I am not entirely sure what you mean by canonical forms studies but it seems very unlikely that they are possible with gaussian. Several canonical forms or resonance structures are a way to depict the same delocalized state using localized representations in written form. They are used if the state of the molecule can't be represented by a single Lewis formula. Quantum mechanically they don't exist. Any SCF method, if Gaussian or Gamess or Mopac or any other software, will produce the delocalized state.
It might be possible to study different resonance structures using a force field approach. But I am no expert on that and it still sounds not very useful.
If you could explain why you would do these studies it would be easier to find a solution to your problem.
What you are understanding under canonical forms, really?.
If you are discussing over 2nd derivative numerical solution, Than, such as solution it is possible to do this with Gaussian, employing semi empirical levels of the theory. Just, for example, studying the
vibrational frequencies, you should add Freq(Numer,Step= X) as shown below: