers, so we must use it too. GUI's do offer one major advantage over traditional DOS-based systems, multitasking. This is the ability to run more then one application simultaneously. This would allow a person to be composing a document in a word processor and easily look up something in a database without leaving the word processor. At the same time he could also be downloading a file from an on-line service or BBS. As an example, this portion of the document is being written in WordPerfect 5.1 for DOS, which is running under Microsoft Windows 3.1. Also running on the same computer, at the same time, are Telix--a communications program, Allfind--a database program we use at the NFB, Allserve--an in-house program we use to handle phone calls and literature orders, and SDIR--a directory management program. The phone just rang with a question about Artic Business Vision. This writer was able to bring up the directory management program with two or three keystrokes, look up the needed information, and instantly return to this point in this document. Under DOS, he would have had to save the file, exit the program, start the SDIR program, look up the information, exit SDIR, re-start WordPerfect, load this document and return to where he left off. Finally, the other advantage that GUI's offer, in conjunction with multitasking, is memory management. MS-DOS systems were essentially restricted to using 640K of memory. Windows, and other GUI systems, automatically handle large pools of memory allocating it to different programs as it is needed. This is in part what makes multitasking possible. There is yet another advantage to Windows that is often cited, that of "standardization. " That is that standard procedures for common tasks, like opening and saving files, are employed. Thus, when you start to learn a new program, you will know many of the necessary actions automatically. It is not our purpose to debate whether or not this is true, especially for sighted computer users. It is not yet completely true for blind computer users. There are two major reasons for this: (1) the access software still needs improvements so many, if not all, of the packages don't work as well as they might and ultimately will, and (2) software developers do not always use standard controls such as dialog boxes, even though they may look standard on the screen. This nonstandardization causes some of the problems for access software, mentioned in point (1). Guidelines Because some of the concerns of DOS-based programs don't apply, and because there is a whole new set of concerns stemming from the use of the GUI, the guidelines used for evaluation were modified for this portion of the tests. These guidelines are as follows: 1. General description: How does the program work? How are commands laid out? What is the logic? Description? 2. What speech synthesizers are supported? Is indexing supported, and if so how well does it work? Does it shut up the synthesizer efficiently and quickly? Does it use the SSIL drivers from Arkenstone or another system? Does it support standard "sound cards" such as the Sound Blaster? Does it support refreshable Braille displays, and if so can they be run simultaneously with a speech synthesizer or separately? Does it support large-print facilities or is there a separate large-print program available? 3. Requirements: Are there any hardware and/or software requirements? A particular version of DOS, etc.? Is a DOS-based screen review program required? Can the program be used with any video and/or keyboard driver, or do specific drivers from the manufacturer have to be used? Does the product work with devices such as video accelerator cards, or are there known conflicts? 4. Installation: How do you install it? Is it easy? Do you need sighted assistance? Can you do it if you don't already have some speech up? Do you need to have Microsoft Windows pre-installed? Can you use the access package to install Windows? Is there any copy protection, and if so does it affect the installation process? Are there special installation considerations such as whether you are using a memory manager, and if so is the needed information readily available? 5. Memory requirements: How much memory does it take? Can it be moved high? Can it run in expanded or extended memory? Can you reduce the memory footprint? Can you remove it from memory? This question may be partially irrelevant under Windows since Windows handles much memory stuff automatically. However, it may not completely be so since the screen review program may affect your DOS use. Also, does the package use large amounts of memory so that you would want to increase available memory over the recommended minimums for OS/2, Windows 3.1, or the Mac? 6. Navigation: How do you navigate around the screen? Discuss reading units, character, word, sentence, line, paragraph, screen, entire document, etc. Is there a review mode? An audio cursor? Can you route the real cursor or the review/audio cursor? How do you move the mouse pointer and click the mouse? How do you find the pointer or does the software ignore it? Can you use the real mouse? If so, what aids are provided to assist you, and how do they work? Do regular Windows navigating commands work properly? Does the program have a method for telling you where an object is on the screen, such as in inches from the edge, or by the number of pixels from the edge? Is this information useful and reliable? Does the package preserve the "look and feel" of Windows, or does it interpret Windows metaphors for you into steps that seem more speech-oriented? 7. Settings: How do you set things: speech rate, tone, pitch, punctuation level, capitalization alert, etc.? 8. Configuring: Do you need to configure? How do you do it? How difficult is it? How do you define windows, silent areas, monitors etc.? How does it handle unknown icons and other unknown graphical objects? How do you define these unknown/ objects? How can or do you configure for Windows applications? 9. Monitoring: Can it monitor different areas of the screen and notify you of changes? How do you set it up? How well does it work? What options are there: beeping, chaining windows, etc.? Some of this is going to be unimportant in Windows. However, additions could include, whether it reads dialog boxes automatically and reads other messages, wizards, and other text items that appear on the screen at different places? 10. Alternative cursors: Can the program track an alternate cursor, highlight bar, ASCII character, soft cursor, etc.? How do you set it up? How well does it work? How well does it track the Mouse Pointer, Insertion Bar, and other graphical mouse pointers? 11. Attributes: How do you identify attributes: background and foreground color, extended ASCII characters, highlighted text, bold and/or underlined text, font size and appearance, grayed out objects, etc.? How can you use this information? How much of it is important in Windows and other GUI's? If you change the color, position, etc., of objects on the screen, does the system still work? 12. Searching: Can you search for things on the screen, character strings, attributes, icons, mouse pointers, etc.? How well does it work? What can you do with it? 13. Keyboard: How is the key response? Are there keyboard conflicts, and how can you get around them? Is there a pass-through key? Can you tell if the CAPS LOCK key is down? Can you find out the status of other toggle keys? Do the Windows keyboard commands all work? Are there keyboard provisions for mouse movements, and how do they work? If the application requires a mouse, what can be done? Can you "drag and drop" objects, and if so, how? 14. Macros: Does the program have built-in macro capability? If so, describe. Is there provision for macros from an external program? 15. Pre-defined configurations: Does the program come with pre-defined configurations for popular software packages? If so, how many and what? Are there other configurations available via the company, dealers, or a BBS? How well do any available configurations work? Are there configurations available for DOS, Windows, and/or other programs from other operating systems? 16. Autoloading: Does the program autoload configurations? If so, how does it work, and how well does it work? If not, what other provision is there? 17. Pronunciation: Can you change how things are pronounced? Is there an exceptions dictionary? If so, how does it work, and how well does it work? 18. Responsiveness: How responsive is the software? Does it require a fast machine? Does it eat up lots of CPU cycles? Does it adversely affect other applications? How well does it handle DOS and Windows sessions simultaneously? 19. DOS Sessions: How does the program work with DOS applications? Do you have to or need to load a DOS screen reader first? Does it support full DOS applications itself? Does it work with DOS applications without loading Windows? 20. Applications: How well does the access product work with the leading Windows software? How well does it work with applications for which there is not a specific configuration? Does it keep the focus, or does it ever get lost? Does it work with the Windows and/or OS/2 Applets? 21. Stability: some of the Microsoft Windows access products have problems with stability. Does the product under question lock up? If so, can you tell when, how and/or why? 22. Support for sighted people: Do the menus appear on the screen, or just in speech? How well does it help a sighted assistant, highlighting review cursor, etc.? Can it easily be shut up if a sighted person must use your computer? If the menus, dialogues, etc. appear on screen, do they look and act like standard Windows objects, like DOS sessions, or what? 23. Documentation: How is the documentation? What formats does it come in? How good is it? Is on-line documentation only available once you have speech and/or Windows up and running? 24. On-line help: Is there on-line help? How does it work? How good is it? Is it context-sensitive? 25. Technical support: Is there technical support and help available from the company? Is it toll-free? How good is it? Does the company have an update policy? If so, how are updates made available and how often do they appear? Procedures In addition, we used somewhat different procedures for evaluating most of the GUI programs, excluding outSpoken for the Macintosh. We acquired ten of the best-selling Windows applications and installed them on a 486 DX2/66 Mhz computer which was used for most of the testing. Our reviews will in part discuss how well the different Microsoft Windows access packages worked with these commercial off-the-shelf applications. The applications we tested are:  A. Lotus Smart Suite release 2 for Windows 1. 