• Conference Home
• About the Conference and Proceedings
  • Preface
  • Foreword
  • Accessibility
• Author Index
• Policy
• Practice
• Research
  • Computer Applications and Communications
  • Job & Environmental Accommodations, including Ergonomics
  • Outcomes and Quantitative Measurement
  • Public Policy
  • Seating and Wheeled Mobility
  • Technology for Cognitive and Sensory Impairment
  • Other
• Student Scientific Paper Competition
  • Computer Applications and Communications
  • Job & Environmental Accommodations, including Ergonomics
  • Outcomes and Quantitative Measurement
  • Seating and Wheeled Mobility
  • Public Policy
  • Technology for Cognitive and Sensory Impairment
  • Other

Patrick Clerkin, Stan Cronk, PhD, Puneeth Nimmagadda
Center for Biomedical Engineering and Rehabilitation Science
Louisiana Tech University, Ruston, LA 71270

Abstract

Morse code has many applications in augmentative and alternative communication. This form of indirect selection is useful because it requires very little physical skill and is a fast method compared to scanning (3). The disadvantage of Morse code is that it requires a large amount of cognitive effort to learn. It is often difficult for a clinician to determine if Morse code will be suitable for a client in the limited amount of time that he or she has with the client (1) (5). We propose that by building a model of how Morse code is input into the computer, we will gain a greater understanding of the intricate timings involved and the mental preparation time required to input each character. We have also developed a testing procedure that minimizes the learning times necessary to gauge communication rate. The result will allow clinicians to predict the communication rate that the client will be able to achieve once he or she has mastered Morse code.

Background

GOMS modeling is a technique in existence since its creation in 1983 by Stuart Card, Thomas Moran, and Allen Newell (2). It was originally intended to be a way of mathematically evaluating expert human performance pertaining to computer interfaces. Previous research has been done to model scanning with and without word prediction using the GOMS model (4). The GOMS model itself is an acronym for Goals, Operators, Methods, and Selection Rules. A goal is any task to be achieved and can be thought of as an objective of the user. An operator is a basic perceptual, motor, or cognitive act that allows the user to manipulate the task. A method is the sequence of steps that a user makes to accomplish a goal. A selection rule is a rule that determines which method will be used to achieve a goal if more than one is available.

When Card, Moran, and Newell introduced the GOMS modeling technique, they created two different models. The first was CMN-GOMS which was a loosely defined model of how to express goals and subgoals along with methods, operators, and selection rules in a hierarchy. The three authors also created KLM-GOMS which stands for Keystroke-Level Model. This technique breaks all model components into keystroke-level operators and does not include goals, methods, or selection rules. We chose the latter technique because of its simplicity, and the lack of methods or selection rules necessary when using Morse code.

The Model

Before we could understand the timings and mental load of Morse code, we needed to develop a model that would accurately represent how a user uses the code. Using the framework of KLM-GOMS, we developed Equation 1.

T = M + N_d • K_d + N_t • K_t + R (Equation 1)

T is the time to enter a single character. M is the mental preparation time required to access the code for that character. N_d and N_t are the number of dashes and dots in a character respectively. K_d and K_t are the time to initiate the key press plus the key press time itself for dashes and dots. Finally, R is the system response time, or the time for the computer to recognize that a character has been entered which was set to 700ms for our tests. We calculated N_d and N_t by taking a frequency of use table for the English language, and determining the average number of dashes or dots in a character. We found N_d and N_t to be 0.962 and 1.588 respectively. The remaining variables: M, K_d and K_t, are user independent and had to be determined through testing.

Testing

The GOMS model was designed to be used to evaluate expert performance, but we did not have enough expert users to evaluate our model. Instead we developed our tests to simulate expert use as closely as possible. We limited the number of characters being tested to six letters: E, T, N, H, O, and P. We chose these characters because they had a good mix of dots and dashes that were close to the ratio that we found for the entire English language. We also only tested words that contained these six characters. We allowed the user to practice entering these characters into the computer until they felt comfortable and their reaction times had leveled off. With this method, we felt comfortable that our subjects would accurately represent Morse code expert users for the purposes of our investigation.

