1. Evolutionary emergence. The slowest moving emergent structures are those that

are encoded in the genes. These structures, which are subject to more variability and

competition than is frequently acknowledged, are typically the result of glacial

changes resulting from the pressures of evolutionary biology. Stipulationist accounts

of language evolution (Bickerton, 1990; Chomsky, 1980) typically emphasize how

evolutionary discontinuities (Gould, 1977) have led to the construction of specific

modules. Emergentist accounts (MacWhinney, 2001) in this area emphasize

continuity and the ways in which evolution has reused older forms for new functions.

2. Epigenetic emergence. Translation of the DNA in the embryo triggers a further set

of processes from which the initial shape of the organism emerges. Some structures

are tightly specified by particular genetic loci. For example, the recessive gene for

phenylketonuria or PKU begins its expression_r prenatally by blocking the production

of the enzymes that metabolize the amino acid phenylalanine. Although the effects of

PKU occur postnatally, the determination of this metabolic defect emerges prenatally.

Other prenatal emergent structures involve a role for physical forces in the developing

embryo. The formation of the stripes of the tiger is an example of this type.

Epigenetic emergence does not cease at birth. To the degree that the brain maintains a

level of plasticity, epigenetic processes allow for recovery of function after stroke

through rewiring and reorganization.

3. Developmental emergence. Jean Piaget’s genetic psychology (Piaget, 1954) was

the first fully articulated emergentist view of development. Impressively complete in

its coverage, it was often incomplete in terms of its underspecification of particular

mechanisms of development. Attempting to provide this missing mechanistic detail,

current emergentist accounts of development rely on connectionism, embodiment, and

dynamic systems theory.

4. Online emergence. The briefest timeframe for the study of emergent processes is

that of on-line language processing. Emergentist accounts are now showing how

language structure emerges from the pressures and loads imposed by online

processing (MacWhinney, 1999). These pressures involve social processes, memory

mechanisms, attentional focusing, and motor control.

5. Diachronic emergence. We can also use emergentist thinking to understand the

changes that languages have undergone across the centuries (Bybee, 1998). These

changes emerge from a further complex interaction of the previous three levels of

emergence (evolutionary, developmental, and online).

With these five timeframes in mind, we can provide a revised interpretation of the standard question “Is it innate or learned?” What this question really means is “Across what timeframe does this ability emerge?” If we could consistently replace the earlier form of this basic question with this newer form, I believe that much of our scientific dialog regarding the nature of human language would become clarified.

Even within this newer framework, there is still an enormous amount to discuss and debate. First, we can easily disagree regarding the timeframe for a given ability. Consider the case of the “KE” family in East London studied by Gopnik and Crago (1990). This family has members who exhibit problems with the marking of regular suffixes on verbs. Specifically, the affected family members tend to use “jump” as the past tense of “jumped” more than their languagematched controls. Genetic analysis (van der Lely & Stollwerk, 1996) points to a pattern of autosomal dominant inheritance, since about half of the members of three generations descending from a particular grandmother are affected. Researchers such as van der Lely and Gopnik have interpreted this deficit in non-emergentist terms. They see it as involving a specific mutation on a specific gene that somehow controls the process of regular suffixation and perhaps other aspects of linking (Van Der Lely & Christian, 2000).

Emergentist accounts of this familial pattern provide a more complete picture. Emergentist accounts are able to deal with the fact that this disability impacts many aspects of motor control apart from language (Alcock, Passingham, Watkins, & Vargha-Khadem, 2000). Affected family members have problems with swallowing, finger tapping, mouth control, and other fine motor actions. Their speech is effortful and strained, as if they were dealing with a major disconnection to the control of motor output. This pattern of impairment suggests that we are not dealing with a“grammar gene” or a module for regular inflection (Pinker, 1991), but with a general motor impairment that impacts regular morphology, perhaps because of the omissibility of the regular marker (Labov, 1986; Leonard, 1998). From an emergentist perspective, this particular disability could be linked evolutionarily to the recent consolidation of motor control for language in the human species (Donald, 1991; MacNeilage, 1998). However, the fact that people with language impairments tend to marry and reproduce less than the general population would make it difficult for a disability like this to propagate. The fact that this particular disability has not been reported from other families indicates this must be either a very recent mutation or the result of a virus with an effect on gestation. A virus of this type, which can also be inherited, may have impacted the embryological formation of pathways for motor control.

