Reaffirming the poverty of the stimulus argument: a reply to the replies

Lidz, Waxman, and Freedman [Lidz, J., Waxman, S., & Freedman, J. (2003). What infants know about syntax but couldn’t have learned: Evidence for syntactic structure at 18-months. Cognition, 89, B65-B73.] argue that acquisition of the syntactic and semantic properties of anaphoric one in English relies on innate knowledge within the learner. Several commentaries have now been published questioning this finding. We defend the original finding by identifying both empirical and logical flaws in the critiques.     

Structural packaging in the input to language learning: Contributions of prosodic and morphological marking of phrases to the acquisition of language

The linguistic input to language learning is usually thought to consist of simple strings of words. We argue that input must also include information about how words group into syntactic phrases. Natural languages regularly incorporate correlated cues to phrase structure, such as prosody, function words, and concord morphology. The claim that such cues are necessary for successful acquisition of syntax was tested in a series of miniature language learning experiments with adult subjects. In each experiment, when input included some cue marking the phrase structure of sentences, subjects were entirely successful in learning syntax; in contrast, when input lacked such a cue (but was otherwise identical), subjects failed to learn significant portions of syntax. Cues to phrase structure appear to facilitate learning by indicating to the learner those domains within which distributional analyses may be most efficiently pursued, thereby reducing the amount and complexity of required input data. More complex target systems place greater premiums on efficient analysis; hence, such cues may be even more crucial for acquisition of natural language syntax. We suggest that the finding that phrase structure cues are a necessary aspect of language input reflects the limited capacities of human language learners; languages may incorporate structural cues in part to circumvent such limitations and ensure successful acquisition.     

Précis of Foundations of language: brain, meaning, grammar, evolution

The goal of this study is to reintegrate the theory of generative grammar into the cognitive sciences. Generative grammar was right to focus on the child's acquisition of language as its central problem, leading to the hypothesis of an innate Universal Grammar. However, generative grammar was mistaken in assuming that the syntactic component is the sole course of combinatoriality, and that everything else is "interpretive." The proper approach is a parallel architecture, in which phonology, syntax, and semantics are autonomous generative systems linked by interface components. The parallel architecture leads to an integration within linguistics, and to a far better integration with the rest of cognitive neuroscience. It fits naturally into the larger architecture of the mind/brain and permits a properly mentalistic theory of semantics. It results in a view of linguistic performance in which the rules of grammar are directly involved in processing. Finally, it leads to a natural account of the incremental evolution of the language capacity.     

Learning the unlearnable: the role of missing evidence

Syntactic knowledge is widely held to be partially innate, rather than learned. In a classic example, it is sometimes argued that children know the proper use of anaphoric one, although that knowledge could not have been learned from experience. Lidz et al. [Lidz, J., Waxman, S., & Freedman, J. (2003). What infants know about syntax but couldn't have learned: Experimental evidence for syntactic structure at 18 months. Cognition, 89, B65-B73.] pursue this argument, and present corpus and experimental evidence that appears to support it; they conclude that specific aspects of this knowledge must be innate. We demonstrate, contra Lidz et al., that this knowledge may in fact be acquired from the input, through a simple Bayesian learning procedure. The learning procedure succeeds because it is sensitive to the absence of particular input patterns--an aspect of learning that is apparently overlooked by Lidz et al. More generally, we suggest that a prominent form of the "argument from poverty of the stimulus" suffers from the same oversight, and is as a result logically unsound.     

The learnability of abstract syntactic principles

Children acquiring language infer the correct form of syntactic constructions for which they appear to have little or no direct evidence, avoiding simple but incorrect generalizations that would be consistent with the data they receive. These generalizations must be guided by some inductive bias – some abstract knowledge – that leads them to prefer the correct hypotheses even in the absence of directly supporting evidence. What form do these inductive constraints take? It is often argued or assumed that they reflect innately specified knowledge of language. A classic example of such an argument moves from the phenomenon of auxiliary fronting in English interrogatives to the conclusion that children must innately know that syntactic rules are defined over hierarchical phrase structures rather than linear sequences of words (e.g., [Chomsky, 1965] , [Chomsky, 1971] , [Chomsky, 1980] and [Crain and Nakayama, 1987] ). Here we use a Bayesian framework for grammar induction to address a version of this argument and show that, given typical child-directed speech and certain innate domain-general capacities, an ideal learner could recognize the hierarchical phrase structure of language without having this knowledge innately specified as part of the language faculty. We discuss the implications of this analysis for accounts of human language acquisition.     

