One brought the tools of computational geometric morphometrics — an automated algorithm trained on CT scans of living French adults, producing a mathematically consistent face from a mathematically described skull.

Each of them was doing their best.

Each of them was working from the same evidence.

Each of them was constrained by the same skull, the same genome, the same scientific understanding of how skulls and faces relate to each other.

And they produced three different faces.

The differences between those faces tell us more than any single face could have told us.

They tell us where the evidence is strong, because that is where the faces agree.