Another thing that can affect the way your brain interprets the bump map is the underlying texture colors. The problem with veins is that they're usually darker than the surrounding skin and that tends to encourage the brain to see them as depressions (though not always). Could you give us some screen shots of the successful and unsuccessful veins? Might give us some ideas on how to change the visual context. My first guess is to make sure there's a gradient across the width of the vein, so that the center is noticeably lighter than both edges -- that might provide a good depth cue. But what you're noticing, as heyas says, is standard bump map behavior -- to your brain it sometimes looks protruding, sometimes indenting, in itself; all you can do to get the illusion working in the right direction is fiddle with the context, light and color.