Search

Measures of narrative microstructure are used to assess the linguistic features of narrative samples. The most common are measures of syntactic complexity and lexical diversity (e.g., Mäkinen et al. 2014). Previous studies have shown a positive correlation between age and measures of syntactic complexity and lexical diversity (Frizelle et al. 2018 ; Hržica & Roch 2019 ; Wood et al. 2018), suggesting that these measures can be used to better understand the language abilities of children, including children with language and communication disorders (Botting 2002 ; Watkins et al. 1995). There have been no such studies in Croatia. The aim of this study was to investigate the relationship between measures of syntactic complexity, lexical diversity, and age. We expected to find correlations between the measures in narratives of children with different developmental language and speech disorders while controlling their results of standardized tests. The sample consisted of 38 children with developmental speech and language disorders (developmental language disorder, childhood apraxia of speech, and social communication disorder). Children narrated a story based on a sequence of pictures. We calculated correlations between measures of lexical diversity, syntactic complexity, and age. There was a significant correlation between age and syntactic complexity (ρ = .613, p < .01 ; ρ = .637, p < .01), but no correlation was found between age and lexical diversity measures. When interpreting these results, several factors should be considered that may influence the relationship between narrative microstructure and age, such as differences in a population sample, narrative elicitation task, and sample size.

A wide range of tools have been used to assess the language proficiency of bilingual speakers, including heritage language speakers. There are subjective forms of assessment, such as self-reports or self-assessments, and objective forms of assessment, such as analysis of written production elicitations, samples of spontaneous spoken conversations, oral narratives, and standardized tests. Two types of tools were not that often combined. Although self-assessment ratings across single or multiple language domains are a widely used method, their accuracy, reliability, and usefulness have been debated (e.g., Brantmeier, 2006 ; Treffers-Daller, 2015). Thus, there is a need for indices that combine subjective and objective measures and allow for a more in-depth examination of the relationship between these measures (Treffers-Daller, & Korybski, 2015). This study aims to contribute to this line of research by investigating the relationship between self-assessment data and language sample analysis. We focus on heritage language speakers of Italian that live in Istria County, a statutory bilingual county in Croatia. Italophone heritage language varieties in this area (the most prominent being Istrovenetian) date back several centuries and are still used in everyday spoken communication, while standard Italian is mainly used in formal contexts and education (Blagoni et al., 2016). The use of self-assessment and language sample analysis will allow us to obtain information about Istrovenetian, since other objective measures (e.g., standardized tests) are rarely used to assess spoken language varieties. Measures of lexical diversity and syntactic complexity were used to examine whether they reflect the self-assessment of language proficiency of 30 functional Istrovenetian heritage speakers in four domains (reading, writing, speaking, and understanding). Participants’ data were extracted from the Istrian spoken multilingual corpus (Hržica et al., in preparation), which was created using the conversational sampling method. Measures of lexical diversity (Moving-average type-token ratio – MATTR ; Covington, & McFall, 2010) and syntactic complexity (Mean length of communication unit – MLCU ; Heilmann, 2010) were calculated using Computerized Language Analysis (CLAN – MacWhinney, 2000). A Pearson correlation was calculated to test the relationship between self-assessment and language measures. The average self-assessment score correlated positively and moderately with MLCU (r(30) = .513, p < .005). MATTR did not correlate with the average self-assessment score, but the results indicate a trend suggesting this relationship (r(30) = .351, p = 0.57). A Spearman correlation was calculated for each domain of self-assessment. MATTR showed weak positive correlations with self- assessment of understanding (rs(30) = .364, p < .05) and speaking (rs(30) = .388, p < .05), but not with self-assessment of reading and writing. MLCU showed moderate positive correlations with self-assessment of reading (rs(30) =.484, p < .05) and writing (rs(30) = .563, p < .005), and weak positive correlations with speaking (rs(30) = .373, p < .05) and understanding (rs(30) = .364, p < .05). Descriptive statistics are presented in Table 1. The correlations found between measures of lexical diversity and syntactic complexity and self-assessment scores indicate that these measures are related, suggesting that heritage language speakers may be able to objectively assess their language proficiency in spoken language variety.

Early lexical development is characterised by changes in the size and composition of children’s vocabulary. Previous studies have reported cross-linguistic differences in the composition of children’s vocabulary, which could be due to methodological differences (parent reports vs. longitudinal corpora). The aim of this study was to investigate the relationship between vocabulary composition and age of acquisition (AoA) in Croatian. The analysis was conducted on language samples of two female children aged 1 ; 5 to 2 ; 8 using the Croatian corpus of child language. We calculated the ratios of nouns (PN), predicates (PP), and function words (PF) in the lemmas for each month. Antonija’s data showed a strong correlation between PN nouns and AoA (rs(14) = .78), and between predicates PP and AoA (rs(14) = .79). There was a strong correlation between PF function words and AoA (rs(14) = -.94) (all p < .01). In Marina’s data, we found a strong correlation between PN and AoA (rs(15) = -.70, p < .01) and a moderate correlation between PP and AoA (rs(15) = .62, p < .05). The correlation between PF and the AoA did not reach the significance (rs(15) = .08, p > .05). The present study shows that the ratio of predicates increases from an early age. However, the ratios of function words and nouns show individual differences, decreasing with age or showing no correlation with age. These results differ slightly from previous studies on Croatian, especially regarding the ratio of nouns. Our study contributes to cross-linguistic knowledge of vocabulary acquisition.

The acquisition of reference is one of the central topics in the study of language development. Referential expressions convey different morphosyntactic, semantic, and pragmatic information, so children need to learn which expressions to use for referring in specific communicative situations. Previous studies have mostly focused on the emergence and production of reference in the 1st and/or 2nd person singular. The common assumption is that children acquiring pro-drop languages go through an initial stage of null subjects, followed by the increasing use of overt pronominal subjects, namely personal pronouns. Chronologically, the cut-off point between the two stages appears to be around the age of two. However, it has also been found that children’s referential behaviour differs due to cross-linguistic differences in the availability of referential expressions. The present study investigates the emergence and production of the 1st person singular in children acquiring Croatian. We addressed the following research questions: 1) At what age do children begin to produce overt pronominal and null subjects in the 1st person singular? 2) What is the prevalence of overt pronominal versus null subjects in the 1st person singular marking? 3) Does the prevalence of overt pronominal versus null subjects in the 1st person singular marking change with age? The analysis was conducted on language samples from three children, recorded from the beginning of speech until the age of about 3 years old. All utterances with a verb in the 1st person singular were extracted from the Croatian corpus of child language (Kovačević 2002) and coded by subject type (overt pronominal vs null). The results show that children begin to use pronominal and null subjects in 1st person singular before the age of 2. Null subjects are more frequent than overt pronominal subjects from the beginning of 1st person singular production. Two children show a null-subject stage at the onset of 1st person singular production, while one child shows a stage of overgeneralization of the overtly produced 1st person singular pronoun. The prevalence of null subjects over overt pronominal subjects in 1st person singular is constant throughout the period studied, with the exact ratios changing with age.

Aphasia is an acquired language disorder resulting from damage to the language centers in the brain (ASHA, 2020). Narrative discourse analysis provides a better understanding of the language abilities of people with aphasia. One way to achieve discourse cohesion is through the use of appropriate reference devices to indicate that a piece of information is new (generally indefinite forms, e.g., boy), is given (pronoun), or is assumed (definite forms, e.g., Peter) (Givón, 1989). These pragmatic categories are universal, but the referential devices can vary across languages. The successful use of referential devices depends on linguistic and extra-linguistic factors, including cognitive demands that increase with the number of characters in the story and the gender of the characters (Hendriks et al., 2014). The aim of this study was to investigate reference in the discourse of people with aphasia (PwA) and typical speakers (TS) through the introduction, maintenance, and reintroduction of characters in situations with different cognitive demands (characters of the same gender vs.characters of different gender). Narrative samples of PwA (n = 17) and TS (n = 17) were retrieved from the Croatian discourse corpus of speakers with aphasia (Kuvač Kraljević et al., 2017). The proportions of pronouns and nouns (both in the grammatical role of subject) were calculated for each participant. A 2 x 2 mixed design ANOVA was conducted to test the main effects and interaction of group (PwA vs. TS) and situation (same gender vs. different gender). We found no main effects of group and situation in introduction. In maintenance, there was a main effect of situation (F(1, 32) = 5.993, p < .05). The proportions of pronouns in both groups were significantly higher in situations with characters of different gender, while the proportions of nouns were significantly higher in situations with characters of the same gender. In reintroduction, we found a main effect of group (F(1, 32) = 4.784, p < .05), and the interaction of group and situation was also significant (F(1, 32) = 4.157, p = .05). PwA produced significantly more pronouns compared to TS and the difference was significant in situations with characters of the same gender. There was also a main effect of group (F(1, 32) = 4.621, p < .05) on the proportion of nouns and the interaction between group and situation (F(1, 32) = .364, p = .550). TS produced significantly more nouns than PwA. The descriptive statistics is available in Table 1. In cognitively demanding situations, when characters of the same gender need to be reintroduced with nouns, PwA produced a greater number of pronouns than TS, leading to ambiguity and redundancy. Our findings are consistent with studies showing that PwA are prone to erroneous and repetitive referencing patterns, such as the overuse of pronouns (see Arslan et al., 2021 ; Martinez- Ferreiro et al., 2019 ; Zhang et al., 2020). To conclude, the present study contributes to the limited knowledge on reference in the discourse of PwA, especially in Croatian, by revealing in which functions and/or situations PwA show the greatest differences compared to TS.

