Since the brain manipulates probabilities (I will argue) then it should do so according to Bayes’ Law. After all, it is normative, and Darwin would not expect us to do anything less. Furthermore there is a lot to be learned from taking Bayes seriously. I consider myself a nativist despite my statistical bent, and it tells me how to combine an informative prior with the evidence of our senses – compute the likelihood of the evidence. It then tells us that this likelihood must be a very broad generative model of everything we encounter. Lastly, since Bayes says nothing about how to do any of this, I presume that the computational methods themselves are not learned, they must be innate, and I will argue there seems to be very few options on how this can be done, with something like particle filtering being one of the few. I will illustrate these ideas with work in computational linguistics, both my own and that of others.
All adults seem to have amnesia about much that happened in their childhood. Does early memory simply wither away through massive synaptic pruning and cell death in early brain development? Or, is it just masked by interference from later experience? This talk explores these questions in the specific case of childhood language memory. Research into the re-learning of long-disused childhood languages turns out to have much to offer. It provides relatively objective evidence for access to early childhood memory in adulthood via re-learning. It complements linguistic deprivation research to highlight the special status of childhood language experience in phonology and morphosyntax acquisition. It thereby suggests a strategy to salvage seemingly forgotten childhood languages, which are often also heritage languages. Equally importantly, re-learning childhood languages may well open a window onto how language affects cognitive development not only during, but also well beyond, the childhood years.
An Integrative Approach to Understanding the Neuroscience of Language
Language serves a specialized purpose: to translate thoughts to sound (or sign) and back again. The complexity and relative uniqueness of linguistic knowledge reflects this specialization. But language evolved in the context of a brain that was already performing functions that are broadly important for language: perceiving, acting, remembering, learning. From an evolutionary standpoint, then, we should expect to find some architectural and computational parallels between linguistic and non-linguistic neural systems. Our work has indeed uncovered such parallels. Language processes are organized into two broad neural streams—a ventral auditory-conceptual stream and a dorsal auditory-motor stream—functionally analogous to that found in vision. And the dorsal auditory-motor language stream uses computational principles found in motor-control more broadly. This approach to understanding the neural basis of language does not replace traditional linguistic constructs but integrates them into a broader neuro-evolutionary context and provides a richer, comparative source of data.
Hundreds of millions of people live and work while using a language that is not their native tongue. Given that using a foreign language is more difficult than using a native tongue, one would expect an overall deleterious effect on their mental and physical performance. We have discovered that the opposite is often true. We argue that a foreign language provides psychological and emotional distance, thereby allowing people to be less biased in their decision-making, more willing to take smart risks and to be guided more by hope than by fear of loss. We show that a foreign language also affects ethical behavior such as cheating and moral choice. But we also find that when emotions are crucial for learning from experience, native tongue is crucial for improving choice over time. Living and functioning in a foreign tongue, then, has surprising consequences for how individuals think, feel and operate, and it has important implications for social policy, negotiation, diplomacy and immigration issues.
Information theoretic approaches to language universals
Finding explanations for the observed variation in human languages is the primary goal of linguistics, and promises to shed light on the nature of human cognition. One particularly attractive set of explanations is functional in nature, holding that language universals are grounded in the known properties of human information processing. The idea is that grammars of languages have evolved so that language users can communicate using sentences that are relatively easy to produce and comprehend. In this talk, I summarize results from explorations into several linguistic domains, from an information-processing point of view.
First, we show that all the world’s languages that we can currently analyze minimize syntactic dependency lengths to some degree, as would be expected under information processing considerations. Next, we consider communication-based origins of lexicons and grammars of human languages. Chomsky has famously argued that this is a flawed hypothesis, because of the existence of such phenomena as ambiguity. Contrary to Chomsky, we show that ambiguity out of context is not only not a problem for an information-theoretic approach to language, it is a feature. Furthermore, word lengths are optimized on average according to predictability in context, as would be expected under and information theoretic analysis. Then we show that language comprehension appears to function as a noisy channel process, in line with communication theory. Given si, the intended sentence, and sp, the perceived sentence we propose that people maximize P(si | sp ), which is equivalent to maximizing the product of the prior P(si) and the likely noise processes P(si → sp ). We discuss how thinking of language as communication in this way can explain aspects of the origin of word order, most notably that most human languages are SOV with case-marking, or SVO without case-marking.