1-2-3 release 4.01 2. AmiPro release 3.01 3. Freelance Graphics release 2.01 4. Adobe type manager 5. Organizer release 1.1 6. Approach release 1.2 B. Borland Office 2.0 for Windows 1. WordPerfect release 6.0 2. Quattro Pro release 5.0 3. Paradox release 4.5 C. Microsoft Office Professional 4.3 for Windows 1. Word release 6.0a 2. Excel release 5.0a 3. Access release 2.0 D. Microsoft Money release 3.0 E. CompuServe Information Manager for Windows release 1.2 F. Quicken for Windows release 3 As mentioned in the initial introduction to this document, the application tests were done by a sighted consultant. Unfortunately, not all of the programs catch everything that is happening on the screen, so sighted assistance can still be helpful. Some day soon we hope this is no longer the case. We also employed more of a narrative style in this section. We described some of our problems and experiences in detail because they were new to us and will be new to most of you as well. Time did not permit us to test adequately database programs. We plan on adding those results to this document some time in the future. ----- outSpoken for Windows Version 1.0 Conclusions OutSPOKEN is a very solid screen review program. The process of capturing and storing the screen text seems flawless. It never reported anything that was not on the screen, nor did it fail to report anything that was on the screen. OSW should announce window titles as they appear, especially error message boxes. It would be very helpful if OSW could follow the visual focus in spreadsheets. It would make them accessible as well. It would also be nice to have some customization facility for hot spots, special keys, outSPOKEN Preferences used with a specific application, etc. Response time was good throughout and no product failures of any kind were experienced. OutSPOKEN for Windows has taken a unique but overall a successful approach to Microsoft Windows 3.1 access. Like its Mac big brother, it primarily uses the mouse pointer as the navigation metaphor. This makes for a largely manual process, that is, there is a lot of pointer movement that is required with most applications; however, it leaves decisions up to the end user, not the software--which may not make the right decisions. It is also a good way to learn about Windows, since exploration is necessary when moving the pointer. Berkeley is in the process of releasing version 1.1 of outSPOKEN for Windows as we write this. It offers the ability to read dialogue boxes automatically as they pop up, which should make for a little less pointer manipulation. Version 1.1 also allows for the continuous reading of a document when working in a word processor. All in all, outSPOKEN for Windows is our current favorite for Windows screen review programs. However, we have yet to test JAWS for Windows and Screen Power for Windows. Slimware Window Bridge has just released a major upgrade, and others are making improvements all the time. There is, as with DOS screen review programs, no one best universal package. Contact Information Berkeley Systems Inc. 2095 Rose St. Berkeley, CA 94709 Phone (510) 540-5535 FAX (510) 849-9426 Internet: osw@berksys.com ----- Protalk for Windows Version 1.0 Conclusions Protalk provides reasonable access to those programs that it lists in its documentation. With the exception of Excel, we were able to complete all the tasks listed in our tests. However, there are many times that little things are a hindrance and some major flaws exist even in the approved applications, such as General Protection Faults that cause a reboot. It does not seem that any applications other than the approved ones are very accessible. It would also be nice to have some special functions in certain areas of applications, like spell checking. While the Item Manager and the Menu Manager always seem to give the user complete access to Windows commands, using regular Windows keyboard commands does not always work. For example, when switching from group to group with the CONTROL-TAB command, Protalk does not always say the name of the window you land on. Further, sometimes when using an application such as Write, speech disappears. Pushing the Second Function key twice refreshes the screen and gets speech back, but these things shouldn't be happening. Biolink has not updated the product recently and no new "PDF" files have been offered since the program was introduced. The company reports that they are working on versions for Windows NT and Windows 95. It is our belief that they are putting their energy and efforts into these products and are unlikely to improve the Windows 3.1 version of Protalk for Windows in any substantial way, if at all. Finally, the package is on the expensive side. Except for the use of multimedia sound cards, Protalk has little else to recommend it. Contact Information Biolink Computer Research and Development Ltd., 4770 Glenwood Avenue, North Vancouver, B.C., Canada V7R 4G8; telephone (604) 984-4099; fax (604) 985-8493; BBS (604) 985-8431. Price: $895.00 ----- SLIMWARE WINDOW BRIDGE Version 1.32 Conclusions Slimware Window Bridge is a good Windows access product. It has an almost overwhelming number of commands and functions and provides an extensive configuration capability. Those configurations shipped with the product are good, but much more could be done with read areas and monitors. Spreadsheets especially could use configuration help to make them more accessible. It does seem that SWB's off screen model (data base of screen contents) gets confused from time to time. It is necessary to remember that a lot of complex activity goes on regularly in Windows. The "refresh" key helps a lot, but it depends on the user's knowing that the model and screen are not the same. Overall, SWB provides a good means of using Windows for the user who is blind, even with programs that have no supplied configuration. Contact Information Syntha-Voice Computers, Inc., 800 Queenston Road, Suite 304, Stony Creek, Ontario, Canada L8G 1A7; telephone (905) 662-0565; fax (905) 662-0568; BBS (905) 662-0569; E-mail help@synthavoice.on.ca Price: $695.00 ----- SCREEN READER/2 Version 1.2 Conclusions Even though they are not native Windows programs, OS/2 and Screen Reader/2 warrant a serious look by those who need to run Windows applications. The system is stable and reliable for the most part. Its off-screen model, the underlying software that converts GUI elements and text into a form that speech software can use, is one of the best on the market, if not the best. It rarely gets confused, as some do. IBM has put a lot of time and effort into making this system operate well. More profiles are needed, and technical support remains a question, but on the whole, we like it very much. Contact Information IBM Corporation, Special Needs Systems, 1000 Northwest 51st Street, Boca Raton, Florida 33432; telephone (800) 426-4832; Support (800) 426-8637. Price: *$600 to $743 approximately ----- Windows Master Version 1.4 Conclusions Deane Blazie, President of Blazie Engineering, has stated that it is his intent to make Windows Master the best access program on the market. Unfortunately, the program has a long way to go. Its major problems include poor documentation, instability with some applications, excessive speech, the inability to recognize toolbar icons, and a lack of features and configurations. Blazie has a reputation for prematurely releasing software and allowing users to do the beta testing. The company also generally sticks with things until they are right. We hope that they are able to do so with Windows Master, because as it now stands, the program is virtually useless. It would be very difficult to accomplish any amount of serious work with it. Contact Information Blazie Engineering, Inc., 109 East Jarrettsville Road, Forest Hill, Maryland 21050; telephone (410) 893-9333; fax (410) 836-5040; E-mail Telnet to Blazie.com or write to (first name)@blazie.com where (first name) is the name of the person you wish to reach. Price: $495.00 ----- Artic WinVision Version 1.54 Conclusions We found WV to be "rock solid" in its access to the groups of items contained in the Program Manager. Standard dialogues like that for the Desktop were completely accessible. Moving the mouse pointer over icons causes WV to speak their associated text, even though the WV manual discourages the use of the physical mouse. WV even figures out words that go together, like multi-word titles for group boxes, and speaks them together as the mouse pointer moves over them. Menu items are reported correctly, including checked items; but the fast path keys are not spoken, nor were ellipses at the end of menu items. These ellipses are important to let the user know that more is to come if that item is selected. WinVision is a good Windows screen reader. It always seems to have a good model of the screen, so the user is never subject to incorrect reading of screen information. WV really suffers from a lack of application configurations. If those were provided, this product would really do a good job. Artic also needs to figure out why cell contents were not read in 1-2-3 and Quattro Pro. Artic has released Version 2.x of WinVision. It offers additional configurations, the ability to read titles and field names above the data entry box, and many other improvements. We did not have time to test completely the beta version of the 2.16 release but hope to do so in the future. This new release should make WinVision a very strong competitor. Contact Information Artic Technologies International, 55 Park Street, Suite 2, Troy Michigan 48083-2753; telephone (810) 588-7370; fax (810) 588-2650; BBS (810) 588-1424; technical support East Coast (810) 588-1425; technical support West Coast (209) 291-3645. Prices: Business Vision $495, WinVision $495, SynPhonix 210 and 215 $395, SynPhonix 315 for Microchannel computers $495, Transport $895, Transport with Businessvision $1295. ----- OutSPOKEN for the Macintosh Conclusions Overall, outSPOKEN for the Macintosh does an adequate job of providing blind and visually impaired people with access to the Apple Macintosh. However, it would be nice if Berkeley Systems would pay more attention to the Mac. It seems as though the bulk of their time is spent on developing outSPOKEN for Windows and that updates for the Mac have been virtually nonexistent. It's understandable that Berkeley is devoting time to the IBM, given its popularity in the blind community, but more and more companies are using the Macintosh. We have also recently become somewhat concerned because of organizational changes at Berkeley. For those who may not know, Berkeley Systems writes and sells commercial software primarily. Their flagship product is the "After Dark" screen saver line. They have recently separated the access division that makes the outSpoken and enLarge products from the main corporation. The access division will now have to be self-supporting, without major cash infusions from the parent company. Since Berkeley has been disappointed with sales anyway, this move can't bode well for the future of Berkeley Access. Contact Information Berkeley Systems, Inc., 2095 Rose Street, Berkeley, California 94709; telephone (510) 540-5535; fax (510) 849-9426; E-mail access@berksys.com Price: $495.00 for Macintosh $595.00 for Windows ----- LAST WORDS This document has attempted to provide the reader with comprehensive and comparative data on