Are lexical representations graded or discrete?

Most research on mental lexical representations (lemmas) assumes they are discrete and correspond in number to a word's number of distinct meanings. Thus, homophones (bat), whose meanings are unrelated, have separate lemmas for each meaning (one for baseball bat, another for flying bat), whereas polysemes (paper), whose senses are related, have shared lemmas (the same lemma for printer paper and term paper). However, most aspects of cognition are thought to be graded, not discrete; could lemmas be graded too? We conducted a preregistered picture-word interference study with pictures of words whose meanings ranged from unrelated (homophones) to very related (regular polysemes). Whereas semantic competitors to picture names slow picture naming, semantic competitors to non-depicted meanings of homophones facilitate naming, suggesting distinct lemmas for homophones' meanings. We predicted that competitors to non-depicted senses of polysemes would slow naming, as polysemes' depicted and non-depicted senses presumably share a lemma. Crucially, we aimed to examine the transition from facilitation to inhibition: two groupings (where competitors to non-depicted senses led to facilitation for words with two lemmas but inhibition for words with one lemma) would imply that lemmas are indeed discrete. But a transition that varies continuously by sense relatedness would imply that lemmas are graded. Unexpectedly, competitors to non-depicted senses of both homophones and polysemes facilitated naming. Although these results do not indicate whether lemmas are graded or discrete, they do inform a long-standing question on the nature of polysemes, supporting a multiple-lemma (vs. core-lemma) account.

The "Small World of Words" free association norms for Rioplatense Spanish.

Large-scale word association datasets are both important tools used in psycholinguistics and used as models that capture meaning when considered as semantic networks. Here, we present word association norms for Rioplatense Spanish, a variant spoken in Argentina and Uruguay. The norms were derived through a large-scale crowd-sourced continued word association task in which participants give three associations to a list of cue words. Covering over 13,000 words and +3.6 M responses, it is currently the most extensive dataset available for Spanish. We compare the obtained dataset with previous studies in Dutch and English to investigate the role of grammatical gender and studies that used Iberian Spanish to test generalizability to other Spanish variants. Finally, we evaluated the validity of our data in word processing (lexical decision reaction times) and semantic (similarity judgment) tasks. Our results demonstrate that network measures such as in-degree provide a good prediction of lexical decision response times. Analyzing semantic similarity judgments showed that results replicate and extend previous findings demonstrating that semantic similarity derived using spreading activation or spectral methods outperform word embeddings trained on text corpora.

How trial-to-trial learning shapes mappings in the mental lexicon: Modelling lexical decision with linear discriminative learning.

Trial-to-trial effects have been found in a number of studies, indicating that processing a stimulus influences responses in subsequent trials. A special case are priming effects which have been modelled successfully with error-driven learning (Marsolek, 2008), implying that participants are continuously learning during experiments. This study investigates whether trial-to-trial learning can be detected in an unprimed lexical decision experiment. We used the Discriminative Lexicon Model (DLM; Baayen et al., 2019), a model of the mental lexicon with meaning representations from distributional semantics, which models error-driven incremental learning with the Widrow-Hoff rule. We used data from the British Lexicon Project (BLP; Keuleers et al., 2012) and simulated the lexical decision experiment with the DLM on a trial-by-trial basis for each subject individually. Then, reaction times were predicted with Generalized Additive Models (GAMs), using measures derived from the DLM simulations as predictors. We extracted measures from two simulations per subject (one with learning updates between trials and one without), and used them as input to two GAMs. Learning-based models showed better model fit than the non-learning ones for the majority of subjects. Our measures also provide insights into lexical processing and individual differences. This demonstrates the potential of the DLM to model behavioural data and leads to the conclusion that trial-to-trial learning can indeed be detected in unprimed lexical decision. Our results support the possibility that our lexical knowledge is subject to continuous changes.

El estudio de los efectos de la modalidad lingüística en el léxico mental / The study of the effects of linguistic modality on the mental lexicon

Resumen El estudio se cuestiona sobre un abordaje fructífero en el estudio de los efectos de la modalidad lingüística (de señas versus oral) en la organización de la memoria semántica. Para responder la pregunta de investigación, un aspecto central es definir criterios y procedimientos capaces de distinguir entre los factores transportados por la modalidad lingüística de los que le son propios. Entre los primeros, se encuentran la edad de exposición a la lengua natural y su distribución funcional, etaria o geográfica. Entre los aspectos propios de la modalidad se incluyen la iconicidad y el grado de concreción/abstracción de las piezas léxicas. Se sugiere centrarse en el uso de dicho gradiente para evocar dominios conceptuales. Si lo concreto/abstracto es un continuo, entonces hay que preguntarse qué forma de organización es predominante en el léxico mental de las poblaciones señantes u oyentes y qué factores modulan el perfil de esa organización.

