Research Article: Behavioural and neuroanatomical correlates of auditory speech analysis in primary progressive aphasias

Date Published: July 27, 2017

Publisher: BioMed Central

Author(s): Chris J. D. Hardy, Jennifer L. Agustus, Charles R. Marshall, Camilla N. Clark, Lucy L. Russell, Rebecca L. Bond, Emilie V. Brotherhood, David L. Thomas, Sebastian J. Crutch, Jonathan D. Rohrer, Jason D. Warren.


Non-verbal auditory impairment is increasingly recognised in the primary progressive aphasias (PPAs) but its relationship to speech processing and brain substrates has not been defined. Here we addressed these issues in patients representing the non-fluent variant (nfvPPA) and semantic variant (svPPA) syndromes of PPA.

We studied 19 patients with PPA in relation to 19 healthy older individuals. We manipulated three key auditory parameters—temporal regularity, phonemic spectral structure and prosodic predictability (an index of fundamental information content, or entropy)—in sequences of spoken syllables. The ability of participants to process these parameters was assessed using two-alternative, forced-choice tasks and neuroanatomical associations of task performance were assessed using voxel-based morphometry of patients’ brain magnetic resonance images.

Relative to healthy controls, both the nfvPPA and svPPA groups had impaired processing of phonemic spectral structure and signal predictability while the nfvPPA group additionally had impaired processing of temporal regularity in speech signals. Task performance correlated with standard disease severity and neurolinguistic measures. Across the patient cohort, performance on the temporal regularity task was associated with grey matter in the left supplementary motor area and right caudate, performance on the phoneme processing task was associated with grey matter in the left supramarginal gyrus, and performance on the prosodic predictability task was associated with grey matter in the right putamen.

Our findings suggest that PPA syndromes may be underpinned by more generic deficits of auditory signal analysis, with a distributed cortico-subcortical neuraoanatomical substrate extending beyond the canonical language network. This has implications for syndrome classification and biomarker development.

The online version of this article (doi:10.1186/s13195-017-0278-2) contains supplementary material, which is available to authorized users.

Partial Text

The primary progressive aphasias (PPAs) continue to present substantial problems of classification and diagnosis. A number of patients do not meet consensus diagnostic criteria for particular PPA syndromes [1], while the major syndromes show clinical and anatomical overlap [2]. Accumulating evidence suggests that abnormalities of speech processing in these ‘language-led’ dementias may reflect broader deficits of non-linguistic auditory signal decoding; indeed, presentations with progressive word deafness were among the first descriptions of PPA [3] and have since been expanded upon in some detail [4–13]. Abnormalities of non-verbal auditory processing have been most consistently documented in the canonical non-fluent variant (nfvPPA) and semantic variant (svPPA) syndromes of PPA. These syndromes have relatively distinct clinico-anatomical profiles [1, 14]: nfvPPA presents with impaired speech production and/or agrammatism associated with asymmetric, predominantly left-sided peri-Sylvian atrophy; while svPPA characteristically presents with vocabulary loss and impaired word comprehension associated with asymmetric anterior temporal lobe atrophy. Consistent with these syndromic profiles, nfvPPA is associated with more prominent deficits of early perceptual auditory analysis including impaired temporal (rhythm) perception, while svPPA is particularly associated with auditory associative deficits and impaired sound meaning [4–6, 9–13]. The processing of certain auditory information (such as higher-order spectrotemporal statistics) is affected in both syndromes [5, 13]: this may reflect a fundamental computational deficit affecting the linkage of perceptual and semantic auditory object data [15, 16]. However, most studies of auditory processing in PPA have focused on non-verbal sounds and elementary acoustic patterns, rather than the acoustic analysis of speech signals per se. Moreover, the brain substrates that mediate auditory processing in PPA largely remain to be defined.

We have demonstrated behavioural and neuroanatomical correlates of the defective analysis of generic speech signal attributes in two canonical PPA syndromes. In line with previous neuropsychological evidence concerning the processing of non-verbal sounds in PPA [5, 6, 9–13], processing of speech signal temporal regularity (an early perceptual property) was impaired in the patient group with nfvPPA, while processing of phonemic spectral structure and prosodic predictability (higher-order auditory properties) was impaired in both patient groups. Taken together, our findings substantiate an emerging picture of more generic, extra-linguistic deficits that may contribute to the hallmark neurolinguistic syndromes of PPA. The psychoacoustic deficits identified in our patient cohort had separable structural neuroanatomical substrates within distributed cerebral cortico-subcortical networks previously implicated in the analysis of auditory object and multimodal sensory information [17, 20–28].

From a clinical perspective, our findings show that generic auditory processing deficits in PPA syndromes extend to the processing of speech signals and suggest that such deficits may correlate with overall disease severity as well as standard measures (here, phonemic discrimination) of linguistic competence in these syndromes. With respect to the nosology of PPA, these findings suggest that certain measures of speech signal analysis (such as temporal coding) may stratify syndromes, whereas other measures (such as spectral and statistical coding) may cross conventional syndrome boundaries. These behavioural measures capture regional atrophy within a distributed fronto-temporo-parietal network that overlaps but extends beyond canonical language areas (compare Table 2 and Additional file 11), involving striatal structures implicated in non-verbal pattern decoding. This study requires substantiation in larger patient cohorts, ideally with longitudinal tracking of deficits and, ultimately, histopathological correlation. The relations between linguistic and pre-linguistic impairment in PPA will only be fully defined through more comprehensive neuropsychological correlation and functional neuroimaging techniques that address underlying neural mechanisms directly [17]. We regard the present work as a prima-facie case for the systematic exploration of non-verbal signal processing functions in PPA, with a view to re-evaluating conventional syndrome definitions and new biomarker discovery.




Leave a Reply

Your email address will not be published.