Do grammar checkers work?
A report on research into the effectiveness of Grammatik V
based on samples of authentic essays by EFL students
Paper presented at EUROCALL 96 Dániel Berszenyi College, Hungary
Yu Hong Wei and Graham Davies

Links checked 12 October 2011

This paper was originally published in Kohn J., Rüschoff B. & Wolff D. 1997 (eds.) New horizons in CALL: proceedings of EUROCALL 96, Dániel Berzsenyi College, Szombathely, Hungary, a copy of which can be obtained from Grammar checkers have continued to improve since this article was written, but they are still far from perfect. See:

Biographical information

Yu Hong Wei was registered at the time of writing this article as a research student at Thames Valley University. where she was working on a PhD thesis under the supervision of Graham Davies and Peter Skehan. Graham Davies is Visiting Professor at Thames Valley University. Peter Skehan is now working at King's College London. At the time of writing Graham Davies was President of EUROCALL. He was recently Chair of the WorldCALL 2003 Steering Committee and is Head of the Working Party that aims to set up WorldCALL as an official organisation:


This paper seeks to assess the effectiveness of a popular grammar and style checker, Grammatik V, based on empirical evidence. The paper describes the main features of Grammatik V and summarises the results of trialling the program with students of English as a Foreign Language. Data was collected by means of questionnaires, tape recordings and interviews, and by submitting authentic, unmarked students' essays to analysis by the program. A copy of each piece of writing that was 'marked' by the program was compared with the original piece of work and re-examined and compared by the researcher and a lecturer. The major part of this paper summarises the data collected, with reference to specific examples that illustrate both the drawbacks and advantages of Grammatik V. It is hoped that the insights gained and their implications will be useful to those who are interested in developing CALLware and to those who are interested in integrating CALL into their own teaching.

1. Grammatik V: description of the facilities offered by the program

Grammar and style checkers are just one category of a range of tools that are available for people who use computers for writing. The range includes software designed to train writers, software to support professional writers, and a variety of pre- and post-writing and composing tools. Grammatik V is a well-established post-writing tool which is designed to help the user to proof-read writing. The range of potential problems that Grammatik V can detect falls into three 'rule class groups':

i. Mechanics: e.g. capitalisation, punctuation and spelling.

ii. Grammar: e.g. parts of speech and subject/verb agreement.

iii. Style: e.g. words and expressions which 'set a tone' for the ten types of writing pre-set in the program: general, a business letter, a report, fiction or a technical document and so on.

Messages about the nature and types of problems are flagged and presented under these three 'rule class groups', which are further divided into more than fifty different 'rule classes'. A general summary of the readability of the text both in words and in statistical format are given at the end of proof-reading, and a comparison chart can be called up to enable the writer to compare his/her writing with one of these three types of writing: a novel, a speech and a life insurance policy.

Grammatik V can carry out proof-reading in one of the following ways:

i. Interactive: the program opens a document and takes the user through it, pausing at 'errors'. In this way it interacts with the user, referring to the text being checked. Problems are highlighted in the original text and the user can see comments on the nature of these problems and suggestions for change. Online editing is readily available, so immediate adoption of feedback from the program into the working document is possible. The user decides whether to accept or ignore any comments or suggestions the program makes. The user can also call up references when needed.

ii. Mark only: the program proof-reads a text without stopping and will mark all the 'suspicious' points. It then provides comments and suggestions for change in a separate document.

iii. Read only: the program presents the user with an interactive screen in which problematic words, expressions and sentences are highlighted in their original form. The user can read the original text and the comments and suggestions given but cannot change the text on the screen.

iv. Grammar and mechanics checks only: the program ignores style problems but checks only grammatical problems and mechanical problems such as punctuation and capitalisation.

v. Spelling only: a straightforward spelling check.

In all of these checking modes a general summary of the readability of the text and a statistical score are available. Compared to other similar programs, Grammatik V offers a wider range of functions and facilities, especially flexible user control.

2. How the study was conducted

This study is part of an investigation of the potential role that the computer can play in an integrated approach to help non-native students to acquire and develop ability and strategies in writing for academic purposes. Data was therefore collected to establish the usefulness of Grammatik V in an ELT context, namely whether it can help non-native students to solve their problems and to what extent it can help to improve the overall quality of their writing.

