No single definition of friendship has traveled intact across all human cultures. The word arrives in every language bearing different cargo — different obligations, different emotional registers, different assumptions about who qualifies, for how long, and under what conditions the relationship can be said to exist at all. When social scientists attempt to count friendships across cultures — to build something like a global friendship census — they run immediately into this problem: the category being counted is not stable. What gets reported as friendship in the United States, where the word is casually applied to coworkers, gym acquaintances, and people met once at a conference, bears only partial resemblance to what gets reported as friendship in Germany, where the term is reserved for relationships tested over years and never applied lightly.
This is not merely a semantic problem for survey researchers. It is a substantive finding. The variation in how cultures define, count, and weight friendship tells us something important about the relationship between social structure and personal life — about which forms of connection a given society values enough to protect with language, norms, and institutional support, and which it leaves to informal maintenance or deliberate neglect.
Cross-cultural surveys of friendship since the 1980s have produced a striking dataset: reported friendship counts vary enormously by country, but self-reported loneliness does not correlate neatly with reported friend counts. Countries with high reported friend counts sometimes show high loneliness; countries where close friendships are fewer but more intensively maintained show lower loneliness. The implication is that quantity and quality are separable dimensions, and that cultures organize the trade-off between them differently. Americans tend toward networks that are large, loosely connected, and low-maintenance; Germans, Japanese, and some Nordic populations tend toward networks that are small, tightly bonded, and demanding. Neither is obviously superior for wellbeing outcomes, but the differences are real and have distributional consequences — they determine who is most likely to fall through the network when crisis arrives.
The friendship census, when attempted seriously, is therefore less a headcount than a map of social ontology. It reveals what a society believes relationship is for, who deserves it, and what work it is expected to do. The data is messiest exactly where those beliefs are most divergent from the survey categories being used to capture them.