The Transition from One-word to Multi-word Utterances. A Case-study of an Italian Child

The Transition from One-word to Multi-word Utterances. ACase-study of an Italian Child Luca Cilibrasi ABSTRACT: Previous work has suggested that infants start producing multi-word utterances while they still produce single-word utterances, and they make generalizations about the order of constituents from early stages. This article describes the transition from one-word to multi-word utterances in an Italian child recorded every two weeks for 45 minutes from the age of 1;05 to the age of 2;05, and it examines specifically whether the child’s productions respect ahead–complement generalization (in Italian, the generalization that heads precede complements). The analysis was conducted on files available in the Childes database. The study shows that one-word utterances are indeed not abandoned when the child learns to combine words, and the first two-word productions reflect adult utterances, whether or not these comply with the general ordering of heads and complements. These results are compatible with approaches that see first multi-word utterances as being syntactic, but they also show that the level of generalization is not fully compatible with predictions from experimental work on head-directionality (which predicts awider generalization than the one observed in this child). KEY WORDS syntax, infants, acquisition, head–complement, merge 1 INTRODUCTION 1.1 THEOrETICAL BACKGrOUND Every grammatical sentence consists of a sensible combination of at least two words, an operation described with the term “merge” (Adger, 2003; Zaccarella & Friederici, 2015; Fukui, 2017). The ability to combine words to form new units has been apillar in linguistics within the so-called generative approach, where it is conceived as the main property of human syntax and askill that differentiates our cognition from that of other animals (Chomsky, 2014). In different theoretical environments this ability is seen as less crucial and as emerging from domain-general cognition (Bybee & McClelland, 2005): for example, researchers in the usage-based tradition agree that there is arole of combinatorial skills in language, but they undermine considerably their importance, and point to the fact that most of the sentences we produce are not created ex-novo, but rather they are repetitions of what we have already said or heard (Ibbotson, 2013). The main aim of this article is to describe the emergence of simple syntax (combination of at least two units) in the spontaneous productions of an Italian child. The two theoretical approaches just mentioned offer different interpretations of early OPEN ACCESS LUCA CILIBrASI 27 child productions, as aconsequence of their different approach to the role of combination in language. Researchers in usage-based grammar suggest that the first multiword utterances are not created using combinatorial skills, but rather by storing sequences of words as if they were units (Tomasello, 2009; Diessel, 2013). According to this view, when children start producing small sequences of words, such as “article + noun”, or “copula + adjective”, they are not combining these elements but simply treating them as units. Researchers in the generative tradition suggest instead that the first small sequences are basic syntactic constructions, where the child is combining two elements to form anew unit (Guasti, 2017). Disentangling these two proposals is beyond the scope of this article. However, this introduction is necessary because the notion of syntactic development Iuse here is bound to the generative tradition. In this article, Ilook at the emergence of sequences such as the ones just described (combinations of two words), which Iinterpret as being syntactic. It should be stressed, however, that this is apersonal choice and usage-based approaches would not interpret these sequences as being necessarily syntactic. In addition, while the article is grounded in the generative tradition in that it treats small sequences as syntactic, the analysis of the results on head–complement order does take into account the tenets of usage-based theory, and it shows that usage-based theories of language acquisition may offer relevant insights into the description of the current data. The article is organized as follows: first Ipresent an introduction to previous studies dealing with the transition from one-word to multi-word utterances and with the development of head–complement generalizations. Then Ipresent two research questions (stemming from these two topics), followed by two analyses that address them. Finally, Idiscuss the obtained results in relation to the literature presented in the introduction. 