We devised three sets of tests for each user in order to determine the mental preparation time as well as the K_d and K_t for each user. The first set of tests that we ran was a simple reaction time test. In this test we displayed the letter E on the screen and had the user enter the Morse code for it every time. The same character was displayed for each user at random time intervals for 50 times. The second set of tests was a Morse code letter test. Before starting this test we allowed the user to practice with a Morse code cheat sheet in from of him or her. Once the user felt that he or she had memorized every code we let the test begin. For this test we displayed our six letters on the screen and had the user enter the Morse code for that particular letter. We displayed the characters at random time intervals and recorded 50 data points with the user given a choice of looking at the sheet, and 50 without the sheet. The final test was a word test where the user was asked to input the Morse code for a particular word on the screen once it was displayed. The words were comprised of only the six letters that were previously tested, and they were displayed at random intervals. We collected 120 data points for each user for this final test.

Results and Analysis

We found that users had similar mental preparation times for single characters being entered, for characters at the beginning of a word, and for characters in the middle of a word. This finding leads us to believe that our equation holds for single characters as well as characters in a word.

Table 1 clearly shows the wide variety of mental preparation times for different subjects. Because all of the other timings are set, these differences, along with the differences in K_d and K_t will increase or decrease the communication rates of the users. Table 2 shows all of the variables along with the predicted time T to enter a character.

Table 1 : Mental Preparation Times of Users

User	M in ms
1	17.3111
2	293.6041
3	189.6669
4	19.5050
5	287.8111
6	332.4569
7	249.7833

 

Table 2: Inputs into Equation and Predicted Time

	M in ms	Kd in ms	Kt in ms	T in MS
User 1	17.3111	859.5366	422.0000	2214.7870
User 2	293.6041	832.7632	372.5161	2386.6960
User 3	189.6669	804.1463	470.5000	2410.9140
User 4	19.5050	875.5000	464.7778	2300.3090
User 5	287.8111	793.1622	461.0000	2483.3950
User 6	332.4569	724.3846	429.2963	2411.4960
User 7	249.7833	824.3947	420.3000	2410.7480

Conclusion

In this study Morse code was modeled with mental preparation time and key initiation/press times as the only variables in the equation. It would be possible for a clinician who is determining which device to use for augmentative or alternative communication to get these parameters from these simple tests and then determine what the client’s predicted communication rate would be. In this way Morse code can be looked at as a more viable option when determining which interface or device is appropriate for a certain user.

Further testing must be done to determine how long a user needs to practice using our simple technique to truly become an expert, but we conclude that our current findings are practical nonetheless because it will give a clinician an indication of a client’s predicted communication rate. Future research will also be conducted with various evolutions of the GOMS model including learning times and error rates, which will make our model much more comprehensive.

References

1. Beukelman, D. R., Mirenda, P. Augmentative and Alternative Communication: Management of Severe Communication Disorders in Children and Adults. Baltimore. Paul H. Brookes Publishing Co, 1998.
2. Card, S.; Moran, T.; Newell, A. The Psychology of Human-Computer Interaction. 1983, LEA Publishers.
3. Cook, Hussey. Assistive Technologies: Principles and Practice. St. Louis. Mosby, 2002.
4. Horstman, H.; Levine, S. Modeling of User Performance with Computer Access and Augmentative Communication Systems for Handicapped People. Augmentative and Alternative Communication, 1990, 6:4, 231-241.
5. King, Thomas W. Modern Morse Code in Rehabilitation and Education: New Applications in Assistive Technology. Boston. Allyn and Bacon, 2000.

Author Contact Information:

 Patrick Clerkin,
Center for Biomedical Engineering and Rehabilitation Science,
Louisiana Tech University,
711 South Vienna. PO Box 3185.
Ruston, LA 71270

EMAIL: pcl004@latech.edu