Using Monte-Carlo simulation to test predictions about the time-course of semantic and lexical access in reading.

One of the main theoretical distinctions between reading models is how and when they predict semantic processing occurs. Some models assume semantic activation occurs after word-form is retrieved. Other models assume there is no-word form, and that what people think of as word-form is actually just semantics. These models thus predict semantic effects should occur early in reading. Results showing words with inconsistent spelling-sound correspondences are faster to read aloud if they are imageable/concrete compared to if they are abstract have been used as evidence supporting this prediction, although null-effects have also been reported. To investigate this, I used Monte-Carlo simulation to create a large set of simulated experiments from RTs taken from different databases. The results showed significant main effects of concreteness and spelling-sound consistency, as well as age-of-acquisition, a variable that can potentially confound the results. Alternatively, simulations showing a significant interaction between spelling-sound consistency and concreteness did not occur above chance, even without controlling for age-of-acquisition. These results support models that use lexical form. In addition, they suggest significant interactions from previous experiments may have occurred due to idiosyncratic items affecting the results and random noise causing the occasional statistical error.

Graphemes are used when reading: Evidence from Monte Carlo simulation using word norms from mega-studies.

Recently, Chetail (Journal of Memory and Language, 2020) has claimed there is no strong evidence that multi-letter graphemes are used in reading tasks with proficient adult readers, with most studies being statistically weak or having confounds in the stimuli used. Here, I used Monte Carlo simulation with data from reading mega-studies to examine the extent to which the number of multi-letter graphemes matters in words when letter length is held constant. This was done by simulating thousands of experiments using different sets of items for each of a small number of comparisons (e.g., words with only single-letter graphemes versus words with one multi-letter grapheme). The results showed that words with two multi-letter graphemes tended to cause slower reaction times than words with one or no multi-letter graphemes, with effects found in both naming and lexical decision tasks. Interestingly, when words with no multi-letter graphemes were compared with words with one multi-letter grapheme, the differences were much weaker. Simulations of naming results using two computer models, the connectionist dual-process (CDP) model and the dual-route cascaded (DRC) model, showed only CDP predicted this pattern. Since CDP learns simple associations between graphemes and phonemes whereas DRC uses a set of grapheme-phoneme rules, this suggests that the results may have been caused by simple associations between spelling and sound being relatively easy to learn with words with one compared with two multi-letter graphemes. More generally, the results suggest that graphemes are used when reading, but they often produce relatively weak effects and thus differences in some studies may not have been found due to a lack of power.

Using electrophysiological correlates of early semantic priming to test models of reading aloud.

The speed at which semantics is accessed by words with consistent (simple) and inconsistent (difficult) spelling-sound correspondences can be used to test predictions of models of reading aloud. Dual-route models that use a word-form lexicon predict consistent words may access semantics before inconsistent words. The Triangle model, alternatively, uses only a semantic system and no lexicons. It predicts inconsistent words may access semantics before consistent words, at least for some readers. We tested this by examining event-related potentials in a semantic priming task using consistent and inconsistent target words with either unrelated/related or unrelated/nonword primes. The unrelated/related primes elicited an early effect of priming on the N1 with consistent words. This result supports dual-route models but not the Triangle model. Correlations between the size of early priming effects between the two prime groups with inconsistent words were also very weak, suggesting early semantic effects with inconsistent words were not predictable by individual differences. Alternatively, there was a moderate strength correlation between the size of the priming effect with consistent and inconsistent words in the related/unrelated prime group on the N400. This offers a possible locus of individual differences in semantic processing that has not been previously reported.

What Is Going on with Visual Attention in Reading and Dyslexia? A Critical Review of Recent Studies.