1. 2 EArLY SYNTACTIC DEVELOPMENT. frOM ONE-WOrD TO MULTI-WOrD UTTErANCES In everyday life, attending to child verbal productions, one can observe that children do not combine words like adults. However, despite these early stages, child productions gradually transform into adult productions, suggesting that the unconscious skills of infant and adult speakers are linked. The question of how children’s productions transform into adults’ productions has been the subject of an intense debate for several decades (Deuchar, 2013). One proposal argues that children use simplified rules and their competence partly overlaps with that of adults. According to this view, children leave some positions empty at the level of realization, while keeping similar underlying structures, and this leads to speech that differs from adult speech. According to adifferent view, children’s rules (or generalizations) are different than those found in adult grammars. These asymmetries are observed particularly in intermediate stages of development (so called U-shaped development, Siegler, 2004). In English preschoolers, for example, it is common to hear expressions such as “Igoed” or “it breaked”, asymptom of the fact that children are applying arule which in the adult grammar does not apply to irregular verbs such as “to go” or “to break” ( Marcus, 1996; Taatgen & Anderson, 2002). Before producing such sentences, OPEN ACCESS 28 STUDIE ZAPLIKOVANÉ LINGVISTIKY 2/2021 children experience various developmental phases in which they communicate with different systems, such as screams, cries, pointing and visual coordination (Lock & Zukow‐Goldring, 2010). The first proto-syntactic productions arrive around the first birthday and are made up of single items described with the term “holophrases”. Holophrases are interesting elements in achild’s speech: They are phrases made up of only one unit, which however already express arelatively complex concept (Bloom, 2013). Achild saying “ball” to their parents might mean “let’s play ball”, while achild saying “dog” might mean “there is adog”. There are good reasons to think that singleword production does not mean the absence of syntactic combination (Guasti, 2017). For example, several studies show that even when children produce verbs in isolation, they make use of mini-paradigms, which essentially allow them to inflect verbs in the present tense in the first, second and third person singular, in languages that have these inflections (Bittner et al., 2011). It is reasonable to assume that children have some knowledge of the fact that the verb may be divided into aroot and abound morpheme, and that therefore their syntactic representation of verbs reflects this competence. Asentence like the Italian “accendo luci” ‘Iturn on lights’ is likely to have an underlying structure as in Figure 1, when uttered by achild. As one can see, the inflectional element (T = tense), in this case assuming the value “present”, is displayed, since the verb is inflected for tense. It should also be noted that the structure contains the underlying subject of the sentence, since the appropriate verbal inflection suggests that the child is correctly assuming who the subject is. Figure 1:Syntactic tree for “accendo luci”. Similarly to work on verbs produced in isolation, work on omissions leads to conclusions suggesting that when children produce sentences that miss some elements, these elements may be present at amore abstract level (Rizzi, 2002; Carter & Gerken, 2004; Guasti et al., 2004; Wexler et al., 2004). These observations lead to the first prediction of this study: single-word utterances and multi-word utterances overlap, since single-word utterances do not necessarily indicate alack of syntax. This claim is compatible with both generative and usage-based accounts. Diessel (2013) notices that even when the child produces new combinations of words, these tend to be combinations of holophrases with other elements, and there is no specific trade-off between the two stages. Similar conclusions are reached by Guasti (2017) in her review of the generative literature. TP NP T’ (Ø) T VP “present” V NP accendo luci OPEN ACCESS LUCA CILIBrASI 29 1.3 HEAD DIrECTION AND BEYOND. PrEVIOUS STUDIES ON TWO-WOrD UTTErANCES According to generative analyses, one generalization both children and adults make is that of head directionality: When two units combine, one of the two takes the head role and gives its properties to the whole phrase. Averb merging with its complement, for instance, will