The Mental Lexicon - Volume 18, Issue 2, 2023

Form predictability has long been known to influence speaker behaviour in language learning and use. However, this observation has largely remained dissociated from the question of the most apt theoretical framing of the effects observed. We set out to seek evidence that speakers’ relationship to form predictability is best characterised in paradigmatic terms: in an experimental task comparable to prediction of one word form from a related one, speakers appear sensitive to the probabilistic, implicative relations that make up a morphological paradigm. We find this effect to be omnidirectional, from any paradigm cell to any paradigm cell. Form predictability does not impact speaker behaviour in a vacuum, but instead works together with aspects of memory and learning to organise the mental lexicon and inform language use. In a corpus study, we map out the complex relationships that exist between paradigmatic form predictability, lexeme frequency and cell frequency in the context of naturalistic language use. Speakers appear to exploit all available probabilistic relationships between the word forms of a language in a way that is predicted by Word and Paradigm theories of morphology, with memory and predictive processing playing a mediating role in all aspects of language use.

Morphological processing has been extensively studied in English and European languages, but there is a growing interest in extending the research to other languages. Here we examined Malay, an Austronesian language that is morphologically rich. We investigated the effects of morphological constituents on lexical decisions for prefixed words. Specifically, we explored whether readers are sensitive to any distributional properties of the prefix and root morphemes. Variables investigated included length and family size for both prefixes and roots, as well as number of allomorphs, consistency, and productivity for prefixes. Decision latencies were collected for 1,280 Malay words of various morphological structures. Data from the 640 prefixed words were analyzed in a series of GAMM models. We observed a facilitative effect of root family size and an effect of several distributional properties of prefixes on decision latencies after accounting for word frequency and length. Furthermore, a larger interaction between frequency and several distributional properties of prefixes was found for words with three-letter prefixes than for those with two-letter prefixes. These findings provide insight into the types of distributional properties to which Malay readers are sensitive in multimorphemic words.

A key question in research concerning the typing production of morphologically complex words is whether the whole multimorphemic word is output ballistically or whether individual constituents are accessed during typing. To address this question, we examined keystroke latencies during the production of English compounds (e.g., snowball) to test whether the initiation and continued typing of each constituent (e.g., snow and ball) are influenced by its linguistic properties (length, frequency, and semantic transparency). Participants identified and then typed a compound word. We found that the initiation and continued typing of each constituent was influenced by the linguistic properties of that constituent. However, the linguistic properties of the second constituent also influenced the typing latency of the final letter of the first constituent, suggesting that production of the first constituent overlapped with accessing and planning the keystrokes of the second constituent. The influence of the linguistic properties of the first constituent on its own initiation and continued typing suggests that accessing and planning the keystrokes of the first constituent occurred as the compound word was being identified. Our findings indicate that individual constituents are accessed during production and influence the typing of compound words.

Two types of discontinuous morphemes are thought to be the basic building blocks of words in Semitic languages: roots and templates. However, the role of these morphemes in lexical access and representation is debated. Priming experiments, where reaction times to target words are predicted to be faster when preceded by morphologically-related primes compared to unrelated control primes, provide conflicting evidence bearing on this debate. We used meta-analysis to synthesise the findings from 229 priming experiments on 4710 unique Semitic speakers. With Bayesian modelling of the aggregate effect sizes, we found credible root and template priming in both nouns and verbs in Arabic and Hebrew. Our results show that root priming effects can be distinguished from the effects of overlap in form and meaning. However, more experiments are needed to determine if template priming effects can be distinguished from overlap in form and morphosyntactic function.