Language acquisition in the absence of explicit negative evidence: can simple recurrent networks obviate the need for domain-specific learning devices?

On the matter of essence

A large literature suggests that people attribute the inborn properties of living things to their essence. Here, I explore the possibility that the essence of living things must be further embodied, and that this presumption guides intuitive reasoning about all of an organism's inherited properties, physical and psychological. Accordingly, when people reason about agentive living things (animals and humans), they presume that (a) Their essence must exhibit the properties of bodily matter- it must occupy a certain location in space, and it must be comprised of some appropriate organic substance that is anchored in the body; (b) Inborn (essentialized) traits must be embodied, and conversely, embodied traits are likely innate; and (c) The identity of biological kinds and by extension, one's psychological core, are defined by the material properties of their essence. I show that the embodiment hypothesis can capture numerous phenomena, ranging from laypeople's intuitions about which psychological traits are plausibly innate to their perception of the self and its capacity to migrate to humanoids and reemerge after death.

Syntax acquisition

Every normal child acquires a language in just a few years. By 3- or 4-years-old, children have effectively become adults in their abilities to produce and understand endlessly many sentences in a variety of conversational contexts. There are two alternative accounts of the course of children's language development. These different perspectives can be traced back to the nature versus nurture debate about how knowledge is acquired in any cognitive domain. One perspective dates back to Plato's dialog 'The Meno'. In this dialog, the protagonist, Socrates, demonstrates to Meno, an aristocrat in Ancient Greece, that a young slave knows more about geometry than he could have learned from experience. By extension, Plato's Problem refers to any gap between experience and knowledge. How children fill in the gap in the case of language continues to be the subject of much controversy in cognitive science. Any model of language acquisition must address three factors, inter alia: 1. The knowledge children accrue; 2. The input children receive (often called the primary linguistic data); 3. The nonlinguistic capacities of children to form and test generalizations based on the input. According to the famous linguist Noam Chomsky, the main task of linguistics is to explain how children bridge the gap-Chomsky calls it a 'chasm'-between what they come to know about language, and what they could have learned from experience, even given optimistic assumptions about their cognitive abilities. Proponents of the alternative 'nurture' approach accuse nativists like Chomsky of overestimating the complexity of what children learn, underestimating the data children have to work with, and manifesting undue pessimism about children's abilities to extract information based on the input. The modern 'nurture' approach is often referred to as the usage-based account. We discuss the usage-based account first, and then the nativist account. After that, we report and discuss the findings of several studies of child language that have been conducted with the goal of helping to adjudicate between the alternative approaches to language development. WIREs Cogn Sci 2012, 3:185-203. doi: 10.1002/wcs.1158 For further resources related to this article, please visit the WIREs website.

Evolutionary psychology and the brain

The human brain is a set of computational machines, each of which was designed by natural selection to solve adaptive problems faced by our hunter-gatherer ancestors. These machines are adaptive specializations: systems equipped with design features that are organized such that they solve an ancestral problem reliably, economically and efficiently. The search for functionally specialized computational adaptations has now begun in earnest. A host of specialized systems have recently been found, including ones designed for sexual motivation, social inference, judgment under uncertainty and conditioning, as well as content-rich systems for visual recognition and knowledge acquisition.