create averb phrase; apreposition merging with anoun phrase will give rise to aprepositional phrase (Chomsky, 2014). According to this view, languages display aregular organization of their constituents: certain languages prefer aheadinitial order, while other languages prefer ahead-final order. English, French and Italian, for example, display ahead-initial order: verbs appear before their complements, prepositions appear before their nouns, and copulas appear before their adjectives. Japanese and Turkish, on the other hand, show the opposite pattern. Research on child development has suggested that children make generalizations about head directionality from very early stages during development. Using the highamplitude sucking procedure, Christophe et al. (1997) showed that 2–3-month-old infants are able to discriminate languages based uniquely on the prosodic contour created by head–complement order. In this study, infants from amonolingual French background were presented with two kinds of semi-artificial stimuli paired on all phonetic properties except the phonetic contour associated with head–complement directionality. By measuring their interest (sucking rhythm) when the phonetic contour was changed from one kind to the other, the authors showed that infants were sensitive to achange in head–complement directionality. Gervain et al. (2008) investigated these preferences comparing Italian and Japanese infants presented with artificial stimuli. This study is based on an important property of function words, which is that of being more frequent than content words. If infants are making ageneralization about word-order directionality, they should have apreference based on frequency distributions: In two-unit sequences, Italian children should prefer frequent items before infrequent items, while Japanese children should show the opposite pattern. Gervain et al. (2008) tested this idea on 8-month-olds in an artificial grammar learning experiment. By measuring looking times, the authors showed that at this stage children from the two linguistic environments behave differently, with Italians displaying longer looking times for initial-frequent sequences, and Japanese showing the opposite pattern. These findings suggest that infants do indeed operate with generalizations about head–complement directionality in early stages. While these findings are interesting, some researchers have criticised the interpretation that the authors of these studies have given to them, particularly the proposal that these results imply generalisations in the child’s mind. One problem with these ideas is that the generalizations proposed are too wide and do not mirror the actual linguistic environment. As Tomasello (2001) and Dryer (1992) notice, head directionality generalization is not necessarily observed in languages. First, the theory of head directionality was developed on asmall number of languages, and its applicability to all human languages is debatable, with some languages displaying considerable flexibility in the order of heads and complements. Second, even within the languages analyzed, the generalization makes predictions of word order that are simply not observed. For OPEN ACCESS 30 STUDIE ZAPLIKOVANÉ LINGVISTIKY 2/2021 example, in English adjectives occur before nouns, even if nouns are described as heads of aphrase that contains anoun and an adjective. Similarly, articles occur before nounsin English and all Romance languages, again against the intuition that nouns are heads in an article–noun sequence (even though some have proposed that articles, and not nouns, are heads in those sequences (Hudson, 2004)). Tomasello (2001) claims that early productions are less abstract than what generalizations would suggest, and he proposes that alarge component of early productions must be stored sequences of words. Several studies described the occurrence of specific head–complement structures in anumber of children, abundle of data that can help evaluate this idea: Seminal work from Braine and Bowerman (1976) showed that most of children’s productions respect the order of heads and complements as required by their language, and early productions mostly boil down to variations on asimple structure where one element is kept, and the other is changed (i.e., more-water vs more-juice vs more-chicken). This pioneering work was conducted on 6 American children and anumber of case studies on other languages (Hebrew, Samoan, Finnish and Swedish) and still remains one of the most thorough analyses of this kind. As the authors stress, their findings do not hint at wide generalizations