Rabbit ears: if you start with two identical resonant circuits that are weakly coupled, the response is narrower than that of a single circuit. When the coupling increases, the response goes from a single peak to a double-humped curve, with a local minimum at the original center frequency.
More interesting: start with two similar individually tuned circuits, and tune no. 1 to a lower frequency than no. 2. With coupling, you will see two peaks very close to the original resonant frequencies. Now, start slowly tuning no. 1 towards the frequency of no. 2. You will see the lower peak increase in frequency, but not merge with no. 2. Instead, as no. 1 approaches the second frequency, the higher-frequency peak starts moving away from the lower peak (the two peaks in your rabbit-ear curve).
About coupled resonant circuits, there are simulators that can attach sliders to components values, so to drag the sliders while live watching the circuit response, and thus to build an intuition.
The behavior of coupled resonators was quite an epiphany to me when TimFox (thank you!) told me something similar (some years ago, in a very different topic about an array of high Q resonators). In fact, I was so impressed about the response of the coupled resonators that I even saved on YT this 1 minute video, about how the two peaks "push" each other when only one of the resonators is changed:
Line splitting in two coupled LC resonators and moving peaks - QucsStudio
This video might not cover all the aspects of coupled resonators behavior, so I highly encourage everybody to try the steps described by TimFox. The live sliders are such a helpful feature. Aside from QucsStudio, I've heard MicroCap (which is now free) can have sliders, too, but I don't know MicroCap. Both simulators are for Windows, but they can be run from Linux, too (through WineHQ).