Phonotactic awareness deficit following left-hemisphere stroke

• ¹ Georgetown University Medical Center, Neurology Department, United States
• ² MedStar National Rehabilitation Hospital, United States
• ³ NIDCD, Brain Imaging and Modeling Section, United States

Introduction Sublexical processing is an important sub-process involved in speech perception. Despite the amount of effort devoted to the study of sublexical processing, the neurobiology of this process has remained unclear. Although the left hemisphere has been suggested to play an important role in sublexical processing, neuroimaging studies have mostly reported bilateral cortical activations associated with this process (Gow, 2012; Hickok & Poeppel, 2004, 2007; Majerus et al., 2002; Scott & Wise, 2004; Turkeltaub & Branch Coslett, 2010). Studies examining sublexical speech processing in aphasic patients can shed lights on its cortical lateralization. However, there are few such studies in the literature. The aim of this study was to determine whether sublexical speech processing might be affected by left-hemisphere lesions. Specifically, we tested the hypothesis that individuals with left-hemisphere stokes might be less sensitive to phonotactic regularities compared to healthy adults. Methods Participants included 34 left-hemisphere stroke survivors aged 37 to 77 years (M =59.27, SD =12.49), and 27 age-matched healthy participants aged 28 to 83 years (M =59.93, SD =12.36). All participants were native speakers of English. Patients included in the study had left-hemisphere ischemic or hemorrhagic stroke at least 6 months prior to the study. Participants completed an “Englishness”-rating experiment adapted from Dewitt (2013). They were required to listen to a series of non-words and rate, on a 5-point likert scale, how similar each non-word sounded to a plausible English word. Stimuli included 60 six-phoneme non-words, divided into three phonotactic regularity conditions ¬(ranging from 1, irregular in all syllables, to 3, regular in all syllables). Results and Discussion Likert-type scale responses were z-transformed and coded accurate for positive z-values in condition 3 and negative z-values in condition 1 trials. Accuracy was analyzed using binomial mixed effects models and z-transformed scale responses were analyzed using linear mixed effects models. For both analyses, the fixed effects of stimulus, trial number, group (patient/control), education, age, response time, phonotactic regularity (1/3), and gender were examined along with all relevant interactions. Random effects for participant and stimuli as well as random slopes were also included. Model fitting was performed in a backward-stepwise iterative fashion, followed by forward fitting of maximal random effects structure. Models were evaluated by model fitness comparisons using Akaike Information Criterion and Bayesian Information Criterion. Accuracy analysis revealed that healthy participants were significantly more accurate than patients [β = 0.47, p<0.001] in Englishness rating. Scale response analysis revealed a significant effect of phonotactic regularity [β = 1.65, p<0.0001] indicating that participants were sensitive to phonotactic regularity differences among non-words. However, the significant interaction of group and phonotactic regularity [β = -0.5, p= 0.02] further demonstrated that, compared to healthy adults, patients were less able to recognize the phonotactic regularity differences between non-words. Results suggest that left-hemisphere lesions cause impaired phonotactic processing and that the left hemisphere might be necessary for phonotactic awareness. These preliminary findings will be followed up by further analyses investigating the interactions between phonotactic processing and participants’ scores on other linguistic/cognitive tasks as well as lesion-symptom mapping.