Pseudowords (words without semantic meaning) are often used as a control condition in linguistic cognitive experiments, with the expectation that such words, unlike real words, do not activate higher cognitive processes in the brain. However, other theories assume that pseudowords are perceived as new words, leading to an even higher cognitive response. The aim of this study was to investigate the cognitive load of processing a pseudoword by observing event-related potentials in a 3-stimulus oddball paradigm using real target words, real non-target words, and pseudowords as oddball stimuli. The results show a clear task-related P3b triggered by target words, but also a prominent P600 component triggered by pseudowords, indicating difficulty in the classification task due to unknown words. Surprisingly, N400 was decreased for pseudowords compared to target and non-target words at the locations where P3b and P600 were observed, suggesting that task-related effects might inhibit other aspects of cognitive processing. These results could lead to better understanding of the components that may overlap temporally and topographically, and to the more precise control of different cognitive generators involved in event-related potential experiments in pseudowords.

This study investigates how children acquire inflectional systems with morphological overabundance, meaning morphological paradigm cells which can be realised by more than a single lexical form. Four- to six-year-old children acquiring either Croatian or Estonian, two morphologically rich languages with abundant overabundance, will be tested on noun forms with and without overabundance in the target language to investigate how they navigate the parallel forms and to what extent they overgeneralise with nouns which do not exhibit overabundance in adult language use.

Language sample analysis provides rich information about the language abilities in the written or spoken text produced by a speaker in response to a language task. Language sample analysis is generally used to assess the abilities of children during language acquisition, but also the abilities of adult speakers across the lifespan. Its wide range of uses also allows for the assessment of language abilities in educational contexts such as second language acquisition or fluency, the abilities of bilingual speakers in general, and it is also used for diagnosis in speech and language pathology. Various computer programs have been developed to assist in the language sample analysis. However, these programs have been developed mainly for English and are often not fully open-access or do not provide data on population metrics, history of data uploaded by a user, and/or improvements in basic language measures. The time needed for transcription and the linguistic knowledge required for manual analysis are considered to be the main obstacles to its implementation The goal of this paper is to present a web-based application MultiDis intended for the analysis of language samples at the microstructural text level in Croatian. The application is still under development, but the current version fulfils its main purpose – it enables the (semi-) automatic calculation of measures reflecting language productivity, lexical diversity, syntactic complexity, and discourse cohesion in spoken language, and provides users with socio-demographic and linguistic metadata as well as the history of uploaded transcripts. We will present the challenges we have faced in developing the application (e.g., annotation system, text standardisation), future improvements we plan to make to the application (e.g., syntactic parsing, speech-to-text, multilingual analysis), and the possibilities of its use in the wider scientific and professional community.

Označavanje vremena u iskazima ostvaruje se kombinacijom glagolskog vremena, glagolskog vida i konteksta (Bos, 2014). Prema Past DIscourse LInking Hypothesis (engl., PADILIH ; Bastiaanse i sur., 2011) za razumijevanje i proizvodnju glagolskih oblika, tj. za morfološko označavanje glagola koji se odnose na prošlo vrijeme potrebno je uspostaviti vezu između vremena radnje i vremena pripovijedanja jer ona nisu istovremena (Zagona, 2003), što glagole u prošlom vremenu čini kompleksnijima od glagola u neprošlom vremenu (Bastiaanse i sur., 2011). Istraživanja proizvodnje osobnih narativa u različitim jezicima (pr. Aboum i Bastiaanse, 2012 ; Arslan, Bam aci i Bastiaanse, 2016) ukazuju na teškoće označavanja prošlog vremena kod osoba s afazijom netečnog tipa te na povezanost vrsta pogrešaka i tipologije jezika. Cilj je ovog rada opisati označavanje vremena u osobnim narativima osoba s afazijom netečnog tipa u hrvatskom, s naglaskom na prošlo vrijeme i glagolski vid. Ovaj rad nastoji odgovoriti na pitanja kako osobe s afazijom netečnog tipa u osobnim narativima u hrvatskom jeziku (1) označavaju glagolsko vrijeme i vid, i (2) oblikuju morfološke oblike glagola s obzirom na glagolsko vrijeme. Na temelju PADILIH hipoteze pretpostavlja se da osobe s afazijom netečnog tipa u osobnim narativima u hrvatskom jeziku (1) češće rabe neprošlo vrijeme, (2) češće rabe svršeni glagolski vid, (3) češće griješe u označavanju prošlog vremena u usporedbi s neprošlim vremenom, (4) prilikom morfološkog označavanja prošlog vremena češće čine pogreške supstitucije nego omisije morfoloških sufiksa, a (5) najčešće izostavljaju pomoćni glagol u složenim glagolskim oblicima. Budući diskursni uzorci odražavaju funkcionalnu uporabu jezičnog znanja, s ciljem provjere hipoteza, provedena je kvantitativna i kvalitativna analiza uporabe glagola u 10 transkripata osobnih narativa osoba s afazijom netečnog tipa, preuzetih iz Hrvatskog diskursnog korpusa govornika s afazijom (Kuvač Kraljević, Hržica i Lice, 2017). Navedeni je korpus dio korpusa jezičnih uzoraka osoba s afazijom AphasiaBank (Mac Whinne i sur., 2011) te su uzorci prikupljeni kroz strukturirane intervjue prema zadanom protokolu. Ispitanici su upitani da ispričaju nešto o svom govoru, moždanom udaru i oporavku te da prepričaju jedan važan životni događaj. Iz transkripata su izdvojeni glagoli te su grupirani prema kategoriji glagolskog vremena u dvije skupine: prošlo i neprošlo (sadašnjost i budućnost) vrijeme. Glagoli u prošlom vremenu grupirani su prema glagolskom vidu (svršeni, nesvršeni). Preliminarni rezultati statističke analize podataka potvrđuju pretpostavku da osobe s afazijom netečnog tipa u hrvatskom jeziku češće griješe u označavanju prošlog vremena u osobnim narativima te da je najčešća vrsta pogreške omisija pomoćnog glagola u složenim glagolskim oblicima.

During language acquisition, children often overgeneralize transparent morphological patterns to irregular words (e.g., bring to bringed instead of brought in English). Overgeneralized forms reveal strategies children use when confronted with morphological complexity. Highly inflected languages exhibit features of language acquisition not entirely consistent with findings in less inflected languages (Dressler 2005), especially English, which is usually taken as a model to draw conclusions about this phenomenon. The Croatian conjugation system displays various degrees of complexity based on the number of inflectional morphemes and on an elaborate system of stem changes. For most verbs the infinitive stem and the simple present stem differ in phonological features (cf. šet-a-ti.INF ‘to walk’ and šeć-e-m.PRS.1SG ‘I walk’). During early language development, children are likely to use overgeneralized forms interchangeably with adult-like forms (e.g., plesati.INF 'to dance' > plesam / plešem.1SG 'I dance'). In the research on the acquisition of inflectional morphology, two main factors influencing the occurrence of overgeneralization can be identified – token frequency (how often a child is exposed to a particular verb form) and phonological neighbourhood density or class size (the number of verbs with phonologically similar word stems and inflection). Not many studies investigated the influence of these factors in morphologically rich languages, especially on the acquisition of verbal morphology (cf. Kirjavainen, Nikolaev & Kidd 2012 ; Engelmann et al. 2019). The aim of this study was to investigate the production of overgeneralized verb forms in Croatian-speaking children aged 2 ; 6 to 5 ; 11 using a questionnaire in which parents report children’s overgeneralizations. We hypothesized that parents will report overgeneralized forms in all verb classes, but that the frequency of overgeneralizations will depend on the features of the input, i.e. it will be predicted by the frequency of verbs (higher rate of overgeneralizations for infrequent verbs) and class size (higher rate of overgeneralizations for verbs with smaller class size). A total of 36 verbs were investigated. For each verb, two items were created: the 'adult-like' form and the overgeneralized form. The frequency of verbs was calculated from the longitudinal child language corpus (CDS, Kovačević, 2002) and corpus of written adult language (hrWaC, Ljubešić & Klubička 2014). The class size was calculated from the same sources in lemmas, tokens and tokens of selected types. Parents were asked to indicate how often their child produces a particular verb form using a 5-point Likert scale. Our results show overgeneralized forms in child language for all verbs included in the questionnaire. We found a negative relationship between the proportion of overgeneralized forms and the overall verb frequency in both CDS and hrWaC. Verbs with a lower frequency have a higher proportion of overgeneralized forms. We also found a negative relationship between the class size in tokens (CDS) and both the frequency of overgeneralized forms and the proportion of overgeneralized forms. The frequency of overgeneralized forms and the proportion of overgeneralized forms are lower in larger classes. Our results are consistent with previous studies (e.g. Engelmann et al. 2019). Preschool children still resort to mechanisms of overgeneralizations in highly inflected language.