17 students, from Thames Valley University and Richmond College, the American International University in London, were selected by means of distributed questionnaires. The students came from six different subject disciplines. The students then trialled the program and proof-read finished copies of their own writing using the predefined style for general writing. The writing samples consisted mainly of students' written assignments for their subjects of study or essays written as part of an EAP training course in writing.

During the trials, students' queries and comments on the program were tape-recorded. The interaction between the students and the researcher was also recorded in order to provide detailed information on any problems the students might have had during the trials. To clarify the students' comments and questions during the trials and to ascertain how they evaluated Grammatik V, post-trial interviews were conducted. A 'marked' copy of each student's writing sample, with all the points noted and considered inappropriate by the program, was printed. Copies of two versions of each student's writing sample, one prior to using the program and one with the changes made during the trial, were collected in order to provide information on the effects of using the program. The following points were focused on during the trial sessions. They account mainly for students' needs and problems identified in the main part of this research.

i. Whether the program provided useful comments and advice about the appropriateness of the language and style used in the writing samples.

ii. Whether the feedback given by the program on grammatical, syntactic and linguistic features of the writing samples was accurate and informative enough for the user to determine what should be improved.

iii. Whether the language used in the interaction between the program and the user was clear, easy to understand and follow.

iv. Whether the program would be more useful in proof-reading one type of writing than another, e.g. writing at different proficiency levels or writing on different subjects;

v. Whether the use of the program led to an overall improved quality of writing.

In the trials the interactive proof-reading mode was used. That is to say that the program took the user through the process of proof-reading, and highlighted and commented on problematic areas on an 'interactive screen'.

3. Findings and summary of the study

3.1. Trials of Grammatik V

The seventeen essays and assignments that were analysed by the program comprised a total of 17213 words - an average of 1012 words per essay/assignment. The program checked the writing samples and yielded a large number of messages. 1394 words and phrases were commented on - about one message per 12.4 words. Descriptive data for 'errors' and 'inappropriate usage' that the program detected and marked are shown in the chart below (Figure 1). To demonstrate the actual performance of the program, we arranged the messages given by the program as: i. Useful, ii. Misdetected, iii. Questionable, according to the participants' choices and opinions.

Figure 1: Overall performance of Grammatik V

These figures show that the program made a detailed check of the writing assignments and yielded a considerable number of messages. But the next issue to consider is the quality of feedback given in these messages.

.i. Mechanical problems and spelling

Under this heading we have two types of problems: one is concerned with mechanical problems, e.g. capitalisation, punctuation, infinitive forms, number style etc., and the other with spelling. In the analysis below, spelling errors have been separated from the others in order to highlight specific issues and to avoid misleading conclusions.

a. Mechanical problems

The program detected and commented on 142 mechanical errors, out of which 92 were correctly identified and 50 were misdetected. The percentage of accuracy is shown in the following table (Figure 2):


Mechanical problems Correctly detected Misdetected
Capitalisation 26% 74%
Number style 59% 41%
Quotation marks 50% 50%
Split words 100% 0%
Punctuation 81% 19%


.Figure 2: Mechanical problems

b. Spelling errors

Generally, the program proved to be sensitive to spelling and punctuation errors. The program spotted 376 spelling 'errors'. Among them, however, there were a large number of 'false' errors, which included 244 correctly spelt proper nouns and words considered as Americanisms. Words that were not in the program's dictionary were also treated as 'errors', for instance: anti-social, counter-culture, initiate, solvents, typology, decontextualise. The following table (Figure 3) shows the breakdown.

1 = total 2 = proper nouns

3 = false errors 4 = misdetected 5 = errors

Figure 3: Breakdown of spelling errors

Although it may seem unfair or even unreasonable to expect a program to be so comprehensive as to have any word in its dictionary it is, however, reasonable and perhaps possible to expect the spell checker in such a highly-rated and widely-used program to be able to recognise commonly used proper nouns, for instance: electorate, gown, lampstand, non-swimmer, etc.

