The second answer is correct, some problems need wave-treatment, others may be treated with the particle approach.
Let's make clear the specific of each case: in the quantum theory an object (which typically is microscopic) is described by a wave-packet.
In some experiments, and for some particles, this wave-packet is small (we say in this case that the particle is well-localized).
As long as we pass it through beam-splitters, but we just examine its passing through bubble chambers or ionization chambers or different materials, we may be able to treat it as a particle. Also, when the object is big, i.e. its linear dimensions >> wave-length, we can treat it as a classical object.
But if the wave-packet is sufficiently long, and we pass it through a beam-splitter, e.g. Mach-Zender beam-splitter, the two instanciations of the wave-packet meet on the screen at the exit a sufficiently long time for producing an interference pattern. For understanding why packet-length is important you need to read about coherence length.
Now, there is also another issue about particle vs. wave behavior. Even if the wave-packet is split into two, i.e. a transmitted and a reflected packet, when you place detectors on the transmitted path and on the reflected path, in only one of the two detectors will be produced a click. Only one of the two wave-packets gives a response. Why? The particle is one, we have a single particle, not two.
Do you understand how it is possible, why one wave-packet gives a response and the other doesn't? Don't you understand? Well, welcome to the club! Nobody understands! The quantum objects behave in a strange way.
And though, this is the answer, the particle is one. This is also a particle-like feature of the quantum object.