Svrha je knjige Čestotni rječnik hrvatskog dječjeg jezika: morfološki i razvojni oblici omogućiti uvid u morfološke oblike riječi koje upotrebljavaju djeca do treće godine života. Uz oblike i njihove morfološke opise navodi se i čestotnost svakog oblika. Rječnik je nastao na temelju stvarnih uzoraka govornog jezika troje djece govornika hrvatskog jezika u dobi od 13 do 28 mjeseci. Uzorci su objedinjeni u Hrvatskom korpusu dječjeg jezika (HKDJ - Kovačević, 2002), koji sadrži oko 136 000 pojavnica. Korpus je dio svjetske baze dječjeg jezika CHILDES, s otvorenim pristupom i predstavlja vrijedan izvor lingvističkih podataka o ranom jezičnom razvoju.

Proučavanje pripovijedanja koristan je izvor podataka o djetetovim jezičnim sposobnostima i omogućava otkrivanje teškoća u jezičnom razvoju (Aksu-Koç i Aktan-Ercyes, 2018). Postoje različiti strukturirani pripovjedni programi koji podržavaju razvoj pripovjednih vještina kod djece (McGregor, 2000, Hayward i Schneider, 2000). Kako bi se provjerila učinkovitost takvog programa, u jednoj je zagrebačkoj osnovnoj školi 2019. u sklopu projekta Višerazinski pristup govornom diskursu u jezičnom razvoju (HRZZ-UIP-2017-05-6603) proveden pripovjedni program Pričolovci (Hržica i sur., 2019). Dosadašnje spoznaje idu u prilog uspješnosti intervencija u obliku narativnih programa, premda su rezultati istraživanja nejednoznačni (Clancy, 1980 ; Michaels, 1981 ; Hayward i Schneider, 2000 ; Spencer i Slocum, 2010). Cilj je ovog rada provjeriti učinak provedene narativne intervencije na makrostrukturalnu razinu pripovijedanja djece rane školske dobi. Program je provođen tri puta tjedno unutar dvanaest tjedana, a razvijao je pripovjedne sposobnosti na mikro- i makrostrukturi te rječnik djece. Uzorak je činilo 55 djece kronološke dobi osam godina – 27 u tretmanskoj te 28 u kontrolnoj skupini. Za procjenu makrostrukture, odnosno gramatike priče korištene su fikcionalne priče Koze i Ptice unutar MAIN-a (Multilingual Assessment Instrument for Narrative, Gagarina i sur., 2012), materijala za procjenjivanje razine pripovijedanja. Za svakog sudionika su prije i nakon provođenja narativne intervencije procijenjeni makrostruktura priče i broj riječi kojima se izražavaju unutarnja stanja likova. Obrada podataka učinjena je u programu IBM SPSS Statistic – inačici 26. Nisu pronađene statistički značajne razlike između skupina, što znači da tretmanska skupina nije napredovala više od kontrolne. Dobivene je rezultate potrebno promatrati u svjetlu značajki dosadašnjih istraživanja: trajanje je dosad provedenih programa obično kraće i intenzivnije, uzorci sudionika manji i točno probrani (djeca s jezičnim poremećajima te rizične skupine nižeg SES-a, ili dvojezična djeca), a način praćenja učinka programa varira. Nalazi ovog istraživanja su vrijedni jer mogu poslužiti kao temelj za unaprjeđenje programa narativne intervencije te jezičnih intervencija općenito, koje su u Republici Hrvatskoj tek u začetku. Stoga je ostavljen prostor za daljnja istraživanja i nadograđivanje narativnih intervencija ove vrste. Buduća istraživanja svakako bi trebala istražiti i druge aspekte učinka programa, kao što su mjere mikrostrukture, rječnik, opće jezične sposobnosti te povratna informacija roditelja.

The acquisition of inflectional morphology is a complex process that is often accompanied by overgeneralization. Children overgeneralize because they cannot cope with the complexity of the system they are confronted with. The phenomenon of overgeneralizations is still poorly understood in highly inflected languages. The factors influencing overgeneralizations may be interrelated and it is difficult to identify their respective contributions. One of the factors known to influence overgeneneralizations is class size. Class size is the number of verbs with phonologically similar word stems that carry the same corresponding inflectional morpheme. These clusters of verbs correspond to conjugation or inflectional classes. Patterns with many "class neighbours" are predicted to have lower error rates. Previous findings of class size effects at the level of inflectional morphology could in principle be disguised as effects of morphological complexity. The aim of this study was to test overgeneralizations of verbs with different class sizes, controlling for frequency and morphological complexity. Data were collected using a parental questionnaire. The participants were parents of Croatian monolingual children aged 2 ; 6 to 5 ; 11 years (N=96). The verbs included in the questionnaire were taken from the Croatian Corpus of Child Language. Each verb was assigned to a class based on its morphological properties. For each class, we calculated the class size in tokens based on the child-directed speech. We also calculated the complexity for each class, measured by the number of phonemes that are different in the two verb stems. The results showed a facilitating effect of class size on the acquisition of inflectional morphology - verbs with fewer class neighbours have a lower frequency of correct forms and a higher rate of overgeneralization. Using different methods, both in terms of parental questionnaire and measures of frequency and class size, similar trends to previous studies were observed. Our findings support theoretical approaches that emphasise the properties of input as the primary source of information for child language development.

Motivacija za izradom rječnika proizlazi iz naših istraživanja. Puno smo se puta našli u potrebi za brzim i jednostavnim dohvatom podataka o čestoti riječi u dječjem jeziku. Izrada bilo kakvih eksperimentalnih jezičnih zadataka, dijagnostičkih testova, materijala za terapiju pa i bilo kakvog obrazovnog teksta namijenjenog djeci podrazumijeva kontrolu čestoće riječi koje su u njih uključene. Ovi su podatci sadržani u korpusima, ali iz njih nisu lako dohvatljivi, niti su razvrstani na načine prikladne za brzu uporabu. Ubrzo smo uvidjeli da bi od takvog rada mogla prosperirati i šira znanstvena i stručna zajednica. No, pripremiti podatke tako da su prikladni za široku primjenu zahtijevalo je posebna znanja. Stoga su u pripremi rječnika od samih početaka uključeni stručnjaci za računalnu lingvistiku, a metode i alati računalne lingvistike omogućili su prikladno označavanje i razvrstavanje korpusnih podataka. Dakle, ovi su rječnici plod petogodišnje interdisciplinarne suradnje autora različitih znanja i vještina koji su u ove dvije knjige uspješno objedinili viziju potrebnu za daljnji objektivni rad u području logopedije i lingvistike. Oba rječnika nastala su u sklopu dvaju projekata Računalni asistent za podršku pri unosu teksta osobama s jezičnim poremećajima (EU - Strukturni fond ; RC.2.2.08-050) u okviru Poziva Jačanje kapaciteta za istraživanje, razvoj i inovacije (2014-2016) te Višerazinski pristup govornom diskursu u jezičnom razvoju (UIP-2017-05- 6603) Hrvatske zaklade za znanost.

Spoken language corpora represent a comprehensive data source for investigating contact phenomena in multilingual interactions (e.g. code-switching, amount of languages used, number of languages used, etc.). Multilingual spoken corpus design including standard language and dialectal variation requires many methodological decisions and have been mainly built on elicitation tasks. However, such a methodology does not allow us to see variation in naturally occurring communication (Ruhi and Isik Tas, 2014). Therefore, spontaneous and high interaction speech are essential for representing the nature of discourse in real life situations (Čermak, 2009). The goal of this paper is to describe the sociolinguistic features of the Istrian multilingualism, aiming at shedding light over a potential correspondent contact linguistic model. For that purpose, we present the Istrian spoken multilingual corpus (in preparation), for which the conversational sampling method has been employed. The corpus’ recordings involve native speakers of Istrovenetian/Istriot and/or bilingual/native speakers of three different generations that are being representatively recruited in all the bilingual towns of the Istria County. In the current state, the corpus consists of 51 transcripts and sound recordings with 42 participants of different age (from 22 to 73 years). Current data reveals that the social interactions are determined by double diglossic/imperfectly polyglossic relationships and that the Istrian multilingual Italophone community is mostly a dialectophone community and/or vernacularophone one. Therefore, Italian has never achieved sociolinguistic capillarity and ethnolinguistic vitality. Istrovenetian and Istriot are ethnolinguistically (much) more vital (particularly the Istrovenetian koine’, as Istriotshows signs of language shift towards it) than the Italian language. Furthermore, Croatian and Chakavian, idioms of undeniable social prestige, may cover almost all the entire range of functional communicative needs, serving as constant resources for the regular use of codeswitching.

