The first question to ask is: am I going to do a ZORA calculation to include scalar relativistic effects? This is recommendable for systems containing heavy nuclei, but there is no objection to doing a ZORA calculation for a system with light atoms only, as the required CPU time does not increase very much if ZORA is used.
If you are doing a ZORA calculation, you will need the ZORA basis sets which can be found in $ADFHOME/atomicdata/ZORA. You may also use the all electron basis sets from the ET or AUG directory, but be aware that these were optimized to non-relativistic calculations. Currently the ZORA basis sets cover the entire periodic table and besides all electron basis sets offer a choice of frozen cores. At present the ZORA directory does not contain basis sets with very diffuse functions, which may be required in calculations for hyperpolarizabilities or high-lying excitation energies, but for the lighter elements (H-Kr) you can certainly use the all-electron basis sets from the ET or AUG directory. Warning: in a ZORA calculation use only the frozen core basis sets coming from the $ADFHOME/atomicdata/ZORA directory, or use all electron basis sets.
If you do not use ZORA, your basis sets should come from the directories SZ, DZ, DZP, TZP, TZ2P (old names I-V), or one of the ET or AUG basis sets. Alternatively, you could make your own new ET basis set using the utility programs described below. For many of the heavy elements only ZORA basis sets are available, but for such elements it would be inadvisable to do nonrelativistic calculations anyway. For light elements the ZORA and normal basis sets should be identical except for the description of the frozen core. Usually the ZORA basis sets contain much steeper basis and fit functions to accurately describe the core region.