Statistical learning of non-native morphology in adult learners : effects of L1 morphological richness, redundancy, and salience

Zhu, Liuqi and Monaghan, Padraic and Rebuschat, Patrick (2025) Statistical learning of non-native morphology in adult learners : effects of L1 morphological richness, redundancy, and salience. PhD thesis, Lancaster University.

Text (2025ZhuPhD)
Liuqi_Yuki_Zhu_PhD_thesis.pdf - Published Version
Available under License Creative Commons Attribution-NonCommercial-NoDerivs.
Download (2MB)

Abstract

A body of statistical learning studies supports the view that language development is influenced by learners’ sensitivity to the frequency and distributional properties of linguistic features in their input (see review: Saffran, 2020). Grammatical morphemes, based on this view, pose unique learning challenges: they are often acoustically subtle, redundant with other cues, and lack transparent form-meaning mappings, making it difficult to extract reliable patterns from the input (Ellis, 2022). Additionally, prior language experience may shape the ease with which morphological features are acquired (Van der Slik et al., 2019). While these factors have typically been identified through correlations with acquisition order (DeKeyser, 2005) or examined in isolation through tasks based on single sensory processing such as reading or auditory processing (Ellis & Sagarra, 2011), the current thesis explores their effects within a cross-situational statistical learning (CSL) paradigm, where audio-visual processing ability was examined. Learners track co-occurrent patterns between audio and visual cues, across multiple exposures, solving the referential ambiguity, and learned the meanings of linguistic forms indicated by co-occurring visual referents. To date, CSL has provided robust evidence for adults’ statistical learning ability to acquire novel words by keeping track of words and their visual referents (Yu & Smith, 2007; Ge et al., 2025). However, whether similar mechanisms can support the learning of more complex morphological cues in sentences remains less thoroughly investigated. To address the gaps outlined above, this thesis, across three studies, examines whether adults can acquire grammatical morphology in a novel language through exposure to sentence-level input under a CSL paradigm. In addition to this overarching goal, the thesis also explores how specific factors, namely L1 morphological background, cue redundancy, and perceptual salience, shape the success of morphological learning. Each of these factors is examined in a dedicated study, allowing for a systematic investigation of their individual and comparative influence on non-native morphological acquisition within a CSL framework. The key features of the CSL paradigm simulate the referential ambiguity problem typical in natural language learning. In each CSL trial, participants were exposed to two visual referents and an auditory sentence in an artificial language. The artificial language contained nouns and verbs, and morphologically marked tense and number, with a subjectverb agreement on number cues. Participants were asked to infer the referential meaning of the sentence by selecting the image that best matched it, thereby simulating the referential ambiguity problem typical in natural language learning. No explicit instruction or feedback was provided throughout the learning phase. Study 1 examined the influence of L1 morphological complexity on morphological learning. Participants were native speakers of Mandarin, English, and German - languages that vary in morphological richness. The results indicated that all groups were able to track statistical regularities and acquire both lexical and morphological patterns in the artificial language, demonstrating robust statistical learning ability. However, differences emerged in relation to L1 background: morphological learning outcomes were significantly higher for German-L1 speakers compared to Mandarin-L1 speakers, suggesting a possible facilitative effect of L1 morphological richness. Interestingly, the English-L1 group did not conform to this trend, showing learning performance that did not align neatly with their intermediate position on the morphological richness scale. This deviation may reflect the influence of additional factors such as prior non-native learning experience or overall linguistic proficiency, which potentially modulate the impact of L1 structure in CSL contexts. Study 2 investigated the effects of cue redundancy and availability. Based on the artificial language built in study 1, sentences in study 2 included an additional adverbial cue for temporal reference, which either occurred consistently with the morphological tense marker or variably across trials. Results indicated a blocking effect: the presence of a temporal adverb reduced reliance on morphological tense cues. The learning was overall better in the consistent condition. However, the availability of cues did not modulate the blocking effect. Study 3 explored the roles of cue salience. Morphological cues were systematically manipulated in their phonological prominence and structural transparency. Learners demonstrated significantly greater success in acquiring syllable cues (e.g., /ti/) compared to the single-consonant cues (e.g., /d/), indicating that perceptual salience significantly affects the accessibility of morphological information. However, the variance of results cannot be explained simply by the syllabicity of the cue, which is discussed in study 3. Overall, the findings demonstrate that while adults can rapidly extract morphological regularities through cross-situational exposure, learning outcomes are shaped by cue salience, cue competition, and cue availability (Ellis, 2006). Contrary to predictions from L1 transfer accounts, L1 morphological complexity did not robustly explain variance in learning success, underscoring the view of L1 being a dominant transfer source.