This critical review examined current issues to do with the role of visual attention in reading. To do this, we searched for and reviewed 18 recent articles, including all that were found after 2019 and used a Latin alphabet. Inspection of these articles showed that the Visual Attention Span task was run a number of times in well-controlled studies and was typically a small but significant predictor of reading ability, even after potential covariation with phonological effects were accounted for. A number of other types of tasks were used to examine different aspects of visual attention, with differences between dyslexic readers and controls typically found. However, most of these studies did not adequately control for phonological effects, and of those that did, only very weak and non-significant results were found. Furthermore, in the smaller studies, separate within-group correlations between the tasks and reading performance were generally not provided, making causal effects of the manipulations difficult to ascertain. Overall, it seems reasonable to suggest that understanding how and why different types of visual tasks affect particular aspects of reading performance is an important area for future research.

Effects of Orthographic Consistency on Bilingual Reading: Human and Computer Simulation Data.

English serves as today's lingua franca, a role not eased by the inconsistency of its orthography. Indeed, monolingual readers of more consistent orthographies such as Italian or German learn to read more quickly than monolingual English readers. Here, we assessed whether long-lasting bilingualism would mitigate orthography-specific differences in reading speed and whether the order in which orthographies with a different regularity are learned matters. We studied high-proficiency Italian-English and English-Italian bilinguals, with at least 20 years of intensive daily exposure to the second language and its orthography and we simulated sequential learning of the two orthographies with the CDP++ connectionist model of reading. We found that group differences in reading speed were comparatively bigger with Italian stimuli than with English stimuli. Furthermore, only Italian bilinguals took advantage of a blocked presentation of Italian stimuli compared to when stimuli from both languages were presented in mixed order, suggesting a greater ability to keep language-specific orthographic representations segregated. These findings demonstrate orthographic constraints on bilingual reading, whereby the level of consistency of the first learned orthography affects later learning and performance on a second orthography. The computer simulations were consistent with these conclusions.

Learning to Read and Dyslexia: From Theory to Intervention Through Personalized Computational Models.

How do children learn to read? How do deficits in various components of the reading network affect learning outcomes? How does remediating one or several components change reading performance? In this article, we summarize what is known about learning to read and how this can be formalized in a developmentally plausible computational model of reading acquisition. The model is used to understand normal and impaired reading development (dyslexia). In particular, we show that it is possible to simulate individual learning trajectories and intervention outcomes on the basis of three component skills: orthography, phonology, and vocabulary. We therefore advocate a multifactorial computational approach to understanding reading that has practical implications for dyslexia and intervention.

Young Children's Indiscriminate Helping Behavior Toward a Humanoid Robot.

Young children help others in a range of situations, relatively indiscriminate of the characteristics of those they help. Recent results have suggested that young children's helping behavior extends even to humanoid robots. However, it has been unclear how characteristics of robots would influence children's helping behavior. Considering previous findings suggesting that certain robot features influence adults' perception of and their behavior toward robots, the question arises of whether young children's behavior and perception would follow the same principles. The current study investigated whether two key characteristics of a humanoid robot (animate autonomy and friendly expressiveness) would affect children's instrumental helping behavior and their perception of the robot as an animate being. Eighty-two 3-year-old children participated in one of four experimental conditions manipulating a robot's ostensible animate autonomy (high/low) and friendly expressiveness (friendly/neutral). Helping was assessed in an out-of-reach task and animacy ratings were assessed in a post-test interview. Results suggested that both children's helping behavior, as well as their perception of the robot as animate, were unaffected by the robot's characteristics. The findings indicate that young children's helping behavior extends largely indiscriminately across two important characteristics. These results increase our understanding of the development of children's altruistic behavior and animate-inanimate distinctions. Our findings also raise important ethical questions for the field of child-robot interaction.

Effects of a mirror on young children's transgression in a gift-delay task.

