In September 2008, archaeologists from the University of Tuebingen in Germany discovered an inch-long ivory carving of a nubile woman that by carbon-dating is on the order of thirty-five thousand years old (to be published in this week's Nature). After reveling in the impact of looking at something carved by cognitively active humans living that long ago, my next thought was whether sculpture pre-dates flat visual art.
The oldest sculptures we have are thousands of years older than than oldest visual art. The oldest cave drawings we have are from Chauvet, which are at most 30,000 years old. This is an outlier. We have over a hundred Venus-type sculpture artifacts older than that, dating from the period 25,000-29,000 B.C. (see Venus of Dolni Vestonice). This still isn't a huge sample, and maybe images don't age as well as sculpture - but more importantly, it's reasonable to expect that even older images and sculptures are both waiting to be found not just around the Alps and Pyrenees but throughout Central Africa.
Alternatively, maybe sculpture really did come first. This makes sense. To us today, drawing is easier than sculpture because, while we're are familiar with both, creating a drawing recognizably representing an animal is less time-, material-, and skill-intensive than creating a sculpture. Now that we have lightweight drawing mediums (nowadays including electrons!) drawings are also more portable. But to someone who has never seen either medium, a sculpture may seem less difficult to "get your head around". A woman is a three-dimensional thing; so is a piece of ivory. The leap to representing a woman (or horse) as a flat, untouchable image on a surface - as if they're trapped under ice - must have seemed a real abstraction.
Phonetic writing systems today offer another example of a mode of representation that is anything but obvious to those never before exposed to it. Ask any literate native Chinese speaker. Representing utterances phonetically rather than semantically - that is, using sound-based rather than meaning-based symbols - is so abstract, when though about from first principles, that it's amazing anyone thought of it at all. It's therefore not surprising that there was a gap of millennia between the idea of regular correspondence between visual symbols for words and phonetically defined ones, although the non-obviousness of alphabets is difficult for modern literate Westerners or Middle Easterners to fully appreciate today.
The benefit of phonetic systems, once you invent them, is that they are much easier to learn - hence why in Japanese book stores I go to the children's section and read about cats and bears in the 42 basic phonetic characters of Hiragana, but I'm hopeless if I try to read the grown-up prose in yesterday's Asahi Shimbun. Learning those 42 basic characters in Hiragana are on the same order of difficulty as the 26 in the English Roman alphabet and 33 in Cyrillic, but wholly different from the number (and visual complexity) of the 3,000+ you need to for the newspaper. It is also not surprising that as adults, it's more likely for monoliterate ideogram-readers to learn an alphabet than it is for monoliterate alphabet-readers to learn an ideogram system. While there may be some influence behind this asymmetry from the current realities of history and economics, I doubt it will change very much. Similarly, now that we have drawing in addition to sculpture, we demonstrate a preference for it - because it's easier and cheaper.
So did sculpture pre-date image? Carbon-dating of the available evidence supports this so far, and it's consistent with the development of human representation systems for which we have more direct evidence. I expect that there are more caves around the Pyrenees for us to discover with paleolithic artifacts. Chauvet Cave was only found in 1994. I predict that some of these caves will yield more carved figures that are older than any known drawings.