What speaking under load actually costs

The first false picture to remove is that fluent people think clearly and hesitant people do not.

Speech production runs in stages. You decide what to say, which the field calls conceptualisation. You turn that into words and grammar, which is formulation. You produce sound, which is articulation (Levelt, 1989). The staged architecture is one of the better-supported models in psycholinguistics, resting on converging evidence from speech errors, timing studies and lesion data over thirty-five years.

The consequence that matters here is that load at the first stage shows up at the third. When you are still deciding what you think, the hesitation appears in your mouth.

There is a clean demonstration of it. In a corpus of 48 pairs, 96 speakers and about 192,000 words, participants completed a task where one person had to describe a set of figures and the other had to identify them. Same conversation, same people, but one role carried the burden of generating and structuring content and the other did not.

The director, doing the planning, produced 7.00 disfluencies per hundred words, of which 3.30 were fillers. The matcher, responding, produced 4.93, of which 1.81 were fillers. Roughly 42 per cent more hesitation on the side that was thinking (Bortfeld, Leon, Bloom, Schober & Brennan, 2001).

Two things follow, and the second one reorganises the whole course.

The first is that your disfluency rate is not primarily a fact about your confidence. It is a fact about how much conceptual work you are doing while your mouth is running. The same person, on the same day, in the same room, produces markedly different rates depending on whether they are generating or responding. If you hesitate more in meetings where you are asked to assess something than in meetings where you are reporting something, that is not a nerve problem.

The second is a resource-allocation argument. Most communication training targets articulation, because articulation is what an observer can see. But the load originates two stages upstream. Working on your voice while your conceptual stage is unmanaged is polishing the output of a process that is starving.

This is also the mechanism behind the one thing in this level that transfers reliably from the language-learning literature: pre-task planning improves fluency. That work is on second-language production rather than professional speaking, so the transfer is an assumption rather than a finding, and it is stated as an assumption here. But the direction is exactly what the staged model predicts. Do the conceptual work before the articulation starts and the articulation improves without being touched.

One caution about the numbers, because you will use them in Exercise 0.A. The corpus figures come from a referential communication task, not from a board meeting, and disfluency rates vary with age, gender, role and topic within the same study. Your own figure is a comparison against yourself twelve weeks later, not against 5.97.