While the development of self-recognition in a mirror by toddlers is well documented, less is known about how the presence of a mirror affects young children's behaviour. Here, we explored how the presence of a mirror affected 2.5- to 3.5-year-olds' behaviour in a gift-delay task. Behaviour was assessed for a five-minute test period during which children sat in front of a gift bag that was not to be touched until an experimenter returned. Transgressive behaviour by adults is reduced in the presence of a mirror, so we hypothesized that children faced with a mirror would be less likely to touch the gift than children tested without a mirror. We found that the mirror reduced transgressions in children starting from around 3 years of age. We conclude that the presence of a mirror facilitated self-monitoring in 3-year-old children, such that deviations from a behavioural standard are noticed and corrected immediately. Statement of contribution What is already known on the subject? Children's self-recognition in a mirror has been well documented. Adults' behaviour can be affected by the presence of a mirror. There is a lack of research investigating how the presence of a mirror affects young children's behaviour. What does this study add? We show that the presence of a mirror decreases young children's likelihood to transgress in a gift-delay task. This effect appears to emerge at around three years of age. These findings raise interesting questions regarding the development of self-awareness and how it relates to other mechanisms.

Understanding Dyslexia Through Personalized Large-Scale Computational Models.

Learning to read is foundational for literacy development, yet many children in primary school fail to become efficient readers despite normal intelligence and schooling. This condition, referred to as developmental dyslexia, has been hypothesized to occur because of deficits in vision, attention, auditory and temporal processes, and phonology and language. Here, we used a developmentally plausible computational model of reading acquisition to investigate how the core deficits of dyslexia determined individual learning outcomes for 622 children (388 with dyslexia). We found that individual learning trajectories could be simulated on the basis of three component skills related to orthography, phonology, and vocabulary. In contrast, single-deficit models captured the means but not the distribution of reading scores, and a model with noise added to all representations could not even capture the means. These results show that heterogeneity and individual differences in dyslexia profiles can be simulated only with a personalized computational model that allows for multiple deficits.

Working memory load affects early affective responses to concrete and abstract words differently: Evidence from ERPs.

Early posterior negativity (EPN) is an early-occurring, event-related, potential that is elicited by pictures and words that have highly arousing characteristics. Whilst EPN has been found with words presented in isolation several times, different types of words have shown quite different effects across different types of tasks. One possible reason for this is that memory and attentional demands may affect the way semantic features of words are processed, and this may modulate EPN. This was investigated in a silent reading task using abstract and concrete words of negative and neutral valence and a dual phonological working memory task to manipulate memory load. The results showed that abstract but not concrete words elicited EPN, and this may have affected downstream processing. Further analyses examining alpha desynchronization showed that negative concrete words appeared to be significantly affected by the memory load manipulation, unlike negative abstract words. These results provide evidence that the processing of features in negative concrete words is more affected by working memory and attentional demands than the processing of features in negative abstract words, and this may be responsible for the failure of negative concrete words to elicit EPN in this study. Thus, the extent to which words elicit EPN appears to be dependent on both their semantic representations and competing cognitive processes. These results provide a potential explanation for some of the differences that have been reported in previous experiments as well as insight into how memory and attention can affect the processing of the semantic features of words.

Toddlers prefer to help familiar people.

Young children's willingness to spontaneously help others is the subject of a large body of research investigating the ontogeny of moral behavior and thought. A developing debate centers around the extent to which social factors influence the desire to help. Familiarity with the person needing help is one such factor that varies across many studies but has not been systematically investigated. In Experiment 1, we show that toddlers were significantly more likely to assist a person on an out-of-reach clothespin task when they had previously become familiar with that person. Moreover, and in contrast to previous work, we found that becoming familiar with a person increases helpfulness only toward that person and does not transfer to an unknown person. We further demonstrate, in Experiment 2A, that children were equally likely to approach and take a sticker from an experimenter with whom they were familiar or unfamiliar-thereby ruling out wariness of strangers as the key driver for familiarity effects in Experiment 1. Moreover, in Experiment 2B, we show that children were more likely to help the previously unfamiliar partner (from Experiment 2A) after the partner gave the child the sticker. We conclude that familiarity is an ecologically important social influencer of toddler helping behavior.

Examining the N400m in affectively negative sentences: A magnetoencephalography study.

