The Google Ngram Viewer is a neat tool. It allows you to check the relative frequency with which words and phrases appear, over time, in Google’s corpus of millions of digitized books. It follows that the Ngram Viewer can help you build a case that a given word or phrase is the more modern choice than another word or phrase and so is the better choice to use in a contract. I used Ngrams for that purpose in this post and this post.
But I wouldn’t ask the Ngram Viewer to do more than that. That’s why Bryan Garner’s recent article in the ABA Journal (here) caught my eye. In it, Garner offers what amounts to a shaggy-dog story about using the Ngram Viewer when deciding among contract usages.
In the portion of the article dealing with the Ngram Viewer, Garner notes that an Ngram shows that starting in 1970, indemnify and hold harmless started being used more frequently in print sources than indemnify and save harmless. He goes on to say that as regards whether hold harmless is ever necessary, an Ngram shows that indemnify the has always been used more frequently than indemnify and hold harmless the.
He then says this:
But this graph isn’t dispositive because (1) the references are probably mostly to printed books, including lay books, as opposed to quotations of contracts; and (2) the usual wording in contracts is indemnify the company and hold it harmless, not indemnify and hold harmless the company. So the data aren’t rock-solid here. They provide only a weak indication of whether indemnify alone suffices without and hold harmless. Big data alone won’t always give you an easy answer.
I agree with Garner’s two points. After all, if you want to explore contract usages, you’re much better off rummaging in the contracts filed by public companies on the SEC’s EDGAR system, and Westlaw allows you to run more sophisticated (although imperfect, see this post) searches of EDGAR than those you can run on the Ngram Viewer.
But I find perplexing his assertion that “the data aren’t rock-solid here” and “provide only a weak indication.” The data are in fact utterly irrelevant. Contract usages aren’t subject to a popular vote. The frequency with which a usage appears in a contract is no indication of whether it serves a useful purpose.
He then says, “Just so you know the answer to this question,” and goes on to explain that indemnify and hold harmless were treated as meaning the same thing until the late 19th century, when a few American courts created a distinction, and that the distinction still applies in about a third of the states. That’s interesting, but it doesn’t tell drafters what they want to know—What should I say in my contract!
Well, just so you—ahem—know the answer to this question, use just indemnify, and make sure that what follows specifies exactly what’s covered. That way, you avoid a fight over the meaning of indemnify or hold harmless or both, even in those jurisdictions that have succumbed to idiocy and held that there’s a meaningful distinction between the two. Explicit meaning trumps addled code.
By the way, you can find my indemnification language for nonparty claims here.