Grammatical production deficits in PPA: Relating narrative and structured task performance

Introduction Grammatical production impairments in primary progressive aphasia (PPA) have been investigated using structured language tasks and analysis of narrative language samples (for review see Thompson & Mack, 2014; Wilson et al., 2012). However, little research has examined the relationship between them in PPA. Whereas structured tasks often assess production accuracy at different levels of syntactic complexity (e.g., Thompson et al., 2013), narrative measures typically assess overall lexical and grammatical usage (e.g., % grammatical sentences; noun-to-verb ratio), with lesser emphasis on complexity. The present study investigated the relationship between narrative measures of grammatical production and performance on structured language tests in the domains of syntax, verb morphology, and verb-argument structure (VAS). Materials and methods Data from 101 individuals with PPA were included. Participants completed a test battery including the Northwestern Assessment of Verbs and Sentences (NAVS, Thompson, 2011), the Northwestern Assessment of Verb Inflection (NAVI, Lee & Thompson, experimental version) and the Northwestern Anagram Test (NAT, Thompson, Weintraub, & Mesulam, 2012). Grammatical production deficits were quantified as follows: for syntax, accuracy of non-canonical sentence production on the NAVS Sentence Production Priming Test (SPPT) and the NAT; for morphology, the accuracy on finite verbs on the NAVI; for VAS, the accuracy of sentences produced with 2- and 3-argument verbs on the NAVS Argument Structure Production Test (ASPT). Cinderella narrative samples were analyzed using the Northwestern Narrative Language Analysis system (e.g., Thompson et al., 2012). For syntax, complexity was measured by the ratio of syntactically complex to simple sentences produced, whereas accuracy was indexed by computing the proportion of words with a locally grammatical lexical category. Morphological complexity was measured by mean number of verb-morphological elements (i.e., the mean verb morphology index), whereas morphological accuracy was measured by the proportion of correctly inflected verbs. VAS complexity was measured by the mean number of arguments produced per verb (i.e., the verb-lemma complexity index), whereas accuracy was assessed by the proportion of verbs with correct argument structure. Within each grammatical domain, multiple regression was used to determine if narrative measures were predictive of structured task performance. Results For syntax, measures of narrative complexity and accuracy significantly predicted performance on both the NAVS SPPT and NAT. Performance on the NAVI was predicted by the proportion of correctly inflected verbs produced in narratives, but not the verb morphology index. Similarly, VAS production on the NAVS ASPT was predicted by the measure of narrative accuracy, but not complexity. Scores obtained on structured tasks were all highly correlated, with the exception of NAVI and NAVS ASPT. Discussion The results indicate significant relationships between measures of grammatical production ability from structured tasks and narratives in PPA. Measures of narrative accuracy predicted structured task performance across all domains, suggesting that the two methods may yield similar results in quantifying grammatical impairments. However, measures of narrative complexity predicted structured task performance only in the syntactic domain, suggesting that the complexity of verb phrases produced in narratives may not reflect impairments in verb-morphology and VAS as measured by structured tasks.