How speakers adapt object descriptions to listeners under load

Figures & data

Figure 1. Higher resolution version of an example stimulus with minimal description length 2 to describe image #4.

Figure 2. The driving simulator used for this experiment. Speaker-passengers described a target image from the image array displayed above the road. Listener-drivers had to follow the yellow bar back and forth across the road while identifying the intended referent by number.

Figure 3. Schematic representation of the experimental design. The order of the blocks was counterbalanced across participants.

Figure 4. Degree of overspecification in speakers who described first (left) and speakers who had driven first (right) by driving condition and block. For an individual participant pair, if Block 1 had the easy driving condition, Block 2 had the difficult driving condition and vice versa, as indicated by the swapping of the colours in Block 2.

Table 1. Cumulative link mixed effect model output for the degree of overspecification. Threshold coefficients are the intercepts for each of the response categories.

	β	SE	z	p
Threshold coefficients
0|1	−0.4816	0.2286	−2.1068	0.0351
1|2	3.7023	0.255	14.5162	<.001***
Coefficients
Driving condition	0.0416	0.1512	0.2753	0.7831
Role order	0.6606	0.4524	1.4603	0.1442
Minimal number of required attributes	−2.1248	0.3434	−6.1875	<.001***
Driving condition : Role order	−0.4258	0.3028	−1.4064	0.1596

Table 2. Cumulative link mixed effect model output for the degree of overspecification with condition order as main predictor. Threshold coefficients are the intercepts for each of the response categories.

	β	SE	z	p
Threshold coefficients
0|1	−0.4689	0.2202	−2.1298	0.0332
1|2	3.7144	0.2477	14.9977	<.001***
Coefficients
Condition order	0.5402	0.3000	1.8008	0.0717
Role order	0.6347	0.4354	1.4579	0.1449
Minimal number of required attributes	−2.1240	0.3432	−6.1887	<.001***
Condition	0.0417	0.1543	0.2704	0.7868
Condition order : Role order	1.3520	0.6000	2.2536	0.0242*

Table 3. Linear mixed effect model output for log modifier speech rate (overspecified descriptions only).

	β	SE	t	p
(Intercept)	1.4526	0.034	42.7242	<.001***
Modifier redundancy	0.0276	0.0343	0.8064	0.4223
Condition	0.0490	0.0159	3.0769	0.0021**
Role order	−0.0025	0.0675	−0.0374	0.9703
Modifier redundancy : Condition	0.0301	0.0318	0.9466	0.3440
Modifier redundancy : Role order	−0.0107	0.0684	−0.1560	0.8764
Condition : Role order	0.0661	0.0321	2.0601	0.0395*
Modifier redundancy : Condition : Role order	0.0657	0.0641	1.0248	0.3056

Table 4. Linear mixed effect model output for log speech onset latency.

	β	SE	t	p
(Intercept)	0.8590	0.0366	23.4558	<.001***
Condition order	−0.0132	0.0722	−0.1824	0.8561
Role order	−0.0877	0.0728	−1.2036	0.2349
Overspecified vs. minimally specified	0.0311	0.0138	2.2568	0.0252*
Overspecified by 2 vs. 1 attributes	−0.0027	0.0209	−0.1306	0.8965
Condition	−0.0049	0.0153	−0.3227	0.7486
Condition order : Role order	−0.2705	0.1443	−1.8750	0.0674