Online conference

Reading is a complex multifaceted process that critically relies on phonological processing (Ramus et al., 2013). The well-known theoretical model of phonological processing developed by Wagner and colleagues (Wagner et al., 1994) separates phonological skills into three main components: phonological awareness, phonological working memory, and rapid naming. Despite extensive research, it remains elusive which phonological tests are the most effective for studies of reading (Georgiou, Parrila & Papadopoulos, 2008). Even when they address the same component of phonological processing, some tests predict reading skills better than others. One of the most plausible explanations is that phonological tests may differ in their cognitive requirements (Vandervelden & Siegel, 1995; Yopp, 1988). However, the idea that the overall complexity of a phonological task affects its potential to predict reading skills has never been strictly tested. To fill the gap, we developed seven phonological tests varying in linguistic complexity, e.g. the number and type of involved linguistic processes, and examined the relations between the level of complexity in phonological tasks and reading.

The research was conducted in a group of typically developing Russian-speaking children (N=90, 7-to-11 years of age). Data were analyzed using (generalized) linear mixed models estimated in a Bayesian framework (Bürkner, 2017). For each child we extracted the individual estimate of decrease in accuracy associated with introducing one more linguistic process and assessed whether the cost of a linguistic process estimated for each child is associated with the child’s reading fluency. We revealed, that greater individual cost of a linguistic process required for a phonological test was associated with a reliable decrease in reading fluency. These results suggest that phonological processing needs to be considered in models of reading not only in terms of traditional types, but also regarding the quantity and types of involved cognitive processes.

A highly ecological ability such as reading is often investigated through non-ecological experimental paradigms. The present work aims at providing the community with an ecological tool for future research, a developmental database of eye movement measures during natural reading. Eye movements were recorded from a large sample of Italian developing readers (N=140), aged 8–11, as they read 12 multi-lined passages taken from story books for children (1566 tokens and 762 distinct types). Eye-tracking data were also recorded from a group of skilled adult readers (N=33), for comparison.

In line with previous literature (e.g., Rayner, 1986; Blythe & Joseph, 2011), our results show well-known developmental changes in reading behaviour, including a significant increase in reading rate as expressed in number of words per minute (p<.001; fig. 1a), as well as a decrease in saccade length (p<.001; fig. 1b) and in fixation duration (p<.001; fig. 1c). Benchmark effects of word length and frequency also emerge: first-of-many fixation duration (FoM, figs. 2a, 2b) and gaze duration (GD, figs. 2c, 2d) are greater for longer, compared with shorter words (length effect on FoM: p=.02; on GD: p<.001) and for low-frequency, compared with high-frequency words (frequency effect on FoM: p<.001; on GD: p<.001). Furthermore, the effects of word length and frequency on gaze duration are modulated by age, with both effects getting progressively smaller as a function of age (p<.001).

These results corroborate evidence from single sentence reading tasks, and validate the database as a valuable resource for the investigation of reading acquisition. The large amount of data, as well as the naturalistic nature of the materials and the task, will allow researchers to complement findings from controlled experiments and potentially address questions that are left open by those.

This paper discusses the competence of 8 Italian high-school students with developmental dyslexia (DD: mean age 17;11) in the repetition and the production of oblique relative clauses, compared to 16 Italian typically developing peers (TD).

These structures, typical of the formal register, are frequently substituted by more colloquial alternatives, namely sentences containing resumptive clitic pronouns. Typically developing children start producing oblique relatives at the age of 10 (Guasti and Cardinaletti 2003). Data on adolescents are not available.

Individuals with dyslexia may manifest difficulties with movement-derived sentences in oral tasks at school age (Arosio et al. 2016), or even at the University (Cardinaletti and Volpato 2015), beyond their disability in reading and writing skills.

Complex syntactic structures are examined through a sentence repetition task (Del Puppo et al., 2016) and an elicited production task (a modification of Mulas, 2000).

In the repetition task, the rates of correct experimental sentences repeated by DD are significantly lower than those of TD. 2 students with dyslexia behaved like the TD peers, while the rest of the group had significantly lower results (p< .001).