It is also necessary to point out that such a limitation could have been overcome by teaching the program new words, i.e. if the user had used function key F7. Such a facility was available in the program in interactive checking mode, but during the trials none of the users employed this facility.

ii. Grammatical errors

The program was fairly sensitive in detecting mismatches between subject and verb agreement in simple sentences. Unfortunately, it proved much less successful in identifying and relating subjects with verbs when they had modifiers in between or when they appeared in complex sentences. For instance, sentences with the structure of subject followed immediately by verb were generally recognised. But the program would fail to identify subject and verb not immediately adjacent to each other but with modifiers in between; and it would sometimes fail to identify parts of speech, e.g. perceiving nouns, adjectives or adverbs as verbs, or vice versa. This may have been caused by the limitation of the program's parsing ability, nevertheless, some of the commonly used sentence structures and collocations were neither recognised nor accepted. The following sentences are just a few examples of the program's failure to parse the source text correctly. Grammatik V's comments are in square brackets.

a. S V O agreement:

"... the term itself suggests..."

[The object pronoun 'itself' can't be used as a subject.]

"The two key factors in his 'Confluent Education Theory' were awareness and responsibility."

[(a) After 'two' you need a plural noun, not the singular noun 'key'. (b) The singular subject 'Theory' takes a singular verb, not the plural verb 'were.]

b. Non-standard S V O pattern

"I just remember how excited I was ..."

[ The subject pronoun 'I' shouldn't be in the object position.]

"It is hardly surprising that to find a certain reluctance ..."

[This verb requires an object.]

Interrogative sentences were sometimes not recognised and therefore were considered sentences with lack of agreement between subject and verb.

"How much cholesterol do I need?"

[The singular subject 'cholesterol' takes a singular verb, not the plural verb 'do'.]

c. Parts of speech

"She still is ill but lives proudly ..."

[An article or other modifier usually proceeds the word 'ill'.]

"People become infected by the virus everyday."

[Normally an adjective like 'everyday' doesn't modify a verb.]

Assuming that a noun should usually be proceeded by an article "the" or "a", any words appearing immediately after articles were perceived as incorrect or incomplete forms of nouns or vice versa.

"It was a very, very beautiful painting."

[This noun phrase appears incomplete.]

"an even more far-reaching development ..."

[The adjective 'even' is not usually used with the singular modifier 'an'.]

d. Collocations

"This animal could be extinct in a short period of time."

[Since 'period' and 'time' overlap in meaning, you may want to choose one or the other.]

"I was looking forward to seeing his self portrait."

[An article or other modifier usually proceeds the word 'looking'.]

iii. Stylistic errors

Problems at this level are of a different nature from what has been discussed so far. Stylistic problems, as perceived by Grammatik V, have more to do with semantics and register and, for example, the use of the passive voice and length of sentences.

a. Register

Grammatik V made many recommendations to replace words and expressions which were considered wordy, jargon, cliché and overstated. The following table (Figure 4) contains words and phrases (in italics) that the program objected to, with recommended replacements in square brackets.


Wordy Jargon Cliché Overstated

one of the

[one, a]

in the absence of

[without, lacking]


[extensive or broad]



on the part of

[by. of, for]


[response, react, opinion]

in the context of




in such a way

[so, so that, as]


[do, act]

in fact

[Try to avoid this phrase unless there is actually some real question involved.]


[aid, help]


.Figure 4: Words to be avoided

Although simplicity of language is a virtue when a simple expression can convey the same meaning as a more complex expression, this does not always work. For instance, if the words or phrases in the above examples had been replaced by what was suggested, it would be debatable whether the replacements would convey the same subtlety, emphasis or caution the writers had intended, or even what the writer actually meant.

b. Formalism

This category consists mainly of the use of certain adverbs and conjunctions. There is good advice, such as avoiding adverbs like indeed, rather and actually in formal writing and not to begin sentences with the conjunctions but and and. There are some other adverbs which are considered as 'vague' by the program: for instance, certainly, indeed, reasonable, relatively. It is suggested that words and expressions of this kind are to be 'used sparingly' or to be 'omitted or replaced by a specific modifier'. This proves to be prescriptive. The use of the relative pronoun which was constantly queried. The program picked up every which in all 17 pieces of writing and suggested that it should be replaced by which ones or that, for instance:

"Ealing is like many other London boroughs which have a significant minority ethnic population."

"These words which were chosen to be introduced are essential for the general comprehension of the text."

The advice given was:

[Unless you could substitute 'which one' ('s') for 'which' here, you should probably use 'that' instead. When 'which' begins a clause, it is usually preceded by a comma. Use help key for more information.]

When consulting help, the user could get such information under the rule class of relative pronouns:

[Many people use the relative pronouns 'that' and 'which' incorrectly to begin clauses:

i. Use 'which' to begin clauses that are not essential to the meaning of a sentence.

ii. Use 'that' to begin clauses that are essential to the meaning of the sentence. These clauses aren't set off from the rest of the sentence by a comma.']

