Abstract
In contrast to technologically deterministic approaches that focus on how communication technology affects social relationships, this paper examines how individuals draw on a variety of commonly used communication media in conjunction with in-person contact to stay connected to their personal networks. I term this use of multiple communication media the ‘personal communication system’. Findings are based on a random sample telephone survey of 2200 adults living throughout the continental USA. Descriptive statistics show that despite the popularity of email and mobile phones, in-person and landline phone contact are still the most common ways of connecting with personal networks. Multivariate analysis reveals a more complex picture of media use, showing that the extent to which each medium is used varies to differing degrees with the size and diversity of personal networks. Hierarchical cluster analysis is used to explore the possibility that individuals may have different types of personal communication systems. Results show only two distinct clusters: those who draw heavily on all types of media to connect with their personal networks and those who draw less heavily on all types of media. Heavy communicators typically have larger and more diverse personal networks than light communicators. When taken together, the results presented in this paper suggest that rather than radically altering relationships, communication technology is embedded in social networks as part of a larger communication system that individuals use to stay socially connected.
Notes
1 Although ‘personal communication system’ is sometimes used in the technology industry to refer to cellular technology in America, it is not widely used in academic writing about the social significance of communication technology.
2 Core and significant tie results are kept separate for density variables in this analysis because the nature of these measures does not allow them to be summed together.
3 Zero inflated count regression is more suitable than regular OLS regression because the dependent variables are positive count numbers with a strong positive skew and a substantial number of zero values. Negative binominal count regression was used instead of Poisson count regression because the standard variation of each dependent variable was greater than its respective mean, indicating an over-dispersion that is better handled by negative binomial count regression than Poisson count regression. STATA's likelihood-ratio test confirmed the suitability of this choice.
4 Ward's Method is an efficient way of sorting data that uses an analysis of variance approach to minimize the Sum of Squares between separate clusters.
5 A Calinski and Harabasz index gives a value reflecting the overall distinctiveness of clusters at each stage of a cluster analysis. Larger values reflect more distinct clustering.