Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Constructed languages are processed by the same brain mechanisms as natural languages
Poster D127 in Poster Session D, Wednesday, October 25, 4:45 - 6:30 pm CEST, Espace Vieux-Port
Saima Malik-Moraleda1,2, Maya Taliaferro2, Steve Shannon2, Niharika Jhingan2, Sara Swords2, David Peterson, Paul Frommer, Marc Okrand, Jessie Sams, Ramsey Cardwell3, Cassie Freeman3, Evelina Fedorenko1,2; 1Harvard University, 2Massachusetts Institute of Technology, 3Duolingo
Many domains share similarities with language and have been argued to draw on the same or overlapping neural mechanisms: for example, like language, math and logical reasoning require manipulation of symbolic representations; like language, music is highly structured; and like language, gestures are a critical communication channel. Despite these similarities, these and many other information systems do not engage the brain mechanisms that support language comprehension (e.g., Fedorenko et al., 2011; Monti et al., 2012; Ivanova et al., 2020; Chen et al., 2023). In the present study, we examined a new domain, previously untested: constructed languages, or conlangs, like Esperanto or Klingon. Do we use our language-processing mechanisms to understand conlangs? Or does conlang processing instead draw on the system that supports our processing of math, logic, and computer languages: the Multiple Demand system (e.g., Duncan, 2010; Monti et al., 2012; Ivanova et al., 2020; Amalric & Dehaene, 20xxNeuroimage)? We tested 38 proficient speakers of conlangs: Esperanto (n=19), Klingon (n=10), Na’vi (n=8), High Valyrian (n=3), and Dothraki (n=3; five participants were proficient in 2 conlangs). While in an fMRI scanner, participants performed a language localizer task (Fedorenko et al., 2010) in their native language (English for most participants) and a critical conlang comprehension task, where they listened to conlang sentences vs. to acoustically degraded version of those sentences (e.g., Malik-Moraleda, Ayyash et al., 2022). Additionally, each participant completed a Multiple Demand network localizer based on a spatial working memory task (e.g., Fedorenko et al., 2013; Shashidara et al., 2019). Using individual-subject fMRI analyses, we identified language functional regions of interest (fROIs) in each participant using their native-language localizer, and then examined the responses to the conlang task conditions. We found that both frontal and temporal language fROIs respond robustly to conlang comprehension, evidenced by a large and significant conlang>degraded effect. The magnitude of response to conlang sentence condition is similar to the magnitude of response to the participant’s native language, and a Dice coefficient analysis on the individual-level whole-brain maps revealed strong overlap between the language localizer contrast in the native language vs. in the conlang. In contrast, the Multiple Demand network fROIs showed little to no response during conlang comprehension. The results hold across all five conlangs, which can be construed as five independent replications. To summarize: Constructed languages like natural languages, can express a wide range of meanings related to the external and internal worlds. On the other hand, conlangs differ in critical ways: they were created much more recently, often for different (sometimes esoteric), purposes, and are not shaped by learning and processing pressures, at least not to the same degree (e.g., Okrent, 2010; Goodall, 2022). In spite of these differences all five conlangs recruit the same brain mechanisms as those that support comprehension of natural languages, which suggests that the language network is robust to variation along these dimensions.
Topic Areas: Writing and Spelling, Multisensory or Sensorimotor Integration