all of the screen review systems that are currently and generally available. We view this as a first effort and as a living document. We intend to add new products to it as they become available. We hope to add new information and comparisons as time permits and as the International Braille and Technology Center for the Blind develops this information. This could include an Appendix on speech synthesizers, tables and charts more directly comparing competing features, and maybe even a rating system. It is our hope that we can provide you with enough information to make wise purchasing decisions. Because of the sheer amount of information present, this could be difficult. If you need help, please give us a call at the International Braille and Technology Center for the Blind. We will help you sort out the issues and tell you of new information we have learned. If you are in the Baltimore area, make an appointment to drop in and see us. We are more than happy to show you any technology in the International Braille and Technology Center for the Blind. David Andrews, Director International Braille and Technology Center for the Blind National Federation of the Blind June 28, 1995 ---------- This file contains two documents that could be of interest to any current or potential user of screen access software for Microsoft Windows. Neither document has been edited and is presented in its entirety. There is a line of asterisks at the start of each document for ease of access. The first document was issued by Henter-Joyce Inc., manufacturer of Jaws for Windows (JFW), describing itself as a "report" of the Dueling Windows pre-conference session held at the California State University - Northridge conference in March, 1996. It was written by Cathy Kaiser of Henter-Joyce. The second document is a letter from Mr. Ian Robinson of England regarding Henter-Joyce's document. The Henter-Joyce document was published in England in a periodical called "Access It" in June, 1996. The letter from Mr. Robinson is to Paul Blankhorn, the publisher of "Access It". While Mr. Robinson is a user of Artic WinVision, he is not associated with Artic Technologies nor any of our representatives in any way. Neither Artic Technologies nor any of our representatives solicited either Mr. Robinson's actions or his letter. We are however, grateful for his permission to re-print his letter and thereby present the reality of WinVision's capabilities. This information is being provided in the same spirit as that of the Henter-Joyce report, in the hope that it will (as Ms. Kaiser so aptly puts it) "be extremely helpful to the blind and visually impaired populace unable to attend the competition". We further encourage anyone interested in the effectiveness of screen access for either Windows 3.1 or Windows 95 to do their own evaluation, using whatever application software they would normally use. We will be happy to provide a free, fully functional, fully documented demonstration version of our WinVision or WinVision 95 programs. Simply call, FAX, or write: Artic Technologies 55 Park Street Troy, Michigan 48083 Telephone: (810) 588-7370 FAX: (810) 588-2650 ********************************************************************** Dueling Windows Report From: Cathy Kaiser <76622.1535@compuserve.com> There has been much discussion over the last few weeks about Dueling Windows, a pre-conference session held at the C S U N conference in Los Angeles, California, March 19th, 1996. For the few of you fortunate enough to attend the conference you were able to see seven of the major screen reader development companies dueling it out in a head to head competition. The format was very fair. No company knew the questions beforehand; each company used an I B M computer provided for them of identical speed and each vendor used software already available to the public. In a round robin style format questions were asked from the panel of each vendor and the audience scored the contestants according to their own judgment. A winner was never proclaimed by the people running Dueling Windows. While declaring a winner would probably lend itself to argument, it is our feeling that at least a reporting of the proceedings would be extremely helpful to the blind and visually impaired populace unable to attend the competition. We decided to release our report because there is no "official" report and we have had many inquiries from people who could not be there. This document was compiled from notes taken separately by Dan Clark and Ted Henter from Henter-Joyce who were in attendance. While this is our opinion, and obviously not without bias, we have tried to be fair. Cathy J. Kaiser, Henter-Joyce, Inc., 800 336 5658, 813 528 8900, info@hj.com "Dueling Windows", C S U N Conference, Los Angeles, C A, March 1996 (Report from Henter-Joyce) We decided to release this report because there is no "official" report and we have had many inquiries. This document was compiled from notes taken separately by two people from Henter-Joyce, Inc., Dan Clark and Ted Henter, who were in attendance. The "duel" involved seven different Windows screen reader software products performing identical tasks in the Windows 3.X environment. Each machine used was identical, being an I B M Pentium with 16 MB of Ram, many thanks to I B M for providing them. The software on each hard drive was identical, with the exception of the windows screen reader which was installed by each vendor prior to the duel. Also, I B M used a different program for internet access, Web Explorer, whereas the rest of the participants used Netscape. TeleSensory added a PowerBraille display during the afternoon presentation, and although several other screen readers support Braille display devices, the other participants did not use one. During the presentation, a large screen projector displayed for the audience what was on each vendor's screen. Also, the persons doing the presenting could neither see their monitor, which had been placed on the floor in front of their table facing the audience, nor the large screen. Each question was asked by one member of the panel of each vendor, one after the other, and the panelists rotated the asking of questions among themselves. Additionally, the order of which vendor went first was rotated with each question, so that no one vendor was able to benefit from always hearing the other vendors' attempts to answer. The panelists consisted of the following people: Mark Nelson, Foundation for Blind Children, Phoenix, A Z; Greg Meise, Missing Link Technologies, Lenexa, K S; Kerry Malone Cook, Southern Oregon Goodwill, Medford, O R; Jay Stiteley, The Seeing Eye, Inc., Morristown, N J. The software companies represented were, in alphabetical order: Artic WinVision 2, ver 2.16, Paul Hendrikson presenting. Berkeley outSPOKEN for Windows, ver 1.2 vb26 public beta, Josh Miele presenting. G W Micro, Window-Eyes ver 1.03, Mike Lawler presenting. Henter-Joyce, Inc., JAWS for Windows ver 1.2, Ted Henter presenting except for Netscape, which was done by Glen Gordon. I B M Screen Reader/2, ver 2.0, Guido Corona presenting. Syntha-Voice Window Bridge, ver 2.21, Neil Cooper presenting. TeleSensory Screen Power for Windows ver 1.47 B, Scott Duncan presenting. Overview. There were a total of 15 questions which were asked over the course of the day. There was one question which no screen reader could do at this time. It involved Netscape, and went something like this: If hypertext links are on a page, but are not currently visible on the screen, can you read them without hitting the page down key to bring them into view? Since no one was able to do this, it is not counted in the tally below. The screen readers are listed below in order of the least amount of errors. The panel recommended using the terms "accomplished task", "accomplished task with difficulty", and "failed task". The scores listed in this report are simply our opinion of what transpired. JAWS for Windows successfully answered 13 questions, and answered one question with difficulty involving reading columns in Netscape. Screen Reader/2 successfully answered 11 questions; they answered 2 questions with difficulty (one in Access, and one in Excel), and failed on one question in Web Explorer. outSPOKEN for Windows successfully answered 8 questions; they answered 5 questions with difficulty (one in Access, 3 in Excel, and one in Netscape) and failed on one question in Excel, where they had to reboot the computer. Window Bridge successfully answered 6 questions; they answered 2 questions with difficulty in Netscape, and failed on 6 questions, two in Access (one where they had to reboot the computer), three in Excel, and one in Netscape. WinVision successfully answered 6 questions; they answered one question with difficulty in Excel, and failed on 7 questions, two with Access, three with Excel, and two with Netscape. Screen Power for Windows successfully answered 3 questions; they answered 4 questions with difficulty (2 in Excel, one in Netscape, and one in Word for Windows) and failed on 7 questions, two in Access, two in Excel, and three in Netscape. Window-Eyes successfully answered 5 questions and failed on 9 questions, two in Access, four in Excel, and three in Netscape. We will outline the questions and answers, or attempts to answer, in detail below. The questioning began with Microsoft Access, a Windows database program. 1. Use the File Open dialog box in Access to load a sample database file. Everyone was able to accomplish this task with little or no problems. 2. The sample Access database was in Table view when opened. The question was, if you are in the body of the table, can you read the title of the column you are in? WinVision failed the task. It was not reading contents of fields as they tabbed through them and could not read the column heading. outSPOKEN for Windows accomplished the task. They used the mouse cursor to read it. Window-Eyes failed the task. It was not reading the contents of fields as they tabbed through them without using the mouse cursor. Were unable to get their mouse cursor to go outside the table to read column header. JAWS for Windows accomplished the task. It read the cell contents by moving between cells with "TAB" key and could read the column header using the mouse cursor. Screen Reader/2 accomplished the task. It was able to read the contents of each cell and had a special function to read the column header. Window Bridge failed the task. It began reading the cells okay, but had a General Protection Fault error and had to exit windows completely and reboot the computer. Screen Power for Windows failed the task. Tried exploring and could not find the column header. 