in the child’s mind, but they do hint at the presence of some simplified rules. In another well-known study, Poeppel and Wexler (1993) show that 2-year-old German children produce verbs and objects in the order required by the specific sentence they are uttering. In German the order of verbs and objects varies according to the properties of the verb, and infants appear to be sensitive to these properties, suggesting, according to the authors, an internal representation of syntactic rules. Sugisaki (2005) showed that Japanese children produce objects and verbs consistently in the right order at age 2, suggesting again that children master the correct ordering of these elements from avery early stage. In astudy of aGerman child, Behrens and Gut (2005) showed that the earliest two-word productions are an interesting mix of structures: The child is either producing ungrammatical repetitions of two nouns, grammatical sequences of determiners and nouns, or sequences of verbs and nouns in both possible orders (the authors do not report the prevalence of each order). Finally, alarge corpus study on French (Le Normand et al., 2013, p.655) investigated “whether or not increases in language complexity (as measured by MLU) could be predicted on the basis of lexical, pragmatic, and grammatical diversity (word types)”. This study showed that the mean length of utterances (MLU) is best predicted by grammatical diversity, and not by lexical diversity, afinding that is in line with the generative approach (Guasti, 2017) and not in line with the usagebased approach (Diessel et al., 2013). These findings, taken together, do not seem to point in one specific direction, since they are not all compatible with one another. Interestingly, none of these studies specifically addressed the order of head–complement structures when these differ within the same language (for example, “adjective + noun” vs “verb + complement” in English). The current study partly fills this gap, since it includes both structures that are canonical head–complement sequences in Italian (for instance “verb + complement”), and structures that require the opposite order in adult grammar (for instance “object clitic + verb”). The studies mentioned in this section bring us to the second prediction of the current article, which, based on the less unified literature on this topic, is more open OPEN ACCESS LUCA CILIBrASI 31 than the first prediction. If children rely on awide head–complement generalization during the transition from one-word to multi-word utterances (as predicted by Gervain et al., 2008), we will observe asignificant number of occurrences where the child inverts heads and complements in comparison to adults, for example in the context of “object clitic + verb” sequences. If, instead, productions are less bound to generalizations (as predicted by Tomasello, 2001), we will observe children producing two-word utterances in the same order as their parents, whether or not these respect the generalization of head–complement directionality. 2 RESEARCH QUESTIONS This article aims at contributing new data to two open research questions: 1. Do children abandon one-word utterances once they start producing syntactic sequences? 2. Do children rely on ahead–complement generalization once they start combining words? 3 METHOD The analysis reported in this article was carried out on files extracted from the Childes international database (MacWhinney, 2014).1 To choose which files to use among those available in the database, Ihave implemented asystematic procedure. Ineeded early productions, and Ineeded files from achild who had been followed consistently for along time, in order to see changes from the lack of syntax to acertain level of syntactic complexity. Ithen made the first reduction by selecting the folders in which the oldest file did not indicate an age greater than about ayear and ahalf. Ithen used amathematical tool, the mean length of utterance (MLU), to select achild for the analysis. MLU is afunction that counts the number of words and divides it by the number of expressions that the subject chosen for analysis has pronounced and thus gives an estimate of the length of the uttered sentences (Rice et al., 2010). By calculating MLU on the first and last file of the subjects who had passed the first screening, Iselected the (only) subject that initially had aMLU value close to 1 and reached avalue close to 3. The child thus selected was Marco (MAR) and the data were those collected by Tonelli (2004). These are recordings of child–mother interaction lasting 45 minutes each, conducted at home during