I've already written a little about why the late Gunther Schuller's The Compleat Conductor is such a mess. The main premise - that you should do what the score says - fails to hold up under even cursory examination; virtually every page of the book (especially Parts 1 and 2) contains some kind of logical flaw. (Some of his opinions in Part 3 about how various pieces in the standard repertoire should be performed are interesting, and his thorough analysis of recordings is valuable, but beyond that I can't find much else good to say.) The notion that we need not do what's on the page, or that reproducing the composer's intent shouldn't be the goal, strikes some people as heretical, and I'll save the less concrete philosophical stuff for another time. The purpose here is to illustrate another repudiation of the notation-equals-intent myth which underpins Schuller's thesis.
To reiterate my argument in the Sibelius post linked above, if composers really put what they wanted on the page, that ought to be reflected in their recordings of their own music. But it isn't: composers constantly deviate from their scores in ways that Schuller would consider unacceptable. These deviations, far from being the exceptional one-offs that Schuller makes them out to be, are par for the course. Given this reality, in order to justify Schuller's approach, we have to believe that composers were painstaking with their notation but got bonked in the head prior to their recording sessions and suddenly forgot how their own music was supposed to go.
(One of Schuller's logical inconsistencies: while he dismisses composers' own recorded vagaries as irrelevant [pg. 43], he nonetheless considers a hypothetical "early 19th-century recording" as "indisputable evidence" of how a piece from that era should be performed [pg. 16]. Why 20th-century recordings made by the composer shouldn't be held as similarly definitive isn't made clear. Given that, if anything, notation has become more and more precise over the centuries, Schuller should probably be grateful that these hypothetical 19th-century recordings don't exist; they would almost certainly blow his thesis out of the water.)
This time, we're comparing Howard Hanson's score and 1939 recording of his Symphony No. 2. As always, the usual caveats apply: the recording may not reflect his intent for any number of reasons, and the score may include mistakes. (I'm looking at what is purported to be a "Critical Edition," though it does contain some errors. But the relevant features here seems consistent with other editions, and the editors claim to be using definitive sources including Hanson's own corrections.) Unlike the shorter and sparsely notated Sibelius example, the Hanson gives us lots of opportunities to compare page to sound. Here is a table of what I found:
By my count, there are 53 "tempo events" in the piece, 28 of which (in red) deviate from the notation - more than half. While I can't be sure that Schuller would agree entirely with my list of "errors," I'm doing as much as possible to match his criteria based on what he writes in the book. He allows for gentle ebbs and flows of tempo; thus I only list departures from the MM of more than 10% as errors. Similarly, unwritten tempo changes of more than 10% are also flagged - after all, if Hanson wanted a ritardando he would have written one, according to Schuller. (Even determining what constitutes a "tempo event" is problematic; if the tempo stays the same for 15 bars, is that one data point, or should Hanson get credit for not changing tempo in every bar where no tempo change is marked?)
Additionally, I haven't listed other errors of sorts that Schuller laments in his book. There are numerous rhythmic distortions in the recording - the ascending figure that first appears in mvmt 1/ms. 107 and comes back several times is almost never played with a real triplet; the half note triplets starting at mvmt 3/ms. 126 are inconsistent; and the English horn player seems to have no idea what to do with the triplets at mvmt 1/ms. 131 and 134. The fermata over the horn whole note in mvmt 1/ms. 332 is if anything shorter than the note should be even without a fermata. There are numerous places where the dynamics fail to correspond to the page; given the limitations of the recordings of that age, I've elected not to include those either.
In any case, I can't imagine any interpretation of this recording that doesn't include literally dozens of what Schuller would consider errors. Schuller may primarily address 19th-century standard repertoire, but I don't see any reason why Hanson should be exempted from his thesis. At one point, Schuller makes a barb about "trendy, anachronistic, neo-Romantic stylistic pleasantry" not qualifying as true "contemporary music" [pg. 53], and it's very possible that Hanson would fall into that category in Schuller's mind. But Hanson is a more experienced conductor than many composers, and even if he wrote in a style Schuller didn't necessarily care for, that doesn't suggest that Hanson was incapable of getting his notation down the way he wanted it.
Schuller writes that the conductor's ego, at its most harmful,
assumes it can impose its own ideas, its own fancies and whims, on the score and the work of the composer. Conductors who perpetrate these impositions really think they know better than the composer what the composer 'had in mind,' and how the composer 'should have notated the work.' [pg. 48]
But given what we've seen from Hanson conducting Hanson (and the myriad other composers in similar situations), for such a statement to be true, there must exist a remarkable psychological phenomenon: the composer who thinks he knows better than himself.