Christiansen & Chater (C&C) argue that language is shaped by the constraints of the brain, and that, over time, language changes to fit those constraints. They demonstrate the implausibility of an innate Universal Grammar (UG) that encodes arbitrary principles of language, and suggest that the features of language we observe are those best suited to the domain-general constraints humans bring to the problem of language acquisition. Both the innate UG that C&C reject and the domain-general constraints that they propose are thought to arise through evolution – they are encoded in the genome. However, genes do not function in isolation from their environment, and domain-general learning mechanisms do not exist in the absence of external input (see, e.g., Spencer et al., in press). Therefore, understanding the mechanisms that shape language change involves understanding how domain-general processes are shaped by the interaction of genes and environment. This interaction is tied to development over multiple timescales. Research suggests that even domain-general processes, like those C&C outline, change over the course of in-the-moment learning and developmental time. Unless we provide an explanation of language evolution with respect to developmental process, we have done little more than substitute one set of genetically encoded principles for another.

Interactions occur at multiple levels, from gene expression and cell signaling to the formation of neural networks and behavior (Tomblin & Christiansen, in press). The rise of molecular genetics has greatly changed how we investigate cognition, neurodevelopment, and behavior. At present we understand very little about how genes are involved in these processes. However, we do know that genes make regulatory factors and signaling molecules that interact in highly complex frameworks, modulated by environmental influences, to build and maintain the brain. These processes can have an impact on the nature of domain-general learning mechanisms that may drive language change. For example, brain-derived neurotrophic factor (BDNF) is a protein that is important for synapse formation, and, in particular, modulates hippocampal learning. Egan et al. (Reference Egan, Kojima, Callicott, Goldberg, Kolachana, Bertolino, Zaitsev, Gold, Goldman, Dean, Lu and Weinberger2003) found that BDNF polymorphisms are associated with differences in verbal episodic memory. Given that language learning is reliant upon long-term storage of arbitrary mappings, BDNF is widely believed to contribute to vocabulary learning. Thus, inheritance of a particular polymorphism may influence word learning abilities. At the cellular level, secretion of BDNF leads to changes in neurotransmitter activity, synaptic connectivity, and long-term potentiation (LTP). In addition, external stimulation that produces LTP influences the expression of BDNF (Patterson et al. Reference Patterson, Grover, Schwartzkroin and Bothwell1992). Thus, the interaction between genes, their expression, and the environment is critical for determining the properties of learning systems relevant to language.

At higher levels, we see evidence of complex interactions in the development of domain-general capacities. MacDonald and Christiansen (Reference MacDonald and Christiansen2002) demonstrated this in a neural network that was trained to predict the next word in a sentence. Performance in this task is dependent on the network's processing capacity for word sequences. Their simulations showed that processing capacity is an emergent property dependent on both the architecture of the network and its experience. Similarly, Ross-Sheehy et al. (Reference Ross-Sheehy, Oakes and Luck2003) showed that 4-month-olds have a visual short-term memory capacity for only one object, whereas by 13 months, infants' capacity increases to four objects. Thus, infants are not born with an adult-like short-term memory. Rather, this capacity depends on developmental processes and is likely to be modulated by biological and environmental factors. Those factors include the way memory representations are brought to bear on in-the-moment behavior; Spencer et al. (Reference Spencer, Simmering, Perone and Ross-Sheehy2008) have shown that the visual working memory capacity of preschoolers depends on the specifics of how memory representations are probed. This provides evidence of change in domain-general processes at the level of the in-the-moment timescales from which learning and development emerge (see also, Samuelson & Horst Reference Samuelson and Horst2008).

How might developmental changes in these domain-general mechanisms impact language processing? We know that age and previous linguistic experience influence the process of language acquisition and that there are large differences in the ability of children and adults to learn language. Acquisition of a first language is accomplished with relative ease; in fact, children are remarkable language learners. Second language acquisition as an adult, however, is quite difficult. One way to explain this difference is to consider the different developmental histories children and adults bring to the problem of learning a language. If developmental history affects future learning, we might suppose that information learned earlier will be weighted more heavily than information learned later. Under this assumption, adults will have difficulty learning a new language because of their developmental history (i.e., they assign less weight to new information given previous language experience). Even though adults and children bring the same domain-general learning constraints to the problem, differing developmental histories can affect their ability to learn a language.

Experimental work by Hudson Kam and Newport (Reference Hudson Kam and Newport2005) has demonstrated that developmental history can impact the learning processes that shape language change. When adults and 5- to 7-year-old children were exposed to an artificial language with an inconsistent grammar, the children regularized the language, producing more consistent constructions than the ones they had heard, while the adults produced only the constructions they had heard. Hudson Kam and Newport hypothesized that this difference in regularization is caused by differences in the way children's and adults' learning processes are constrained by prior developmental history. Thus, the temporal aspect of development, that is, that children have less accumulated developmental history, can have an impact on how languages will form and change over time.

These examples demonstrate that development is essential for describing the learning processes that drive language acquisition and language evolution. Factors that determine the learnability and expressivity of language are shaped by developmental history. This history is determined by an interaction between an organism and its input and between genes and the environment at multiple timescales. Therefore, we agree with C&C that language has changed to fit the brain. However, in order to understand how language changes, we must also understand how it develops in the individual.