The Three Terms Task - an open benchmark to compare human and artificial semantic representations.

Word processing entails retrieval of a unitary yet multidimensional semantic representation (e.g., a lemon's colour, flavour, possible use) and has been investigated in both cognitive neuroscience and artificial intelligence. To enable the direct comparison of human and artificial semantic representations, and to support the use of natural language processing (NLP) for computational modelling of human understanding, a critical challenge is the development of benchmarks of appropriate size and complexity. Here we present a dataset probing semantic knowledge with a three-terms semantic associative task: which of two target words is more closely associated with a given anchor (e.g., is lemon closer to squeezer or sour?). The dataset includes both abstract and concrete nouns for a total of 10,107 triplets. For the 2,255 triplets with varying levels of agreement among NLP word embeddings, we additionally collected behavioural similarity judgments from 1,322 human raters. We hope that this openly available, large-scale dataset will be a useful benchmark for both computational and neuroscientific investigations of semantic knowledge.

Naming unique entities in the semantic variant of primary progressive aphasia and Alzheimer's disease: Towards a better understanding of the semantic impairment.

While the semantic variant of primary progressive aphasia (svPPA) is characterized by a predominant semantic memory impairment, episodic memory impairments are the clinical hallmark of Alzheimer's disease (AD). However, AD patients also present with semantic deficits, which are more severe for semantically unique entities (e.g. a famous person) than for common concepts (e.g. a beaver). Previous studies in these patient populations have largely focused on famous-person naming. Therefore, we aimed to evaluate if these impairments also extend to other semantically unique entities such as famous places and famous logos. In this study, 13 AD patients, 9 svPPA patients, and 12 cognitively unimpaired elderly subjects (CTRL) were tested with a picture-naming test of non-unique entities (Boston Naming Test) and three experimental tests of semantically unique entities assessing naming of famous persons, places, and logos. Both clinical groups were overall more impaired at naming semantically unique entities than non-unique entities. Naming impairments in AD and svPPA extended to the other types of semantically unique entities, since a CTRL>AD>svPPA pattern was found on the performance of all naming tests. Naming famous places and famous persons appeared to be most impaired in svPPA, and both specific and general semantic knowledge for these entities were affected in these patients. Although AD patients were most significantly impaired on famous-person naming, only their specific semantic knowledge was impaired, while general knowledge was preserved. Post-hoc neuroimaging analyses also showed that famous-person naming impairments in AD correlated with atrophy in the temporo-parietal junction, a region functionally associated with lexical access. In line with previous studies, svPPA patients' impairment in both naming and semantic knowledge suggest a more profound semantic impairment, while naming impairments in AD may arise to a greater extent from impaired lexical access, even though semantic impairment for specific knowledge is also present. These results highlight the critical importance of developing and using a variety of semantically-unique-entity naming tests in neuropsychological assessments of patients with neurodegenerative diseases, which may unveil different patterns of lexical-semantic deficits.

Semantic dementia and persisting Wernicke's aphasia: linguistic and anatomical profiles.

Few studies have directly compared the clinical and anatomical characteristics of patients with progressive aphasia to those of patients with aphasia caused by stroke. In the current study we examined fluent forms of aphasia in these two groups, specifically semantic dementia (SD) and persisting Wernicke's aphasia (WA) due to stroke. We compared 10 patients with SD to 10 age- and education-matched patients with WA in three language domains: language comprehension (single words and sentences), spontaneous speech and visual semantics. Neuroanatomical involvement was analyzed using disease-specific image analysis techniques: voxel-based morphometry (VBM) for patients with SD and overlays of lesion digitized lesion reconstructions in patients with WA. Patients with SD and WA were both impaired on tasks that involved visual semantics, but patients with SD were less impaired in spontaneous speech and sentence comprehension. The anatomical findings showed that different regions were most affected in the two disorders: the left anterior temporal lobe in SD and the left posterior middle temporal gyrus in chronic WA. This study highlights that the two syndromes classically associated with language comprehension deficits in aphasia due to stroke and neurodegenerative disease are clinically distinct, most likely due to distinct distributions of damage in the temporal lobe.

