…if a speaker marks a delay with “um,” that delay will be longer than if it had been marked by “uh.” If “uh” is used, the delay that follows before resumption of fluent speech is roughly a quarter of a second; if “um” is used, the anticipated delay approaches three-quarters of a second.
In one case (um) the delay is inserted as “a deliberate signal”, in the other (uh) it is “an unavoidable effect of having processing problems.”
In male speech, either um or uh accounts for about one out of fifty words; for female speech it is only one out of seventy words.
One study found that over 40 percent of transitions — the switch in who is talking — have gaps of within a quarter-second of zero. In another similar study, the two parties are talking at the same time only 3.8 percent of the time, and then usually not for very long.
So says N.J. Enfield, in his new book How We Talk: The Inner Workings of Conversation.