Search Abstracts | Symposia | Slide Sessions | Poster Sessions
De novo motor learning in adults with developmental language disorder
Poster Session B, Wednesday, September 30, 4:30 - 6:30 pm, Wangari Maathai
Gabriel Cler1,2, Abigail Searle1, Graciela Knudsen1, Noelle Abbott1; 1Basque Center on Cognition, Brain and Language, 2University of Washington
Converging behavioral evidence suggests that learning is a primary deficit in developmental language disorder (DLD). Emerging neural evidence implicates the striatum, a subcortical structure key for both learning and motor control (Ullman et al., 2024). Here we evaluated learning and motor control with a non-linguistic, de novo visuomotor learning task in which skin-surface muscle activity captured with surface electromyography (sEMG) was mapped in real-time to cursor movements on a computer screen. Nine adult participants with DLD and 18 adult participants with typical development (TD; group determined through protocol from Fidler et al., 2011) completed two sessions of training with the sEMG cursor. sEMG sensors were placed on participants’ faces to capture activity of six facial muscles, mapped non-intuitively to six non-cardinal directions. Participants activated facial muscles to move the sEMG cursor and select pseudorandomly-placed targets across two sessions. When two or more sEMG channels exceeded thresholds, the cursor moved in the corresponding directions (360° control), with speed proportional to the magnitude of activation. Learning and performance were quantified with throughput, an index of movement time that accounts for target difficulty. Throughput was averaged across trial bins (15 trials per bin; 10 bins per session), and group performance was compared at key points with t-tests. Both groups had identical performance during the first bin (p = 0.98). Early learning rate was assessed by percent improvement from bin 1 to bin 3, which was significantly greater in the TD group compared to DLD (87.2% vs. 12.9%; p < 0.001). To evaluate later learning in Session 1, the slope of throughput across bins 4–10 was compared and did not differ between groups (p = 0.33), indicating convergence in learning rates. Retention was assessed as the percent change in throughput from the end of Session 1 to the start of Session 2, with no significant group difference observed (p = 0.57). These results indicate impairment in the initial learning of arbitrary sensorimotor mappings, followed by typical procedural learning (refining the motor activations; coordinating multiple muscles) and retention after the mappings have been discovered. Interestingly, there is evidence from animal literature (Wise & Murray, 1999) and fMRI studies in humans (Toni et al., 2001) that learning arbitrary sensorimotor mappings initially involves the hippocampus, while later learning becomes more reliant on the basal ganglia. We propose that this type of learning is analogous to word learning, as both involve the rapid formation of novel, initially fragile mappings between previously unrelated inputs and outputs. In both cases, early performance requires error-driven updating and transient representations that gradually stabilize into more efficient, automatized mappings through repetition and consolidation. Our findings are therefore in accordance with models of DLD suggesting a specific impairment of encoding (McGregor et al., 2017), while they contradict the procedural deficit hypothesis, which suggests that learning involving the medial temporal lobe should be spared and procedural learning should be impaired. Behavioral testing is still ongoing and task learning during fMRI scanning is planned.
Topic Areas: Disorders: Developmental, Language Development/Acquisition