1 Introduction

An ongoing debate within generative discussions of syntactic acquisition concerns how and when functional categories (i.e., non-lexical categories encoding grammatical properties) are acquired.

A traditional distinction in the literature is between so-called maturational and continuity approaches; these contrast with respect to whether all/most knowledge encoded in Universal Grammar (UG) is available from the start, with the former positing gradual development of UG principles and hierarchies (see, e.g., de Villiers 2001, for a review). Among the many issues grappled with in these frameworks, two stand out as central:

    1. (1)
    1. Are functional categories innate or (partly) emergent?
    1. (2)
    1. If innate, to what extent is syntactic category formation directed by a biological program (maturation)?

Given the wider commitments of the generative framework, (1) has thus far usually been answered by favouring innate categories (Diercks et al. 2023 is a recent exception). Nonetheless, the use of emergent categories in generative syntax and typology also has its proponents (e.g., Ramchand & Svenonius 2014; Ritter & Wiltschko 2014; Wiltschko 2014; Biberauer 2019; Larson 2021). With respect to (2), the choice between continuity and maturation remains highly debated; the former assumes complete access to UG content from the start, while the latter proposes instead gradual (maturation-guided) development of UG knowledge. Generative proposals work under a tripartite model of the structure of clauses where these are organised into three key clausal domains (e.g., Grohmann 2003) – vP/VP, which encodes thematic and argument-structural relations, TP, which establishes tense and/or other deictic properties, and CP, which encodes discourse information. Maturational and continuity proposals differ, then, in how many of these domains (if any) are available to learner at the start of grammar development.

Within maturational approaches, the dominant proposed directionality of development has arguably been bottom-up, where structurally lower elements mature first, e.g., vP/VP matures first, before TP, which in turn matures before CP (i.a., Radford 1990; Rizzi 1993; Friedmann et al. 2021). However, inward perspectives (where CP and/or vP mature first) have also been put forward, adducing empirical data that challenges bottom-up development (i.a., Galasso 2003; Tsimpli 2005; Heim & Wiltschko 2025). Therefore, different implementations of both continuity and maturation have been proposed, also reflecting the progressive addition of new theoretical machinery (e.g., cartography), to account for new developmental patterns. The wide range of approaches proposed to date (Strong Continuity, Weak Continuity, several articulations of maturation, etc.; see section 2) plausibly indicate the non-negligible, but still somehow constrained, variation in crosslinguistic developmental patterns and undoubtedly also the difficulty of modelling acquisition with this toolkit.

These contentious issues have remained to the present day. However, their prevalence has led to other, equally interesting empirical questions being addressed less. One of these questions forms the point of departure for this paper:

    1. (3)
    1. What is the categorial granularity of children’s representations at various developmental stages?

As for (3), the granularity (roughly, detailedness) of children’s syntactic knowledge in the literature thus far has typically been dictated by UG specifications. Work of cartographic orientation generally postulates a cartographic spine in UG, with the resulting acquisition theory being based on its projections (e.g., Westergaard 2009; De Lisser et al. 2017; Friedmann et al. 2021). On the other side, where acquisitionists posit more pared-down spines (e.g. those centred on a basic CP>TP>vP; i.a., Galasso 2003, Tsimpli 2005, Vainikka & Young-Scholten 2011), cartographic-type syntactic patterns are set aside in their acquisition studies. Two examples of more minimalist and cartographic-type functional projections are given in (4)–(5), respectively. This illustrates differences across frameworks in categorial granularity: cartographic-style approaches postulate richer functional structure that elaborates each of the three clausal domains. In both cases, children’s granularity is fixed across all stages, bound by the functional hierarchy in UG. As a result, (3) has generally not been treated as a direct object of study thus far.

    1. (4)
    1. (5)

But it does not follow that granularity should merely be taken for granted on a priori grounds. This is especially true if one acknowledges that there may be crosslinguistically variable degrees of elaboration of functional structure, like the left periphery, as underlined by several proposals (i.a., Ramchand & Svenonius 2014; Biberauer & Roberts 2015; Biberauer 2018; see section 2.4). Indeed, after positing a given fixed UG spine (e.g., a cartographic one), evidence in children’s data is needed, across all stages, to motivate the postulation of, for example, an articulated CP-domain. Otherwise, a given granularity is simply presupposed prior to the relevant developmental study, and is thus not falsifiable unless explicitly interrogated. More specifically, we can ask, if we accept (some of) the arguments for fine-grained cartographic functional sequences: how and when do children acquire the rich functional hierarchies posited typologically (e.g., Rizzi 1997; Cinque 1999; Svenonius 2006) and when does evidence for this surface in their data? (3) is, then, a closely linked, but separate, question from (1) and (2).

Against this background, this paper has three goals:

Firstly, I propose that categorial granularity (henceforth, simply ‘granularity’) should not just be a theoretical prior, but must be probed, just like (1)–(2) are. The paper’s primary goal, then, is to present an initial case study demonstrating the importance of considering (3). I summarise a study on the development of the left periphery in 10 children across 5 languages. A contrast emerges from the data: while CP-structures emerge from the earliest files, evidence for co-occurrence of CP-elements (motivating a Split CP-type structure) appears systematically at a later stage. This observation forms the basis of this paper.

The paper’s second goal is to problematise the directionality component of (2), particularly once cartography is adopted: do crosslinguistic developmental patterns align with a directionality-driven approach to the acquisition of functional spines? I will argue that multilingual corpus data provides significant insights into the absolute pathways predicted by maturation, with the present data suggesting a negative answer to this question. Topicalisation and illocutionary complementisers provide the crucial empirical evidence to this effect.

The resulting empirical picture challenges, to different extents, most extant approaches to development, which either undergenerate or overgenerate. I suggest that granularity provides the flexibility needed to interpret the data.

The third and final goal is to provide an account of the data following neo-emergentist generative perspectives on development (following Biberauer 2011, et seq. and Biberauer & Roberts 2015),1 as well as to highlight some predictive similarities and differences with other existing frameworks. This paper aims to show how, by conceiving of categorial development as (at least partly) emergent, we arrive at an approach to development where these otherwise problematic patterns fall into place. This is my final theoretical contribution.

The paper is organised as follows: section 2 summarises extant approaches to syntactic development. Here, I also identify the underexplored and testable aspect of syntactic development addressed above, namely a possible role for categorial granularity. I highlight the contrast with the focus of the (maturational) literature thus far – developmental directionality. Section 3 introduces the multilingual corpus study, which addresses the three important aspects of development outlined in (1)–(3); I summarise its methodology, diagnostics and results. This section demonstrates that current approaches (especially bottom-up maturational approaches; section 2.1) face important challenges: the data show very early emergence of several CP-structures, including structurally very high ones, but systematically late emergence of structures indicating the child is harnessing an elaborate, “split” CP. I take these patterns as initial evidence to motivate the adoption of granularity as a potentially productive tool when modelling development. I develop this proposal in section 4, where I also evaluate the approaches discussed in light of the data. Section 5 concludes.

2 Background: approaches to the acquisition of functional categories

Traditionally, approaches to syntactic development have been divided into two main kinds – maturational and continuity – depending on the amount of innate linguistic structure assumed to be accessible to the child at the start of the learning path. I begin first with maturational approaches.

2.1 Maturational approaches

“Incomplete structure” or maturational approaches argue that children’s grammatical knowledge is not (fully) available initially and that grammatical categories and principles appear gradually. The key proposal is to operationalise this “gradual” development in terms of syntactic maturation: a UG-given spine (whose exact articulation varies across proposals) and its categories develop (“mature”) with a specific directionality, determined by UG. The dominant directionality harnessed in maturational approaches has been a bottom-up one, whereby domains projected by structurally lower heads (e.g., vP/VP) are predicted to be acquired before higher ones (e.g., CP). This hypothesis finds its roots in proposals such as Radford’s (1988; 1990) Small Clause Hypothesis and Rizzi’s (1993) Truncation Hypothesis.

Since the focus of this paper lies on the left periphery, I will concentrate on a more recent cartographic and left-periphery-centred interpretation of bottom-up maturation, namely Friedmann et al.’s (2021) Growing Trees Hypothesis. However, the results in section 3 will have implications not just for Growing Trees, but for any version of bottom-up maturation. In Growing Trees, the development of the left periphery is composed of two stages: at an initial Stage 1, the CP-domain is yet to develop. Only IP/TP and VP are available to the child, permitting, for instance, the production of inflection and A-movement. A later Stage 2 sees the development of the lower left periphery, up to QP (hosting, i.a., wh-elements). The entire cartographic hierarchy, from TopP until ForceP, becomes available at Stage 3, when, e.g., topicalisations and embedding are expected to appear.2 The full developmental trajectory proposed is schematised in Figure 1. This thereby predicts intermediately late emergence of wh-questions (before higher projections, but after VP/IP) and very late emergence of elements like topics or complementisers.

Figure 1: Stages in the Growing Trees Hypothesis (Friedmann et al. 2021: 12).

The above has discussed bottom-up maturation. Now I turn to inward maturation,3 which has received much less attention than the dominant bottom-up approaches. Simplifying grossly, inward approaches can be defined as approaches that assume development to begin at vP/VP and CP (the phasal edges, from the perspective of the CP>TP>vP functional sequence), before maturation of TP. Under inward maturation, then, we expect early development of (some) CP knowledge (alongside (some) vP knowledge).

Arguably, the first maturational approach that, in retrospect, argues for this progression can be traced back to Galasso’s (2003) “Empty Middle” approach for child English, which proposes a revision of Radford’s (1990) Small Clause hypothesis: first, CP>VP are acquired; then, the full CP>IP>VP matures. The approach assumes a VP structure for basic ‘telegraphic’ declaratives (with Radford), CP>VP for basic interrogatives at his stage 1 and, finally, the full blown CP>IP>VP at stage 2, when inflection, auxiliaries and inversion start being acquired. For a slightly similar hypothesis, see Tsimpli (2005), whose maturational pathway hinges instead on the development of uninterpretable vs. interpretable features.

More recently, Heim & Wiltschko (2025) outline an explicit inward-growing maturational approach, the Inward Growing Spine Hypothesis, adducing support from the development of the English discourse particle huh. This hypothesis draws on two associated models of clausal organisation: the Universal Spine Hypothesis (Wiltschko 2014) and the Interactional Spine Hypothesis (Wiltschko 2021). These are jointly represented in Figure 2.

Figure 2: The Bridge Model (Hinzen & Wiltschko 2023: 75).

In this approach, the thematic and categorisation-based domain (centring on Wiltschko’s 2014 Classification head) and the interactional and speech-act-based domain (Wiltschko’s 2021 Responding head) are hypothesised to develop first, along with Linking (broadly speaking, the CP, hosting, e.g., wh-words and complementisers). The subsequent steps involve gradual ‘unfolding’ into an adult-like spine: first, the child develops the Linking layer further into two projections (a Grounding layer, above Linking and an Anchoring layer, below Linking and roughly equivalent to TP). Finally, the Anchoring layer is divided into a deictic and a perspectival anchor (e.g., Tense and Aspect); Grounding is divided into a speaker-oriented (GroundSpkr) and an addressee-oriented layer (GroundAddr) (Heim & Wiltschko 2025: 15ff).

This is visualised in Figure 3. Since the present paper largely does not discuss interactional language concerning most of the speech-act layers assumed by Heim & Wiltschko, I will not be able to test this specific hypothesis, but a more general version of an inward-growing approach (e.g., Galasso 2003), discussed in section 2.5. However, I will briefly return to Heim & Wiltschko (2025) in section 4 in light of the results observed.

Figure 3: Stages of the Inward Growing Spine Hypothesis, applied to huh (Heim & Wiltschko 2025: 23).

2.2 Continuity approaches

Continuity approaches, on the other hand, claim that the functional structure of children’s initial grammar is not significantly different from adults’ grammars. The extent to which child and adult grammars are taken to be isomorphic varies, with both so-called Weak Continuity (i.a., Clahsen 1990; Clahsen et al. 1994; 1996) and Strong Continuity approaches (i.a., Pinker 1984; Boser et al. 1992; Hyams 1992; 1994; Verrips & Weissenborn 1992; Poeppel & Wexler 1993) having been put forward. For example, Poeppel & Wexler (1993) and Boser et al. (1992), both on the basis of V2 data in child German, suggest that the child grammar includes the categories and operations in the adult grammar (a CP domain which hosts V2-topics in its specifier, and moved finite verb in C; den Besten 1983) from the start, with no period in the production displaying a target-deviant stage with finite verbs in OV position.4

A relevant cartographic continuity perspective on the left periphery is Westergaard’s (2009) Micro-cues approach. Westergaard (2009) argues for sensitivity to fine-grained cartographic and information-structural content from the earliest stages of acquisition; in particular, children seek “micro-cues” in the Primary Linguistic Data (i.e., the input) that signal the specific configuration of the adult grammar they are exposed to (e.g., the micro-cue for Norwegian (non-V2) embedded wh-questions is hypothesised to be [whP wh … [VP V]], ch. 3). The claim, then, is that children are successful in deploying the knowledge distilled from these micro-cues from the earliest multi-word combinations.

As a result, all continuity approaches share the intuition that functional structure is an early phenomenon, with all of CP, TP and vP/VP being accessible (in some form or another) from the start. More specifically for present purposes, they predict early attestation of CP-structures.

2.3 Interim summary: a prediction with innate categories

Taking stock of the past two sections (maturation and continuity), maturation postulates universal orders of acquisition of structures: UG not only specifies the functional spine shaping all languages, but also the specific order in which its categories will be acquired. Notice, crucially, that what changes throughout development is not the UG-given tree; it is how much of that UG-given spine children have access to at different stages, subject to a given maturational directionality (bottom-up or inwardly). The granularity of the syntactic categories and the spine postulated is fixed or static: if a cartographic spine is the UG prior, then children will acquire cartographic domains (internally layered with multiple projections) once they mature. Mutatis mutandis, the same point holds for the continuity approaches just discussed, which assume full access to UG content from the start; Westergaard (2009: 104), for example, argues that children harness cartographic left-peripheral knowledge “from the onset of multiword utterances”. As above, children operate, constantly, with the granularity specified by the analyst as being provided by UG (here: cartographic).

The one exception here is Heim & Wiltschko (2025), who already adopt a more explicit notion of granularity, via successive divisions into their full, adult-like speech-act projections; however, they explicitly do not adopt a cartographic CP. This is something I return to in the next section and section 4.

Importantly, as also noted in passing by Westergaard (2009), Moscati & Rizzi (2021: 16) and Moscati (2023: 3), this theoretical commitment makes a testable prediction with respect to children’s categorial granularity: once (or soon after) CP develops, we should observe (some) evidence that the child is indeed harnessing a CP with elaborate structure. Our expectation, therefore, is that we should see evidence in the child data, once they start producing/comprehending CP-structures, that postulating a fine-grained CP-domain is justified, and is not merely taken for granted on theory-internal grounds. This prediction will be a key focus of the corpus study in section 3.

2.4 Neo-emergentist approaches

The approaches introduced thus far generally adopt what we could term a “fixed granularity” perspective on acquisition, with children’s categorial granularity being specified by UG. This assumption that categorial granularity never changes throughout development is, however, not a necessary one. This section discusses so-called neo-emergentist approaches (Biberauer 2011, et seq. and Biberauer & Roberts 2015), one of whose central claims is that increasing (i.e., flexible) granularity is, in fact, a key aspect of syntactic development.

Capitalising on the consequences and opportunities that a minimal UG brings, Biberauer (2019) endorses a Three Factors approach, in turn proposing an explicit third-factor principle guiding acquisition (Maximise Minimal Means, MMM). Importantly for present purposes, she introduces the essence of a neo-emergentist approach: acquirers are steered by a minimal UG, containing Merge, Agree and some notion of [F], and by MMM. The latter is a general-cognitive bias leading learners to make maximal use of minimal resources. This leads acquirers to generalise already-acquired knowledge as much as possible and only postulate new formal content (e.g., [F]s) if required by the Primary Linguistic Data and if it fulfils a contrastive role in the system (for greater detail on this approach and MMM, see Biberauer 2019). The essence crucially remains fundamentally generative: systems are “guided” by a minimal UG, making systems both Merge- and formal-feature-regulated (Biberauer 2019: 59). The emphasis, however, lies on exploring the empirical productivity of impoverishing UG, in line with Chomsky (2005). In contrast to maturational approaches, neo-emergentism suggests that development is best modelled via emergent syntactic categories.

Specifically, I will focus on Biberauer & Roberts’ (2015) emergent categorial hierarchy. Firstly, according to Biberauer & Roberts (2015), crosslinguistically, clauses can be analysed at different levels of ‘magnification’ or ‘granularity’. Having briefly discussed granularity in broad conceptual terms thus far, I define it more explicitly below, following Song (2019: 21):

(Syntagmatic) granularity: A sequence of combinatorially oriented categories may be collapsed into a single category (e.g., X–Y–Z→A), and a single category may be split into a sequence of categories (e.g., X→X1–X2–…–Xn).

For example, the relation between the Core Functional Category C (Chomsky 2000; 2001) and the cartographic sequence Force–Topic–Focus–Fin (Rizzi 1997) is one of (syntagmatic) granularity; the former can be split into the latter and, vice versa, the latter can be collapsed into the former. The major granularity levels I assume throughout this paper are the following (Biberauer & Roberts 2015: 4), from less to more fine-grained (as the nominal domain is orthogonal to the present study, I restrict exposition to the clausal domain):

    1. (6)
    1. Extended Projection(V) > phase (C, v) > Core Functional Category or CFC (C, T, v) > “cartographic field” (e.g. Tense, Mood, Aspect, Topic, Focus) > semantically distinct head (as in Cinque 1999; Speas & Tenny 2003; Frascarelli & Hinterhölzl 2007; Haegeman & Hill 2013, among others).

At the highest level of organisation, we find the Extended Projection (EP) in the sense of Grimshaw (1991: et seq.), referring to a maximal projection which projects from a category that it shares categorial features with. Thus, the whole clausal EP is C–T–V, with all categories sharing the features [+V, –N]. Here, the only featural contrast in the system may be between predicates and nouns (±V or ±N). At a second granularity level, we encounter Phasal distinctions (C, v; Chomsky 2008). These are located at a coarser-grained level than the three Core Functional Categories C, T and v (Chomsky 2000; 2001), given that T is not inherently a phase. Then we encounter the finer-grained cartographic fields (e.g., Force–Topic–Focus–Fin) and semantically more fine-grained heads such as those in Cinque (1999). The former represent iterations of C-positions (i.e., sequences of heads with [C]), each associated with a grammaticalised semantic feature (e.g., Topic with [topic]). The exact elaboration of the left periphery is therefore taken to be a matter of crosslinguistic variation, just like in earlier and contemporary work (i.a., Giorgi & Pianesi 1997; Bobaljik & Thráinsson 1998; Walkden 2017; Hsu 2017; Larson 2021). I will focus in particular on the contrast between C as a CFC/Phasal category vs. its potential cartographic elaborations. In parametric terms, the cartographic level of granularity corresponds to the microparametric level of variation, while coarser-grained levels recruit macro/mesoparametric options.

In turn, (6) can be reformulated as an emergent categorial hierarchy, which unifies and identifies overall commonalities between independently proposed hierarchies and categorial sequences in the literature. The hierarchy thus begins at an initially ‘undivided’ categorial space and gradually increments the level of grain with which the system is analysed (Biberauer & Roberts 2015: 6):

The key for present purposes is that Figure 4 can be conceptualised as a learning path. The emergent hierarchy is regulated in a contrastivist manner (i.a., Hall 2007; Dresher 2009; Cowper & Hall 2014), with acquirers postulating ever finer-grained categorial distinctions once contrastive features are detected in the input. Although not represented in the original hierarchy in Biberauer & Roberts (2015: 6) for simplicity, it is not until a later stage, after the emergence of the CFCs, that the approach expects cartographic heads and semantically distinct heads (as in Rizzi 1997, Cinque 1999, respectively) to emerge and further subdivide each of the CFCs. Being a phasal domain, this approach predicts CP to develop before TP, like inward approaches, but proposes this CP will remain “basic” (non-split) in organisation until a later stage. I will centre on the prediction that acquirers will first entertain a “basic” CP-domain (of the Phasal or CFC kind) at an early stage, before cartographic-type knowledge is considered (space considerations preclude discussion of TP, but for discussion see Bosch 2023a, where the emergence of TP-structures is analysed in light of this hypothesis). Put differently, acquisition is such that microparametric distinctions are not posited unless a macro/mesoparametric analysis of the Primary Linguistic Data cannot be maintained (Biberauer & Roberts 2015: 14). The acquirer is conservative when it comes to positing higher levels of granularity and formal features, but liberal when generalising already-existing formal distinctions (viz. Feature Economy, Roberts & Roussou 2003, and Input Generalisation, Roberts 2007/2021; and Maximise Minimal Means, which unifies the two).5

Figure 4: Biberauer & Roberts’s (2015) emergent categorial hierarchy, adapted to include a cartographic CP.

