He could not believe that Broca's patients' brains were preserved and in our possession, that Hindi/Urdu सब्ज़ी / سبزی sabzī 'vegetables' was etymologically related to Bangla সবুজ sôbuj 'green', and that Urdu قمیص qamīs 'kameez' was cognate with Portuguese/Spanish camisa 'shirt'
My neurologist is Bengali, so my appointment this afternoon was twice as long since he wanted to talk neuro shop with me and then ask me etymology questions about Perso-Arabic borrowings into Bangla
A very interesting research setup on grammatical differences and how brain processes them. Done with a couple of understudied languages, Hindi and Nepal. Finnish has a similar differential object marking depending on quantitative definiteness of the object.
And a thorough Bsky-summary, thanks!
I'm very proud of this work; to my wit, it's the first cog neuro study in Nepali, & the first MEG study in sentence processing in Hindi. Thanks to NYUAD and NeLLab for supporting this hare-brained idea, two *VERY* patient & thorough reviewers, & editor @kemmorey.bsky.social for making this possible!
A figure showing that MEG activity results in a reduced positive field around 300–500ms and a reduced negative field 650–900ms in Nepali for differential object-marking
A figure showing nearly the same sensor-space pattern as in Nepali but in Hindi. Except, we then found in a post hoc, exploratory source-space analysis a p = 0.09 cluster reflecting a difference in activity between the two languages, with a reduction in activity in supramarginal gyrus for object-agreement constructions in Hindi but not Nepali.
Here's the (sensor space) results showing a biphasic wave corresponding to differential object marking in Nepali (left) and Hindi (right). In the time window of the second 'phase' in the Hindi data, we did the exploratory source-space analysis and found the p = 0.09 cluster in supramarginal gyrus s
This is important since our field has been very enamored with one-size-fits-all approaches to language, like LLMs, that might blur the lines between these subtle grammatical differences between languages.
A lil slogan that I've started saying is that we have good theories of 'langauge in the brain', but I'm not sure we have good theories of 'languages in the brain', like our friends who study syntax and phonology might have.
This means we have to be very careful and clear about the details of the lgs that we study, the extent to which findings from lg A can (and should) be ported over to lg B, and that we need theories of language-in-the-brain that are flexible and detailed enough to 'fit' different grammatical systems
What *is* important, though, is the Marrian calculation that we present in this paper – the brain is instantiating an algorithm to process language, and the problem that that algorithm is trying to solve will differ depending on the grammatical properties of that language.
So, the main idea has not been demonstrated as easily as I thought it would be in this paper.
But, an admittedly very post hoc analysis did observe a n.s. cluster in the supramarginal gyrus that seemed to have a reduction in activity while processing an agreement-controlling object NP in Hindi, but not in Nepali. What's going on here? Stay tuned, I'm working on a 'sequel' paper.
Did we find that? Well... not quite. We did find pretty systematic MEG responses to differentially object marked NPs in both languages, so the aspects of the grammatical systems that are the same in the two languages elicit excitingly similar brain responses.
We reasoned that, somewhere, the brain should be doing something special while processing 'candy' in Hindi in 'the girl-ERG candy ate', since 'candy' controls agreement, that should NOT be occurring in 'the girl candy eats', or in any sentence in Nepali.
Previous work by Sakshi Bhatia and @linguistbrian.bsky.social showed that Hindi readers likely identify and encode the agreement controller NP in a special way in memory for the purposes of agreement computation. If so, how/where/when does that work in the brain?
A schematic showing that object NPs in Hindi agree with the verb in the perfective, if the object NP is not differentially object marked. In Nepali, there is always subject agreement.
But, in Hindi, these features interact not seen in Nepali; in Hindi, the verb agrees with the highest NP that does not have a case ending. So, 'the girl candy eats' has subject-verb agreement, 'the girl-ERG candy ate' has OBJECT-verb agreement, and 'the girl-ERG candy-ACC ate' has no agreement.
Both languages also have differential object marking, such that specific/definite objects are marked with an accusative morpheme – 'the girl candy-ACC eats' means something like 'the girl eats *THE* candy'.
This paper had a very ambitious hypothesis that didn't quite pan out the way we expected. We looked at simple SOV sentences in Hindi and Nepali, like 'the girl candy eats'. Both languages have aspect-based split ergativity, so the perfect has a case morpheme on the subject: 'the girl-ERG candy ate.'
In other news, the proofs are up for my new paper, with a dream team of Subhs Shrestha, @linguistbrian.bsky.social, Diogo Almeida, Alec Marantz, and Rajesh Bhatt!
direct.mit.edu/nol/article/...
This is my second time teaching it, and each year, each project has been (to my knowledge) a genuinely novel finding and a potentially unique contribution to psycholinguistics.
Although all of these are really preliminary results collected from a convenience sample (i.e., the rest of the Psycholing 2 class, plus roommates/friends/families), I encourage them to try to pursue these as genuine research projects.
Lastly, one group showed that the 'sentence superiority effect' – people's ability to recall a short phrase ('dogs bark') flashed ~200ms is greater than a non-phrase ('dogs cats') – persists even when one of the words is presented far in the parafovea and thus outside the center of visual attention
The other group found that exocentric compounds show semantic priming mostly for the first stem in the word more than the second stem; 'airhead' primes words associated with 'air' but not 'head' or 'idiot'. So, not all stems are the same in compounds (in English)
Two projects were on morphology. One group found that morphologically well-formed-but-absent wordforms ('gloriosity', 'badder', 'Japanish') have exactly the same priming patterns as their present alternatives ('glory', 'worse', 'Japanese')
The second showed that the famous effect of PP attachment on ambiguous reference ('put the frog on the napkin...') can be elicited with linguistic context – it suffices to have a plural NP ('frogs') prior in the discourse; no visual world needed!
Two projects were on ambiguities. One project showed that 'lingering' garden path effects (reported by Ferreira et al.) are not observed for local coherence effects, i.e., not all ambiguities are the same.
A photo of myself and the class (approximately 20 undergraduate linguistics students) standing in front of two posters.
At UCSC, our Psycholing 2 is a crash course on sentence processing for about 5 weeks, and then a practicum to do a novel research project for another 5 weeks. The 'final' is a poster presentation where each group gets to learn about what each other did.
New team:
Yamper x 3
Pombon x 3