playtime. In these sessions, the child produced an average of 355 outputs per recording (standard deviation: 80,54). Outputs include utterances, but also gibberish sounds, 1 The term Childes is an acronym that stands for Child Language Data Exchange System. The database was created in 1984 by Brian MacWhinney and Catherine Snow and it contains spontaneous productions of children from many different linguistic backgrounds ( MacWhinney, 2004). OPEN ACCESS 32 STUDIE ZAPLIKOVANÉ LINGVISTIKY 2/2021 word plays, and attempts at producing words that were, however, distant from any meaningful output. As part of the main analysis of this study, Ifirst proceeded with acount of the utterances produced by the child throughout his development. The first insights were obtained using an automatic procedure on the dataset, aimed at obtaining acount of strings of 2, 3, 4 and 5 words produced by the child in each datafile. These were obtained with aGrep command. The command used is the following: grep -i-c -n -B3 -A3 -E'MAR: [[: space:]] * [@ a-zA-Z] [@ a-zA-Z] * ([[: space:]] | [.,;:?! ']) [@ a-zA-Z] [@ a-zA-Z] * ([[: space:]] | [.,;:?!']) * $ ‘ This command counts the number of expressions that contain two words, and translated into natural language reads: “count the number of expressions starting with MAR, followed by aspace, followed by one or more uppercase or lowercase letters, followed again by aspace or apunctuation symbol, followed again by one or more uppercase or lowercase letters, ending with aspace or apunctuation mark”. Similar strings were used to get the count of 3, 4 and 5 words. Importantly, this command does not count word plays, off-target productions and sequences that are not aword in Italian, since these were tagged with specific symbols in the file. This command entails aspecific sense of the term “word”. The precise definition of the term “word” is asubject of debate and it appears that researchers have used it equivocally (Haspelmath, 2017). Here we use the term to refer to monoor plurimorphemic items that are conventionally written in isolation. In this sense, both corro ‘Irun’ and burro ‘butter’ are words, even if corro contains two morphemes, and burro only one.2 This specification is important because it underlines an intrinsic limit of this count as ameasure of (morpho)syntactic complexity, since such count does not differentiate between monomorphemic and plurimorphemic words. This approach has consequences on the count of MLU. MLU is conventionally calculated on morphemes (mean number of morphemes per utterance) in the English data of Childes. This is possible because English is amostly analytic language (Josiah & Udoudom, 2012) and has alimited number of bound inflectional morphemes. In the Italian data, MLU is instead calculated on words, using the notion of “word” outlined above. Despite this difference, it should be stressed that previous work indicates that there is an almost perfect correlation between morpheme-MLU and word-MLU in the English Childes data, suggesting that the use of word-MLU in languages other than English does not compromise the validity and reliability of this measure and its comparison to previous analyses of English and other languages (Parker & Brorson, 2005). Asecond manual analysis was additionally performed to better understand the content of the strings counted with the automatic analysis (more detail is presented in the results section). 2 In corro ‘Irun’, corris the stem and -ois abound inflectional morpheme indicating first person singular, present tense. In burro ‘butter’, there is only one morpheme that cannot be decomposed. OPEN ACCESS LUCA CILIBrASI 33 4 RESULTS Counts of utterances can offer an important insight into linguistic processing, and measures of how counts change over time (so called trend analyses) are asimple yet powerful system to investigate child language development (Altman, 1991; Parker & Brorson, 2005). Results obtained with the automatic Grep count are summarized in Table 1. Areport of utterances divided by record is preferred in the table to an average across records, because the average would fail to capture the developmental component of the data. Record 1 word 2 words 3 words 4 words 5 words Age 1 47 2 0 0 0 01;05,04 2 55 13 0 0 0 01;05,18 3 94 14 1 0 0 01;06,02 4 79 12 4 1 0 01;06,22 5 104 27 2 0 1 01;07,06 6 104 33 1 0 1 01;07,19 7 134 40 7 0 0 01;08,03 8 88 37 7 2 0 01;08,17 9 122 26 7 1 0 01;09,01 10 89 29 5 0 0 01;09,15 11 90 32 11 0 0 01;09,29 12 80 36 6 5 1 01;10,12 13 107 34 10 3 0 01;11,15 14 103 32 10 3 0 02;00,00 15 111 57 24 4 3 02;00,14 16 84 35 14 3 0 02;00,27 17 109 44 16 5 4 