Pauses and linguistic structure

Since April, I've been spending my sabbatical leave at the Spoken Language Systems Laboratory (L2F) at INESC-ID in Lisbon, Portugal. One of my projects here has been to look at the signalling effect of silent versus filled pauses to listeners. This is building on work done by Karl Bailey and Fernanda Ferrreira (2003). Their study (which is actually Karl's doctoral dissertation) was a series of investigations to explore how filler disfluencies (specifically uh) influence the way that listeners process incoming speech. They asked participants to listen to sentences in which a filled pause disfluency was placed at (or very near to) a clause boundary versus when the disfluency was placed at a non-boundary (i.e., inside a clause), as follows.

1. While the man hunted the uh uh deer ran into the woods.
2. While the man hunted the deer uh uh ran into the woods.

In (1), the disfluency occurs at a clause boundary. As such, it interrupts the flow of speech in a manner that is consistent with the syntactic structure of the sentence. In (2), on the other hand, the disfluency occurs at a location which may lead the listener to believe that it is the closure of a clause but, in fact, is not. Hence, the prediction is that sentences like (2) are more likely to confuse listeners than those like (1).

Their experimental task was a relatively simple one: Participants listened to each sentence and then judged whether it was grammatical or not—a simple binary choice. The results were actually quite clear: Participants judged (2) sentences as grammatical less often than they did (1) sentences.

So, my interest was to extend this study to look at two factors more deeply. First, whether silent pauses could evoke the same effect. This would make sense if the explanation for the effect is that the additional time afforded by the pauses (filled or silent) is what enables the listener to make the judgments they make.

Second, I wanted to see whether the effect could replicated with nonnative listeners. Bailey and Ferreira looked only at native speakers and listeners. Different languages may have quite different patterns of filled and silent pause use, so this could affect how they are process these pauses when listening to English speech.

So, as a first step in this study, I extended Bailey and Ferreira's stimulus set to include 90 items (they used 50). I also simplified some of their original stimuli because I feared that some of the vocabulary or US cultural knowledge might be too obscure for nonnative listeners in other places. Then, I inserted filled pauses or silent pauses in the clause boundary or non-boundary positions. Otherwise, I used exactly the same grammaticality judgment task.

Thanks to the Linguistics faculty at University of Lisbon, I was able to recruit almost 40 students to participate in the experiment. The results are really nice. They perform just like the native English listeners in Bailey and Ferreira's study with the filled pause stimuli: non-boundary cases are judged grammatical less often.

But the really interesting results come with the silent pauses. Basically, they pattern the same as the filled pauses but in a more distinct manner. When a silent pause occurs, they really expect a clause boundary there. Thus, the non-boundary case (i.e., bad condition) is really bad for them and the boundary case (i.e., good condition) is really good for them. So, the silent pause actually played a facilitative role in helping listeners understand the sentence. Bailey and Ferrerira did not have this result (when comparing to a baseline no pause condition).

Fig. 1. Results (n=30): Proportion judged grammatical

So, the results overall seem to suggest that while both silent and filled pauses are used by listeners as signals of sentence structure, silent pauses are stronger signals than are filled pauses.

Of course, this isn't the end of the story. Next, I plan to look at nonnative listeners of another language (French), and then native English speakers (to replicate Bailey and Ferreira), and then finally native speakers of Japanese (after I get back to Japan at the end of my sabbatical).

[Note: This post was written in September, 2020. However, in order to preserve the chronology of the blog, it has been dated to reflect when the described events actually took place.]