In (1), a comment is introduced by ‘a man’ in apposition with Lewis; ‘a man’ is antecedent to a defining relative clause; separate them, and the antecedent is meaningless. But next remove the connecting words ‘a man’, and the relative changes at once its antecedent and its nature: the antecedent is ‘Lewis’; the relative is non-defining; and the clause is a comment, and does not merely contribute to one.

b. ‘That’ and ‘who’ or ‘which’.

‘That’ is evidently regarded by many writers as nothing more than an ornamental variation for ‘who’ and ‘which’, to be used, not indeed immoderately, but quite without discrimination. The opinion is excusable; it is not easy to draw any distinction that is at all consistently supported by usage. There was formerly a tendency to use ‘that’ for everything: the tendency now is to use ‘who’ and ‘which’ for everything. ‘That’, from disuse, has begun to acquire an archaic flavour, which with some authors is a recommendation. De Quincey, for one, must certainly have held that in exalted prose ‘that’, in all connexions, was the more dignified relative; his higher flights abound in curious uses of the word, some instances of which are quoted below.

This confusion is to be regretted; for although no distinction can be authoritatively drawn between the two relatives, an obvious one presents itself. The few limitations on ‘that’ and ‘who’ about which every one is agreed all point to ‘that’ as the defining relative, ‘who’ or ‘which’ as the non-defining. We cannot say ‘My father, that left Berlin last night, will shortly arrive’, and an examination of instances would show that we can never use ‘that’ where the clause is unmistakably non-defining. On the other hand, we cannot say ‘All which I can do is useless’; this time, it is true, the generalization will not hold; ‘which’ can, and sometimes must, be used, and ‘who’ commonly is used, in defining clauses. But that is explained partly by the obvious inconvenience sometimes attending the use of ‘that’, and partly by the general tendency to exclude it from regular use, which has already resulted in making it seem archaic when used of persons, except in certain formulae.

The rules given below are a modification of this principle, that ‘that’ is the defining, ‘who’ or ‘which’ the non-defining relative; the reason for each modification is given in its place. We must here remind the reader of the distinction drawn in a. between defining and non-defining clauses: a defining clause limits the application of the antecedent, enabling us to select from the whole class to which the antecedent is applicable the particular individual or individuals meant.

1. ‘That’ should never be used to introduce a non-defining clause; it is therefore improperly used in all the following examples:

‘That I represent’ is right: ‘that daily share’ is wrong.

‘That’ and ‘which’ should change places.

If this is a defining clause, excluding ‘the social and economic sciences that’ do not interest the writer, the comma after ‘sciences’ should be removed.

2. ‘Who’ or ‘which’ should not be used in defining clauses except when custom, euphony, or convenience is decidedly against the use of ‘that’. The principal exceptions will be noted below; but we shall first give instances in which ‘that’ is rightly used, and others in which it might have been used with advantage.

‘That’ should have been used in both clauses.

In the following examples, ‘that’ is to be preferred to ‘which’; especially with antecedent ‘it’, and after a superlative or other word of exclusive or comprehensive meaning, such as ‘all’, ‘only’, ‘any’.

‘Which consisted’ is right: but we should have ‘that was held’; the clause defines.

‘Of which’ in a defining clause is one of the recognized exceptions; but we ought to have ‘that subserve’.

3. We come now to the exceptions. The reader will have noticed that of all the instances given in (2) there is only one—the last—in which we recommend the substitution of ‘that’ for ‘who’; in all the others, it is a question between ‘that’ and ‘which’. ‘That’, used of persons, has in fact come to look archaic: the only cases in which it is now to be preferred to ‘who’ are those mentioned above as particularly requiring ‘that’ instead of ‘which’; those, namely, in which the antecedent is ‘it’, or has attached to it a superlative or other word of exclusive meaning. We should not, therefore, in the Spectator instance above, substitute ‘the person that desires’ for ‘who desires’; but we should say

Outside these special types, ‘that’ used of persons is apt to sound archaic.