The acquisition of inflectional morphology can be particularly challenging for children and often leads to the production of incorrect forms, including overgeneralization of a given rule to irregular word forms (e.g., bring to bringed instead of brought in English). Overgeneralized forms reflect the complexity of the morphological system and reveal the strategies children use when confronted with it. Highly inflected languages exhibit some features related to early language acquisition that are not entirely consistent with findings in less inflected languages (Dressler 2005), especially English, which is usually taken as a model to draw inferences about this phenomenon. The Croatian conjugation system displays different degrees of complexity based not only on the number of inflectional morphemes but also on an elaborate system of stem changes. For most verbs it is not sufficient to attach the corresponding affix to the infinitive stem because the infinitive stem and the simple present stem differ in phonological features (cf. šet-a- ti.INF ‘to walk’ and šeć-e-m.PRS.1SG ‘I walk’). During early language development, children are likely to use overgeneralized forms to overcome this complexity. These forms are often used interchangeably with ‘adult-like’ forms (e.g., plesati.INF ‘to dance’ > plesam / plešem.PRS.1SG ‘I dance’), that is, children use more than one form per slot in the morphological paradigm. Both the correct and overgeneralized forms are attested simultaneously, but not with the same frequency. In the research on the acquisition of inflectional morphology, two main factors influencing overgeneralization can be identified – token frequency (how often a child is exposed to a particular verb form) and phonological neighbourhood density or class size (the number of verbs with phonologically similar word stems bearing the same corresponding inflectional morpheme). However, not many studies investigated the influence of these factors in morphologically rich languages, especially on the acquisition of verbal morphology (cf. Kirjavainen, Nikolaev & Kidd 2012 ; Engelmann et al. 2019). The aim of this study was to investigate the production of overgeneralized verb forms in preschool Croatian- speaking children aged 2 ; 6 to 5 ; 11 using a questionnaire in which parents report overgeneralizations in the language of their children. We hypothesized that parents will report overgeneralized forms in all verb classes in which the stem changes, but that the frequency of overgeneralizations will depend on the features of the input, i.e. it will correlate with the frequency of verbs (higher rate of overgeneralizations for infrequent verbs) and class size in lemmas, tokens and tokens of selected verb types (higher rate of overgeneralizations for verbs with smaller class size). To date, studies have used a corpus-based method to retrieve overgeneralizations in child language, which has had limited success in capturing this phenomenon due to the low density of language sampling and low-frequency phenomena such as overgeneralizations tend to be underrepresented. A parental questionnaire could provide more precise information about the overgeneralization of verb forms. The verbs for the questionnaire were selected from the longitudinal Croatian corpus of child language (Kovačević 2002) to ensure that they occur in child-directed speech (CDS). To refine the selection, we used two lexical databases that contain information on the estimated age of acquisition and subjective word frequency retrieved from native speakers. A total of 36 verbs were included in the study. For each verb, two items were created: the ‘adult-like’ form and the overgeneralized form. The frequency of verbs was calculated from the longitudinal child language corpus (CDS) and corpus of written adult language (hrWaC, Ljubešić & Klubička 2014). We obtained two types of frequency information for each of the preselected verbs: the overall frequency of all morphological types of a verb and the frequency of specific morphological types of a verb that were central to our study. The class size was calculated from the same sources in lemmas, tokens and tokens of selected types. Parents were asked to indicate how often their child produces a particular verb form using a 5- point Likert scale. Altogether, 87 parents completed the online questionnaire, therefore, we obtained data for 87 children. The results showed that parents reported overgeneralized forms in child language for all verbs included in the questionnaire. We found a negative relationship between the proportion of overgeneralized forms and the overall verb frequency in both CDS and hrWaC. Verbs with a lower frequency have a higher proportion of overgeneralized forms. We also found a negative relationship between the class size in tokens (CDS) and both the frequency of overgeneralized forms and the proportion of overgeneralized forms. The same results were found for the class size in specific morphological types of a verb (hrWaC). The frequency of overgeneralized forms and the proportion of overgeneralized forms are lower in larger classes, that is, in classes that are more frequent in the language surrounding children. The frequency and class size showed better correlations with the measures in the questionnaire when calculated on tokens of specific morphological forms, i.e. those where overgeneralization is expected. Our results are consistent with previous studies that have confirmed a facilitating effect of frequency and class size on the acquisition of inflectional morphology (e.g., Kirjavainen, Nikolaev & Kidd 2012 ; Engelmann et al. 2019). The present study reveals that preschool children still resort to the mechanism of overgeneralizations to overcome the complexity of verbal morphology.

Označavanje vremena u narativima jedna je od sposobnosti koja se razvija kroz djetinjstvo (Silva, 1991), a jezični sustavi označavanja vremena i vida doprinose učvršćivanju i organizaciji informacija u diskursu. Vremenski odnosi između situacija i događaja u diskursu uspostavljaju se glagolskom morfologijom, priložnim oznakama vremena i vremenskim konektorima. Osim uspostavljanja vremenskih odnosa, kategorija vida se povezuje i s dovođenjem informacija u prednji plan. Glagoli perfekivnog vida povezuju se s informacijama u diskursu koje su u prednjem planu. Priložne oznake vremena i vremenski konektori služe integraciji informacija u narativni kontekst (Gernsbacher, 1996), a kao kohezivna sredstva mogu izražavati vremenski odnos uzastopnosti ili istovremenosti (djelomične ili potpune). Osim toga za nastavljanje radnje mogu koristiti svenamjenske koordinacijske veze poput veznika ‘’i’’. Prema Hickmann (2003) djeca imaju problema s linearnom organizacijom prostorno- vremenskih informacija i tipološke karakteristike jezika utječu na njihovu organizaciju diskursa. Cilj ovog rada bio je opisati označavanje vremena na Hrvatskom pripovjednom korpusu (Hržica i Roch, 2020) kroz analizu upotrebe glagolskog vremena i vida te priložnih oznaka vremena i vremenskih konektora. Prema podacima iz dosadašnjih istraživanja postavljene su tri hipoteze: H1) u narativima šestogodišnjaka prevladava sadašnje glagolsko vrijeme, H2) šestogodišnjaci upotrebljavaju više glagola perfektivnog vida i H3) češće izražavaju odnos uzastopnosti a najčešće koriste veznike koji nastavljaju radnju bez izražavanja vremenskog odnosa. U istraživanju je analizirano 40 transkripata narativa šestogodišnjaka. Podaci u korpusu prikupljeni su hrvatskom inačicom Višejezičnog instrumenta za ispitivanje pripovijedanja (MAIN ; Hržica i Kuvač Kraljević, 2012). Transkriptima se pristupalo online putem baze CHILDES u sustavu TalkBank. Iz transkripata izdvojeni su svi pojedinačni glagoli, vremenski veznici, vremenski veznički izrazi i vremenski prilozi. Glagoli su analizirani prema glagolskom vremenu i vidu, a priložne oznake vremena i vremenski konektori prema vrsti vremenskog odnosa. Glagoli su razvrstani prema kategoriji glagolskog vremena (Barić i sur., 2003) na: sadašnje (prezent), prošlo (perfekt, imperfekt, aorist i pluskvamperfekt) i buduće (futur prvi i drugi). Glagoli prema vidu dijele se na svršene i nesvršene te je ta raspodjela učinjena za sve glagole u prošlom vremenu. Priložne oznake vremena i vremenski konektori podijeljeni su u tri kategorije: sekvencijalna, regionalna i svenamjenska sredstva prema Hickmann (2003). Navedene kategorije potom su analizirane s obzirom na pojavnost unutar dječjih narativa. Preliminarni rezultati istraživanja potvrđuju H2 i djelomično H3. Rezultati ovog istraživanja djelomično potvrđuju dosadašnja istraživanja, a uočene razlike mogle bi biti posljedica tipoloških karakteristika hrvatskog jezika.

Temeljni je cilj nastave Hrvatskoga jezika u osnovnoj školi ovladati jezično-komunikacijskom kompetencijom koja podrazumijeva usvajanje funkcionalnoga znanja, odnosno uporabno znanje jezika na svim normativnim razinama: gramatičkoj, leksičkoj, stilskoj, pravopisnoj i pravogovornoj. Budući da je razvoj rječnika kontinuiran proces koji traje gotovo cijeli život (Peretić, Padovan i Kologranić Belić, 2015), u tradicionalnoj nastavi jezika nisu se koristile posebne strategije za poticanje njegova razvoja. No, novija su istraživanja pokazala kako se osnovnoškolci i na pismenoj i na usmenoj razini izražavaju sve lošije, s puno poštapalica, ponavljanja, upotrebljavanja riječi u pogrešnim kontekstima ili imaju poteškoće s razumijevanjem pročitanoga teksta (Aladrović Slovaček i Rimac Jurinović, 2021). Također, postoje posebno osmišljeni programi kojima je cilj razvoj specifičnoga rječnika (usvajanje prijedloga, imenovanje boja i sl.) i koji se najčešće realiziraju u predškolskoj dobi te se djeca tako pripremaju za polazak u školu i početak procesa opismenjavanja, no oni su vrlo rijetki i često nedostatni. Sve navedeno utjecalo je i na promjenu u sustavu obrazovanja pa je u Kurikulumu nastavnoga predmeta Hrvatski jezik (2019) u području jezika i komunikacije poseban ishod namijenjen upravo poticanju razvoja rječnika. Budući da je razvoj rječnika prediktor razumijevanja pročitanoga, pripovijedanja, sposobnosti stvaranja pisanoga teksta te kasnijih akademskih vještina, napravljeno je eksperimentalno istraživanje u kojemu je primijenjen program poticanja rječnika „Pričolovci“ s učenicima prvoga razreda osnovne škole. Cilj je stoga ovoga rada opisati jezične osobitosti riječi (dužina, čestotnost, glasovna struktura, funkcija u rečenici, vrsta riječi) kojima su učenici kroz program ovladali značajno bolje te prikazati koji su izvanjezični čimbenici utjecali na rezultate, poput: spola djeteta, obrazovanja roditelja, broja članova kućanstva, čestotnosti čitanja i pripovijedanja u roditeljskom domu, te dobi roditelja. Očekuje se da će usvajanje značenja riječi ponajviše ovisiti o njezinoj frekventnosti dok će od izvanjezičnih čimbenika najvažniju ulogu imati čestotnost čitanja i pripovijedanja, odnosno čitalačke navike koje se njeguju u obitelji.