Magnetoencephalography was used to examine the effect on the N400m of reading words that create emotional violations in sentences. The beginnings of the sentences were affectively negative and were completed with either a negative congruous, positive incongruous, or neutral incongruous adjective (e.g., "My mother was killed and I felt bad/great/normal"). The task conditions were also manipulated to favor semantic over affective processing. Compared to the sentences with the congruous negative adjectives, the results of sensor space analysis showed that there was an N400m effect with the sentences with the neutral but not positive adjectives, despite both types of sentences containing an emotional violation. Source localization results showed a similar pattern where the sentences with the incongruous positive versus congruous negative adjectives showed no significant N400m effect in the temporal and frontal areas examined, but the sentences with the incongruous neutral versus incongruous positive adjectives in the temporal areas did, particularly the left middle temporal gyrus. These results suggest that (a) the N400m effect was likely to be caused by the incongruous neutral adjectives being comparatively harder to integrate into a negative emotional context than the incongruous positive ones, (b) emotional context created by the negative sentence stems caused deeper semantic processing of the incongruous positive adjectives to be bypassed, and (c) negative affective context was generated from reading the sentences even in task conditions where it has not been generated with isolated words.

Phonotactic constraints: Implications for models of oral reading in Russian.

The present article investigates how phonotactic rules constrain oral reading in the Russian language. The pronunciation of letters in Russian is regular and consistent, but it is subject to substantial phonotactic influence: the position of a phoneme and its phonological context within a word can alter its pronunciation. In Part 1 of the article, we analyze the orthography-to-phonology and phonology-to-phonology (i.e., phonotactic) relationships in Russian monosyllabic words. In Part 2 of the article, we report empirical data from an oral word reading task that show an effect of phonotactic dependencies on skilled reading in Russian: humans are slower when reading words where letter-phoneme correspondences are highly constrained by phonotactic rules compared with those where there are few or no such constraints present. A further question of interest in this article is how computational models of oral reading deal with the phonotactics of the Russian language. To answer this question, in Part 3, we report simulations from the Russian dual-route cascaded model (DRC) and the Russian connectionist dual-process model (CDP++) and assess the performance of the 2 models by testing them against human data.

What Do You Mean by That?! An Electrophysiological Study of Emotional and Attitudinal Prosody.

The use of prosody during verbal communication is pervasive in everyday language and whilst there is a wealth of research examining the prosodic processing of emotional information, much less is known about the prosodic processing of attitudinal information. The current study investigated the online neural processes underlying the prosodic processing of non-verbal emotional and attitudinal components of speech via the analysis of event-related brain potentials related to the processing of anger and sarcasm. To examine these, sentences with prosodic expectancy violations created by cross-splicing a prosodically neutral head ('he has') and a prosodically neutral, angry, or sarcastic ending (e.g., 'a serious face') were used. Task demands were also manipulated, with participants in one experiment performing prosodic classification and participants in another performing probe-verification. Overall, whilst minor differences were found across the tasks, the results suggest that angry and sarcastic prosodic expectancy violations follow a similar processing time-course underpinned by similar neural resources.

CDP++.Italian: modelling sublexical and supralexical inconsistency in a shallow orthography.

Most models of reading aloud have been constructed to explain data in relatively complex orthographies like English and French. Here, we created an Italian version of the Connectionist Dual Process Model of Reading Aloud (CDP++) to examine the extent to which the model could predict data in a language which has relatively simple orthography-phonology relationships but is relatively complex at a suprasegmental (word stress) level. We show that the model exhibits good quantitative performance and accounts for key phenomena observed in naming studies, including some apparently contradictory findings. These effects include stress regularity and stress consistency, both of which have been especially important in studies of word recognition and reading aloud in Italian. Overall, the results of the model compare favourably to an alternative connectionist model that can learn non-linear spelling-to-sound mappings. This suggests that CDP++ is currently the leading computational model of reading aloud in Italian, and that its simple linear learning mechanism adequately captures the statistical regularities of the spelling-to-sound mapping both at the segmental and supra-segmental levels.

Modelling reading development through phonological decoding and self-teaching: implications for dyslexia.

