Journal of memory and language最新文献

Reading research robustly finds that shorter and more frequent words are recognized faster and skipped more often than longer and less frequent words. An empirical question that has not been tested yet is whether languages within the same writing system would produce similarly strong length and frequency effects or whether typological differences between written languages would cause those effects to vary systematically in their magnitude. We analyzed text reading eye-movement data in 12 alphabetic languages from the Multilingual Eye-Movement Corpus (MECO). The languages varied substantially in their word length and frequency distributions as a function of their orthographic depth and morpho-syntactic type. Yet, the effects of word length and frequency on fixation durations and skipping rate were highly similar in size between the languages. This finding suggests a high degree of cross-linguistic universality in the readers’ behavioral response to linguistic complexity (indexed by word length) and the amount of experience with the word (indexed by word frequency). These findings run counter to influential theories of single word recognition, which predict orthographic depth of a language to modulate the size of these benchmark effects. They also facilitate development of cross-linguistically generalizable computational models of eye-movement control in reading.

Models of eye-movement control during reading, developed largely within psychology, usually focus on visual, attentional, lexical, and motor processes but neglect post-lexical language processing; by contrast, models of sentence comprehension processes, developed largely within psycholinguistics, generally focus only on post-lexical language processes. We present a model that combines these two research threads, by integrating eye-movement control and sentence processing. Developing such an integrated model is extremely challenging and computationally demanding, but such an integration is an important step toward complete mathematical models of natural language comprehension in reading. We combine the SWIFT model of eye-movement control (Seelig et al., 2023) with key components of the Lewis and Vasishth sentence processing model (Lewis and Vasishth, 2005). This integration becomes possible, for the first time, due in part to recent advances in successful parameter identification in dynamical models, which allows us to investigate profile log-likelihoods for individual model parameters. We present a fully implemented proof-of-concept model demonstrating how such an integrated model can be achieved; our approach includes Bayesian model inference with Markov Chain Monte Carlo (MCMC) sampling as a key computational tool. The integrated Sentence-Processing and Eye-Movement Activation-Coupled Model (SEAM) can successfully reproduce eye movement patterns that arise due to similarity-based interference in reading. To our knowledge, this is the first-ever integration of a complete process model of eye-movement control with linguistic dependency completion processes in sentence comprehension. In future work, this proof of concept model will need to be evaluated using a comprehensive set of benchmark data.

For decades, researchers have debated whether readers benefit from translating visual word forms into phonological codes. A focus of this debate has been on the earliest moments of processing when a word is perceived in parafoveal vision (i.e., phonological preview benefit). A recent meta-analysis (Vasilev et al., 2019) concluded that the phonological preview benefit may be small and unreliable but they did not take into account potentially important stimulus-level or participant-level factors that varied across the included studies. Therefore, we conducted two well-powered experiments that systematically investigated the effects of sentence constraint, preview lexicality, and participant language skills on the phonological preview benefit effect. We found phonological preview benefits that were larger in high versus low constraint sentences, larger for words than pseudowords, and larger for better spellers. We conclude that phonological codes do facilitate early word recognition during reading, but that the phonological preview benefit magnitude depends on subject- and stimulus-level factors.

In two structural priming experiments, we investigated the representations of lexically-specific syntactic restrictions of English verbs for highly proficient and immersed second language (L2) speakers of English. We considered the interplay of two possible mechanisms: generalization from the first language (L1) and statistical learning within the L2 (both of abstract structure and of lexically-specific information). In both experiments, L2 speakers with either Germanic or Romance languages as L1 were primed to produce dispreferred double-object structures involving non-alternating dative verbs. Priming occurred from ungrammatical double-object primes involving different non-alternating verbs (Experiment 1) and from grammatical primes involving alternating verbs (Experiment 2), supporting abstract statistical learning within the L2. However, we found no differences between L1-Germanic speakers (who have the double-object structure in their L1) and L1-Romance speakers (who do not), inconsistent with the prediction for between-group differences of the L1-generalization account. Additionally, L2 speakers in Experiment 2 showed a lexical boost: There was stronger priming after (dispreferred) non-alternating same-verb double-object primes than after (grammatical) alternating different-verb primes. Such lexically-driven persistence was also shown by L1 English speakers (Ivanova, Pickering, McLean, Costa, & Branigan, 2012) and may underlie statistical learning of lexically-dependent structural regularities. We conclude that lexically-specific syntactic restrictions in highly proficient and immersed L2 speakers are shaped by statistical learning (both abstract and lexically-specific) within the L2, but not by generalization from the L1.

The present research shows that language comprehenders are sensitive to multiple states of target and semantically related objects. In Experiments 1 to 2B, participants (total N = 273) read sentences that either implied a minimal change of an object’s state (e.g., “Jane chose a mango”) or a substantial change (e.g., “Jane stepped on a mango”) and then verified whether a subsequently pictured object was mentioned in the sentence. Crucially, the picture either showed the original/modified state of an object that was mentioned in the sentence (e.g., “mango” in Experiment 1) or not (e.g., “banana” in Experiments 2A and 2B). The results of Experiment 1 demonstrated that the objects in a modified state were verified faster when a sentence implied a substantial state-change rather than a minimal state-change. In contrast, the reverse was true for the objects in the original state. Importantly, verification latencies of pictures depicting original and modified states of an object in the substantial state-change condition were approximately similar, thus suggesting that language comprehenders maintain multiple representations of an object in different states. The results of Experiments 2A and 2B revealed that when participants had to indicate that a pictured object (e.g., banana) was not mentioned in the sentence, their verification latencies were slowed down when the sentence contained a semantically related item (e.g., mango) and described this item as being changed substantially by the action. However, these verification latencies varied continuously with the degree of change: the more dissimilar the states of a semantically related item, the less time participants needed to verify a pictured object. The results are discussed through the prism of theories emphasizing dynamic views of event cognition.