The aim of this study was to compare the morphological complexity in a corpus representing the language production of younger and older children across different languages. The language samples were taken from the Frog Story subcorpus of the CHILDES corpora, which comprises oral narratives collected by various researchers between 1990 and 2005. We extracted narratives from typically developing, monolingual, middle- class children. Additionally, samples of the Lithuanian language, collected according to the same principles, were added. The corpus comprises 249 narratives evenly distributed across eight languages: Croatian, English, French, German, Italian, Lithuanian, Russian and Spanish. Two subcorpora were formed for each language: a younger children corpus and an older children corpus. Four measures of morphological complexity were calculated for each subcorpus: Bane, Kolmogorov, Word entropy and Relative entropy of word structure. The results showed that younger children's corpora had lower morphological complexity than older children's corpora for all four measures for Spanish and Russian. Reversed results were obtained for English and French, and the results for the remaining four languages showed variation. Relative entropy of word structure proved to be indicative of age differences. Word entropy and relative entropy of word structure show the potential to demonstrate typological differences.

The resolution of third-person pronouns has attracted substantial interest in psycholinguistics in recent decades. The most influential claim is that pronoun resolution is guided by the saliency of syntactic roles. In pro- drop languages such as Italian and Croatian, the null pronoun often refers to the most salient subject antecedent, and the overt pronoun to the less salient object antecedent (Carminati, 2002 ; Kraš, 2008). In other pro-drop languages such as Chinese, however, both overt pronouns and null pronouns refer to the subject (Li, Mak, & Sanders, 2016). Moreover, pronoun resolution might be subject to the influence of implicit causality (IC) bias (Fedele & Kaiser, 2015). For a sentence with an IC verb (e.g., John angered/praised Peter because...), the semantic bias of the verb indicates which referent (NP1 or NP2) is likely to be the cause and therefore is the preferred agent of the second clause. The aim of this study is to investigate how implicit causality interacts with pronoun-type bias in two typologically distant pro- drop languages, Croatian and Chinese. We hypothesised that null pronouns are more acceptable in NP1-biased sentences than in NP2- biased sentences in Croatian, whereas overt pronouns will be more acceptable in NP2-biased sentences. In contrast, both null pronouns and overt pronouns should be better accepted in NP1- biased sentences in Chinese. The participants were adult monolingual speakers of Croatian (n=76) and Chinese (n=67). They completed an acceptability judgement task in an online questionnaire. Twenty sentences in 4 conditions were constructed with a similar structure in both languages. We manipulated the direction of the verb bias (IC-NP1 vs IC-NP2) and the pronoun type (null vs overt). The strength of the verb bias was held constant, with minimum bias values of 70%. Participants were instructed to rate the naturalness of the sentences on a Likert scale ranging from 1 (completely unacceptable) to 6 (completely acceptable). Data was analysed using mixed-effect linear regression models. Results show that null pronouns are in general more acceptable than overt pronouns in Croatian, while the opposite was observed in Chinese. The effect of verb bias was similar, however: for overt pronoun sentences in both languages, NP2-biased sentences were more acceptable than NP1 ones. Unexpectedly, for sentences with null pronouns, NP1 and NP2 sentences were of the same acceptability. There is a cross-linguistic difference between Croatian and Chinese in pronoun-type bias. Croatian is more dependent on syntactic features, with the null pronoun sentences being more natural to native speakers. Chinese pro-drop has been considered more discourse-dependent, and thus, sentences with null pronouns are in general less acceptable than those with overt pronouns. For null pronoun sentences, the implicit causality verb bias is strong enough to overwrite the null pronoun bias. For sentences with overt pronouns, the verb bias does not play a similar role. We will discuss the implications of these results for future online processing experiments.

In the course of acquiring verbal morphology, children develop certain rules based on forms they have already acquired and extend them to new forms. However, not all verb forms are predictable and cannot be produced simply by a general rule. There is evidence that children often overgeneralize. Overgeneralization consists of extending a regular pattern to an irregular form. For example, goed instead of went in English. These errors are a temporary side effect of the early stages of language acquisition and should disappear as children get older. There are two main factors that are considered to influence overgeneralization - token frequency and class size. Not many studies have closely examined the influence of these factors on overgeneralization, and some conflicting results have been found. In the present study, we aimed to investigate the relationship between overgeneralized forms and specific characteristics of the input. Participants were 92 parents of Croatian monolingual children aged 2 ; 6 to 5 ; 11 years, recruited through social media. We used the parental questionnaire in which parents were asked to estimate how often their child produced a particular form of the verb. We selected verbs with stem changes from six conjugation classes, controlling for all relevant characteristics (i.e., age of acquisition, frequency). Frequency and class size were calculated from two sources: child-directed speech in the Croatian Corpus of Child Language and the Croatian Web Corpus. The results showed that verbs that are less likely to be overgeneralized have a higher frequency and belong to larger classes, i.e., classes that are more frequent in the language surrounding children. Our novel method, in which data are collected from parents, shows similar trends to other studies (corpus-based or experimental), suggesting that children acquire morphological systems gradually in a way that is highly sensitive to the properties of the input. This supports theoretical approaches that emphasize input properties as the primary source of information for child language development.

Spontani govor govornika urednoga jezičnoga i govornoga statusa rijetko je potpuno tečan. Dvije su temeljne skupine netečnosti. Mucajuće netečnosti javljaju se unutar riječi i karakteristične su za govorne poremećaje. Normalne netečnosti pojavljuju se između riječi i nisu nužno povezane s govornim poremećajima. Većina istraživanja pokazuje da se u uobičajenomu spontanome govoru na 100 riječi javlja šest netečnosti. Mucajuće netečnosti mogu biti prisutne i kod tečnih govornika, iako su rjeđe od normalnih. Kao najčešće vrste normalnih netečnosti kod tečnih govornika navode se umetanja, revizije te, od mucajućih netečnosti, ponavljanje jednosložnih riječi. Cilj je ovoga istraživanja ponuditi opis netečnosti odraslih govornika hrvatskoga jezika pri pripovijedanju. Zanimao nas je ukupan broj netečnosti, vrste netečnosti te u kojim se omjerima pojavljuju. Rezultati su sukladni ranijim istraživanjima jer je prosječan broj 6,5 netečnosti na 100 riječi. Sudionici su najviše proizvodili normalne netečnosti, najčešće umetanja popunjivača te revizije fraza i revizije riječi. Od mucajućih netečnosti najviše je bilo duljenja glasova, a nešto manje ponavljanja jednosložnih riječi. Iako je mucajućih netečnosti puno manje, pojavile su se sve vrste osim zastoja ili bloka.

Self-assessment ratings across single or multiple linguistic domains represent one of the methods for assessing bilingual language knowledge (Treffers-Daller, 2015). However, self-assessment scores alone might not be the most appropriate assessment tool (Zell & Krizan, 2014, Treffers- Daller 2015), thus more comprehensive indices are needed, combining subjective and objective measures (Treffers­Daller and Korybski, 2015). This approach is being developed in the present study, combining self-assessment data and language sample analysis, aiming at providing an insight into the multilingual Istrian reality. Measures of lexical diversity and syntactic complexity have been used to determine whether they reflect self-assessment of language proficiency of 30 functional multilingual speakers of Croatian, Chakavian, Italian and Istrovenetian across four language skills. Participants’ data were extracted from the Istrian spoken multilingual corpus (Hržica, Poropat Jeletić, Moscarda Mirković, in preparation), created using the conversational sampling method. Measures of lexical diversity (moving-average type-token ratio - MATTR ; Covington & McFall, 2010) and syntactic complexity (mean length of communication unit - MLCU) were calculated using the Computerized Language Analysis (CLAN – MacWhinney 2000). The average self-assessment score positively and moderately correlated with MLCU (r(30)=.513, p

Linguistic adaptations and graphic adjustments, such as highlighting keywords, facilitate reading and comprehension, which can be used as a strategy to support reading tasks, especially for people with dyslexia (Rello et al., 2014). There is a great deal of research on natural language processing that focuses on the extraction of keywords for the purpose of text summarisation (D’Avanzo and Magnini, 2005). In the education of people with dyslexia, teachers are encouraged to highlight keywords to make texts more accessible (Hargreaves, 2007). However, highlighting keywords is often based on the personal choices rather than on general, scientifically based rules. To date, little attention has been paid to the effectiveness of highlighting keywords to facilitate the readability and comprehensibility of texts. The aim of this study was to investigate the general rules experts apply when highlighting keywords in texts. The participants were speech and language pathologists (SLP) who are experts in adapting materials for people with dyslexia (n=50). The control group consisted of other highly educated individuals from various academic disciplines (n=50). The participants were instructed to highlight keywords in the text. The material was a text fragment from the novel The Return of Filip Latinovicz written by Miroslav Krleža, which was used as the starting text for the Croatian language essay in the 2019 State Matura exam. This study is ongoing ; we will discuss the possible differences in the number and type of keywords highlighted by SLPs and the control group.

