@part(assum,root"thesis.mss") @chapter(Testing assumptions about the use of the program). @label(assumptions) @section(Introduction) @label(ass-intro) In chapter @ref(design) a number of assumptions were made about the way in which the program described would be used. These assumptions were tested in two studies, described in this chapter. Children from the Reading Unit took part in these studies. There were three main purposes of the first study: @begin(enumerate) to test whether children with spelling disabilities would be both able and willing to use a simple text editor to write stories; to examine the ways in which the text editor was used by the children, and to consider ways it might be extended; to provide samples of the children's errors, made in free writing on pre-specified topics. @end(enumerate) In the second study, the editcost spelling corrector was used to test whether or not: @begin(enumerate) a child would use the spelling corrector to check the spelling of words; he would be able to select the required word from those presented by the correction program; he would recognize when the required word was not presented by the correction program. @end(enumerate) @newpage @section(Study 1) @label(ass-pilot1) @subsection(Introduction) @label(pilot1-intro) The dyslexic child is not usually highly motivated to write using pencil and paper. He is also unwilling to proof-read his work; to look for errors, check them and correct them. It is argued, in chapter @ref(theory), that the provision of computer based tools, such as text editors, would increase his motivation to write. He would use simple text editing commands to draft his composition, to make changes to it, and to correct the errors. The final copy that he produces would not reveal errors made in earlier attempts: it would be a piece of work that he would not feel the need to conceal. Additionally, by not constraining the subject matter and vocabulary that the child uses in his writing, he will be more interested in, and see more purpose to, his production. The child must be able to cope with the commands needed to use the text editor, and to use the keyboard to input his text and commands. Typing itself may create problems: different skills are involved in typing and handwriting. In typing, any spelling information available through the kinetic sense is lost: the child is not concerned with 'making the shape of the letter' but with recognising it and pressing the appropriate key on the keyboard. 'Typing errors' may be introduced, but these are more likely to occur with typists typing at speed. 'Handwriting errors' will be unimportant. Additional confusion may be caused by the fact that the keyboard is labelled in upper case: the child is more familiar with lower case. The added feature of the delete key may outweigh the difficulties of using an eraser. Text editors have been used successfully by children who have no specific spelling difficulties @cite(sharples). Evidence is needed to demonstrate that children with these difficulties can also use them. The argument that the child will be able and motivated to write, using the computer, has to be supported: if evidence cannot be provided then the tool, as designed, would be of no use. Information gathered through the observation of the children in this study influenced the design of the program described in chapter @ref(design). It should be noted that no spelling correction was provided for use in this study. The way in which the correction programs might be incorporated was considered. The child was encouraged to guess the spelling of words initially and to ask for help with correcting errors he spotted when proof-reading. In effect, the investigator performed the function of the 'spelling corrector' here. All the spelling errors made by children in this study were recorded. The corpus of errors is given in appendix @ref(app-corpus). These errors were used in the design and implementation of the spelling correction programs (see chapter @ref(detail)). @subsection(Method) @label(pilot1-method) @begin(verbatim) @b(Subjects) @end(verbatim) @label(pilot1-subjects) The children attending the Reading Unit usually did so for one or more sessions a week, each session lasting approximately one and a half hours. Those normally attending during two particular sessions were selected to take part in the study. There were four children in the first group (group 1), and three in the second group (group 2). Those children in group 2 (J.M., G.Q., and M.W. - all boys) were considered by the Reading Unit teacher to be moderately able (when compared with all children attending the Unit). Three in group 1 (S.S. - the only girl, N.M., and C.M.) were considered by the teacher to be the least able ("hopeless cases?"). The fourth child in group 1 (L.B.) was thought to be very bright. Exact details of age, I.Q. scores, and reading test results were not provided. All the children had difficulties with writing, and some also with reading (although their reading had generally improved whilst attending the Unit). Specific problems were mentioned in the cases of two children: S.S - dysphasia, and C.M. - auditory perceptual difficulties. Several children also had behavioural problems (G.Q., M.W., and L.B.). The aim of this study