We use multiple lines of evidence.
When scientists talk about the past, we can't directly observe what happened, any more than we can directly observe the building of the Pyramids. So we reconstruct history from the various lines of evidence that we have available.

In this case, we use evidence from phylogeny (relationships) and from functional morphology (how the parts work) as well as from aerodynamics, to help us construct explanations that seem to account best for all the evidence that we observe.

We use evolutionary relationships.
One big line of evidence you've already seen is the evolution of the group: their relationships to other animals that they evolved from. Because birds share so many features with theropod dinosaurs that no other animals have, we think it's a reasonable conclusion that birds evolved from this group.

If we look at that hypothesis of relationships, which we call a cladogram, we see how all the birds and various theropods are related to each other. But we also see the specific features that these animals share, and this is what our cladogram is based on. Looking at these features more carefully, we notice that different features appear at different points in the evolutionary picture. Each of these features helps us to establish relationships at different levels in the cladogram.

We use functional morphology.
The main question is: what's flight, and how do you do it? You've already seen what it takes to fly: you need a good wing, or airfoil; you need the muscle power to move it; and you need to be able to perform a stroke with this airfoil that will get you up in the air and keep you there.

This is really the central problem of flight: how do you evolve this flight stroke? It doesn't just come from nowhere. Here's where the cladogram is important. Once we identify the features that we need in order to fly, we can see just where on the cladogram they appeared. And, because they didn't all appear at once, we can ask what they were doing for the animals in the first place, before they were used in flight.

We think about what the features were used for, if not for flight.
When we first find feathers, they're not in birds, but in a larger group of theropods called coelurosaurs. They start out being filamentous (as in the fossil image at right), and later they diversify into many different feather types. As you've seen, there are several feathered dinosaurs that aren't birds and didn't fly. Their feathers were too short and too sparse to make a good airfoil. However, they could have been used in thermoregulation, or in camouflage, display, or species recognition. Evidence from nesting oviraptors even suggests that they might have sheltered the eggs during brooding season. It's hard to be sure, but we do know that the first feathers weren't used for flight.

We think about what the features were used for, if not for flight.
Some maniraptorans could move their arms in a motion very similar to the flight stroke, even if they couldn't fly. The sternum also becomes larger, and the wishbone becomes well developed, in these same maniraptorans. The wishbone anchors muscles that help to pull the arms forward. What were they doing with this movement? Why would a predatory dinosaur want to pull its arms forward? Catching prey is a good bet. That's not difficult to imagine.