In his life of Filippo Brunelleschi, Vasari tells a story about how the artist won the commission to build the dome for the cathedral of Florence:

They [others competing for the commission] would have liked Filippo to speak his mind in detail, and to show his model, as they had shown theirs; but this he refused to do, proposing instead to those masters, both the foreign and the native, that whosoever could make an egg stand upright on a flat piece of marble should build the cupola, since thus each man's intellect would be discerned. Taking an egg, therefore, all those masters sought to make it stand upright, but not one could find the way. Whereupon Filippo, being told to make it stand, took it graciously, and, giving one end of it a blow on the flat piece of marble, made it stand upright. The craftsmen protested that they could have done the same; but Filippo answered, laughing, that they could also have raised the cupola, if they had seen the model or the design.

Vasari's story nicely illustrates hindsight bias, the tendency to believe that something is obvious in retrospect, whereas at the time it might actually have been much more challenging to accomplish. When we look back at the history of grammar study, we find a great deal that now seems absurdly wrong. Yet it's worth remembering that what is obvious to us now may not always have been so.

One victim of this hindsight bias is linguist Geoff Pullum. Writing at Lingua Franca, he complains about the virtually useless definition of pronouns to be found in traditional grammar books, even those published very recently, when tend to define the pronoun as a word that "replaces a noun."

It's astonishing that people ever accepted this drivel. Look at Murray's example of a pronoun "used instead of a noun" to avoid repetition: The man is happy; he is benevolent. If he stood for the noun, we'd predict *The man is happy; the he is benevolent, wouldn't we? The lesson is that when pronouns do stand for other constituents, it's noun phrases, not nouns. It's amazing that generations of scholars could miss something that obvious.

Pullum's grammatical analysis is, of course, exactly right. And the traditional definition of a pronoun surely needs updating. Yet it's not at all astonishing that generations of textbook writers could miss this point, because the clear distinction that Pullum assumes between a word-level and a phrase-level category does not exist in traditional grammar.

For anyone familiar with syntactic structure as elucidated by modern linguistics, a label like noun is a word-level category, whereas a phrase is a grammatical constituent that exists "above" the word level. A noun phrase is a phrase headed by a noun (and need not consist of more than one word), an adjective phrase is one headed by an adjective, etc.

In traditional grammar the same terms are handled in a distinctly different way, and when you run across terms like "noun phrase" or "verb phrase" in an old grammar book, they do not mean what linguists mean by the same terms.

Often, the expression "X phrase", when encountered in a traditional grammar book means something acting "in place of" X. In other words it's a label of function, not form. Thus an "adjective phrase" means any phrase that modifies a noun, regardless of its form. For example, in "the boy in the corner," "in the corner" is called an adjective phrase because it modifies boy, whereas if the same phrase appears in a sentence like "The boy worked in the corner," it is called an adverbial (or sometimes adverb) phrase.

Traditional grammar books would also call "in the corner" a prepositional phrase, which is the closest that they ever come to using a label based on grammatical form. Notice, though, that this use creates an inconsistency in their categorization, as in this one case they label by form, whereas in instances like "adjective phrase" and "adverbial phrase" they are labeling by function.

A still further inconsistency appears with the "verb phrase." When the term is used at all (and it's not all that common) it doesn't indicate the verb and all its complements and adjuncts (direct objects, etc.) the way it does in modern syntax. Instead, it simply means the string of auxiliaries and the main verb (e.g., must have been working). This usage doesn't indicate either form or function. It really depends on a loose notion of a phrase as merely a string of words that we can group together conceptually in some way.

When you see the term "noun phrase" in traditional grammar books, it actually means a phrase that is not headed by a noun and yet is functioning where we canonically expect there to be one. (Note that this use in the nineteenth century is less common than the equivalent "substantive phrase.") In other words, it functions analogously to the traditional use of "adjective phrase" or "adverbial phrase." For example, in "To sing at Carnegie Hall was Cecilia's main ambition.", the subject ("To sing at Carnegie Hall") would be called a noun phrase (or a substantive phrase, depending on the book). What we think of as a noun phrase (i.e., a noun and all its dependents) is, importantly, never called a noun phrase. The term noun could be used both for the simple word and for the phrase headed by a noun.

So when the old writers talked about pronouns replacing nouns, we have to keep in mind that they would not have seen the objection that the pronoun substitutes for a whole phrase to be a significant point. For them, a noun could be what we think of as a phrase, and so their definition is, in fact, consistent with their own terminology in a way that Pullum doesn't give credit for.

Don't get me wrong. It's certainly true that modern views of the phrase provide a more accurate (not to mention more elegant, and theoretically consistent) explanation how how language is structured. I can see no reason at all for a modern writer of grammar books to cling to the old, nineteenth-century way of looking at things. But for the older writers, at least, I don't find it helpful to blame them for not spotting what seems so obvious to us today, or to dismiss it as "drivel," or to think of their work as entirely static and entirely unreflective about the consequences of their definitions. Between the work of Robert Lowth in the eighteenth century and the mid-nineteenth century, the understanding of the phrase actually changes significantly (although that's a long story best for another time). With a modern understanding of the phrase, the weaknesses of the traditional definitions become glaringly obvious. But without that theoretical preliminary in place, the grammarians of the day were like the other builders trying to balance Brunelleschi's egg. They were groping towards a solution, but they were predisposed to think within the schemata they had available to them, and so they couldn't quite get the egg to stand up.