The origin of and evolutionary transitions among the extraordinary diverse forms of parental care in teleost fish remain largely unknown. The "safe harbor" hypothesis predicts that the evolution from a "guarding" to a "brooding" form of care in teleost fish is associated with shifts in reproductive and life-history features such as reduced fecundity, and increased egg volume with higher parental investment. Robust phylogenetic hypotheses may help to identify evolutionary changes in key traits associated with differences in the form of parental care. Here, we used reconstruction of ancestral character states to study the evolution of the two forms of parental care, bubble nesting and mouthbrooding in the fighting fish genus Betta. We also applied a comparative analysis using the phylogenetic generalized least-squares method to test the "safe harbor" hypothesis by evaluating differences between the two forms of parental care in standard length, life-history traits, and three habitat variables. Evolutionary hypotheses were derived from the first molecular phylogeny (nuclear and mitochondrial DNA sequence data; 4448 bp) of this speciose group. Ancestral character state reconstructions of the evolution of the form of parental care in the genus Betta, using the methods of unweighted parsimony and maximum likelihood, are uncertain and further indicate a high rate of evolutionary transitions. Applying different weights for the suspected directionality of changes, based on the consistent phenotypic and behavioral differences found between bubble nesters and mouthbrooders, recurrent origin of mouthbrooding in the genus Betta is favored using parsimony. Our comparative analyses further demonstrate that bubble nesters and mouthbrooders do not have a consistent set of life-history correlates. The form of parental care in Betta is correlated only with offspring size, with mouthbrooders having significantly bigger offspring than bubble nesters, but is not correlated with egg volume, clutch size, and broodcare duration, nor with any of the three habitat variables tested. Our results thus challenge the general predictions of the "safe harbor" hypothesis for the evolution of alternative brood care forms in the fighting fish genus Betta.