Communicative language, one of the most important qualities distinguishing humans from other species of animal, involves a Gordian knot of intertwined neural systems within the brain. Only recently have we begun to unravel these systems with the advent of imaging and measuring techniques that can coax out language's patterns of activation. In the process, the study of inner structures of the mind has enjoined with the theories of its outer workings, especially in relation to the field of linguistics. Because language theorists were hitherto limited to observing the output of a 'black box' and deriving its workings based only upon those external signs, competing hypotheses concerning the fundament of communication have remained untested. The tools and expertise of neurology have begun to remove this obstacle, affording linguists the opportunity to open the lid of the black box and finally get a sense of what lays within.

Sin → sinned, Win → winned

A recent study by Ullman et. al. (2005) has used this method of neuroscientific study to provide evidence for the dual-system theory of morphology. The dual-system theory addresses the question of the interaction between the lexicon—memorized parings of vocal or visual utterance and meaning commonly known as 'words—'and grammar—the system of rules that predictably builds lexical forms into composite words, phrases, and sentences that communicate complex meanings.

The examples of this phenomenon are diverse; English indicates a third-person subject of a verb by the suffix -s (as in 'Jane walk-s'), Arabic prefaces nouns with the syllable al- as an equivalent to 'the' (as in al-Jazeera)*, and Japanese negates an adjective by changing its final syllable to -kunai. There are exceptions, however, to the rules governing lexicon/grammar interface in every language. For example, the past tense of the verb 'to sing' is not 'singed,' but 'sang.'

The dual-system theory proposes that lexicon and grammar are processed in different regions of the brain, with basic forms of words stored in declarative memory localized to the temporal and temporal-parietal regions of the left hemisphere, and the rules governing their modification stored in non-declarative memory localized to the left frontal cortex and especially Broca's area. Irregular forms are given their own places within the lexicon, allowing them to override the application of regular grammatical rules.

In his book Words and Rules (1999), Steven Pinker proposes that this double-dissociation allows quicker mental access to irregular words, sidestepping the process of taking the word from the lexicon and determining the correct rule to apply to it. Since irregular words are generally used more frequently than their regular counterparts, double dissociation would bridge the gap between the necessities of consistent grammar and fluid speech.

Words and Rules? Or just Words and Statistics?

Contrasting with this explanation is the single-system theory, which proposes that language depends upon a single computational system without localization of specific functions, depending upon statistical interactions within a neural network to produce grammar and lexicon in speech. Ullman et al. sought to provide evidence for localization in order to support the dual-system theory and erode the basis of the single-system theory. They interviewed two sets of aphasics—patients who have suffered speech impairment due to brain damage. Eleven non-fluent aphasics, characterized by damage to the left frontal cortex and speech lacking grammatical coherency, were compared to nine fluent aphasics, characterized by damage to the temporal and temporal-parietal cortices and difficulty retrieving content words in speech, in addition to sixteen cognitively unimpaired subjects.

It was proposed that for the task of producing the regular and irregular past tense forms of real and novel English verbs, non-fluent aphasics would demonstrate difficulty producing regular in comparison to irregular past tense forms, with a tendency not to overregularize (i.e. digged instead of dug). Fluent aphasics, in contrast, would demonstrate difficulty with irregular past tense forms, overregularizing real verbs and avoiding probable irregular past tense forms of novel verbs (strinked for strink, rather than strank or strunk).

Subjects were given a series of real and novel verbs chosen for their diversity and lack of homonyms in other parts of speech and directed to provide their real or plausible past tense forms. Researchers transcribed the subjects' first responses and judged them either correct or incorrect, with partial or unformed answers coded 'other.'

Yes Virginia, you still have to learn grammar separately from vocabulary.

The results of the study confirmed the hypothesis, lending credence to the notion that grammar and lexicon are localized to different regions of the brain as suggested by the dual-system theory. Significantly, even controlling for factors that could explain the non-fluent/fluent specific errors according to the single-system theory of a neural network, such as the phonological complexity of the verbs contributing to speech impairment, the differences between the non-fluent and fluent aphasic groups were still statistically significant.

The findings not only confirm the localization of lexicon and grammar, but furthermore imply that the unified phenomenon of language utterance can be broken down into neurostructural components within the brain, with matters of content coding such as words (or gestures in the case of sign language) stored in declarative memory localized to the left temporal and temporal-parietal lobes, and matters of structure such as grammar stored in non-declarative (procedural) memory localized to the left frontal cortex.

Besides the light this study sheds on a contentious area of linguistic theory, it may also have implications for the study of language acquisition and learning. While children appear to have a biological predilection to learning language, adults require different techniques to achieve near-fluency. Learning methods that take into account the dissociation of grammar from lexicon in the case of regular forms and the lexicon override for irregular forms may show better results.

Further research is still necessary to better determine how localization functions in morphology, including examination of secondary firing patterns in non-localized regions. Improvements to the neural network single-system theory may yield a supplementary or perhaps even superior model as closer observation provides linguists and neuroscientists with more thorough information. The black box of language has only just opened its lid to human eyes. There are sure to be greater depths left to probe.


Pinker, Steven. 1999. Words and Rules: the Ingredients of Language. New York: Perennial.

Ullman M.T., Pancheva R, Love T, Yee E, Swinney D, Hickok G. 2005. Neural correlates of lexicon and grammar: Evidence from the production, reading, and judgement of inflection in aphasia. Brain and Language 93:185-238

Log in or register to write something here or to contact authors.