Such advice may be correct in some instances and incorrect in others; but which cannot be replaced by which one(s) in either of the above sentences.

c. Commonly confused words

Comments and advice given here mainly concerned homonyms and words which had similar meanings, e.g.

principle [principal]

['Principle' is the main item, a school director, or an amount of money. 'Principle' is a basic law or doctrine.]

accept [except]

['Accept' means to receive: Please accept my offer.' 'Except' as a verb means to leave out: 'We'll except the last provision of the contract.']

affect [effect]

['Affect' as a verb means 'influence'. You may want the verb 'effect', meaning 'bring about' or 'cause'.]


['Affective' means 'caused by or causing an emotional response'. You may want 'effective', meaning 'adequate' or 'operative'.]

Some of them, for instance principle and principal, may be useful reminders, but others such as accept and except, affect and effect are probably unnecessary.

d. Long Sentences

Long sentences were usually commented on:

['Long sentences can be difficult to read or understand.]

and it was suggested that they should be divided into separate sentences:

['Consider revising so that no more than one complete thought is expressed in each sentence.']

It should be pointed out that the length of a 'long sentence' could have been 'set' by the user: that is to say the flexibility of the program allows the user to decide how long a sentence will be considered 'long'. This can be done by setting the number of words in the rule class.

3.2. Two Case Studies

Taking into consideration that the program was designed with native speakers of English as the target users, it was assumed that the program might perform differently when used with EFL users. It was assumed that there would be different reactions and responses from EFL students to the feedback given by the program than from native speaker students, and vice versa. We examined in detail two extracts from samples of students' writing which had been proof-read and 'marked' by the program: one by an EFL student on an EAP writing course and the other by a native English speaker writing about his subject of study on a MA course. Extracts from Grammatik V's analysis of the two writing samples follow.

Column 1 contains extracts from each student's writing sample.

Column 2 contains comments and advice given by Grammatik V.

Column 3 contains the researcher's comments on the interaction between the program and the user, i.e. the validity of the advice.

Case Study 1: Essay written by a Spanish MT student on Effects of television on children and young people.


Context Grammatik V's advice Validity of advice
Par example, some children who have never seen violence on TV before and then start watching violente programes, they will as a result at the first moment ask themselves why of such violence.



i. Correct but overkill. Missed par: Spanish MT interference?

ii. The main problem is that the sentence is badly constructed, not long.

... they will as a result at the first moment ask themselves why of such violence. iii. THE SINGULAR SUBJECT 'RESULT' TAKES A SINGULAR VERB, NOT THE PLURAL VERB 'ASK'. iii. Wrong: sentence incorrectly parsed: result is not the subject of ask.
... they will as a result at the first moment ask themselves why of such violence. iv. REFLEXIVE PRONOUNS SHOULD AGREE IN NUMBER AND PERSON WITH THE SUBJECT. DID YOU MEAN TO USE A SIMPLE OBJECT PRONOUN? iv. Wrong and confusing: they will ask themselves is OK.
... they will as a result at the first moment ask themselves why of such violence. v. CONSIDER REVISING WITH 'SO' AND THE ADJECTIVE FORM OF THE WORD AFTER 'SUCH' REPLACE "OF SUCH URGENCY" WITH "SO URGENT" v. Wrong: such violence is OK.
... violente programes ... vi. SPELLING ERROR. vi. Good advice: violente and programes correctly spotted.


.Case Study 2: Essay written by an English MT student on The humanistic approach to ELT.


Context Grammatik V's advice Validity of advice
A main point of focus will be Krashen's 'Natural Approach', as this can involve a broad range of 'humanistic' techniques, and I shall be discussing the pros and cons of endeavouring to utilize such an approach in my own classroom situation. i. THERE APPEARS TO BE AN EXTRA NOUN PHRASE IN THIS SEQUENCE. CHECK FOR MISSING WORDS OR PUNCTUATION. i. Wrong: saw error in cluster main point of focus.
A main point of focus will be Krashen's 'Natural Approach', ... ii. SPELLING ERROR. ii. Perceived proper noun Krashen as spelling mistake.