was to observe how well the children coped with the system, and to collect data on the children's errors: it was not to assess improvement in the children's abilities, and so no control group was used. @begin(verbatim) @b(Apparatus) @end(verbatim) @label(pilot1-apparatus) A simple text editing program, called Walter (Word ALTERer), was used by the children. This program, written in POP-2, was developed by Sharples, and has been used by him to investigate children's use of written language @cite(sharples). Walter allowed each child to write new stories, to edit them, and to store and retrieve them. A summary of the commands available is given in appendix @ref(app-assum). The program was run on the Artificial Intelligence (University of Edinburgh) Department's PDP-11/60 computer under UNIX. Each child had his own user number and copy of the program. Each used a keyboard V.D.U. to type text into the computer and a printer provided final copies of the stories. Everything which appeared on the V.D.U. screen during a session was recorded. (These records were not available for the first session, and occasionally lost in later sessions due to technical malfunctions). Worksheets were used in the first few sessions, providing instructions for the use of Walter. A summary sheet of commands was used for most later sessions. @begin(verbatim) @b(Procedure) @end(verbatim) @label(pilot1-procedure) Each group of children attended for six sessions. Group 2 started two weeks before group 1, so the project ran for eight weeks in total. All children attended all sessions, with the exception of S.S who missed two sessions. The actual time spent using the computer, for each child, was 45-55 minutes per session. The children were collected from the Unit and brought to the A.I. Department, and then returned to the unit after the session, (hence the reduced time spent using the computer). The first part of each session was spent discussing the topic on which the stories were to be based. The first session was concerned with "How to use Walter". Later sessions involved either preselected topics (e.g. turtle logo, an adventure) or topics arising from general conversation on the way to the A.I. Department (e.g. horror stories). After some discussion of a topic, the group were taken through to the terminal room, and were logged onto the machine by the investigator. During the first session worksheets were used to familiarise the children with Walter. In later sessions they wrote stories based on the topics discussed at the start of the session. They typed each story straight into the machine. They were encouraged to write whatever they wished, and to attempt to spell words correctly, but not to worry unduly if they were unsure of a spelling - to guess it if they didn't know it. When the story was complete (that is, when they were sure that they had finished it, or at the end of the allocated time), they saved it in a "memory" file. At this point they were asked by the investigator to indicate those words in their story that they thought that they might have misspelt. (The amount of prompting at this stage varied). Each child was then prompted to correct those words misspelt. When all spelling errors (and some of the grammatical errors) had been corrected the story was saved and then printed out. Each child received a copy of all of his stories, corrected and neatly printed. Notes were made after each session (based upon observation and recordings) of how each child coped with the keyboard and with Walter. Each child's spelling errors and problems arising during the session - with the program, or with the machine - were also recorded. Features such as the amount of help each child was given with spellings when first writing a story, and how much he was prompted when using different commands, were varied. Different children required varying amounts of help. Consideration of the most effective ways of using the program was influenced by the results of these variations. The children's attitudes were assessed informally throughout the project. There were asked how they felt about the project whilst going to and from the department, and during the last session. Some feedback on the children's attitudes was also obtained from the Reading Unit teacher. @subsection(Results) @label(pilot1-results) Over the six sessions, five short stories were written by N.M., three by S.S. and four by each of the other children. Copies of the corrected versions of the stories are provided (in appendix @ref(app-assum)). Lengths of the stories varied from twenty words (N.M.,"Peter (my cousin)"), to ninety-eight words (G.Q.,"Blackhand"). (The story "Scream" by J.M. was one hundred and ninety-four words, but was actually written away from the terminal). Information about the length of their pencil and paper stories was not provided. The children were caused some frustration initially by the problems in logging on, machine crashes, and by being thrown out of Walter when they leaned on the keys whilst thinking! These problems were resolved in later sessions. It had previously been suggested that making the children type their stories might cause then to make errors. They were very slow typists, and tended to hover over the keyboard searching for a letter. Because they were slow, however, they did not make many actual typing errors (and were able to correct those that they did make). There were also no problems caused by using the upper case keyboard (whilst seeing lower case on the v.d.u.), and the 'qwerty' keyboard presented no special difficulties. The children learned to use the special function keys easily, (e.g. 