Biberauer & Roberts (2015) converge, then, with the already-discussed inward-growing and continuity approaches (section 2.1) which have also, in various forms, argued for early emergence of the CP-domain, and against a strictly bottom-up approach to development. However, and in contrast to the approaches discussed thus far, evidence for cartographic-type knowledge should come visibly late in the production data – this is because categories granularise in development, elaborating on developmentally-prior knowledge. Therefore, this perspective explicitly envisages changes in granularity during development, which correlate with the acquisition of macro/microparametric knowledge. Crucially, some previous work already offered very suggestive speculations in this direction. Roeper & de Villiers (2011: 9) noted that “one possibility […] is that the child begins with a proto-CP, and it undergoes an actual process of Splitting as suggested by Hollebrandse & Roeper (1998) to end up with a distinct series of nodes with different functions”. A similar logic is adopted in Soares (2006) for European Portuguese data.

The current approach is also in principle synergetic with Heim & Wiltschko (2025), which, as noted above, incorporates some notion of granularity, e.g., their final stage involves “subdividing” their Ground head into GroundAddr and GroundSpkr. Given that their hypothesis is not cartographic and largely concerns projections above CP, I will mostly set it aside. However, future work should consider the clear overlap between the approach proposed here, Wiltschko’s (2014; 2021) categorial template and Heim & Wiltschko’s (2025) Inward Growing Spine.

Overall, then, the intuition we will pursue in the corpus study is as follows: granting that various degrees of granularity are needed to capture crosslinguistic variation (per Biberauer & Roberts 2015), we will not, a priori, assume a specific granularity (in UG or elsewhere) as the starting point for children. Instead, we “let the data decide”, taking child productions to be informative as to the granularity acquirers are most likely to be harnessing at various stages. This is important, insofar as it inquires into not just the “earliness” of functional domains (e.g., CP, TP, etc.), but also their granularity throughout development. The latter represents an almost undiscussed aspect of syntactic representations in development and a novel way of approaching corpus data.

2.5 Predictions

Summarising the preceding discussion, the architectural differences of the approaches introduced, as well as their predictions for the left periphery, can be broken down in terms of the following four criteria:

  • (i) Biological locus of syntactic categories (innate vs. emergent).

  • (ii) Postulation of syntactic maturation and its directionality (if applicable).

  • (iii) Acquisition of CP (early vs. late).

  • (iv) Fixed vs. flexible granularity.

These are compared in Table 1, which incorporates the approaches introduced thus far.

Table 1: Architectural differences between approaches to syntactic development.

Approach Maturation Locus Acquisition of CP Granularity
Bottom-up approaches (Growing Trees) Bottom-up Innate Late (two-stage) Fixed
Inward approaches Inwardly Innate Early Fixed
Continuity ✗ Innate Early Fixed
Neo-emergentism ✗ Emergent Early Flexible

All four groups of approaches make clearly testable predictions. Inward-growing, continuity and neo-emergentist approaches anticipate early acquisition of (some) CP-structures, irrespective of their structural height in the left periphery. Bottom-up approaches, on the other hand, maintain that CP is a late(r) development, relative to TP. More specifically, the hypothesis under consideration here (Growing Trees) suggests this development is two-stage, such that the structurally highest heads (Topic, Force, etc.) mature after left-peripherally lower elements (like QP/FocP, which hosts, among others, wh-questions). I focus in particular on the committment that there will be a relative order of emergence of left-peripheral structures correlating with structural height. Importantly, neo-emergentism is the only approach making explicit predictions with respect to granularity. In this approach, granularity increases throughout development, by virtue of its emergent categories, and this should accordingly be reflected in children’s competence and production: as such, CP-structures (on their own) may emerge early, but we do not anticipate more specialised (granular) CP-structures to be combined until a later stage. The relevant split CP-structure is assumed to be featurally more complex, so acquisitionally harder. Notice therefore that neo-emergentism does not adopt syntactic maturation,6 in the sense that the relevant finer-grained categories are hypothesised to be emergent, not innate. All other approaches typically do not entertain changes in granularity; the granularity children operate with is fixed (hard-wired) by the functional structure in UG. This distinction between predicting “fixed” vs. “flexible” granularity is operationalised in the diagnostics introduced in section 3.2.

3 The corpus study

3.1 Children studied

The spontaneous speech of 10 children across 5 languages in the CHILDES database MacWhinney (2000) was analysed; this involved 2 children per language, which included Catalan, Italian, Spanish, German and Dutch. Overall, the period in which children’s production was analysed ranged from 0;11 to 5;02, across all children.

The details of the children studied, including their word-based MLU (MLUw), are reported in Table 2. Henceforth, all examples and tables will be presented in the order of languages and children given below.

Table 2: Children studied in the CHILDES database and summary information (Serra & Solé 1989; Cipriani et al. 1989; Llinàs-Grau & Ojea 2000; Montes 1987; Miller 1979; Bol 1995; van Kampen 2009).

Language Corpus Child Files analysed Age range MLUw range Total words Total utterances
Catalan Serra-Solé Laura 19 1;07–4;00 1.03–3.47 16651 7576
Gisela 21 1;07–4;02 1.02–3.51 11132 4753
Italian Calambrone Martina 13 1;07–2;07 1.26–2.69 7569 3586
Rosa 21 1;07–3;03 1.27–3.24 14659 6614
Spanish Llinàs-Ojea Irene 59 0;11–3;02 1.0–5.13 38234 11908
Montes Koki 13 1;07–2;11 1.96–3.61 11773 4276
German Miller Kerstin 37 1;03–3;04 1.09–2.89 28609 16686
Simone 507 1;09–2;09 1.52–4.89 45407 21371
Dutch Groningen Josse 28 2;0–3;04 1.2–4.01 30066 11120
van Kampen Sarah 50 1;06–5;02 1.07–6.07 49465 17408

I will adopt MLUw as the guiding developmental metric throughout, over age, given the high variability observed among children of the same age with respect to their linguistic development (see, for example, the discussion in Paradis & Genesee 1997; Caprin & Guasti 2009). This will also facilitate crosslinguistic and inter-speaker comparisons, as well as the replicability of these findings. The choice of word-based MLUs over morpheme-based metrics is due to two considerations: (i) not all corpora in CHILDES are morphosyntactically tagged, meaning morpheme-based MLUs are not available in practice; (ii) morpheme-based MLUs are more prone to annotation errors and subjectivity/arbitrariness, as the annotator has to establish which morphemes are productive at various stages (Ezeizabarrena & García Fernández 2017). Nonetheless, age will still be reported. Some notes of caution about the stability of MLU as a metric (either word- or morpheme-based) will be offered in section 4.

3.2 Research questions and diagnostics used

I conducted a multilingual corpus study to probe the following empirical and theoretical questions, synthesised from the preceding theoretical review in section 2:

  • i. How early do CP-structures emerge in the child production data? Are these timings in accord with the theoretical approaches introduced (continuity, inward maturation, bottom-up maturation, neo-emergentism)?

  • ii. Do left-peripherally high structures emerge later than lower CP-structures, as in Friedmann et al. (2021)?

  • iii. Do children display knowledge of cartographic-type distinctions within CP from their earliest uses of CP-material? If not, does the data follow the ‘increasing granularity’ prediction by neo-emergentist approaches, where evidence for Split CP-structure should emerge comparatively late in production?

As emphasized in sections 1 and 2, justifying the granularity of the representations assumed for children at various stages is arguably as crucial as determining which domains (whether cartographic or not) are acquired first. This is the novel research question I seek out to address in the next sections. I do so by introducing a new methodological distinction in the data collection between ‘CP-structures’ vs. ‘Split CP-structures’, described below. Research question (iii) is clearly ambitious, I do not aim to provide a conclusive answer to (iii) in this paper, but rather to introduce relevant empirical data that serves as a springboard for future work on this topic.

3.2.1 CP-diagnostics

The study contrasted two sets of structural diagnostics for children’s knowledge of the left-periphery. First, I searched for structures standardly taken to involve the CP-domain. These included V-to-C (V2), wh-questions, yes/no questions, topicalisation and focalisation, illocutionary complementisers and embedding markers.

Evidence for V-to-C movement (or a distributional distinction between finite/non-finite verbs in V2/V-final position) in Dutch and German children will be taken to indicate an emerging CP-domain. More precisely, the study investigates evidence in the same recording of both non-finite verbs in a non-raised (or base) position with OV-order as well as of finite verbs which can be shown to have raised outside of the verbal domain. If the initial constituent is the subject or some non-object topic, these display surface VO-order when the verb is finite. This is illustrated below, where (7a), (8a), (8b) exemplify verb-raising of haben/lust/eet, shown by the absence of surface OV-order, and (7b) displays a combination of a finite modal in its V2-position (willen) and a non-finite verb trinken following its object Kola:8

    1. (7)
    1. German
    1.  
    1. a.
    1. Sie
    2. they
    1. haben
    2. have.3pl
    1. Glück
    2. luck
    1. ‘They are lucky.’
    1.  
    1. b.
    1. Ich
    2. I
    1. will
    2. want.1sg
    1. Kola
    2. Cola
    1. trinken
    2. drink.inf
    1. ‘I want to drink coke.’
    1. (8)
    1. Dutch
    1.  
    1. a.
    1. Ik
    2. I
    1. lust
    2. desire.1sg
    1. geen
    2. no
    1. kaas
    2. cheese
    1. ‘I don’t like cheese.’
    1.  
    1. b.
    1. Nu
    2. now
    1. eet
    2. eat.1sg
    1. ik
    2. I
    1. een
    2. a
    1. boterham
    2. sandwich
    1. ‘Now I eat a sandwich.’

(7a)–(8a) instantiate cases of subject-initial structures with finite verbs. These structures have been analysed as involving raising of the subject and verb to the T-domain only (Zwart 1993; 1997; et seq.). I will, nonetheless, include subject-initial declaratives as evidence of knowledge of V2 (following, i.a., den Besten 1983; Schwartz & Vikner 1989; 1996; van Craenenbroeck & Haegeman 2007), on the grounds that exclusion of subject-initial sentences would effectively reduce the V2-diagnostic to the (non-default) topicalisation diagnostic. This is undesirable, as it would not control for a scenario where knowledge of the V2-system emerges before the first productions of topicalised non-subject elements. This point also becomes important in light of literature noting an initial stage in the acquisition of V-to-C in Germanic where children omit the first-position element, yielding surface V1 (Evers & van Kampen 2001; van Kampen 2010).

Below I illustrate examples of wh-questions (with a wh-element raised to SpecCP, with subject-verb inversion):

    1. (9)
    1. Catalan
    1. Quan
    2. when
    1. comença
    2. begin.3sg
    1. el
    2. the
    1. partit?
    2. match
    1. ‘When does the match start?’
    1. (10)
    1. Italian
    1. Perché
    2. why
    1. l’hai
    2. cl.do=aux.have.2sg
    1. cambiato?
    2. change.ptcp
    1. ‘Why did you change it?’
    1. (11)
    1. Spanish
    1. Como
    2. how
    1. cantó
    2. sing.prfv.3sg
    1. María?
    2. Mary
    1. ‘How did Mary sing?’
    1. (12)
    1. German
    1. Wo
    2. where
    1. hast
    2. aux.have.2sg
    1. du
    2. you
    1. gegessen?
    2. eat.ptcp
    1. ‘Where have you eaten?’
    1. (13)
    1. Dutch
    1. Wat
    2. what
    1. doen
    2. do.1pl
    1. we
    2. we
    1. vanavond?
    2. tonight
    1. ‘What are we doing this evening?’

Although analysed and quantified, wh-less questions of the sort in (14), attested in child Germanic languages, are not considered as sufficient evidence for emergence of the CP-domain. I required other CP-diagnostics to have also emerged for us to conclude that the CP-domain might be in place (this decision only became relevant in one child, Sarah).9 Questions with a wh-word but no verb were excluded.

    1. (14)
    1. German (Schmerse et al. 2013: 658)
    1. Macht
    2. do.3sg
    1. das
    2. the
    1. Pferd?
    2. horse
    1. ‘What is the horse doing?’

In Germanic languages, yes/no questions were included, given that these involve verb-raising to the C-domain. In contrast, polar interrogatives in Romance are typically marked via intonation only, with inversion having been lost for most modern Romance varieties (Giurgea & Remberger 2016). Note that VS-order as a marker of polar interrogatives has been retained in Spanish, coexisting with intonation-only marking (e.g., ¿Está María en casa? vs. ¿María está en casa? ‘Is Mary at home?’; Giurgea & Remberger 2016: 865). However, this order is syntactically ambiguous between a structure with raising to C (inversion properly understood) and one where the verb does not move from T and the subject remains in a lower position. For this reason, too, I exclude all polar interrogatives in Romance.

Topicalisations and focalisations are exemplified in (15)–(19). In Romance, these also include Clitic Left-Dislocation (CLLD), where the topic is accompanied by a clitic (15)–(16), and other cases of contrastive or informational foci (Laka 1990; Rizzi 1997), where a clitic is not required (17).

    1. (15)
    1. Italian, CLLD
    1. Questa
    2. this
    1. la
    2. cl.do=
    1. compro
    2. buy.1sg
    1. io
    2. I
    1. ‘This one I’m buying.’
    1. (16)
    1. Catalan, CLLD
    1. I
    2. and
    1. a
    2. to
    1. mi
    2. me
    1. em
    2. cl.io=
    1. donaràs
    2. give.fut.2sg
    1. un
    2. a
    1. regal?
    2. present
    1. ‘And will you give ME a present?’
    1. (17)
    1. Spanish, Contrastive focus
    1. Arroz
    2. rice
    1. he
    2. aux.have.1sg
    1. comido
    2. eat.ptcp
    1. (no
    2.   not
    1. pasta)
    2. pasta
    1. ‘I have eaten RICE (not pasta).’
    1. (18)
    1. German, Topicalisation
    1. Das
    2. that
    1. glaube
    2. believe.1sg
    1. ich
    2. I
    1. nicht
    2. not
    1. ‘That, I don’t believe.’
    1. (19)
    1. Dutch, Topicalisation
    1. Nu
    2. now
    1. heb
    2. aux.have.1sg
    1. ik
    2. I
    1. mijn
    2. my
    1. bal
    2. ball
    1. vergeten
    2. forget.ptcp
    1. ‘Now I have forgotten my ball.’

Knowledge of embedding was diagnosed via finite embedding structures containing overt embedding markers (e.g., complement clauses, relative clauses, conditionals, temporal, causal, purpose clauses, etc.). Non-finite embedded clauses were not counted as evidence for CP. Embedding markers are abbreviated in the tables in section 3.4.1 as ‘Embed’.

Finally, a further type of main-clause complementisers tracked in the Romance languages are so-called illocutionary complementisers, present primarily in Ibero-Romance (see Corr 2016). Illocutionary complementisers involve instances of main-clause complementisers (such as Cat./Sp. que) that do not function as subordinators heading a complement clause, and instead surface in main-clause contexts with a range of illocutionary functions (Corr 2016; 2022). Adopting Corr’s (2016) terminology, the item que in (20a) generates an exclamation (exclamative que); a quotation in (20b) (quotative que); and in (21a), it contextualises the preceding utterance information, resulting in a conversational move (conjunctive que). In Standard Eastern Ibero-Romance, matrix polar interrogatives can also be introduced with que, as in (21b) (interrogative que). This means interrogative que is sanctioned in Catalan, but is rarer in Spanish.

    1. (20)
    1. Spanish
    1.  
    1. a.
    1. ¡Que
    2. that.excl
    1. casi
    2. nearly
    1. sube
    2. climb.3sg
    1. y
    2. and
    1. rompe
    2. break.3sg
    1. la
    2. the
    1. estantería!
    2. shelves
    1. ‘He/she nearly climbed and broke the shelves!’
    1.  
    1. b.
    1. Que
    2. that.quot
    1. no
    2. not
    1. lo
    2. cl.do=
    1. pienso
    2. think.1sg
    1. hacer
    2. do.inf
    1. ‘(I’ve said) I’m not going to do it.’
    1. (21)
    1. Catalan
    1.  
    1. a.
    1. Assenta’t
    2. sit.down.imp.2sg=cl.refl
    1. que
    2. that.conj
    1. et
    2. cl.io=
    1. pentinaré
    2. comb.fut.1sg
    1. ‘Sit down, I’ll brush/comb you.’
    1.  
    1. b.
    1. Que
    2. that.int
    1. em
    2. cl.io=
    1. pots
    2. can.2sg
    1. ajudar?
    2. help.inf
    1. ‘Can you help me?’

Although most robustly attested in Ibero-Romance, some main-clause complementisers are also attested in Italian (e.g., Dammene un altro che lo mettiamo qui ‘Give me another one, we’ll put it here.’, Che ciascuno scelga una carta! ‘Everyone choose a card!’; Giurgea & Remberger 2016: 874). Therefore, they will also be included in the data analysis. Illocutionary complementisers are typically analysed as being located very high in the left periphery; in a cartographic approach, in Force (like embedded-clause complementisers; Rizzi 1997), and, very frequently, in higher Speech Act heads, above Force (see, e.g., Corr 2016; 2022; Cruschina & Remberger 2018; Kocher 2022). These complementisers are abbreviated in the tables as “Illoc”.

3.2.2 Split CP-diagnostics

The diagnostics in the previous section would thus help establish children’s command of CP. Note, crucially, that they do not per se provide direct insights into the granularity of children’s representations, bar via theory-internal assumptions. They simply provide indications for CP knowledge and, as such, any results obtained could be compatible with cartographic and more minimalist frameworks. Therefore, a second set of novel diagnostics was used to probe specifically syntactic granularity. As Moscati & Rizzi (2021) note, the most direct evidence for knowledge of a fine-grained CP-domain would involve cases of co-occurrence of left-peripheral elements. This is the only diagnostic type adopted here for these structures and I take them to provide evidence for a CP-system that is additionally populated by a set of separate projections. I will refer to these diagnostics as “Split CP-diagnostics”.

In Romance, these will most frequently involve instances of topics preceding wh-phrases (22) or cases of recursive topics (24)–(25). A possible structural representation of (22) is given in (23).

    1. (22)
    1. Catalan
    1. La
    2. the
    1. Júlia,
    2. Júlia
    1. on
    2. where
    1. ha
    2. aux.have.3sg
    1. anat?
    2. go.ptcp
    1. ‘Júlia, where has she gone?’
    1. (23)
    1. [TopP La Júlia [Q/FocP on ha [TP…[VP anat]]]]
    1. (24)
    1. Italian
    1. Questo,
    2. this
    1. a
    2. to
    1. te,
    2. you
    1. ti
    2. cl.io=
    1. spaventa
    2. scare.3sg
    1. ‘This, it scares YOU.’
    1. (25)
    1. Spanish
    1. Aquí,
    2. here
    1. esto
    2. this
    1. no
    2. not
    1. lo
    2. cl.do=
    1. pones
    2. put.2sg
    1. ‘Here, this, you’re not putting it.’

Other instances of cartographic-type structures will include cases of complementisers or subordination markers being followed by a topic or a wh-element, such as those in (26a), and, in Ibero-Romance, instances of quotative complementisers preceding wh-elements (26b). In Catalan, furthermore, interrogative complementisers can be preceded by topics (27):

    1. (26)
    1. Spanish
    1.  
    1. a.
    1. Me
    2. cl.io=
    1. dijo
    2. tell.prfv.3sg
    1. que
    2. that
    1. el
    2. the
    1. puzle
    2. puzzle
    1. no
    2. not
    1. lo
    2. cl.do=
    1. quería
    2. want.impf.3sg
    1. comprar
    2. buy.inf
    1. ‘He/she told me that he/she didn’t want to buy the puzzle.’
    1.  
    1. b.
    1. ¿Que
    2. that.quot
    1. cuánto
    2. how.much
    1. te
    2. cl.io=
    1. han
    2. aux.have.3pl
    1. costado
    2. cost.ptcp
    1. estas
    2. these
    1. bambas?
    2. trainers
    1. ‘How much have you said these trainers have cost you!?’
    1. (27)
    1. Catalan
    1. I
    2. and
    1. en
    2. the
    1. Joan,
    2. John
    1. que
    2. that.int
    1. no
    2. not
    1. vindrà
    2. come.fut.3sg
    1. a
    2. to
    1. sopar?
    2. dine.inf
    1. ‘Is John not coming to dinner?’

In Catalan and Spanish, Topic + sí que ‘yes that’ and que sí que ‘that yes that’ constructions, the former shared with Italian sì che, are also analysed (see (28)). The examples below are accompanied by the (adult-like) cartographic structures assumed in Villa-García & González Rodríguez (2020) (the label of XP is left open to debate, see p. 26). Although I remain agnostic as to exact implementation of their cartographic analysis, the key assumption that I adopt from this work and others below is that these constructions require more elaborate structure than a basic CP (in the CFC sense).

    1. (28)
    1. Spanish
    1.  
    1. a.
    1. Pablo
    2. Paul
    1. sí
    2. yes
    1. que
    2. that
    1. vino
    2. come.prfv.3sg
    1. ‘Paul did come.’
    1.  
    1. b.
    1. Creo
    2. think.1sg
    1. que
    2. that
    1. sí
    2. yes
    1. que
    2. that
    1. ganaron
    2. win.prfv.3pl
    1. la
    2. the
    1. carrera
    2. race
    1. ‘I think that they did win the race’.
    1. (29)
    1. [ForceP (que) [TopicP [XP sí [X que [TopicP [FocusP [ΣP [TP … ]]]]]]]]

As far as Germanic languages are concerned, however, the V2-system makes a Split CP-domain substantially harder to detect and, indeed, not all approaches agree that the left periphery of Germanic V2-languages is “cartographic” in the traditional sense (cf. Giorgi & Pianesi 1997; Biberauer & Roberts 2015; Cormany 2015; Walkden 2017; Hsu 2017; 2021: for various views that assume a weakened left periphery in V2-languages, e.g., via conflation or bundled functional heads). However, limited instances of V3-orders are present in Germanic languages (these are more widespread in all colloquial and some vernacular varieties of German, although less so in Dutch; see Freywald et al. 2015, though cf. Meelen et al. 2020, on Dutch urban youth varieties). These surface V3-orders, apparent violations of the V2-rule, are often modelled via an additional projection above the landing site for V2-subjects (e.g., Haegeman & Greco’s 2016; 2018, FrameP). This additional projection accommodates these restricted V3-patterns and helps capture why a restricted set of elements would seemingly be invisible to the general Germanic V2-rule.