4. It will also have been noticed that all the relatives in (2) were either in the subjective case, or in the objective without a preposition. ‘That’ has no possessive case, and cannot take a preposition before it. Accordingly ‘the man that I found the hat of’ will of course give place to ‘the man whose hat I found’; and ‘the house in which this happened’ will generally be preferred to ‘the house that this happened in’. The latter tendency is modified in the spoken language by the convenient omission of ‘that’; for always in a defining clause, though never in a non-defining, a relative in the objective case, with or without a preposition, can be dropped. But few writers like, as a general rule, either to drop their relatives or to put prepositions at the end. ‘The friends I was travelling with’, ‘the book I got it from’, ‘the place I found it in’, will therefore usually appear as

5. Euphony demands that ‘that that’ should become ‘that which’, even when the words are separated; and many writers, from a feeling that ‘which’ is the natural correlative of the demonstrative ‘that’, prefer the plural ‘those which’; but the first example quoted in (2) seems to show that ‘those ... that’ can be quite unobjectionable.

6. A certain awkwardness seems to attend the use of ‘that’ when the relative is widely separated from its antecedent. When, for instance, two relative clauses are coordinate, some writers use ‘that’ in the first, ‘which’ in the second clause, though both define. This point will be illustrated in c., where we shall notice that inconsistency in this respect sometimes obscures the sense.

It may seem to the reader that a rule with so many exceptions to it is not worth observing. We would remind him (i) that it is based upon those palpable misuses of the relatives about which every one is agreed; (ii) that of the exceptions the first and last result from, and might disappear with, the encroachment of ‘who’ and the general vagueness about the relatives; while the other two, being obvious and clearly defined, do not interfere with the remaining uses of ‘that’; (iii) that if we are to be at the expense of maintaining two different relatives, we may as well give each of them definite work to do.

In the following subsections we shall not often allude to the distinction here laid down. The reader will find that our rules are quite as often violated as observed; and may perhaps conclude that if the vital difference between a defining and a non-defining clause were consistently marked, wherever it is possible, by a discriminating use of ‘that’ and ‘which’, false coordination and other mishandlings of the relatives would be less common than they are.

c. ‘And who’; ‘and which’.

The various possibilities of relative coordination, right and wrong, may be thus stated: (i) a relative clause may be rightly or wrongly coordinated with another relative clause; this we shall call ‘open’ coordination; (ii) it may be rightly or wrongly coordinated with words that are equivalent to a relative clause, and for which a relative clause can be substituted; ‘latent’ coordination; (iii) a clause that has obviously no coordinate, open or latent, may yet be introduced by ‘and’ or other word implying coordination; for such offenders, which cannot be coordinate and will not be subordinate, ‘insubordination’ is not too harsh a term.

The following are ordinary types of the three classes:

A coordination in which ‘and’ is the natural conjunction may also be indicated simply by a comma; there is safety in this course, since the clause following the comma may be either coordinate or subordinate. But we have to deal only with clauses that are committed to coordination.

‘Insubordination’ will not detain us long; it is always due either to negligence or to gross ignorance; we shall illustrate it in its place with a few examples, but shall not discuss it. With regard, however, to open and latent coordination opinions differ; there is an optimist view of open coordination, and a pessimist view of latent, both of which seem to us incorrect. It is held by some that open coordination (provided that the relatives have the same antecedent) is never wrong, and by some—not necessarily others—that latent coordination is never right: we shall endeavour to show that the former is often wrong, and the latter, however ungainly, often right.

The essential to coordination is that the coordinates should be performing the same function in the sentence. It is not necessary, nor is it enough, that they should be in the same grammatical form: things of the same form may have different functions, and things of different forms may have the same function. If we say ‘Unambitious men, and who have no experience’, ‘unambitious’ and ‘who have no experience’ are not in the same form, but they have the same function—that of specifying the class of men referred to. Their grammatical forms (vocabulary permitting) are interchangeable: a defining adjective can always take the form of a relative clause, and a defining relative clause can often take the form of an adjective: ‘inexperienced men, and who have no ambition’. ‘Unambitious’ is therefore the true grammatical equivalent of ‘who have no ambition’, and latent coordination between it and a relative clause is admissible.

On the other hand, among things that have the same grammatical form, but different functions, are the defining and the non-defining relative clause. A non-defining clause, we know, can be removed without disturbing the truth of the predication; it has therefore no essential function; it cannot therefore have the same function as a defining clause, whose function we know to be essential. It follows that open coordination is not admissible between a defining and a non-defining clause; and, generally, coordination, whether open or latent, is admissible between two defining or two non-defining coordinates, but not between a defining and a non-defining.

