Our findings suggest that syntactic processing depends primarily on dorsal language tracts. This was demonstrated by strong correlations between reduced FA in the SLF/Arcuate and deficits in syntactic comprehension and production. In contrast, we found that damage to ventral tracts—the extreme capsule fiber system or the uncinate fasciculus—does not result in syntactic deficits.
When other potentially important factors were included as covariates, the integrity of the SLF/Arcuate continued to be associated with syntactic processing function. Specifically, we observed relationships between FA in the SLF/Arcuate and syntactic comprehension and production when we took into account PPA variant, overall severity, executive function, motor speech, and gray matter atrophy in the left IFG, the cortical region most associated with syntactic deficits. These analyses indicate that although these factors certainly may contribute to syntactic deficits, the SLF/Arcuate makes a unique contribution to syntactic processing even when these other factors are accounted for. Furthermore, the fact that we found robust correlations with both syntactic comprehension and production measures makes it less likely that the deficits resulting from SLF/Arcuate damage reflect component processes such as executive functions or motor speech.
A key role for the SLF/Arcuate in syntactic processing has been suggested previously based on indirect evidence from fiber tracking connecting regions activated in an fMRI study of syntactic processing (Friederici et al., 2006
). Our findings provide more direct evidence for the importance of dorsal tracts for syntactic processing, by showing that damage to these tracts results in syntactic deficits. Syntax is perhaps the most uniquely human component of language, due to its hierarchical structure, unparalleled complexity, and recursion, which gives rise to infinite generativity. Therefore it might be expected that the neural substrate(s) for syntactic processing might have been significantly modified over the course of human evolution. A recent comparitive DTI study reported that the arcuate branch of the SLF is indeed strongly modified in humans relative to non-human primates; it projects much more densely to posterior temporal cortex than it does in macaques or chimpanzees, especially in the left hemisphere (Rilling et al., 2008
Recent studies have established the importance of ventral tracts including the ECFS and UF in language processing (Friederici et al., 2006
; Friederici, 2009
; Saur et al., 2008
; Weiller et al., 2009
). Our results support the importance of these tracts in language processing, indicating that they may play a role in lexical processing at the single word level. Ventral tracts are most severely affected in patients with semantic variant PPA (Galantucci et al., 2011
), who present with profound lexical deficits encompassing lexical retrieval, single word comprehension, and semantic knowledge (Hodges and Patterson, 1996
). Furthermore, a role for ventral tracts in single word processing is consistent with the observation that regions connected by ventral tracts are activated by language comprehension (Saur et al., 2008
), since language comprehension typically involves both lexical and syntactic processes. However our results do not support suggestions that these tracts play a direct role in processing of grammar (Weiller et al., 2009
) or computation of local phrase structure (Friederici, 2009
). Many patients with significant degeneration of these ventral tracts showed normal or near-normal syntactic processing, and in general, there were no correlations between damage to these tracts, and syntactic deficits. These observations would be difficult to account for if ventral tracts play a key role in syntactic processing.
Although we have argued that the left SLF/Arcuate is the most important tract for syntactic processing, this is not to imply that this tract is important only for syntactic processing. The SLF/Arcuate is clearly also crucial for other aspects of speech/language processing and other cognitive functions. For instance, vascular lesions and neurodegenerative volume loss in the SLF/Arcuate have been associated with motor speech deficits (Ogar et al., 2006
; Wilson et al., 2010b
), and in this study we found that reduced FA in the SLF/Arcuate was associated with motor speech deficits (see Supplemental Text
Two limitations of our study are noteworthy. First, the SLF/Arcuate has multiple subcomponents (Catani et al., 2005
; Frey et al., 2008
; Makris et al., 2005
), which are often damaged in parallel, for instance in non-fluent PPA (Galantucci et al., 2011
). For this reason, we could not determine whether syntactic processing depends differentially on particular subcomponents of the SLF/Arcuate.
Second, fibers passing through the extreme capsule connect wide regions of frontal cortex with wide regions of temporal and occipital cortex (Makris and Pandya, 2009
), raising the possibility that a subset of ECFS fibers might be important for syntactic processing, which we might not have identified because we quantified FA in the whole ECFS. However, this concern is mitigated by the secondary analysis where the ECFS was constrained to connect fMRI-derived ROIs, and we continued to observe no relationship between the ECFS and syntactic processing.
In conclusion, we used a multimodal imaging approach, combining DTI with voxel-based morphometry and fMRI, to show that the dorsal and ventral language pathways linking frontal and temporal language regions have distinct functional roles. Only the dorsal pathway (SLF/Arcuate) plays a critical role in syntactic processing. Our findings suggest that syntactic deficits (Amici et al., 2007
; Gorno-Tempini et al., 2004
; Grossman and Moore, 2005
; Grossman et al., 2005
; Hodges and Patterson, 1996
; Thompson et al., 1997
; Wilson et al., 2010b
) and functional abnormalities related to syntactic processing (Wilson et al., 2010a
) in PPA may reflect not only gray matter damage, but also disruption of communication between frontal and temporal language regions (Sonty et al., 2007
), specifically via the dorsal pathway.