OBSERVING PHYSICAL TRACES I63 strictly hospital attendants follow fire-safety rules will learn more from counting the fire exits blocked by stretchers than from interviewing atten- dants, who may want to paint a rosier picture than actually exists. School principals who want to avoid showing they are not doing a good job may report less damage to school property than a researcher might observe directly, while principals who want the school committee to increase the budget for maintenance may magnify the damage. If a respondent at home knows a researcher is coming, she may tidy up the house beforehand, put— ting away such physical traces as toys in the living room, which might indi- cate how different rooms are used. Observing or measuring traces does not require being present when the traces are created. The method is therefore particularly useful to find out about rare events, hard-to-see events, private behaviors, and behavior of groups who cannot be interviewed. Zeisel’s school study (1976a) provides an example of using physical traces to document private behavior that is hard to observe directly. During the day teenagers can be seen hanging out around schools, playing stickball against walls, and sometimes climbing onto rooftops. At night they sometimes find out—of—the—way places around back to sit together, drink, and smoke. Boston teenagers treat these half-hidden settings as clubhouses where outsiders are not allowed. The first hint of such nighttime clubhouse activity came from physical traces: empty beer cans, discarded playing cards, cigarette butts, graffiti, and broken lights. Durable Many traces have the advantage for researchers that they do not quickly disap- pear. Investigators can return to a research site for more observations or count- ing and can document traces with photographs or drawings. Of course, the more permanent a trace is, the greater its chance of being observed at all. For example, rock gardens and paving stones in someone’s garden will be visible for years, long after grass and flowers have virtually disappeared. There is, however, the problem of selective deposit. Some activities are more likely to leave traces than others. The extent of beer drinking that takes place behind a school can be detected by counting the number of cans the next day. Playing poker or smoking nonfilter cigarettes may leave no traces at all. Another consequence of the durability of traces is their cumulative qual- ity; earlier traces can encourage later ones. A large number of people may feel free to cross a lawn because people who did so before left a path, whereas fewer people would do so were there no path. This cumulative quality can cause problems for investigators who overlook it and think each act is inde- pendent of earlier ones. But if traces are not taken out of context, their cumu— lative character can provide insights for data gathering and analysis. The find- ing, for example, that litter tends to beget litter (Finnie, 1973) is particularly useful if you want to arrange maintenance schedules in parks and around schools.