# Social Networks and Linguistic Variation Social groupings are indicative of one’s linguistic exposure and can provide additional insight into the clustering of linguistic features beyond classical measures of social membership such as age, sex, and socioeconomic status. In the case of the current study, social network analyses can be utilized to explore the relationship between an individual’s social connections and their sociodemographic reality as well as the importance of exposure from members of one’s social network in predicting one’s participation in the target sound changes in-progress. In this section, I utilize a graph-theoretic approach to social network analysis that, while compatible with previous work, makes use of distinct conventions in order to explore the role of an individual’s embeddedness in a local network in their adaptation to ambient linguistic variation. ## Methods The social network was constructed by identifying two kinds of social connections: familial connections (i.e., through consanguinity or marriage) and friendship connections (identified during participant recruitment). Two measures of the importance of a given node in the network (centrality) were used: betweenness centrality and eigenvector centrality. **Betweenness centrality**: measures how often a given node is a member of the shortest path between two other nodes. * Index of an individual’s “control of communication” within a network (Freeman, 1979, p. 224)—making it ideal for the current analysis. * Identifies those with wider social networks. Their nature as bridges among different networks could increase the likelihood of exposure to linguistic variation, which could be propagated through their extensive social network (cf. Labov, 2001, p. 364; Milroy & Milroy, 1985). [Social Network by Betweenness Centrality][1] **Eigenvector centrality**: measures the degree to which a node is connected to other well-connected nodes (Bonacich & Lloyd, 2015). * Individual nodes with high eigenvectors will be connected to one another. Used for website rankings and the calculation of some academic journals’ impact metrics. * Places importance on the linguistic variation present within an individual’s immediate subnetwork. Individuals with stronger eigenvector centrality may be more likely to (1) retain more traditional features when novel features are not present in their social network; but (2) quickly acquire patterns of variation when it is present in their network. [Social Network by Eigenvalue Centrality][2] ## Results ### EY-Raising The highest and most fronted values of /eɪC/ are found in members of Lisa’s subnetwork: Jessica, Tonya, Tina, and Lisa herself. Intermediate values are found in those connected to this network through Jessica and Leticia H., demonstrating a clear network effect in the spread of this variable within the Puerto Rican sample. [EY-Raising in the Social Network][3] There is a strong relationship between eigenvector centrality and the adoption of Ey-Raising, to the effect that those who are closely connected to the densest subnetwork (i.e., Lisa’s group) both raise /eɪ/ the most in favorable contexts and make the greatest distinction by phonetic context. ### Canadian Raising (AY-Raising) Canadian Raising is diffuse across the social network. For example, Lisa’s subnetwork is spread across those who raise and those who appear not to and, while Jake is connected to three individuals with higher-than-average vowels in the favorable context (Cristian, Susie, and Jocelyn), he does not share the same production patterns. Additionally, Leticia H., who is one of the strongest adopters of Ey-Raising, appears not to participate in Canadian Raising to the same degree. [Canadian Raising in the Social Network][4] This agrees with the findings that Canadian Raising is found in all social classes in the Puerto Rican dataset and is widespread in Philadelphia more generally (Labov, 2001, pp. 300–303), but it suggests that adoption of one of these changes does not necessarily imply adoption of the others within the Puerto Rican sub-community. * As one’s embeddedness within a subnetwork increases, the degree of separation between the two phonetic contexts diminishes. * All speakers are participating in the change to some degree, decreasing an effect of social network. ### OH-Lowering For the stereotype, /ɔ/, an influence of social network is nigh inexistent. [OH-Lowering in the Social Network][5] For example, despite Jessica’s strong COT-CAUGHT merger, few people she is directly connected to follow suit, suggesting that other social factors influence the employment of this strategy (male, younger, higher SES). # Correlation Among Sound Changes In-Progress Labov (2001, pp. 373–381) finds that correlations vary considerably depending on the sound change. In the current data, there is some evidence for correlation among the sound changes in-progress. **Correlation among variables in favorable contexts** **(conversational speech)** ||/eɪC/| /aɪ0/| /ɔ/| |:---:|:---:|:---:|:---:| |/eɪC/ |1.00|*p* > .05|*p* = .06 |/aɪ0/|0.20|1.00|*p* > .05 |/ɔ/ |0.31 |0.19 |1.00| ## By Age Group **Born before 1985**: variables are strongly correlated: in most cases, the 95% confidence intervals for each of the three variables overlap. In general, the older speakers have production patterns that are unlikely to deviate from the grand mean. [Correlation among sound changes in older group][6] **Born after 1985**: Values skewed above the grand mean, suggesting overall participation in the sound changes in-progress. * Only about a quarter of the younger speakers (n = 4/14 = 29%) have all confidence intervals that cross the grand mean for all three variables compared to nearly half of the older speakers (n = 10/23 = 43%). [Correlation among sound changes in younger group][7] Individuals appear to vary considerably in the change in which they are most strongly participating. * No case where a speaker participates in all three changes in-progress at a rate above average, * No case where a speaker participates in all changes in-progress at a rate below average. Puerto Ricans in Philadelphia are advancing these sound changes in concert with one another as members of a community. [1]: https://osf.io/jgra5/ [2]: https://osf.io/52xhj/ [3]: https://osf.io/w57qa/ [4]: https://osf.io/jaf2d/ [5]: https://osf.io/5rcbs/ [6]: https://osf.io/qg68v [7]: https://osf.io/cnwqs/