Language input and child syntax

Existing work on the acquisition of syntax has been concerned mainly with the early stages of syntactic development. In the present study we examine later syntactic development in children. Also, existing work has focused on commonalities in the emergence of syntax. Here we explore individual differences among children and their relation to variations in language input. In Study 1 we find substantial individual differences in children's mastery of multiclause sentences and a significant relation between those differences and the proportion of multiclause sentences in parent speech. We also find individual differences in the number of noun phrases in children's utterances and a significant relation between those differences and the number of noun phrases in parent speech. In Study 2 we find greater syntactic growth over a year of preschool in classes where teachers' speech is more syntactically complex. The implications of our findings for the understanding of the sources of syntactic development are discussed.     

A computational learning model for metrical phonology

One of the major challenges to linguistic theory is the solution of what has been termed the "projection problem". Simply put, linguistics must account for the fact that starting from a data base that is both unsystematic and relatively small, a human child is capable of constructing a grammar that mirrors, for all intents and purposes, the adult system. In this article we shall address ourselves to the question of the learnability of a postulated subsystem of phonological structure: the stress system. We shall describe a computer program which is designed to acquire this subpart of linguistic structure. Our approach follows the "principles and parameters" model of Chomsky (1981a, b). This model is particularly interesting from both a computational point of view and with respect to the development of learning theories. We encode the relevant aspects of universal grammar (UG)--those aspects of linguistic structure that are presumed innate and thus present in every linguistic system. The learning process consists of fixing a number of parameters which have been shown to underlie stress systems and which should, in principle, lead the learner to the postulation of the system from which the primary linguistic data (i.e., the input to the learner) is drawn. We go on to explore certain formal and substantive properties of this learning system. Questions such as cross-parameter dependencies, determinism, subsets, and incremental versus all-at-once learning are raised and discussed in the article. The issues raised by this study provide another perspective on the formal structure of stress systems and the learnability of parameter systems in general.     

Bootstrapping in miniature language acquisition

Given the difficulties in learning meanings of words by observing the referent, it has been suggested that children use the syntactic context of the word to predict part of its meaning, a hypothesis known as syntactic bootstrapping. Semantic bootstrapping is the opposite theory that the knowledge of semantics helps in acquiring syntax. While there is evidence that children can apply their knowledge of correlations between syntax and semantics to perform bootstrapping, it is not clear how they come to know about these correlations in the first place. Here, a connectionist network is presented that learns to comprehend a miniature language by associating sentences with the corresponding scenes. In doing so, it learns the syntactic/semantic correlations and exhibits bootstrapping behavior. It is argued that such specialized phenomena can emerge when general mechanisms are applied to a specific task, and it is not always necessary to endow the learner with pre-existing specialized mechanisms.     

Discovering syntactic deep structure via Bayesian statistics

In the Bayesian framework, a language learner should seek a grammar that explains observed data well and is also a priori probable. This paper proposes such a measure of prior probability. Indeed it develops a full statistical framework for lexicalized syntax. The learner's job is to discover the system of probabilistic transformations (often called lexical redundancy rules) that underlies the patterns of regular and irregular syntactic constructions listed in the lexicon. Specifically, the learner discovers what transformations apply in the language, how often they apply, and in what contexts. It considers simpler systems of transformations to be more probable a priori. Experiments show that the learned transformations are more effective than previous statistical models at predicting the probabilities of lexical entries, especially those for which the learner had no direct evidence.     

