phrases - Is it correct to say Person A is the "spitting image" or the "splitting image" of Person B?

I understand that when trying to describe a person who has a resemblance to another, the common term is spitting image. As in:

Person A is a spitting image of Person B.

Here's my issue, I've recently heard some people saying splitting image as opposed to spitting image and upon thinking about it, splitting image makes more sense. That may be because I am unaware of the etymology of spitting image.

My thinking is that Person A looks so much like Person B, that they have appeared to "split" from the same person, hence splitting image. Wouldn't you agree?

Answer

To expand a little on Claudiu’s excellent answer, there seems to be an interesting progression/evolution here:

• 
  

  metaphor: “it’s like he was spat out of his father’s mouth” (1689).

  
  


• 
  

  metonymy: “he’s the very spit of his father” (1825) — when the metaphor is commonplace enough, it no longer gets spelled out in full.

  
  


• 
  

  idiom/cliché: “the spit and image of his father” (1859) — a particularly effective wording of the metonymy solidifies into a widely re-used phrase.

  
  


• 
  

  corruption: “the spitten image” (1878) — the original analysis of the phrase is lost.

  
  


• 
  

  reanalysis: “the spitting image” (1901) — this strange new word “spitten” gets replaced by something which is at least syntactically more comprehensible.

  
  


• 
  

  further reanalysis/eggcorning: “the splitting image” (1880(!?), 1939) — the phrase changes to something which is more semantically plausible — it’s easier to imagine ways that “splitting image” could have arisen than “spitting image”.

  
  

(NB: my examples are not quotations; these dates given are when each form seems to have arisen)

Huh: so in chasing up the dates of all these forms (in the OED), I got a surprise! It turns out that while “splitting image” comes in 1939, postdating “spitting image” as we’d expect, the form “splitten image” appears in 1880, in Specimens of Westmorland Dialect: “Soa t’kersmas up i’t’fells Et just be t’splitten image Ov a kersmas ’mang yersells.” This is only just after the first citation of “spitten” (1878, in “spitten picter”), and well before the first citation for “spitten image” (1901)!

Whats going on here? Written language tends to lag behind spoken language by quite a long time, so it could well be that actually in 1878 the progression had gone further than the cited sources go, and so “spitten” had been actually already been around for a while before getting reanalysed to “splitten” — and it just so happens that “splitten image” struck it lucky with an early dialect-collector and got recorded more promptly than most coinages do. Or perhaps it had gotten reanalysed at an earlier stage, “spit and image” → “split and image”, and then corrupted “split and” → “splitten” in parallel with the “spit” forms? Or maybe even the “splitting image” proponents we laughed at were right all along — the Westmorland folk coined this form entirely separately, while splitting their firewood for Christmas up in the fells?

My guess is that it’s something like the first option: “splitten” does indeed follow “spitten”, and it’s just chance that it got recorded remarkably early. But this isn’t a terribly authoritative guess… anyone better-educated in historical linguistics have any thoughts on this?