4. Thus we must be very careful how we apply the conception of “parts of speech.” The grammatical analysis which classifies words as substantives, adjectives, adverbs, verbs, and the like, is not to be taken as telling us what words are by themselves, but just the opposite, viz. what they do when employed in a significant sentence. They are studied separately for convenience in attending to them, as we may study the wheels and pistons of an engine; but the work which gives them their names can only be done when they are together. This truth is often expressed by saying that “the sentence is the unit of language,” _i.e._ a word taken by itself cannot have a complete meaning--unless it is a verb, or used with verbal force, for a verb is an unanalysed sentence. If any one uses a substantive or adverb by itself, we think that he has not finished his sentence, and no meaning is conveyed to our minds. We ask him, “Well, what about it?” The same is true, as we saw, of a relative clause. If we read in a newspaper such a clause as this, “The epidemic of influenza, which has appeared in England for three successive seasons,” followed by a full stop, we should infer, without hesitation, that some words had dropped out by accident. Of course such a {86} combination of words would make us think something, but the meaning which we might ascribe to it would be conjectural; we should necessarily complete the thought for ourselves by some affirmation--some relation to reality--while recognising that no such relation was given in the clause as we read it. Nothing less than a sentence, or, omitting the wish and the command, nothing less than a proposition, conveys a meaning in which the mind can acquiesce as not requiring to be supplemented conjecturally. There are traces in language that indicate the sentence to have been historically prior to the word. I question whether the word could be certainly distinguished within the sentence in early languages that have not been reduced to writing. The tendency of reflective analysis, as in grammar and dictionaries, is to give it a more and more, artificial isolation. The Greeks did not separate their words in writing, and they wrote down the change in a terminal consonant produced by the initial letter of the next word, just as if it was within a compound word. Nor had they really any current term co-extensive with our “word.” Where we should say “the word ‘horse’” they most commonly use the neuter article “the” followed by the word in question as if in quotation-marks (”the ‘horse’“). In defining noun and verb, Aristotle has no simple class name like ”word“ to employ as a common element of the definition, but uses the curious description ”a portion of discourse, of which no part has a meaning by itself.“