'return', 'delete', 'shift', and 'space'), although they occasionally forgot to press the 'return' key at the end of a command line, and caused problems when using 'back-space' instead of 'delete'. The Walter editor commands were all used by most children. After the first session, little difficulty was encountered with them. The main problem was that of typing a filename instead of a command after the prompt e.g. "W: train" instead of "W: recall train". There was also some misspelling of commands. In general, all the children in the two groups got on well with the system. The children in group 1 worked particularly well together, often offering suggestions and advice to each other. This was the less able of the two groups, but once they had got used to the system, they were able to use the editing commands. They had more difficulty in judging whether words that they had written were correct or not, and in correcting them (with the exception of LB), but all they all appeared to be well motivated. The children in group 2 were generally better at using commands, wrote longer stories, were better at correcting errors, but were more easily distracted, and not as enthusiastic and hard-working as those in group 1. More detail is given about individuals in appendix @ref(app-assum). The children said they enjoyed using the computer to write stories, and appeared to do so. Some (especially in group 2) were less willing to go back and read over their work and make changes, whilst others were very keen to correct their errors. All, however, persevered to produce a 'good error-free copy' of their compositions. They also requested extra printed copies, on a number of occasions, to show to friends and family. One child commented that he preferred writing with the computer to writing on paper, because if he made an error it was easy to correct it without it showing with the computer, whereas if you had to rub it out there was always a hole in the paper! Group 1 were especially excited about using the computer, and children in both groups were keen to come back to do further work. The Reading Unit teacher sat in on several occasions: she was pleased with the work produced and happy for the children to take part in later studies. It was found that the children generally coped well with using the text editor. There were some problems: the 'change' command changed all occurrences of a string in the text: if only one specific occurrence was to be changed the string had to be put in context; the Walter program decapitalised all first words in the 'change' function, causing some confusion when changing capitalised words - this was solved by including the preceding (lower case) word in the change; each new piece of text had to be added as a separate 'new' story, and could not be added to an existing story; the original story might be saved, then changes made and a copy printed, but no copy of the corrected version saved. The results obtained from the collection of errors made by the children are summarised and discussed in chapter @ref(detail), figures 7-5 and 7-6. @subsection(Discussion) @label(pilot1-discuss) The children learned to cope with problems with the editor, as any user learning to use an unfamiliar text editor would have to. A different text editor might have been more suitable for their use, or Walter altered. Over a long period of use the children would get more practice with the keyboard and editor. A screen editor might, however, have alleviated some of these problems. The children were asked to guess the spelling of words, and to ask the investigator for help when correcting them. It was sometimes difficult to resist giving the child the correct spelling initially, without encouraging him to guess it. It was also difficult to deal with all the children who wanted help with correction at any one time. It usually took two sessions to complete a story - one to write it and the second to proof-read and correct it. This meant that sometimes a child forgot what a word was meant to be; for example, one child had used the word 'fapyer' in one session but had forgotten by the next session what it was meant to be - it was never corrected! The incorporation of the spelling corrector, as a 'check' function, would solve these problems. The child would be able to exit from writing text temporarily and ask the program to 'check' the spelling of a word, and would be given the correction (or possible corrections). He would get immediate help, and not have to wait until the teacher/invstigator was free. He would then be able to write his story, checking the spellings he was uncertain of as he went, and could also use the facility when proof-reading. The children expressed an interest in having such a facility. From the results it can be seen that the children were able to write using the computer. They had a large amount of control over what they wrote, and their errors were not seen after they had