02;01,11 18 56 16 18 6 4 02;01,27 19 82 44 21 8 5 02;02,11 20 57 32 22 8 4 02;03,02 21 44 29 15 5 3 02;03,02 22 56 37 20 9 3 02;03,15 23 67 40 23 8 7 02;03,29 24 75 36 22 18 5 02;04,13 25 100 54 22 8 3 02;04,25 26 98 54 32 11 7 02;05,10 27 53 46 14 10 6 02;05,24 Table 1:Counts of utterances. The data obtained with the Grep count were analyzed using correlations. To understand whether these counts were significantly changing over the child’s development, correlations were run between age and scores in the different kinds of utterances (see OPEN ACCESS 34 STUDIE ZAPLIKOVANÉ LINGVISTIKY 2/2021 Altman, 1991 for ajustification of this method in the analysis of trends over time). Results can be found in Table 2. r t p 1W –0,237 –1,217 0,235 2W 0,686 4,710 < 0,001* 3W 0,893 9,900 < 0,001* 4W 0,868 8,744 < 0,001* 5W 0,833 7,539 < 0,001* Table 2:Correlation coefficients with age across different utterance lengths. Asmall negative coefficient was obtained for the 1-word utterances, and large and very large coefficients were obtained for all multi-word utterances (see Hemphill, 2003, for guidelines on how to interpret the magnitude of correlation coefficients). ABonferroni adjusted alpha of 0,01 was used as threshold for significance (0,05/5). All correlations reached significance except for the one between age and the count of one-word utterances. This result suggests that all multi-word utterances increased in number (based on the positive sign of the coefficient) over the time period recorded, while one-word utterances did not decrease significantly (the negative coefficient indicates that they decreased in absolute terms). The results are shown visually in Figure 2: For each specified length, Ipresent the trending line to make the development easier to read. From this graph, one can see that the number of expressions made up of one word tends to decrease over time overall, even though the trend is not significant, while the number of expressions made up of two, three, four and five words tends to increase, with significant trends. This finding is not surprising and could indicate the child’s gradual abandonment of holophrases in favor of syntactic constructions, but the decrease in the use of oneword utterances appears to be slow and it certainly does not happen automatically with the first multi-word productions. The automatic count of words is auseful measure, and it offers an excellent estimate of the competence and the productivity of the child. However, it has the intrinsic limit of not analyzing structures for what they contain, but only for how long they are (Vasilyeva et al., 2008). Asecond analysis was then performed looking for specific structures that the child may be developing at this age. The chosen structures, with Italian examples and English translations, are provided below: — Determiner + Noun (i.e., la casa ‘the house’) — Negation + Verb (i.e., non voglio ‘Idon’t want’) — Whquestions (i.e., cosa è? ‘what is that?’) — Subject + Verb (i.e., io mangio ‘Ieat’) — Verb + Adverb (i.e., voglio ancora ‘Iwant more’) — Copula + Adjective (i.e., è piccolo ‘it is small’) — Auxiliary + Verb (i.e., hanno fatto ‘they did’) — Verb + Complement (i.e., voglio la palla ‘Iwant the ball’) OPEN ACCESS LUCA CILIBrASI 41 University Press. https://doi.org/10.1093/ oxfordhb/9780195396683.013.0019 Dryer, M.S. (1992). The Greenbergian word order correlations. Language, 68, 81–138. https://doi.org/10.2307/416370 Everitt, B., & Skrondal, A. (2010). The Cambridge Dictionary of Statistics. Cambridge University Press. Fukui, N. (2017). Merge in the Mind-brain: Essays on Theoretical Linguistics and the Neuroscience of Language. Taylor & Francis. Gervain, J., Nespor, M., Mazuka, R., Horie, R., & Mehler, J. (2008). Bootstrapping word order in prelexical infants: AJapanese–Italian cross-linguistic study. Cognitive psychology, 57(1), 56–74. https://doi.org/10.1016/j. cogpsych.2007.12.001 Grewendorf, G. (2012). Wh-movement and topic movement. In L.Brugé, A.Cardinaletti, G.Giusti, N.Munaro & C.Poletto (Eds.), Functional Heads: Cartography of Syntactic Structures, Vol.7 (pp. 91–120). Oxford University Press. https://doi.org/10.1093/ acprof:oso/9780199746736.003.0004 Guasti, M. T. (2017). Language Acquisition: The Growth of Grammar. MIT press. Guasti, M.T., Lange, J.D., Gavarró, A., & Caprin, C. (2004). Article omission: Across child languages and across special registers. LOT Occasional Series, 3, 199–210. https://doi. org/10.1080/10489220801937346 Haspelmath, M. (2017). The indeterminacy of word segmentation and the nature of morphology and syntax. Folia linguistica, 51, 