.
Grammar-checking software sometimes suggests correct grammar is wrong, especially passive verbs and acceptable uses of normally-wrong structures
THE FAILINGS OF COMPUTERISED GRAMMAR CHECKING
Most people who write with a computer word processor are likely to have used a grammar-checking facility to find and correct their grammar mistakes. This is generally a useful thing to do because awareness of the mistakes we make can be the first step towards eradicating them (see 202. Some Strategies for Successful English Learning).
However, grammar-checkers are not as reliable as spellchecking software. They often miss grammar mistakes altogether (see 138. Test your Command of Grammar 1), or suggest good grammar is bad. These weaknesses can have some fairly serious consequences for writers. A program’s failure to highlight an erroneous structure can reinforce the writer’s belief that it is possible or correct. Unjustified highlighting of correct structures leads many writers who cannot see why their grammar has been questioned to accept the computer’s incorrect alternative, quite reasonably trusting that their own poor understanding of grammar is the cause of their confusion. Over time, this can create uncertainty where previously there was none, thus undermining confidence.
My belief is that the way to enjoy the benefits of computerised grammar checking whilst minimising its weaknesses is through developing a more critical (but not totally negative) attitude to the grammatical suggestions, and that this can be achieved by means of extensive practice in analysing questioned wording and suggested alterations. In this post, I wish to analyse a number of examples of computer software wrongly questioning English grammar usage, and to seek some general features in them that might explain the computer struggles and also help writers to recognise similar examples as deserving of scepticism. For further examples, see 68. How Computers Get Grammar Wrong1 and 275. How Computers Get Grammar Wrong 3.
.
COMPUTER CRITICISMS OF PASSIVE VERBS
Sometimes sentences are underlined because they are quite long and have their main verb in the passive voice. The advice is usually to make the verb active. The “problem” that the programmers see here is not so much one of grammar as of style: the criticised passive verbs usually break no grammar rules but are perceived instead to be over-complicated, “clumsy” or “unnatural” (see 100. What is a Grammar Error? and 142. Grammar Errors with Passive Verbs).
This advice is questionable because it is based on opinion rather than scientific fact. I have argued elsewhere (see 27. How to Avoid Passive Verbs) that the very existence of the passive voice in English is taken by linguistics experts to be evidence of its value. Much work has been done to establish exactly what this value is (see, for example, my own article within these pages entitled Active-Passive Paraphrases in English and What They Mean for Teaching). For a specific example of a desirable passive, see 265. Grammar Tools for Better Writing, #5.
It is true that the active voice is more common than the passive in English, but the mere fact that a computer word processor has underlined a passive voice verb does not necessarily mean that the verb should be replaced by an active.
.
STRUCTURES THAT ARE USUALLY WRONG BUT SOMETIMES RIGHT
Computers seem to have been programmed with a list of English grammar combinations that they must highlight as wrong. The problem is that a surprisingly large number of usually-wrong grammar combinations can become correct in particular circumstances (see 124. Structures with a Double Meaning 1), a possibility that the programmers seem not to have given enough recognition to.
An example of a usually-wrong combination is the attachment of BE…-ing (normally a marker of the present continuous tense in English) to a verb like KNOW, SEEM or OWN, which cannot usually be in this tense. Here is a case where this combination is definitely wrong:
(a) *Even the youngest of the children is knowing complicated algebra.
The underlined verb should, of course, have the present simple tense form knows, not the present continuous.
Yet combinations like is knowing are of a kind that can become correct in the right circumstances:
(b) The key to sounding formal in writing is knowing which words to avoid.
This is a perfectly acceptable sentence, with the message “the key = knowing which…”. In grammatical terms, is is not an auxiliary verb combining with a participle to make the present continuous tense of KNOW, but the verb BE by itself (meaning “equals”) combining with a noun-like use of -ing (i.e. a “gerund” rather than a “participle” – see 71. Gerund and Participle Uses of “-ing”), making it into a “complement” (see 220. Features of Complements, #3). Most verbs are usable in this alternative way after BE. And knowing in particular can also be an adjective rather than participle – another possibility after BE (see 254. Tricky Word Contrasts 10, #5)
My word processor, however, does not recognise such alternatives, instead highlighting is knowing and suggesting that I “correct” it to knows. It seems to have been programmed to make no exceptions in singling out instances of BE -ing around verbs like KNOW, ignoring the alternative ways of using BE, despite the fact that small words with alternative uses are common in English (see 3. Multi-Use Words).
