AbouBenAdhem@lemmy.world
on 18 Mar 2025 15:09
nextcollapse
In previous work, Fedorenko and her students have found that computer programming languages, such as Python — another type of invented language — do not activate the brain network that is used to process natural language. Instead, people who read computer code rely on the so-called multiple demand network, a brain system that is often recruited for difficult cognitive tasks.
I’m curious if there’s some overlap between conlangs and programming languages, on the region level if not the network level. IIRC, the multiple demand network is a bit ill-defined and every region doesn’t necessarily activate for every task; and Fedorenko et al have their own idiosyncratic definition of the language network that omits anything that might also have other functions (including canonical regions like Broca’s and Wernike’s areas).
I’m curious if there’s some overlap between conlangs and programming languages, on the region level if not the network level.
I predict that, if there is such overlap, natural languages will also overlap the same way. Because in practice there’s no individual difference between a highly developed conlang and a natlang.
AbouBenAdhem@lemmy.world
on 18 Mar 2025 17:51
collapse
Sure.
I suspect the motivation behind this study was to try to narrow down the deciding factor in the earlier study showing a difference between natural and programming languages—the next logical step would be looking at a more “artificial” conlang like Lojban (and/or a more “natural” programming language like ACE).
The end result will probably be some broader category of “language-network-interpretable” languages including natural and (some but maybe not all) conlangs.
All of those constructed languages were modelled after natural languages, either to give a tongue to some fictitious people (most of the listed) or to perform like an auxiliary language (Esperanto). So the difference between conlang and natlang here is just one of origin.
In the meantime, those programming sets of instructions (Python, C, etc.) were created for something else, issuing instructions in code. On the very best you could analyse them as doing a fraction of what a language does, but in practice it’s something else entirely.
AbouBenAdhem@lemmy.world
on 18 Mar 2025 19:01
collapse
I suspect their ultimate goal is to confirm or refute the common theory that there’s a general mental faculty for recursion that’s used both for natural language and for other recursive tasks (implying that language and recursive thought evolved together).
If that’s their goal then it was a really dumb idea to pick these conlangs. It simply won’t show any surprising data, since all of those languages implement recursion in one or another way.
AbouBenAdhem@lemmy.world
on 18 Mar 2025 22:02
collapse
all of those languages implement recursion in one or another way
Yeah—Python and English are both recursive, so that doesn’t account for why the brain processes them differently. But they need to figure out what other feature does account for it—ideally by finding a pair of (probably artificial) languages that differ only in the exact feature which triggers the language network. Then they can figure out how that feature relates to recursion or any other mental abilities that might have co-evolved with language.
threaded - newest
I’m curious if there’s some overlap between conlangs and programming languages, on the region level if not the network level. IIRC, the multiple demand network is a bit ill-defined and every region doesn’t necessarily activate for every task; and Fedorenko et al have their own idiosyncratic definition of the language network that omits anything that might also have other functions (including canonical regions like Broca’s and Wernike’s areas).
I predict that, if there is such overlap, natural languages will also overlap the same way. Because in practice there’s no individual difference between a highly developed conlang and a natlang.
Sure.
I suspect the motivation behind this study was to try to narrow down the deciding factor in the earlier study showing a difference between natural and programming languages—the next logical step would be looking at a more “artificial” conlang like Lojban (and/or a more “natural” programming language like ACE).
The end result will probably be some broader category of “language-network-interpretable” languages including natural and (some but maybe not all) conlangs.
I’m not surprised.
All of those constructed languages were modelled after natural languages, either to give a tongue to some fictitious people (most of the listed) or to perform like an auxiliary language (Esperanto). So the difference between conlang and natlang here is just one of origin.
In the meantime, those programming sets of instructions (Python, C, etc.) were created for something else, issuing instructions in code. On the very best you could analyse them as doing a fraction of what a language does, but in practice it’s something else entirely.
I suspect their ultimate goal is to confirm or refute the common theory that there’s a general mental faculty for recursion that’s used both for natural language and for other recursive tasks (implying that language and recursive thought evolved together).
If that’s their goal then it was a really dumb idea to pick these conlangs. It simply won’t show any surprising data, since all of those languages implement recursion in one or another way.
Yeah—Python and English are both recursive, so that doesn’t account for why the brain processes them differently. But they need to figure out what other feature does account for it—ideally by finding a pair of (probably artificial) languages that differ only in the exact feature which triggers the language network. Then they can figure out how that feature relates to recursion or any other mental abilities that might have co-evolved with language.