been corrected. They were able to produce a corrected, printed copy of their stories, which they could show to other people. Their writing had a function, in communicating their ideas, and was not just 'an exercise in a book'. No assessment was attempted of improvements in their spelling ability as a result of interaction with the computer, however. @newpage @section(Study 2) @label(ass-pilot2) @subsection(Introduction) @label(pilot2-intro) In order to use the spelling checking facility described in chapter @ref(design), the child types in his spelling of the word he wishes to check, and then selects the required word from those options offered. Evidence from the first study supports the need for this on-line checking facility. The child cannot ask the teacher for help with the spelling of every word, and it would be difficult for him to check the spelling using a conventional dictionary. It is necessary to test whether or not the child would be able to select the required word from those offered by the checking program. If the child is unable to recognize the intended word when presented with it, the checking facility would be of no use. If the program were always to offer the correct word, and no others, then this would not need to be tested. It is not possible to guarantee that the closest match to the checked word will be the required word, however. If the word is in the dictionary, and the checked word is a reasonable enough approximation to it, then the four closest matches should include it (section @ref(perform) gives detail of the program's success in this). 'Reasonable enough approximation' here means that the spelling could normally be recognized, by a competent speller, as the intended word. It is not desirable for him to select the wrong word from those being offered: this might reinforce learning of a misspelling. If bizarre and unrecognizable words (relative to the children's usual spellings) are produced, then it is not wholly desirable that the program should correct them: there would be incentive for the spelling to be 'as close to correct as possible'. It was hypothesized in chapter @ref(background) that the child would be able to recognize correct spellings even if he could not produce them. This hypothesis is further discussed in chapter @ref(theory). It was tested in this study. Study 2 tested whether the child would use the spelling correction facility, and also examined the ways in which it was used. The text editor was not incorporated, however, nor was specific phonetic information about words used (see chapter @ref(design)) to select options for corrections. The editcost program, plus lookup and addword facilities was used. The reason for use of the editcost spelling corrector, but not the phoncode corrector, was that it was desirable to test whether the checking program could be used at a stage when the phoncode program was not fully developed. The result was that fewer options were provided for the misspellings: this was, therefore, a more stringent test. This study was designed to answer a number of specific questions: @begin(enumerate) If the child is unsure about the spelling of a word does he use the computer spelling program to check it, or does he guess the spelling? If he checks the spelling, is he able to select the required word from those offered by the program in cases where the word is present? If the required word is not offered when checked, does the child realise this? More generally, in what ways is the program used? @end(enumerate) Information obtained in this study was also used in evaluating the performance of the editcost program (reference chapter @ref(perform)). @subsection(Method) @label(pilot2-method) @begin(verbatim) @b(Subjects) @end(verbatim) The children who took part in this study were selected on the basis of their availability at the time of the study. The first group consisted of four children: D.R., F.R., D.V., and T.E., all boys aged between nine and eleven years. They were considered to be of average ability by the Reading Unit teacher (in relation to other children attending the Unit). In the second group there were four children: three boys, D.S., S.T., and G.R., and a girl, M.A., all aged between eleven and twelve years. The Reading Unit teacher considered the children in this group to be exceptionally bright (reported IQ`s : G.R.>128, others > 135). All the children had difficulty with spelling, and some also experienced some difficulty in reading. @begin(verbatim) @b(Apparatus) @end(verbatim) The children used a version of the editcost spelling checking program. This program enabled them to check the spelling of any one word by comparing it with words stored in an on-line dictionary and displaying the four closest matches to the child`s word/misspelling. This checking algorithm is discussed fully in chapter @ref(detail). The main example used in chapter @ref(detail) is based on a session in which one of the children was using the checker. The on-line dictionary differed for each child, for each session (see section @ref(design-dictsize)). Each dictionary consisted of a file of commonly used words, including those words used frequently by the particular child, and a "topic vocabulary" based on the words likely to be needed for the particular session's