Read on Twitter
Are some languages more easily learned than others?What makes a language easy to learn?In a new study /w @mariannedhk & Meyer, we tested whether language learnability is shaped by community size and the degree of systematic grammar.
<Thread
> https://www.sciencedirect.com/science/article/pii/S0010027721000391
First, some background.
There's a wide-spread axiom in the field of linguistics, which states that all languages are equally difficult to learn and take the same effort to acquire. But is this true? Kids can clearly master every language, but it doesn't mean it's always as easy..
There's a wide-spread axiom in the field of linguistics, which states that all languages are equally difficult to learn and take the same effort to acquire. But is this true? Kids can clearly master every language, but it doesn't mean it's always as easy..
Actually, many scholars & language learners share a basic intuition that some languages are just harder to learn.
Be it by children (who indeed acquire some languages slower than others, e.g., @PuzzleOfDanish), or by adults (who struggle more with some languages, e.g., #Georgian)
Be it by children (who indeed acquire some languages slower than others, e.g., @PuzzleOfDanish), or by adults (who struggle more with some languages, e.g., #Georgian)
(We review evidence for this in the paper, so don't just take my word for it!)
But WHY are some languages harder to learn?
This is can be depended on many features, but one good candidate is the degree of *morphological complexity* in the language - that is - the degree to which inflectional morphemes are informative, productive, and clearly marked.
This is can be depended on many features, but one good candidate is the degree of *morphological complexity* in the language - that is - the degree to which inflectional morphemes are informative, productive, and clearly marked.
Specifically, languages with opaque and irregular structures are expected to be harder to learn when compared to languages with more regular, compositional, and transparent structures, where there are systematic one-to-one mappings between meaning units and form units.
This postulated link between grammatical systematicity and language learnability is really important - not only for many theories on languages learning, but also for theories that try to explain language evolution and language diversity.
But unfortunately, this causal relationship (= more systematic linguistic structurebetter learning) is poorly attested, and only a handful of studies tried to examine it in natural languages and in the lab...
better learning) is poorly attested, and only a handful of studies tried to examine it in natural languages and in the lab...
To makes things more complicated, we know that bigger communities tend to have more systematic languages compared to small communities. This correlation (big population ~ more systematic) was confirmed in natural languages, and also in our own experiment: https://royalsocietypublishing.org/doi/10.1098/rspb.2019.1262
If languages with more systematic grammars are indeed easier to learn, and if such languages tend to develop more in bigger communities, then this suggests that learning advantages can be *traced back* to community size and to the degree of systematicity in the language!
We decided to test this hypothesis using an artificial language learning experiment.
In this experiment, we took the final languages that were created by big and small groups of interacting participants in the previous group communication game, and taught them to 100 new people.
In this experiment, we took the final languages that were created by big and small groups of interacting participants in the previous group communication game, and taught them to 100 new people.
These languages varied in their group size origin (i.e., whether they come from big or small groups) and in their degree of systematic structure (ranging from languages with little to now structure, to languages with very high systematic structure), but were otherwise comparable.
Adult learners were trained on these input languages, and then were tested on their knowledge of the input language using a memory test (measuring their reproduction accuracy on learned items) and a generalization test (measuring their ability to label new, unseen items).
We then analyzed adults' learning of the input languages (their accuracy, speed, generalization) with 2 questions in mind:
(1) are more systematic languages easier to learn?
(2) are languages created by bigger groups easier to learn, above and beyond their systematic structure?
(1) are more systematic languages easier to learn?
(2) are languages created by bigger groups easier to learn, above and beyond their systematic structure?
To promote #OpenScience and increase the credibility of our research, the entire study (design, procedure, predictions, analyses plans) was preregistered on @OSFramework and is available at https://osf.io/9vw86 . All data & scripts are also available at https://osf.io/d5ty7
First, across all measures and all analyses, there was no significant effect of community size. That is, we found no evidence that languages created by bigger groups are easier to learn than those of small groups when equating for the degree of systematicity in their languages.
Second, we found that languages with more systematic grammars were indeed learned better and faster by our participants: learners benefitted from high systematicity in their input language, and struggled with learning languages that were semi-structured or unstructured.
Yet this relationship was not linear: depending on the exact measure of accuracy, languages with *no structure whatsoever* were learned equally well or even slight better than languages that contained *some* patterns (but also multiple irregulars and inconsistencies).
This result is puzzling and quite unexpected, and we discuss it at length in the paper. In essence, we think this has to do with participants' learning strategies (rule-searching vs. memorization) - inconsistent patterns can be even more confusing than having no patterns at all!
Third, systematicity promoted more generalizations. In an exploratory analysis, we found that participants who learned more structured languages generalized more: learners were able to exploit the regular patterns in their language to make informed guesses about unknown meanings.
However, these results are based on a preliminary & exploratory measure, and as such should be taken with caution and require further validation. If you'd like more details you can look at the code and read the detailed discussion in my thesis (pp.189–191)
https://pure.mpg.de/rest/items/item_3215438_4/component/file_3215439/content
https://pure.mpg.de/rest/items/item_3215438_4/component/file_3215439/content
Finally, we looked at the generalizations of different participants who learned the same language and asked: Are they similar? Do different people generalize in the same way and use the same words for new meanings, despite never learning them & never interacting with each other?
We found that participants who learned more structured languages were more likely to produce the SAME generalizations. This suggests that systematicity allows strangers to converge effortlessly: people who never met can communicate about new events, and immediately be understood!
Going back to the original group communication experiment, these results support the mechanism we postulated to be driving the creation of more systematic languages in big groups to begin with: having more structured languages can help facilitate convergence with more people!
We think these results can potentially extend to language learning by children. Despite major differences between kids and adults (see work by @CarlaHudsonKam), both should benefit from having more regularity in their input. If anything, kids may be even more structure-sensitive!
