Read on Twitter
What is the role of network structure in shaping the grammar of languages?
In a new paper with Antje Meyer and @shirilevari, we tested this question using a group communication experiment. <Thread>
Full #openaccess paper here: https://onlinelibrary.wiley.com/doi/full/10.1111/cogs.12876?af=R
In a new paper with Antje Meyer and @shirilevari, we tested this question using a group communication experiment. <Thread>
Full #openaccess paper here: https://onlinelibrary.wiley.com/doi/full/10.1111/cogs.12876?af=R
Why are languages different from each other? One possible explanation is that pressures associated with social dynamics and language use can influence the emergence and distribution of different linguistic properties - making language typology a mirror of the social environment.
According to this hypothesis ("the Linguistic Niche Hypothesis" @glupyan) the structure of languages is shaped by the structure of the community in which they evolved. Specifically, there seems to be a difference between languages developed in two types of social environments:
(1) Esoteric communities, which are generally small and tightly knit societies with little contact with outsiders
VS.
(2) Exoteric communities, which tend to be much bigger and sparser societies, and have a higher degree of language contact and more interaction with strangers
VS.
(2) Exoteric communities, which tend to be much bigger and sparser societies, and have a higher degree of language contact and more interaction with strangers
The distinction between exoteric and esoteric communities relies on differences in community size (small vs. big), network structure (highly connected vs. sparsely connected), and language contact, or the proportion of adult nonânative speakers in the community (low vs. high).
But these social parameters are naturally confounded in realâworld environments (=smaller groups also tend to be highly connected). This makes it hard to evaluate the unique contribution of each of these factors to the process of language diversity.
Here, we tried to experimentally tease apart the specific role of *network structure* in shaping languages' structure. We did this using a group communication paradigm, where people needed to create a new artificial language to communicate with other members of their group.
In a previous paper, we used the same paradigm to look at the role of community size, and found that larger groups created more systematic languages: https://royalsocietypublishing.org/doi/full/10.1098/rspb.2019.1262
But does network structure also have a unique effect, above and beyond community size? To answer this question, we examined the formation of new languages that were developed by different groups that varied in their network structure, while keeping community size constant (8ppl).
We tested language emergence in 3 types of networks: fully connected, small-world, and scale-free. The networks differed in their degree of connectivity (i.e., how many people each member interacted with) and their homogeneity (i.e., whether all members were equally connected).
We analyzed the emerging languages over time on:
- Communicative success (did people understand each other?)
- Convergence (did they align on a shared language?)
- Stability (how much did languages change?)
- Linguistic structure (are there systematic labelâtoâmeaning mappings?)
- Communicative success (did people understand each other?)
- Convergence (did they align on a shared language?)
- Stability (how much did languages change?)
- Linguistic structure (are there systematic labelâtoâmeaning mappings?)
Our main prediction (which was based on theoretical and computational models) was that sparser networks would develop more structured languages, as a result of higher levels of input variability in such networks (which increase the pressure for generalization and systematization)
We found that, over the course of the experiment, all groups developed languages that were highly systematic, communicatively efficient, stable, and shared across members.
Crucially, there were no significant differences between the 3 network conditions with respect to our original predictions: All networks had similar degrees of input variability and reached similar levels of linguistic structure, stability, convergence, and communicative success.
Nevertheless, we did find several differences in the networks' behavioral patterns, which were not directly predicted. Most notably, smallâworld networks showed the most variance in their behaviors. This suggests that smallâworld networks may be more sensitive to random events.
That is, different small-world groups behaved very differently from one another in terms of their convergence, stability, and linguistic structure levels. In contrast, different fully connected & scale-free groups were quite similar to other fully connected & scale-free groups.
Since we found no other significant differences between networks, it is possible that network structure actually has little to no effect on the formation linguistic trends, at least not in relatively small network as used in this experiment.
However, we think this interpretation is less likely considering the multiple theoretical and computational models that argue in favor of network structure effects. Instead, we believe it is more likely that we did not sufficiently capture the potential role of network structure.
In the discussion, we propose several possible explanations for our null results, mostly related to design limitations. Specifically, it is likely that network structure had no effect in the current design because our selected networks did not differ sufficiently from each other.
This possibility is supported by the lack of differences in input variability across conditions, which stands in contrast with the general consensus that sparser networks are more diversified. The networks' similar levels of input variability may therefore explain our results.
To conclude, we propose several ways of dealing with these methodological issues (e.g., online experiments), and suggest that more research is needed in order to confirm or refute the role of network structure on language evolution and language change. But yey to first steps!