task. For a large number of the words, definitions and examples of their use were added to the dictionary. This information was either provided automatically or on request. The version of the spelling checking program used in this study was written in PASCAL and was run on the Artificial Intelligence Department`s VAX 11/750 under UNIX. Each child had his own copy of the program and used a keyboard v.d.u. Everything that appeared on the v.d.u. screen during a session was recorded. Notebooks were provided for the children to write their stories in. @begin(verbatim) @b(Procedure) @end(verbatim) Group 1 attended for eleven sessions, over a period of 3 weeks. Of the four children D.R. was absent for 5 sessions and F.R. was absent for session 11. T.E. and D.V. attended all sessions. Approximately forty-five minutes per session were spent at the computer terminal. Group 2 attended three (weekly) sessions, each of one and a half hours. All four children attended all three sessions. Each group focussed on a particular project during the study. Group 1 focussed on producing articles for a magazine: topics included a sports review, a horror story, a visit to the departmental workshop and an interview. In the first session they were shown how to use the "check" program (the interactive version of the editcost program), whilst writing a description of themselves. Group 2's project was a report on the Department of Artificial Intelligence. They interviewed members of staff and were given demonstrations of computers and programs. They were shown how to use the check program, whilst writing up their interviews, in the first session. At the start of each session some stimulation for writing was provided: e.g. each child carried out an interview, or saw a computer demonstration, or there was a group discussion focussing on a particular task. Once they had decided what they would be writing about, each child sat next to a computer terminal and started writing. The spelling checking program was running on the computer. They wrote in pencil in an exercise book. They were told that whenever they were unsure of the spelling of a word they were to try to write it anyway, and then to use the checking program to check it. The program would offer them four possibilities for the word that they wanted. If they thought that the word they needed was among these four they were to cross through their written attempt (if it was not the same) and write in the correct word. They were requested not to rub out or scribble over any words. If the word was correct (the same as the selected option) they were to leave it as it was. It was emphasized to them that this was to be their working draft, in the exercise book, and that it did not matter if it was a bit messy as it was to be typed later anyway by the investigator. If they thought that the word they needed was not present in the four offered by the checking program, they were told that they should look at their spelling again. If they were not certain it was correct, they were to check it again with a different spelling. This will be referred to as 're-checking' the spelling. However, if they believed that it was correct but not in the dictionary they should tell the investigator who would then add it to the dictionary. In these cases, the word was added immediately without the child seeing it, and the child checked his original misspelling again. When they had completed the piece of writing they were asked to proof-read it and to check any words that they were still unsure of. Varied instructions were given regarding the use of the lookup facility. For the first week (group 1, sessions 1 to 4; group 2, session 1) the children were not told about the facility, as the dictionary definitions were not complete. For the following sessions, group 1 were told that they could "lookup" a word either to see if it was actually in the dictionary or to find out its definition. It was suggested to them that if they thought that one of the options offered by the checker was the correct one, but were not certain, they could "lookup" the word to confirm that it was the required word (or otherwise). In addition, TE and FR were automatically given a one-line definition or example with each word offered by the checker. In group 2, DI and GR were told the same as DR and DV in group 1, and were not provided with definitions automatically. MA and ST were told that they could 'lookup' a word to see if it was in the dictionary, but that definitions were not provided for all words. Everything that appeared on each child`s v.d.u. screen was recorded for later analysis. The investigator was present in each session and also observed the children and made notes on these observations. Printed copies of each corrected piece of writing were produced by the investigator: if errors still remained in these the child was prompted to check and correct them (either by using the program or by self-correction). @subsection(Results) @label(pilot2-results) For each child, from his writing in his exercise book and from the recordings of the v.d.u screen @foot(These will be referred to as dribble files.) each session, words that had been altered and those that were incorrectly spelt were noted. Additionally, a small number of words that appeared