31–80. https://doi.org/10.1515/flin-2017-1005 Hemphill, J.F. (2003). Interpreting the magnitudes of correlation coefficients. The American Psychologist, 58(1), 78–79. https://doi. org/10.1037/0003-066X.58.1.78 Hudson, R.A. (2004). Are determiners heads? Functions of Language, 11(1), 7–42. https://doi. org/10.1075/fol.11.1.03hud Ibbotson, P. (2013). The scope of usage-based theory. Frontiers in Psychology, 4, 255. https:// doi.org/10.3389/fpsyg.2013.00255 Josiah, U.E., & Udoudom, J.C. (2012). Morphophonemic analysis of inflectional morphemes in English and Ibibio nouns: Implications for linguistic studies. Journal of Education and Learning, 1(2), 72–81. https://doi. org/10.5539/jel.v1n2p72 Le Normand, M.T., Moreno‐Torres, I., Parisse, C., & Dellatolas, G. (2013). How do children acquire early grammar and build multiword utterances? Acorpus study of French children aged 2 to 4. Child Development, 84(2), 647–661. https://doi.org/10.1111/j.14678624.2012.01873.x Lock, A., & Zukow‐Goldring, P. (2010). Preverbal communication. In J.G.Bremner & T.D.Wachs (Eds.), The Wiley‐Blackwell Handbook of Infant Development (pp. 394–425). Whiley-Blackwell. Marcus, G.F. (1996). Why do children say “breaked”? Current Directions in Psychological Science, 5(3), 81–85. https://doi. org/10.1111/1467-8721.ep10772799 MacWhinney, B. (2014). The CHILDES project: Tools for analyzing talk. Volume I: Transcription format and programs. Psychology Press. Parker, M.D., & Brorson, K. (2005). Acomparative study between mean length of utterance in morphemes (MLUm) and mean length of utterance in words (MLUw). First Language, 25(3), 365–376. https://doi. org/10.1177/0142723705059114 Poeppel, D., & Wexler, K. (1993). The full competence hypothesis of clause structure in early German. Language, 69(1), 1–33. https:// doi.org/10.2307/416414 Rice, M.L., Smolik, F., Perpich, D., Thompson, T., Rytting, N., & Blossom, M. (2010). Mean length of utterance levels in 6-month intervals for children 3 to 9 years with and without language impairments. Journal of Speech, Language, and Hearing Research, 53, 333–349. https://doi.org/10.1044/10924388(2009/08-0183) Rizzi, L. (2002). On the grammatical basis of language development. In G.Cinque & R.S.Kayne (Eds.), The Oxford Handbook of Comparative Syntax (pp. 70–109). Oxford University Press. https://doi.org/10.1093/ oxfordhb/9780195136517.001.0001 Siegler, R.S. (2004). U-shaped interest in u-shaped development— and what it means. OPEN ACCESS 42 STUDIE ZAPLIKOVANÉ LINGVISTIKY 2/2021 Journal of Cognition and Development, 5(1), 1–10. https://doi.org/10.1207/s15327647jcd0501_1 Sugisaki, K. (2005). Early acquisition of basic word order: New evidence from Japanese. In A.Brugos, M.R.Clark-Cotton & H.Seungwan (Eds.), Proceedings of the 29th Annual Boston University Conference on Language Development (pp. 582–591). Cascadilla Press. Taatgen, N.A., & Anderson, J.R. (2002). Why do children learn to say “broke”? Amodel of learning the past tense without feedback. Cognition, 86(2), 123–155. https://doi. org/10.1016/S0010-0277(02)00176-2 Tomasello, M. (2001). The item-based nature of children’s early syntactic development. In M.Tomasello & E.Bates (Eds.), Language Development: The Essential Readings. Essential Readings in Developmental Psychology. (pp. 169–186). Blackwell Publishing. Tomasello, M. (2009). The usage-based theory of language acquisition. In E.L.Bavin (Ed.), The Cambridge Handbook of Child Language (pp.69–87). Cambridge University Press. Tonelli, L. (2004). The Tonelli corpus. CHILDES Database. Vasilyeva, M., Waterfall, H., & Huttenlocher, J. (2008). Emergence of syntax: Commonalities and differences across children. Developmental Science, 11(1), 84–97. https://doi.org/10.1111/ j.1467-7687.2007.00656.x Vygotsky, L.S. (2004). Imagination and creativity in childhood. Journal of Russian & East European Psychology, 42(1), 7–97. https:// doi.org/10.1080/10610405.2004.11059210 Wexler, K., Gavarró, A., & Torrens, V. (2004). Feature checking and object clitic omission in child Catalan and Spanish. In R.BokBennema, B.Hollebrandse, B.KampersManhe and P.Sleeman (Eds.), Romance Languages and Linguistic Theory 2002. Selected papers from ‘Going Romance’, Groningen, 28–30November 2002 (pp. 253–269). John Benjamins. https://doi.org/10.1075/ cilt.256.15wex Zaccarella, E., & Friederici, A.D. (2015). Merge in the human brain: Asub-region based functional investigation in the left pars opercularis. Frontiers in Psychology, 6, 1818. https://doi.org/10.3389/fpsyg.2015.01818 Luca Cilibrasi | Department of English Language and ELT Methodology, Faculty of Arts, Charles University <[email protected]> OPEN ACCESS