Structures generating surface V3 include frame-setters, Hanging Topic Left-Dislocation (HTLD) and Contrastive Left-Dislocation (CLD). Frame-setters are information-structural categories that provide specifications that restrict the domain in which the proposition is valid (such as temporal or local adverbs and phrases). HTLD permits DPs in first position to correspond with a resumptive personal pronoun in the V2-clause, often (though not invariably) in the middle field. CLD instead involves an utterance-initial XP with contrastive topicalisation or focus alongside a resumptive pronoun (typically a d-pronoun), which is preferably located in the left periphery (again, not obligatorily; see Frey 2004, for a review of the syntactic differences between HTLD and CLD). Some examples of these are offered in (30)–(32).

    1. (30)
    1. Dutch, Frame-setter (Haegeman & Greco 2020: 65)
    1. In
    2. in
    1. alle
    2. any
    1. geval,
    2. case
    1. ik
    2. I
    1. had
    2. aux.have.pst.1sg
    1. het
    2. it
    1. niet
    2. not
    1. verwacht
    2. expect.ptcp
    1. ‘Anyway, I had not expected it.’
    1. (31)
    1. German, HTLD
    1. Dem
    2. the.dat
    1. Alex,
    2. Alex
    1. ich
    2. I
    1. habe
    2. aux.have.1sg
    1. ihm
    2. him.dat
    1. geholfen
    2. help.ptcp
    1. ‘Alex, I helped him.’
    1. (32)
    1. German, CLD
    1. Diesen
    2. the.acc
    1. Kuchen
    2. cake.acc
    1. hier,
    2. here
    1. den
    2. pron.acc
    1. möchte
    2. want.1sg
    1. ich
    2. I
    1. probieren
    2. try.inf
    1. ‘This cake here, I want to try.’

Lastly, another kind of V3-orders in Germanic include conditional and temporal clauses followed by a resumptive adverb (Dutch als…dan and German wenn…dann, ‘if/when…then’). These are illustrated below and will also be analysed (following Haegeman & Greco 2020 and Meklenborg 2020, who also analyse them cartographically; see also De Clercq et al. 2023, for recent discussion):

    1. (33)
    1. German
    1. Wenn
    2. when/if
    1. du
    2. you
    1. ausrutschst,
    2. slip.2sg
    1. dann
    2. then
    1. kannst
    2. can.2sg
    1. du
    2. you
    1. dich
    2. cl.refl=
    1. verletzen
    2. hurt.inf
    1. ‘When/if you slip, then you can hurt yourself.’
    1. (34)
    1. Dutch
    1. Als
    2. when/if
    1. het
    2. it
    1. niet
    2. not
    1. zo
    2. so
    1. warm
    2. hot
    1. is,
    2. be.3sg
    1. dan
    2. then
    1. ga
    2. go.1sg
    1. ik
    2. I
    1. naar
    2. to
    1. buiten
    2. outside
    1. ‘When/if it isn’t so hot, then I’ll go out.’

The rationale for these Germanic diagnostics thus remains the same as for Romance: the structures involve two left-peripheral constituents and have been analysed as requiring an elaborate CP of some kind.

A final note regarding the status of V3-orders is in order. Given that V3-orders in Germanic, and Germanic V2 more generally, can be modelled in a non-cartographic manner (see Walkden 2017, and also references above), it is thus justified to wonder to what extent these V3-structures can properly be called “cartographic”. This raises issues about what these kinds of V3-orders are diagnosing in our context and its formal nature. As noted earlier, however, I simply take these Split CP-diagnostics (including V3-orders) to be diagnosing further internal elaboration within the CP-domain. The exact richness or granularity of this structural elaboration (e.g., if cartographic in the sense of Rizzi 1997; Cinque 1999; and subsequent work or “weaker”, as in Giorgi & Pianesi 1997; Cormany 2015; Walkden 2017, i.a.) is expected to be language-variant, and so secondary for our purposes so long as this granularity is greater than that of a CP with a single head. For this reason, I will continue to name these diagnostics “Split CP” or “cartographic-type” diagnostics, not “cartographic” diagnostics.

It is important to recognise that, firstly, the set of Split CP-structures introduced includes both structures that have been argued to be moved (e.g., wh-questions) vs. others that have received base-generation or adjunction analyses (e.g., Hanging Topics, V3 in Germanic; Grohmann 2000, den Dikken & Surányi 2017). Secondly, not all of them have been argued to support an articulated CP (e.g., Villa-García & Ott 2024). Other structures (e.g., recomplementation) have received both left-periphery-based and paratactic treatments (cf. Villa-García 2012a; Villa-García & Ott 2024). This is simply to acknowledge the breadth-first approach in my methodology: not all left fields with co-occurring constituents need imply an articulated CP under some analyses. Additionally, moved vs. base-generated left-peripheral constructions could follow different developmental pathways (Moscati 2023). The lack of distinctions in this paper was due to a methodological consideration: maximising the data sample with two co-occurring CP-elements was prioritised here over differentiation across subtypes of Split CP-structures (e.g., by including only moved constituents in the latter case). In this paper, I explore the development of co-occurring left-peripheral elements in broad terms, and make the (preliminary) assumption that these may require articulated CPs (following, e.g., Haegeman & Greco 2016; Walkden 2017). By acknowledging their heterogeneous nature, I am therefore open to the possibility that these may indeed develop at different stages. As will be shown in section 3.4, however, it is striking that all structures identified as Split-CP for the purposes of this study displayed the same developmental trends. Making subdivisions within these structures – and tracing their respective development – would nonetheless be a welcome future extension of this work, if sample sizes permit.

3.3 Summary of diagnostics, criteria for emergence and utterances excluded

Table 3 summarises all the left-peripheral structures searched for in every recording.

Table 3: Summary of structural diagnostics used.

Functional domains Structures analysed
CP-domain    • V-to-C movement (Germanic only)
   • Wh-questions
   • Yes/no questions (Germanic only)
   • Topicalisations and focalisations
   • Illocutionary complementisers (Romance only)
   • Finite embedding markers
Split CP    • Top > Wh (Romance only)
   • Top > Top/Foc (e.g., recursive topics; Romance only)
   • Comp > Wh/Top (Romance only)
   • quot que ‘that’ > Wh (Ibero-Romance only)
   • Top > int que ‘that’ (Catalan only)
   • Topic + sí que/sì che ‘yes that’ and que sí que ‘that yes that’ structures (Romance and Ibero-Romance only, respectively)
   • Frame-setters, HTLD, CLD (Germanic only)
   • Conditional/temporal clauses with resumptive dann/dan ‘then’ (Germanic only)
   • Other possible combinations of the above

For each diagnostic, I manually checked whether the structure appeared at least once in each file. A given functional domain (e.g., CP) is assumed to have emerged after the first attestation of a structure corresponding to that domain (e.g., CP or TP) and followed soon after (within ∼2 files) by other uses of the same or other structures searched for in that domain. That is, isolated instances of a structure were not deemed sufficient evidence for emergence, and no conclusive statements were made in such cases – a “repeated uses” criterion for emergence. This criterion is necessarily somewhat arbitrary. There is no justified motivation for why ∼2 files should be the relevant window of consideration. However, to avoid reaching theoretical conclusions on the basis of isolated examples, it is methodologically important to adopt a conservative measure of emergence above (as done, for example, in Villa-García 2012b), modulo the fact that it is not fully satisfactory. In practice, however, “repeated uses” as per above and the more liberal metric of “first file of emergence” (Stromswold 1990) overlapped to a significant degree for this dataset, insofar as most first uses of a structure were almost always followed by further uses in subsequent files; this becomes clear in the tables in the Appendices, and is also discussed in passing at various points in section 3.4.

Additionally, depending on the recordings’ length for a given child, this criterion had to be weakened. This was only the case for the early-talker Spanish child Irene. Her recordings at the earliest stages are extremely brief, often including <50 words by the child, meaning they contain very few to no verbs and, therefore, often none of the structures analysed. It was also relaxed for the Split CP-diagnostics, which, by their very nature, are less frequent than the CP-structures out of which they are formed. This is crucial to ensure structures with rarer contexts of use are not underreported in the study. For a full breakdown of the counts obtained per file and structure, I refer the reader to Appendix A.

A quantitative analysis of all structures was also carried out, including an analysis of the variety in types for each structure (e.g., the types of wh-questions or wh-words used, verbs with which V2-order was found, among others). This was necessary to determine the productivity of the earliest CP-structures and to discard possible instances of rote-learned formulae. Full details can be found in Appendix B.

Identical repetitions of an adult utterance or the child’s previous utterance were excluded, as were lyrics and unintelligible utterances (either due to pronunciation or transcription).

3.4 Results

The analysis of the longitudinal data revealed a transparent order of emergence of the two sets of diagnostics. I begin by presenting an outline of the findings comparing the structures analysed as groups (CP vs. Split CP-structures). Then, I consider the relative order of emergence of the various CP-structures analysed. Overall, I introduce three theoretically consequential generalisations. These shed new light on the two theoretical topics introduced in sections 1 and 2 – directionality and granularity in development. Some theoretical commentary is offered, but the implications are discussed in section 4.

3.4.1 Order of emergence of structures and the role of MLUw

Based on the diagnostics used, the findings reflect two transparent stages, across the ten children. These are listed below, and exclude a period, in some children, where none of the structures are attested or, if they are, they represent isolated cases with unclear status.

  1. Stage 1 – CP-structures only: these typically included wh-questions, V2 (for Germanic), illocutionary complementisers (for Romance) and, more rarely, topicalisations/focalisations (typically MLUw ≈ 1.5). These structures become consolidated soon after their emergence.

  2. Stage 2 – Split CP-structures emerge: Split CP-structures emerge well after the first CP-structures (in these children, generally at MLUw ≥ 2.5).

Stage 1 is generally long in duration (see Table 12 later) and also includes the emergence of other structures, TP-elements (e.g., auxiliaries, set aside here; Appendix A and Bosch 2023a, for the full data). The emergence of Split CP-structures, however, comes systematically late, often after other complex syntactic configurations like embedding. To ease readability, I present the results of the corpus study via summarised tables, which indicate the point of emergence of these stages and the structures attested (per the criteria outlined earlier). Table 4 summarises the data obtained for the Catalan, Italian and Spanish children and Table 5 outlines the results for German and Dutch. CP-diagnostics are presented first, in blue, and then Split CP-diagnostics, in green. Cells which are not colour-shaded, but which feature a structure, indicate isolated, potentially unproductive utterances, per the emergence criterion above. Each table condenses the order of emergence of each of the CP-structures studied. The age/MLU ranges for each Stage are determined based on the first file in every child that satisfies the criteria for emergence for each domain, as defined above. In Germanic, where no wh-words are attested in a given period, but wh-less questions are present, this is noted as “wh-less”. The full tables are available in the Appendices.

Table 4: Summary of production data by the Romance children.

Language Child Age MLUw Wh-Q Top/Foc Illoc Embed Split CP Stage
Catalan Laura 1;07.20 1.03–1.15
1;10.22 1.15–1.3 ✓ ①
2;02.13 1.3–1.88 ✓ ✓
2;08.30 1.88–2.42 ✓ ✓ ✓
3;00.02 2.42–2.54 ✓ ✓ ✓ ✓
3;03.21 2.5410–3.18 ✓ ✓ ✓ ✓ ✓ ②
Gisela 1;07.14 1.04–1.12
1;08.24 1.13–1.58 ✓
2;04.25 1.58–2.32 ✓ ①
2;06.23 2.32–2.61 ✓ ✓
2;08.00 2.61–3.41 ✓ ✓ ✓ ✓ ✓ ②
Italian Martina 1;07.18 1.26–1.57
1;08.02 1.5711–1.57 ✓ ✓ ①
1;08.17 1.57–1.99 ✓ ✓ ✓
1;11.20 1.99–2.69 ✓ ✓ ✓ ✓
2;04.13 2.69–2.55 ✓ ✓ ✓ ✓ ✓ ②
Rosa 1;07.13 1.27–1.75 ✓ ①
2;04.23 1.75–2.6 ✓ ✓
2;06.29 2.6–2.87 ✓ ✓ ✓
2;09.04 2.87–2.512 ✓ ✓ ✓ ✓
2;10.14 2.5–3.24 ✓ ✓ ✓ ✓ ✓ ②
Spanish Irene 0;11.01 1.42–1.32
1;04.16 1.32–1.88 ✓ ①
1;08.09 1.88–2.95 ✓ ✓ ✓
1;09.10 2.28–2.95 ✓ ✓ ✓ ✓
1;11.13 2.95–3.38 ✓ ✓ ✓ ✓ ✓ ②
Koki 1;07.20 1.96–2.69 ✓ ①
1;11.25 2.47–2.47 ✓ ✓ ✓
2;02.27 2.47–2.69 ✓ ✓ ✓ ✓
2;04.18 2.69–3.38 ✓ ✓ ✓ ✓ ✓ ②

Table 5: Summary of production data by the Germanic children.

Language Child Age MLUw V2 Wh-Q Y/N-Q Top/Foc Embed Split CP Stage
German Kerstin 1;03.22 1.09–1.36
1;05.24 1.36–1.28 ✓
1;10.03 1.28–1.38 ✓ ①
1;10.05 1.38–1.76 ✓ ✓
2;00.05 1.76–1.68 ✓ ✓ ✓
2;00.10 1.68–2.13 ✓ ✓ ✓ ✓
2;07.23 2.13–2.32 ✓ ✓ ✓ ✓ ✓
2;09.11 2.32–2.89 ✓ ✓ ✓ ✓ ✓ ✓ ②
Simone 1;09.11 1.54–1.62 ✓ ①
1;10.20 1.62–2.31 ✓ ✓ ✓
2;00.23 2.31–1.96 ✓ ✓ ✓ ✓
2;04.20 1.96–2.78 ✓ ✓ ✓ ✓ ✓
2;06.23 2.78–3.46 ✓ ✓ ✓ ✓ ✓ ✓ ②
Dutch Josse 2;00.07 1.2–1.94 ✓ ✓ ①
2;03.28 1.94–2.42 ✓ ✓ ✓ ✓
2;09.02 2.42–3.57 ✓ ✓ ✓ ✓ ✓
2;11.09 3.57–3.2 ✓ ✓ ✓ ✓ ✓ ✓ ②
Sarah 1;06.16 1.12–1.17
1;07.21 1.17–1.09 Wh-less
1;10.05 1.09–1.68 ✓ Wh-less ①
2;00.17 1.68–2.11 ✓ Wh-less ✓ ✓
2;02.18 2.11–3.52 ✓ ✓ ✓ ✓
3;00.19 3.52–4.92 ✓ ✓ ✓ ✓ ✓ ✓ ②

Overall, MLUw proves to be a relatively stable metric, with the two stages emerging on average at the following MLUws. The average length (in months) of the first stage is also noted:

  • Stage 1: MLUw 1.4 (range 1.09 – 1.96; average length of 10.5 months).

  • Stage 2: MLUw 2.82 (range 2.32 – 3.57).

The analysis of the emergence of the various structures studied indicates that there exists a relationship between the MLUw of the child (understood as a metric of syntactic development) and the types of structures produced, with a clear relative order among those structures. The MLUw variances of all three stages also appear fairly homogenous (with the variances for Stages 1 and 2 being 0.0701 and 0.1755, respectively). Finally, a paired-samples t-test comparing the MLUw at the point of emergence of the two stages output statistically significant differences (t(19)=6.7188, p < .001). This suggests Split CP-structures are produced at later stages, with significantly higher MLUw.

Crucially, age is not a reliable predictive factor. Table 6 displays the age of emergence of the three sets of structures analysed. Age of emergence presents high variance. The variances for each variable (Stage 1 and 2) when measured in terms of age (in months) vary substantially (being 10.4556 and 17.25, respectively).

Table 6: Age of emergence across the two stages.

Stage 1 Stage 2
Laura 1;10.22 3;03.21
Gisela 2;04.25 2;08.00
Martina 1;08.02 2;04.13
Rosa 1;07.13 2;10.14
Irene 1;04.16 1;11.13
Koki 1;07.20 2;04.18
Kerstin 1;10.03 2;09.11
Simone 1;09.11 2;06.23
Josse 2;00.07 2;11.09
Sarah 1;10.05 3;00.19

The age at which children produce specific structures can thus vary significantly from child to child. There is vast variation: at 2;00, Josse is still at Stage 1, while Irene reaches Stage 3 at 1;11. Similarly, Rosa and Laura reach Stage 2 by the same age (2;04), but Laura reaches Stage 3 around half a year later (2;10 vs. 3;03). The most obvious outlier is Irene, who reaches Stage 3 by 1;11, well before several children reach Stage 1. This echoes previous literature showing high variability in age of acquisition (Paradis & Genesee 1997; Caprin & Guasti 2009; Friedmann et al. 2021). These results further endorse the need to make crosslinguistic comparisons with another less variable metric, such as MLUw, or centre the attention on stages observed, rather than ages (Friedmann & Reznick 2021: 1). Guttman Scales have proven a useful heuristic tool in this latter respect; I incorporate them in the following section.

3.4.2 The development of CP-structures (Stage 1): an early phenomenon

I now go on to unpack the broad observations above, outlining first the specific patterns of acquisition observed at each Stage and, then, introducing three theoretically consequential generalisations.

First, Stage 1 suggests that CP might be acquired early. I provide evidence from various diagnostics that shows this conclusion is likely corroborated.

As for Romance, the earliest recordings contain wh-questions, illocutionary complementisers and some topics/foci. Some examples of these early CP-structures are given below:

    1. (35)
    1. Catalan
    1.  
    1. a.
    1. Què
    2. what
    1. és?                                                                                 (Laura, MLUw 1.3)
    2. be.3sg
    1. ‘What is it?’
    1.  
    1. b.
    1. Ai,
    2. ouch
    1. que
    2. that.excl
    1. crema!                                                        (Laura, MLUw 1.35)
    2. burn.3sg
    1. ‘Ouch, it’s burning!’
    1.  
    1. c.
    1. Que
    2. that.excl
    1. cau!                                                                          (Laura, MLUw 1.3)
    2. fall.3sg
    1. ‘It’s falling!’
    1. (36)
    1. Italian
    1.  
    1. a.
    1. Dov’è?                                                                                       (Rosa, MLUw 1.39)
    2. where-be.3sg
    1. ‘Where is it?’
    1.  
    1. b.
    1. Pello
    2. that
    1. oio
    2. want.1sg
    1. io
    2. I
    1. [:
    2.  
    1. quello
    2.  
    1. voglio
    2.  
    1. io]                      (Rosa, MLUw 1.75)
    2.  
    1. ‘That, I want.’
    1.  
    1. c.
    1. Ove
    2. where
    1. va
    2. go.3sg
    1. chetto?                                                      (Martina, MLUw 1.57)
    2. this
    1. ‘Where does this one go?’
    1.  
    1. d.
    1. Che
    2. that.excl
    1. gira                                                                       (Martina, MLUw 1.9)
    2. stir.3sg
    1. ‘He/she stirs (it).’
    1. (37)
    1. Spanish
    1.  
    1. a.
    1. ¿Quién
    2. who
    1. es?                                                                             (Irene, MLUw 1.32)
    2. be.3sg
    1. ‘Who is it?’
    1.  
    1. b.
    1. ¿Qué
    2. what
    1. hay
    2. have.3sg
    1. ahí
    2. there
    1. mamá?                                            (Irene, MLUw 1.69)
    2. mum
    1. ‘What is there mum?’

German and Dutch children likewise present evidence of CP-structures at this stage, and to a greater extent than Romance children. Firstly, there is evidence of early wh-questions and some topicalisations (of both adverbs and objects), as shown in (39).

    1. (38)
    1. Dutch
    1.  
    1. a.
    1. Wo
    2. where
    1. ’s
    2. be.3sg
    1. den
    2. the.acc
    1. Nina?                                          (Kerstin, MLUw 1.28)
    2. Nina
    1. ‘Where is Nina?’
    1.  
    1. b.
    1. Wat
    2. what
    1. doet
    2. do.3sg
    1. ie
    2. he
    1. nou?                                                       (Josse, MLUw 1.59)
    2. now
    1. ‘What is he doing now?’
    1. (39)
    1. German
    1.  
    1. a.
    1. Jetzt
    2. now
    1. kommt
    2. come.3sg
    1. ein
    2. a
    1. Affe                                                (Kerstin, MLUw 1.76)
    2. monkey
    1. ‘Now a monkey comes.’
    1.  
    1. b.
    1. Blume
    2. flower
    1. eßt
    2. eat.3sg
    1. nicht                                                        (Simone, MLUw 1.69)
    2. not
    1. ‘Flower, he/she doesn’t eat.’
    1.  
    1. c.
    1. Bau(e)n
    2. build.inf
    1. will
    2. want.1sg
    1. ich                                                     (Simone, MLUw 1.71)
    2. I
    1. ‘To build (something), I want.’