Implicit statistical learning is directly associated with the acquisition of syntax

This article reports on an individual differences study that investigated the role of implicit statistical learning in the acquisition of syntax in children. One hundred children ages 4 years 5 months through 6 years 11 months completed a test of implicit statistical learning, a test of explicit declarative learning, and standardized tests of verbal and nonverbal ability. They also completed a syntactic priming task, which provided a dynamic index of children's facility to detect and respond to changes in the input frequency of linguistic structure. The results showed that implicit statistical learning ability was directly associated with the long-term maintenance of the primed structure. The results constitute the first empirical demonstration of a direct association between implicit statistical learning and syntactic acquisition in children.     

What language says about the psychology of events

Despite the variety of verb meanings, linguistic research on their syntax and semantics has shown that they can be categorized into a finite and surprisingly small number of event types. More recently, research in the psycholinguistics of language acquisition and processing has emphasized the relevance of event type. The wider implication of these findings is that the conceptual fluidity of verbal concepts is confined by the fundamental structures of mental grammar, shedding light on this important interface between cognition and syntactic organization.     

句子理解的关键——对句法和语义关系的再探讨

对于句子理解中句法和语义的关系,传统的模块化观点认为,句法信息优先于语义信息,句法加工支配着句子的理解过程。近期,这种强调句法优先的理论受到了一系列ERP（Event-related potentials）研究结果的挑战,这些研究从不同角度论证了语义信息对句法加工及句子整合加工的重要作用。基于这些研究发现,研究者提出句子理解过程中句法和语义的关系并非一成不变：一些情况下可能是句法支配语义加工,而另一些情况下则可能是语义引导句法加工     

Automatic evaluation of syntactic learners in typologically-different languages

Human syntax acquisition involves a system that can learn constraints on possible word sequences in typologically-different human languages. Evaluation of computational syntax acquisition systems typically involves theory-specific or language-specific assumptions that make it difficult to compare results in multiple languages. To address this problem, a bag-of-words incremental generation (BIG) task with an automatic sentence prediction accuracy (SPA) evaluation measure was developed. The BIG–SPA task was used to test several learners that incorporated n-gram statistics which are commonly found in statistical approaches to syntax acquisition. In addition, a novel Adjacency–Prominence learner, that was based on psycholinguistic work in sentence production and syntax acquisition, was also tested and it was found that this learner yielded the best results in this task on these languages. In general, the BIG–SPA task is argued to be a useful platform for comparing explicit theories of syntax acquisition in multiple languages.     

What artificial grammar learning reveals about the neurobiology of syntax

In this paper we examine the neurobiological correlates of syntax, the processing of structured sequences, by comparing FMRI results on artificial and natural language syntax. We discuss these and similar findings in the context of formal language and computability theory. We used a simple right-linear unification grammar in an implicit artificial grammar learning paradigm in 32 healthy Dutch university students (natural language FMRI data were already acquired for these participants). We predicted that artificial syntax processing would engage the left inferior frontal region (BA 44/45) and that this activation would overlap with syntax-related variability observed in the natural language experiment. The main findings of this study show that the left inferior frontal region centered on BA 44/45 is active during artificial syntax processing of well-formed (grammatical) sequence independent of local subsequence familiarity. The same region is engaged to a greater extent when a syntactic violation is present and structural unification becomes difficult or impossible. The effects related to artificial syntax in the left inferior frontal region (BA 44/45) were essentially identical when we masked these with activity related to natural syntax in the same subjects. Finally, the medial temporal lobe was deactivated during this operation, consistent with the view that implicit processing does not rely on declarative memory mechanisms that engage the medial temporal lobe. In the context of recent FMRI findings, we raise the question whether Broca’s region (or subregions) is specifically related to syntactic movement operations or the processing of hierarchically nested non-adjacent dependencies in the discussion section. We conclude that this is not the case. Instead, we argue that the left inferior frontal region is a generic on-line sequence processor that unifies information from various sources in an incremental and recursive manner, independent of whether there are any processing requirements related to syntactic movement or hierarchically nested structures. In addition, we argue that the Chomsky hierarchy is not directly relevant for neurobiological systems.     

