Essays_, p. 196.

OBS. 20.--The two _special_ rules commonly given by the grammarians, for
the construction of relatives, are not only unnecessary,[382] but faulty. I
shall notice them only to show my reasons for discarding them. With whom
they originated, it is difficult to say. Paul's Accidence has them, and if
Dean Colet, the supposed writer, did not take them from some earlier
author, they must have been first taught by _him_, about the year 1510; and
it is certain that they have been copied into almost every grammar
published since. The first one is faulty, because, "_When there cometh no
nominative case between the relative and the verb, the relative shall_ [not
always] _be the nominative case to the verb_;" as may be seen by the
following examples: "Many are the works of human industry, _which_ to begin
and finish are [say _is_] hardly granted to the same man."--_Dr. Johnson's
Adv. to Dict._ "They aim at his removal; _which_ there is reason to fear
they will effect."--"_Which_ to avoid, I cut them off."--_Shak., Hen. IV_.
The second rule is faulty, because, "_When there cometh a nominative case
between the relative and the verb, the relative shall_ [not always] _be
such case as the verb will have after it_;" as may be seen by the following
examples: "The author has not advanced any instances, _which_ he does not
think _are_ pertinent."--_Murray's Gram._, i, 192. "_Which_ we have reason
to think _was_ the case with the Greek and Latin."--_Ib._, 112. "Is this
your son, _who_ ye say _was born_ blind?"--_John_, ix, 19. The case of the
relative cannot be accurately determined by any rules of mere location. It
may be nominative to a verb afar off, or it may be objective with a verb
immediately following; as, "_Which_ I do not find that there ever
_was_."--_Knight, on the Greek Alphabet_, p. 31. "And our chief reason for
believing _which_ is that our ancestors did so before us."--_Philological
Museum_, i, 641. Both these particular rules are useless, because the
general rules for the cases, as given in chapter third above, are
applicable to relatives, sufficient to all the purpose, and not liable to
any exceptions.

OBS. 21.--In syntactical parsing, each word, in general, is to be resolved
by some _one_ rule; but the parsing of a pronoun commonly requires _two_;
one for its agreement with the noun or nouns for which it stands, and an
other for its case. The rule of agreement will be one of the four which are
embraced