Sparrows

Surprisingly this behavour is not innate. This article should help you understand this question (attached)

Bird songs of related species are predicted to diverge in sympatry to reduce species recognition errors when hybrids have lower fitness...

Some reasons why

The four most important ones are:

(i) The habitat a species occupies. The physical characteristics of the habitat should select for particular frequencyspectra in the songs, such that the spectra of arboreal species should be predictably different from terrestrial species due to frequency-dependent attenuation of soundover distance

(ii) The number and kinds of signals in itsrepertoire; the greater the number of signals in its repertoire, the more difficult it will beto modify calling signal in a given direction,since all of the songs are functionally, if notstructurally linked, and a change in one willnecessitate a change in the others.

(iii) Thenature of the minimal information carryingunits its signals.

(iv) The particular number and sort of acoustical neighbors.

For info can be found in http://entnemdept.ufl.edu/walker/buzz/k340lo92.pdf

Also check out: http://www.jstor.org/pss/4532867

Some analogies are the different accents in humans.