Morpho-Phonetic Effects in Speech Production: Modeling the Acoustic Duration of English Derived Words With Linear Discriminative Learning

Abstract

Recent evidence for the influence of morphological structure on the phonetic output goes unexplained by established models of speech production and by theories of the morphology-phonology interaction. Linear discriminative learning (LDL) is a recent computational approach in which such effects can be expected. We predict the acoustic duration of 4,530 English derivative tokens with the morphological functions DIS, NESS, LESS, ATION, and IZE in natural speech data by using predictors derived from a linear discriminative learning network. We find that the network is accurate in learning speech production and comprehension, and that the measures derived from it are successful in predicting duration. For example, words are lengthened when the semantic support of the word's predicted articulatory path is stronger. Importantly, differences between morphological categories emerge naturally from the network, even when no morphological information is provided. The results imply that morphological effects on duration can be explained without postulating theoretical units like the morpheme, and they provide further evidence that LDL is a promising alternative for modeling speech production.

Keywords: acoustic duration; derivation; linear discriminative learning; mental lexicon; morphological theory; speech production.

Figure 1

Comprehension and production mapping, adapted from Baayen et al. (2019b). For comprehension, transformation matrix F transforms the cue matrix C into the semantic matrix S. For production, transformation matrix G transforms the semantic matrix S into the cue matrix C.

Figure 2

Toy example of an articulatory path for the word lawless. Each connection between a triphone node is assigned a probability of being selected against other triphones.

Figure 3

Effects on duration difference in the standard linear regression models for the Idiosyncratic Network variables (left column), the Morphology Network variables (middle column) and the Base Network variables (right column).

Figure 4

Effects on duration difference in the mixed-effects regression models for the Idiosyncratic Network variables (left column), the Morphology Network variables (middle column) and the Base Network variables (right column). Red regression lines indicate significant effects from the final models, gray regression lines indicate non-significant effects from the initial models before the non-significant predictors were excluded.

Figure 5

Density distributions of variables by derivational function in the Idiosyncratic Network models (left column), the Morphology Network models (middle column), and the Base Network models (right column). Note that in the first two panels in the top tow, the density curves around 1.0 are calculated over a single value.

Figure 6

Type count of top 8 neighbors by network and morphological function.