
When you search for an image in Google the image spider(s) looks at the text before and after the images on webpages, and it uses these to get its listings. It is estimated that the amount of words before or after is somewhere between twenty and forty. However this all depends on how big your words are.
Of course many of you well educated search buffs will tell me that all the text on websites are spidered. That's true, but the text surrounding the images will have more influence.