The Mental Lexicon - Current Issue

Aging involves a variety of cognitive and language changes. Word association norms (WANs), which describe the first word people say or write after reading or hearing a stimulus word, reflect the organization of semantic knowledge. The aim of this study was to create a WAN corpus for Mexican older adults, with 114 participants responding to 117 nouns. The results of eight measures of word-word relationships showed that participants generated an average of 38.14 different response words for each stimulus word. Of these responses, 41.88% were of high associative strength and 20.70% were idiosyncratic, demonstrating the uniqueness of responses of older adults. In addition, we compared their responses to a corpus for younger adults (Barrón-Martínez & Arias-Trejo, 2014). A qualitative analysis categorizing the responses into syntagmatic and paradigmatic types showed that the older group tended to respond with words from a different grammatical class. Responses of the younger adults were also more cohesive and less varied than those of the older group. This corpus for older adults is an essential resource for evaluating age-related changes in semantic memory, and it provides a point of comparison for responses from people with neurodegenerative diseases.

Thul et al. (2020) called attention to problems that arise when chronometric experiments implementing specific factorial designs are analysed with the generalized additive mixed model (GAMM), using factor smooths to capture trial-to-trial dependencies. From a series of simulations incorporating such dependencies, they conclude that GAMMs are inappropriate for between-subject designs. They argue that in addition GAMMs come with too many modeling possibilities, and advise using the linear mixed model (LMM) instead. As clarified by the title of their paper, their conclusion is: “Using GAMMs to model trial-by-trial fluctuations in experimental data: More risks but hardly any benefit”.

We address the questions raised by Thul et al. (2020), who clearly demonstrated that problems can indeed arise when using factor smooths in combination with factorial designs. We show that the problem does not arise when using by-smooths. Furthermore, we have traced a bug in the implementation of factor smooths in the mgcv package, which will have been removed from version 1.8–36 onwards.

To illustrate that GAMMs now produce correct estimates, we report simulation studies implementing different by-subject longitudinal effects. The maximal LMM emerges as slightly conservative compared to GAMMs, and GAMMs provide estimated coefficients that can be less variable across simulation runs. We also discuss two datasets where time-varying effects interact with numerical predictors in a theoretically informative way.

Language experience is essential for SLA. Yet, studies comparing the role of L2 proficiency and L2 use on L2 processing are scant, and there are no studies examining how these variables modulate learners’ ability to generalize grammatical associations to new instances. This study investigates whether L2 proficiency and L2 use affect L2 stress-tense suffix associations (a stressed syllable cuing a present suffix, and an unstressed syllable cuing a preterit suffix) using eye-tracking. Spanish monolinguals and English learners of Spanish varying in L2 proficiency and L2 use saw two verbs (e.g., firma-firmó ‘(s)he signs/signed’), heard a sentence containing one of the verbs, and chose the verb they had heard. Both groups looked at target verbs above chance before hearing the suffix, but the monolinguals did so more accurately and earlier than the learners. The learners recognized past verbs faster than present verbs, were faster with higher than lower L2 proficiency, and later with higher than lower L2 use. Finally, higher L2 proficiency yielded earlier morphological activation but higher L2 use produced later morphological activation, indicating that L2 proficiency and L2 use affect L2 word processing differently. We discuss the contribution of these findings to language acquisition and processing models, as well as models of general cognition.

The current study investigates two lexical explanations for variation in L2 production: approximate (‘fuzzy’) representations vs dual URs. The focus is on Quebec francophone (QF) production of English /θ ð/ and /h/, which a reading-aloud task shows to be highly variable. Variation is problematic for the assumption that, due to perceptual illusions, URs are inaccurate. How is accurate output generated from inaccurate URs? Approximate representations employ diacritics rather than distinctive features. Arguably, these representations do not consistently generate accurate output. Under dual URs, lexical entries contain both inaccurate URs due to initial misperceptions and accurate URs generated when learners become capable of perceiving L2 phonemes. These URs compete for selection, leading to variation. Perception findings from oddball and semantic incongruity tasks provide conflicting support for the explanations: perception is variable, as predicted under approximate representations; but typical L2→L1 substitutions are harder to detect than atypical L1→L2 substitutions, an asymmetry expected under dual URs. To resolve the contradiction, we reinterpret the latter findings as revealing an implicit strategy of corrective adjustment acquired through experience with L2 errors. While we conclude that the L2 lexicon employs approximate representations, an enduring enigma concerns the considerably higher rates of hypercorrect [h] than [θ ð].

Psycholinguists have yet to reach a consensus on what role constituent morphemes play in the processing of compound words, although some recent work suggests that morphemes are activated obligatorily during processing. In the current study, we investigate whether people use morphemes to attempt meaning construction even for pseudo-compounds which are words that appear to have a compound structure, but in fact do not (e.g., carpet is not car + pet). We obtained relational entropies (a measure of potential relational competition) for a set of pseudo-compound words based on responses from a possible relations task. The relational entropy values as well as frequency of the prime (e.g., carpet) and target (e.g., car) were then used to predict the processing of the pseudo-first constituents after exposure to the pseudo-compound masked primes. We observed a significant three-way interaction between entropy, target frequency, and prime frequency. Our results suggest that meaning construction is attempted for pseudo-compound words.

Ideophones are marked words that depict sensory imagery and are hypothesized to be structurally marked, i.e., exhibiting unique structural properties. In this paper, “marked” is broadly used to mean phonologically marked (Dingemanse, 2021: Akita and Dingemanse, 2019). Using Cantonese ideophones as our case study, this paper measures sequential predictability within ideophones and non-ideophones, as a way to test their relative degree of structural markedness. We created a database of non-ideophones and ideophones from the Hong Kong Cantonese Corpus (HKCC) (Luke and Wong, 2015) and Mok (2001) and calculated the sequential predictability of each phoneme in various phonological contexts. The results indicate that Cantonese ideophones exhibit lower degrees of sequential predictability than non-ideophones, lending empirical support to the structural markedness of ideophones. We argue that non-ideophones exhibit a higher degree of sequential predictability because they follow the phonotactic regularities of Cantonese, whereas ideophones, to some degree, flout these regulations in favor of sequences of sounds that might better depict a given referent or percept.