*p<0.05 **p<0.01

The results in Table 1 show a significant increase in correct spelling performance (18%) and in number of times the subject read the sentence before answering (66%). A significant decrease was registered in mean response-time (23 seconds) and in number of repeated dictations of the exercise the subject requested (25%©).

Data collected during the training sessions indicate similar trends for the different variables. H's remedial plan included work on 39 words involving seven rules, for which the system generated 31 tasks (total 394 exercises). Table 2 shows data for the relevant variables by the first and the last task of each rule, e.g., 111/16 - 111/21 indicates that task number 16 was the first treatment assignment for rule 111 and task number 21 the last one for that rule. The next column in the Table presents the number of occurrences of words related to a given rule until the rule was considered stable by the system, e.g., 47 occurrences (namely exercises) for rule 111.

Table 2: Data collected during remedial treatment

The results show a significant increase in mean grade for three rules, and no change in the (already high mean) for two rules. On the other hand a decrease in performance was observed for two rules (101 and 201), but it should be noted that H. worked on these rules for only two tasks (about 30 exercises each rule) which makes for a very short treatment period. Response time decreased between the first and last task of each rule. The request for repeated dictation was low in general and even decreased in all cases but one (rule 209).

 

Research question 2: Does the remedial plan support automation of correct spelling?

Two variables supply relevant information for this question: response time and correctness of the answer. Table 2 shows a cross tabulation between these variables for a total of 39 words.

The results show that for six words out of the nine for which H's spelling improved between the pre- and post-assessment, response time decreased. For a total of 21 words (out of 28 correct answers) correctly spelled by H. in the post-assessment response time decreased.

These data can be considered in line with the assumption that improvement in spelling performance together with decrease in response time implied an automation trend in H's spelling skills.

Research question 3: What characterized the remedial process in terms of the stabilization of correct spelling and the patterns of use of supporting tools?

Regarding this research question we will refer to data collected for the variables grade¨response-time and tool-use. For practical and clarity purposes, the great amount of data collected will be summarized in the following analyses in either quantitative or qualitative form.

Figure 1 shows the overall pattern of the time-response curve for all training sessions. The curve is a succession of peaks and troughs which appear in two main blocks: the basic and most massive block at the bottom of the curve ranging from 5 to 15 seconds, and a series of peaks reaching response times of 25-60 seconds.

Figure 1: Overall pattern for the response-time during the remedial treatment

This result will be best understood by looking closely at a segment of the curve with focus on a given word, as in Table 3. The focus word is (scorn) with some of its derivatives . The word first appeared in session 14 exercise 3 (or 14/3), and was correctly spelled by H. in response-time (rt) of 15 seconds. The word appeared again three exercises later (according to the session generation mechanism) in 14/6, with rt=7". Word appeared in 14/8, rt=17". Word in 15/8, rt=68", misspelled, and again in 15/11 rt=22", 15/14 rt=9", 22/5 rt=20", all three times correctly spelled. We can learn from this segment that peaks in response-time occurred each time a new word appeared (as in 14/3, 14/8 or 15/8), or an already practiced word reappeared several exercises away from its last occurrence (as with which did not appear between tasks 15/14 and 22/5). Moreover, as a word becomes more stable (namely, correct spelling along successive exercises) rt decreases, until reaching steady value within the range of 5-15 seconds.

Table 4: Stabilization process of a word and its derivatives

Data in the table suggest a two-stage pattern. First there is a correspondence between misspelling and recurrent use of a support tool ("picture" or "read-me-again") as in task 15/8. Following the use of a tool after misspelling, the word is then correctly spelled and rt decreases.

Looking at the complete set of data of H's spelling performance on 394 word occurrences, we observed that the pattern takes two different forms. During the first part of the stabilization process the pattern can be depicted as in Figure 2a. Leaks in spelling performance (misspelling) correspond to peaks in rt and in tool use. During the treatment sessions H. asked for a support-tool 51 times, 84% of these after misspelling a word. Once words reach stable status the pattern can be depicted as in Figure 2b. Recourse to tools decreases, as well as rt, which remains in steady low value.

Figure 2: Patterns of correspondence between spelling, response time and use of support tools

2a

2b

Observations and interview data

Technical features. H. learned to work freely with the computer system within a short period of time. She defined the system settings according to her needs (letter size, foreground and background colors, sound volume) in the very first session and did not change these throughout the treatment. At about the third task H. showed confident and skillful manipulation of the system's tools and work modes, and even expressed these feelings to the observer.

Spelling strategies. Through the remedial process H. developed varied strategies for accomplishing the spelling tasks. In correspondence with her level of awareness to spelling mistakes, and mastery of appropriate (formal or intuitive) spelling rules, these strategies evolved and became more complex and accurate. Following are some meaningful stages in the development of H.'s strategies.

During the first stages H.'s response to spelling mistakes was to ask first for repeated dictation. Then she typed letter by letter hesitantly, while speaking each letter out loud. In these stages H. strongly relied on contextual information (namely the whole sentence into which the target practice word fits) for solving the spelling tasks.

By about the eighth task H's attitude towards contextual information started to change. When a target word first appeared she focused on its location within the sentence, carefully typing it while listening the auditory feedback. But in subsequent occurrences of the same word she typed fast typing, no longer needing the recurrent reading of the whole sentence.

During task 23 H. first turned to use a support tool before typing a word which she found difficult to spell. In general terms it seems that in the last portion of the remedial sessions H. adopted a more cautious approach in solving the tasks. When a target word which H. misspelled in the past appeared as part of aUnew exercise, she typed it carefully and revised its spelling before pressing the confirmation key.

