Performance and competence intertwined: A computational model of the Null Subject stage in English-speaking children

Soumik Dey, William Gregory Sakas · 2025

Read Performance and competence intertwined: A computational model of the Null Subject stage in English-speaking children by Soumik Dey, William Gregory Sakas (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-25545.txt.

Introduction

Soumik Dey and William Gregory Sakas build a computer model of a stage where young children drop sentence subjects. Their 2025 paper covers English-speaking children up to about age four. Children at this stage often say verb phrases with no subject, which clashes with adult grammar rules.

Earlier work argued that children confuse command forms, which need no subject, with plain statements. That mix-up may push them toward a grammar that allows dropped subjects for a time. Dey and Sakas add a new measure for this mix-up to a known learning model. Their simulations back the earlier claim. The study offers a way to join computer models of grammar learning with other known factors in how children grow.

Catalog record

CREATORSoumik Dey, William Gregory Sakas
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE45 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-25545

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Performance and competence intertwined: A computational model of the Null Subject stage in English-speaking children Soumik Dey, William Gregory Sakas Abstract The empirically established null subject (NS) stage, lasting until about 4 years of age, involves frequent omission of subjects by children. Orfitelli and Hyams 2012 observe that young English speakers often confuse imperative NS utterances with declarative ones due to performance influences, promoting a temporary null subject grammar. We propose a new computational parameter to measure this misinterpretation and incorporate it into a simulated model of obligatory subject grammar learning. Using a modified version of the Variational Learner Yang 2012 which works for superset-subset languages, our simulations support Orfitelli and Hyams’ hypothesis. More generally, this study outlines a framework for integrating computational models in the study of grammatical acquisition alongside other key developmental factors. 1 Introduction The Null Subject ( NS ) stage is a well-researched phenomenon in child language acquisition, characterized by young children sometimes forming declarative sentences without subjects. This is expected in children exposed to null subject languages but contentious in obligatory subject language environments. The NS stage challenges the Subset Principle Gold 1967 ; Berwick 1985 ; Manzini and Wexler 1987 ; Valian 1990 ; Déprez and Pierce 1993 ; Fodor and Sakas 2005 — children learning obligatory subject languages exhibit NS-like sentences (a superset language), which gradually shift to non-NS (subset language) with time. This phenomenon puzzles learning theorists. Explanations vary, with some attributing the NS stage to differences between children’s internal grammar and adult target grammars Yang 2012 ; Orfitelli and Hyams 2008 ; Valian 1990 , while others cite extrasyntactic factors like memory and processing constraints Bloom 1970 ; Bloom 1990 ; Valian 1991 ; Wang et al. 1992 . Rizzi 2005a ; Rizzi 2005b connects a performance account of a limited production system with its consequence of the varying grammatical competence we see in children. In this paper, we model the grammatical theory of the NS stage in children using a developmental parameter and the Variational Learner (VL) Yang 2012 , a well-known computational model of language acquisition. More generally, this study outlines a framework for integrating computational models in the study of language acquisition alongside other key developmental factors. 2 Background 2.1 Orfitelli & Hyams (2012) Experiment 2 The two distinct theories of performance and grammatical competence present distinct explanations for children’s comprehension of subject-lacking sentences ( NS sentences , such as imperatives in English). Grammatical theories propose that young English speakers view NS sentences akin to grammatically correct declaratives, similar to adults in null subject languages. Conversely, performance theories attribute omissions to production limitations, suggesting children interpret NS sentences as adults do in obligatory subject languages, limiting English-speaking children’s interpretations to imperatives or diary forms. To explore this, Orfitelli and Hyams 2012 used a truth-value judgment (TVJ) experiment Crain and McKee 1985 ; Crain and Fodor 1993 . In Experiment 2, a child