Recent studies of spoken language in healthy populations
The central goal of our spoken language research is to develop an account of the neurobiology of language. Our work with healthy volunteers forms the basis for our theoretical model. Spoken language is underpinned by two distinct neural systems. The first occupies both left and right hemispheres of the brain, especially the temporal lobe, is shared with non-human primates, and enables the interpretation of incoming auditory signals with respect to social signals. The second is a left-hemispheric frontal and temporal system dedicated to core linguistic functions found only in humans, such as processing sentential syntax (grammar). The dual system model is described in more detail in a Research Features article written for the University of Cambridge website by Professor Tyler and Professor William Marslen-Wilson (see also Tyler & Marslen-Wilson, 2008, and Bozic et al., 2010, in the Further Reading section).
Our functional neuroimaging studies in healthy controls have consistently associated the left fronto-temporal system with sentential syntax processing. Critically, our experiments are carefully design to isolate syntactic processing from other processes that may activate frontal cortex: those elicited by task-related effects such as memory and decision making, or error-checking processes which may be elicited by unnatural or ungrammatical sentences. We can elicit syntactic processing without using a task or artificial sentences by introducing syntactic ambiguity. For example, the sentence beginning "He noticed landing planes ..." may continue using "landing" as either an adjective (e.g. "... are noisy overhead") or a verb (e.g. "... is daunting to novice pilots"). These sentences elicit greater activity in the left fronto-temporal system than unambiguous sentences (Rodd et al., 2010; Tyler et al, 2011). These results are similar to those obtained using a different language comprehension paradigm that engages syntax (see Tyler et al., 2010, discussed in the Ageing section).
We have also investigated the temporal properties of syntactic processing. By combining measures of brain activity recorded with magnetoencephalography (MEG) and multivariate statistical analysis we have sought to characterise the dynamics of syntactic computations in the fronto-temporal language system (Tyler et al., 2013). Using representational similarity analysis, we tested a variety of lexico-syntactic and ambiguity similarity models against the MEG data, and showed that the left MTG represents and transmits lexical information to the LIFG, which responds to and resolves syntactic ambiguity. This points to a clear differentiation in the functional roles of these two regions:
Summary of the MEG representational similarity analysis results showing effects in the LpMTG and LIFG during the central phrase and subsequent disambiguation. From Tyler et al., 2013, Frontiers in Psychology.
Recent studies of spoken language in brain-damaged populations
In addition to our work with healthy volunteers, we gain unique insights into how the brain processes spoken language by studying people with brain injuries. When we find a relationship between damage in a particular network of brain regions and performance on a specific task, we can infer that this network is essential for preserved performance. This kind of inference is not possible in functional neuroimaging studies of the intact brain, which can only tell us which regions are correlated with task performance.
We are not only interested in how damage to the language system affects performance, but also in how the brain may adapt or reorganise to preserve performance after injury. To investigate this we perform functional neuroimaging experiments with patients.
Our studies have found that damage to either the frontal or temporal regions in the left hemisphere - activated during syntactic processing in healthy controls - impairs performance on tests of syntactic processing in patients (Tyler et al., 2010 & 2011). Moreover, performance is also impaired by damage that affects communication between these regions (Papoutsi et al., 2011; Rolheiser et al., 2011). Even though patients with damage in the left hemisphere may activate frontal regions in the right hemisphere that are not usually activated in controls, this additional activity does not correlate with performance. This confirms that the left fronto-temporal network is essential for syntax, so performance can only be preserved by the intact parts of the left hemisphere, not by recruiting corresponding regions in the right hemisphere.
We know that people with large injuries in their left hemisphere can still understand the gist of speech, despite having impaired syntax. This may be possible because the bilateral comprehension system (shared with nonhuman primates) may continue to function using intact tissue in the right hemisphere of the brain. Further analysis of our patient data showed that although comprehension task performance based on syntax alone correlated with tissue integrity and brain activation in the left frontotemporal system, a distinct system in the right temporal lobe correlated with comprehension performance supported by semantic information (meaning) in addition to syntax (Wright et al., 2012). We interpret this right temporal region as reflecting the intact part of the bilateral comprehension system.
This work was funded by the Medical Research Council (UK) programme grant (grant number G0500842) to Professor Tyler. We are immensely grateful to our volunteers for their essential contribution to this work.
Patients' performance on a test of syntactic comprehension correlates consistently with a left frontotemporal network whether looking at brain activity during syntactic processing (left) or tissue integrity (right). From Tyler et al. (2011) Brain, 134(2), 415-431
The importance of connectivity in spoken language comprehension
Although the neural language system and other neurocognitive systems can be described in terms of functionally segregated brain regions, it is communication between these regions that allows them to operate as a functionally integrated system. Frontal and temporal regions play a particularly salient role in language comprehension and production. We study dynamic, on-line (functional/effective) connectivity during language perception tasks using fMRI and MEG, as well as the anatomical connections underlying this using diffusion-weighted MR imaging and tractography. A key conclusion that has come from our tractography work in stroke patients is that distinct linguistic abilities do not rely on separate neural pathways, but rather multiple pathways interact dynamically to support multiple linguistic abilities (Rolheiser et al., 2011). Two distinct fronto-temporal white matter tracts - a dorsal pathway (the arcuate fasciculus), and a ventral pathway (the extreme capsule fibre system) - were both recruited by multiple levels of psycholinguistic analysis (morphology, phonology, syntax, semantics). Although these cognitive functions loaded differentially on the two fibre systems, our findings argue strongly against any strict regional specialization for different psycholinguistic component processes, either within the frontal and temporal lobes or within the white matter structures that interconnect them.
Two white matter tracts essential for syntax. Dark red blobs show regions where white matter integrity (Fractional Anisotropy) correlates with performance on syntactic comprehension in patients with left hemisphere damage. Pale blobs show reference locations of key language tracts, arcuate fasciculus (yellow) and extreme capsule (blue). From Rolheiser et al. (2011), J. Neurosci. 31(47), 16949-16957.
Syntactic comprehension depends upon functional interaction between frontal and temporal regions. Left: red and yellow blobs show regions in left posterior middle temporal gyrus where the effective connectivity (PPI) with a seed region in left inferior frontal gyrus (green) correlated with performance on syntax in patients with left hemisphere damage. Adapted from Papoutsi et al. (2011) NeuroImage, 58(2), 656-664.