Pripovijedanje podrazumijeva sposobnost proizvodnje i razumijevanja uzročno i vremenski povezanih događaja priče organiziranih oko glavne teme. Kada se promatra razvoj pripovjednih sposobnosti, u obzir se uzima sposobnost povezivanja događaja priče na globalnoj makrostrukturnoj razini i lokalnoj mikrostrukturnoj razini. U pripovijedanju se fiktivnih priča na makrorazini redovito promatra gramatika priče (engl. story grammar), ponekad nazivana i struktura priče (engl. story structure). To je univerzalni okvir priče prema kojem se pripovijeda, a sadrži nekoliko sastavnica (situaciju (zajednička za priču), uvodni događaj, cilj, pokušaj, ishod (za svaku epizodu)). Unutar svake epizode priče mogu se ostvariti samo neki ili svi elementi, što se naziva strukturna složenost (engl. structural complexity). Potpune epizode s nizom cilj – pokušaj – ishod upućuju na najviši stupanj složenosti priče i pokazatelj su napretka dječje pripovjedne sposobnosti. Dječje priče postaju sve složenije i koherentnije kako napreduju njihove jezične i kognitivne sposobnosti. Trogodišnjaci urednoga jezičnog razvoja imaju neku ideju o tome što je priča. Međutim, pripovjedne se sposobnosti razvijaju sporo jer dječje razumijevanje priča i jezičnih vještina koje im omogućavaju da tvore koherentan i kohezivan tekst također se razvijaju sporo, sve do adolescencije (Berman i Slobin 1994, Berman i Verhoeven 2002). To znači da dječje priče pružaju mnogo mogućnosti za proučavanje kasnijega jezičnog razvoja. Cilj je ovog istraživanja bio utvrditi razlikuju li se djeca različite dobi u gramatici priče te u strukturnoj složenosti. Za istraživanje su iz Hrvatskoga pripovjednog korpusa MAIN (Hržica i Roch 2020) izdvojeni jezični uzorci: 48 predškolske djece (Mdob = 6 ; 2 ; SDdob = 0, 39, min = 5 ; 1, max = 6 ; 9) te 50 djece školske dobi (Mdob = 8 ; 3, SDdob = 0, 45, min = 7 ; 1, max = 9). Priče su prikupljene hrvatskom inačicom instrumenta Multilingual Assessment Instrument for Narratives (MAIN, Hržica i Kuvač Kraljević 2021). MAIN je prikladan za procjenu pripovjednih sposobnosti djece od 3 do 10 godina. Struktura priče i strukturna složenost procijenjene su za svakog sudionika. Na mjeri gramatike priče dobivena je statistički značajna razlika (p < .05) između skupina djece predškolske dobi (M = .195, SD = .459) i školske dobi (M = .54, SD = 734) (p < .05). Rezultati strukturne složenosti pokazuju da djeca školske dobi tek ponekad proizvode potpune epizode, a rjeđe ih proizvode djeca predškolske dobi.

Jezikovne prilagoditve in oblikovne prilagoditve, kot je poudarjanje ključnih besed, olajšajo branje in razumevanje, kar se lahko uporablja kot strategija za podporo pri bralnih nalogah, zlasti pri osebah z disleksijo (Rello idr., 2014). Na področju obdelave naravnega jezika obstaja veliko raziskav, ki se osredotočajo na pridobivanje ključnih besed za namen povzemanja besedila (D‘Avanzo in Magnini, 2005). Pri izobraževanju oseb z disleksijo se učitelje spodbuja, naj poudarjajo ključne besede, da bi bila besedila dostopnejša (Hargreaves, 2007). Vendar izbira ključnih besed pogosto temelji na osebnih odločitvah in ne na splošnih, znanstveno utemeljenih pravilih. Doslej je bilo učinkovitosti poudarjanja ključnih besed za lažjo berljivost in razumljivost besedil posvečeno le malo pozornosti. Namen te študije je bil raziskati splošna pravila, ki jih strokovnjaki uporabljajo pri poudarjanju ključnih besed v besedilih. Sodelovali so logopedi, ki so na Hrvaškem strokovnjaki za prilagajanje gradiv osebam z disleksijo (n = 50). Kontrolno skupino so sestavljali drugi visoko izobraženi posamezniki z različnih akademskih področij (n = 50). Vsem udeležencem je bilo naročeno, naj v besedilu označijo ključne besede. Gradivo je bil odlomek besedila iz romana Vrnitev Filipa Latinoviča, ki ga je napisal Miroslav Krleža, in je bil uporabljen kot izhodiščno besedilo za esej iz hrvaškega jezika na državni maturi leta 2019. Ta študija je v teku ; razpravljali bomo o morebitnih razlikah v številu in vrsti ključnih besed, ki so jih izpostavili logopedi in kontrolna skupina. / Linguistic adaptations and graphic adjustments, such as highlighting keywords, facilitate reading and comprehension, which can be used as a strategy to support reading tasks, especially for people with dyslexia (Rello et al., 2014). There is a great deal of research on natural language processing that focuses on the extraction of keywords for the purpose of text summarisation (D’Avanzo and Magnini, 2005). In the education of people with dyslexia, teachers are encouraged to highlight keywords to make texts more accessible (Hargreaves, 2007). However, the choice of keywords is often based on personal choices rather than on general, scientifically based rules. To date, little attention has been paid to the effectiveness of highlighting keywords to facilitate the readability and comprehensibility of texts. The aim of this study was to investigate the general rules experts apply when highlighting keywords in texts. The participants were speech and language pathologists (SLP) who are experts in adapting materials for people with dyslexia (n=50). The control group consisted of other highly educated individuals from various academic disciplines (n=50). The participants were instructed to highlight keywords in the text. The material was a text fragment from the novel The Return of Filip Latinovicz written by Miroslav Krleža, which was used as the starting text for the Croatian language essay in the 2019 State Matura exam. This study is ongoing ; we will discuss the possible differences in the number and type of keywords highlighted by SLPs and the control group.

U rječniku se nalaze riječi evidentirane u jezičnom razvoju djece do početka treće godine života. Uz čestotu riječi, vidljiv je i morfološki opis natuknice. Podaci su organizirani prema nekoliko kriterija, uključujući čestotu, abecedu, vrstu riječi te dob pojavljivanja.

Purpose: The MultiDis application is a new, web- based application designed for the analysis of spoken and written language samples, which provides information about the language abilities of children and adults, thus facilitating language assessment. The aim of this paper is to present the selection, implementation, and testing of language measures in the MultiDis application. We will present the application, the process of selecting the measures we implemented, the language resources needed to calculate them, and the results of testing. MultiDis is currently being developed for Croatian, but it could be scaled up for multilingual analysis. Design/methodology/approach: Language samples can be analyzed according to several dimensions, such as productivity, lexical diversity, and syntactic complexity. A set of (semi-) automatic measures has been selected to assess language abilities (e.g., number of lemmas, mean-average type-token ratio, mean length of communication unit). The next step was the integration of an open-source Python library for lemmatization, part-of-speech tagging, and syntactic parsing (Stanza ; Qi et al., 2020). To test whether these tasks and the subsequent calculation of language measures can be successfully performed on spoken language samples, we uploaded 150 short narrative samples produced by children as a result of a storytelling task. Findings: Lemmatization and part-of-speech tagging are fairly accurate (>85% of cases), as they do not interfere with the calculation of the currently implemented measures of productivity and lexical diversity. The process of syntactic parsing has been an obstacle that is currently being resolved. Research limitations/implications: The MultiDis web application is still under development, although the current version fulfils its main purpose – it allows for (semi-)automatic spoken language analysis. Practical implications: There is an increasing awareness of the importance of language sample analysis as a complementary method in language assessment. The time needed for transcription and the linguistic knowledge required for manual analysis are considered to be the main obstacles to its implementation (Pezold et al., 2020). Therefore, the development of a tool for automatic calculation of language measures such as the MultiDis application could make naturalistic language assessment more feasible. Originality/Value: The value of this study lies in proposing a new application for lemmatization and part-of-speech tagging that allows for more reliable calculation of measures of productivity, lexical diversity, and syntactic complexity. Selecting appropriate measures for language assessment is a challenging task because there are many available. Implementing language technologies developed for large bodies of written texts to spoken language is also challenging. Success in some parts of automated tagging (lemmatization and part-of-speech tagging) allows for the reliable calculation of measures of productivity and lexical diversity. Future work on syntactic parsing will lead to the successful implementation of measures of syntactic complexity.