Renewed awareness to visual information. Of particular interest was H's response to exposure to visual information and her discovery of its functional contribution to her work. The visual properties of the user interface were specifically designed for low-vision users (e.g., size and location of icons, scroll window for very large fonts), and individual users could adapt interface features (e.g., size of characters, foreground and background colors) to their specific needs. H. showed interest in these visual features and willingness to benefit from them from the very first stages of the remedial sessions; she also made this explicit to the observer.

The adaptation of display features to her needs allowed her to read the target words as many times as needed before typing. By the seventh session she had already developed her own navigation strategies through the icons and tools in the screen. When looking for a support tool, she managed to move efficiently among the icons (working at 8-10 cm distance from the screen) until reached the one she wanted. This visual mode allowed her to access and activate many different and powerful features effectively and inUa very short time. By the eighth session H. already followed visually and by sound both the computer dictation of the target word as well as her own typing of it.

Motivation. During all stages H. was completely concentrated on her work with the "Pupil" system. Whenever she had the option between starting a new task and quitting the system she chose to continue to work. She often asked for additional work; in her own words:..."the letters and the sounds helped me very much...I would like to continue myUwork with the computer". H. was a shy girl, and even if she did not often speak directly about the work process, her body language and gestures revealed her satisfaction.

When asked whether she felt that the remedial process had helped her in writing also in class and at home, she answered that she feels that ..."it helped, ...my spelling mistakes decreased...I enjoyed working with the Pupil system".

Discussion

The lack of appropriate instruction and of quality support tools seriously affect the acquisition process of correct spelling by low-vision children. Given this population's high level of heterogeneity regarding vision capability and spelling knowledge and skills, it seems obvious that any solution to be devised should satisfy a comprehensive setUof requirements: individualized treatment according to vision and spelling levels; rich and varied alternative information channels to compensate for the impairment in the visual channel; adaptive generation of remedial plans to adjust to individual improvement paths; flexibility in time demand and composition of remedial tasks to accommodate to individual pace; continuous feedback by varied means to allow students to evaluate their own performance while they accomplish the tasks; high sensitivity toUnuances in students' learning and improvement process to ensure utmost possible adaptation to their needs.

The case study reported in this paper focuses on one student's work with a computer adaptive remedial tool developed along these lines and requirements. The first level of observation looked at the results of the system's comprehensive diagnostic test. The test offered a detailed picture of the child's spelling performance in a wide range of cases and spelling rules. In line with our previous pilotUdata we found that the most frequent category of misspelling in the diagnostic tasks was letter-shift, in particular among the Hebrew characters .

The next level of observation considered the effect of the work with the computer tool on the student's spelling performance. The pre- and post-test comparison showed significant improvement in spelling after completion of the remedial plan. A more detailed view however was obtained by collecting data on particular variables of the student's performance, e.g., response time, stabilization stages for words and rules, frequency and context of use of support tools. The more detailed picture thus obtained was equally encouraging. Let us consider several salient issues regarding changes in the student's performance.

Saimon and Saimon (1973, 1979) suggested three operational spelling modes, in correspondence with the amount and quality of knowledge a subject owns about a target word. These modes are (a) automatic retrieval, (b) spell-and-check cycle, and (c) phonetic writing (based on sound mimicry). The last mode has the highest potential for misspelling, particularly for letters with phonemical similarity. It is also the most frequently used by low-vision children, who do not rely on reference information from the visual channel but on incomplete (and often incorrect) word models stored in memory. The detailed data obtained in this case study show a clear evolution path in the student's performance. From phonetic writing in the first sessions theUstudent moved to process-writing, typing the word and checking it. Initially the student typed the word using her partial spelling knowledge, then she looked for help from the system's support tools, and finally evaluated the result according to correctness criteria stored in her memory. In the advanced stages of learning H. was able to automatically retrieve the spelling of the word. This process is reflected in the overall trend observed by correlating correctness (increasing) and time-response (decreasing) of the student's answers along the training process and between the pre- and post tests.

Another central issue concerns the role of the support tools in the system. The tool most frequently used by H. was "spelling" (literal auditory spelling of the target word), which she considered the fastest help. This tool seemed to have a twofold effect. The first and most obvious was that of supplying immediate help in solving the task. But in addition it served H. to develop models of the words by relating auditory information to visual information about particular letters and letter configurations. As H's work advanced she started to replicate the system's spelling by herself before and during the actual typing of the target word, in the process of the cognitive assimilation of the word as automatically-retrievable unit.

H's use of alternative information channels deserves special attention. As could be expected, support relying on the auditory channel was of great use. But as the work proceeded the use of the visual channel increased. The student spent increasing time reading the sentence before typing the target word (reinforcing the auditory information she got from the system). At the same time the demand for auditory support (e.g., repeated uttering of the target word by the system) decreased. This readiness for effective utilization of existent visual capability could be attributed both to features of the system enabling the adaptation of the working environment to individual needs and constraints (e.g., font size, font and background color), and to motivational aspects (e.g., sense of improvement, of effective interaction with the system, of varied support in solving the task).

A final remark should be made about our research agenda. This comprehensive case study led us to a preliminary set of conclusions which will serve for the formulation and planning of forthcoming systematic studies. We have already collected additional case-study information, as well as started a systematic study with a larger group of low-vision children. Examples of issues we are currently looking at are, among others: the differential contribution of the diverse types of tools (e.g. perceptual, conceptual, intuitive rules) on the students' spelling performance; a detailed tracing of the process by which automation in spelling and model word retrieval evolves; students' strategies of spelling acquisition and tool use; ability to generalize and apply spelling problem-solving, rules, and words models beyond the spellingUset targeted by the current version of the system.

As regard the role of advanced technology in supporting populations with special needs, we have no doubt that the "Pupil" system has the potential to contribute to low-vision children's' quality of learning and functioning in the literate environment.

References