It's not really possible to say how many voices there are (I hear), there appears to be some confusion about "doubling", plus some missing information, "harmonizing".
"Doubling" and "harmonizing" both have two different meanings. In musical terminology, "doubling" means to double the line, either two different musicians singing/playing the same line or the same musician playing it twice (multi-tracked). Same with "harmonizing" either two different musicians playing the same line but one of them shifted harmonically (say by a 3rd, 5th or an octave) or again, one musician playing it twice (multi-tracked). In recording terminology, it's effectively the same thing except there's only one musician and only one performance/recorded track. That "track" is duplicated, shifted a little in time and pitch and then both played together. This is "doubling" and now we have one voice that sounds like two voices, although to be honest it's often used with such a small shift in time and pitch that it still sounds like one voice but a "phatter" voice. Incidentally, this studio "doubling" was invented by Sir George Martin (and Geoff Emerick) on the Sgt. Pepper album. "Harmonizing" is effectively the same thing, one musician and one recorded track which is then duplicated, shifted in time and pitch shifted by a harmonically related amount (say a 3rd, 5th or octave). In the 1970's you didn't even have to physically do this duplication, you just routed your track through a "Doubler" or "Harmonizer" effects unit. In the mid 1970's a company called Eventide released their "Harmonizer" (the first harmonizer effects unit, and by far the best). Within a few years Eventide's Harmonizer became a required/ubiquitous effect unit in pretty much every commercial pop/rock studio on the planet, was used by pretty much every music producer on a large percentage of all commercial music releases and it (or it's plugin equivalent) is still used fairly widely today. In fact, the studio term "Harmonizer" (or to "harmonize") would generally be taken to specifically mean Eventide Harmonizer.
Getting back to the question then: It could be just one voice/track that's been put through Eventide's Harmonizer, so that it sounds like 2 voices and in some places about 5-6 voices. Or, it could be Gabriel multi-tracking himself several times. My guess is that it's a bit of both. I think where there's two parts (with one part an octave lower than the main line) that's been achieved with one track that's been "harmonized" (an octave lower) and later, where there's several voices, we've got the main line, another track that Gabriel's sung higher and then a harmonizer used to create the 5-6 lines/voices. Incidentally, Gabriel was always at the forefront of recording technology and employing it innovatively/creatively in his productions, although the Harmonizer wasn't particularly new technology when "So" was produced. When I was in his studio ("Real World Studios") in the early 1990's, he had two Eventide Harmonizers if I remember correctly.
G