... as this can involve a broad range of 'humanistic' techniques, and I shall be discussing the pros and cons of endeavouring to utilize such an approach in my own classroom situation.

iii. THERE APPEARS TO BE AN EXTRA NOUN PHRASE IN THIS SEQUENCE. CHECK FOR MISSING WORDS OR PUNCTUATION. iii. Wrong: saw error in cluster broad range of 'humanistic' techniques.
... and I shall be discussing the pros and cons of endeavouring to utilize such an approach in my own classroom situation.

iv. SIMPLIFY with do, use

iv. The meanings of do and use differ from utilize.
Moskowitz (1978) expressed the essence of the 'humanistic quest': "throughout the ages man has been striving to become more human". v. SPELLING ERROR. v. Perceived proper noun Moskowitz as spelling error.
Moskowitz (1978) expressed the essence of the 'humanistic quest': "throughout the ages man has been striving to become more human".



vi. Wrong: saw error in cluster essence of the 'humanistic' quest.

vii. Wrong: did not understand humanistic.

... "throughout the ages man has been striving to become more human". viii. THERE MAY BE A POSSESSIVE ERROR IN THE WORD 'AGES'. viii. Wrong: sentence incorrectly parsed.


.Analysis of the Case Studies

In order to ascertain whether the use of the program resulted in any improvement in the quality of the two pieces of writing, possible changes consequent upon the advice and suggestions given by the program were examined. It was considered that some of the messages given by the program would not result in immediate changes in the text, e.g. reminders about commonly confused words, passive verbs, sentence variety and sentence length.

In Case Study 1 the program generated 54 messages, out of which 33 resulted in the student making changes of one kind or another. Spelling and punctuation errors were at the top of the list, and further examination revealed that most of them were real errors and that changes following the advice resulted in improved accuracy. So the advice was 61% effective.

In Case Study 2 there were 38 messages, which resulted in only one change: a word the student had accidentally typed twice, which he then deleted. So the advice was 0.03% effective. Spelling was also top of the list in Case Study 2, but most of the 'errors' were false as they were either proper nouns or words not found in the program's dictionary. Second on the list was advice concerning style problems, e.g. usage and meanings of words and expressions, and sentence length. The user was asked to simplify words or expressions and cut down sentence length. None of the advice was accepted by the user since he considered what was suggested was inadequate in conveying what he had intended. Consequently no changes of this kind were made.

Comparing the two case studies, it is clear that the program was more specific and efficient in detecting problems at the lower, mechanical levels. It was much less effective in dealing with problems at a more sophisticated level, for instance, dealing with semantic issues. The result of these two case studies hence gave us a good idea of how the program dealt with problems in texts written by students from different language backgrounds.

4. Evaluation of Grammatik V by participants in trials

To ascertain whether the program's feedback was helpful to the users and thus effective in achieving its goals, post-trial interviews were conducted to see how the participants evaluated the program from their own experience of trialling it. Each of the participants commented on the effectiveness of the program, which included accuracy of the feedback, clarity and effectiveness of advice and comments given; in other words, whether the advice and comments were considered clear and advisable. The writers of the two extracts in the case studies gave completely contradictory answers, the first commenting 'very useful' - 'The program is good. It told me almost all the mistakes I made.' - and the latter 'useless'. Most of the 17 participants did not give the program a high rating, but some were kinder:

'The program is useful in detecting spelling or punctuation errors, but not very helpful in general.'

'The program could be useful if you know what you want to say in the writing.'

'It is good in drawing your attention to some of the things like voice of verbs or sentence length or similar words.'

'It can work if used with caution when the user is competent in using the language; otherwise it can be misleading.'

The students' comments indicated that the program dealt with lower-level problems satisfactorily. At advanced level or above, e.g. in dealing with lexical, semantic and stylistic problems, it did not provide the users with what they had expected: for example, feedback often did not pin down the precise problem or the nature of the problem, advice or comments given were not clear and accurate and they were not very helpful to the users.

5. Conclusions and implications

There are three kinds of useful insights we can draw from the trials of the program.