This reinterpretation of the impairment in the KE family provides several important lessons. First, when interpreting disabilities, we need to be careful about assuming the existence of astipulated gene that controls a stipulated module. We should examine the complete manifestation of the disability, asking ourselves questions about timeframe and mechanism. Second, if we want to postulate phylogenetic emergence, we need to see that the trait has been widely distributed in the species. Third, if suspect phylogenetic emergence, we need to also understand how the particular genetic pattern influences embryological development. Fourth, we need to examine the extent to which the disorder may be linked genetically to some form of compensatory adaptive advantage (Bradshaw & Sheppard, 2000). For example, we know that the gene for sickle cell anemia, while maladaptive in homozygotes, provides protection against malaria in heterozygotes.

What would be a parallel function for language disorders? Finally, when we look at the behavioral expression_r of the postulated disorder, we need to consider the ways in which general cognitive and motor disabilities can have differential impacts on specific linguistic patterns. For example, Bates, Wulfeck, and MacWhinney (1991) have shown that all forms of aphasia tend to lead to the omission and misinterpretation of grammatical morphology. In fact, we know that grammatical morphology is subject to loss in non-neurological patients such as those suffering from lower-back injuries. To understand these patterns, we need to focus on models of information load during online processing.

Emergentist accounts must specify particular mechanisms that operate on particular timeframes.As we move to replace the earlier stipulationism with the new emergentism, we need to focus on developing a fuller understanding of the arsenal of basic emergent mechanisms. In the end, all emergentist accounts must be grounded on these core mechanisms. If we attempt to postulate specialized mechanisms for single problems, we are returning to stipulationism. Some examples of general mechanisms include:

1. Learning through error propagation. A good example of this type of mechanism is the back-propagation algorithm used in PDP modeling (Rumelhart & McClelland, 1986).

2. Self-organization. Mechanisms such as the self-organizing feature map (Kohonen, 1990) provide alternatives to mechanisms based on error propagation.

3. Item-based learning. In the area of grammatical development, the theory of item-basedlearning (MacWhinney, 1975; Tomasello, 2000b) relies on general concepts from Construction Grammar (Goldberg, 1999).

4. Reorganization of cognitive function to the contralateral hemisphere. Children with early left focal lesions are able to recover language function by reorganizing language to the right hemisphere. This plasticity in development is a general mechanism that supports a wide variety of emergent responses to injury or sensory disability (Booth et al., 1999; Corina, Vaid, & Bellugi, 1992; MacWhinney, Feldman, Sacco, & Valdes-Perez, 2000).

5. Physical pressures on cognitive structures. Phonologists have shown that the shape of the vocal mechanism has a wide-ranging impact on phonological processes (Ohala,

1974). Rather than stipulating phonological rules or constraints (Bernhardt &

Stemberger, 1998; Kager, 1999), we can view them as emergent responses to these

underlying pressures.

6. Conversational emergence. Linguistic structures seem to be adapted to specific

conversational patterns as they emerge online. For example, Du Bois (1987) has argued that ergative marking in languages emerges from the fact that speakers tend to delete the actor in transitive sentences, because it is already given or known.

7. Perceptual recording. Recent studies of infant auditory perception (Jusczyk, 1997) have revealed that, even in the first few months, infants apply some general-purpose

mechanism to record and learn from auditory input.

8. Constituent structure. All syntactic theories need to assume that related words cluster together in units and that the head of those units then serves to cluster with higher argument slots. This fundamental process of constituent structuring must be based on a set of basic mechanisms for motor control and planning (Donald, 1999).