The Language Network and Its Place within the Broader Architecture of the Human Mind and Brain
Although many animal species have the ability to generate complex thoughts, only humans can share such thoughts with one another, via language. My research aims to understand i) the system that supports our linguistic abilities, including its neural implementation, and ii) its interfaces with the rest of the human cognitive arsenal. I will begin by introducing the “language network”, a set of interconnected brain regions that support language comprehension and production. With a focus on the subset of this network dedicated to high-level linguistic processing, I will then consider two questions. First, what is the internal structure of the language network? In particular, do different brain regions preferentially process different levels of linguistic structure (e.g., sound structure vs. syntactic/semantic compositional structure)? And second, how does the language network interact with other large-scale networks in the human brain, like the domain-general cognitive control network or the network that supports social cognition? To tackle these questions, I use behavioral, fMRI, and genotyping methods in healthy adults, as well as intracranial recordings from the cortical surfaces in humans undergoing presurgical mapping (ECoG), and studies of patients with brain damage. I will argue that: i) Linguistic representations are distributed across the language network, with no evidence for segregation of distinct kinds of linguistic information (i.e., phonological, lexical, and combinatorial – syntactic/semantic – information) in distinct regions of the network. Even aspects of language that have long been argued to preferentially rely on a specific region within the language network (e.g., syntactic processing being localized to parts of Broca’s area) turn out to be distributed across the network when measured with sufficiently sensitive tools. Further, the very same regions that are sensitive to high-level (e.g., syntactic) structure in language show sensitivity to lower-level (e.g., phonotactic) regularities. This picture is in line with much current theorizing in linguistics and the available behavioral psycholinguistic data that shows sensitivity to contingencies spanning sound-, word- and phrase-level structure. And: ii) The language network necessarily interacts with other large-scale networks, including prominently the domain-general cognitive control system. Nevertheless, the two systems appear to be functionally distinct given a) the differences in their functional response profiles (selective responses to language vs. responses to difficulty across a broad range of tasks), and b) distinct patterns of functional correlations. My ongoing work aims to characterize the computations performed by these systems – and other systems supporting high-level cognitive abilities – in order to understand the division of labor among them during language comprehension and production.
The Role of the Body in Structuring Sociophonetic Variation
Scholars of gesture and bodily hexis have long recognized the centrality of the body in speech production (Bourdieu 1984, McNeill 1992, Kendon 1997). Yet theories of variation have generally been constructed based on analyses of what can be observed in the audio channel alone (cf. Mendoza-Denton and Jannedy 2011). This paper draws on a multimodal analysis of audiovisual data to illustrate that voice quality and vowel quality are strongly constrained by body movement and facial expression.
Dyadic interactions between friends were recorded in a sound-attenuated environment staged like a living room. The acoustic analysis focuses on the incidence of creaky voice (using Kane et al.’s 2013 neural network model) and vowel quality (the lowering and retraction of the front lax vowels, in accordance with the California Vowel Shift). Computer vision techniques were applied to additionally quantify the magnitude of body movements (movement amplitude) and identify when speakers were smiling.
Results show that body movement and facial expression predict the realization of both linguistic variables. Creaky voice was more common in phrases where speakers moved less, in phrases where they were not smiling (for women), and in interactions where speakers reported feeling less comfortable. The front lax vowels were lower (more shifted) among women, and in phrases where speakers (regardless of sex) were smiling.
Speakers use their bodies in non-random ways to structure linguistic variation, so analysts can improve quantitative models of variation by attending to forms of embodied affect. Focusing on the body can also facilitate the development of more comprehensive social analyses of variation, many of which rely solely on correlations between linguistic practice and social category membership. I conclude by discussing the implications of an embodied view of variation for language change.
The Usefulness of Useless Utterances: Why Um, Like, and Other Disparaged Phenomena are not Superfluous
Spontaneous communication differs from prepared communication in what is said, how it is said, and how talk develops based on addressee responses. Spontaneously produced phenomena such as ums, likes, and rising intonation on declarative sentences, or uptalk, are often vilified, but they have specific functions. In addition to what is said and how it is said, spontaneous communication involves responding to contributions from interlocutors. Even the shortest of addressee responses, such as the choice between uh huh versus oh, affects speaker production and overhearer comprehension. Differences between quotation devices, such as said versus like, also reflect functional choices. Because many spontaneous phenomena do not appear in carefully constructed communication, there has been a mistaken conclusion that they are uninformative. In fact, however, spontaneous phenomena are solutions to problems encountered in unplanned, unrehearsed communication.
Speakers mean more than their sentences do, because they can take a lot about their audience for granted. This talk explores how presuppositions and pragmatic enrichments play out in acquisition. How do children untangle semantic from pragmatic contributions to what speakers mean? The case study I will focus on is how children learn the meaning of the words think and know. When and how do children figure out that think but not know can be used to report false beliefs? When and how do they figure out that with know, but not think, speakers tend to presuppose the truth of the complement clause? I will suggest that the path of acquisition is traced by the child’s understanding both of where such verbs occur, and of why speakers use them. (joint work with Rachel Dudley and Jeff Lidz)
The challenges of event cognition: Object representation at the interface of episodic and semantic memory
Language is often used to describe the changes that occur around us – changes in either state (“I cracked the glass…”) or location (“I moved the glass onto the table…”). To fully comprehend such events requires that we represent the ‘before’ and ‘after’ states of any object that undergoes change. But how do we represent these mutually exclusive states of a single object at the same time? I shall summarize a series of fMRI studies which show that these alternative states compete with one another in much the same way as alternative interpretations of an ambiguous word might compete. This interference, or competition, manifests in a part of the brain that has been implicated in resolving competition. Moreover, activity in this area is predicted by the dissimilarity, elsewhere in the brain, between sensorimotor instantiations of the described object’s distinct states. Connectivity analyses show that hippocampus is also implicated in these cases of language/event comprehension, as a function of when episodic or semantic knowledge must be accessed. I shall end with the beginnings of a new account of event representation which does away with the traditional distinction between actions and participants, which maintains instead that object state representations across time are the fundamental representational primitive of event cognition, and which addresses how we instantiate individuated objects (tokens) from semantic memory (about types) on-the-fly. [Prior knowledge of the brain is neither presumed, required, nor advantageous!].