The elicited production task confirms the results obtained in the repetition task. DD had significantly more difficulties in the production of oblique relative clauses than TD peers.

This study shows that, overall, individuals with DD have difficulties in mastering complex structures derived by syntactic movement. TD students also showed some difficulties in the use of oblique relative clauses compared to the other structures investigated: at high-school age, constructions of the formal register are not fully mastered yet. This may be due to the frequency at which these constructions are used at school or in formal contexts.

6 students with DD resorted to sentences of the colloquial register much more often than the other participants. This may be due to their low experience with the formal register and their difficulties in reading.

This study focuses on the acquisition of 1^st, 2^nd and 3^rd person singular direct (DO) and indirect object (IO) clitic pronouns in Italian. We tested typically developing children (TD) and children with Learning Difficulties (LD) with a repetition task of sentences containing restructuring verbs, to observe the effects of person, case features, and position on accuracy. In Romance languages, there is an asymmetry in the acquisition of 1^st and 2^nd person versus 3^rd person singular clitic pronouns: 3^rd person pronouns are acquired later and are more difficult to process. Concerning case features, DO clitics are often omitted, while IO clitics are better preserved (1; 2; 3; 4). Children with LD often display difficulties with DO clitics (5; 6; 7). No data have been systematically collected yet on the acquisition of restructuring. For this study, a repetition task was designed to investigate both comprehension and production abilities (8; 9).

63 monolingual Italian children (age 6;6-9;11, mean 8;6) took part in the experiment. They were divided into five groups: a group of children with LD (dyslexia or dysorthography); a group of children at risk for developing LD; three control groups of TD children. A control group of TD adults was also tested.

The test includes 49 sentences containing clitics and restructuring verbs. Each sentence contains one or two restructuring verbs (modal/motion). The clitic appears in the two (proclitic/enclitic) or three (proclitic/enclitic intermediate/enclitic final) available positions. Case and person features did not influence accuracy, clitic positions did. The most frequent error was clitic misplacement: the clitic was moved most frequently from the enclitic to the proclitic position in 2-verbs sentences, or from the final to the intermediate enclitic position in 3-verbs sentences. Adults almost performed at ceiling.

Both TD children and children with LD seem to prefer proclisis, therefore mono-clausality. In 3-verbs sentences, some children dispreferred the final position, but they did not move the pronoun to the proclitic position. This can be explained with Jakubowicz’s Derivational Complexity Hypothesis (10). Note that enclisis on the intermediate verb implies that the highest verb is analysed as lexical (11). Both types of structures entered by restructuring verbs seem available to children

The problem. Italian deaf children accurate linguistic competence assessment is problematic using standard comprehension tests (Cardinaletti 2019). This is because these tests either rely on the auditory modality (TROG-2, Bishop 2009) or they target constructions which are too complex for children (Comprendo, Cecchetto et al. 2012). Moreover, simple behavioral measures, like reading times, are hardly recorded by standard tests, but those measures, once normalized by subject (to minimize individual differences), correlate effectively with the perceived difficulty (Traxler 2017).

The proposal. Elaborating on a 3-year longitudinal study (Chesi, Ghersi & Musola 2019) on minimal pairs perception (Radelli 1998, Franchi & Musola 2011), we designed a test (COnVERSA) based on simple grammaticality judgements targeting 3 marco-phenomena (agreement and inflection, thematic role assignment and non-local dependencies) articulated in 14 specific phenomena on the basis of cartographic studies (Rizzi & Cinque 2016 a.o.) and on typical language acquisition (Belletti & Gusti 2015); the complexity score attributed to each item is proportional to the height of the functional projection and the locality of the dependency involved (Moscati & Rizzi 2014, Chesi & Canal 2019). Three versions of the test have been validated so far: two off-line (Fig.1a) and one on-line (Fig.1.b) using jsPsych platform (de Leeuw 2015) with reading times recording. The final version consists of 240 items divided in 2 parts to be delivered one at the beginning, the other at the end of a period of observation for documenting competence progress and/or the efficacy of a specific linguistic training program.

