1. Introduction and the Goal of AAC
An estimated ten to twenty million individuals worldwide experience a speech communication disorder so severe that they cannot express themselves effectively to others. This number is the extension of commonly referenced demographics (Beukelman & Ansel, 1995). Causes can range from congenital conditions such as cerebral palsy to acquired conditions such as amyotrophic lateral sclerosis (ALS), or motor neuron disease (MND), stroke, and injury. For this population some form of augmentative and alternative communication (AAC) intervention would be indicated. AAC solutions range from unaided systems such as sign language to aided systems such as speech output assistive technology. Figure 1 shows a speech output AAC system.
Communication has a significant impact on personal achievement. Therefore, the goal of AAC must be the most effective communication possible for the individual. This goal must be honored in the processes of selecting and implementing an AAC system. However, clinical decisions frequently are made without much assistance from a research base (Beukelman & Miranda, 1998) and without quantitative data on the communication performance of the individual being served.
Automated performance monitoring is a relatively new concept in AAC service delivery. Language sample logfiles collected with these modern tools and methods now are being analyzed to report quantitative summary measures of performance. This data is being used to guide therapy and to measure outcomes.
Figure 1. Speech output AAC system.
2. AAC Assistive Technology
AAC assistive technology is comprised of three essential elements: a means of selection, a language representation method, and outputs (Romich, Vanderheiden & Hill, 2000; Lloyd, Fuller & Arvidson, 1997). The means of selection is a function of the physical ability of the individual who will be using the system. Direct selection refers to those methods that are pointing in nature. The most common of these would be the standard keyboard or touch screen. An alternative might be head pointing systems in which head position determines the position of an indicator on a screen or in an array. Also, a two dimensional proportional control, such as a joystick or mouse, could be used to make selections.
A generally slower method, but one that is less demanding in terms of physical ability, is scanning (Koester & Levine, 1994a). In row-column scanning a single switch can be used to control a scanning process in which rows of selections and then individual items are presented to the user in sequence. Activation of the control switch changes the scanning sequence or makes the selection.
Finally, coding methods such as Morse Code can be used to make choices. Coding methods can be implemented in ways that do not require either visual or auditory observation of feedback from the system. Therefore, they can become automatic and efficient.
The language representation method is the interface between the means of selection and the generated communication. Language representation methods commonly used in AAC systems are single meaning pictures (including graphic symbols), alphabet-based methods, and semantic compaction (Hill, 2000; Romich, Vanderheiden & Hill, 2000). These methods provide access to the words in the vocabulary of the individual. Core vocabulary words are those few hundred words that constitute the vast majority (85-95%) of what is said. Extended vocabulary words are the remaining words, generally numbering in the thousands, that make up the remaining 5-15% of communication.
With single meaning pictures, as the name implies, a picture represents a single word. While a simple concept, it requires a number of pictures equal to the number of words in the vocabulary. A normally developing three-year old has a vocabulary of about 1000 words (Owens, 1996). A set of 1000 pictures becomes very difficult to access in a way that allows fluent communication to occur. For example, this relatively small vocabulary would require twenty screens of fifty pictures each. For just 250 words, five screens would be needed.
Alphabet-based methods include spelling, word prediction, and letter coding. Spelling is attractive from the perspective of the size of the symbol set. In many languages, 25-30 letters (depending on the language) and a space allow anything to be spelled. The adverse side of spelling is in the number of selections that must be made to convey meaning. This may be slow, resulting in a low communication rate. One effort to expedite the process was the development of word prediction systems in which a computer would guess the word that was being spelled. While this technique reduced the keystrokes, communication rate was not enhanced (Koester & Levine, 1994b). Letter codes, or abbreviations, can be used with some success. However, conflicts arise with even a small vocabulary. Figure 2 shows the display of an AAC system configured for spelling.
Semantic compaction uses short sequences of symbols from a small symbol set to define words and commonly used phrases (Baker, 1986). These multi-meaning icons fit on a single selection area, so that multiple pages are not necessary. As with other methods, training is needed for proficient use of semantic compaction. However, performance in access to core vocabulary has been demonstrated as superior to that of other methods (Burger, 1997; Hill, 2001). One reason for this is that the static (non-changing) user interface of the semantic compaction method encourages the selection process to become automatic. Unlike word prediction, the semantic compaction keystroke reduction relative to spelling results in enhanced communication rate. The adverse side of this method is that only those words previously stored in the system can be selected and spoken. Figure 3 shows the display of an AAC system configured for semantic compaction.
For many people who rely on AAC, two or three of the language representation methods are used. For example, core vocabulary may be accessed using semantic compaction or whole words, and extended vocabulary may be accessed using either alphabet-based methods or single meaning pictures, depending on literacy skills.