3. Still in Access, the task was to do a query in the database. Close the current Table view child window, go to the word "View" in the menu bar and press "enter" to pull down a menu. The word "Tables" was the first item in this menu and was checked, because that was the view we had just been in. Go down one to the word "Queries" in this menu and select that with the "enter" key, then press "enter" again in the query list box for "D W-Query". This was a query that the panel had set up prior to the duel. This pulled up the "Enter Parameter Value" dialog box which had a single edit field labeled "Enter Position" and two buttons, an "O K" button and a "Cancel" button. The task was to enter in this field "Sales Representative" and when you pressed "enter" you got the "Enter Parameter Value" dialog box again with another (secondary) input field, labeled "Enter country" which you had to fill in. The end result is to get a few records which matched the query for "Position" and "Country", as opposed to a whole list of records with all information. WinVision failed the task. It could read the dialog when it opened the first time. Could not read the edit field labels. outSPOKEN for Windows accomplished the task with difficulty. It did not say the word "checked" when going into the View menu and landing on the word "Tables". It did not read the dialogs automatically, but was able to read the first edit field label with the "minus" key on the number pad. Then when they tried to read the label on the second dialog box edit field for "Enter country" it said "no label". They were, however, able to route the mouse cursor there and read it. Window-Eyes failed the task. It did not read the edit field labels automatically. They tried using their own keyboard command ("ALT" + "SHIFT" + N), without success, and thus did not know what to enter in the edit fields. JAWS for Windows accomplished the task. It read all of the edit field labels automatically, and was able to use a JAWS command ("CTRL" + "DOWN ARROW") to re-read the edit field labels and edit field contents. Screen Reader/2 performed the task with difficulty. It did not read the edit field label automatically when the dialog box opened. They could not read the edit field labels without moving around. The "TAB" key did read the edit field labels then, but said "no active point" when the cursor was actually in the edit field. Window Bridge failed the task. It did not read the edit field labels in the dialog box and therefore did not tell the operator what to enter in each field. Screen Power for Windows failed the task. It did not read the edit field labels in the dialog box and therefore did not tell the operator what to enter in each field. They tried to use navigation techniques and got lost. Lost the focus on the query dialog box and could not get the focus back there. The next series of questions deal with the program Microsoft Excel, a Windows spreadsheet program. 4. The first task in Microsoft Excel was to use the "File Open" dialog box and press "TAB" to go to the "directories" list box and change to a different directory. Then press "TAB" or "SHIFT" + "TAB" to go back to the "files" list box and load a file called "invoice". WinVision accomplished the task with difficulty. It did not automatically read the labels for the file name list box or the directories list box. When they got the file open, it did not announce the cell coordinates of the active cell. outSPOKEN for Windows failed the task. It read everything in the entire dialog box when it opened. When they pressed "TAB" it did not read the labels for the list boxes or edit fields as they moved from control to control. Then, when they pressed "TAB" to go to the "directories" list box, they got into an endless loop and had to reboot the computer. They could not open the file and had to pass. Window-Eyes failed the task. It was not able to work with Excel and therefore they chose to pass on all Excel questions. JAWS for Windows accomplished the task. When the dialog box opened J F W automatically read the title of the dialog box and the label of the File Name edit box. As "tab" was pressed it read the labels of the file name list box and the directories list box. It read the directories as they were moved to, and reported "no item selected" when no file was yet selected in the files list box. It spoke the cell coordinates when the spread sheet was loaded. Screen Reader/2 accomplished the task with difficulty. It used chorded keystrokes on their separate numeric keypad to find things and did not automatically read the edit field labels or directories list box labels. It did not say "no selected item" in files list box to let you know that no files were selected. Window Bridge failed the task. They pressed "tab" to move between controls and this did not work. They tried using the arrow keys to move up and down the lists in various controls and that did not work either. Screen Power for Windows failed the task. Tried using navigation keys and got lost. Could not get back to Excel and had to re-boot the computer. 5. The task was to read the cell coordinates where the cursor was in the Excel spreadsheet called "invoice". WinVision failed the task. Tried, but could not do it. outSPOKEN for Windows accomplished the task with difficulty. They had to load the file through File Manager rather than the File Open dialog box, then had to hunt around the screen for the cell coordinates. Window-Eyes failed the task. They passed on all Excel questions without attempting them. JAWS for Windows accomplished the task. It was able to read the cell coordinates with a single JAWS keystroke ("INS" + "C"). Audience applauded. Screen Reader/2 accomplished the task. They used a key combination on their separate numeric keypad to read the cell coordinates. Window Bridge accomplished the task. They were able to do it with a keystroke. Screen Power for Windows accomplished the task with difficulty. It was able, after much exploring with navigation techniques, to get the spreadsheet loaded (had failed earlier in this task). Then needed more exploring to find the cell coordinates. 6. The next task was to read the text in cells A 1 to A 6. This was "Foundation for Blind Children" in A 1, with address, phone, and fax number information in the cells below. WinVision failed the task. It was able to move to cell A 1, but the operator did not know, because nothing was spoken. "Up" and "down arrow" said "UP" or "DOWN" but did not read contents of the cell moved to. They ended up passing on the question. outSPOKEN for Windows accomplished the task with difficulty. It could not confirm they were in the correct cell because it did not read the cell coordinates when moving around. It did not track the cell pointer. Used their reading cursor to review what was on the screen. Window-Eyes failed the task. They passed on all Excel questions without attempting them. JAWS for Windows accomplished the task. It read each cell's coordinates as the cursor moved from one cell to the next (optional, depending on "Verbosity" level). It also read what was visible within each cell as the cursor moved from cell to cell. With a separate JAWS keystroke ("INS" + "F 2"), it read the edit field which contains the formula or overflow data. Screen Reader/2 accomplished the task. It took several keystrokes to do it, but we did not record it all. Window Bridge failed the task. It could not read the text in the cells. Screen Power for Windows accomplished the task with difficulty. They used navigation mode and were able to read what was on the screen, but were not able to say what cell the cursor was in. 7. The next question was also in Excel. The task was to move to cell F 2 2 using the "go to" command. The column was one with a narrow width. Next, participants had to enter a number which was too long to fit in the cell. The result visually on the screen is that Excel fills the cell with pound signs. Then the participants had to read the cell contents and verify that pound signs were there so that they would know that the number was too large to fit in the space provided. WinVision failed the task. They were not able to read the data in the cell. outSPOKEN for Windows accomplished the task with difficulty. It was very verbose in the "go to" dialog box. It did not confirm that they were in the proper cell nor did it track the cell pointer. After entering the data, they searched for the row label "22" and then read across that line to find the pound signs. Window-Eyes failed the task. They passed on all Excel questions without attempting them. JAWS for Windows accomplished the task. J F W said "go to" when the dialog box was opened. When the coordinates were entered and the "enter" key was pressed it said the cell coordinates landed on to confirm the new position. The numeric data was input, the "enter" key was pressed, and J F W said the new cell coordinates moved to, one row down. The "up arrow" key was pressed to move to the previous cell and J F W read the pound signs. Screen Reader/2 accomplished the task. It performed very similar to J F W, did not read the cell coordinates automatically. Window Bridge failed the task. It could get to the cell and input the numeric data, but could not read the pound signs in the cell. Screen Power for Windows failed the task. It was able to input data in the cell, but was unable to read the pound signs in the cell. The next series of questions dealt with Netscape, a World Wide Web browser program for Windows. For this series of questions, Screen Reader/2 used I B M Web Explorer and not Netscape. 8. The question was to open a sample Web page into Netscape and read the U R L address near the top of the screen on the web page. All of the screen readers were able to do this with relative ease. 9. The next question was "how do you know when an entire Web page has been received?" Everyone basically went to the bottom of the Web page and looked for the words "document done" on the status line. All of the screen readers except for one were able to accomplish this task. Screen Power for Windows failed the task. They used navigation mode and got lost. 10. This Netscape question was "how do you know when you're at the top or bottom of a web page?" Most said they would page down or up until nothing else happened when reading. Everyone did this with about the same rate of success when they demonstrated it. 11. The next task in Netscape was to find a hypertext link on a Web page and then select it. WinVision accomplished the task. They used a pop-up dialog box, selected the underline feature to search for, started the search and found a link. Then used a mouse click to go to the new Web page. outSPOKEN for Windows accomplished the task. They used a pop-up dialog box, selected the underline feature to search for, started the search and found a link. Then used a mouse click to go to the new Web page. Window-Eyes failed the task. They pressed the "TAB" key and it spoke the link. However, when they pressed the mouse key to click on the link, it was not actually on the link as they thought it was, so it did not work. JAWS for Windows accomplished the task. When the "TAB" key was pressed, it spoke the link and moved the cursor there. They pressed enter on the link (or could use a mouse click) and it selected the link and automatically read the next Web page. Screen Reader/2 accomplished the task. They used the I B M Web Explorer program, had a key to move to the link, and then selected it. Window Bridge accomplished the task with difficulty. They had some special keys to find a link, but it did not work at first. Tried again and it worked. Screen Power for Windows accomplished the task with difficulty. They used navigation mode. Had to read through all the text to find a link. Did not have a key or function to automatically find the links. 