ETA: On further thought I think that’s the paper I was looking for after all, I was just thrown off by the reference to Monty Hall for some reason. My thanks.
Sadly, that isn’t the paper I was looking for. I found a vague reference here but it looks like I either made the experiment up entirely, it uses rats or mice instead of pigeons (I could have sworn it was pigeons though!), or it was on another website (unlikely—this kind of topic is far more likely to be on LW than anything else I read).
Hm, I was thinking you probably meant the experiment talked about in this old Language Log post. Not sure where an original reference might be found, though.
Ha! I didn’t think to check Language Log, and that post does look very familair. I must have attributed it to LW based on the subject matter.
Upvoted for some sound detective work, I think I have enough information to go reference chasing now :)
Oh and in case anyone was wondering: I was looking for this experiment because I’m currently writing about the difference between child langage acquisition and adult second language acquisition, and some experiments I came across showed that exact same pattern—that the children will quickly over- or undergeneralise based on the data, while the adults engage in probability matching.
Cognitive and linguistic development is something that continues over our whole lives, but at the earlier ages there are a whole lot of basic concepts that aren’t in place yet, like theory of mind (anecdotally, most roleplayers who try to roleplay with their young <10 year old kids find that the kids have trouble pretending to be someone else), and there are lots of linguistic structures that have certain thresholds of complexity.
So children pick up new linguistic structures at around the same rate as they develop the cognitive machinery to deal with them, and they try to regularise everything. Adults on the other hand already have all that machinery in place, plus they already know what language is supposed to be like based on prior experience, so a lot of structures are much easier for them to acquire (and interesting, the order of acquisition for adults in a particular language is largely the same regardless of their first language, so it’s prior experience with language in general that’s important, not experience with similar structures). Oh, and there’s some not-yet-repeated results that also say that education/literacy is a big factor in individual linguistic complexity, which would probably have knock-on effects for acquisition.
the order of acquisition for adults in a particular language is largely the same regardless of their first language, so it’s prior experience with language in general that’s important, not experience with similar structures
Suppose language X inflects verbs for present/past tense, but has no definite article, whereas for language Y, it’s the other way around. Assume now the a native speaker of X and a native speaker of Y are learning some third language Z, which has both features. Are you saying that both learners are going to acquire the use of the article and the past tense in Z in the same order?
This sounds strikingly implausible to me, though I might be wrong.
According to the book I’m referencing from, one of the studies was other language → English, where the other languages were: Greek, Persian, Italian, Turkish, Japanese, Chinese, Thai, Afghani, Hebrew, Arabic, Vietnames.So a decent spread of different types, with lots of points of similarity and difference between them and English. There have also been other studies showing similar findings in other L2 languages and from other researchers in general confirming those results, although this particular book doesn’t go into much detail about the others. Oh and the aggregate of the results also show that the order is the same regardless of the method/environment of acquisition, too.
Could you please write what book that is? I’d be curious to look it up.
One issue I see with the study (according to your description) is that English, by historical accident, doesn’t have any related languages that would be close enough to preserve some mutual intelligibility and make the analogous grammatical rules immediately obvious for their speakers (like e.g. between Slavic or Romance languages). I would expect that the conclusion doesn’t hold when comparing with learners speaking a closely related vs. a distant or unrelated language.
Another issue is how “acquisition” is defined. How accurate does one’s use of a particular grammatical feature have to be so that it qualifies as “acquired”? For many subtler grammatical rules, nearly all adult learners will never master them fully; for example, I still don’t know how to use the English definite article accurately, and neither does any other non-native English speaker I know. Formulating these criteria in a way that makes sense seems like a very tricky problem.
ok, the main book is “Second language learning theories” by Rosamond Mitchell and Florence Myles, p32-33 in my library copy. The main studies you’ll want to fact check are by Dulay and Burt (various years and publications) for children acquiriing a second language. Bailey, Madden and Krashen (1974) for the Dulay and Burt results replicated in adults, and (from a different textbook) Zobl and Liceras “functional categories and acquisition order” (1994) in the journal language learning which I seem to remember provides a good summary of the studies to date
I would expect that the conclusion doesn’t hold when comparing with learners speaking a closely related vs. a distant or unrelated language.
Probably not, no, but I did find it striking that the results held across such typologically different languages. For example, English has plenty of derivational morphology and so do most European languages, but most Asian languages don’t. But the order or acquisition was still pretty much the same.