Preliminary results. Version 1 of the test has been validated on 59 hearing children (age-range: 7-8 yearold, M=7.6, SD=0.28); version 2 on 89 hearing children (age-range: 7-10 yo, M=8.7, SD=0.9), 19 deaf children (range: 8-14 yo, M=12, SD=2,68) and two deaf adults (32 e 38 yo). The results (Growing Curve Analysis, Mirman 2014) show that the performance significantly increases with age only in hearing children (Fig. 2.a, estimate=0.92, z=8.06, p<0.001) and this correlates with standard development based on Belletti & Guasti (2015) evidence. This suggests that the test is sensitive enough to capture the “maturation” of the linguistic competence in the analyzed morphosyntactic areas. Deaf children show individual typical patterns that do not correlate with age (Fig. 2.b). Only the older children group (8 yo) performs significantly better than deaf children group (estimate= +0.51, z=1.91, p = 0.05). This indicates that the test is also accurate in documenting a delay in deafs acquisition, independently from age.

Introduction. One of the most fascinating properties of syntactic comprehension is that subjectrelative clauses are easier to process than their object-relative counterparts. However, in the last few decades, research sought theoretical knowledge about this syntactic property mostly in English (e.g., Ford, 1983; Just & Carpenter, 1987; Frazier, Clifton & Randall, 1983; Gibson, 1998, 2000; Lewis, Vasishth & Van Dyke, 2006), whereas languages with different syntactic preferences could provide different results. Furthermore, this phenomenon could be studied not only in monolingual, but also bilingual population, which combines the knowledge of two languages. This study aimed at identifying the mechanisms of monolinguals and bilingual syntactic processing with the focus on the subject (SRC) and object-relative clauses (ORC).

Method. Three groups of participants were investigated: monolingual Russians, monolingual Germans and Russian-German bilingual speakers. The comprehension was investigated with two sentencepicture-matching experiments: one based on Russian and another one based on German sentences. The participants were asked to identify the agent, choosing the picture with the correct representation of thematic roles while their accuracy and reaction time were measured.

Results. First, we found no credible evidence that SRCs are easier to process than ORCs in German, however, this effect approached significance. Second, we observed no indication of RC asymmetry, but a strong evidence of word order preference in Russian. Third, we observed no monolingual advantage over bilingual. Finally, no effect of transfer from second to first language was found in the bilingual group. Most crucially, we observed RC asymmetry in the bilingual group that was not present in the monolingual groups. The theoretical substantiation of these results in the framework of the dominant theories of syntactic processing in monolingual and bilingual speakers is discussed.

Introduction: Infant’s cry may be considered as an early biomarker of neurodevelopmental disorder (NDDs). Acoustical properties of cry vocalizations have been shown to differ in infants at risk for Autism Spectrum Disorder (ASD) compared to a control group at 6 months.

The Italian network for early detection of ASD (NIDA) aims to identify early signs of NDDs, including atypicalities in cry features, in High-Risk (HR, siblings of children with a diagnosis of ASD) and Low-Risk infants (LR, siblings of typically developing children) from birth to 36 months of age.

Methods: The NIDA Network enrolled 314 HR infants and 136 LR. For each infant, audio-recordings have been collected at 10 days, 6-12-18-24 weeks. Acoustical features from cry vocalizations were analyzed: F0 is the lowest pitch of periodic signals; formants (F1–F4) are frequency peaks with wavelengths multiples of the fundamental frequency. Four infants’ groups based on clinical outcome were defined: HR diagnosed with ASD (HR‐ASD, n=3); HR diagnosed with NDDs (HR‐NDD, n=14), HR with typical development (HR-TD, n=25) and LR with typical development (LR, n=59).

Results: Analysis of the cry episodes detected significant differences between HR-TD, HR-NDD; HR-ASD and LR. HR-NDD infants differed in comparison to HR-TD, and HR-TD in comparison to LR at each timepoint in cry formants. HR-ASD infants emitted cry with less intensity compared to HR-NDD, HR-TD and LR at 10 days and 6 weeks. No significant differences were observed in F0.