12. In Netscape, on a Web page with a form to fill in, find and fill in the Name and E-mail fields. The input field boxes were several spaces over from where the field labels were. WinVision failed the task. They searched the screen manually for the "name:" edit field. Their screen showed differently than their speech. They gave up after trying. outSPOKEN for Windows accomplished the task with difficulty. While looking for the input field, they accidentally switched to another application in Windows, but then recovered. They searched for the "name:" field label and found it, but the software did not indicate that the cursor was not in the edit field for them to input data. They did eventually locate the edit field by moving to the right and were able to input the data. Window-Eyes failed the task. They were able to find the edit field label "name:" and put their mouse cursor just to the right of it, hoping to find the edit field, but never did get the cursor into the edit field. They gave up after trying. JAWS for Windows accomplished the task. Used a JAWS keystroke ("INS" + "TAB"), JAWS found the field and said "Name: edit" to let him know the cursor was there in the edit field, ready to enter data. He typed the data in the first field. He then pressed "INS" + "TAB" again and landed in the second field, the label was spoken automatically, and the data was entered. The audience applauded. Screen Reader/2 accomplished the task. They used Web Explorer, not Netscape, and "TAB" moves between fields and automatically puts the cursor in the edit field where it is needed. Window Bridge accomplished the task with difficulty. Read around, pressed many keys, got lost once but did it. Screen Power for Windows failed the task. Used navigation mode, got lost, found an edit field, but could not get the cursor there. 13. In Netscape, read a single column in a multiple column Web page. WinVision, Window-Eyes, Screen Reader/2, Window Bridge, and Screen Power for Windows failed to accomplish this task. outSPOKEN for Windows accomplished the task. They have a "read within column" feature. JAWS for Windows accomplished the task with difficulty. Used the mouse cursor to find the spaces between columns and was able to read by going down one line at a time and reading "from beginning of line to cursor". 14. This question was a Microsoft Word for Windows task. Open up the spell checker and demonstrate it, then tell how you know the short cut keys for the various buttons in this spell checker dialog box? WinVision accomplished the task. Read the entire dialog box when it popped up. Do not have a feature to read all short cut keys in the entire dialog box. Able to read the short cut keys if they tabbed to the control. Did not spell the misspelled or suggested word. OutSPOKEN for Windows accomplished the task. Read the entire dialog box when it popped up. Do not have a feature to read all short cut keys in the entire dialog box. Able to read the short cut keys if they tabbed to the control. Did not spell the misspelled or suggested word. Window-Eyes accomplished the task. Read the entire dialog box when it popped up. Do not have a feature to read all short cut keys in the entire dialog box. They did not demo reading the short cut keys if they tabbed to the control. Did not spell the misspelled or suggested word. JAWS for Windows accomplished the task. Reads and spells the misspelled word, and reads and spells the suggested word. Was the only one that demoed reading the misspelled word in context, and the only one that had a "hot key help" message that read all the hot keys with one command. Screen Reader/2 accomplished the task. Read and spelled the misspelled word and the suggested word. Able to determine shortcut keys by tabbing to each control. Window Bridge accomplished the task. Read the dialog box and misspelled word. Able to determine shortcut keys by moving to the controls. Screen Power for Windows accomplished the task with difficulty. Used explore mode, got lost, was able to recover. ********************************************************************** The following is Mr. Robinson's letter to Paul Blankhorn, the publisher of "Access It". ********************************************************************** 18 Whitehead Close, Barrow-in-Furness, Cumbria, LA14 1AP. Thursday 20 June 1996 Dear Paul, I write with reference to the article in the June copy of Access It, supposing to give an unbiased report from the C S U N, Dueling Windows pre-conference session in California during March. Having used Windows 3.11 with Winvision on a daily basis since the beginning of last year I feel obliged to convey to your readers the inaccuracies of the printed article. The original tests were done using Winvision 2.16; I have tried to replicate them using my slightly earlier version of 2.06. Original Winvision score: 6 successful; 1 successful with difficulty; 7 not successful. Task 2. Winvision will usually automatically read the contents of a Microsoft Access field in table view. However, the success rate can be increased to near 100% if the table's fields are formatted to the correct width on the screen. As for reading the column heading using the mouse keys, which they did with some other packages, this will work with Winvision azwel. Task 3. As yet I have not replicated this task and cannot therefore comment at this time. Task 4. In the "File Open" dialogue box in MS Excel Winvision will happily read the labels for "File name" and "Directory" if the user chooses to have them spoken automatically. Alternatively, they can be read manually with no difficulty whatsoever. Task 5. I cannot understand how anyone can suggest that Winvision will not read MS Excel co-ordinates - either automatically or manually. Having worked with Excel regularly for 18 months I can honestly say that this is not and never has been the case. task 6. As with the previous task, the suggestion that Winvision will not read the contents of the active cell is complete and utter rubbish. The contents are read automatically and can be repeated manually if necessary. Speaking as someone who has designed spreadsheets for sighted users to work on I can genuinely say that it is perfectly possible to design and work with an Excel spreadsheet using Winvision and anyone who says otherwise simply doesn't know what they are talking about. Task 7. Regarding the reading of a cell whose contents are to wide to fit. Upon making the relevant cell active Winvision will automatically read the pound signs, although it calls them "number signs". With the press of a Winvision key combination the true contents of the cell (displayed in the formulae box above the spreadsheet) will be read accurately every time. Tasks 8 to 13. As these related to the use of Netscape, I am unable to comment. Task 14. Winvision has the facility for having the last spoken word spelled. Contrary to the report this will work perfectly well when using the spell-checker in MS Word. True Winvision score: 12 successful; 2 with difficulty. This more accurate assessment of Winvision puts it on a par with JAWS, the suggested leader in the original report. JAWS is produced by Henter-Joyce who also employ the two people who took the notes to produce the article. Isn't life full of strange little coincidences! Yours sincerely Ian Robinson ---------- Sensory Access Foundation has published a "Dueling Windows" report in the Spring 1996 issue of Access Review. It summarizes a competition among Windows screen readers that was held at the 1996 CSUN conference in March. I am prefacing the article with some tabulations I did with its data. Since the article does not quantify the performance of each screen reader according to its criteria, I did such an analysis for the interest of readers. I assigned a value of 2 for accomplishing a task, 1 for accomplishing it with difficulty, and 0 for failing it. If anyone observes an error in my arithmetic, please let me know. Artic's WinVision accomplished: 7 accomplished with difficulty: 2 failed: 6 total points: 16 average: 1.06 Berkeley's OutSpoken for Windows accomplished: 9 accomplished with difficulty: 5 failed: 1 total points: 23 average: 1.53 GW Micro's Window-Eyes accomplished: 5 accomplished with difficulty: 1 failed: 9 total points: 11 average: .73 Henter-Joyce's JAWS for Windows accomplished: 13 accomplished with difficulty: 1 failed: 1 total points: 27 average: 1.8 IBM's Screen Reader/2 accomplished: 13 accomplished with difficulty: 1 failed: 1 total points: 27 average: 1.8 Syntha-Voice's Window Bridge accomplished: 5 accomplished with difficulty: 4 failed: 6 total points: 14 average: .93 TeleSensory's Screen Power for Windows accomplished: 5 accomplished with difficulty: 7 failed: 3 total points: 17 average: 1.13 Based solely on this analysis, JAWS for Windows and Screen Reader/2 were tied for top. They were followed by OutSpoken for Windows, Screen Power for Windows, WinVision, Window Bridge and Window-Eyes. Let me remind readers that Windows access products are rapidly developing so that new versions might lead to significantly different results today. In fact, some developers have already made enhancements to address problems encountered at this competition. ---------- Dueling Windows: An Overview by Kenneth Frasse (re-printed with permission of Sensory Access Foundation from Access Review, Vol. I, No. 1, Spring 1996) Abstract In March, at the CSUN International Conference on Technology and Persons with Disabilities held in Los Angeles, California, Mark Nelson of the Foundation for Blind Children hosted a pre-conference seminar called "Dueling Windows". This seminar attempted to contrast and compare graphical user interface (GUI) screen review software in applications for Windows. The format of the seminar was to pose identical tasks to each of the vendors in alternating order and have the vendors demonstrate the ability of their products to perform the specified tasks. Since SAF does not sell, distribute, or endorse any single GUI screen reader, we attempted to generate an objective review of the GUI screen readers demonstrated. Editor Similar to the 1995 Closing the Gap presentation of Dueling Windows, Mark Nelson of the Arizona-based Foundation for Blind Children coordinated another Dueling Windows at the March CSUN conference in Los Angeles. The Los Angeles seminar differed in several respects from the first seminar at Closing the Gap, not the least of which was that the seminar was six hours long. It was similar to the Closing the Gap seminar in that it was composed of questions and tasks that each of the vendors had to answer or perform. Notably, Greg Meise of the Kansas-based Missing Link Technologies, Kerry Malone-Cook of the Southern Oregon Goodwill, and