Reading disorders in primary progressive aphasia: a behavioral and neuroimaging study.

Previous neuropsychological studies on acquired dyslexia revealed a double dissociation in reading impairments. Patients with phonological dyslexia have selective difficulty in reading pseudo-words, while those with surface dyslexia misread exception words. This double dissociation in reading abilities has often been reported in brain-damaged patients, but it has not been consistently shown in patients with neurodegenerative diseases. In this study, we investigated reading impairments and their anatomical correlates in various neurodegenerative diseases. First, we performed a behavioral analysis to characterize the reading of different word types in primary progressive aphasia (PPA). Then, we conducted a voxel-based morphometry neuroimaging study to map the brain areas in which gray matter volume correlated with the accurate reading of exception and pseudo-words. The results showed a differential pattern of exception and pseudo-word reading abilities in different clinical variants of PPA. Patients with semantic dementia, a disorder characterized by selective loss of semantic memory, revealed a pattern of surface dyslexia, while patients with logopenic/phonological progressive aphasia, defined by phonological loop deficits, showed phonological dyslexia. Neuroimaging results showed that exception word reading accuracy correlated with gray matter volume in the left anterior temporal structures, including the temporal pole, the anterior superior and middle temporal and fusiform gyri, while pseudo-word reading accuracy correlated with left temporoparietal regions, including the posterior superior and middle temporal and fusiform gyri, and the inferior parietal lobule. These results suggest that exception and pseudo-word reading not only rely upon different language mechanisms selectively damaged in PPA, but also that these processes are sustained by separate brain structures.

Atrophy progression in semantic dementia with asymmetric temporal involvement: a tensor-based morphometry study.

We performed a longitudinal anatomical study to map the progression of gray matter atrophy in anatomically defined predominantly left (LTLV) and right (RTLV) temporal lobe variants of semantic dementia (SD). T1-weighted MRI scans were obtained at presentation and one-year follow-up from 13 LTLV, 6 RTLV, and 25 control subjects. Tensor-based morphometry (TBM) in SPM2 was applied to derive a voxel-wise estimation of regional tissue loss over time from the deformation field required to warp the follow-up scan to the presentation scan in each subject. When compared to controls, both LTLV and RTLV showed significant progression of gray matter atrophy not only within the temporal lobe most affected at presentation, but also in the controlateral temporal regions (p<0.05 FWE corrected). In LTLV, significant progression of volume loss also involved the ventromedial frontal and the left anterior insular regions. These results identified the anatomic substrates of the previously reported clinical evolution of LTLV and RTLV into a unique 'merged' clinical syndrome characterized by semantic and behavioral deficits and bilateral temporal atrophy.

The logopenic/phonological variant of primary progressive aphasia.

OBJECTIVE: Primary progressive aphasia (PPA) is characterized by isolated decline in language functions. Semantic dementia and progressive nonfluent aphasia are accepted PPA variants. A "logopenic" variant (LPA) has also been proposed, but its cognitive and anatomic profile is less defined. The aim of this study was to establish the cognitive and anatomic features of LPA. METHODS: Six previously unreported LPA cases underwent extensive neuropsychological evaluation and an experimental study of phonological loop functions, including auditory and visual span tasks with digits, letters, and words. For each patient, a voxel-wise, automated analysis of MRI or SPECT data were conducted using SPM2. RESULTS: In LPA, speech rate was slow, with long word-finding pauses. Grammar and articulation were preserved, although phonological paraphasias could be present. Repetition and comprehension were impaired for sentences but preserved for single words, and naming was moderately affected. Investigation of phonological loop functions showed that patients were severely impaired in digit, letter, and word span tasks. Performance did not improve with pointing, was influenced by word length, and did not show the normal phonological similarity effect. Atrophy or decreased blood flow was consistently found in the posterior portion of the left superior and middle temporal gyri and inferior parietal lobule. CONCLUSIONS: Logopenic progressive aphasia (LPA) is a distinctive variant of primary progressive aphasia. Cognitive and neuroimaging data indicate that a deficit in phonological loop functions may be the core mechanism underlying the LPA clinical syndrome. Recent studies suggest that Alzheimer disease may be the most common pathology underlying the LPA clinical syndrome.

