Prenominal adjective order preferences in Chinese and German L2 English

Abstract

This study presents a contrastive analysis of 3624 instances of prenominal adjective order retrieved from the Chinese and German sections of the International Corpus of Learner English and the International Corpus of English. The data was annotated for nine determinants of adjective order, including semantic, frequency-related, and articulatory features. Applying a two-step regression procedure called MuPDAR (Multifactorial Prediction and Deviation Analysis Using Regressions), the present study finds that overall, the intermediate-advanced level learners are well-aligned with native speakers' preferences. However, we also see that while the German learners seem generally better aligned with regard to frequency-related factors, the Chinese learners behave more target-like with regard to the effect of adjective gradability, and they seem more sensitive to segmental alternation constraints. In discussing these and other results, the study hopes to illustrate how corpus-based methods can make a valuable contribution to contemporary SLA research, specifically with regard to multifactorially determined phenomena such as adjective order.

Particle Placement in Learner Language

“…Much as speakers prefer to alternate stressed and unstressed syllables, there seems to be a universal tendency to alternate consonant and vowels wherever possible (Couper‐Kuhlen, ; Gries, , ; Gries & Wulff, ; Schlüter, ; Wulff & Gries, ). Applying segment alternation to particle placement, one can hypothesize that Example 7a should be preferred to Example 7b because climbed a pine tree up contains a consonant–vowel alternation at the boundary between the verb and the direct object, and a vowel–vowel sequence at the boundary between the direct object and the particle, but climbed up a pine tree contains consonant–vowel alternations at both junctures.…”

Profiling verb complementation constructions across New Englishes

“…To sum up, in this paper we are extending Gries and colleagues' (Gries & Adelman 2014, Gries & Deshors 2014, Gries & Bernaisch 2016, Wulff & Gries 2015 MuPDAR approach from one based on (two) regressions to one based on two random forests (MuPDARF, with RF standing for random forests). Specifically, we: i. do a random forests analysis on only the native BrE and AmE speakers and test whether its fit is good enough to proceed; this analysis uses COMPLPATTERN as the dependent variable and the following as predictors: FINITEMATRIX, VOICEMATRIX, NEGMATRIX, OBJECTFORM, VERBSEMCOMP, VERBTYPECOMP, VERBSEMMATRIX, VERBTYPEMATRIX, and COUNTRY (for the variety); ii.…”

“…In the best of cases, both of these regressions involve mixed-effects models (e.g. Gries & Adelman 2014, Gries & Deshors 2015, Wulff & Gries 2015. 1 In this paper, we apply a variant of the MuPDAR approach to an alternation which has not been targeted much outside of native English, the to vs ing alternation in Example (1), and whose study, if not targeting native speakers, has not been corpus-based, has focused mostly on lexically-specific preferences of the two alternating patterns and has rarely included other predictors (with the exception of Deshors 2015 and Khamis 2015).…”

There’s more to alternations than the main diagonal of a 2×2 confusion matrix: Improvements of MuPDAR and other classificatory alternation studies

“…the coefficient that indicates how much a variable (level or increase) changes the predicted outcome) may be significant even if it leads to the same predictions for NS and NNS alike. 1 MuPDAR has led to many interesting results, including Gries and Deshors (2014) on may vs. can, Gries and Adelman (2014) on subject realization vs. omission in Japanese, Wulff and Gries (2015) on prenominal adjective order in English, Deshors and Gries (2016) and Kolbe-Hanna and Baldus on ing vs. tocomplements, Heller, Bernaisch and Gries (2017) on the genitive alternation in British vs. Singaporean English, Wulff and Gries (2019) on particle placement, Wulff and Gries (to appear) on genitives in learner data, Kruger and De Sutter (2018) and Lester (2019) on that-omission, Werner, Fuchs, and Götz (to appear) on present perfect vs. simple past choices, etc. However, in this paper we want to discuss two areas in which we see room for improvement of both MuPDAR in particular and general regression/classifier approaches in corpus linguistics in general.…”