Various other English grammar structures are likewise questioned by my computer when in fact they are right but are following a rather unusual grammar rule. Examples are:
(c) Not all active verbs with no following noun are a problem.
(d) Both words can act as conjunctions, which means having a following subject + verb.
(e) Also useful to know is the fact that share prices can fall.
My computer’s suggested correction of (c) was to change no into a, on the grounds that only one negative (not) is necessary. The common error that the computer thought was present was the use of two negative words to express the meaning of just one, like this:
(f) *The refugees did not have no money (= The refugees had no money).
Such sentences are correct in some non-standard English dialects but rarely acceptable in writing. The indicators of this non-standard usage are not and no used together. The computer did not seem to “know” that they occasionally go correctly together in formal writing, cancelling each other out in sentences like (c) (see 9. Double Negatives). The way a human writer knows when not … no is correct is by thinking about the resultant meaning – something that computers seem very weak at.
In sentence (d), the computer wanted the singular-showing -s of the verb means to be removed, on the grounds that its subject which had been given plural meaning by the noun immediately before it (conjunctions). The reality is that which does not have this meaning, but rather stands for the whole of the preceding statement can act as conjunctions, a singular idea. The computer seemed not to have been programmed to bear in mind this less common possibility of a whole statement before which determining its meaning rather than a single word (see 200. Special Uses of Relative Clauses).
In (e), the computer called for a comma after the starting also. The “rule” that it appeared to be following was that starting adverbs need a following comma – a rule that certainly does apply often but has exceptions. The rule is more accurately that starting adverbs relating to a whole sentence rather than to any particular part of one (usually) need a following comma (see 121. Sentence-Spanning Adverbs). Although also does very often relate to a whole sentence (functioning as a “connector”: see 40. Conjunctions versus Connectors), in (e) it just relates to useful to know.
In sentence (e), also is not being used as a “sentence” adverb. Instead, it is closely linked to the word after it, the adjective useful. It is rare for an adjective to be placed near the start of a sentence before the subject of the verb, but very possible when it has the role of complement (cf. is), the purpose being to highlight its meaning as the main message of the sentence (see 220. Features of Complements, #6). Also cannot have a following comma when used like this.
.
SPELLINGS THAT USUALLY MEAN ONE THING BUT SOMETIMES MEAN SOMETHING ELSE
English spellings representing two or more different words are examined in detail in this blog in 11. Homonyms and Homographs. I suspect that, when one of the words is rare or unconventional, they can be as problematic for grammar-checking programs as they are for English learners reading. Consider the following extract, which my computer wrongly said contained a “fragment” (“fragments” tend to be mentioned a lot in mistaken criticisms by computers):
(g) Each new birth confronted my parents with a major childcare problem, since my father never seemed able to take time off from his very demanding work. Leo’s arrival, some months after our change of address, particularly sticks in my memory.
The computer thought something vital was absent from the second of the two sentences here. Typically, no indication was given of what exactly this vital element was, but knowing that it usually turns out to be either the subject or the verb of the sentence, I followed the useful first step of looking carefully at the main verb of the sentence (sticks), and was soon able to hypothesise that the computer analysed this word as a noun, since it often is one and is slightly informal as a verb. This meant the sentence seemed to lack a main verb, and hence had to be labelled a “fragment”.
A different kind of alternative spelling is involved in the following:
(h) Schools start the new year in September.
The common error that the computer thought it recognised here was the use of new year without starting capital letters. It seemed unaware that this phrase can sometimes have small letters, and that the choice once again depends on meaning: if the new year in question is the calendar event of 1st January, or its celebration, then capitals are normal; but otherwise they are not. For more about this variability in the use of some capital letters, see 62. Choices with Capital Letters.