The most influential theory of learning to read is based on the idea that children rely on phonological decoding skills to learn novel words. According to the self-teaching hypothesis, each successful decoding encounter with an unfamiliar word provides an opportunity to acquire word-specific orthographic information that is the foundation of skilled word recognition. Therefore, phonological decoding acts as a self-teaching mechanism or 'built-in teacher'. However, all previous connectionist models have learned the task of reading aloud through exposure to a very large corpus of spelling-sound pairs, where an 'external' teacher supplies the pronunciation of all words that should be learnt. Such a supervised training regimen is highly implausible. Here, we implement and test the developmentally plausible phonological decoding self-teaching hypothesis in the context of the connectionist dual process model. In a series of simulations, we provide a proof of concept that this mechanism works. The model was able to acquire word-specific orthographic representations for more than 25 000 words even though it started with only a small number of grapheme-phoneme correspondences. We then show how visual and phoneme deficits that are present at the outset of reading development can cause dyslexia in the course of reading development.

A computational and empirical investigation of graphemes in reading.

It is often assumed that graphemes are a crucial level of orthographic representation above letters. Current connectionist models of reading, however, do not address how the mapping from letters to graphemes is learned. One major challenge for computational modeling is therefore developing a model that learns this mapping and can assign the graphemes to linguistically meaningful categories such as the onset, vowel, and coda of a syllable. Here, we present a model that learns to do this in English for strings of any letter length and any number of syllables. The model is evaluated on error rates and further validated on the results of a behavioral experiment designed to examine ambiguities in the processing of graphemes. The results show that the model (a) chooses graphemes from letter strings with a high level of accuracy, even when trained on only a small portion of the English lexicon; (b) chooses a similar set of graphemes as people do in situations where different graphemes can potentially be selected; (c) predicts orthographic effects on segmentation which are found in human data; and (d) can be readily integrated into a full-blown model of multi-syllabic reading aloud such as CDP++ (Perry, Ziegler, & Zorzi, 2010). Altogether, these results suggest that the model provides a plausible hypothesis for the kind of computations that underlie the use of graphemes in skilled reading.

Relationships between receptive vocabulary in English and Cantonese proficiency among five-year-old Hong Kong Kindergarten children.

There is little consensus among different early childhood education stakeholders in Hong Kong on whether it is beneficial or detrimental for children to receive an English bilingual education before the age of 6. This longitudinal study investigated the issue of potential 'detrimental effects of learning English' on Hong Kong kindergarten children's performance in L1 (Cantonese) and L2 (English) over a six-month period. The sample consisted of 53 children, 29 of whom went to international schools and received 90 minutes of daily in-class English instruction, and 24 of whom went to local schools, and received 20 minutes of daily in-class English instruction. Analyses of the relationship between L1 and L2 development showed no evidence that learning a second language is detrimental to the learning of the first. This was despite the large difference in the amount of in-school instruction time the children who went to local vs. international schools received. Children in the international schools vastly outperformed those in the local schools in English. We found no evidence that learning a second language is detrimental to learning more general cognitive skills. The results provided very weak evidence for the opposite. Thus, learning English as a second language in Hong Kong before the age of 6 did not harm children's learning in any way.

Beyond single syllables: large-scale modeling of reading aloud with the Connectionist Dual Process (CDP++) model.

Most words in English have more than one syllable, yet the most influential computational models of reading aloud are restricted to processing monosyllabic words. Here, we present CDP++, a new version of the Connectionist Dual Process model (Perry, Ziegler, & Zorzi, 2007). CDP++ is able to simulate the reading aloud of mono- and disyllabic words and nonwords, and learns to assign stress in exactly the same way as it learns to associate graphemes with phonemes. CDP++ is able to simulate the monosyllabic benchmark effects its predecessor could, and therefore shows full backwards compatibility. CDP++ also accounts for a number of novel effects specific to disyllabic words, including the effects of stress regularity and syllable number. In terms of database performance, CDP++ accounts for over 49% of the reaction time variance on items selected from the English Lexicon Project, a very large database of several thousand of words. With its lexicon of over 32,000 words, CDP++ is therefore a notable example of the successful scaling-up of a connectionist model to a size that more realistically approximates the human lexical system.