Our object, however, in pointing out what seems to be the true principle of relative coordination is not by any means to encourage the latent variety. It has seldom any advantage over full coordination; it is perhaps more apt to lead to actual blunders; it is usually awkward; and it does violence—needless violence, as often as not—to a very widespread and not unreasonable prejudice. Many writers may be suspected of using it, against their better judgement, merely for the purpose of asserting a right; it is their natural protest against the wholesale condemnation of ignorant critics, who do not see that latent coordination may be nothing worse than clumsy, and that open coordination may be a gross blunder. For the benefit of such critics it seems worth while to examine the correctness of various examples, both open and latent; on the other merits and demerits of the latent variety the reader will form his own judgement.

(i) Open coordination.

All these are correct: in the first both clauses are non-defining, in the others both define.

Wrong: the first clause defines, the second not.

Wrong. The first clause defines, the second is obviously one of comment: the ‘scene’ is not distinguished from those that the leader did take steps to avert.

Wrong. The communes that ‘will not permit’ are not meant to be distinguished from those that will. The second clause is comment, the first defines.

All these are wrong. Thackeray would probably have been saved from these false coordinations if he had observed the distinction between ‘that’ and ‘which’: ‘In the best French (that) he could muster, which in sooth was...’.

Probably the second clause, like the first, is meant to define: if so, the coordination is right; if not, it is wrong. We have alluded to the tendency to avoid ‘that’ when the relative is widely separated from its antecedent; here, the result is ambiguity.

Wrong: ‘what (that which)’ defines, the ‘and which’ clauses do not.

(ii) Latent coordination, between relative clause and equivalent, is seldom correct when the relative clause is non-defining; for the equivalent, with few and undesirable exceptions, is always a defining adjective or phrase, and can be coordinate only with a defining clause. The equivalent must of course be a true one; capable, that is, of being converted into a relative clause without altering the effect of the sentence. Neglect of this restriction often results in false coordination, especially in one particular type of sentence. Suppose that a historian, after describing some national calamity, proceeds: ‘In these distressing circumstances....’ Here we might seem to have two possible equivalents, ‘these’ and ‘distressing’. First let us expand ‘these’ into a relative clause: ‘In the distressing circumstances that I have described’. This, in the context, is a fair equivalent, and as often as not would actually appear instead of ‘these’. But next expand ‘distressing’: ‘In these circumstances, which were distressing’, a non-defining clause. To this expansion no writer would consent; it defeats the object for which ‘distressing’ was placed before the antecedent. That object was to record his own sensibility without disparaging the reader’s by telling him in so many words (as our relative clause does) that the circumstances were distressing; and it is secured by treating ‘distressing’ not as a separate predication but as an inseparable part of the antecedent. ‘Distressing’, it will be observed, cannot give us a defining clause; it is obviously meant to be co-extensive with ‘these’; we are not to select from ‘these’ circumstances those only that are ‘distressing’. Moreover, as ‘these’, although capable of appearing as a relative clause, can scarcely require another relative clause to complete the limitation of the antecedent, it follows that in sentences of this form coordination will generally be wrong. We have examples in the Cowper quotation below, and in the anonymous one that precedes it.

‘Rules that shall be general, and that can....’

‘The man who was goldsmith to ... and who’.

All these are correct, with defining coordinates throughout.

‘Who has ... views, and who....’ ‘Sic’ is the comment of the Times writer. The coordination is correct.

These are the instances of false expansion alluded to above. The former is based on the non-defining expansion ‘in all its moods, which are varied and capricious’; the true expansion being ‘in all the varied and capricious moods in which it reveals itself’, a defining clause, which will not do with the ‘and which’. Similarly, the second is based on the non-defining expansion ‘in my uncle’s present state, which is an infirm one’; the true expansion is ‘in the infirm state in which my uncle now is’. In both, a non-defining clause is coordinated with words that can only yield a defining clause.

If the writer means us to distinguish, among the innovations introduced by the Tudors, those that had also been taken away, the ‘and which’ clause defines, and the coordination is right. But more probably the clause conveys independent information; the coordination is then wrong.

Wrong. ‘A plural substantive’ can yield only the defining clause ‘a substantive that is plural’. Now these words contain an inference from a general grammatical principle (that a plural verb must have a plural subject); and any supplementary defining clause must also be general, not (like the ‘and which’ clause) particular. We might have, for instance, ‘Amabant, being plural, and finite, must have for its subject a plural substantive, and which is in the nominative case’. But the ‘and which’ clause is evidently non-defining; the inference ends at ‘substantive’; then comes the application of it to the particular case.

