Ah, that's not a statement you want taken out of context...
But yes, I think too much logic often gets embedded into the view. I always ask myself "would the result of this condition change between..."
- Different visual layouts.
- Rendering on different devices.
- Rendering in different languages or locales (including e.g. right-to-left languages).
- Rendering for users with different accessibility needs.
If the answer to any of the questions is "yes", I think it's at least potentially sensible to put the logic in the view. If all the answers are "no" then I think it's pretty unequivocal that it should go into the model.
Of course, issues like full i18n are tricky and often still require model input anyway (even if it's just to select one of a set of views based on the locale). But I find thinking in those terms tends to focus me away from what's easy to implement towards what's the cleanest interface.