Jay Stiteley of The New Jersey-based Seeing Eye, Inc. joined Mr. Nelson as the three panelists who would present the questions of the day; these panelists also formed the impetus for creating Dueling Windows. I wish to take a moment to explain the presentation and approach of this article so that the reader may be as clear on the issues presented as possible. I will present Dueling Windows in five sections: 1) Who was there: Which GUI software were represented at the seminar and what are some of the fundamental equipment differences between them that the seminar results might not make obvious for the reader; 2) Technical Set-up: What was the technical set-up for the seminar; 3) Environment: What are the advantages and disadvantages of the Dueling Windows environment; 4) Results: The Dueling Windows results; 5) Summary and commentary. Who was there Seven GUI screen reader developers were represented at the seminar. These included: 1) Artic Technologies's WinVision 2 v2.16; 2) Berkeley Access's outSPOKEN for Windows v1.2 (vb26 public beta) 3) GW Micro's Window-Eyes v1.03; 4) Henter-Joyce's JAWS for Windows (JFW) v1.2; 5) IBM's Screen Reader/2 (SR/2) v2.0; 6) Syntha-Voice's SLIMWARE Window Bridge v2.21; 7) TeleSensory's ScreenPower for Windows (SPW) v1.47B. Nearly all GUI screen readers were represented by their respective owners/developers/staff except for WinVision which was represented by one of their California authorized dealers, Paul Hendrikson. Window Bridge also opted to have a student user demonstrate the latter portion of the seminar in lieu of Niel Cooper who demonstrated the first segment. The other participants were Josh Miele of Berkeley Access, Mike Lawler of GW Micro, Ted Henter and Glen Gordon of Henter-Joyce, Inc., Guido Corona of IBM, and Scott Duncan of TeleSensory, Inc. Several of the GUI screen readers claim to possess electronic braille display support, but TeleSensory was the only vendor who brought and used their own PowerBraille 40-cell display; unfortunately, the airlines misplaced TeleSensory's PB-40, and Mr. Duncan was unable to demonstrate the braille display ability until later in the afternoon. (TeleSensory claims that one of the strengths of ScreenPower for Windows is its braille display support. SAF will soon be testing and comparing braille display support, and the results will be published in ACCESS REVIEW. Technical Set-up The seminar was performed on 7 identical Pentium computers with 16 megabytes of RAM; these machines were generously donated by IBM. Each of the computers had a 17-inch monitor that was turned 180 degrees away from the participant and facing the audience so that the audience could review where the participant was in their manipulations. Additionally, a large overhead screen displayed to the audience the monitor of the individual vendor who was performing tasks; vendors were not allowed to visually review the monitor. All audible feed from the computers was sent to DECtalk voice synthesizers which were in turn fed into the main PA of the room. Each vendor installed their own GUI screen reader and peripherals associated with their software. Aside from TeleSensory's eventual use of the PB-40, IBM's Screen Reader/2 used an external keypad from which all operations were performed. Mr. Corona did have one piece of software that differed from the other vendors; IBM uses their own Web Explorer for Internet access while the other GUI screen readers use Netscape for their access to the Internet. Since real-time access to the Internet posed a tactical impasse, the vendors had to download prepared pages from their own hard drives; it was assumed that this would mostly resemble real-time operation on the Internet. Lastly, the seminar was split into two three-hour sessions with an emphasis on spread sheets in the morning, and an emphasis on Netscape and word processing in the afternoon. The applications used included Microsoft Access, Microsoft Excel, Netscape (Web Explorer for IBM), and Microsoft Word. Environment It is important to understand the context and limitations of the Dueling Windows environment with respect to GUI screen readers. In so doing, the reader can better establish where more inquiry may be appropriate, as well as where, and when, GUI screen readers can directly address the user's needs. Generally speaking, the fifteen questions from the Dueling Windows panel posed common, direct tasks for the vendors to perform. The reader can determine the general state of GUI affairs through the entire seminar, but there are several cases in which the reader cannot determine the cause of a certain result. In several cases, vendors' systems crashed, and they were obligated to re-boot their systems; in some cases, it was not possible to determine exactly what the cause of the crash was. It is important to acknowledge that there are numerous ways to program in Windows for a desired result, and, consequently, there are inherent problems with the way Windows applications have been programmed. Since it is SAF's belief that most beginner users do not know how to configure the speech parameters or profiles for various Windows environments, it was also difficult to determine whether a GUI screen reader could not access something on the screen, or whether this problem could have been solved by altering the speech configuration; in short, there are minor and major obstacles, and Dueling Windows could not afford the vendors the time to really address this other than to allow them to verbally state whether they could, or could not, perform a task. Naturally, the ease with which a user can access their application or configure their speech software is important, even critical, but this detail could not be discussed because of time constraints as well. Though the general approach and philosophy of each vendor became increasingly apparent as the seminar passed, the more subtle distinctions of each vendor's approach could not be related through the Dueling Windows context. None of this is to say that the seminar should, or could, have been planned differently. It is simply necessary to understand that any competitive context for any product is sometimes misleading in only reviewing the statistical results. The seminar was well-planned, well-organized, and extremely thoughtful and reasonable in its approach and inquiry. Results The following results are displayed in a similar fashion as did the Dueling Windows panel. There were three categories of scoring; the vendor either achieved the task, achieved the task with difficulty, or failed to perform the task. For the purpose of our review of the seven vendors, "achieved task" means that the vendor was able to achieve the task with good efficacy and few keystrokes; "achieved task with difficulty" means that the vendor had to try repeatedly, the task required an inordinate number of keystrokes or searching, or multiple modes or cursors were required to achieve a task that was simpler in design than performing it; "failed task" simply means that the task was not accomplished. The panel rotated through the vendors so that no single vendor could always have the benefit of his predecessor's insight; the results here shall be by vendor in alphabetical order and will include additional comment for clarification. A comment "none" can mean just that, or it can mean that there was insufficient data to determine the cause of the outcome. It should be noted that this article is not to promote any single vendor, so it will concentrate its comments on GUI screen reader's ability rather than their failure. (More comment on the evolution of GUI screen readers will be presented in the Summary section of this article.) Each of the fifteen questions shall be presented, and followed with each vendor's performance. Note: The first 4 questions regard Microsoft Access, MS's database. 1. Find and launch Microsoft Access. WinVision: achieved task Comments: none outSPOKEN: achieved task Comments: none Window-Eyes: achieved task Comments: none JFW: achieved task Comments: none Screen Reader/2: achieved task Comments: none Window Bridge: achieved task Comments: none SPW: achieved task Comments: none 2. Using the file menu and dialog box, open the file called dwdb from the \dwfiles directory. WinVision: achieved task Comments: none outSPOKEN: achieved task Comments: none Window-Eyes: achieved task Comments: none JFW: achieved task Comments: none SR/2: achieved task Comments: none Window Bridge: achieved task Comments: none SPW: achieved task Comments: none 3. Open the employee table in view, arrow down and tab into the employee data, and demonstrate how to read the field data and field title. WinVision: failed task Comments: Could not read either cell contents or header. outSPOKEN: achieved task with difficulty Comments: Had to invoke mouse cursor to read header. Window-Eyes: failed task Comments: Could not read either cell contents or header; very verbose. JFW: achieved task with difficulty Comments: Read cell contents but had to enter mouse cursor mode and search to read header. SR/2: achieved task Comments: Screen Reader/2 was the only program that could read the field and header without entering into an additional mode or review mode, or having to press numerous keystrokes just to confirm the header. Window Bridge: achieved task with difficulty Comments: Had to re-boot. SPW: failed task Comments: none 4. Close the employee table window (CTRL+F4), and from the View menu select Query (Alt+V, Q). Now select the DWquery, and in two separate dialog boxes you will be prompted to enter data . In the first you will be asked to enter a position. In the second you will be asked to enter a country. Demonstrate your program's ability to find and read the appropriate prompts. (Editor's note: the prompt here is located on the line above the entry field.) WinVision: failed task Comments: Could not re-read edit fields. outSPOKEN: achieved task with difficulty Comments: Could read dialogue box, but did not speak automatically and needed to enter mouse cursor mode to re-read the field. Window-Eyes: failed task Comments: Very verbose; could not re-read edit field. JFW: achieved task Comments: none SR/2: achieved task with difficulty Comments: Read fields, but did not speak automatically. Window Bridge: failed task Comments: none SPW: failed task Comments: none Note: The following questions refer to operations in Excel, Microsoft's spread sheet. 