Measures of narrative microstructure are used to assess linguistic features in narrative samples. The most common are measures of production (reflecting how much is produced), measures of syntactic complexity (reflecting the complexity of syntactic clauses in the narrative process) and measures of lexical diversity (reflecting vocabulary variety in narration process) (e.g., Mäkinen et al., 2014). These measures can provide insight into different linguistic levels of narrative sample and can reflect the basic linguistic skills of the narrator. Some previous studies have shown a positive correlation between comprehension and production of various syntactic structures and measures of syntactic complexity (Frizelle et al., 2018). Also, some studies have reported a moderate positive correlation between tests of vocabulary knowledge and lexical diversity in narrative (Hržica & Roch, 2019 ; Wood et al., 2018). These results support the use of narrative microstructure measures to better understand children’s language abilities, including the language abilities of children with language and communication disorders (Botting, 2002 ; Watkins et al., 1995). In Croatia, the relationship between vocabulary comprehension, syntactic structures and narrative microstructure in children with language and speech disorders have not been investigated. The aim of this study was to examine the relationship between measures of syntactic complexity and comprehension of syntactic structures, and measures of lexical diversity and vocabulary comprehension. We expect that measures of syntax, as well as measures of vocabulary, will correlate in children with speech and language disorders, as was shown for children with typical language development in previous studies. In this ongoing study up so far twenty- nine children (83% boys) with speech and language disorders (developmental language disorder, childhood apraxia of speech and social communication disorder) were tested. They narrated a story elicited by a sequence of pictures. Test for Reception of Grammar (TROG-2:HR) and Peabody picture vocabulary test (PPVT-III-HR) were administered. Narrative samples were transcribed in the program CLAN. Microstructure measures were calculated: mean length of communication unit (MLCU ; syntactic complexity), moving average type-token ratio (MATTR) and vocabulary measure D (lexical diversity). Spearman correlation coefficients were calculated in SPSS 23.0. There were no significant correlations between the results on PPVT-III-HR test and measures of lexical diversity (ρMATTR = - .09, p = .65 ; ρD = - .12, p = .52) or the results on TROG-2:HR test and syntactic complexity measure (ρMLCU = .05, p = .80). Interpretation of these results should take into account several factors that potentially affect the results (e.g., primary difference in population sample, narrative elicitation task, and sample size). Mjerama mikrostrukture u pripovijedanju procjenjuju se jezične karakteristike pripovjednog uzorka. Najčešće su mjere proizvodnje (odražavaju koliko je proizvedeno), mjere sintaktičke složenosti (odražavaju kompleksnost sintaktičkih struktura rabljenih u pripovijedanju) te mjere rječničke raznolikosti (odražavaju raznovrsnost rječnika koji se rabio u pripovijedanju) (npr. Mäkinen i sur., 2014). Mjere mikrostrukture, dakle, pružaju uvid u različite jezične razine pripovjednog uzorka te tako, posredno, mogu odraziti temeljne jezične sposobnosti pripovjedača. Dosadašnja su istraživanja pokazala kako postoji pozitivna povezanost ovladanosti različitim sintaktičkim strukturama i mjerama sintaktičke složenosti (Frizelle i sur., 2018). Također, postoje istraživanja koja upućuju na umjerenu pozitivnu povezanost rezultata na testovima rječničkog znanja i rječničke raznolikosti u pripovijedanju (Hržica i Roch, 2019 ; Wood i sur., 2018). Takvi rezultati govore u prilog uporabe pripovjednih mjera u boljem razumijevanju jezičnih sposobnosti djece, između ostaloga i djece s jezičnim ili komunikacijskim teškoćama (Botting, 2002 ; Watkins i sur, 1995). U Hrvatskoj još nisu provedena istraživanja povezanosti razumijevanja rječnika i sintaktičkih struktura te obilježja mikrostrukture u pripovijedanju djece jezičnim i govornim poremećajima. Cilj je ovog istraživanja ispitati povezanost između mjera sintaktičke složenosti i razumijevanja sintaktičkih struktura te mjera rječničke raznolikosti i razumijevanja riječi. Očekuje se da će mjere usmjerene na sintaksu te one usmjerene na rječnik biti međusobno povezane kod djece s različitim poremećajima jezika i govora, kao što se to pokazalo kod djece urednog jezičnog razvoja. U istraživanju koje je u tijeku do sada je sudjelovalo 29 djece (83% dječaka) s jezičnim i govornim poremećajima (razvojni jezični poremećaj, govorna apraksija, artikulacijski poremećaj te socijalnokomunikacijski poremećaj). Ispričali su priču prema slikovnom predlošku Jež i gljiva. Ispitani su testovima Test razumijevanja gramatike (TROG-2:HR) te Peabody slikovni test rječnika (PPVT-III-HR). Ispričane priče transkribirane su u programu CLAN, pomoću kojega su izračunate mjere mikrostrukture: prosječna duljina komunikacijske jedinice (MLCU ; sintaktička složenost), pomični prosječni omjer različnica i pojavnica (MATTR) te mjera rječnika D (rječnička raznolikost). U programu SPSS 23.0 izračunati su Spearmanovi koeficijenti korelacije između navedenih mjera mikrostrukture i rezultata na standardiziranim jezičnim testovima. Nije dobivena značajna povezanost između rezultata na testu PPVT-III-HR i mjera rječničke raznolikosti (ρMATTR = - .09, p = .65 ; ρD = - .12, p = .52), kao ni između rezultata na testu TROG-2:HR i prosječne duljine komunikacijske jedinice (ρMLCU = .05, p = .80). Prilikom tumačenja rezultata u obzir je potrebno uzeti nekoliko čimbenika koji su mogli utjecati na dobivene korelacije: značajke ispitivane populacije (prvenstveno razlike među sudionicima), materijal za poticanje pripovijedanja te veličinu uzorka.

In the acquisition of verbal morphology, children develop certain rules based on the forms they have already acquired and extend them to produce new forms. This is an important step because generalizations can increase the number of forms acquired. However, not all verb forms are predictable and cannot be easily generated by a general rule. This gives rise to unacceptable so-called overgeneralizations. Children overgeneralize because they cope with the complexity of the system they are confronted with. To figure out the patterns of overgeneralizations, one must examine what increases or decreases the likelihood of overgeneralization. Two factors can be discerned: frequency and class size. Most of the studies investigated the effects of frequency and class size in the domain of English past tense. But there is still a debate on how exactly these factors affect the acquisition of verbal morphology in highly inflected languages like Croatian. Morphological features of Croatian make it very suitable for producing a large number of deviations from canonicity within a conjugation paradigm made by children during the early stages of language acquisition. Parental questionnaires are designed and (often) standardized to obtain a general measure of language development. However, they have not been used to investigate more specific phenomena like overgeneralizations in child language. In this study, we aim to test the relationship between overgeneralized forms and specific properties of input (frequency and class size). Participants were parents of Croatian monolingual children ages 2 ; 6 to 5 ; 11 (N=87). An online parental questionnaire was developed. Parents were asked to indicate how often their child produces a particular verb form using a 5-point Likert scale. Different measures of frequency and class size were used to examine the relationship between overgeneralized forms and the properties of the input. Our results are in line with studies that showed the facilitative effect of frequency and sometimes class size on the acquisition of inflectional morphology. Similar tendencies were observed by using a different methodology, both regarding the parental questionnaire and regarding the usage of diverse measures of frequency and class size.

Analiza jezičnih uzoraka (pisani ili govoreni tekst nekog govornika, obično potaknuti predloškom, primjerice usmeno pripovijedanje ili pisani esej) pruža podatke o usvajanju prvog i inog jezika te o jezičnim sposobnostima, što znači da se može upotrebljavati kao dio standardne jezične procjene. Mogu je rabiti nastavnici inog jezika, logopedi, nastavnici u razrednoj nastavi te profesori prvog jezika. No najviše se upotrebljava u usvajanju prvog i drugog jezika, što u stručnom što u znanstvenom radu. U oba se područja upotrebljavaju jednake ili slične mjere, pri čemu je u usvajanju prvog jezika naglasak na govornim uzorcima, a u usvajanju drugog jezika na pisanima. Analiza jezičnih uzoraka u nekim je zemljama dio standardne jezične procjene, no često nisu prepoznate sve prednosti takve procjene ili nisu dostupni alati za njezinu provedbu. Tijekom posljednjih desetljeća prošlog stoljeća razvijeni su različiti programi za transkripciju i analizu jezičnih uzoraka (pregled: Pezold i sur., 2020). Uporaba takvih programa može biti složena te često ostaje ograničena na uporabu u znanstvenoj zajednici. Posljednjih se godina razvijaju mrežni resursi za automatsku analizu jezičnih uzoraka (primjerice, Coh-Metrix ; Greasser i sur., 2004). Uz tradicionalne mjere (primjerice, prosječna duljina iskaza, omjer različnica i pojavnica) nude i naprednije mjere temeljene na jezičnim tehnologijama (morfološko označavanje, sintaktičko označavanje, semantičko označavanje…). Takve su aplikacije razvijene prvenstveno za engleski jezik ili druge veće svjetske jezike, ali često imaju tek djelomično otvoren pristup. Aplikacija MultiDis ima otvoren pristup te je dostupna svim korisnicima, a uporabom dostupnih jezičnih tehnologija (Peng i sur., 2020) prilagođenih za hrvatski jezik omogućena je morfosintaktička osnova za računanje mjera u širokom rasponu. Aplikacija uključuje automatski izračun mjera rječničke raznolikosti (računate na temelju natuknica), sintaktičke složenosti te analizu konektora. Dodatna joj je prednost mogućnost analize uzoraka govorenog jezika, to jest pojednostavljen i olakšan način transkripcije uzoraka. U nastavku razvoja aplikacije planira se proširenje izračuna mjera te implementacija dodatnih jezičnih tehnologija kao što su automatsko prepoznavanje i transkripcija govora. Tijekom izlaganja cilj nam je predstaviti izradu aplikacije, buduća proširenja te mogućnosti njezine uporabe u široj znanstvenoj i stručnoj zajednici.

Anaphora resolution is the process of determining which previously mentioned element in the sentence or text, called the antecedent, the anaphoric element refers to. In psycholinguistics, the most studied anaphoric element is the personal pronoun in the third person singular. Compared to other anaphoric elements, such as proper nouns, the personal pronoun can be ambiguous in the absence of morphosyntactic cues. Much of the research has focused on investigating what factors guide the reader in choosing an antecedent of the ambiguous personal pronoun at the sentential level. It has been found that world-knowledge factors, as well as the biases that the pronoun itself induces, play a crucial role. This is particularly interesting from a cross-linguistic perspective since languages differ in terms of their pronominal inventory. In pro-drop languages, the personal pronoun can be omitted in the subject position, and two forms are available, null and overt. It is generally assumed that the pro-drop feature influence pronoun ambiguity resolution similarly across languages. The null pronoun prefers the subject antecedent, while the overt pronoun prefers the object antecedent. However, recent studies have questioned the generalizability of this division of labour and have shown that differences in preferences exist even between typologically closer pro-drop languages. The conflicting results can be interpreted in light of the characteristics of the languages themselves, but also the research method that was used in previous studies. Not many studies investigated how readers resolve pronoun ambiguity in real-time using more sophisticated online methods such as the eye-tracker. The pro-drop feature is also important when studying pronoun ambiguity resolution in Croatian, a pro-drop language that is rather understudied in this regard. The use of the eye-tracker should shed more light on the complexity of the phenomenon in Croatian.