The past 10 years have seen rapid growth of online (web-based) data collection across the behavioural sciences. Despite the many important contributions of such studies, some researchers have concerns about the reduction in experimental control when research moves outside of laboratory conditions. This paper provides an accessible overview of the issues that can adversely affect data quality in online experiments, with particular focus on cognitive studies of memory and language. I provide checklists for researchers setting up such experiments to help improve data quality. These recommendations focus on three key aspects of experimental design: the technology choices made by researchers and participants, participant recruitment methods, and the performance of participants during experiments. I argue that ensuring high data quality for online experiments requires significant effort prior to data collection to maintain the credibility of our rapidly expanding evidence base. With such safeguards in place, online experiments will continue to provide important, paradigm-changing opportunities across the behavioural sciences.

We investigate the onomasiological question of which words speakers actually use and produce when trying to convey an intended meaning. This is not limited to selecting the best-fitting available existing word, but also includes word formation, the coinage of novel words. In the first two experiments, we introduce the taboo game paradigm in which participants were instructed to produce a single-word substitution for different words so that others can later identify them. Using distributional semantic models with the capability to produce quantitative representations for existing and novel word responses, we find that (a) responses tend to be semantically close to the targets and (b) existing words were represented closer than novel words, but (c) even novel compounds were often closer than the targets’ free associates. In a final third experiment, we find that other participants are more likely to guess the correct original word (a) for responses closer to the original targets, and (b) for novel compound responses as compared to existing word responses. This shows that the production of both existing and novel words can be accurately captured in a unified computational framework of the semantic mechanisms driving word choice.

Recall from episodic memory has been shown to enhance both memory for the retrieved information (e.g., relative to a restudy control condition; the testing effect, or TE) and memory for different, subsequently studied materials (the forward testing effect, or FTE). Hence, the TE may be subject to an FTE confound when training in a TE experiment involves either testing prior to restudy or when restudied and tested items are randomly mixed. Across two cued-recall TE experiments, we show that (1) a potent FTE confound exists in the test-first but not the mixed training design, (2) there are no other evident learning related interactions between restudied and tested items across three frequently used training phase task orderings, and (3) the predictions of the dual-memory model of test-enhanced learning – which posits that a test trial creates a memory that is separate from the initially encoded study memory, yielding two routes to retrieval for tested items – are held both when there is and is not a confounding FTE. Further, our results yielded no evidence for two accounts of the FTE (the proactive interference and reset of encoding hypotheses) as applied to cued recall but are consistent with two alternative accounts (the strategy change and increasing effort hypotheses). Through distribution analyses we identify a novel and potent FTE individual differences effect that can be accommodated by the latter accounts. Finally, we show that at least three large-n studies exploring individual differences in the TE are confounded by the FTE, compromising conclusions in those papers about the efficacy of the TE across individuals.

This paper investigates whether agreement attraction is modulated by distributional properties determining subject-likelihood by examining the degree to which bare nouns and full determiner phrases (DPs) cause agreement attraction effects in Romanian. Romanian represents an ideal testing ground for this, given two distributional constraints making bare nouns less subject-like: Locative Determiner Omission, preventing locative prepositions from taking nouns with definite articles (unless modified by adjectives), and the Naked Noun Constraint, disallowing bare nouns as preverbal subjects. We predicted that bare nouns should be less likely to trigger agreement attraction than overt DPs. We conducted four speeded forced-choice sentence continuation tasks on Romanian native speakers to test this prediction. We observe that overt DPs cause significantly more attraction than bare nouns. We suggest that the results are consistent with a cue-based retrieval mechanism for forming agreement dependencies, where cues that determine subjecthood are used to reactivate elements in working memory upon processing a verb. These cues can be language specific, and in Romanian, this means that agreement attraction is sensitive to the morphophonological overtness of the determiner.

Two eye-tracking experiments in alphabetic Finnish and two in logographic Chinese examined the recognition of two-constituent compound words in reading. In Finnish, two-constituent compound words vary greatly in length, whereas in Chinese they are identical in length. According to the visual acuity principle (Bertram & Hyönä, 2003), short Finnish compound words and all two-character Chinese compound words that fit in foveal vision are recognized holistically, whereas long Finnish compound words are recognized via components. Experiment 1 in Finnish provided evidence consistent with the account, whereas the results for long compound words presented in condensed font in Experiment 2 were inconsistent with it. In Chinese, the first-character frequency effect was non-significant even when the compound words were presented in large font. The Finnish results suggest that componential processing is necessary when the compound word entails more than 10 letters. The Chinese results are compatible with the Chinese Reading Model (Li & Pollatsek, 2020) that assumes whole-word representations to overrule the activation of components during compound word recognition.