Another issue is how “acquisition” is defined
I believe the bar for ‘acquired’ is usually set somewhere around 80-95% accuracy, but I haven’t looked into this aspect much. Another sensible way of measuring it would be perfect or near-perfect use in common contexts, and ok accuracy in less common contexts
ETA: On further thought I think that’s the paper I was looking for after all, I was just thrown off by the reference to Monty Hall for some reason. My thanks.
Sadly, that isn’t the paper I was looking for. I found a vague reference here but it looks like I either made the experiment up entirely, it uses rats or mice instead of pigeons (I could have sworn it was pigeons though!), or it was on another website (unlikely—this kind of topic is far more likely to be on LW than anything else I read).
Hm, I was thinking you probably meant the experiment talked about in this old Language Log post. Not sure where an original reference might be found, though.
Ha! I didn’t think to check Language Log, and that post does look very familair. I must have attributed it to LW based on the subject matter.
Upvoted for some sound detective work, I think I have enough information to go reference chasing now :)
Oh and in case anyone was wondering: I was looking for this experiment because I’m currently writing about the difference between child langage acquisition and adult second language acquisition, and some experiments I came across showed that exact same pattern—that the children will quickly over- or undergeneralise based on the data, while the adults engage in probability matching.
This is a topic I’m really interested in. What arae you writing about it?
Cognitive and linguistic development is something that continues over our whole lives, but at the earlier ages there are a whole lot of basic concepts that aren’t in place yet, like theory of mind (anecdotally, most roleplayers who try to roleplay with their young <10 year old kids find that the kids have trouble pretending to be someone else), and there are lots of linguistic structures that have certain thresholds of complexity.
So children pick up new linguistic structures at around the same rate as they develop the cognitive machinery to deal with them, and they try to regularise everything. Adults on the other hand already have all that machinery in place, plus they already know what language is supposed to be like based on prior experience, so a lot of structures are much easier for them to acquire (and interesting, the order of acquisition for adults in a particular language is largely the same regardless of their first language, so it’s prior experience with language in general that’s important, not experience with similar structures). Oh, and there’s some not-yet-repeated results that also say that education/literacy is a big factor in individual linguistic complexity, which would probably have knock-on effects for acquisition.
erratio:
Suppose language X inflects verbs for present/past tense, but has no definite article, whereas for language Y, it’s the other way around. Assume now the a native speaker of X and a native speaker of Y are learning some third language Z, which has both features. Are you saying that both learners are going to acquire the use of the article and the past tense in Z in the same order?
This sounds strikingly implausible to me, though I might be wrong.
According to the book I’m referencing from, one of the studies was other language → English, where the other languages were: Greek, Persian, Italian, Turkish, Japanese, Chinese, Thai, Afghani, Hebrew, Arabic, Vietnames.So a decent spread of different types, with lots of points of similarity and difference between them and English. There have also been other studies showing similar findings in other L2 languages and from other researchers in general confirming those results, although this particular book doesn’t go into much detail about the others. Oh and the aggregate of the results also show that the order is the same regardless of the method/environment of acquisition, too.
Could you please write what book that is? I’d be curious to look it up.
One issue I see with the study (according to your description) is that English, by historical accident, doesn’t have any related languages that would be close enough to preserve some mutual intelligibility and make the analogous grammatical rules immediately obvious for their speakers (like e.g. between Slavic or Romance languages). I would expect that the conclusion doesn’t hold when comparing with learners speaking a closely related vs. a distant or unrelated language.
Another issue is how “acquisition” is defined. How accurate does one’s use of a particular grammatical feature have to be so that it qualifies as “acquired”? For many subtler grammatical rules, nearly all adult learners will never master them fully; for example, I still don’t know how to use the English definite article accurately, and neither does any other non-native English speaker I know. Formulating these criteria in a way that makes sense seems like a very tricky problem.
ok, the main book is “Second language learning theories” by Rosamond Mitchell and Florence Myles, p32-33 in my library copy. The main studies you’ll want to fact check are by Dulay and Burt (various years and publications) for children acquiriing a second language. Bailey, Madden and Krashen (1974) for the Dulay and Burt results replicated in adults, and (from a different textbook) Zobl and Liceras “functional categories and acquisition order” (1994) in the journal language learning which I seem to remember provides a good summary of the studies to date
Probably not, no, but I did find it striking that the results held across such typologically different languages. For example, English has plenty of derivational morphology and so do most European languages, but most Asian languages don’t. But the order or acquisition was still pretty much the same.
I believe the bar for ‘acquired’ is usually set somewhere around 80-95% accuracy, but I haven’t looked into this aspect much. Another sensible way of measuring it would be perfect or near-perfect use in common contexts, and ok accuracy in less common contexts
Thanks for the references!