Discussion: Results suggest that infants' cry formants of HR infants later diagnosed with NDDs differ from those of TD infants. Interestingly, HR infants later diagnosed with NDDs, including ASD, emit cry with less intensity from birth. Data are promising for considering cry intensity and frequency peaks with wavelengths multiples of the fundamental frequency, as very early markers of future atypical neurodevelopmental.

Production of 3rd person Direct Object Clitic pronouns (DOcls) is a clinical marker of developmental Language Disorders (LD) (Bortolini, 2002) but is also problematic in several other conditions of atypical development (Guasti, 2013; Sorianello, 2012; Tuller et al., 2011). DOcl elicitation showed that children produce fewer target DOcls under conditions of gender mismatch between the sentential subject and the DOcl antecedent (Arosio et al., 2019; Casani & Cardinaletti, 2019; Délage et al., 2016). Nevertheless, gender, as well as number effects on DOcl production have not been explored so far under balanced conditions of gender and number.

A test (Casani, Cardinaletti, & Volpato, in preparation) was implemented to elicit 48 DOcl pronouns with calibrated combinations of gender and number with the sentential subjects. A pilot experiment was conducted on 94 children attending 1st (N=49), 2nd (N=36), and 3rd (N=9) grade of an Italian primary school. Forty-two were bilinguals coming from migrant families with different mother tongues; the other fifty-two were Italian monolinguals with the same instructional background. Eight monolinguals and two bilinguals had a diagnosis of mixed LD within different profiles of comorbidity. Responses were analyzed by adapting the coding scheme by Casani (2020). A mixed multinomial logistic regression was performed to analyze the effects of different gender-number combinations, bilingualism, LDs, and school grade on responses.

The production of the singular and plural feminine DOcl under conditions of gender mismatch and number matching with the subject predicts higher rates of clitic errors, probably for the interference of the subject gender or the use of the default masculine singular form. No significant effects of number emerged, confirming a higher prominence of number than gender in anaphora resolution (Carminati, 2005; De Vincenzi & Di Domenico, 1999). The disambiguation of number seems to help gender agreement of the marked singular feminine DOcl. This can be explained by assuming a syntactic model that entails the existence of an independent functional projection of gender whose realization is helped by the presence of a higher functional projection of number (Volpato, 2010). The new test showed to be sensitive and specific enough towards LDs in the monolingual sample. Bilingual children face language difficulties that are proportional to the typological distance of their mother tongue from Italian but their clitic performance is significantly improved by instruction. DOcl production as a marker of LDs deserves further exploration in larger more homogeneous samples of bilingual children.

Automatic speech analysis and natural language processing have enabled more exact measures of human states and traits and skills that are manifest in spoken language. The authors have built several speech-interactive apps that elicit spoken responses and extract multiple measurements from these responses. These analysis methods have been tested and validated with adults, with children, and with other special populations for applications in education, psychology and medicine. Our focus is on current applications that are deployed on devices with just audio I/O and (optionally) a screen (e.g. mobile phone, tablet computer), without special instrumentation.

The fully automated measures analyze content (words in sequence; semantic embeddings), while also producing fine-grained data in the time domain (segment, syllable, word, and phrase times; latency and pause times; all in 10 msec. units) and in the frequency/amplitude domain (energy; formant tracks; voicing, F0, jitter & shimmer, harmonics/noise, PLP components). These measures are applied in ways that generate new insights from traditional, well-studied tasks, for example: answer open or closed questions; say missing words (cloze), read text aloud, shadow spoken utterances, and inhibit cognitive interference (Stroop task). These fine-grain analyses have refined the measurement of constructs including language fluency/proficiency, reading development, cognitive coherence, perseverance, and vigilance or executive control, as well as the valence and intensity of emotional states.

Following an overview of these methods, we describe four specific applications, with technical detail and validation data: childhood second language development (ages 5-12 years); reading development (accurate rate, fluency, perseverance, comprehension), assessments of attention (Stroop) and verbal memory (retelling stories) with psychiatric outpatients. The near-term value of these assessments for atypical children and adults will be discussed. Ultimately, these measures may help inform new causal (neuroanatomical) studies of atypical language acquisition and processing.