GARLI was used to search for the ML tree and to conduct likelihood bootstrap analyses. All analyses of gap data assumed equal state frequencies and a four-category discrete approximation to the īGLVWULEXWLRQZLWKWKHVKDSHSDUDPHWHUHVWLPDWHGIURPWKHGDWD7KLVFRUUHVSRQGVWRWKH&)1Yī PRGHOIRUDQDOVHVZLWKDGGHGLQYDULDQWFKDUDFWHUVRUWKH&)1īPRGHOIRr analyses without added invariant characters). Up to 200 searches were conducted in GARLI to evaluate the ability of that program to identify the ML tree.