Firstly, the design of the program itself. The effectiveness of using the current program to check students' writing for academic purposes, on the whole, is not very satisfactory, although the different ratios of correctly and misdetected and questionable problems did suggest that there were variations of performance in the three areas the program focused on: mechanics, grammar and style. However, it is also fair for us to say that the program has its own strengths and weaknesses. Its strengths can be seen in its user-friendliness, i.e. attractiveness of the interface, evidence of focused problems, ease of use of online help and references, user control (i.e. flexibility of using different checking modes, styles and formality) and availability and easiness of online editing, as well as its sensitiveness to potential errors. Nevertheless, the effectiveness of the program in dealing with complex grammatical and stylistic problems was weak; and for the students we have in mind, this was affected by the lack of an inbuilt style in the program, with customised rules to check the kind of errors commonly occurring in their writing. Therefore, if used with non-native students for the kind of writing they do, there is a need for re-examination and improvement of program design, and this should be preceded by a careful analysis of what the students need, what proof-reading programs should aim to do and what is feasible within the current limitations of the technology and linguistic science.

Secondly, an important insight we draw from the trials is that there are issues other than program design itself which affected the effectiveness of the program. This concerned the way the program was used. There were certain things that emerged during the trials which showed that the program could have been more efficient if certain facilities had been used or used more appropriately, e.g. the various command keys. For instance, the user could tell the program whether to activate or inactivate certain rules or simply ignore them, e.g. passive verbs and the length of 'long sentences'; or the user could tell the program to 'learn word' and add new vocabulary to its dictionary. For example, function key F6 can be used to turn 'on' or 'off' any rule class in all the styles preset in the program if they are considered unsuitable for or irrelevant to the kind of writing being proof-read, e.g. passive verbs or long sentences. F7 ('learn word') enables the user to tell the program to remember certain new words or add new spellings to the program's dictionary, so that specific words or specialised vocabulary will be recognised and accepted by the program. In both cases, if such function keys had been used properly the number of questionable comments relating to stylistic problems and to spelling 'errors' could have dropped dramatically. It could also have saved the user the trouble of having to skip them. During the trials, however, most of the participants did not use the function keys at all.

This leads to the third issue which has been raised in the course of the research: the part students should play in using a computer program. It demonstrates that their different assumptions and actual role-playing affects the result of using a program and, in turn, affects their understanding of software. For example, when invited to comment on the program and to suggest useful and desirable features of writing tools or aids, most of the trial participants expressed a certain dissatisfaction and also the hope that the software could be specific in providing comments, e.g. 'uses' of certain lexical items and functional words, 'meanings of words and phrases about which they were not sure' or 'telling them exactly what was wrong'. Some of them hoped that the program could provide suggestions, telling them 'what to use' when they were short of words and when unsure of 'what to say'. It shows that the students did have problems and they hoped they could get help for them. It also reveals students' perceptions and expectations of the computer, i.e. as a machine which not only delivers information about words and grammar when required, but which also provides information for formulating ideas that they want to put across but do not know how to.

The implications we can draw are that, in evaluating the effectiveness of a program, we need to examine not only the quality of the program design and to re-examine the possibility of getting what is desirable and what is feasible, but also to consider any variables which may affect and determine the result we may get. In this study, we need to consider the learner variables involved, namely the role of learners themselves and how their role-playing - led by the program or their behaviour when using it - has affected the effective use of the program.

The recommendations we make are that we need to consider and analyse students' specific needs, to find and offer appropriate technical tools and aids and then integrate them into a writing course supervised by a writing teacher. The teacher should make sure that the pedagogical applicability of the software is appropriate and also ensure that the students know the functions of the software and how it can be used efficiently.

Considering the fact that Grammatik V may not provide sufficiently relevant information and references for the specific problems L2 students have, it will be not only beneficial but necessary to make use of other relevant programs and to incorporate certain features into the present program to accommodate the target students' needs, instead of using one particular program for a particular task. For instance, Grammatik V may be used as an aid for awareness-raising, or the program could be adapted and used with a particular group to help solve their specific problems. Some of the advice given on style, word usage and also sentence length was constructive in raising the user's awareness of the formality of style that was assumed and the language that was actually used. Commenting on long or passive sentences was often a useful reminder to the writer to reconsider or re-arrange them for the sake of brevity and clarity. In addition, programs such as outliners, thesauruses, concordancers and composing tools can used in conjunction with Grammatik V to enable students to get help throughout the whole writing process.

Integration of software into a writing course is a complex process, and there will be great demands on the teacher, both pedagogically and organisationally. The crucial issues, above and beyond the quality of the software itself, are the pedagogical considerations in the choice and adoption of software and an appropriate understanding of the functions of the software and sufficient knowledge of its application from both the teacher's and the students' points of view.

© Yu Hong Wei & Graham Davies 1997. This work is licensed under a
Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.

Creative Commons Licence