This is, of course, just a small sampling of the many mechanisms and pressures that shape the emergence of language. Others involve the shape of social relations in the young child’s family (Ninio & Snow, 1988), the shape of the input to guest workers learning a second language (Klein & Perdue, 1989), the preference in the brain for short connections (Shrager & Johnson,1995), and the shape of sound dissipation for low frequencies across distances. In each case, the mechanisms we are considering are either corroborated through direct observation or are highly general processes based on lower-level mechanisms that have been directly observed.

4 Domain Generality

Within the language learning community, there is an active debate regarding the extent to which language learning is based on domain-general mechanisms. Sabbagh and Gelman (2000)

present an analysis which equates emergentism with domain generality. This strong formulation

of the emergentist position matches up well with the disembodied connectionism of the 1980s

(Rumelhart & McClelland, 1986). However, the strong version fails to fully appreciate the

degree to which emergentists view cognition as grounded on the body, the brain, and the social

situation.

Consider a simple example from phonological development. There is a universal tendency to

avoid sequences of nasal consonants followed by voiceless obstruents, as might arise in forms

like ‘manpower.’ This constraint is grounded on the facts of speech production (Huffman,

1993) and figures prominently in recent elaborations of Optimality Theory (Kager, 1999).

Languages use at least five phonological processes to deal with this problem. These processes

include nasal substitution, post-nasal voicing, denasalization, nasal deletion, and vowel

epenthesis. Initially, children may apply a variety of these processes (Bernhardt & Stemberger,1998). Which processes are preserved and which are dropped out will depend on the shape of the target language, be it Indonesian, Quechua, Toba Batak, English, or Kelantan Malay. The shape of the vocal tract and the innervation of the muscles of the tongue determine the domainspecific landscape. Domain-general processes sample these constraints and negotiate between them in real time. In the terms used by Sabbagh and Gelman, the overall system of constraint satisfaction is a ‘buzzsaw’ cutting patterns through the local domain of embodied articulatory constraints. This example emphasizes the extent to which emergentism must make reference to the body. To attempt to construct an emergentist psycholinguistics that ignores the body, the brain, and the social situation would be like attempting to build an emergentist account of honeycomb formation that ignores the honey.

Although it is clear that emergentism needs to refer to domain-specific facts about the body, it is not clear that it needs to rely on any domain-specific cognitive mechanisms. Instead, it is likely that evolution reuses general cognitive mechanisms to serve new functions in special areas. For example, Givón (1998) has argued that the major cognitive event that occurred during language evolution involved a linkage of episodic memory to the auditory system through the support or tutelage of the visual system. The visual system had already established general mechanisms for the episodic encoding of spatial position and form. Primates had already developed a mechanism for recording auditory sequences (Hauser, Newport, & Aslin, 2001). Adapting this mechanism to the task of language learning involved reshaping and relinking previously available cognitive mechanisms. It is true that these domain general episodic mechanisms have a specific localized shape for each modality. However, it is likely that the general mechanisms undergo a special tuning when they function at the local level (Caplan & Waters, 1999; Pinker, 1999).

5 Emergence in Grammar

In the next three sections, I will present three specific example emergentist solutions to central problems in language learning. These sections will examine, respectively, grammatical learning, lexical learning, and language evolution. In this section, we will look at how emergentism provides accounts for grammatical learning.

One of the most active areas in recent work on language acquisition has been the study of the

child’s learning of inflectional marking. In English, inflections are short suffixes that occur at the ends of words. For example, the word “dogs” has a final /s/ suffix that marks the fact that it is plural. There are now well over 30 empirical studies and simulations investigating the learning of inflectional marking. The majority of work on this topic has examined the learning of English verb morphology with a particular focus on the English past tense. These models are designed to learn irregular forms such as “went” or “fell”, as well as regular past tense forms such as “wanted” and “jumped”. Other areas of current interest include German noun declension, Dutch stress placement, and German participle formation. Although the learning of inflectional markings is a relatively minor aspect of language learning, our ability to quantify this process has made it an important testing ground not only for the study of child language, but for developmental psychology and cognitive science more generally.