Mapping sensorimotor sequences to word sequences: A connectionist model of language acquisition and sentence generation

In this article we present a neural network model of sentence generation. The network has both technical and conceptual innovations. Its main technical novelty is in its semantic representations: the messages which form the input to the network are structured as sequences, so that message elements are delivered to the network one at a time. Rather than learning to linearise a static semantic representation as a sequence of words, our network rehearses a sequence of semantic signals, and learns to generate words from selected signals. Conceptually, the network's use of rehearsed sequences of semantic signals is motivated by work in embodied cognition, which posits that the structure of semantic representations has its origin in the serial structure of sensorimotor processing. The rich sequential structure of the network's semantic inputs also allows it to incorporate certain Chomskyan ideas about innate syntactic knowledge and parameter-setting, as well as a more empiricist account of the acquisition of idiomatic syntactic constructions.     

What Can Bilingual Children Tell Us About the Developmental Relationship Between Vocabulary and Grammar?

Strong correlations between vocabulary and grammar are well attested in language development in monolingual and bilingual children. What is less clear is whether there is any directionality in the relationship between the two constructs, whether it is predictive over time, and the extent to which it is affected by language input. In the present study, we analyzed data from 100 bilingual children with English as an additional language who were tested on measures of vocabulary breadth and depth, morphology, and syntax at three time points at 6-month intervals from the age of 5;8. We used bivariate growth models to test the directionality of the relationship between vocabulary breadth and depth, and measures of morphology and syntax; testing bilingual children allowed us to use measures of English input as covariates in the analyses. All the models showed a correlation between vocabulary and grammar, but no correlation between their growth slopes, suggesting that vocabulary and grammar grow independently. Three of the four bivariate models showed a significant correlation between the intercept of grammar skills and the slope of vocabulary growth. Length of exposure to English predicted the intercept of vocabulary breadth and grammar, suggesting that children exposed to English earlier had larger vocabularies and better morpho-syntactic skills. Current English input predicted the intercept of both measures of vocabulary as well as the slope for vocabulary depth, the only measure for which there was a significant relationship between intercept and slope, suggesting a Matthew effect for this dimension of vocabulary. All materials, data, and code are available at https://osf.io/x3wht/ . Research highlights

• Vocabulary breadth and morphological and syntactic skills increased linearly for all participants, without any difference between lower and higher achieving children.
• Vocabulary depth grew more over time for those children with deeper vocabulary knowledge and higher levels of current English input at the start of the study.
• All of the bivariate growth models showed a correlation between vocabulary and grammar, but failed to show any correlation between their growth.
• Significant relationships between the intercept of grammar and the growth of vocabulary showed steeper lexical growth in children with better grammar skills.
• Length of exposure to English had an effect on morphological and syntactic skills, while only current English input had an effect on vocabulary depth.

     

Parsing to Learn

Learning a language by parameter setting is almost certainly less onerous than composing a grammar from scratch. But recent computational modeling of how parameters are set has shown that it is not at all the simple mechanical process sometimes imagined. Sentences must be parsed to discover the properties that select between parameter values. But the sentences that drive learning cannot be parsed with the learner's current grammar. And there is not much point in parsing them with just one new grammar. They must apparently be parsed with all possible grammars, in order to find out which one is most successful at licensing the language. The research task is to reconcile this with the fact that the human sentence parsing mechanism, even in adults, has only very limited parallel parsing capacity. I have proposed that all possible grammars can be folded into one, if parameter values are fragments of sentential tree structures that the parser can make use of where necessary to assign a structure to an input sentence. However, the problem of capacity limitations remains. The combined grammar will afford multiple analyses for some sentences, too many to be computed on-line. I propose that the parser computes only one analysis per sentence but can detect ambiguity, and that the learner makes use of unambiguous input only. This provides secure information but relatively little of it, particularly at early stages of learning where few grammars have been excluded and ambiguity is rife. I consider three solutions: improving the parser's ability to extract unambiguous information from partially ambiguous sentences, assuming default parameter values to temporarily eliminate ambiguity, reconfiguring the parameters so that some are subordinate to others and do not present themselves to the learner until the others have been set. A more radical alternative is to give up the quest for error-free learning and permit parameters to be set without regard for whether the parser may have overlooked an alternative analysis of the sentence. If it can be assumed that the human parser keeps a running tally of the parameter values it has accessed, then the learner would do nothing other than parse sentences for comprehension, as adults do. The most useful parameter values would become more and more easily accessed; the noncontributors would drop out of the running. There would be no learning mechanism at all, over and above the parser. But how accurate this system would be remains to be established.     

