Songbirds can be divided into closed-ended learners and open-ended learners. The former don't change their song once it is completely crystallized. Zebra finches for example have a fixed learning period and repeat the same song the rest of their lives. The latter can continue to change their song as long as they live. Common nightingales are a good example, they broaden their repertoire as they get older.
In a way, bird song is like human speech. There's an innate urge to produce sound, but both bird and human need a tutor to successfully sing and talk respectively. In zebra finches, the main model species for birdsong, the father is usually the tutor, though there's some influence from peers. Interestingly, female preference is also partly learned from a tutor or peers in zebra finches.
Because birdsong is a cultural trait, it can show substantial spatial variation. One can imagine that with large-scale changes in distribution and abundance, cultural traits like birdsong also change accordingly. If local populations disappear or individuals from different populations are introduced, local dialects can go extinct. If this happens fast enough and over the substantial parts of the distribution of the species, one can image a huge loss in cultural variation. The honeyeater is obviously an exceptional case, but I wouldn't be surprised if this happened in some form already to many species that have experienced range contractions.