5. Load MS-Excel and, from the file menu, select "open". Using the dialog box, open a file called invoice.xls from the \dwfile directory. (Note: MS-Excel is an example of how Microsoft does not necessarily maintain its programming conventions; this presents an obstacle for GUI speech developers.) WinVision: achieved task Comments: none outSPOKEN: failed task Comments: Had to re-boot and could not open file. Window-Eyes: n/a Comments: GW Micro declined to demonstrate their program in MS-Excel at this time, but would be working on access to it in the future. JFW: achieved task Comments: none SR/2: achieved task Comments: none Window Bridge: failed task Comments: none SPW: achieved task with difficulty Comments: Had to re-boot. 6. Have the speech announce the cell pointer location. WinVision: achieved task with difficulty Comments: Could not easily speak coordinates. outSPOKEN: achieved task with difficulty Comments: Could not easily speak coordinates; used review mode. Window-Eyes: n/a Comments: none JFW: achieved task Comments: Used INS-C macro. SR/2: achieved task Comments: none Window Bridge: failed task Comments: Could not read because of font type. SPW: achieved task Comments: none 7. Read the text in cells A1 to A6. WinVision: failed task Comments: none outSPOKEN: achieved task Comments: none Window-Eyes: n/a Comments: none JFW: achieved task Comments: none SR/2: achieved task Comments: none Window Bridge: failed task Comments: none SPW: achieved task Comments: none 8. Go to cell F22. Enter the 4-digit number provided by the panel; this number is too big for the cell. How does your program let you know this? WinVision: failed task Comments: Could not read cell contents or overflow. outSPOKEN: achieved task Comments: Had to read line instead of cell contents and overflow. Window-Eyes: n/a Comments: none JFW: achieved task Comments: none SR/2: achieved task Comments: none Window Bridge: failed task Comments: Could not read overflow. SPW: achieved task with difficulty Comments: none Note: The next section regards the World Wide Web (WWW) browser, Netscape; IBM used their own WWW access, Web Explorer. 9. Load the web browser; Get the URL of the current sample web site. WinVision: achieved task Comments: none outSPOKEN: achieved task Comments: none Window-Eyes: achieved task Comments: none JFW: achieved task Comments: none SR/2: achieved task Comments: none Window Bridge: achieved task Comments: none SPW: achieved task Comments: none 10. When on-line and going to a new location, many times the new web page takes some time to load. How do you determine when the entire page has been received or downloaded? WinVision: achieved task Comments: none outSPOKEN: achieved task Comments: none Window-Eyes: achieved task Comments: none JFW: achieved task Comments: none SR/2: achieved task Comments: SR/2 has a sound feature that indicates if a download is still in progress. Window Bridge: achieved task Comments: none SPW: achieved task with difficulty Comments: none 11. As you move up and down through the web page, how do you determine when you are at the top or bottom? Demonstrate this by finding the bottom of the current page. WinVision: achieved task Comments: none outSPOKEN: achieved task Comments: none Window-Eyes: achieved task Comments: none JFW: achieved task Comments: none SR/2: achieved task Comments: none Window Bridge: achieved task with difficulty Comments: none SPW: achieved task with difficulty Comments: none 12. Find and move to the link on the screen. Now, click the "Let the Internet duel begin" link. WinVision: achieved task Comments: none outSPOKEN: achieved task Comments: none Window-Eyes: failed task Comments: Found link but could not select as the cursor was not on the link. JFW: achieved task Comments: none SR/2: achieved task Comments: none Window Bridge: achieved task with difficulty Comments: none SPW: achieved task with difficulty Comments: Had trouble distinguishing between hypertext link and standard text. 13. If a link is not visible on the current page, can your program find it on other pages without going to those pages? Note: This question was deemed invalid since sighted users would not be using Windows in this fashion. Though not discussed at the seminar, a link Goto/locator might give blind users an advantage over their sighted colleagues. 14. Perform the following seven-step process to get to a specific page: Click on the netsuite products link; Page down twice; Click on the WebWhacker link; Page down three times; Click on the Windows 3.1 v1.0 link; Page down once; Click on the download link. You are now at a form. Please fill in the last name and first name fields. WinVision: failed task Comments: Could not find entry field. outSPOKEN: achieved task Comments: none Window-Eyes: failed task Comments: Could not get mouse pointer in edit box for data input. JFW: achieved task Comments: none SR/2: achieved task Comments: none Window Bridge: achieved task Comments: none SPW: failed task Comments: Could not get mouse pointer in edit box for data input. 15. When online in a newspaper with multiple columns, read a single column. WinVision: failed task Comments: Could not establish single-column window to read. outSPOKEN: achieved task with difficulty Comments: Was the only program that could read newspaper columns, but with many keystrokes and some column defining. Window-Eyes: failed task Comments: Could not establish single-column window to read. JFW: failed task Comments: Could not establish single-column window to read. SR/2: failed task Comments: Could not establish single-column window to read. Window Bridge: failed task Comments: Could not establish single-column window to read. SPW: achieved task with difficulty Comments: Could not establish single-column with speech, but could use braille display to differentiate between columns. 16. Start MS-Word. From the file menu open a file called dw.doc from the \dwfiles directory. From the tool menu, select spellcheck. Using the tab key, now demonstrate your program's ability to read the items in this dialog box. WinVision: achieved task with difficulty Comments: Had to tab to short keys to read them; did not spell misspelled or optional words. outSPOKEN: achieved task with difficulty Comments: Had to tab to short keys to read them; did not spell misspelled or optional words. Window-Eyes: achieved task with difficulty Comments: Had to tab to short keys to read them; did not spell misspelled or optional words. JFW: achieved task Comments: none SR/2: achieved task Comments: none Window Bridge: achieved task with difficulty Comments: Had to tab to short keys to read them; did not spell misspelled or optional words. SPW: achieved task with difficulty Comments: Had to tab to short keys to read them; did not spell misspelled or optional words. Summary and Commentary We gleaned a few conclusions from Dueling Windows. It showed that we have come a long way in GUI screen reader development since only a few, short years ago when access to GUI appeared a dismal prospect. Dueling Windows further showed, however, that we have a long way to go in GUI screen reader development. Few tasks, by any vendor, could be easily performed with a single keystroke or short series of keystrokes; though certain individual tasks could be performed relatively easily, access to an entire process or series of tasks seemed burdensome and never completely straight-forward. The seminar brought to light some important points to consider regarding this. DOS was not accessible in its early state, and is often still inaccessible. Access to Windows, however, is occurring at an earlier stage in Windows development than did access in DOS's development. Also, if certain programming conventions are upheld in Windows, Windows may become much more predictable and workable than DOS did; remember that GUI access is in its fledgling state and will become much more sophisticated. One key issue for access to Windows is the need, or lack thereof, to build one's own speech configurations or profiles. If an access program works well with a strict, conventional set-up of an application program, but fails to easily be configured for a slight variation of the original application program, this is obviously not desirable. Also, key issues for GUI screen reader users are the ability to re-read text and fields that have been spoken once as a window opens, and the ability to spell words in any area of the screen where they are only spoken in context. Since Windows is somewhat predictable and generally constant in its presentation, and, since certain Windows application programs have become more conventional than others, it is not unreasonable to assume that the differences between access programs will become more philosophy-based, rather than task-based. The conventions of Windows, or domination of Microsoft, essentially requires all GUI screen readers to eventually have full and complete access to these programs. The distinctions are likely to appear in the interface where more or less transparency is desired, etc. Lastly, Dueling Windows may be providing a marketing aspect that could benefit the users. Though certainly not all-encompassing, it injects a competitiveness into the screen reader market that might not otherwise be there. Much product development for the blind has suffered as a result of a lack of competition, so the user may benefit from this current surge in GUI access competitiveness. The only danger with this added competitiveness is if the small pool of developers and vendors in this market lose respect for another's product, or cease to work with other developers. Judging from the March CSUN event, this is certainly not the case yet, and SAF looks forward to the next Dueling Windows. Post Scriptum: SAF welcomes any or all of the GUI screen reader vendors to respond to any misrepresentations or errors this article may infer regarding their product. ---------- I'm redistributing an article that Ken Frasse, Editor of Access Review, posted to the GUISPEAK list. I made some formatting and typographical improvements for email circulation, but did not alter the content. ---------- This document is copyrighted by Sensory Access Foundation. As per request of Mark Nelson, SAF has published this review of Dueling Windows 95. Dueling Windows 95: GUI in Minneapolis by Kenneth Frasse (Published from ACCESS REVIEW, Vol. I, No. 4, Winter' 96 with permission of Sensory Access Foundation) Abstract At the Closing the Gap conference on technology and persons with disabilities held in Minneapolis, Minnesota, Mark Nelson of the Foundation for Blind Children hosted a pre-conference session entitled "Dueling Windows". This session compared graphical user interface (GUI) screen review software in applications for Windows 95. The format of the session was to pose identical tasks to each of the vendors in alternating order and have the vendors demonstrate the ability of their products to perform the specified tasks. Since SAF does not sell, distribute, or endorse any specific GUI screen reader, we were asked to generate an objective review of the GUI screen readers demonstrated. Editor Similar to the 1996 CSUN presentation of Dueling Windows, Mark Nelson of the Arizona-based Foundation for Blind Children coordinated another Dueling Windows at the October Closing the Gap conference in Minneapolis. The session was increased in duration to 8 hours, and it paralleled the 1996 CSUN presentation in that it was composed of questions and tasks that each of the vendors had to answer or perform. Notably, Kerry Malone-Cook of the Southern Oregon Goodwill, Greg Meise of the Kansas-based Missing Link Technologies, and Doug Wakefield of the General Services Administration joined Mr. Nelson as the four panelists who would present the tasks. This article contains 6 sections: 1. GUI Vendors: Which GUI software were represented at the session. 