Abstract The efforts to understand the relationship of the meaning of a lexical piece to the object to which it refers, the relationship between the lexical pieces and the concepts, and the understanding of the meanings shared between two speakers who use the same lexical pieces have all constituted major problems for the semantic memory models. According to contemporary literature, perceptual-motor, linguistic, and social information have different weights in the formation of concepts, whether concrete or abstract, stored in the aforementioned memory. Regardless of the models developed so far, it is interesting to note that semantic knowledge is represented by various ways of relating the concepts and the types of relationships between them. In this context, studies in sign languages and comparative studies between spoken and sign languages are scarce. Thus, little is known about the effect of linguistic modality on the semantic networks. After all, the theory on semantic networks and norms for the production of features has been grounded on theories of language and its processing adjusted to spoken languages. As the incorporation of the sign language and the Deaf population has shown in other psycholinguistic and linguistic topics, the importance of including these languages and populations, and comparatives with spoken languages, might increase the explanatory power of the theory to account for the universal and contextual aspects of language and its processing. In this effort, there is a latent risk: the linguistic modality can be only a vehicle for more well-known or studied cross-modal variables (e. g., age of acquisition, functional distribution of language, size of the available lexicon, etc.). If it is considered that languages are not stored together, but similar processes can occur in them, it is essential to find out what may be a singular feature of each modality (spoken versus sign) that might ground differentiated processes. Considering the high iconicity of the sign languages and the possibility of a high concreteness of the lexical pieces in the sign language as distinctive features -not collapsible into well-known variables such as the aforementioned-, this article suggests a careful approach to avoid the aforementioned risk in the study of the effects of the linguistic modality (sign versus oral) in the organization of semantic memory. Since perceptual-motor and social information are the main sources of iconicity, a balanced instrument is necessary in the evocation of perceptual-motor, social, and linguistic information. Repeated free word association tasks seem like an appropriate paradigm for a suggested approach. The reasons for this are that, by not censoring the types of response, then free association tasks allow capturing all kinds of concepts (concrete or abstract), all kinds of semantic relationships/organization (paradigmatic versus thematic) and all kinds of processes (comparison versus interaction). This type of task therefore makes it possible to collect meanings related to linguistic information and non-linguistic experience because affective and experiential information is accessible by doing the task in different repetitions. The approach and the tool are exemplified by an ongoing comparative study between Deaf signing and hearing populations. The partial findings of this study also serve to focus on the expected effects of the difference in iconicity and the level of concreteness/abstractness of the lexical pieces of each linguistic modality; namely, the differences between an abstract and a concrete conceptualization of the conceptual domains. Taxonomic and introspective labels might appear as indicative of paradigmatic relationships, of a taxonomic organization, and of underlying comparison processes. On the other hand, the situational and entity labels, indicative of syntagmatic relationships, of a thematic organization and of underlying interaction processes, might suggest a predominantly concrete organization.

Frequency of word usage by Hebrew preschoolers: implications for AAC core vocabulary.

This study's goal was to inform the selection of the most frequently used words to serve as a reference for core vocabulary selection for Hebrew-speaking children who require AAC. The paper describes the vocabulary used by 12 Hebrew-speaking preschool children with typical development in two different conditions: peer talk, and peer talk with adult mediation. Language samples were audio-recorded, transcribed, and analyzed using the CHILDES (Child Language Data Exchange System) tools to identify the most frequently used words. The top 200 lexemes (all variations of a single word) in the peer talk and adult-mediated peer talk conditions accounted for 87.15% (n = 5008 tokens) and 86.4% (n = 5331 tokens) of the total tokens produced in each language sample (n = 5746, n = 6168), respectively. A substantially overlapping vocabulary of 337 lexemes accounted for up to 87% (n = 10411) of the tokens produced in the composite list (n = 11914). The results indicate that a relatively small set of words represent a large proportion of the words used by the preschoolers across two different conditions. General versus language-specific implications for core vocabulary selection for children in need of AAC devices are discussed.

A tale of two lexica: Investigating computational pressures on word representation with neural networks.

