Anticipation is the problem, because overreliance implies there is some level of reliance where anticipation is helpful. If you want a window into how fluent people experience the process of speaking in regards to this, grab a text you haven't read before and start reading out loud. Even if you do not stutter when alone, you will inevitably stumble at one point at some complicated tongue-twister of a word or sequence as all people do. You then correct the error on the go and keep reading as usual. In contrast, read a sentence or paragraph first in your mind through your inner speech. Silently observe where you might stumble, and then start reading out loud. Now, what I consider to be the problem is that in the first scenario, you stumble on the word you did not anticipate and keep reading. There is negligible spike activity on the part you stumble but no associative synaptic plasticity there for long-term potentiation of the synapse to occur. Long-term depression occurs instead. You will attach no particular significance to pronunciation of that part for future instances, and thus no anticipatory feedback. In the latter scenario, the initial predictive/anticipatory inner speech model makes the prediction on where you might stumble based on the past. If you indeed stumble, there is now some associative synaptic plasticity in the form of long-term potentiation because you made a successful prediction, even though it needn't be there in first place and is counter-intuitive to the process of speaking. To note, spike-timing dependent plasticity, LTP, LTD, act at the level of the neuron with signals milliseconds apart, so this explanation is not completely literal. It's just to illustrate the dynamic. Ultimately, I would argue anticipation is the problem because developmental stuttering in itself is a result of associative synaptic plasticity. With each successful anticipation we make, the fucking synapse further strengthens, and the anticipatory feedback messes with the executive motor function. If we forgot we stutter and fully stopped anticipation, the synaptic connections that lead us to stuttering would undergo long-term depression and voilà. 😂 I could get more technical to explain neural structures and dynamics that lend credence to this explanation, but this comment is already too long.