Webster's Dictionary of English Usage (1989) includes this discussion (part of a longer comment) on preventative versus preventive:
You may wonder how preventative came to be objected to. The earliest attack is in [Richard Meade] Bache 1869 [Vulgarisms and Other Errors of Speech]; he said that there was no such word. Bache's book was one of those used by [Alfred] Ayres 1881 [The Verbalist], who also criticized preventative. From Ayres it went to [Frank] Vizetelly 1906 [A Desk-Book of Errors in English], [Ambrose] Bierce 1909 [Write It Right], [H. N.] MacCracken & [Helen] Sandison 1917 [Manual of Good English], and [Charles] Lurie 1927 [How to Say It]. [Henry] Fowler 1926 [A Dictionary of Modern English Usage] picked it up too, and so it has gone right down to the 1980s. A couple of commentators—[Bergen] Evans 1962 [Comfortable Words] and [William] Watt 1967 [A Short Guide to English Usage]—realize that preventative is acceptable. Probably none of the recent objectors realizes that his opinion goes back to Bache 1869. But the moderns do not claim that preventative is nonexistent anymore; nowadays they say it is wrong because it is "irregularly formed." That is not so, of course. It is formed in the same way as authoritative, quantitative, normative, talkative, and other words to which no one objects.
The same comment appears in the (very slightly) updated Merriam-Webster's Dictionary of English Usage (1994).
But whatever the validity of preventative as a word may be today, the notion that disapproval of it originated with Bache's 1869 book is demonstrably wrong. In fact, the earliest such instance that I'm aware of occurs in the first Merriam-Webster edition of An American Dictionary of the English Language (1847). There you'll find the following regular-size entries for preventable and prevented:
PREVENTABLE, a. That may be prevented or hindered. Reynolds.
PREVENTED, pp. Hindered from happening or taking effect.
And between these two entries, in smaller type, this bracketed note appears:
[PREVENTATIVE is a gross blunder.]
Though the version of the relevant dictionary in Google Books is dated 1850, that edition was first published in 1847 (I have the 1852 version), which makes its denunciation of preventative 22 years older than Bache's.
With regard to the historical legitimacy of preventative, I think it's somewhat misleading to focus on the fact that its first known occurrence is just 30 years younger than the first known occurrence of preventive. The latter term appears (both as a noun and as an adjective) in every dictionary I've consulted, starting with Samuel Johnson's Dictionary of the English Language (1756), and ranging through the various nineteenth-century editions of Webster's (1806, 1828, 1840, 1847, 1867, and 1890) and the Webster's Collegiate series (1898, 1910, 1916, 1931, 1936, 1949, 1963, 1973, 1983, 1993, and 2003). In contrast, preventative makes its bracketed debut in the 1847 Webster's and then reappears in the 1864 Webster's with this full-scale entry:
Preventative, n. That which prevents;—incorrectly used instead of preventive, q. v.
This dictionary appeared five years before Bache's book. The entry for preventative in Webster's International Dictionary (1890) is identical to the one that appears in the 1864 dictionary. Webster's New International Dictionary (1910) updates the entry to read as follows:
preventative, n. That which prevents;—an unnecessary and irregularly formed doublet of preventive.
The first and second editions of Webster's Collegiate Dictionary (1898 and 1910) show no such disapproval, despite being based on the 1890 Webster's International:
preventative, n. A preventive.
But the Third (1916), Fourth (1931), and Fifth (1936) Collegiates omit preventative altogether. The term reappears in the Sixth Collegiate (1949), with the old cautionary note restored (and for the first time with acknowledgment of the word's adjective form):
preventative, n. Preventive;—an irregular formation. —preventative, adj.
Finally, the Seventh Collegiate (1963) removes the stigmatizing "—an irregular formation" from the definition; and for the past 50 years, preventative has enjoyed equal status with preventive in Merriam-Webster's estimation.
Returning to the coverage of preventative in Webster's Dictionary of English Usage, it seems not at all unlikely that early grammar commentators relied as much on Merriam-Webster's hostile view of the term from 1847 through 1910 as on Richard Bache's hostile view in 1869. Similarly the "moderns" who criticized the term in more-recent times as "irregularly formed" might have found their opinion comfortingly corroborated in Merriam-Webster's Sixth Collegiate, which was the Collegiate Dictionary of record for the years between 1949 and 1962.
As for the WDEU's assertion that the preventative is just as regularly formed as authoritative, quantitative, normative, and talkative, the author of that statement neglects to observe that those four terms are adjectives that lack an earlier, identically spelled noun form and consequently do not owe their existence to it; historically, the criticism of preventative as "irregularly formed" referred to preventative as a noun (which was the only form of the word that Merriam-Webster's Collegiate Dictionary series acknowledged until 1949). A better example for the WDEU's purposes would have been recitative (from the Italian recitativo), which appears as a noun in Johnson's 1756 dictionary and in all Webster's dictionaries, and as an adjective in Webster's 1828 dictionary and all subsequent ones. But whether the noun recitative establishes a regular pattern of formation for the noun preventative—a word not derived from Italian (the corresponding Italian noun, according to Google Translate, is preventiva)—seems at least debatable.
Follow-up: What did critics mean by “irregularly formed?
Seven months later, it occurs to me that the WDEU may simply have misunderstood what earlier Merriam-Webster lexicographers meant by the phrases “irregularly formed” and “irregular formation.” The author of the WDEU article seems to think that the criticism is specifically about attaching the ending -ative instead of the ending –ive to preserved roots that end in the letter t.
But if that were so, why didn’t those lexicographers attack the word representative, too? It was certainly available for criticism: The Eleventh Collegiate reports that the word has been around as an adjective since the fourteenth century and as a noun since 1635.
Either representative owes its immunity from criticism to the laughable inconsistency of nineteenth- and (pre-1963) twentieth-century language experts, or those experts were applying a different notion of “regular formation” from the one that the WDEU attributes to them. The latter possibility seems more likely to me.
What would their notion have been, then? My guess is that they considered a group of words to be regularly formed if certain members of the group maintained consistent endings. Thus, for instance, each of these six word groups would qualify as regularly formed:
collect (v.), collection (n.), collective (adj.) collective (n.)
compete (v.), competition (n.), competitive (adj.)
prevent (v.), prevention (n.), preventive (adj.), preventive (n.)
relate (v.), relation (n.), relative (adj.) , relative (n.)
represent (v.), representation (n.), representative (adj.), representative (n.)
sedate (v.), sedation (n.), sedative (adj.), sedative (n.)
What makes each of these words regularly formed within its word group isn’t that it follows some blanket rule about adding –ive instead of –ative or –etive to certain noun and adjective forms; it’s that the same decision about how to handle those endings applies to all relevant forms in the specific group.
In other words, once representative is established as the preferred adjective form in the represent word group, the later regular noun forms are representative and representation; and representive and represention are irregularly formed. Likewise, if preventative were the first preferred noun form for the prevent word group, the later regular noun and adjective forms would be preventation and preventative, and prevention and preventive would be irregular forms.
I’m not claiming that this line of reasoning is persuasive or that it proves that preventative is “irregularly formed” as a matter of modern grammatical interpretation. I’m merely pointing out that such an understanding of “irregular formation” is internally consistent (at least in the subset of examples I’ve listed above) and that, as a basis for analysis, it is far more descriptive than prescriptive.