Introduction: The notion of a single localized store of word representations has become increasingly less plausible as evidence has accumulated for the widely distributed neural representation of wordform grounded in motor, perceptual, and conceptual processes. Here, we attempt to combine machine learning methods and neurobiological frameworks to propose a computational model of brain systems potentially responsible for wordform representation. We tested the hypothesis that the functional specialization of word representation in the brain is driven partly by computational optimization. This hypothesis directly addresses the unique problem of mapping sound and articulation vs. mapping sound and meaning. Results: We found that artificial neural networks trained on the mapping between sound and articulation performed poorly in recognizing the mapping between sound and meaning and vice versa. Moreover, a network trained on both tasks simultaneously could not discover the features required for efficient mapping between sound and higher-level cognitive states compared to the other two models. Furthermore, these networks developed internal representations reflecting specialized task-optimized functions without explicit training. Discussion: Together, these findings demonstrate that different task-directed representations lead to more focused responses and better performance of a machine or algorithm and, hypothetically, the brain. Thus, we imply that the functional specialization of word representation mirrors a computational optimization strategy given the nature of the tasks that the human brain faces.

Frequency effects in linear discriminative learning.

Word frequency is a strong predictor in most lexical processing tasks. Thus, any model of word recognition needs to account for how word frequency effects arise. The Discriminative Lexicon Model (DLM) models lexical processing with mappings between words' forms and their meanings. Comprehension and production are modeled via linear mappings between the two domains. So far, the mappings within the model can either be obtained incrementally via error-driven learning, a computationally expensive process able to capture frequency effects, or in an efficient, but frequency-agnostic solution modeling the theoretical endstate of learning (EL) where all words are learned optimally. In the present study we show how an efficient, yet frequency-informed mapping between form and meaning can be obtained (Frequency-informed learning; FIL). We find that FIL well approximates an incremental solution while being computationally much cheaper. FIL shows a relatively low type- and high token-accuracy, demonstrating that the model is able to process most word tokens encountered by speakers in daily life correctly. We use FIL to model reaction times in the Dutch Lexicon Project by means of a Gaussian Location Scale Model and find that FIL predicts well the S-shaped relationship between frequency and the mean of reaction times but underestimates the variance of reaction times for low frequency words. FIL is also better able to account for priming effects in an auditory lexical decision task in Mandarin Chinese, compared to EL. Finally, we used ordered data from CHILDES to compare mappings obtained with FIL and incremental learning. We show that the mappings are highly correlated, but that with FIL some nuances based on word ordering effects are lost. Our results show how frequency effects in a learning model can be simulated efficiently, and raise questions about how to best account for low-frequency words in cognitive models.

The Mental Lexicon Features of the Hakka-Mandarin Dialect Bilingual.

The current study investigated the mental lexicon features of the Hakka-Mandarin dialect bilingual from two perspectives: the structural features of lexicons and the relations between lexicons. Experiment one used a semantic fluency task and complex-network analysis to observe the structural features of lexicons. Experiment two used a cross-language long-term repetition priming paradigm to explore the relations between lexicons, with three sub-experiments focusing on conceptual representation, lexical representation, and their relations, respectively. The results from experiment one showed that the dialect bilingual lexicons were small-world in nature, and the D2 (Mandarin) lexicon was better organized than the D1 (Hakka) lexicon. Experiment two found that D1 and D2 might have partially shared conceptual representations, separate lexical form representations, and partially shared lemma representations. Based on the findings, we tentatively proposed a two-layer activation model to simulate the lexicon features of dialect bilingual speakers.

Mental Representation of Word Family Structure: The Case of German Infinitives, Conversion Nouns and Other Morphologically Related Forms.

This study investigates how two non-finite forms, infinitives and conversion nouns, are represented in the mind of L1 and L2 speakers and what is their relationship to other members of the corresponding word family. German native speakers and proficient German learners with Czech as L1 participated in four overt priming experiments involving a grammatical judgement task. We investigated the relationship between infinitives (Experiment 1) and conversion nouns (Experiment 2) and formally identical verbal or noun forms. We further focussed on the relationship between conversion nouns and regular nominal derivation forms with two derivational suffixes: -er and -ung (Experiments 3 and 4). Our results show that the two non-finite forms differ in their relations to other members of a word family and do not constitute a special class of non-finites as suggested in previous literature. While German infinitives seem to be closer related to finite verbal forms, conversion nouns behave in the same way as other regular nominal derivatives within the same word family. As for the German L1 and L2 contrast, no significant difference in the mental representation of the examined forms was found. This finding suggests that with respect to the explored phenomena, proficient learners rely on the same linguistic organisation as L1 speakers.

Characterizing the semantic and form-based similarity spaces of the mental lexicon by means of the multi-arrangement method.

