I'm at school, so I can access the Science journal article I mentioned before. Phillips-Silver and Trainor have made a good experimental design for exploring the effects of movement on the development of rhythm perception. However I do find fault with their training stimulus. All the babies listened to the "ambiguous" beat pattern while being bounced every two beats or every three beats. You can listen to the beat pattern here, and see the pattern in Part A of the figure shown below.
Phillips-Silver, J. and L.J. Trainor. "Feeling the Beat: Movement Influences Infant Rhythm Perception." Science, Vol. 308, 5727, 1430, 3 June 2005.
I do not think this pattern can be heard as either duple or triple, and therefore ambiguous. Rather, this pattern is clearly a duple pattern, given the weak beats on the 2nd and 6th beats. Because beat 2 is much weaker than beat 3, (beat 3 being just as strong as beat 1), beat 3 sounds like a new downbeat. The rest of the pattern confirms this, with beat 5 much stronger than beat 6. This paper by Justin London illustrates my claim, as do the cited articles by Mari Riess Jones. Thus I think there is a bias towards the duple pattern, which could have masked the bouncing effect somewhat. The babies trained in the triple pattern may not have felt as strong an affinity for the congruent condition, weakening the difference between the two conditions.
For fun, the other two beat patterns are found here and here. The addition of dynamic accents do intensify the duple feel in the second pattern, and do make the triple feel possible in the third pattern. But absent of these strong accents, the timbral accents take over to create the unambiguous meter of the training stimulus.