One of the most important questions is of course whether these results are produced by selection effects. This is not likely because most selection effects are independent of spectral index. Two effects are, however, worth mentioning.
If source structures become more nonuniform with steeper spectra (e.g. the flux ratios of double sources become more unequal) it is possible (i) to miss identifications for steeper spectrum sources and (ii) measure smaller angular sizes for these sources. This effect has been examined for the 3C sample (for which there is good structural information available) and this effect has not been seen.
A second selection effect can occur when spectral indices are calculated for
very large sources. The uncertainty in the spectral index can be large enough
for these sources to misplace them in another spectral group (but not more than
one). There are only 4 sources in the 3C sample with angular sizes larger than
. If these sources were misplaced because of this then the total
influence of this effect on the general behaviour of spectral index dependent
properties shown in this paragraph would would be negligible.