There are, also, noticeable signs of a distinction between non-finite verbs (which remain in-situ and yield OV-order where an object is present) and finite verbs (which can be shown to have raised outside the vP).13

    1. (40)
    1. German
    1.  
    1. a.
    1. Ich
    2. I
    1. will
    2. want
    1. de(r)
    2. the
    1. Ball                                                        (Kerstin, MLUw 1.58)
    2. ball
    1. ‘I want the ball.’
    1.  
    1. b.
    1. Wieder
    2. again
    1. aufräumen                                                               (Kerstin, MLUw 1.34)
    2. up.tidy.inf
    1. ‘To tidy up again.’
    1.  
    1. c.
    1. Ma(ch)
    2. make.imp.2sg
    1. ma(l)
    2. once
    1. Schuh
    2. shoe
    1. ab                                        (Kerstin, MLUw 1.53)
    2. off
    1. ‘Take off (the) shoe.’
    1.  
    1. d.
    1. Mag
    2. like.1sg
    1. nicht
    2. not
    1. Kuche(n)
    2. cakes
    1. backe(n)                                  (Simone, MLUw 1.62)
    2. bake.inf
    1. ‘I don’t like to bake cakes.’
    1.  
    1. e.
    1. Bluse
    2. blouse
    1. auszieh(e)n                                                              (Simone, MLUw 1.62)
    2. off.take.inf
    1. ‘To take off (the) blouse.’
    1. (41)
    1. Dutch
    1.  
    1. a.
    1. Komt
    2. come.3sg
    1. ie                                                                             (Josse, MLUw 1.2)
    2. he
    1. ‘He comes.’
    1.  
    1. b.
    1. Koffie
    2. coffee
    1. drinken                                                                        (Josse, MLUw 1.44)
    2. drink.inf
    1. ‘To drink coffee.’
    1.  
    1. c.
    1. Doet
    2. do.3sg
    1. niet
    2. not
    1. meer                                                                 (Josse, MLUw 1.55)
    2. more
    1. ‘He/she doesn’t do (it) anymore.’
    1.  
    1. d.
    1. I(k)moe(t)
    2. I-must.1sg
    1. ete(n)                                                                   (Sarah, MLUw 1.16)
    2. eat.inf
    1. ‘I have to eat.’
    1.  
    1. e.
    1. I(k)wi(l)
    2. I-want.1sg
    1. farkj [: varkentje]                                                 (Sarah, MLUw 1.25)
    2. pig.dim
    1. ‘I want piglet.’
    1.  
    1. f.
    1. Boekje
    2. book.dim
    1. lezen                                                                      (Sarah, MLUw 1.68)
    2. read.inf
    1. ‘To read book/booklet.’

Finally, at the end of this stage finite embedding markers also emerge for all children. Some instances are given in (42)–(46):

    1. (42)
    1. Catalan
    1. Saps
    2. know.2sg
    1. que
    2. that
    1. no
    2. not
    1. vindrà
    2. come.fut.3sg
    1. ningú?                                 (Gisela, MLUw 2.61)
    2. no.one
    1. ‘Do you know no one is going to come?’
    1. (43)
    1. Italian
    1. È
    2. be.3sg
    1. l’omo
    2. the-man
    1. che
    2. that
    1. ccrive                                                      (Martina, MLUw 2.55)
    2. write.3sg
    1. ‘It/he is the man that writes.’
    1. (44)
    1. Spanish
    1. Deja
    2. leave.imp.2sg
    1. esto
    2. this
    1. ahí
    2. there
    1. para
    2. so
    1. que
    2. that
    1. no
    2. not
    1. se
    2. cl.refl=
    1. rompe    (Koki, MLUw 3.08)
    2. break.3sg
    1. ‘Leave this there so that it doesn’t break.’
    1. (45)
    1. German
    1. Weil
    2. because
    1. der
    2. the
    1. Tommy
    2. Tommy
    1. gebißen
    2. bite.ptcp
    1. hat                                         (Simone, MLUw 2.52)
    2. aux.have.3sg
    1. ‘Because Tommy bit (something/someone).’
    1. (46)
    1. Dutch
    1. Ik
    2. I
    1. krijg
    2. get.1sg
    1. deze
    2. this
    1. a(l)s
    2. when
    1. ik
    2. I
    1. uit
    2. out
    1. de
    2. the
    1. crèche
    2. nursery
    1. ben               (Sarah, MLUw 3.52)
    2. be.1sg
    1. ‘I get this when I am out of the nursery.’

I substantiate this with Tables 7 and 8. Due to space considerations, only two children are given, but the patterns recur across the ten children (Appendix A).

Table 7: Production of structures by Laura.

Age MLU Wh-Q Top/Foc Illoc Embed Split CP
1;07.20 1.03
1;09.07 1.09
1;10.22 1.15 ✓
1;11.12 1.15 ✓
2;02.05 1.35 ✓
2;02.13 1.3 ✓
2;04.11 1.44 ✓
2;05.08 1.64
2;06.25 1.76 ✓
2;07.20 1.78 ✓ ✓
2;08.30 1.88 ✓ ✓ ✓
2;11.17 1.98 ✓ ✓ ✓
3;00.02 2.42 ✓ ✓ ✓ ✓
3;03.21 3.47 ✓ ✓ ✓ ✓ ✓
3;05.13 2.54 ✓ ✓ ✓ ✓ ✓
3;10.00 2.97 ✓ ✓ ✓ ✓ ✓
3;10.01 2.91 ✓ ✓ ✓ ✓ ✓
3;11.12 3.0 ✓ ✓ ✓ ✓ ✓
4;00.10 3.18 ✓ ✓ ✓ ✓ ✓

Table 8: Production of structures by Sarah.

Age MLU V2 Wh-Q Y/N-Q Top/Foc Embed Split CP
1;06.16 1.12
1;07.21 1.17 Wh-less
1;08.28 1.07
1;09.10 1.17 Wh-less
1;10.05 1.09 ✓
1;10.13 1.17 ✓
1;11.01 1.25 ✓
1;11.15 1.37 ✓ Wh-less
2;00.17 1.68 ✓ ✓ ✓
2;01.10 1.88 ✓ ✓ ✓
2;02.18 2.11 ✓ ✓ ✓
2;03.16 2.05 ✓ ✓ ✓
2;04.02 2.53 ✓ ✓ ✓ ✓
2;04.09 2.34 ✓ ✓ ✓
2;04.27 2.46 ✓ ✓ ✓ ✓
2;05.09 2.47 ✓ ✓ ✓ ✓
2;05.22 2.59 ✓ ✓ ✓ ✓
2;06.04 2.74 ✓ ✓ ✓ ✓
2;06.11 2.45 ✓ ✓ ✓
2;06.18 2.8 ✓ Wh-less ✓ ✓
2;07.16 2.51 ✓ ✓ ✓ ✓
2;08.06 2.66 ✓ ✓ ✓ ✓
2;08.19 2.97 ✓ ✓ ✓ ✓
2;09.02 2.59 ✓ ✓ ✓ ✓
2;09.07 3.15 ✓ ✓ ✓ ✓
2;10.18 2.88 ✓ ✓ ✓ ✓
2;11.03 2.87 ✓ ✓ ✓ ✓
2;11.27 3.64 ✓ ✓ ✓ ✓
3;00.19 3.52 ✓ ✓ ✓ ✓ ✓ ✓
3;01.17 3.06 ✓ ✓ ✓ ✓ ✓ ✓
3;02.13 3.82 ✓ ✓ ✓ ✓ ✓
3;03.21 3.05 ✓ ✓ ✓ ✓ ✓ ✓
3;04.13 3.15 ✓ ✓ ✓ ✓ ✓ ✓
3;05.30 2.89 ✓ ✓ ✓ ✓ ✓ ✓
3;05.30 2.89 ✓ ✓ ✓ ✓ ✓ ✓
3;07.25 3.24 ✓ ✓ ✓ ✓ ✓
3;10.07 3.71 ✓ ✓ ✓ ✓ ✓ ✓
3;11.04 4.07 ✓ ✓ ✓ ✓ ✓ ✓
4;00.11 3.81 ✓ ✓ ✓ ✓ ✓ ✓
4;00.30 4.08 ✓ ✓ ✓ ✓ ✓
4;01.11 4.66 ✓ ✓ ✓ ✓ ✓ ✓
4;03.04 5.37 ✓ ✓ ✓ ✓ ✓ ✓
4;04.28 4.28 ✓ ✓ ✓ ✓ ✓
4;05.29 4.7 ✓ ✓ ✓ ✓ ✓ ✓
4;06.12 5.06 ✓ ✓ ✓ ✓ ✓ ✓
4;07.25 4.62 ✓ ✓ ✓ ✓ ✓ ✓
4;08.03 5.03 ✓ ✓ ✓ ✓ ✓ ✓
4;09.13 6.07 ✓ ✓ ✓ ✓ ✓ ✓
4;09.29 5.2 ✓ ✓ ✓ ✓ ✓ ✓
4;11.15 4.01 ✓ ✓ ✓ ✓ ✓ ✓
5;02.13 4.92 ✓ ✓ ✓ ✓ ✓ ✓

A summary of the types and tokens of all CP-structures attested throughout Stage 1 is given in Table 9. The first value in parentheses next to the number of wh-questions indicates the proportion of copular wh-questions and the second one in German and Dutch signals the proportion of wh-less questions. This illustrates the variety of predicates observed with wh-questions (the variety in wh-words is available in Appendix B). For simplicity, only attestation of V2 is indicated below (with a check mark); for a full quantitative break-down of V2 and V-final orders, see Appendix B.

Table 9: CP-structures produced at Stage 1 and its length.

V2 Wh-Q Y/N-Q Top/Foc Illoc Embed Length
Laura 15 (9) 4 42 4 1;10.22–3;03.21 (MLUw 1.15–2.54)
Gisela 1 (0) 0 6 0 2;04.25–2;08.00 (MLUw 1.58–2.61)
Martina 21 (16)14 3 7 8 1;08.02–2;04.13 (MLUw 1.57–2.69)
Rosa 133 (131) 12 3 8 1;07.13–2;10.14 (MLUw 1.27–2.5)
Irene 18 (6) 3 10 4 1;04.16–1;11.13 (MLUw 1.32–2.95)
Koki 32 (27) 7 2 4 1;07.20–2;04.18 (MLUw 1.96–2.69)
Kerstin ✓ 16 (14, 2)15 21 27 1 1;10.03–2;09.11 (MLUw 1.28–2.32)
Simone ✓ 166 (154, 7) 3 105 24 1;10.03–2;06.23 (MLUw 1.54–2.78)
Josse ✓ 62 (43, 25)16 37 68 1 2;00.07–2;11.09 (MLUw 1.2–3.57)
Sarah ✓ 124 (100, 46) 104 116 0 1;10.05–3;00.19 (MLUw 1.09–3.52)

Given this data, it is reasonable to suggest that productive knowledge of structures like wh-questions, V2, topics and illocutionary complementisers is in place very early on. CP-structures are abundant from the start of Stage 1, with multiple structures instantiated across recordings, as shown by Table 9. Importantly, the knowledge of CP and V2 systems established at Stage 1 involves at most a basic CP. There is virtually no production data supporting the availability of an internally elaborate CP. The specific articulation of CP, including the specific V2 system (e.g., ‘Force-V2’ or ‘Fin-V2’ in Poletto’s 2002, and Wolfe’s 2015 typology; see also Biberauer & Roberts 2015) is expected to be consolidated at Stage 2, when cartographic-type structures appear. It follows, too, that V3-configurations requiring finer-grained cartographic-type structures (e.g., those introduced in section 3.2.2) are expected not to emerge until Stage 3 – an expectation which is borne out and will be discussed in section 3.4.3.

This is, therefore, Generalisation 1: CP is an early phenomenon in this data.

Generalisation 1 (Early Acquisition of CP). CP-structures emerge early on in the developmental data.

This generalisation poses a challenge to bottom-up maturation,17 but is coherent with all inward-growing, continuity and neo-emergentist approaches (see section 4). This first result, then, aligns with the fundamental shared assumption in these frameworks – (some) CP knowledge develops early.

This conclusion is further underscored once the development of several left peripheral structures is examined more carefully. We noted above that embedding markers emerge late, across all children. This stands in contrast with the rest of the CP-diagnostics studied. Implicit in the above is, thus, another important result, echoed by at least two separate empirical points. Structurally very high elements emerge early on in several children; this is the case of topics and illocutionary complementisers.

First, the (near-)simultaneous emergence of embedding markers and topicalisation in Friedmann et al. (2021) (their Stage 3) is, for many children, not replicated. At most, it only holds for some of the Romance children. Topicalisation emerges considerably before finite embedding in several of the children studied (Table 10). The difference between topics and embedding is particularly stark in Germanic children, who use topics from the earliest files and MLUw ranges. The MLUw difference between topics and embedding is highly statistically significant, according to a paired-samples t-test (t(9)=3.3386, p = .009). Note that, in order to facilitate a comparison with Friedmann et al.’s (2021) conjecture (which bears on topicalisation and embedding, not the left-peripherally lower focalisation), this analysis excluded fronted objects in Romance that are possible instances of foci or that are ambiguous between focalisation/topicalisation.18

Table 10: Emergence of topicalisation vs. embedding markers.

Topicalisation Embedding
Laura 2;11.17 3;00.02
1.98 MLUw 2.42 MLUw
Gisela 2;08.00 2;08.00 (same file)
2.61 MLUw 2.61 MLUw
Martina 1;08.17 1;11.20
1.56 MLUw 1.99 MLUw
Rosa 2;04.29 2;06.29
1.77 MLUw 2.6 MLUw
Irene 1;08.09b 1;09.10
2.24 MLUw 3.28 MLUw
Koki 1;11.25 1;11.25 (same file)
2.47 MLUw 2.47 MLUw
Kerstin 2;00.05 2;07.23
1.76 MLUw 2.13 MLUw
Simone 1;10.20 2;04.20
1.62 MLUw 1.96 MLUw
Josse 2;03.28 2;09.02
1.94 MLUw 2.42 MLUw
Sarah 2;00.17 3;00.19
1.68 MLUw 3.52 MLUw
Average 1.96 MLUw 2.54 MLUw

Examples (47)–(49) illustrate some of the early utterances containing topics, in Romance and German, some repeated from (39).

    1. (47)
    1. Italian
    1.  
    1. a.
    1. Ove
    2. where
    1. va
    2. go.3sg
    1. chetto?                                                      (Martina, MLUw 1.57)
    2. this
    1. ‘Where does this one go?’
    1.  
    1. b.
    1. Bere
    2. drink.inf
    1. voglio                                                                  (Martina, MLUw 1.66)
    2. want.1sg
    1. ‘To drink, I want.’
    1.  
    1. c.
    1. Que’
    2. that
    1. lo
    2. cl.do=
    1. l’apri?
    2. cl.do=open.2sg                                              (Rosa, MLUw 1.78)
    1. (48)
    1. German
    1.  
    1. a.
    1. Jetzt
    2. now
    1. kommt
    2. come.3sg
    1. ein
    2. a
    1. Affe                                                (Kerstin, MLUw 1.76)
    2. monkey
    1. ‘Now a monkey comes.’
    1.  
    1. b.
    1. Bau(e)n
    2. build.inf
    1. will
    2. want.1sg
    1. ich                                                     (Simone, MLUw 1.71)
    2. I
    1. ‘To build (something), I want.’
    1. (49)
    1. Dutch
    1. Daar
    2. there
    1. gaat
    2. go.3sg
    1. ie                                                                               (Josse, MLUw 2.14)
    2. he
    1. ‘There he goes.’

Second, illocutionary complementisers also consistently emerge very early: Laura, Gisela, Irene and Koki display illocutionary complementisers at MLUw 1.15, 1.58, 1.88 and 1.96, respectively, and from the earliest files. Some examples of early illocutionary complementisers are given in (50)–(51). Importantly, this pattern is not exclusive to these children or languages, but was already supported with bigger Catalan/Spanish samples in Bosch (2023b) and for Romanian in Stoicescu (2025).

    1. (50)
    1. Catalan
    1.  
    1. a.
    1. Ai,
    2. ouch
    1. que
    2. that.excl
    1. crema!                                                        (Laura, MLUw 1.35)
    2. burn.3sg
    1. ‘Ouch, it’s burning!’
    1.  
    1. b.
    1. Que
    2. that.excl
    1. cau!                                                                          (Laura, MLUw 1.3)
    2. fall.3sg
    1. ‘It’s falling!’
    1.  
    1. c.
    1. Que
    2. that.quot
    1. no
    2. not
    1. la
    2. cl.do=
    1. puc
    2. can.1sg
    1. treure                           (Gisela, MLUw 2.32)
    2. take.out.inf
    1. ‘(I’ve said) I can’t take it out.’
    1. (51)
    1. Spanish
    1. Que
    2. that.quot
    1. rompió                                                                              (Irene, MLUw 1.88)
    2. break.3sg.pst
    1. ‘(What happened is that) it broke / he broke it.’

Recall that these main-clause complementisers are typically assumed to be located high in the CP-domain (often in Rizzi’s Force head) or often in even higher Speech Act heads (section 3.2.1). Illocutionary complementisers, then, bring a new, understudied piece of data suggesting structural height does not always correlate with late acquisition.

In sum, this brings us to the second generalisation we will consider. The long-held assumption that acquisition “recapitulates” a UG-given spine cannot be maintained, at least in its stronger, cartographic form. Acquisition and structural height can, and in fact do, mismatch in the present data (Generalisation 2) (from Bosch & Biberauer 2024: 109):

Generalisation 2 (Structural Height and Acquisition Mismatch). There is a dissociation between structural height and order of emergence. Acquisition does not proceed successively upwards; some syntactically very high elements emerge early.

This mismatch creates a Directionality Problem: if a cartographic-type spine is assumed, we cannot account for acquisition patterns with directionality of maturation as our only developmental driving factor. Regardless of the directionality assumed, we predict some structures to be late-acquired, only for them to emerge very early (we take this up again in section 4). Concretely, Generalisation 2 undermines in particular cartographic bottom-up approaches such as the Growing Trees Hypothesis (Friedmann et al. 2021) as universal pathways: insofar as both topics and illocutionary complementisers are hosted at the top of a cartographic left periphery (e.g., TopicP in the former case, and ForceP or higher Speech-Act heads in the latter), both elements should emerge at the end of the learning path (Friedmann et al.’s 2021 Stage 3). The opposite pattern is observed.

Now, the above does not necessarily imply that these developmental patterns will always replicate themselves in other languages or children (a separate empirical question). It remains possible that some structurally high elements, like topics, emerge late with embedding in other languages, as Friedmann et al. (2021) show for Hebrew. What it does imply, however, is that topics and illocutionary complementisers do not universally emerge alongside other (late-acquired) structurally high elements across all languages and children. Additionally, it is worth noting that the timing discrepancies observed across the two language families in Table 10 are also telling. The timing of topics in Romance is fairly variable across children, but is systematically early in Germanic. This is suggestive in light of literature pointing to Germanic topicalisation being formally much simpler than the options available in (some) Romance varieties (see Biberauer & Roberts 2015: 15–18, for an explicit comment), aligning with the earlier emergence of topics in the former.19 This tentative correlation between formal complexity of topicalisation and its age of acquisition was actually corroborated by Bosch & Biberauer (2025), who show for a sample of 10 languages that early topic acquisition systematically correlates with languages whose topicalisation displays either adjunction/base-generation (e.g., French) or ‘pure’ A’-movement (e.g., European Portuguese; following van Urk 2015; Cinque 1990, i.a.). Languages with topics displaying a mixed set of A/A’-properties (e.g., CLLD languages, but also clitic-less topicalisation in Brazilian Portuguese and Hebrew) are late acquired, without exception. Their results again underscore the L1-specificity of the acquisition of topicalisation.

Thus far we have established that there is variation in the emergence of structures located at the top of the left periphery; some emerge early (e.g., illocutionary complementisers, topics), others do not (embedding), with some crosslinguistic differences. I have argued that this variation is consequential, questioning an absolute ban on early production of structurally high elements. Setting variation aside, an additional question that emerges from the previous two sections, however, is if there are any shared pathways of acquisition of the various CP-structures observed. That is, even after establishing problems with directionality-only approaches, what universality is left in learning paths? I now incorporate Guttman Scales to address this point.

Guttman Scales have often been harnessed to reveal shared temporal orders in language acquisition (Barton 1976; Levelt et al. 1999/2000; Fikkert & Levelt 2008; Friedmann & Reznick 2021; Friedmann et al. 2021; van ’t Veer 2023). These scales are hypotheses about the structure of a dataset with respect to an attribute. This type of cumulative scaling can help establish an ordinal scale with an arrangement such that, if a structure X has emerged (e.g., by a given time), any other structures that fall below it in the scale will have also been acquired by that point. Table 11 summarises the relative order of emergence of each of the CP-structures analysed (organised from earliest to latest, left to right).

Table 11: Relative order of emergence of diagnostics studied.