Yes, we still need Universal Grammar

In a recent paper [Lidz, J., Gleitman, H., & Gleitman, L. (2003). Understanding how input matters: Verb learning and the footprint of universal grammar. Cognition, 87, 151-178], we provided cross-linguistic evidence in favor of the following linked assertions: (i) Verb argument structure is a correlate of verb meaning; (ii) However, argument structure is not directly available to learners as a cue for reconstructing verb meaning, owing to the complexity of form-meaning mappings within and across languages; (iii) A major correlate of argument-structure, namely, noun phrase number, is statistically available on the surface in all languages, and serves as a quasi-universal derivative cue to the meanings of verbs; (iv) this cue is privileged, in the sense that it is used early and selectively by learners, despite within- and cross-language differences in its availability. Goldberg [Goldberg, A. (2004). But do we need Universal Grammar? Comment on Lidz, Gleitman and Gleitman 2003. Cognition] suggests that this cue is not linguistic, that it is too sicklied o'er with exceptions and provisos to be useful to learners, and that conversational conspiracies can in any case serve as the alternative theoretical framework for a theory of predicate learning, and language acquisition more generally. In the present reply, we review and further explicate our original position, to wit: A large part of any generative grammar is a formal statement of the complex alignments between predicate-argument structures and the surface forms (linear strings of words) of sentences. Because the several rules for alignment interact, the surface outcomes reveal individual systematicities only abstractly. Therefore, learning would be impossible if infants could not analyze probabilistically available patterns to recover their principled linguistic sources. This statistics-based discovery procedure is in certain relevant regards specific to language learning. Finally, we argue that while pragmatics and theory of mind properties in learner and tutor necessarily frame language acquisition, that these have not been shown—and probably cannot be shown—to be sufficient to this computational problem.     

The P600 in Implicit Artificial Grammar Learning

The suitability of the artificial grammar learning (AGL) paradigm to capture relevant aspects of the acquisition of linguistic structures has been empirically tested in a number of EEG studies. Some have shown a syntax‐related P600 component, but it has not been ruled out that the AGL P600 effect is a response to surface features (e.g., subsequence familiarity) rather than the underlying syntax structure. Therefore, in this study, we controlled for the surface characteristics of the test sequences (associative chunk strength) and recorded the EEG before (baseline preference classification) and after (preference and grammaticality classification) exposure to a grammar. After exposure, a typical, centroparietal P600 effect was elicited by grammatical violations and not by unfamiliar subsequences, suggesting that the AGL P600 effect signals a response to structural irregularities. Moreover, preference and grammaticality classification showed a qualitatively similar ERP profile, strengthening the idea that the implicit structural mere‐exposure paradigm in combination with preference classification is a suitable alternative to the traditional grammaticality classification test.     

Syntax as a reflex: neurophysiological evidence for early automaticity of grammatical processing

It has been a matter of debate whether the specifically human capacity to process syntactic information draws on attentional resources or is automatic. To address this issue, we recorded neurophysiological indicators of syntactic processing to spoken sentences while subjects were distracted to different degrees from language processing. Subjects were either passively distracted, by watching a silent video film, or their attention was actively streamed away from the language input by performing a demanding acoustic signal detection task. An early index of syntactic violations, the syntactic Mismatch Negativity (sMMN), distinguished between grammatical and ungrammatical speech even under strongest distraction. The magnitude of the early sMMN (at <150ms) was unaffected by attention load of the distraction task. The independence of the early syntactic brain response of attentional distraction provides neurophysiological evidence for the automaticity of syntax and for its autonomy from other attention-demanding processes, including acoustic stimulus discrimination. The first attentional modulation of syntactic brain responses became manifest at a later stage, at approximately 200ms, thus demonstrating the narrowness of the early time window of syntactic autonomy. We discuss these results in the light of modular and interactive theories of cognitive processing and draw inferences on the automaticity of both the cognitive MMN response and certain grammar processes in general.