Collecting human similarity judgments is instrumental to measuring and modeling neurocognitive representations (e.g., through representational similarity analysis) and has been made more efficient by the multi-arrangement task. While this task has been tested for collecting semantic similarity judgments, it is unclear whether it also lends itself to phonological and orthographic similarity judgments of words. We have extended the task to include these lexical modalities and compared the results between modalities and against computational models. We find that similarity judgments can be collected for all three modalities, although word forms were considered more difficult to sort and resulted in less consistent inter- and intra-rater agreement than semantics. For all three modalities we can construct stable group-level representational similarity matrices. However, these do not capture significant idiosyncratic similarity information unique to each participant. We discuss the potential underlying causes for differences between modalities and their effect on the application of the multi-arrangement task.

A Study on the Developmental Features of Yi Students' Chinese Mental Lexicon.

Adopting free word association test, the present study investigated the developmental features of Yi students' Chinese mental lexicon. Eighty primary school students and 85 senior high school students in two typical Yi-Han bilingual schools in Yuexi County were recruited as the research subjects. With Yi language as their L1, all the participants started learning Chinese after entering primary school. The stimuli were 108 words selected from the 9,000 most frequently used words in modern Chinese, including 36 nouns, 36 verbs, and 36 adjectives. The responses were analyzed using the framework proposed by the responses were analyzed using a three-layer framework. The results showed that (1) the mental lexicon of Yi primary and middle school students were mainly connected with each other by meanings, with form-based connection followed. Compared with previous studies, their phonological associations accounted for a relatively higher proportion, while semantic associations were relatively weakened. (2) Syntagmatic associations were of primary importance, with paradigmatic and encyclopedic associations followed. (3) While syntagmatic relation was centered on determinative and governing associations, paradigmatic relation was dominated by adjacent and reverse associations. (4) The development rate of syntagmatic associations was faster than that of paradigmatic and encyclopedic associations; no significant improvement was found on students' adjacent, layer, homogeneous and supplement associations from primary school to high school. The findings generate important implications for Yi students' Chinese language education.

Unsupervised Text Segmentation Predicts Eye Fixations During Reading.

Words typically form the basis of psycholinguistic and computational linguistic studies about sentence processing. However, recent evidence shows the basic units during reading, i.e., the items in the mental lexicon, are not always words, but could also be sub-word and supra-word units. To recognize these units, human readers require a cognitive mechanism to learn and detect them. In this paper, we assume eye fixations during reading reveal the locations of the cognitive units, and that the cognitive units are analogous with the text units discovered by unsupervised segmentation models. We predict eye fixations by model-segmented units on both English and Dutch text. The results show the model-segmented units predict eye fixations better than word units. This finding suggests that the predictive performance of model-segmented units indicates their plausibility as cognitive units. The Less-is-Better (LiB) model, which finds the units that minimize both long-term and working memory load, offers advantages both in terms of prediction score and efficiency among alternative models. Our results also suggest that modeling the least-effort principle for the management of long-term and working memory can lead to inferring cognitive units. Overall, the study supports the theory that the mental lexicon stores not only words but also smaller and larger units, suggests that fixation locations during reading depend on these units, and shows that unsupervised segmentation models can discover these units.

Methodological Considerations for Incorporating Clinical Data Into a Network Model of Retrieval Failures.

Difficulty retrieving information (e.g., words) from memory is prevalent in neurogenic communication disorders (e.g., aphasia and dementia). Theoretical modeling of retrieval failures often relies on clinical data, despite methodological limitations (e.g., locus of retrieval failure, heterogeneity of individuals, and progression of disorder/disease). Techniques from network science are naturally capable of handling these limitations. This paper reviews recent work using a multiplex lexical network to account for word retrieval failures and highlights how network science can address the limitations of clinical data. Critically, any model we employ could impact clinical practice and patient lives, harkening the need for theoretically well-informed network models.

Research on extracting mental lexicon from legal cases of players based on text features / Investigación sobre la extracción de léxico mental de casos legales de jugadores en función de las características del texto

Mental lexicon extraction has important theoretical and practical significance in legal case judgment; however, mental lexicon cannot be extracted in time and effectively. Therefore, this paper puts forward the research on the extraction method of mental lexicon in legal cases of players based on text features. In terms of extraction factors of mental lexicon in legal cases of players, quantity and quality are the two most important factors, and form an important reference for mental lexicon recognition. Therefore, it should be combined with text features to identify and extract mental lexicon frequency, and the extraction process is described in detail in this paper by using pattern recognition theory. The psychological test is applied to test the biological reaction of the suspect to determine and judge the relationship between the suspect and the case to ensure the accuracy of the case.(AU)

Representation of Different Types of Adjectival Polysemy in the Mental Lexicon.