Child Order of emergence
Laura Illoc > Wh > Topic > Embed > Split CP
Gisela Illoc > Wh > Topic/Embed/Split CP
Martina Illoc/Wh > Topic > Embed > Split CP
Rosa Wh > Topic > Embed > Illoc > Split CP
Irene Wh > Illoc/Topic > Embed > Split CP
Koki Topic > Wh/Embed > Illoc > Split CP
Kerstin Wh > V2 > Topic/YN > Embed > Split CP
Simone V2 > Wh/Topic > YN > Embed > Split CP
Josse Wh/V2 > Topic/YN > Embed > Split CP
Sarah V2 > Y/N/Topic > Wh > Embed/Split CP

Crucially, the orderings in Table 11 suggest that extremely few individual CP-structures can be ordered into a universal scale. The only, very broad, emergence scales that emerge for the two language families are the ones below. Note that illocutionary complementisers are the only structures that cannot be ordered relative to other structures; they quite often emerge from the start (Laura, Gisela, Martina), but also a bit later (Koki, Rosa).

  • Romance: Wh/Topic > Embed > Split CP

  • Germanic: Wh/V2/Topic/YN > Embed > Split CP

Here, only these much broader scales can be obtained from this data, which at most encode the observation that CP-structures emerge before embedding, which in turn emerges before Split CP-structures. No matter which order we try to establish for other structures (e.g., Wh > Illoc, or Wh > Topic), then, we almost always cannot get a universal one, some child/language will show a different order of emergence. This contrasts with the results observed in Friedmann et al. (2021) for Hebrew children and Meira & Grolla (2022) for Brazilian Portuguese, who argue that structurally higher elements (e.g., topics, embedding) generally emerge significantly later than lower ones (e.g., wh-questions).

This outcome is suggestive, and endorses Generalisation 2 further: if acquisition and structural height often mismatch, then we might expect shared Guttman Scales to be hard to establish. In section 4, I propose to reconceptualise this aspect of development and argue that granularity appears to bear a potentially more consequential role.20

Thus, Generalisations 1 and 2 reinforce the assumption that CP is acquired early (section 2), but provide novel support against strict directionality as recently conceived of in cartographic approaches. This confers a clear advantage to continuity, inward-growing and neo-emergentist approaches alike. Having addressed directionality in development so far, I now introduce the final set of results (Stage 2), which instead concerns granularity. I argue in section 4 that this stage singles out neo-emergentist approaches as better suited for this data.

3.4.3 Split CP-structures (Stage 2): refining the system

Generalisations 1–2 provide new data and theoretical synthesisation corroborating already-existing results; notably, previous work supporting early acquisition of CP (section 2). However, it is at Stage 2 that a novel empirical generalisation emerges. At this stage, evidence for co-occurrence of multiple CP-elements emerges robustly, something that was virtually absent up until this point.

I will argue that this generalisation provides fresh insights into the development of the left periphery and functional categories more broadly – by emphasising an underexplored aspect of child grammatical knowledge, its syntactic granularity. This is discussed in section 4; I focus first on its empirical manifestation. Fundamentally, Generalisation 3 uncovers an apparent discrepancy in the quality and quantity of CP-related productions in the child data at various stages. Table 12 shows the clear “late” nature of Split CP-structures vis-à-vis CP-structures.

Table 12: Emergence of CP vs. Split CP-structures.

CP-structures Split CP-structures
Laura 1;10.22 3;03.21
1.15 MLUw 2.54 MLUw
Gisela 2;04.25 2;08.00
1.58 MLUw 2.61 MLUw
Martina 1;08.02 2;04.13
1.57 MLUw 2.69 MLUw
Rosa 1;07.13 2;10.14
1.27 MLUw 2.5 MLUw
Irene 1;04.16 1;11.13
1.32 MLUw 2.95 MLUw
Koki 1;07.20 2;04.18
1.96 MLUw 2.69 MLUw
Kerstin 1;10.03 2;09.11
1.28 MLUw 2.32 MLUw
Simone 1;09.11 2;06.23
1.54 MLUw 2.78 MLUw
Josse 2;00.07 2;11.09
1.2 MLUw 3.57 MLUw
Sarah 1;10.05 3;00.19
1.09 MLUw 3.52 MLUw

Some structures with multiple elements in the left periphery attested at this stage in Catalan, Italian and Spanish are shown in (52)–(54). A very common kind of structure displaying evidence for a Split CP involves topics preceding wh-questions (52a), (53a), (53b), (54b). Example (52b) includes recursive topics. (54a) presents a quotative complementiser preceding a wh-word.

    1. (52)
    1. Catalan
    1.  
    1. a.
    1. I
    2. and
    1. el
    2. the
    1. ioc’ioc,
    2. ioc-ioc
    1. què
    2. what
    1. fa?                                             (Laura, MLUw 2.91)
    2. do.3sg
    1. ‘And the ioc-ioc (onomatopoeia), what does it do?’
    1.  
    1. b.
    1. Ara,
    2. now
    1. aquest,
    2. this
    1. no
    2. no
    1. puc                                                       (Gisela, MLUw 2.71)
    2. can.1sg
    1. ‘Now, this one, I can’t.’
    1. (53)
    1. Italian
    1.  
    1. a.
    1. Quetto,
    2. this
    1. cosa
    2. what
    1. fa?                                                            (Martina, MLUw 2.37)
    2. do.3sg
    1. ‘This, what does it do?’
    1.  
    1. b.
    1. Questo,
    2. this
    1. dove
    2. where
    1. si
    2. cl.refl=
    1. mette?                                          (Rosa, MLUw 2.5)
    2. put.3sg
    1. ‘This, where do you put it?’
    1. (54)
    1. Spanish
    1.  
    1. a.
    1. ¿Que
    2. that.quot
    1. qué
    2. what
    1. nos
    2. cl.io=
    1. falta?                                           (Irene, MLUw 3.23)
    2. lack.3sg
    1. ‘(I’ve asked) What are we missing?’
    1.  
    1. b.
    1. ¿Esas
    2. those
    1. bonitas
    2. beautiful
    1. cortinas,
    2. curtains
    1. dónde
    2. where
    1. las
    2. cl.do=
    1. compraron?  (Koki, MLUw 2.71)
    2. buy.prfv.3pl
    1. ‘Those beautiful curtains, where did they buy them?’

At Stage 3, at least three separate slots are also detectable in the production data. Example (55a) combines what appears to be an adverbial topic ahir, followed by a second topic/focus, alongside another topic (the CLLD topic a jo). Similarly, example (55b) displays a que sí que structure (see section 3.2.2) and (56a)–(56b) instantiate two topics (ahora and ésto/todos los muchachos) preceding a wh-element ((a) dónde). By MLUw ≥2.5 I conclude, then, that all children can harness two, and, based on some examples, seemingly also three, separate projections within CP.

    1. (55)
    1. Catalan
    1.  
    1. a.
    1. Ahir,
    2. yesterday
    1. tu,
    2. you
    1. a
    2. to
    1. jo,
    2. me
    1. me
    2. cl.io=
    1. vas
    2. aux.go.2sg
    1. dir
    2. tell
    1. que
    2. that
    1. me
    2. cl.io=
    1. l’explicaries
    2. cl.do=explain.cond.2sg
    1. avui                                                                                          (Laura, MLUw 2.91)
    2. today
    1. ‘Yesterday, YOU told ME that you would explain it to me today.’
    1.  
    1. b.
    1. Que
    2. that.quot
    1. sí
    2. yes
    1. que
    2. that.quot
    1. vol                                           (Gisela, MLUw 2.61)
    2. want.3sg
    1. ‘(I’ve said) He/she does want to.’
    1. (56)
    1. Spanish
    1.  
    1. a.
    1. ¿Y
    2. and
    1. ahora,
    2. now
    1. ésto,
    2. this
    1. dónde
    2. where
    1. se
    2. cl.imprs=
    1. mete?                  (Irene, MLUw 4.01)
    2. put.3sg
    1. ‘Now, this, where does it go?’
    1.  
    1. b.
    1. ¿Ahora,
    2. now
    1. todos
    2. all
    1. los
    2. the
    1. muchachos,
    2. boys
    1. a
    2. to
    1. dónde
    2. where
    1. se
    2. cl.refl=
    1. van
    2. go.3pl
    1. a
    2. to
    1. ir?          (Koki, MLUw 3.38)
    2. go.inf
    1. ‘Now, all the boys, where are they going to go?’

In Germanic languages, Stage 3 involves the emergence of the restricted kinds of V3-orders discussed in section 3.2.2. Frame-setters yielding V3-orders are illustrated in (57a)–(58c). (57a) and (58a) contain instances of Hanging Topics, while (57b), (57c), (58b), (58c) present cases of Contrastive Left-Dislocation.

    1. (57)
    1. German
    1.  
    1. a.
    1. Und
    2. and
    1. dann
    2. then
    1. das
    2. that
    1. brauch(e)
    2. need.1sg
    1. ich                                                                    (Kerstin, MLUw 2.68)
    2. I
    1. ‘And then I need that.’
    1.  
    1. b.
    1. Den
    2. the.acc
    1. Teddy,
    2. Teddy
    1. ich
    2. I
    1. halt
    2. hold.1sg
    1. den
    2. pron.acc
    1. so                                              (Simone, MLUw 2.78)
    2. so
    1. ‘The Teddy, I hold him like this.’
    1.  
    1. c.
    1. Die
    2. the
    1. Leute,
    2. people
    1. die
    2. they
    1. ham [: haben]
    2. have.3pl
    1. des
    2. that
    1. puttemacht [: kaputtgemacht]        (Simone, MLUw 4.89)
    2. break.make.ptcp
    1. ‘The people, THEY have broken that.’
    1. (58)
    1. Dutch
    1.  
    1. a.
    1. En
    2. and
    1. toen
    2. then
    1. daar
    2. there
    1. was
    2. aux.be.pst.3sg
    1. Hanneke
    2. Hanneke
    1. weggegaan                                  (Josse, MLUw 3.91)
    2. away.go.ptcp
    1. ‘And then, there Hanneke had left.’
    1.  
    1. b.
    1. De
    2. the
    1. staart,
    2. tail
    1. die
    2. that
    1. had
    2. aux.have.pst.2sg
    1. je
    2. you
    1. goed
    2. good
    1. gevonden                           (Sarah, MLUw 3.15)
    2. find.ptcp
    1. ‘The tail, that you found good.’
    1.  
    1. c.
    1. Andere
    2. other
    1. boekje,
    2. book.dim
    1. die
    2. that
    1. lees
    2. read.1sg
    1. ik
    2. I
    1. nooit                                                   (Sarah, MLUw 2.89)
    2. never
    1. ‘The other book/booklet, I never read (it).’

Finally, cases of conditional/temporal clauses with adverbial resumption are exemplified below. These are attested primarily in the Dutch data and are abundant in both Sarah and Josse after MLUw 2.5 (36 and 12 instances are recorded, respectively). 1 clear-cut example is found in the German data (in Simone).

    1. (59)
    1. Dutch, V3-order in conditional/temporal clauses with resumptive dann/dan
    1.  
    1. a.
    1. Als
    2. when/if
    1. ik
    2. I
    1. naar
    2. to
    1. de
    2. the
    1. grote
    2. big
    1. school
    2. school
    1. gaat,
    2. go.3sg
    1. dan
    2. then
    1. komt
    2. come.3sg
    1. sneeuw                                      (Josse, MLUw 3.78)
    2. snow
    1. ‘When I go to the big school, then snow comes.’
    1.  
    1. b.
    1. Als
    2. when/if
    1. je
    2. you
    1. drie
    2. three
    1. bent,
    2. be.2sg
    1. dan
    2. then
    1. mag
    2. may.2sg
    1. je
    2. you
    1. niet
    2. not
    1. van
    2. from
    1. de
    2. the
    1. politie
    2. police
    1. op
    2. in
    1. school                                                                                                                                                         (Sarah, MLUw 3.05)
    2. school
    1. ‘When/if you’re three, the police don’t let you into school.’
    1. (60)
    1. German
    1. Wenn
    2. when/if
    1. du
    2. you
    1. keine
    2. no
    1. Suppe
    2. soup
    1. aufißt,
    2. up.eat.3sg
    1. dann
    2. then
    1. hau
    2. smack
    1. ich
    2. I
    1. den
    2. pron.acc
    1. dir                                         (Simone, MLUw 4.89)
    2. you.dat
    1. ‘If you don’t eat up the soup, I’ll smack you.’

The data moreover indicates not just a delay, but also abruptness and “explosiveness” in the way these structures emerge. For the vast majority of children no Split CP-structures or extremely few are detectable before MLUw ∼2.5; they abruptly increase in quantity after this point in a statistically very significant manner, as shown by the results of a Wilcoxon Signed Rank test (z = –2.949874, p = .003).

Note that the emphasis here is not on the exact MLUw ranges observed, but rather on the distinct stages observed. It is likely that this impressionistic MLUw ∼2.5 threshold will not replicate itself in other children (note already that some variation is observed in the Dutch children, whose Stage 2 begins at ∼3.5). The overall pattern would remain unaffected so long as Stage 3 remains a clearly separate stage.

Figure 5 plots the “explosive” emergence of Split CP-structures. The x-axis shows all files ordered by MLUw, and the y-axis the number of Split CP-structures produced. A change-point analysis was applied to this dataset. Change-point analysis is a statistical technique used to identify the point(s) in a time series at which a change occurs. It determines the number of significant changes observed in probability distributions and estimates the time of each change. All files from all languages were included, with Split CP-structure counts as the dependent variable. The red horizontal lines mark the estimated mean levels for each detected phase. The analysis reveals a first major change at approximately MLUw ∼2.5 (file 170) – this is the same timeframe used for the comparison in Table 13.

Table 13: Production of Split CP-structures before and after MLUw ∼2.5.

Before MLUw ∼ 2.5 After MLUw ∼ 2.5 %
Laura 1 20 4.8–95.2%
Gisela 0 9 0–100%
Martina 0 5 0–100%
Rosa 1 31 3.1–96.9%
Irene 0 85 0–100%
Koki 0 41 0–100 %
Kerstin 3 4 42.9–57.1%
Simone 2 7 22.2–77.8%
Josse 1 19 5–95%
Sarah 2 51 3.8–96.2%
Total 10 272 3.5–96.5%

Figure 5: Change-point analysis of the distribution of Split CP-structures.

The question then arises why, if some approaches posit cartographic representations throughout development, these do not materialise in the production data until significantly late stages, often after embedding structures emerge (Table 11). And conversely, CP-structures are nonetheless observable from the earliest word combinations. The abruptness and explosiveness with which Split CP-structures emerge is also another important explanandum for any approach. This, then, raises an apparent representational paradox: how can we reconcile early coarse-grained CP-structures, with late Split CP-structures?

The finding from this third stage is striking in light of approaches that assume a UG-given spine to be accessible to the child (either partially at some stages, as in maturation, or totally, as in continuity). Based on the data presented, it appears that, while peripheral positions are used very early on, children operate initially on a coarse-grained discourse domain (interpreted as an emergent CP allowing a single head and specifier). Later developmental stages refine this coarse-grained CP and successively divide it into separate projections, such that this refined CP can now accommodate co-occurring CP-related structures. Crucially, a very similar generalisation was obtained in Soares (2006) for European Portuguese data, presenting a Derivational Complexity-based analysis (Jakubowicz 2011) whereby “the C category splits in the course of the derivation in function of its formal features, in order to create the necessary syntactic space enabling the verification of these features” (p. i).21

I therefore suggest the data is best explained if fine-grained syntactic heads of the sort in Rizzi (1997) are emergent and are not available as soon as CP becomes apparent at Stage 1, contra the expectations of maturational approaches (and recent cartographic continuity approaches like Westergaard 2009). Instead, they can be shown to systematically emerge at this second stage, as part of a refinement process over developmentally-earlier structures (a single-projection CP). The data in this paper offers the first preliminary developmental evidence for this claim. Significantly, this generalisation also corroborates independent syntactic and biolinguistic work advocating for an ‘emergent cartography’; i.a., Ramchand & Svenonius (2014), Svenonius (2016), Scontras et al. (2017), Mišmaš et al. (2018), Leivada & Westergaard (2019), Marušič et al. (2019), Larson (2021) and Ramchand (2023).22 As a result, the following tentative empirical generalisation surfaces:

Generalisation 3 (Cartography is Emergent). Evidence for cartographic-type structure within CP systematically and abruptly emerges at a later developmental stage, elaborating on developmentally-prior structure (a “basic” CP).

A clear alternative hypothesis at this point, however, may be that the patterns observed are due to performance considerations. One notable possibility would be that Split CP-structures emerge late simply because they may be longer than basic CP-structures and thus might be expected to require higher MLUw. I give some brief results here to argue that relative length of the two diagnostics sets is likely not the driving factor of the patterns attested.

Via fixed-effects logistic regression, I addressed the possibility that the main predictor of presence/absence of Split CP utterances is their relative length, as opposed to Age and MLUw;23 specifically, we ask if, having partialed out the effect of length in our model, Age and/or MLUw are still significant predictors. A sample of 264 utterances with CP-structures and with Split CP-structures from 5 children was extracted, including data for all relevant MLUw stages studied. The type of diagnostic reflected in an utterance (CP vs. Split CP) was coded as a binary dependent variable. The length (measured in words) of each utterance was gathered in a Length variable, as well as the MLUw and Age (in months) associated with each utterance.

Model comparisons with ANOVAs established that the best model fit included all three variables in the model as fixed effects (MLUw, Age and Length). Child was not included as a random effect; the AIC (Akaike information criterion) of the baseline model was lower than the model with random intercepts (322.4768 vs. 324.4768, respectively). The Model Likelihood Ratio Test also showed that the mixed-effects baseline model does not explain more variance (p = 1). As a result, the inclusion of random-effect structure in the model was not justified; I report here the results of the fixed-effects model. Thus, I fitted a logistic model (estimated using ML) to predict the use of Split CP utterances given these three variables (formula: splitcp ∼ mlu + length + age). The effect of MLUw is highly statistically significant and positive (β = 1.23, 95% CI [0.63, 1.88], p < .001; Std. β = 0.87, 95% CI [0.44, 1.32]). Age is likewise highly statistically significant and positive (β = 0.08, 95% CI [0.03, 0.13], p < .001; Std. β = 0.59, 95% CI [0.25, 0.95]). Crucially, the effect of Length is statistically non-significant and positive (β = 0.04, 95% CI [–0.10, 0.18], p = 0.563; Std. β = 0.10, 95% CI [–0.24, 0.45]). This indicates both MLUw and Age are strong predictors of use/absence of Split CP-structures, but Length is not.

Therefore, I discard a length-based account of Generalisation 3, as it does not receive support from this data. Note also that an additional requirement on a performance-only account of these patterns is an explanation of the “explosiveness” facet of Split CP-structures. This observation aligns with the conclusions in Snyder (2007; 2021), who independently argues for the development of English verb-particle constructions that “explosive” changes in production (“decisive”, in his terms) are signs of a potential grammatical change. At any rate, however, establishing the role of other possible performance effects in the acquisition of Split CP-structures, above and beyond the length factor considered here, is a very important future task.

An anonymous reviewer also notes that the fact that many Split CP-structures are of the form Top>Wh (in Romance) (see p. 12), wherein both CP-constituents have been moved, may bear some significance under intervention-based accounts of syntactic development (e.g., Moscati 2023). Under these approaches, (at least some) Split CP-structures may be ruled out early on, not because of lack of CP-structure, but because Relativised Minimality (Rizzi 2004) reduces movement possibilities for children – as has been shown for the acquisition of object relatives and object wh-questions (Friedmann et al. 2009). This is a plausible objection to the account proposed here; however, it still overgenerates the kinds of Split CP-structures that should be produced earlier on. Specifically, it only predicts dislocations that generate a surface order deviating from Subject > Object to be problematic, i.e., O[top] S[wh/foc]. This would then predict S[top] O[wh/foc] to be earlier-acquired.

The developmental patterns in Generalisation 3, however, suggest a much wider developmental ‘gap’, namely complete absence of Split CP-structures at earlier stages, either intervention-inducing or not. Indeed, based on a manual look at the Top>Wh structures collected, only 4 appear to involve true O[top]S[wh/foc]V orders, i.e., intervention configurations. Object topics but with non-subject wh-words (where, how and why) seem more common, in contrast (22 utterances attested). The same pattern is in fact reported for the acquisition of object topicalisation in German, only object topics that induce intervention effects are later-acquired (contrast section 3.4.2 with Ruigendijk & Friedmann 2017). This therefore corroborates that intervention does play a role in our patterns, as expected by Moscati (2023): Split CP-structures with intervention are barely attested. Even so, non-intervention configurations are produced abundantly (see examples in (52)–(54)), but nonetheless remarkably late – this latter pattern remains unexplained by syntactic intervention. I suggest, therefore, that intervention in the sense of Moscati (2023) only plays a partial role in the patterns at stake, in a way which is perfectly compatible with the broad proposals of this paper.

I now discuss the theoretical implications of these results in the next section and adopt a representational difference account of the contrast in Stage 1 vs. 2.

4 Discussion and implications

4.1 Theoretical implications: on the need for ontological flexibility

The goal of this paper was to identify potential ways in which incorporating granularity considerations in developmental theorising may be instructive. The results of the corpus study challenge the predictions of maturational and continuity approaches in various respects. Recall again the generalisations obtained, repeated here:

  • i. Generalisation 1 (Early Acquisition of CP). CP-structures emerge early on in the developmental data.

  • ii. Generalisation 2 (Structural Height and Acquisition Mismatch). There is a dissociation between structural height and order of emergence. Acquisition does not proceed successively upwards; some syntactically very high elements emerge early.

  • iii. Generalisation 3 (Cartography is Emergent). Evidence for cartographic-type structure within CP systematically and abruptly emerges at a later developmental stage, elaborating on developmentally-prior structure (a “basic” CP).

