Referee 1's review:

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements.
**********

Review of the original submission:
This paper presents new data on molecular (mtDNA) relationships among freshwater herring and their marine relatives, with particular attention to the endemics of Lake Tanganyika. The data lead to the conclusion that the 3 Tanganyika endemics colonized the lake from an ancestral west African stock during the formation of the lake, rather than via a marine-freshwater connection as proposed earlier. The idea is good and the paper is very well written. With some caveats noted below, the analyses are sound and appropriate to the questions.

The sequence alignments for the 3 genes should be submitted as popsets to Genbank prior to publication. Were the rDNA sequences aligned in relation to a model of secondary structure? If not, they should be.

That the 3 genes produce significantly incongruous phylogenies is a bit of a red flag. Given that all are from the non-recombining mtDNA, this suggests inadequate models of sequence evolution. I recommend that the analyses be repeated using a partitioned analysis in Mr Bayes - ie. allowing for gene specfic models of sequence evolution. This could well affect the branch lengths of the concatenated analysis, and thus the estimates of divergence time that follow.

When presenting the resuts of the divergence time estimates (bottom p6), the confidence limits should be presented in the text, along with the point estimates. Table 2 can then be deleted. The last phrase of this section (end para 1, p7) is confusing as is, and should be re-written, eg. "the split between the two Lake Tanganyika endemics is consistent with a divergence during the time of formation of the lake (9-12 Mya)."

The discussion section is dominated by results concerning rampant polyphyly or paraphyly of currently described families, whereas the primary goal of the paper is to elucidate the history of the LT endemics. This seems rather imbalanced.

Review of the first revised manuscript:
I enjoyed re-reading the MS and the succinct response to previous reviews. In short, I think the authors have done a fine job of revising the MS to meet both my concerns and those of the second reviewer. I think it should now be accepted for publication.

I do have two minor editorial suggestions - the branch support statistics on Fig. 2 are all but invisible and need to be set at a larger font size. On Fig 3, I have no problem with emphasizing the point estimates of divergence time, but horizontal bars should be added to show the 95% confidence limits. This is important as cursory readers will jump straight to the figure without digesting the broad CL's specified in the text.