Concept of the oscillator

For a circuit to oscillate, the gain must be high, the feedback must be positive, and the coupling from output to input must be good. The feedback path must be easy for a signal to follow. The phase of a fed-back signal can be reversed without any trouble, so that common-emitter or common-source amplifiers can be made to oscillate.

Feedback at a single frequency

Recalling the public-address fiasco, some variation of which you’ve doubtless heard many times, could you know in advance whether the feedback would have a low pitch, a midrange pitch, or a high pitch? No. The oscillation was not intended, and might have started at any audio frequency.
The frequency of an oscillator is controlled by means of tuned, or resonant, circuits. These are usually inductance-capacitance (LC) or resistance-capacitance (RC) combinations. The LC scheme is common at RF; the RC method is more often used for audio oscillators.
The tuned circuit makes the feedback path easy for a signal to follow at one frequency, but hard to follow at all other frequencies (Fig. 1). The result is that the oscillation takes place at a predictable and stable frequency, determined by the inductance and capacitance or by the resistance and capacitance.

Fig. 1: Basic concept of the oscillator.