On the one hand, Generalisation 1 undermines the feasibility of bottom-up maturational approaches. The variety of early CP-structures observed suggests that a comprehensive model of these patterns should recognise a key insight of both continuity and inward-growing approaches: CP is an early development. At the same time, Generalisation 2 indicates that this early CP cannot be achieved by recapitulating a given cartographic spine in any directionality. This is at odds with recent approaches like the Growing Trees Hypothesis. The combination of Generalisations 1–2, then, underscores the fragility of a cartographic maturational approach based exclusively on directionality: irrespective of directionality assumed (bottom-up, inwardly), some structures are predicted to be late (e.g., topics/complementisers for a bottom-up approach), but turn out to be remarkably early developments in several children. Directionality of emergence is therefore not a productive tool with which to rationalise the empirical patterns here.

Stepping back, the developmental trends share many similarities with production and behavioural data reported for other languages supporting an early CP: e.g., for wh-questions (i.a., Santelmann 2003, for Swedish; Soares 2003, for European Portuguese; Saikkonen 2018, for Finnish; Seidl et al. 2003, and Perkins & Lidz 2021, for English), V2 (i.a., Boser et al. 1992, for German; van Kampen 2010, for Dutch; Westergaard 2009, for Norwegian), topicalisations and focalisations (Tsimpli 2005, for Greek; de Cat 2009, for French). If (some) interactional language is also hosted in the CP-domain (or a higher Speech Act domain), Shirai et al. (2000) and Liu & Muramatsu (2025) also report very early emergence of (some) Mandarin and Japanese Sentence-Final particles; and Heim (2023) and Stoicescu (2025) provide data from English question tags and illocutionary complementisers in Romanian, respectively. This suggests that there is significant independent support for several of the claims adduced here.

The above leaves us with three extant approaches capable of capturing Generalisations 1–2: neo-emergentism, inward-growing approaches (e.g., Galasso 2003; Heim & Wiltschko 2025) and continuity. All expect early development of CP and do not make explicit commitments as to the order of emergence of CP-structures based on a cartographic left periphery.

However, Generalisation 3, I argue, requires, at least, a partial departure from all current approaches operating with innate categories, both maturation and continuity approaches. The empirical data presented forces us to assume some early representation of the CP-domain, but this representation cannot be finer-grained than Chomsky’s Core Functional Categories at this earlier stage, otherwise the theory overgenerates. Specifically, the generalisations force a theoretical approach which assumes some early knowledge of the CP-domain that does not develop into a cartographic-type left periphery until later. Development cannot simply recapitulate tree geometry for this effect to be generated. It would necessitate categorial systems and their granularity to be “malleable” during development, an analysis which is incompatible with proposed maturational approaches assuming exclusively innate categorial priors, such as a cartographic spine. The maturational proposals tested here, then, either undergenerate (assuming no CP-knowledge at the first stages of acquisition) or overgenerate (assuming too fine-grained knowledge of the CP-domain as soon as the latter emerges). I conclude, then, that a maturational approach is too rigid for this dataset.

Notice, additionally, that non-cartographic approaches do not technically face this contradiction, by virtue of assuming more impoverished clausal spines (such as Heim & Wiltschko’s 2025 Inward Growing Hypothesis and the earlier “Empty Middle” approach by Galasso 2003; or early versions of continuity). However, they arguably do not possess the tools to account for the contrast between Stage 1 and Stage 2. While not necessarily ruled out as possible developmental theories, further elaboration would be required to accommodate the patterns in Stage 3. Put differently, the three problems laid out in what follows would still apply to them. Mutatis mutandis, a similar point holds for continuity approaches: the abrupt increase in Split CP productions is unaddressed. This also carries over to other emergentist approaches that make developmental predictions based instead on “derivational ordering or timing”, such as Diercks & Bossi’s (2021) and Diercks et al.’s (2023) Developmental Minimalist Syntax. To be clear, the problem is not restricted to maturational approaches, but its paradoxical nature becomes most problematic in this type of theory.

With this in mind, I now propose to rethink these challenging empirical points by incorporating granularity into our analytical toolkit, in place of directionality, to address the shortcomings in existing work; this addresses the shortcomings in existing work. Firstly, we can synthesise the ensuing discussion in terms of three interrelated (but dissociable) problems that have to be addressed, given this data:

  1. i. The Directionality Problem

    Developmental stages cannot straightforwardly be characterised as recapitulating a UG-given tree, either bottom-up or inwardly.

  2. ii. The Granularity Mismatch Problem

    Child grammars and adult grammars do not (always) operate at the same level of granularity. Acquisition may be characterised by increasing granularity, thus bridging the initial existing “gap” with the adult system.

  3. iii. The Comparability Problem

    The fine-tuning and structural homology patterns detected are not readily accounted for by innate categorial ontologies. Child categories may not always have direct counterparts in an adult-like UG-given spine.

I propose here that these problems are potentially serious for existent bottom-up maturational approaches, especially those of a cartographic nature. They can be resolved by dropping two assumptions in turn: firstly, foregoing an adult-like cartographic UG-given spine as the point of departure for the child and, by implication, a maturational mechanism that dictates the development of this latter spine (i.e., its developmental directionality). Doing so immediately lends us the additional ontological flexibility needed to account for (i)–(iii). In a neo-emergentist approach to acquisition, later developmental stages refine developmentally-prior categories (structural homology), leading to the increasing granularity patterns described. It also becomes misleading to describe development as recapitulating an innate spine in any given directionality and to assume that child categorial systems always have to be in direct correspondence with the categories postulated in adult-like UG-given spines (see again Heim & Wiltschko 2025: 17, for an approach with a similar ontological direction, applied to speech-act syntax: “these layers [early Speech Act layers – NBM] of structure are qualitatively different in child and adult language. […] Specifically, we propose that the child gradually expands the complexity of the spine to include all the adult layers of structure”). In effect, I propose that only an approach that assumes categories to be ontologically flexible can readily capture the patterns in Generalisations 1–3.24

As a result, no such contradiction or problems arise with Biberauer & Roberts’ (2015) successive-division learning path, which by design expects changes in representational granularity as development unfolds. The results of the corpus study are therefore consistent with it, as it correctly assumes that core macroparametric structural properties like a coarse-grained, “basic” representation of CP (e.g., Phasal or CFC) will emerge early. Only after children have mastered this option, will they acquire fine-grained (“micro-parametric”) knowledge requiring cartographic-type projections. Crucially, I am making no specific claims regarding how and when children get the entire cartographic structure in Rizzi (1997) in languages where it is required (recall the discussion at the end of section 3.2.2). Indeed, the exact richness of CP is something neo-emergentism expects to be subject to crosslinguistic variation (Biberauer & Roberts 2015). The nature of the empirical data is far too coarse to disambiguate precisely how articulated the children’s CP-representations are. The key here is that the ‘tipping point’ detected by Generalisation 3 points to a potential refinement in children’s CP-structures, now allowing featurally finer-grained, minimally part-cartographic structures. This was completely absent in Stage 1, which appeared to allow only a more generic, ‘single-slot’ CP.

The foregoing is also an argument against syntactic (domain-specific) maturation: the data is firstly incompatible with it, but, equally importantly, maturation alone also cannot straightforwardly generate increasing granularity patterns. Reconsidering the value of syntactic maturation as a notion with explanatory weight in a Three Factors context is a conceptually appealing move (in the spirit of Chomsky 2005): it is in line with more recent attempts to minimise the role of UG and it aims to derive developmental patterns from more fundamental properties of complex systems and cognition, as opposed to a (stipulative)25 biological imperative in UG. Specifically, levels of granularity are often suggested to be an important property of biological systems and cognition more broadly (see various perspectives in Zadeh 1997; Ehresmann & Vanbremeersch 2007; Chen & Du 2015; Lorkowski & Kreinovich 2015; Wang 2017; Wang et al. 2017; Kwisthout & van Rooij 2020). Smith et al. (2020) have also implemented computational simulations of structure learning based on the (Bayesian) active inference framework, showing that concept acquisition from coarse- to fine-grained may facilitate better results. In particular, Kwisthout (2013), Kwisthout et al. (2017), Rutar et al. (2022a; b) also critically discuss the need for state-space granularity in Bayesian models, and Ward et al. (2023) address its need in developmental frameworks. The importance attributed to ‘detailedness’ and ‘granularity’ considerations in biology and (computational) cognitive science thus suggests that granularity could be observable in the acquisition or shaping of language. There is some tentative evidence for this outside of syntax, see Mervis & Crisafi (1982), Xu & Tenenbaum (2007) and Murphy (2016) on lexical/concept acquisition; Krifka (2002; 2009), Cummins et al. (2012) and Enguehard (2018) on semantics/pragmatics; and Dresher (2009) and Nie et al. (2019) on phonology.

I reiterate again that I do not claim to have offered definite evidence for Generalisation 3 – this is necessarily a broader enterprise beyond the scope of this paper (see section 3.2). I do nonetheless treat the patterns emerging from the data discussed as unexpected by existing approaches and as worthy of explanation. The objectives of this section, then, are, firstly, to present a novel asymmetry and, secondly, to suggest a promising underexplored theoretical account of the empirical skew in the realisation of CP-structures. This spirit defines a new research agenda for the acquisition of CP.

In general, then, understanding development in this neo-emergentist way proves, at the very least, productive – it does not just accommodate the data observed, but it predicts all patterns: (i) early-established formal distinctions serving as the basis for later developmental stages, (ii) developments towards ever more language-specific categorial granularity (as in complex emergent systems more generally, Gibb et al. 2019) and (iii) the reflexes of these changes in the granularity of developing grammars manifesting abruptly and explosively in the production data.

4.2 Limitations and future extensions

The present work is nonetheless limited in several ways, given the number of children and languages studied. Nonetheless, it lays the ground for future work on the role of UG, maturation and granularity in development. Several tasks that emerge from this paper represent the topic of ongoing research, beyond those already noted earlier. For example, the sample of both children and languages must be expanded, including adding new diagnostics as required by the languages studied and an analysis of their adult input data. Comparative work could prove particularly illuminating: similar timelines of acquisition for “split” or “cartographic-type” structures are presented in Mitrofanova (2018) for Russian PPs, Sanfelici & Gallina (2022) for Italian PPs, De Lisser et al. (2017) for the Jamaican Creole Tense-Aspect-Mood domain, among others, and for finer-grained speech-act syntax (Heim & Wiltschko 2025). Additionally, MLUw has proven to be a fairly stable metric in the 10 children presented; however, its role and stability must also be examined further with bigger samples (see, e.g., Oosthuizen & Southwood 2009; Ezeizabarrena & García Fernández 2017, for critical discussion on MLUw). Importantly, another research strand involves incorporating granularity considerations in experimental studies.

Finally, a future research strand includes establishing parallels with diachronic trends. Recent work on feature bundling, whereby diachronically more stable stages involve maximally bundled heads (e.g., a C head with [topic], [focus], [fin], etc.]; Lee & Cournane 2019; Klævik-Pettersen 2019a; Klævik-Pettersen 2019b; Cournane & Klævik-Pettersen 2023), makes predictions that are compatible with neo-emergentism. Other diachronic proposals are suggestively similar to the ones here, e.g. Sluckin’s (2025: 41) analysis for the development of V2/V3 systems initially involves “a maximally underspecified C-head for which the full set of features are yet to be posited” before “children acquiring Kiezdeutsch and SG [Standard German – NBM] […] decompose C/T into C and T and also separate the lower and higher specifiers of C/T into C1 and C2”. Under the present approach, then, we likewise expect “increases” and “decreases” in the elaboration of CP to be useful tools for the explanation of diachronic stages (see also Madaro et al. 2025) – in ways that may also insightfully map onto developmental data. Studying the diachrony of functional structure will inform us about preferred categorial trends and about which approaches are most suited for generating the relevant patterns.

5 Conclusion

This paper has argued that neo-emergentism offers a new framework to study the nature of language acquisition, one where categorial granularity is considered an object of study, rather than a theoretical prior. In this paper, I have provided empirical evidence to show why such a shift in analytical focus is productive, and how it opens an unexplored research strand on possible changes in granularity in development.

I identified problems with exclusively directionality-driven and fixed-granularity approaches via a multilingual corpus study. Our initial results indicate that CP emerges early across all children and languages, but reveal a novel generalisation, worthy of further investigation: production evidence for cartographic-type organisation in the CP-domain systematically emerges at a later stage. Additionally, I established an important mismatch between structural height and timing of development; some left-peripherally very high elements emerge very early.

I then argued that maturational and continuity approaches cannot capture the three generalisations introduced, as they generate theoretically contradictory requirements. This is because of a (somewhat implicit) assumption in approaches of this sort – fixed granularity – which amounts to stating that, without some further theoretical elaboration, these frameworks, by design, cannot replicate “shifts’’ in mental granularity. I emphasised how neo-emergentism offers a novel perspective on the data: I proposed to resolve this apparent contradiction by eschewing bottom-up syntactic maturation and innate cartographic categories and, crucially, by incorporating granularity in our analytical and conceptual toolkit. Emergent categories, construed in the neo-emergentist approach proposed, allow the crosslinguistic developmental patterns to be rationalised, without unnecessary stipulations. We can, with inward-growing and continuity approaches, maintain that the C-domain emerges early, but also concede that it is not cartographic in its first instantiation, this being a subsequent development.

This work, although preliminary in scope, has potentially significant implications for how developmental stages can be conceptualised and raises challenging questions about categorial ontology and flexibility. It also has implications for new ways of approaching developmental data, by introducing granularity as a testable aspect of development, probed here via Split CP-structures. Irrespective of the ultimate answers to these questions, a main goal of this paper was to bring these issues to the forefront, such that future research can work towards enhancing our understanding of the variety of grammatical factors shaping acquisition.

Data availability

The additional files for this article can be found here: https://osf.io/utf3b. This repository contains 3 Appendices:

  • Appendix A: Full tables containing the structures produced by every child.

  • Appendix B: Quantitative analysis of CP-structures.

  • Appendix C: Fixed-effects model building process and data collected.

Funding information

This research was funded by St John’s College (Hayes MPhil Scholarship) and the Cambridge Trust (Cambridge Opportunity Master’s Studentship) from 2022–23 and, from 2023-present, by the Arts and Humanities Research Council (UKRI) and St John’s College (Open-Oxford Cambridge AHRC – St John’s Doctoral Training Partnership; project reference 2890509).

Acknowledgements

I owe an incredible intellectual debt to Theresa Biberauer, who supervised the MPhil thesis on which this paper is based. My sincerest thanks go to her for her continuous enthusiasm, for her clear-headedness when approaching linguistic questions, and for encouraging me to write this paper. Anything good that may stem from this paper is a testament to her dedication to her students. This work was also presented as joint papers with Theresa at BUCLD 48, IGG 49 and GALA 16; I thank the audiences at all of these conferences for their comments. I am also very thankful to Dora Alexopoulou, Ashton Brown, Cécile de Cat, Karen Lahousse, Barbara Lust, James Morley, Roman Feiman, Emanuela Sanfelici, Itamar Schatz, Julia Schwarz, Carla Soares-Jesel, William Snyder, Virginia Valian and Bert Vaux for comments and/or help on various versions of this project. Thanks also go to audiences at various seminar groups at Cambridge, KU Leuven, Padua and Frankfurt. Last but not least, I would like to thank three anonymous reviewers for their constructive comments that helped improve this paper. Any remaining errors are entirely my own.

Competing interests

The author has no competing interests to declare.

Notes

  1. Neo-emergentist approaches adopt a maximally poor UG, modelling acquisition, to the greatest extent possible, with independently motivated general cognitive principles (so-called ‘third factors’ in Chomsky 2005). Crucially, however, their essence remains generative: systems are regulated by UG, Merge and formal features. These theoretical assumptions set the approaches apart from traditional emergentist approaches such as Construction Grammar and usage-based theories in non-negligible ways, hence the prefix neo-emergentism (see Biberauer 2019: 59). Neo-emergentism is described in more detail in section 2.4. [^]
  2. Unlike Radford (1988; 1990), Stage 1 already contains functional categories (IP/TP). Friedmann et al. (2021: 27) consider the possibility of an earlier phase with a bare VP lacking IP, ultimately concluding there is no evidence for it in their data as inflection is present (age range 1;6–6;1). They do not consider the possibility that the attested inflection does not reflect adult-like representations. Note, too, that in the full clausal spine in Figure 1 only the left periphery is cartographic. How the child would acquire more elaborate versions of the Tense-Aspect-Mood domain (see also Cinque 1999, cf. too De Lisser et al. 2017), or of the vP-domain and event-structure (e.g., Ramchand 2008) remains an open question. [^]
  3. Echoing Heim & Wiltschko’s (2025) Inward Growing Spine Hypothesis. [^]
  4. Note that continuity approaches do not suggest that there is no language-specific grammatical development with respect to functional structure; instead, the universal functional spine on which an L1 grammar is built is accessible from the outset. L1-specific properties may be hypothesised to be acquired, for example, via so-called “parameter setting”. [^]
  5. This approach therefore assumes parts of the input do not qualify as intake at certain points in development (e.g., structures requiring a fully-fledged CP), for example because their featural complexity renders them inaccessible early on (see Gagliardi 2012; Lidz & Gagliardi 2015). The key is that these ‘Split CP-structures’ become acquisitionally possible (and so can be ‘triggered’ by the input) once children have first identified the need for a generic C-domain in their language. This still begs the question of how Split CP-structures would be expected to be syntactically comprehended before their full representation is acquired (if comprehended at all), as an anonymous reviewer points out. The same question arises for other approaches assuming “incomplete” representations during early development (section 2.1). It is well beyond of the scope of this paper to address this, but at the very least, this is also an empirical question, which could be explored once experimental studies probing these theoretical approaches are carried out (see section 4.2). [^]
  6. This assumption of course does not entail a rejection of neurobiological maturation; the above concerns only syntactic maturation, understood as pre-wired development of (some of) the domain-specific repertoire in UG. The approaches presented – both continuity and neo-emergentism – are compatible with approaches that try to account for patterns of syntactic development in terms of maturation of neurobiological substrates (e.g., see Snyder et al. 2024). The key distinction here is between what Lust (1999: 122–123) termed the General Maturational Hypothesis and the Specific Grammatical Maturational Hypothesis. This paper focuses on the latter; the former is not contentious. [^]
  7. Being the densest corpus of all, only the first 50 files in Simone corpus were analysed, out of the 73 recordings in total. All diagnostics studied are produced by Simone over these 50 files. For the remaining children, the entire corpora were studied. [^]
  8. Unless otherwise noted, the examples provided are mine. For Italian, German and Dutch, they were also checked with native speakers. [^]
  9. This requirement may be unduly conservative. Santelmann (2003) argues for wh-less questions in early Swedish that they display evidence of CP-level structure (see also Evers & van Kampen 2001; van Kampen 2009, on Dutch). [^]
  10. The first Split CP-structure in Laura is attested in a file with MLUw 3.47. However, since the rest of the files in that period include substantially lower MLUs (around 2.5) and MLUw 3.0 is not reached until later files, I adopt the next file’s MLU value as the value of emergence for Split CP-structures in Laura. These are the values cited in any future tables and summaries. [^]
  11. The same approach as for Laura (see footnote 10) was adopted for the first CP-structure in Martina (attested at a file with MLUw 1.9, but surrounded by files at MLUw ∼1.5). [^]
  12. The first file of this period has a slightly higher MLUw than the final file, hence the apparently antichronological ordering. The same remark holds for other apparently antichronological timings. [^]
  13. Note that several examples feature Root Infinitives (RIs). It is beyond the scope of this work to provide an analysis of RIs crosslinguistically. Whenever finite verbs are used, though, they are almost without exception located in V2 position (Appendix B). [^]
  14. 2 wh-questions are non-finite. [^]
  15. 1 wh-question is non-finite. [^]
  16. As for footnote 15. [^]
  17. Although one could technically maintain that Generalisation 1 is only problematic for bottom-up approaches if CP emerges before TP, as an anonymous reviewer notes. This relative interpretation is possible (and is more explicitly endorsed in Friedmann et al. 2021). However, Generalisation 1 emphasises that the data would still force a departure from the original spirit of bottom-up approaches (e.g., Radford 1990; Rizzi 1993) where CP was explicitly predicted to emerge “late” (a term loosely defined, but often correlated with age 2–2.5 in Radford 1990). See also Bosch (2023a) and Heim & Wiltschko (2025), who provide some evidence that CP-structures are as early as, if not earlier, than TP-structures. [^]
  18. However, this could be an underestimate of the occurrence of topics and does not control for a scenario where topicalisations superficially emerge later than focalisations simply because they omit clitics in CLLD (see Guasti 1993; Schaeffer 2000; Wexler et al. 2004; Babyonyshev & Marin 2006; Gavarró et al. 2010; though cf. Wexler et al. 2004; Gavarró et al. 2010, on Spanish). [^]
  19. At this point, one might raise an alternative interpretation of the patterns in Generalisation 2; namely, that these “structurally high” elements do not involve structurally high heads at this particular stage, but instead recruit other lower heads like Fin until those higher heads mature. I discard this option on parsimony grounds. There is no clear motivation at present (except theory-internal reasons) for rethinking these structures as reflecting the development of left-peripherally low heads (e.g., Fin). Additionally, introducing “misanalysis” in the account would significantly hamper the falsifiability of existing cartographic proposals. [^]
  20. An anonymous reviewer points out that Generalisation 2 is only incompatible with Growing Trees regarding the timing of illocutionary complementisers; wh-questions generally (though not always) emerge before topics in Table 11, in line with Growing Trees’ Stage 2 and 3. Adopting this reasoning, however, risks being stipulative: illocutionary complementisers are generally treated as high CP-elements in cartographic approaches (see section 3.2), and have been reported to be early acquired for all of Catalan, Spanish, some Italian children (this paper; Bosch 2023b) and Romanian (Stoicescu 2025). All other things being equal, this pattern alone should be taken to contradict a Growing Trees approach. Further, the extent to which the acquisition of topicalisation follows a Growing Trees pathway crosslinguistically is far from straightforward: I refer the reader to Bosch & Biberauer (2025), who show that early topics are observed in a significant range of languages, correlating with featural complexity; to de Cat (2007; 2009), who reports early topics for French, even before wh-questions are acquired; and to Villa-García & Snyder (2009), who compare the emergence of lexical (topical) subjects with wh-questions in Spanish. [^]
  21. Thanks very much to Carla Soares-Jesel for drawing my attention to this work. [^]
  22. See also Lee & Cournane (2019), Cournane & Klævik-Pettersen (2023) for other relevant work on diachronic syntax. [^]
  23. Note that MLUw is crucially distinct from Length here: MLUw is a measure of how many words utterances have on average in a file, across all utterances. MLUw will increase as age increases, irrespective of whether the change is driven purely by sentence length or by grammatical changes. The alternative hypothesis at stake here is whether the difference between CP- and Split-CP-structures is just a difference in the length of these specific utterances (i.e., whether Length explains all the variance). [^]
  24. There still remains the possibility that an innate template of conceptual categories guides this emergence process (e.g., the classification → pointofview → anchoring → linking template in Song 2019, from Wiltschko 2014, which mediates all granularity levels; see also Ramchand & Svenonius 2014). Additionally, continuity and inward-growing approaches could be made compatible with our data so long as emergent refinements on early-existing structures are theorised. I leave this open. The key argument here is that categorial divisions would have to be at least partly emergent to capture the patterns. [^]
  25. Maturation provides a possible description of some developmental patterns, but it is arguably not explanatory. Among other issues, it does not explain why functional spines should develop in the way they are proposed to (e.g., bottom-up, inwardly), nor does it address why the partitions of the spine into stages are the ones they are. It simply encodes the relevant developmental pathway by brute force (see Lust 1999: 124–128, in particular). [^]

References

Babyonyshev, Maria & Marin, Stefania. 2006. Acquisition of Romanian pronominal clitics. Catalan Journal of Linguistics 5. 17–44. DOI:  http://doi.org/10.5565/rev/catjl.71

Barton, David. 1976. The role of perception in the acquisition of phonology. University of London dissertation.