in the dribble files (having been used with check or lookup) but had not been written were noted. Each of these altered and checked words was recorded according to whether: @begin(enumerate) It was initially @begin(enumerate) correct - left as correct; correct - self-altered to an error; correct - self-altered to another correctly spelt word; incorrect - left as an error; incorrect - self-altered to another correctly spelt word; incorrect - self-altered to a different error; @end(enumerate) It was @begin(enumerate) self-corrected by the child checked using the program @end(enumerate) @end(enumerate) If the word was checked (2.b) then it was noted whether: @begin(enumerate) The correct word was offered, noting its position in the four options. The correct word was not offered. @end(enumerate) If the word was offered, whether the word originally written was: @begin(enumerate) left correct; altered to be correct; left as an error; altered to an error; or some other course of action was taken. @end(enumerate) If the correct word was not offered the course of action that the child took was noted,including: @begin(itemize) asking the investigator to add it to the dictionary and checking it again; rechecking it with a different spelling; leaving the word as it was; changing the word to one of those offered; using the lookup facility. @end(itemize) @comment[An example of one child's dribble file from a session and his writing from that session is given in appendix (XX). Comments are included to indicate how each of the words were recorded.] @comment[give part of session and photocopy of writing of a child - FR?] Specific questions that were asked are now considered. @begin(verbatim) @b(Checking the spelling) @end(verbatim) If the child is unsure about the spelling of a word, does he @begin(enumerate) use the computer spelling program to check it? guess the spelling? @end(enumerate) The number of cases in which the child used the checking program to check a word were counted. These included a number of words that were correctly spelt, but that the child checked using the spelling checking program. The words in this category, therefore, could have been words that were correct or incorrect initially, and that were corrected, left as errors or changed, after checking. The number of errors left unchecked or unchanged, plus the number of self-corrected or changed words, were classified as guessing the spelling. There were a small number of words that were left incorrect because the child had no time left to read over and check them, by himself or with the program: these are also included in this category. In total, 395 words were checked using the program including 99 that were intially correct. 173 words were left as errors, including 25 errors that the child had no time to check. Figures for individual children are given in figure @ref(no-checked). @begin(figure) @begin(verbatim) Number of words Number of words checked using left as errors the program or self-altered Total (initially Total (no time left correct) to check) Group 1 TE 88 (18) 35 (11) DV 53 (8) 13 (0) FR 91 (18) 32 (10) DR 52 (11) 18 (0) Group 2 DI 18 (9) 19 (0) ST 29 (14) 12 (0) MA 37 (17) 9 (4) GR 27 (8) 35 (0) Totals 395 (99) 173 (25) @end(verbatim) @caption(Words checked using the correction program) @tag(no-checked) @end(figure) @comment[ Helen - it would be interesting to know what proportion of total words were checked using the program] Considering these results it can be seen that the spelling checker was used in a large number of cases to check spellings. It is also of interest to note that of those 296 incorrectly spelt words (395 total less 99 initially correct) checked using the computer program, 216 were corrected; of the 148 errors self-checked (173 total less 25 unchecked) only 55 were corrected. @begin(verbatim) @b(Selecting the required word) @end(verbatim) Could the child select the required word from the four offered by the checking program in cases where the word was present? The following were counted: @begin(enumerate) the number of times that the child selected the correct word i.e. left the spelling unchanged if it was correct, or corrected it if it was not; the number of times that he did not select the correct word i.e. left an error uncorrected or altered the word to be a different word or an error. @end(enumerate) The former category will be taken as evidence that the child could select the correct word and the latter as that the child could not. Some words did not fall clearly into either category e.g. if the lookup facility was used after checking a word it may be that the child thinks he recognizes the word as correct (or not, if he is "looking up" the wrong word!) but is still not certain. These words are not counted here. Individual figures are given in figure @ref(no-selected). @foot[The total number of words checked and offered in figure @ref(no-selected) plus those checked and not offered in figure @ref(no-realised) does not equal those checked in figure @ref(no-checked) because if a word is not offered, but is then added and re-checked (and is then offered) if may occur in both figure @ref(no-selected) and figure @ref(no-realised).] The figures for TE and FR are not included in the totals in this figure. This is because