2. Technical Set-up: What was the technical set-up for the session? 3. Scoring: How were the GUI screen readers scored? 4. The Duel: The Dueling Windows session. 5. Results: The scores and statistics. 6. Summary and Commentary. 1. GUI Vendors. The GUI vendors represented at the session were: 1) Artic Technologies WinVision for Windows 95 2) Henter-Joyce JAWS for Windows 95 (JFW) v2.0 3) Syntha-Voice SLIMWARE Window Bridge for Windows 95 v2.30 Artic Technologies was represented by Stan Gowin, Henter-Joyce was represented by Glen Gordon, and Syntha-Voice was represented by Neil Cooper. MicroTalk's ASAW and TeleSensory's Screen Power for Windows have been developed for Windows 95, but both declined to participate immediately prior to the session. TeleSensory stated that they had a working version, but that they wanted to make it more robust before launching or demonstrating it into the general market. Alva Access Group's outSPOKEN for Windows (previously Berkeley Access) has been functional in Windows 95 since the beginning of the year, but could not make the session. Apparently, IBM's Screen Reader/2 will not support Windows 95, and no other developers/vendors participated in the session. 2. Technical Set-up. Each of the vendors were given identical IBM 133 Mhz, 16 Mb RAM Pentiums. On each of these machines, identical software was installed, including Windows 95, MS/Word, and MS/Internet Explorer. No online tasks were performed; rather, Internet information was retrieved directly from the respective hard drives. Each of the computers had a 17-inch monitor that was turned 180 degrees away from the participant and facing the audience so that the audience could review where the participant was in their manipulations. Additionally, a large overhead screen displayed to the audience the monitor of the individual vendor who was performing tasks; vendors were not allowed to visually review the monitor. All audible feed from the computers was fed into the main PA of the room, and vendors could only use the speech synthesis for their access. The same tasks were asked of all participants in a "round-robin" format. a different participant would start with each new task so that no one vendor could benefit from always seeing the other vendors' efforts. The questions and tasks were compiled from the panelists, vendors, developers, users, and professionals in the blindness field. All participants were given their computers the day before session to load their software, and they were also offered the source list of tasks and questions from which the final tasks and questions would be chosen. All three participants refused to accept the source list. No braille support questions were asked of the participants, though some vendors do support electronic braille displays. 3. Scoring. The GUI screen readers were not officially evaluated at the session; SAF has implemented its own scoring system for this purpose. The scoring of each GUI screen reader shall be accomplished through a rating system of zero to 4, and are defined as follows: (0)the GUI screen reader failed to accomplish the task; (1)the GUI screen reader failed to accomplish the task but spoke some relevant items; (2)the GUI screen reader accomplished the task with difficulty or numerous keystrokes that would generally prohibit a user from wishing to use the screen reader frequently for this task; (3)the GUI screen reader accomplished the task with little difficulty and few steps or keystrokes; (4)the GUI screen reader accomplished the task with no difficulty, single/few keystrokes, and little verbosity. 4. The Duel. There are three applications in which the session was performed: Windows 95, Microsoft Word, and Microsoft Internet Explorer. There were a total of 36 tasks to be performed, 35 of which could be scored, and the following details the results. Windows 95 1. Install your GUI software. WinVision: 3 Comments: Had to type "blind" since they did not have DOS screen reader to install W95 version of their GUI; system asks user if they want to replace some configuration files, but this is somewhat confusing and doubtful if the average user would know to which configuration files the system is referring; supports most makes of braille displays. JFW: 3 Comments: Can be installed from DOS or Windows 95; asks if user wishes to install the AZMEMORY driver, but this driver is for W3.1, not for W95 and is misleading; somewhat verbose, and had to jump between the JAWS cursor and "PC" cursor. Window Bridge: 4 Comments: Installs from DOS, and speaks installation even if no speech has been invoked; can be installed with a single command line containing specified parameters or by following the full installation procedure; auto-detects synthesizer; no extraneous chatter. 2. Does your program support Microsoft's Active Accessibility? WinVision: (no task involved) Comments: Will be working with AA. JFW: (no task involved) Comments: Supports AA, and will be released when AA is released. Window Bridge: (no task involved) Comments: Will support AA, and seeks to better support HTML documentation. 3. Highlight the first/top item in the start menu. Does your program automatically read the item ? WinVision: 4 Comments: Used CTRL-ESC then arrow keys. JFW: 4 Comments: Read item automatically without arrowing. Window Bridge: 4 Comments: Used CTRL-ESC then arrow keys. 4. Re-read the highlighted item. WinVision: 4 Comments: Used ALT-H to read item; used ALT-I twice to spell the item. JFW: 4 Comments: Used INSERT-8 (numpad) to read line. Window Bridge: 4 Comments: Used CTRL-SHIFT-5 (numpad) to read item. 5. For items that have an associated hot key, how do you get your program to read the letter? WinVision: 0 Comments: JFW: 4 Comments: Used "5" (numpad) to state hot key. Window Bridge: 0 Comments: Does not possess a re-read key for hot keys; reads hot keys while navigating. 6. Move to the "program" item by depressing the letter "P" and hit the right arrow key. Now hit the left arrow key. Does your program automatically read the "program" item name again? WinVision: 4 Comments: none JFW: 4 Comments: none Window Bridge: 4 Comments: Said both "program" and the item. 7. Go to help and go to the find tab. Does your program read the tabs? [CTRL-ESC, H, SHIFT/TAB, right arrow, right arrow] WinVision: 4 Comments: none JFW: 4 Comments: none Window Bridge: 3 Comments: Somewhat verbose; spoke help, but not "Find" tab. 8. Look for help on time. Arrow through the choices. Now, left click select the second time selection. Does your program read the highlighted/outlined text? [ALT-T, type "time", ALT-S, down arrow] WinVision: 3 Comments: Had to roam a little; routed WinVision cursor to PC cursor; used ALT-HOME for left mouse click. JFW: 3 Comments: Needed to route JAWS cursor to PC cursor; somewhat verbose. Window Bridge: 4 Comments: No routing of cursors needed. 9. Display the topics with Alt/c. use the arrow keys to move the highlight. Does your program read the highlighted text? Now, use the arrow keys to highlight changing your computer's time zone. Display help on this topic with Alt/d. Does your program read the contents of the help window? WinVision: 4 Comments: none JFW: 4 Comments: none Window Bridge: 4 Comments: none 10. Run "my computer." Then, demonstrate how to make your program read the window title. WinVision: 3 Comments: Says "dot" when at Desktop, not obvious; somewhat verbose; used CTRL-T to speak title. JFW: 4 Comments: Used INSERT-F6 to minimize all windows, JFW confirms; used INSERT-T to speak current title. Window Bridge: 4 Comments: Used task bar to go immediately to "My Computer"; used SHIFT-DELETE to speak current title. 11. Enter the File Menu and select Format. Set the settings as follows: 1.44 Mb, "Quick Erase", mark "no label", check the "display summary when finished" check box, and uncheck "copy system files". Does your program read the summary information when the format is complete? WinVision: 3 Comments: Some roaming; did not speak summary automatically; used CTRL-A to read summary dialogue contents. JFW: 4 Comments: Read summary contents automatically. Window Bridge: 2 Comments: Did not speak summary, but did speak some parts of the summary screen; has "Dialogue Manager" for "tailoring" dialogue boxes for reading. 12. Diskcopy the labeled disk to the unlabeled disk. Does your program tell you automatically when to insert the second disk and when the process is complete? WinVision: 3 Comments: Said to insert disk 2, but did not read dialogue upon completion; read dialogue with speak key. JFW: 2 Comments: Said to insert disk 2, but did not read dialogue upon completion; had to route JAWS cursor to PC cursor and arrow down to read status; read status and all buttons. Window Bridge: 3 Comments: Used "5" (numpad) to read status; used ALT-HOME to read dialogue box; did not read automatically. 13. Go to the Windows Explorer. Copy the file called "duplicate" from the CTG1 directory to the CTG2 directory. The file already exists in the CTG2 directory. How do you get your program to read the dialog box notifying you of this? WinVision: 4 Comments: none JFW: 4 Comments: none Window Bridge: 4 Comments: none 14. From the Windows Explorer run the "Find" utility and maximize the window. Now, find all files with the "bat" file extension (*.bat). What are the results? [ALT-T, F, F, ALT-SPACEBAR, X, *.BAT, RETURN] WinVision: 3 Comments: Only read first batch file found, but did not read other batch files. JFW: 3 Comments: Only read first batch file found, but did not read other batch files. Window Bridge: 3 Comments: Only read first batch file found, but did not read other batch files. 15. Place MS/Word on the Start Menu. Does your program speak all of the process? [CTRL-ESC, CHOOSE "SETTINGS" THEN "TASK BAR", CTRL-PGUP, CHOOSE "ADD" BUTTON, TYPE IN "C:\MSOFFICE\WINWORD\WINWORD.EXE", UP ARROW, RETURN, GO TO START MENU AND CONFIRM] WinVision: 4 Comments: none JFW: 4 Comments: none Window Bridge: 4 Comments: none 16. Install the program from the floppy we have provided. How well does your product voice the process? WinVision: 4 Comments: Spoke process, but did not speak during tabbing for "OK" button at end of installation. JFW: 4 Comments: Spoke process, but did not speak during tabbing for "OK" button at end of installation. Window Bridge: 4 Comments: Spoke full process. 17. Go to Windows Explorer and run LapLink. Find and label an unrecognized icon. WinVision: 0 Comments: Could not label icons. JFW: 3 Comments: Roaming; routing JAWS to PC cursor, and left/right arrowing to find icon; spoke label; can use "Graphics Wizard". Window Bridge: 4 Comments: Used CTRL-SHIFT-? for graphics menu, chose "auto-label; program labeled icons and stated how many were labeled. MS/Word 18. Start MS-Word. Retrieve the file called Mom.let from the "dwfiles" directory. Tap the page down key and then the up arrow key twice. Demonstrate your program's ability to read to the end of the document. WinVision: 2 Comments: Much jumping; did not confirm