The anatomy of category-specific object naming in neurodegenerative diseases.

Neuropsychological studies suggest that knowledge about living and nonliving objects is processed in separate brain regions. However, lesion and functional neuroimaging studies have implicated different areas. To address this issue, we used voxel-based morphometry to correlate accuracy in naming line drawings of living and nonliving objects with gray matter volumes in 152 patients with various neurodegenerative diseases. The results showed a significant positive correlation between gray matter volumes in bilateral temporal cortices and total naming accuracy regardless of category. Naming scores for living stimuli correlated with gray matter volume in the medial portion of the right anterior temporal pole, whereas naming accuracy for familiarity-matched nonliving items correlated with the volume of the left posterior middle temporal gyrus. A previous behavioral study showed that the living stimuli used here also had in common the characteristic that they were defined by shared sensory semantic features, whereas items in the nonliving group were defined by their action-related semantic features. We propose that the anatomical segregation of living and nonliving categories is the result of their defining semantic features and the distinct neural subsystems used to process them.

Anatomical correlates of early mutism in progressive nonfluent aphasia.

Patients with progressive nonfluent aphasia (PNFA) can become mute early in the course of the disease. Voxel-based morphometry showed that PNFA is associated with left anterior insula and inferior frontal atrophy. In PNFA with early mutism, volume loss was more prominent in the pars opercularis and extended into the left basal ganglia. Damage to the network of brain regions involved in both coordination and execution of speech causes mutism in PNFA.

Diffusion tensor imaging and voxel based morphometry study in early progressive supranuclear palsy.

BACKGROUND: A comprehensive characterisation of grey and white matter changes in progressive supranuclear palsy (PSP), the second most common extrapyramidal syndrome after Parkinson disease, is still not available. OBJECTIVE: To evaluate grey and white matter changes in mild PSP patients by voxel based morphometry (VBM) and diffusion tensor imaging (DTI), respectively. METHODS: 14 mild PSP patients and 14 healthy controls entered the study and underwent a clinical and neuropsychological evaluation according with a standardised assessment. Each subject had a structural magnetic resonance imaging (MRI) study. Processing analysis of MRI data was carried out according to optimised VBM and fractional anisotropy was determined. RESULTS: Compared with the controls, in PSP patients VBM analysis showed a significant clusters of reduced grey matter in premotor cortex, frontal operculum, anterior insula, hippocampus, and parahippocampal gyrus, bilaterally. With regard to subcortical brain regions, the pulvinar, dorsomedial and anterior nuclei of the thalamus, and superior and inferior culliculum were affected bilaterally. A bilateral decrease in fractional anisotropy in superior longitudinal fasciculus, anterior part of corpus callosum, arcuate fascicolus, posterior thalamic radiations, and internal capsule, probably involving the cortico-bulbar tracts, was present in PSP patients. CONCLUSIONS: These data provide evidence for both grey and white matter degeneration in PSP from the early disease stage. These structural changes suggest that atrophy of cortical and subcortical structures and neurodegeneration of specific fibre tracts contribute to neurological deficits in PSP.

Regional reductions of gray matter volume in familial dyslexia.

An in vivo anatomic study of gray matter volume was performed in a group of familial dyslexic individuals, using an optimized method of voxel-based morphometry. Focal abnormalities in gray matter volume were observed bilaterally in the planum temporale, inferior temporal cortex, and cerebellar nuclei, suggesting that the underlying anatomic abnormalities may be responsible for defective written language acquisition in these subjects.