watched a narrative, then listened to a puppet’s (Mr. Bear) comment, and judged the comments’ accuracy relative to the story. The child corrected Mr. Bear by indicating the correctness of his statements, with explanations. Thirty children from Los Angeles daycare centers were involved, divided into three age groups (2;6-2;11, 3;0-3;5, and 3;6-3;11) to represent early, middle, and late NS stages. Details on age range and distribution are in Table 1 , adapted from Orfitelli and Hyams 2012 . Group Age Range Mean Age N 2;6-2;11 2.54-2.96 2.73 10 3;0-3;5 3.12-3.48 3.3 10 3;6-3;11 3.64-3.98 3.82 10 Total 2.54-3.98 3.28 30 Table 1: Orfitelli and Hyams 2012 [Experiment 2] participant details. The children underwent assessment on 24 grammatical items (sentences), equally split between correct and incorrect true/false responses. There were 8 NS condition sentences, while the remaining 16 items were evenly divided among the remaining four conditions. Orfitelli & Hyams (O&H) classified the children’s responses to NS condition sentences into three categories based on interpretation: • Consistently imperative : 7-8 out of 8 NS sentences interpreted as imperative. • Both interpretations allowed : 2-6 out of 8 NS sentences interpreted as imperative. • Consistently declarative : 0-1 out of 8 NS sentences interpreted as imperative. 2;6-2;11 3;0-3;5 3;6-3;11 Imperative (7-8 imp) 0% (0) 40% (4) 80% (8) Both (2-6 imp.) 80% (8) 60% (6) 20% (2) Declarative (0-1 imp.) 20% (2) 0% (0) 0% (0) Table 2: Individual performance on the NS condition sentences in Orfitelli and Hyams 2012 [Table 8]. Additional details regarding the performance of children on the task are illustrated in Figure 1 . The performance on the NS items varied with age as O&H reported that the youngest group assigned an imperative (adult) interpretation to NS items 40% of the time on average, while the middle age group assigned an imperative interpretation 64% of the time. While O&H do not report an average number for the performance on NS items for the oldest age group, from Figure 1 we can estimate the average performance to be close to 90%. For concreteness, we adopt 90% for this age group from this point on. Figure 1: Performance on NS condition sentences from Orfitelli and Hyams 2012 [Figure 5]. The fact that in O&H’s study children comprehend NS sentences differently than an adult, reinforces the grammatical account of the NS stage. However, O&H also argue for performance limitations which create the illocutionary force ambiguity associated with the imperative NS sentences resulting in the NS stage. 2.2 Language Acquisition in P&P Framework The study of language acquisition presents an extraordinary challenge for scientific inquiry. It requires that a child, over a remarkably short period, must develop a grasp of a grammar system capable of producing and interpreting a set of utterances comparable to those produced by adults within their linguistic surroundings. 1 1 1 Or nearly identical, encompassing microvariations within linguistic communities. The child’s cognitive mechanisms for language learning achieve this despite having limited or no exposure to sentence-level linguistic phenomena, and without the capacity to perceive intrinsic properties of the latent structures that generate the surface forms of utterances (see Fodor 1998 for reference). This restricted interaction with sufficient surface forms forms the core of the argument known as the poverty of the stimulus Chomsky 1981 ; Chomsky 1955 ; Chomsky 1965 ; Chomsky 1986 . This argument has been instrumental in advocating for an intrinsic language faculty that imparts universal structural principles (such as the notion that all languages possess subjects) and parameters, which dictate language-specific structural traits (e.g., whether SpecIP is initial or final) that are adjusted during the language acquisition process.,The framework of principles and parameters (P&P), as introduced by Chomsky 1981 , was designed to streamline the language learning process by: • Limiting the potential scope of grammatical possibilities, transitioning linguistic theory from a potentially limitless universe of human grammars to a explicitly finite set • Simplifying complex structural phenomena into parameter values, which vary between languages. This framework comprises a set of foundational principles that “sharply restrict the class of attainable grammars and narrowly constrain their form, but with parameter [values] that have to be fixed by experience” Chomsky 1981 . Essentially, the child is inherently equipped with these principles, while parameter values are influenced by the linguistic input they encounter

Read the complete text