Presentation

Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Poster Slams

A connectivity constrained analysis of the white matter correlates of taxonomic and thematic based naming errors

Poster E57 in Poster Session E, Saturday, October 8, 3:15 - 5:00 pm EDT, Millennium Hall

Alexander M Swiderski^1,2,3, Alexis Laconi^1,2, Emily B Goldberg^1,2,3, Michael Walsh Dickey^1,2,3, William D Hula^2.1; ¹University of Pittsburgh, ²VA Pittsburgh Healthcare System, ³Carnegie Mellon University

Successful word production relies on retrieval of semantic concepts[1–3].which is supported by neurobiological networks of language[4,5] and semantic memory[6–9]. Damage to these overlapping networks results in semantic naming (word retrieval) deficits, e.g., ‘cat’ for ‘dog’. Semantic naming impairments following stroke-induced aphasia are associated with lesions in anterior temporal lobe (ATL), prefrontal cortex, and temporo-parietal junction[10,11] (TPJ). Schwartz and colleagues[12] examined lesion correlates of taxonomic (feature-based: ‘cat’ for ‘dog’) and thematic naming errors (association-based: ‘leash’ for ‘dog’) in a large left-hemisphere stroke sample (n=88) and found all participants made more taxonomic than thematic errors and that these errors were uniquely associated with the ATL and TPJ, respectively. The current study investigated the neural substrates of taxonomic and thematic errors, expanding on previous work in two novel ways. First, it used connectometry[13] to investigate the white-matter correlates of these error types, rather than their cortical substrates[12]. Second, it used Resnik scores[14] and Point Wise Mutual Information (PMI) methods to objectively and reproducibly classify taxonomic and thematic errors, rather than using human classification of these error types [12]. Methods: Data from twenty-nine out of a larger sample of 60 people with aphasia and unilateral left-hemisphere lesion have been analyzed to date (mean months post onset: 81.0; sd: 70.5). Participants’ aphasia severity ranged from severe to mild[15]. Naming responses were collected using the Philadelphia Naming Test[16]. Resnik[14] scores and PMI were estimated with the Natural Language Processing Toolkit[17] implemented in python with text corpora exceeding 1-million unique lemmas[18,19]. These scores were used to classify semantic naming errors as taxonomic vs. thematic: high Resnik and low PMI reflect taxonomic relations between target and naming error, whereas the opposite pattern is consistent with thematic relation between target and error. Deterministic connectometry[13] was used to derive nonparametric Spearman correlations between Quantitative Anisotropy (QA; the amount of anisotropic spins diffusing along fiber orientations[20]) and logit-transformed number of taxonomic and thematic errors. Results: A ratio of 2.4 to 1 taxonomic to thematic errors was observed across participants, with a significant difference between the mean number of taxonomic and thematic errors. Three participants produced more thematic than taxonomic errors. Connectometry analyses Taxonomic errors were associated with dorsal tracts including the L-Arcuate Fasciculus (AF), L-Fronto-Parietal cingulum (FPC) and ventral tracts including the L-Inferior (ILF) and L-Middle Longitudinal (MdLF) fasciculi, and the L/R-Inferior Frontal Occipital (IFOF) Thematic errors were associated with dorsal (L-AF, L/R-FPC) and ventral (L-ILF, L-MdLF, L-IFOF, and L-Extreme Capsule) tracts. Discussion: The present findings are partially consistent with those of Schwartz and colleagues[12]. First, we did not observe a one-way dissociation between taxonomic and thematic errors, with more taxonomic than thematic errors across all participants. Second, we found dorsal pathways (L-AF) terminating in the TPJ and ventral pathways terminating in the ATL (IFOF and ILF) that were associated with both taxonomic and thematic errors. We attribute these differences to the objective measurements utilized to identify taxonomic and thematic relations within this study. Additional analyses will determine whether these associations hold in the full sample of 60 participants.

Topic Areas: Meaning: Lexical Semantics, Disorders: Acquired