We studied mental representations of literal, metonymically different, and metaphorical senses in Russian adjectives. Previous studies suggested that in polysemous words, metonymic senses, being more sense-related, were stored together with literal senses, whereas more distant metaphorical senses had separate representations. We hypothesized that metonymy may be heterogeneous with respect to its mental storage. "Whole-part" metonymy ("sad person" - "sad eyes"), which is cognitively closer to the literal sense and more regular, should be stored differently from temporal, causal or resultative metonymy ("sad person" - "sad time;" "sad person" - "sad news;" "lead.ADJ ball" - "lead.ADJ poisoning"), which is irregular and semantically distant from the literal sense. We conducted an online experiment with semantic clustering task in which the participants were asked to classify sentences with a literal, proximal metonymic, distal metonymic, or metaphorical sense of an adjective into virtual baskets so that sentences with the same perceived sense were put in the same basket. Our results showed that proximal metonymies were grouped together with the literal sense and with each other more often than with distal metonymies and metaphors. Distal metonymies, in turn, were grouped with literal senses more often than with metaphors. Overall, we concluded that literal senses and proximal metonymies were stored in single representations, distal metonymies formed hybrid representations with literal senses, and metaphors were stored separately from literal senses. Additionally, we discovered that perception of semantic differences is affected by the surrounding senses: distal metonymies were more discernible from literal senses when presented with proximal metonymies, and less so when presented with metaphors.

Morpho-Phonetic Effects in Speech Production: Modeling the Acoustic Duration of English Derived Words With Linear Discriminative Learning.

Recent evidence for the influence of morphological structure on the phonetic output goes unexplained by established models of speech production and by theories of the morphology-phonology interaction. Linear discriminative learning (LDL) is a recent computational approach in which such effects can be expected. We predict the acoustic duration of 4,530 English derivative tokens with the morphological functions DIS, NESS, LESS, ATION, and IZE in natural speech data by using predictors derived from a linear discriminative learning network. We find that the network is accurate in learning speech production and comprehension, and that the measures derived from it are successful in predicting duration. For example, words are lengthened when the semantic support of the word's predicted articulatory path is stronger. Importantly, differences between morphological categories emerge naturally from the network, even when no morphological information is provided. The results imply that morphological effects on duration can be explained without postulating theoretical units like the morpheme, and they provide further evidence that LDL is a promising alternative for modeling speech production.

On the Locus of L2 Lexical Fuzziness: Insights From L1 Spoken Word Recognition and Novel Word Learning.

The examination of how words are learned can offer valuable insights into the nature of lexical representations. For example, a common assessment of novel word learning is based on its ability to interfere with other words; given that words are known to compete with each other (Luce and Pisoni, 1998; Dahan et al., 2001), we can use the capacity of a novel word to interfere with the activation of other lexical representations as a measure of the degree to which it is integrated into the mental lexicon (Leach and Samuel, 2007). This measure allows us to assess novel word learning in L1 or L2, but also the degree to which representations from the two lexica interact with each other (Marian and Spivey, 2003). Despite the somewhat independent lines of research on L1 and L2 word learning, common patterns emerge across the two literatures (Lindsay and Gaskell, 2010; Palma and Titone, 2020). In both cases, lexicalization appears to follow a similar trajectory. In L1, newly encoded words often fail at first to engage in competition with known words, but they do so later, after they have been better integrated into the mental lexicon (Gaskell and Dumay, 2003; Dumay and Gaskell, 2012; Bakker et al., 2014). Similarly, L2 words generally have a facilitatory effect, which can, however, become inhibitory in the case of more robust (high-frequency) lexical representations. Despite the similar pattern, L1 lexicalization is described in terms of inter-lexical connections (Leach and Samuel, 2007), leading to more automatic processing (McMurray et al., 2016); whereas in L2 word learning, lack of lexical inhibition is attributed to less robust (i.e., fuzzy) L2 lexical representations. Here, I point to these similarities and I use them to argue that a common mechanism may underlie similar patterns across the two literatures.

Word class and word frequency in the MMN looking glass.

The effects of lexical meaning and lexical familiarity on auditory deviance detection were investigated by presenting oddball sequences of words, while participants ignored the stimuli. Stimulus sequences were composed of words that were varied in word class (nouns vs. functions words) and frequency of language use (high vs. low frequency) in a factorial design with the roles of frequently presented stimuli (Standards) and infrequently presented ones (Deviants) were fully crossed. Deviants elicited the Mismatch Negativity component of the event-related brain potential. Modulating effects of lexical meaning were obtained, revealing processing advantages for denotationally meaningful items. However, no effect of word frequency was observed. These results demonstrate that an apparently low-level function, such as auditory deviance detection utilizes information from the mental lexicon even for task-irrelevant stimuli.