Web Activity 13.2 Mumbling, male nurses, and SVO order

At various points in the textbook, you have come across the idea that speakers selectively omit or reduce linguistic information in just those contexts where the information can most easily be recovered or inferred by the listener. For example:

• Pronouns tend to occur in contexts where there is one highly salient antecedent (discussed in Chapter 11).
• People are most likely to pronounce words with reduced phonetic information (that is, mumble) when the words are common ones or are highly predictable in the context (discussed in Chapter 12).
• People are most likely to leave off information about the details of an event or an entity if that information is highly predictable. For example, they are less likely to specify the instrument if it is one that is commonly used for that action (e.g., stab with a knife versus stab with an ice pick). They are also less likely to specify a property that is highly predictable for that entity (e.g., male firefighter versus male nurse, or yellow banana versus yellow sweater; discussed in Chapters 11 and 12).
• Case marking appears least often in those contexts where it would be easiest to infer the participant’s role in the sentence (differential case marking, discussed in Chapter 13, Section 13.2).

Some researchers have argued that speakers have a general knack for predicting (not necessarily consciously) how much trouble listeners might have in recovering meaning from a linguistic code that has been reduced or corrupted. They claim that speakers adapt their speech accordingly, and that these adaptations contribute to language universals.

Ted Gibson and colleagues (2013) have invoked this argument to explain the prevalence of certain word orders over others. Step-by-step, their logic goes like this:

1. Production constraints help to explain why it is far more common to see word orders in which the subject comes before the object (as discussed on pp. 487–490), but they cannot explain why SOV and SVO orders are so much more common than VSO orders.
2. When creating a linguistic system from scratch, people seem to gravitate toward SOV as a default word order. This is seen in newly-emerging languages like NSL or ABSL, as well as in research on gesturing.
3. The default SOV order can be overridden by speakers’ tendencies to insert a verb between the subject and object in order to preserve important information in the event that the linguistic signal gets corrupted in some way. Here’s the idea: Suppose you utter an SOV sentence like the girl the boy kicks (which means that the girl did the kicking). And suppose that noise in the environment prevented me from hearing both noun phrases, so that I only heard the girl kicks or the boy kicks. Since both the subject and object come before the verb, I have no way of knowing who did the kicking. But now, suppose you uttered an SVO sentence like the girl kicks the boy, and one of the noun phrases is lost to noise, so that I hear the girl kicks. I may not know who or what the girl kicked, but I know that she is the one doing the kicking. Similarly, if I hear kicks the boy, I do not know who kicked the boy, but I do know what role he plays in the kicking event. Therefore, more information is preserved under noise with the SVO order.

Are speakers really sensitive to these possibilities? In order to find out, Gibson and colleagues devised an experiment in which people watched videos of simple events, and then communicated their content through verbal description and then gesture. The events either involved two animate participants (a fireman kicking a girl) or an inanimate direct object (a rollerskater kicking a ball). The authors reasoned that when the event involved one inanimate participant, the pressure to avoid information loss by inserting a verb between the subject and object would be less severe because it would be easy for the addressee to infer which participant was doing what (the ball could not be doing the kicking). Therefore, when using gestures, they should be more likely to default to the SOV order found in previous studies. But for events involving two animate entities, they should be more likely to use the information-preserving SVO order.

Conduct a collaborative class experiment to see whether this pattern holds in people that you test. As a group, come up with a set of simple events involving two participants. Half of these should involve two animate entities, and the other half should involve one animate instigator performing an action on an inanimate object. Try to set up your items so that the same verb appears in both versions. Now have each student take responsibility for testing three people. The student should create a simple video of the event and ask each of their three subjects individually to convey the event by means of gesture. The order of gestures (SVO, SOV, etc.) should be noted. Do you find that in general, people are more likely to produce the SVO order for sentences in which both participants are animate? Compare your results with those found by Gibson et al., and discuss.

Reference

Gibson, E., Piantadosi, S. T., Brink, K., Bergen, L., Lim, E., & Saxe, R. (2013) A noisy-channel account of crosslinguistic word-order variation. Psychological Science, 24, 1079–188.