Wrong. The ‘and which’ clause is non-defining; none of the three possible antecedents (‘Theory’, ‘limit’, ‘imposition’) will give a non-defining clause.

This illustrates an important point. ‘Of Europe’ gives the defining clause ‘that prevails in Europe’; the coordination therefore requires that the ‘and which’ clause should define. Now a defining clause must contain no word that is not meant to contribute to definition; if, then, the ‘and which’ clause defines, the writer wishes to distinguish the religion in question, not only from those European religions that have not been colonially introduced into America, but also from those European religions that have been introduced, but whose introduction is not a matter for regret; that is the only defining meaning that ‘unhappily’ can bear, and unless we accept this interpretation the clause is non-defining.—We shall allude to this sentence again in d., where the possibilities of parenthesis in a defining clause are discussed.

One of our ‘few and undesirable exceptions’, in which the clause-equivalent is non-defining (‘who was already on the spot’); for a person’s name can only require a defining clause to distinguish him from others of the same name. The sentence is an ugly one, even if we remove the ‘and who’ clause; but the coordination is right.

(iii) Insubordination.

To these we may add examples in which the coordinated relatives have different antecedents. In practice, nothing can justify such coordination: in theory, it is admissible when the antecedents are coordinate, as in the following sentence:

But in the following instances, one antecedent is subordinate to another in the same clause, or is in a clause subordinate to that of the other.

It was of Mr. Lovelace that the uncles were afraid.

d. Case of the relative.

Special attention was not drawn, in the section on Case, to the gross error committed in the following examples:

These writers evidently think that in ‘whom we think are hungry’ ‘whom’ is the object of ‘we think’. The relative is in fact the subject of ‘are’; and the object of ‘we know’ is the clause ‘who are hungry’; the order of the words is a necessary result of the fact that a relative subject must stand at the beginning of its clause.

(The same awkward necessity confronts us in clauses with ‘when’, ‘though’, &c., in which the subject is a relative. Such clauses are practically recognized as impossible, though Otway, in a courageous moment, wrote:

Some writers, with a consistency worthy of a better cause, carry the blunder into the passive, renouncing the advantages of an ambiguous ‘which’ in the active; for in the active ‘which’ of course tells no tales.

‘Which it was anticipated would be’. Similarly, the passive of ‘men whom we-know-are-honest’ is the impossible ‘men who are-known-are-honest’: ‘men who we know are honest’ gives the correct passive ‘men who it is known are honest’.

Nor must it be supposed that ‘we know’ is parenthetic. In non-defining clauses (Jones, who we know is honest), we can regard the words as parenthetic if we choose, except when the phrase is negative (Jones, who I cannot think is honest); but in a defining clause they are anything but parenthetic. When we say ‘Choose men who you know are honest’, the words ‘you know’ add a new circumstance of limitation: it is not enough that the men should in fact be honest; you must know them to be honest; honest men of whose honesty you are not certain are excluded by the words ‘you know’. Similarly, in the Guernsey Advertiser quotation above, the writer does not go the length of saying that these are the only gentlemen who would work: he says that they are the only ones of whom he feels sure. The commas of parenthesis ought therefore to go, as well as the comma at ‘island’, which is improper before a defining clause.

The circumstances under which a parenthesis is admissible in a defining clause may here be noticed.

(i) When the clause is too strict in its limitation, it may be modified by a parenthesis:

A whole class, excluded by the defining clause, is made eligible by the parenthesis.

(ii) Similarly, a parenthesis may be added to tell us that within the limits of the defining clause we have perfect freedom of choice:

They must have held office, that is all; it does not matter when.

(iii) Words of comment, indicating the writer’s authority for his limitation, his recognition of the sentiments that it may arouse, and the like, properly stand outside the defining clause: when they are placed within it, they ought to be marked as parenthetic.

The latter sentence is an adaptation of one considered above on p. 91. ‘Unhappily’ there appeared not as a parenthesis but as an inseparable part of the relative clause, which was therefore defining or non-defining, according as ‘unhappily’ could or could not be considered as adding to the limitation. But with the altered punctuation ‘unhappily’ is separable from the relative clause, which may now define: ‘that obtains in Europe and (I am sorry to have to add) in America.’