Biberauer, Theresa. 2011. In defence of lexico-centric parametric variation: two 3rd factor-constrained case studies. Paper presented at the Workshop on Formal Grammar and Syntactic Variation: Rethinking Parameters (Madrid).

Biberauer, Theresa. 2018. Less is More: On the Tolerance Principle as a manifestation of Maximize Minimal Means. Linguistic Approaches to Bilingualism 8(6). 707–711. DOI:  http://doi.org/10.1075/lab.18080.bib

Biberauer, Theresa. 2019. Factors 2 and 3: Towards a principled approach. Catalan Journal of Linguistics (Special Issue) 45–88. DOI:  http://doi.org/10.5565/rev/catjl.219

Biberauer, Theresa & Roberts, Ian. 2015. Rethinking formal hierarchies: A proposed unification. Cambridge Occasional Papers in Linguistics 7. 1–31.

Bobaljik, Jonathan David & Thráinsson, Höskuldur. 1998. Two heads aren’t always better than one. Syntax 1(1). 37–71. DOI:  http://doi.org/10.1111/1467-9612.00003

Bol, Gerard W. 1995. Implicational scaling in child language acquisition: The order of production of Dutch verb constructions. In Verrips, Maaike & Wijnen, Frank (eds.), Papers from The Dutch-German Colloquium on Language Acquisition, Amsterdam series in child language development, 3. Amsterdam: Institute for General Linguistics.

Bosch, Núria. 2023a. Emergent syntax and maturation: A neo-emergentist approach to syntactic development. University of Cambridge MPhil thesis.

Bosch, Núria. 2023b. Not all complementisers are late: A first look at the acquisition of illocutionary complementisers in Catalan and Spanish. Isogloss. Open Journal of Romance Linguistics 9. 1–39. DOI:  http://doi.org/10.5565/rev/isogloss.313

Bosch, Núria & Biberauer, Theresa. 2024. Emergent syntactic categories and increasing granularity: Evidence from a multilingual corpus study. In Proceedings of the 48th Boston University Conference on Language Development (BUCLD). 101–116. Cascadilla Proceedings Project.

Bosch, Núria & Biberauer, Theresa. 2025. On another topic, how do acquisition orders vary? the left-periphery and topicalization in bilingual and monolingual acquisition. In Proceedings of the 49th Boston University Conference on Language Development (BUCLD). 129–144. Somerville, MA: Cascadilla Proceedings Project.

Boser, Katherine & Lust, Barbara & Santelmann, Lynn & Whitman, John. 1992. The syntax of CP and V-2 in Early Child German (ECG): The strong continuity hypothesis. In Proceedings of the Northeast Linguistic Society (NELS) 22, 51–66. Amherst: University of Massachussets.

Caprin, Claudia & Guasti, Maria Teresa. 2009. The acquisition of morphosyntax in Italian: A cross-sectional study. Applied Psycholinguistics 30(1). 23–52. DOI:  http://doi.org/10.1017/S0142716408090024

Chen, Shu-Heng & Du, Ye-Rong. 2015. Granularity in economic decision making: An interdisciplinary review. In Pedrycz, Witol & Chen, Shyi-Ming (eds.), Granular computing and decision-making: Studies in big data, vol. 10, 47–71. Cham: Springer. DOI:  http://doi.org/10.1007/978-3-319-16829-6_3

Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Martin, Roger & Michaels, David & Uriagereka, Juan (eds.), Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 85–156. Cambridge, MA: MIT Press.

Chomsky, Noam. 2001. Derivation by Phase. In Kenstowicz, Michael (ed.), Ken Hale: A Life in Language, 1–50. Cambridge, MA: MIT Press. DOI:  http://doi.org/10.7551/mitpress/4056.003.0004

Chomsky, Noam. 2005. Three factors in language design. Linguistic Inquiry 36(1). 1–22. DOI:  http://doi.org/10.1162/0024389052993655

Chomsky, Noam. 2008. On Phases. In Freidin, Robert & Otero, Carlos P. & Zubizarreta, Maria Luisa (eds.), Foundational issues in linguistic theory, 133–166. Cambridge, MA: MIT Press. DOI:  http://doi.org/10.7551/mitpress/7713.003.0009

Cinque, Guglielmo. 1990. Types of A-bar dependencies (Linguistic Inquiry Monographs). Cambridge: MIT Press.

Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780195115260.001.0001

Cipriani, Paola & Pfanner, Pietro & Chilosi, Anna Maria & Cittadoni, Lorena & Ciuti, Alessandro & Maccari, Anna & Pantano, Natalia & Pfanner, Lucia & Poli, Paola & Sarno, Stefania & Bottari, Piero & Cappelli, Giuseppe & Colombo, C. & Veneziano, Edy. 1989. Protocolli diagnostici e terapeutici nello sviluppo e nella patologia del linguaggio. Italian Ministry of Health: Stella Maris Foundation.

Clahsen, Harald. 1990. Constraints on parameter setting: A grammatical analysis of some acquisition stages in German child language. Language Acquisition 1(4). 361–391. DOI:  http://doi.org/10.1207/s15327817la0104_3

Clahsen, Harald & Eisenbeiss, Sonja & Vainikka, Anne. 1994. The seeds of structure: A syntactic analysis of the acquisition of case marking. In Hoekstra, Tom & Schwartz, Bonnie (eds.), Language acquisition studies in generative grammar, 85–118. Amsterdam: John Benjamins. DOI:  http://doi.org/10.1075/lald.8.07cla

Clahsen, Harald & Kursawe, Claudan & Penke, Martina. 1996. Introducing CP: Wh-questions and subordinate clauses in German child language. In Koster, Charlotte & Wijnen, Frank (eds.), Proceedings of the Groningen Assembly on Language Acquisition, 5–22. Groningen: Center for Language and Cognition.

Cormany, Ed. 2015. Changes in Friulano subject clitics: Conflation and interactions with the left periphery. In Biberauer, Theresa & Walkden, George (eds.), Syntax over time: Lexical, morphological, and information–structural interactions, 249–264. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/acprof:oso/9780199687923.003.0015

Corr, Alice. 2016. Illocutionary complementisers and utterance syntax. University of Cambridge dissertation.

Corr, Alice. 2022. The grammar of the utterance: How to do things with Ibero-Romance. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780198856597.001.0001

Cournane, Ailís & Klævik-Pettersen, Espen. 2023. The role of the conservative learner in the rise and fall of verb-second. Journal of Historical Syntax 7(6–19). 1–48. DOI:  http://doi.org/10.18148/hs/2023.v7i6-19.160

Cowper, Elizabeth & Hall, Daniel Currie. 2014. Reductio ad discrimen: Where features come from. Nordlyd 41(2). 145–164. DOI:  http://doi.org/10.7557/12.3411

Cruschina, Silvio & Remberger, Eva-Maria. 2018. Speaker-oriented syntax and root clause complementizers. Linguistic Variation 18(2). 336–358. DOI:  http://doi.org/10.1075/lv.16009.cru

Cummins, Chris & Sauerland, Uli & Solt, Stephanie. 2012. Granularity and scalar implicature in numerical expressions. Linguistics and Philosophy 35(2). 135–169. DOI:  http://doi.org/10.1007/s10988-012-9114-0

de Cat, Cécile. 2007. French dislocation: Interpretation, syntax, acquisition. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780199230471.001.0001

de Cat, Cécile. 2009. Experimental evidence for preschoolers’ mastery of “topic”. Language Acquisition 16(4). 224–239. DOI:  http://doi.org/10.1080/10489220903190612

De Clercq, Karen & Haegeman, Liliane & Lohndal, Terje & Meklenborg, Christine (eds.). 2023. Adverbial resumption in Verb Second languages (Oxford Studies in Comparative Syntax). Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780197651148.001.0001

De Lisser, Tamirand Nnena & Durrleman, Stephanie & Shlonsky, Ur & Rizzi, Luigi. 2017. The acquisition of Tense, Modal and Aspect markers in Jamaican Creole. Journal of Child Language Acquisition and Development 5(4). 219. DOI:  http://doi.org/10.5281/zenodo.7697250

de Villiers, Jill. 2001. Continuity and modularity in language acquisition and research. Philosophy: Faculty Publications https://scholarworks.smith.edu/phi_facpubs/47. DOI:  http://doi.org/10.1075/arla.1.02vil

den Besten, Hans. 1983. On the interaction of root transformations and lexical deletive rules. In Abraham, Werner (ed.), On the formal syntax of West-Germania, 47–131. Amsterdam: John Benjamins. DOI:  http://doi.org/10.1075/la.3.03bes

den Dikken, Marcel & Surányi, Balázs. 2017. Contrasting Contrastive Left-Dislocation explications. Linguistic Inquiry 48(4). 543–584. DOI:  http://doi.org/10.1162/LING_a_00254

Diercks, Michael & Bossi, Madeleine. 2021. Minimalist syntax is psychologically real: Lessons from (counter)cyclicity. Ms. https://lingbuzz.net/lingbuzz/005671.

Diercks, Michaels & Johns, Katherine & Bar-Server & Bossi, Madeline. 2023. Developmental Minimalist Syntax. Ms. https://lingbuzz.net/lingbuzz/007134.

Dresher, B. Elan. 2009. The contrastive hierarchy in phonology. Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9780511642005

Ehresmann, Andrée & Vanbremeersch, Jean-Paul. 2007. Memory Evolutive Systems: Hierarchy, emergence, cognition (Studies in Multidisciplinarity). Amsterdam: Elsevier.

Enguehard, Émile. 2018. Comparative modified numerals revisited: scalar implicatures, granularity and blindness to context. In Maspong, Sireemas & Stefánsdóttir, Brynhildur & Blake, Katherine & Davis, Forrest (eds.), Proceedings of SALT 28. 21–39. Linguistic Society of America. DOI:  http://doi.org/10.3765/salt.v28i0.4403

Evers, Arnold & van Kampen, Jacqueline. 2001. E-language, I-language and the order of parameter setting. UiL OTS Working Papers 00105-S-S.

Ezeizabarrena, Maria-José & García Fernández, Iñaki. 2017. Length of utterance, in morphemes or in words?: MLU3-w, a reliable measure of language development in early Basque. Frontiers in Psychology 8. 2265. DOI:  http://doi.org/10.3389/fpsyg.2017.02265

Fikkert, Paula & Levelt, Clara. 2008. How does Place fall into place? The lexicon and emergent constraints in children’s developing phonological grammar. In Avery, Peter & Dresher, Bezalel Elan & Rice, Keren (eds.), Contrast in phonology: Theory, perception and acquisition, 231–268. The Hague: Mouton de Gruyter. DOI:  http://doi.org/10.1515/9783110208603.3.231

Frascarelli, Mara & Hinterhölzl, Roland. 2007. Types of topics in German and Italian. In Schwabe, Kerstin & Winkler, Susanne (eds.), On information structure, meaning and form: Generalizations across languages, 87–116. Amsterdam: John Benjamins. DOI:  http://doi.org/10.1075/la.100.07fra

Frey, Werner. 2004. Notes on the syntax and the pragmatics of German left dislocation. In Lohnstein, Horst & Trissler, Susanne (eds.), The syntax and semantics of the left sentence periphery, 203–233. Berlin: De Gruyter Mouton. DOI:  http://doi.org/10.1515/9783110912111.203

Freywald, Ulrike & Cornips, Leonie & Ganuza, Natalia & Nistov, Ingvild & Opsahl, Toril. 2015. Beyond verb second – a matter of novel information-structural effects? Evidence from Norwegian, Swedish, German and Dutch. In Nortier, Jacomine & Svendsen, Bente A. (eds.), Language, youth and identity in the 21st century: Linguistic practices across urban spaces, 73–92. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9781139061896.006

Friedmann, Naama & Belletti, Adriana & Rizzi, Luigi. 2009. Relativized relatives: Types of intervention in the acquisition of A-bar dependencies. Lingua 119(1). 67–88. DOI:  http://doi.org/10.1016/j.lingua.2008.09.002

Friedmann, Naama & Belletti, Adriana & Rizzi, Luigi. 2021. Growing Trees: The acquisition of the left periphery. Glossa: a journal of general linguistics 6(1). 131. DOI:  http://doi.org/10.16995/glossa.5877

Friedmann, Naama & Reznick, Julia. 2021. Stages rather than ages in the acquisition of movement structures: Data from sentence repetition and 27696 spontaneous clauses. Glossa: a journal of general linguistics 39(1). 143. DOI:  http://doi.org/10.16995/glossa.5716

Gagliardi, Ann C. 2012. Input and intake in language acquisition. College Park University of Maryland dissertation.

Galasso, Joseph. 2003. The acquisition of functional categories: A case study. Indiana University: IUCL Publications.

Gavarró, Anna & Torrens, Vicent & Wexler, Kenn. 2010. Object clitic omission: Two language types. Language Acquisition 17(4). 192–219. DOI:  http://doi.org/10.1080/10489223.2010.509262

Gibb, Sophie & Hendry, Robin Findlay & Lancaster, Tom (eds.). 2019. The Routledge handbook of emergence. London: Routledge. DOI:  http://doi.org/10.4324/9781315714370

Giorgi, Alessandra & Pianesi, Fabio. 1997. Tense and aspect: From semantics to morphosyntax. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780195091922.001.0001

Giurgea, Ion & Remberger, Eva-Maria. 2016. Illocutionary force. In Ledgeway, Adam & Maiden, Martin (eds.), The Oxford guide to the Romance languages, 863–878. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/acprof:oso/9780199677108.003.0053

Grimshaw, Jane. 1991. Extended projection. Unpublished Ms., Brandeis University.

Grohmann, Kleanthes. 2000. Prolific peripheries: A radical view from the left. College Park University of Maryland dissertation.

Grohmann, Kleanthes K. 2003. Prolific domains: On the anti-locality of movement dependencies. Amsterdam: John Benjamins Publishing Company. DOI:  http://doi.org/10.1075/la.66

Guasti, Maria Teresa. 1993. Verb syntax in Italian child grammar: Finite and nonfinite verbs. Language Acquisition 3(1). 1–40. DOI:  http://doi.org/10.1207/s15327817la0301_1

Haegeman, Liliane & Greco, Ciro. 2016. Frame setters and the macro variation of subject-initial V2. Ms. https://ling.auf.net/lingbuzz/003226.

Haegeman, Liliane & Greco, Ciro. 2018. West Flemish V3 and the interaction of syntax and discourse. The Journal of Comparative Germanic Linguistics 21(1). 1–56. DOI:  http://doi.org/10.1007/s10828-018-9093-9

Haegeman, Liliane & Greco, Ciro. 2020. Frame setters and microvariation of subject-initial Verb Second. In Woods, Rebecca & Wolfe, Sam (eds.), Rethinking Verb Second, 61–89. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780198844303.003.0004

Haegeman, Liliane & Hill, Virginia. 2013. The syntacticization of discourse. In Folli, Raffaella & Sevdali, Christina & Truswell, Robert (eds.), Syntax and its limits, 370–390. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/acprof:oso/9780199683239.003.0018

Hall, Daniel Currie. 2007. The role and representation of contrast in phonological theory. University of Toronto dissertation.

Heim, Johannes. 2023. Acquisitional trajectories of North American vs British English tags. Talk presented at 2023 LAGB Annual Meeting (Anglia Ruskin University).

Heim, Johannes & Wiltschko, Martina. 2025. Rethinking structural growth: Insights from the acquisition of interactional language. Glossa: a journal of general linguistics 10(1). DOI:  http://doi.org/10.16995/glossa.16396

Hinzen, Wolfram & Wiltschko, Martina. 2023. Modelling non-specific linguistic variation in cognitive disorders. Journal of Linguistics 59(1). 61–87. DOI:  http://doi.org/10.1017/S0022226722000123

Hollebrandse, Bart & Roeper, Thomas. 1998. Point of view operators, features and a theory of barriers. In University of Pennsylvania linguistics colloquium.

Hsu, Brian. 2017. Verb second and its deviations: An argument for feature scattering in the left periphery. Glossa: a journal of general linguistics 2(1). 35. DOI:  http://doi.org/10.5334/gjgl.132

Hsu, Brian. 2021. Coalescence: A unification of bundling operations in syntax. Linguistic Inquiry 52(1). 39–87. DOI:  http://doi.org/10.5334/gjgl.132

Hyams, Nina. 1992. Morphosyntactic development in Italian and its relevance to parameter-setting models: Comments on the paper by Pizzuto and Casselli. Journal of Child Language 19(3). 695–709. DOI:  http://doi.org/10.1017/S0305000900011636

Hyams, Nina. 1994. Commentary: Null subjects in child language and the implications of cross-linguistic variation, Volume 2. In Lust, Barbara & Suñer, Margarita & Whitman, John (eds.), Syntactic theory and first language acquisition: Cross-linguistic perspectives, 287–300. Hillsdale, NJ: Erlbaum.

Jakubowicz, Celia. 2011. Measuring derivational complexity: New evidence from typically developing and SLI learners of L1 French. Lingua 121(3). 339–351. DOI:  http://doi.org/10.1016/j.lingua.2010.10.006

Klævik-Pettersen, Espen. 2019a. Inversion, V-to-C, and Verb Second: An investigation into the syntax and word order of Old French and Late Latin. University of Oslo dissertation.

Klævik-Pettersen, Espen. 2019b. What is Germanic and what is not about Old French V2. Linguistic Variation 19(1). 141–198. DOI:  http://doi.org/10.1075/lv.00011.kla

Kocher, Anna. 2022. Complementizers on edge: On the boundaries between syntax and pragmatics in Ibero-Romance (Open Romance Linguistics 1). Berlin: Language Science Press. DOI:  http://doi.org/10.15460/repohh/sub.2024100030

Krifka, Manfred. 2002. Be brief and vague! and how bidirectional optimality theory allows for verbosity and precision. In Restle, David & Zaefferer, Dietmar (eds.), Sounds and systems, 439–458. Berlin, New York: De Gruyter Mouton. DOI:  http://doi.org/10.1515/9783110894653.439

Krifka, Manfred. 2009. Approximate interpretations of number words: A case for strategic communication. In Hinrichs, Erhard W. & Nerbonne, John (eds.), Theory and evidence in semantics, 109–132. Stanford: CSLI Publications.