they were automatically given the "lookup" information. @begin(figure) @begin(verbatim) Total Number of Number of Percentage offered correct words correct words selected correctly selected not selected when offered Group 1 DV 44 40 4 90.9% DR 37 32 5 86.5% Group 2 DI 10 7 3 70% ST 12 11 1 91.7% MA 24 21 3 87.5% GR 19 17 2 89.5% Totals: 146 128 18 87.7% ( TE 62 53 9 85.5%) ( FR 78 73 5 93.6%) @end(verbatim) @caption(Words selected when offered by the correction program) @tag(no-selected) @end(figure) 146 words were checked and counted. For 128 of them there was evidence that the correct word had been selected when offered. In only 18 cases was the correct word not selected. So, in 87.7% of cases the correct word was selected when offered. @begin(verbatim) @b(Words not offered by the checking program) @end(verbatim) If the required word was not offered when checked did the child realise that it was not presented in the four options? The following were counted: the number of times that the child re-checked a word, or left an error to stand, or left a correct spelling unchanged, or asked for help, was taken to indicate that the child did realise that the word was not presented. The number of times that the child selected one of the presented words, and altered the original word, was taken as indicating that the child did not recognize that the required word had not been presented. The results show that of 115 cases in which the required word was not offered, in only 13 (c.11%) of these did the child not realise that the word had not been presented. Therefore, in 89% of cases the child did recognize that the required word had not been offered. Individual figures are given in figure @ref(no-realised). @begin(figure) @begin(verbatim) Total Child Child does Percentage child realises not realise does not realise word not word not word not presented presented presented Group 1 TE 26 20 6 23.1% DV 11 9 2 18.2% FR 28 28 0 0% DR 13 10 3 23.1% Group 2 DI 5 5 0 0% ST 11 11 0 0% MA 11 10 1 9.1% GR 10 9 1 10% Totals: 115 102 13 11.3% @end(verbatim) @caption(Child realises the correct word is not presented) @tag(no-realised) @end(figure) @begin(verbatim) @b(Other ways the program was used) @end(verbatim) In general the program was used as intended. However, there were some exceptions to this. DI on a number of occasions used the 'lookup' facility to check the spelling of a word: if he spelt the word correctly and it was in the dictionary, the word and definition were printed. The lookup facility was faster than checking the spelling of the word, and he took advantage of this. For cases where the desired word was not in the dictionary, ST added it himself on several occasions. In some cases he spelt it correctly. In others he did not. When it was spelt incorrectly, he would them check his spelling, as before, and his incorrect spelling would be confirmed. The children were not encouraged to use the addword facility. @subsection(Discussion) The evidence from this study shows that the children were able, and willing, to use the spelling checking facility provided. They checked the spelling of a substantial proportion of the words they were unsure about although in some cases the lookup facility was used for effectively the same purpose. One of the main problems was the slowness of the program. In 87.7% of cases where the correct word was offered by the editcost program, it was selected by the child as the intended word. There were a number of occasions, however, when the correction was not offered. The program's performance is assessed in chapter @ref(perform). In cases where the intended word was not offered by the correction program, the child did realise that it was not present: in some cases where the children were unsure they re-checked the word, or looked up the definition of one of those offered. In only 11% of cases where the word was not offered did the child not appear to realise this. The children in group 1 used the program over a greater number of sessions, and therefore checked more words. Individual children varied in their ability to select the correct word, or to detect when the correct word was not offered. The observation that the children would be able to recognize the correct spelling, even if they cannot produce it, was made. This observation needs to be confirmed by further, more rigorous, testing. @newpage @section(Summary of results from the two studies) @label(ass-summary) The results from each of the two studies will be summarised briefly. @begin(itemize) Children with learning difficulties in spelling are both able and willing to use a simple text editor to write stories. The text editor could be usefully extended to incorporate an error checking and detecting facility. If a spelling correction program is provided, it is used to check the spellings that the child is uncertain of. If the required word is offered by the program the child is generally able to recognize it. If it is not offered, he usually realises it is not there. @end(itemize) The assumptions made about the use of the program, therefore, were generally supported by these two studies. In addition, evidence was presented suggesting that, whilst the children might not be able to produce the correct spelling of a word, they were able to recognize it.