Sentence processing necessarily involves ambiguity resolution ; even unambiguous words are temporally ambiguous as linguistic input unfolds. Sometimes ambiguity is not merely temporal and can impede comprehension. One such ambiguity is anaphoric ambiguity, which means that comprehenders are unable to select a unique referent for an anaphor among several candidates. 3rd person anaphoric pronouns are prone to ambiguity because they are lexically impoverished and their grammatical features do not always ensure the retrieval of the unique referent. Languages resort to various means to resolve pronoun ambiguity (e.g. morphologically marked verbs indicating co-reference in switch reference systems). From a psycholinguistic perspective, it is interesting to see what influences pronoun ambiguity resolution (PAR) when disambiguating cues are not available. Much of the research focuses on the salience/accessibility of the referent. Salient entities are more easily retrievable in mental representations and are therefore more likely to be selected as pronoun referents. Factors that confer salience include implicit causality (IC) and structural bias. IC is considered to be the property of certain verbs that draws the attention of comprehenders to the referent that is likely to be the cause of the event or state. Structural bias is influenced by the correlation between the pronoun and the grammatical role of the referent, with the subject being more salient than the object. Previous theoretical approaches considered these biases separately, so PAR has been studied as an isolated phenomenon. The probabilistic approach builds on the assumption that PAR is determined by the interplay of IC and structural biases. These biases are probabilistic, i.e. they do not represent absolute certainty, and their strength may vary. The probabilistic approach is situated in a strongly incremental theory in which sentence processing occurs moment-by-moment. PAR is a predictive process. The probabilistic approach has not been thoroughly investigated in pro- drop languages where a personal pronoun can be omitted in the subject position. A common assumption is that null and overt pronouns show a complementary distribution in terms of structural bias. The null pronoun prefers the subject and the overt pronoun prefers the object. Several studies conducted within the probabilistic approach have shown that the interplay of biases varies across languages. In our ongoing study, we aim to observe PAR with respect to the interplay of IC and structural biases in Croatian, a pro-drop language. The study will be conducted using an eye-tracking method to capture both the early and late stages of referent anticipation during sentence processing. In the presentation, experimental stimuli, participant characteristics, procedure, and eye-tracker measures will be presented and discussed.

Introduction. A macrostructure is an organizational pattern of textual elements in the narrative. According to Westby (2005), only narratives with a clear goal, attempt and outcome (G-A-O sequence) are complete episodes. However, Gagarina et al. (2019) discovered that even typical adult speakers had difficulties producing goals, leading to lower number of G-A-O sequences. They suggest evaluation of story structure that includes the most stable components (not affected by type of episode or story), namely, internal states as an initiating event (IE-A-O sequence). Research questions. This study aims to explore the role of episodic complexity expressed by both G-A-O and IE-A-O in narratives of monolingual Croatian children and adults and to find out whether preschool children differ from school children and weather children differ from adults in expressing episodic complexity. Method. Participants were preschool children (PSC) (n=45 ; age range 5 ; 4 to 6 ; 11, M=6 ; 3, SD=.48), school children (SC) (n=45 ; age range 7 ; 10 to 9 ; 0, M=8 ; 4, SD=.31) and adults (A) (n=45 ; age >20, M=24 ; 7, SD=2.71). They have been tested with the Multilingual Assessment Instrument for Narratives (Gagarina et al. 2012). G-A-O and IE-A-O sequences were extracted per participant. Results and conclusion. IE-A-O. A Kruskal-Wallis test provided strong evidence of a difference (p

Mucanje je složeni poremećaj na čiji razvoj utječe interakcija brojnih čimbenika, a među njima i jezičnih. Iako se istraživanja jezičnih sposobnosti djece koja mucaju provode već niz godina, rezultati nisu jednoznačni. Neka su istraživanja pokazala da su jezične sposobnosti djece koja mucaju slabije, dok druga nisu utvrdila razlike u odnosu na njihove tečne vršnjake (pregled: Ntourou, Conture i Lipsey 2011). Što se tiče rječnika, djeca koja mucaju pokazala su sličnu rječničku raznolikost mjerenu na jezičnim uzorcima, ali slabije rezultate na testovima rječnika u odnosu na djecu koja ne mucaju (Luckman, 2017). Cilj je ovog rada ispitati leksičke sposobnosti djece koja mucaju te utvrditi razlikuju li se one u odnosu na djecu koja ne mucaju, rabeći pritom standardizirane jezične testove te analizirajući uzorke pripovijedanja. U istraživanju je sudjelovalo 15 djece koja mucaju i 15 tečnih vršnjaka, u dobi od 7 ; 0 do 11 ; 11 godina. Ispitivanje je provedeno primjenom Peabody slikovnoga testa rječnika (PPVT-III-HR ; Dunn i sur., 2009). Za dobivanje uzoraka naracije upotrijebljen je Multilingual Assessment Instrument for Narratives (MAIN ; Gagarina i sur., 2012 ; hrvatska inačica: Hržica i Kuvač Kraljević, 2012). Analizirane su dvije mjere rječničke raznolikosti, rječnička raznolikost D (engl. vocabulary diversity D – VOCD) i pomični prosječni omjer različnica i pojavnica (engl. moving average type-token ratio – MATTR). Rezultati su pokazali da se djeca koja mucaju (DKM) na testu PPVT-III-HR ne razlikuju značajno u odnosu na djecu koja ne mucaju (DKNM), iako postižu prosječno viši rezultat (DKM: M=110, SD=15, 45 ; DKNM: M=104, SD=8, 11). Na mjeri VOCD DKM postigla su prosječno nešto niži rezultat (M=14, 91, SD=6, 07 ; DKNM: M=15, 56, SD=8, 18), ali ta se razlika nije pokazala statistički značajnom, kao ni MATTR (DKM: M=0, 912, SD=0, 046 ; DKNM: M=0, 914, SD=0, 47). Rezultati govore u prilog istraživanjima koja ne nalaze razlike u jezičnim sposobnostima djece koja mucaju i njihovih vršnjaka koji ne mucaju.

During language acquisition children are faced with the difficult task of acquiring inflectional morphology, leading to the production of different erroneous forms, among which is overextending a certain rule to irregular words (e.g., bring to bringed instead of brought in English). Overgeneralized forms reflect the complexity of the morphological system and indicate strategies children employ when facing it. They are often used interchangeably with adult-like forms (Pinker 1984). There is an ongoing debate about mechanisms underlying the production of overgeneralized forms (Ambridge et al. 2012). Dual route models posit dual mechanisms: a formal regular rule capable of operating on any verb, and a similarity-based inventory of irregulars (Prasada & Pinker 1993). Single route models posit a single mechanism for both regular and irregular forms: the generalizations in children's use of inflected forms follow from the formation of patterns between all existing forms in the input (Elman et al. 1996). Studies on verbal overgeneralization often focus on languages with low morphological complexity (e.g., English: Ambridge 2010). The Croatian conjugational system displays different degrees of complexity, and that complexity is not primarily based on the number of inflectional morphemes, but rather on an elaborate system of stem changes (Bošnjak Botica & Hržica 2016). During the early language development, children face a challenging task of acquiring that system, presumably by using overgeneralized forms to overcome its complexity. We aim to test the production of overgeneralized verb forms in preschool children, using the parental questionnaire in which parents report overgeneralizations of their children. We hypothesize that parents will report overgeneralized verbs in all verb classes in which the stem changes, but that the frequency of overgeneralizations will depend on the features of the input, i.e. it will be predicted by the frequency of verbs (higher rate of overgeneralizations for infrequent verbs) and phonological neighbourhood density (higher rate of overgeneralizations for verbs with fewer phonological neighbours). Up to now, studies have used a corpus-based method to retrieve overgeneralizations in child language (Maslen et al. 2004), which has limited success in capturing this phenomenon. Due to the low density of language sampling, low-frequency phenomena such as overgeneralizations tend to be underrepresented (Maratsos 2000). A parental questionnaire will provide more precise information. The items selected for the questionnaire are the verbs used by parents in the longitudinal Croatian Corpus of Child Language (Kovačević 2008). First, all verbs from child-directed speech were extracted (N=1352). Then, verbs with stem change were selected (N=397). They were divided into different classes according to Jelaska and Bošnjak Botica (2019). For each verb class phonological neighbourhood density, frequency in the child language corpus, subjective frequency and age of acquisition were reported (last two retrieved from Croatian lexical databases – Kuvač Kraljević & Olujić 2018 ; Peti-Stantić et al. 2018). Only verbs with the age of acquisition up to 5 were selected. Finally, all prefixed verbs were removed. The final questionnaire consists of 50 verbs with the average age of acquisition 3 ; 4, average subjective frequency 3.94 and average corpus frequency 20.