OBJECTIVES: Government reports on violence in developing countries are almost universally based upon police data, which are typically incomplete and unreliable. Violence in Karachi was evaluated using ambulance service, not police, records. SETTING: Karachi, Pakistan. METHODOLOGY: The study was based on a case series of persons suffering from intentional injuries and transported by Edhi, the largest emergency service in Karachi, between October 1993 and January 1996. Main outcome variables were injury and death rates. RESULTS: Edhi Ambulance Service transported 4,091 intentionally injured persons during the 29 month period from October 1993 to January 1996. Ninety five per cent (n=3,864) were males; 74% (n=2,823) were 20 to 40 years of age, and 2,400 (58%) died before reaching the hospital. Firearms were the most common mode of injury (n=3,396, 83%). Forty six per cent (n= 1,828) of violent injuries occurred in four neighborhoods of Karachi--22% in Korangi (n= 884) and 8% each in Orangi (n=337), Malir Colony (n=307), and Nazimabad (n=300). On the 32 days when a political strike was called, more people were injured (mean = 10.4 v 5.0 persons, p=0.01) and killed (mean 6.6 v 3.9 persons, p<0.01) compared with days without a political strike. CONCLUSION: Violence is a major public health problem in Karachi, affecting predominantly wage earners. At least some of the violence is rooted within the political system. Detailed study of the causes of violence that explains the role that political and ethnic tensions play, may suggest strategies to lessen the toll of violence.