Kwisthout, Johan. 2013. Most inforbable explanations: Finding explanations in Bayesian networks that are both probable and informative. In van der Gaag, Linda C. (ed.), Symbolic and quantitative approaches to reasoning with uncertainty, 328–339. Berlin, Heidelberg: Springer Berlin Heidelberg. DOI:  http://doi.org/10.1007/978-3-642-39091-3_28

Kwisthout, Johan & Bekkering, Harold & van Rooij, Iris. 2017. To be precise, the details don’t matter: On predictive processing, precision, and level of detail of predictions. Brain and Cognition 112. 84–91. DOI:  http://doi.org/10.1016/j.bandc.2016.02.008

Kwisthout, Johan & van Rooij, Iris. 2020. Computational resource demands of a predictive Bayesian brain. Computational Brain & Behavior 3(2). 174–188. DOI:  http://doi.org/10.1007/s42113-019-00032-3

Laka, Miren Itziar. 1990. Negation in syntax: On the nature of functional categories and projections. Massachusetts Institute of Technology dissertation.

Larson, Richard K. 2021. Rethinking cartography. Language 97(2). 245–268. DOI:  http://doi.org/10.1353/lan.2021.0018

Lee, Naomi & Cournane, Ailis. 2019. The journey, not the endstate: Finding innovation in the dynamics of L1A. Talk presented at Diachronic Generative Syntax (DiGS) 2019 (Arizona State University).

Leivada, Evelina & Westergaard, Marit. 2019. Universal linguistic hierarchies are not innately wired: Evidence from multiple adjectives. PeerJ 7. e7438. DOI:  http://doi.org/10.7717/peerj.7438

Levelt, Clara & Schiller, Niels O. & Levelt, Willem J. 1999/2000. The acquisition of syllable types. Language Acquisition 8(3). 237–264. DOI:  http://doi.org/10.1207/S15327817LA0803_2

Lidz, Jeffrey & Gagliardi, Annie. 2015. How Nature Meets Nurture: Universal Grammar and Statistical Learning. Annual Review of Linguistics 1(1). 333–353. DOI:  http://doi.org/10.1146/annurev-linguist-030514-125236

Liu, Zixi & Muramatsu, Giulio Ciferri. 2025. Higher, but not later: Early acquisition of Japanese Sentence Final Particles. Paper presented at the 49th Annual Penn Linguistics Conference (PLC 49).

Llinàs-Grau, Mireia & Ojea, Ana Isabel. 2000. PhonBank Spanish Llinàs/Ojea Corpus.  http://doi.org/10.21415/t5k01q.

Lorkowski, Joe & Kreinovich, Vladik. 2015. Granularity helps explain seemingly irrational features of human decision making. In Pedrycz, Witol & Chen, Shyi-Ming (eds.), Granular computing and decision-making. studies in big data, vol. 10, 1–31. Cham: Springer. DOI:  http://doi.org/10.1007/978-3-319-16829-6_1

Lust, Barbara. 1999. Universal grammar: The strong continuity hypothesis in first language acquisition. In Bhatia, Tej K. & Ritchie, William C. (eds.), Handbook of child language acquisition, 111–155. San Diego, CA: Academic Press. DOI:  http://doi.org/10.1163/9789004653023_007

MacWhinney, Brian. 2000. The CHILDES project: Tools for analyzing talk. Mahwah, NJ: Erlbaum.

Madaro, Romano & Tomaselli, Alessandra & Bidese, Ermenegildo. 2025. Deriving CP-expansion in the German enclave varieties across Northeast Italy: Cracks in the bottleneck? In Harchaoui, Sarah & Modicom, Pierre-Yves (eds.), Verb-third phenomena in Germanic verb-second languages: Historical and variational perspectives (Open Germanic Linguistics), 291–344. Berlin: Language Science Press. DOI:  http://doi.org/10.5281/zenodo.17077198

Marušič, Franc & Mišmaš, Petra & Zaucer, Rok. 2019. Looking for cognitive foundations of functional sequences. Studies in Polish Linguistics 1. 53–75. DOI:  http://doi.org/10.4467/23005920SPL.19.006.10986. Special Volume DOI: https://doi.org/10.4467/23005920SPL.19.006.10986

Meelen, Marieke & Mourigh, Khalid & Cheng, Lisa Lai-Shen. 2020. V3 in urban youth varieties of Dutch. In Bárány, András & Biberauer, Theresa & Douglas, Jamie & Vikner, Sten (eds.), Syntactic architecture and its consequences II: Between syntax and morphology, 327–355. Berlin: Language Science Press. DOI:  http://doi.org/10.5281/zenodo.4280657

Meira, Miguel & Grolla, Elaine. 2022. The ‘Growing Trees’ approach and the development of the left periphery of the sentence in Brazilian Portuguese. Talk presented at Generative Approaches to Language Acquisition (GALA) 15 (Goethe Universität Frankfurt), September 22–24.

Meklenborg, Christine. 2020. Adverbial resumptive particles and Verb Second. In Woods, Rebecca & Wolfe, Sam (eds.), Rethinking Verb Second, 90–125. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780198844303.003.0005

Mervis, Carolyn B. & Crisafi, Maria A. 1982. Order of acquisition of subordinate-, basic-, and superordinate-level categories. Child Development 53(1). 258–266. DOI:  http://doi.org/10.1111/j.1467-8624.1982.tb01315.x

Miller, Max. 1979. The logic of language development in early childhood. Berlin: Springer-Verlag. DOI:  http://doi.org/10.1007/978-3-642-67408-2

Mišmaš, Petra & Marušič, Franc & Žaucer, Rok. 2018. Looking for an extralinguistic source of the strict order of adjectives. Paper presented at the A Workshop (Bled, Slovenia).

Mitrofanova, Natalia. 2018. Early Underspecification of Functional Categories: Evidence from the Acquisition of Locative PPs in Russian. Language Acquisition 25(4). 341–365. DOI:  http://doi.org/10.1080/10489223.2017.1344847

Montes, Rosa. 1987. Secuencias de clarificación en conversaciones con niños. Universidad Autónoma de Puebla.

Moscati, Vincenzo & Rizzi, Luigi. 2021. The layered syntactic structure of the complementizer system: Functional heads and multiple movements in the early left-periphery. a corpus study on Italian. Frontiers in Psychology 12. DOI:  http://doi.org/10.3389/fpsyg.2021.627841

Moscati, Vincezo. 2023. The observation of superiority on multiple movements to the Italian left-periphery: Intervention effects on nested dependencies and the role of information-structure features. Journal of Child Language 1–33. DOI:  http://doi.org/10.1017/S0305000923000508

Murphy, Gregory L. 2016. Explaining the basic-level concept advantage in infants…or is it the superordinate-level advantage? Psychology of Learning and Motivation 64. 57–92. DOI:  http://doi.org/10.1016/bs.plm.2015.09.002

Nie, Yining & Ross, Erica & Sayeed, Ollie & Vaux, Bert. 2019. The evolution of phonological scope. Unpublished Ms., University of Cambridge.

Oosthuizen, Helena & Southwood, Frenette. 2009. Methodological issues in the calculation of Mean Length of Utterance (MLU). South African Journal of Communication Disorders 56(1). DOI:  http://doi.org/10.4102/sajcd.v56i1.194

Paradis, J. & Genesee, F. 1997. On continuity and the emergence of functional categories in bilingual first-language acquisition. Language Acquisition 6(2). 91–124. DOI:  http://doi.org/10.1207/s15327817la0602_1

Perkins, Laurel & Lidz, Jeffrey. 2021. Eighteen-month-old infants represent nonlocal syntactic dependencies. PNAS 118(41). e2026469118. DOI:  http://doi.org/10.1073/pnas.2026469118

Pinker, Steven. 1984. Language learnability and language development. Cambridge: Cambridge University Press.

Poeppel, D. & Wexler, Ken. 1993. The full competence hypothesis of clause structure in early German. Language 69(1). 1–33. DOI:  http://doi.org/10.2307/416414

Poletto, Cecilia. 2002. The left-periphery of V2-Rhaetoromance dialects: A new view on V2 and V3. In Barbiers, Sjef & Cornips, Leonie & van der Kleij, Susanne (eds.), Syntactic microvariation, 214–242. Amsterdam: Meertens Institute.

Radford, Andrew. 1988. Small children’s small clauses. Transactions of the Philological Society 86(1). 1–43. DOI:  http://doi.org/10.1111/j.1467-968X.1988.tb00391.x

Radford, Andrew. 1990. Syntactic theory and the acquisition of English syntax: The nature of early child grammars of English. Oxford: Wiley Blackwell.

Ramchand, Gillian. 2023. The symbolic domain. Talk presented at V-NYI 6 Winter School, 23 January.

Ramchand, Gillian & Svenonius, Peter. 2014. Deriving the functional hierarchy. Language sciences 46. 152–174. DOI:  http://doi.org/10.1016/j.langsci.2014.06.013

Ramchand, Gillian Catriona. 2008. Verb meaning and the lexicon: A first phase syntax. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9780511486319

Ritter, Elizabeth & Wiltschko, Martina. 2014. The composition of INFL. Natural Language & Linguistic Theory 32(4). 1331–1386. DOI:  http://doi.org/10.1007/s11049-014-9248-6

Rizzi, Luigi. 1993. Some notes on linguistic theory and language development: The case of root infinitives. Language Acquisition 3(4). 371–393. DOI:  http://doi.org/10.1207/s15327817la0304_2

Rizzi, Luigi. 1997. The fine structure of the left periphery. In Haegeman, Liliane (ed.), Elements of grammar, 281–337. Dordrecht: Kluwer. DOI:  http://doi.org/10.1007/978-94-011-5420-8_7

Rizzi, Luigi. 2004. Locality and the left periphery. In Belletti, Adriana (ed.), Structures and beyond: The cartography of syntactic structures, vol. 3, 223–251. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780195171976.003.0008

Roberts, Ian. 2007/2021. Diachronic syntax. Oxford: Oxford University Press 2nd edn.

Roberts, Ian & Roussou, Anna. 2003. Syntactic change: A minimalist approach to grammaticalization. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9780511486326

Roeper, Tom & de Villiers, Jill. 2011. The acquisition path for wh-questions. In de Villiers, Jill & Roeper, Tom (eds.), Handbook of generative approaches to language acquisition, vol. 41, 189–246. Amsterdam: Springer. DOI:  http://doi.org/10.1007/978-94-007-1688-9_6

Ruigendijk, Esther & Friedmann, Naama. 2017. A deficit in movement-derived sentences in german-speaking hearing-impaired children. Frontiers in Psychology 8. DOI:  http://doi.org/10.3389/fpsyg.2017.00689

Rutar, Danaja & de Wolff, Erwin & van Rooij, Iris & Kwisthout, Johan. 2022a. Structure learning in predictive processing needs revision. Computational Brain & Behavior 5(2). 234–243. DOI:  http://doi.org/10.1007/s42113-022-00131-8

Rutar, Danaja & Wiese, Wanja & Kwisthout, Johan. 2022b. From representations in predictive processing to degrees of representational features. Minds and Machines 32(3). 461–484. DOI:  http://doi.org/10.1007/s11023-022-09599-6

Saikkonen, Taija. 2018. The development of Finnish syntax between the ages of 1 year, 8 months and 3 years. University of Helsinki dissertation.

Sanfelici, Emanuela & Gallina, Camilla. 2022. The timing of production: On the acquisition of Italian prepositions. Isogloss 8(2). 1–22. DOI:  http://doi.org/10.5565/rev/isogloss.137

Santelmann, Lynn. 2003. The acquisition of Swedish wh-questions. In Josefsson, Gunlög & Platzack, Christer & Håkansson, Gisela (eds.), The acquisition of Swedish grammar, 261–307. Amsterdam: John Benjamins. DOI:  http://doi.org/10.1075/lald.33.09san

Schaeffer, Jeannette. 2000. The acquisition of direct object scrambling and clitic placement. Amsterdam: John Benjamins. DOI:  http://doi.org/10.1075/lald.22

Schmerse, Daniel & Lieven, Elena & Tomasello, Michael. 2013. Error patterns in young German children’s wh-questions. Journal of Child Language 40(3). 656–671. DOI:  http://doi.org/10.1017/S0305000912000104

Schwartz, B. & Vikner, Sten. 1989. All verb-second clauses are CPs. Working Papers in Scandinavian Syntax 43. 27–49.

Schwartz, B. & Vikner, Sten. 1996. The verb always leaves IP in V2 clauses. In Belletti, Adriana & Rizzi, Luigi (eds.), Parameters and functionals heads, 11–62. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780195087932.003.0002

Scontras, Gregory & Degen, Judith & Goodman, Noah D. 2017. Subjectivity predicts adjective ordering preferences. Open Mind 1(1). 53–66. DOI:  http://doi.org/10.1162/OPMI_a_00005

Seidl, Amanda & Hollich, George & Jusczyk, Peter W. 2003. Early understanding of subject and object wh-questions. Infancy 4(3). 423–436. DOI:  http://doi.org/10.1207/S15327078IN0403_06

Serra, Miquel & Solé, Rosa. 1989. CHILDES Catalan Serra/Solé Corpus.  http://doi.org/10.21415/t58w3n.

Shirai, Junko & Shirai, Hidetoshi & Furuta, Yoshiteru. 2000. Acquisition of sentence-final particles in Japanese. In Perkins, Michael & Howard, Sara (eds.), New directions in language development and disorders, 243–250. Dordrecht: Springer. DOI:  http://doi.org/10.1007/978-1-4615-4157-8_23

Sluckin, Benjamin L. 2025. Revisiting the syntax and development of Kiezdeutsch V3: A new perspective. Journal of Germanic Linguistics 37(1). 64–124. DOI:  http://doi.org/10.1017/S1470542724000114

Smith, Ryan & Schwartenbeck, Philipp & Parr, Thomas & Friston, Karl J. 2020. An active inference approach to modeling structure learning: Concept learning as an example case. Frontiers in Computational Neuroscience 14. DOI:  http://doi.org/10.3389/fncom.2020.00041

Snyder, William. 2007. Child language: The parametric approach. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780199296699.001.0001

Snyder, William. 2021. A parametric approach to the acquisition of syntax. Journal of Child Language 48(5). 862–887. DOI:  http://doi.org/10.1017/S0305000921000465

Snyder, William & Luthra, Sahil & Hancock, Roeland. 2024. Why are subject-experiencer passives delayed? the Experiencer Externalization Hypothesis. In Proceedings of the 48th Boston University Conference on Language Development (BUCLD). 536–544. Cascadilla Proceedings Project.

Soares, Carla. 2003. The C-domain and the acquisition of European Portuguese: The case of wh-questions. Probus: International Journal of Romance Linguistics 15. 147–176. DOI:  http://doi.org/10.1515/prbs.2003.002

Soares, Catarina. 2006. La syntaxe de la périphérie gauche en portugais européen et son acquisition. Paris University of Paris 8 dissertation.

Song, Chenchen. 2019. On the formal flexibility of syntactic categories. University of Cambridge dissertation.

Speas, Peggy & Tenny, Carole. 2003. Configurational properties of point of view roles. In Di Sciullo, Anna Maria (ed.), Asymmetry in grammar, 315–344. Amsterdam: John Benjamins. DOI:  http://doi.org/10.1075/la.57.15spe

Stoicescu, Ioana. 2025. The illocutionary complementizer că in adult and child Romanian. In Sevcenco, Anca & Stoica, Irina & Stoicescu, Ioana & Tănase-Dogaru, Mihaela & Tigău, Alina & Tomescu, Veronica (eds.), Exploring linguistic landscapes: A Festschrift for Larisa Avram and Andrei Avram, 257–274. Bucharest: Bucharest University Press.

Stromswold, Karin J. 1990. Learnability and the acquisition of auxiliaries: The acquisition of verbal categories. Cambridge, MA Massachusetts Institute of Technology Ph.d. thesis.

Svenonius, Peter. 2006. The emergence of axial parts. Nordlyd 33. 49–77. DOI:  http://doi.org/10.7557/12.85

Svenonius, Peter. 2016. Emergent extended projections. Paper presented at MIT, 4 November.

Tsimpli, Ianthi Maria. 2005. Peripheral positions in early Greek. In Stavrou, Melita & Terzi, Arhonto (eds.), Advances in Greek generative syntax: In honor of Dimitra Theophanopoulou-Kontou, 179–216. Amsterdam: John Benjamins. DOI:  http://doi.org/10.1075/la.76.08tsi

Vainikka, Anne & Young-Scholten, Martha. 2011. The acquisition of German: Introducing Organic Grammar. Berlin: De Gruyter Mouton. DOI:  http://doi.org/10.1515/9783110263848

van Craenenbroeck, Jeroen & Haegeman, Liliane. 2007. The derivation of subject-initial V2. Linguistic Inquiry 38(1). 167–178. DOI:  http://doi.org/10.1162/ling.2007.38.1.167

van Kampen, Jacqueline. 2009. The non-biological evolution of grammar: Wh-question formation in Germanic. Biolinguistics 3(2–3). 154–185. DOI:  http://doi.org/10.5964/bioling.8719

van Kampen, Jacqueline. 2010. Typological guidance in the acquisition of V2 Dutch. Lingua 120(2). 264–283. DOI:  http://doi.org/10.1016/j.lingua.2008.07.005

van ’t Veer, Marijn. 2023. Feature co-occurrence constraints in acquisition. In Breit, Florian (ed.), Primitives of phonological structure, 86–107. Oxford: Oxford University Press. DOI:  http://doi.org/10.1093/oso/9780198791126.003.0004

van Urk, Coppe. 2015. A uniform syntax for phrasal movement: A case study of Dinka Bor. Massachusetts Institute of Technology, Department of Linguistics and Philosophy dissertation.

Verrips, Maaike & Weissenborn, Jürgen. 1992. Routes to verb placement in early German and French. The independence of finiteness and agreement. In Meisel, Jürgen (ed.), The acquisition of verb placement: Functional categories and V2 phenomena in language acquisition, 283–331. Dordrecht: Kluwer. DOI:  http://doi.org/10.1007/978-94-011-2803-2_10

Villa-García, Julio & González Rodríguez, Raquel. 2020. On the contrasts between sí ‘yes’ and sí que ‘yes that’ in Spanish and the structure of the Complementiser Phrase domain. Studies in Hispanic and Lusophone Linguistics 13(2). 451–484. DOI:  http://doi.org/10.1515/shll-2020-2037

Villa-García, Julio. 2012a. Recomplementation and locality of movement in Spanish. Probus: International Journal of Romance Linguistics 24(2). 257–314. DOI:  http://doi.org/10.1515/probus-2012-0011

Villa-García, Julio. 2012b. Spanish subjects can be subjects: Acquisitional and empirical evidence. IBERIA: An International Journal of Theoretical Linguistics 4(2). 124–169.

Villa-García, Julio & Ott, Dennis. 2024. Recomplementation as a paratactic phenomenon: Evidence from Spanish and English. Journal of Linguistics 60(1). 213–244. DOI:  http://doi.org/10.1017/S0022226722000433

Villa-García, Julio & Snyder, William. 2009. On the acquisition of overt subjects, topics and wh-questions in Spanish. In Language acquisition and development: Proceedings of GALA 2009. 493–504.

Walkden, George. 2017. Language contact and V3 in Germanic varieties new and old. The Journal of Comparative Germanic Linguistics 20. 49–81. DOI:  http://doi.org/10.1007/s10828-017-9084-2

Wang, Guoyin. 2017. DGCC: data-driven granular cognitive computing. Granular Computing 2(4). 343–355. DOI:  http://doi.org/10.1007/s41066-017-0048-3

Wang, Guoyin & Yang, Jie & Xu, Ji. 2017. Granular computing: from granularity optimization to multi-granularity joint problem solving. Granular Computing 2(3). 105–120. DOI:  http://doi.org/10.1007/s41066-016-0032-3

Ward, Emma K. & Rutar, Danaja & Zaadnoordijk, Lorijn & Poli, Francesco & Hunnius, Sabine. 2023. Beyond the adult mind: A developmental framework for predictive processing in infancy. DOI:  http://doi.org/10.31234/osf.io/wktz8

Westergaard, Marit. 2009. The Acquisition of Word Order. Amsterdam: John Benjamins. DOI:  http://doi.org/10.1075/la.145

Wexler, Kenn & Gavarró, Anna & Torrens, Vicent. 2004. Feature checking and object clitic omission in child Catalan and Spanish. In Bok-Bennema, Reineke & Hollebrandse, Bart & Kampers-Manhe, Brigitte & Sleeman, Petra (eds.), Romance languages and linguistic theory 2002, 253–270. Amsterdam: John Benjamins. DOI:  http://doi.org/10.1075/cilt.256.15wex

Wiltschko, Martina. 2014. The universal structure of categories: Towards a formal typology. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/CBO9781139833899

Wiltschko, Martina. 2021. The grammar of interactional language. Cambridge: Cambridge University Press. DOI:  http://doi.org/10.1017/9781108693707

Wolfe, Sam. 2015. The nature of Old Spanish verb second reconsidered. Lingua 164. 132–155. DOI:  http://doi.org/10.1016/j.lingua.2015.06.007

Xu, Fei & Tenenbaum, Joshua B. 2007. Word learning as Bayesian inference. Psychological Review 114(2). 245. DOI:  http://doi.org/10.1037/0033-295X.114.2.245

Zadeh, Lofti A. 1997. Toward a theory of fuzzy information granulation and its centrality in human reasoning and fuzzy logic. Fuzzy Sets and Systems 90. 111–127. DOI:  http://doi.org/10.1016/S0165-0114(97)00077-8

Zwart, Jan-Wouter. 1993. Dutch syntax: A Minimalist approach. University of Groningen dissertation.

Zwart, Jan-Wouter. 1997. Morphosyntax of verb movement: A minimalist approach to the syntax of Dutch. Dordrecht: Springer. DOI:  http://doi.org/10.1007/978-94-011-5880-0