One idea is it might not necessarily be either/or -- I'd suggest that you first release one version, whichever payment model your gut is pulling you toward. Wait until you have some data, maybe 4-6 months of sales history. If sales are gangbusters, maybe you'll feel that you've already got the right sales model. If you have a hunch that sales could be better and that a different sales model may be a major driver of additional revenue, you can release another version under the opposite payment model, and observe what happens to overall sales.
Personally I've preferred the Paid model. I have one app in the store, and the main driver of revenue increase has come from delivering updates which give more value to the user, and increasing the price accordingly (monthly revenue is up 8x on average since a major update & a price increase from 2.99 to 8.99). I haven't taken my own advice above yet, though I have built the IAP version -- I haven't released it yet because so far other experiments have been effective in increasing sales and less risky. I recall Marco Arment mentioned a major source of revenue increase for Instapaper was when he killed the free version. [1]
[1] http://www.marco.org/2011/04/28/removed-instapaper-free