In sentences of this last type, the parenthesis is inserted in the defining clause only for convenience: in the others, it is an essential, though a negative, part of the definition. But all three types of parenthesis agree in this, that they do not limit the antecedent; they differ completely from the phrases considered above, which do limit the antecedent, and are not parenthetic.

e. Miscellaneous uses and abuses of the relative.

(i) A relative clause is sometimes coordinated with an independent sentence; such coordination is perhaps always awkward, but is not always incorrect. The question arises chiefly when the two have a common subject expressed only in the relative clause; for when the subject is expressed in both, the independent sentence may be taken to be coordinate, not with the relative clause, but with the main sentence to which the relative clause is attached, as in the following instance:

Retain ‘I’, and ‘I was’ may be coordinate with ‘he had left’: remove it, and the coordination is necessarily between ‘I was’ and ‘I felt’. In our next examples the writers are committed:

The first two of these are wrongly coordinated: the third, a curiosity in other respects, is in this respect right. The reason is that in the first two we have a defining, in the third a non-defining relative clause. A defining clause is grammatically equivalent to an adjective (‘violated laws’, ‘the popular idea’), and can be coordinated only with another word or phrase performing the same function; now the phrase ‘we have been receiving’, not being attached to the antecedent by means of a relative, expressed or understood, is not equivalent to an adjective. We could have had ‘and (which we) have been properly punished for neglecting’, or we could have had the ‘and’ sentence in an adverbial form, ‘with the fitting result’; but coordination between the two as they stand is impossible.

The Burke sentence is a worse offender. Coordination of this kind is not often attempted when the antecedent of the relative is subject of the main sentence; and when it is attempted, the two coordinates must of course not be separated by the predicate. If we had had ‘the idea which mankind most commonly conceive of proportion, and very seldom trouble themselves about anything further’, the coordination would have been similar to the other, and could have been rectified in the same way (‘and beyond which they very seldom ...’, or ‘to the exclusion of any other considerations’). But this alteration we cannot make; for there is a further and an essential difference. The Daily Telegraph writer evidently meant his second coordinate to do the work of a defining clause; he has merely failed to make the necessary connexion, which we supply, as above, either by turning the words into a second defining clause, or by embodying them, adverbially, in the first. Burke’s intention is different, and would not be represented by our proposed alteration in the order. All that a defining clause can do in his sentence is to tell us what idea is going to be the subject. If we were to give a brief paraphrase of the whole, italicizing the words that represent the second coordinate, it would be, not ‘mankind’s sole idea of proportion is the suitableness ...’, but ‘mankind’s idea of proportion is the suitableness ..., and very little else’; for the question answered is, not ‘what is mankind’s sole idea?’ but ‘what is mankind’s idea?’ In other words, the second coordinate belongs in intention not, like the relative clause, to the subject, but to the predicate; to rectify it, we must either make it part of the predicate (‘and is not concerned with ...’), or, by inserting ‘they’, coordinate it with the main sentence. Obvious as the latter correction is, the sentence repays close examination, as illustrating the incoherence of thought that may underlie what seems a very trifling grammatical slip.

But in our third example, the relative clause is non-defining; it is grammatically equivalent to, and could be replaced by, an independent sentence: ‘Many persons are in the habit of using it’. There is nothing grammatically wrong in this type of coordination; it is objectionable only because it seems to promise what it does not fulfil. When the common subject of two coordinates is expressed only with the first, it is natural to assume that all words preceding it are also to be applied to both coordinates; and the violation of this principle, though not of course ungrammatical, is often felt to be undesirable in other than relative clauses.

(ii) In the sentences considered above, the antecedent of the relative did not belong to the second coordinate, and could not have been represented in it without the material alterations there proposed. But it may also happen that the antecedent, as in the following examples, belongs equally to both coordinates, being represented in the first by a relative, in the second by some other pronoun.

In the pair of parallel coordinates from Mr. Bryce, insert the suppressed ‘was’, and it becomes clear that ‘whose’, not ‘their’, is the right pronoun.

In the ‘Pizarro’ sentence, ‘it’ is not only superfluous, but disturbing to the reader, who assumes that ‘which’ is common to both clauses, and on reaching ‘it’ has to glance back and check the sentence. Here, as often, the pronoun seems to be added to restore an ill-balanced sentence; but that can be done in several other ways. In the Richardson sentence also the ‘it’ should go.

More commonly, the repetition of the antecedent in another form results from the superstitious avoidance of a preposition at the end: