Violence is rarely what we expect it to be. “It was so new and senseless that we felt no pain, neither in body nor in spirit,” Primo Levi wrote of his first experience of Nazi brutality. “Only a profound amazement.”

Sociologist and amateur martial artist Randall Collins starts his wonderful, rambling book, “Violence: A Micro-sociological Theory” (Princeton University Press, 584 pages, $45), by pointing out that violence baffles us, and that it rarely resembles our imaginations, or what we see in films. Between individuals, combat often looks goofy and undignified, more slappy flailing than solid punches. For group violence, the saloon brawls in Westerns have taught us to expect bystanders to join in and smash chairs over each other’s backs; but in real life, bar patrons tend to back off from the melee, staring inertly.

If Mr. Collins is right — and let us hope he is — this reluctance to fight is natural, common, and underestimated. Humans abhor violence, he tells us, and they require acute overdoses of fear and tension to overcome that abhorrence and get physically mean. Violence simply does not happen as readily as we suppose, he says, and because it is so exceptional, many of our guesses about its origins and nature have missed their marks.

Previous approaches tried to identify “violent individuals,” and to discover whether a blend of, say, poverty and desperation and broken home life made them violent. Mr. Collins argues that this broad sociological approach won’t explain anything, because the most important factor — the overwhelmingly most important factor — that determines whether one resorts to violence is not one’s past but one’s present. Sociologists could find that a history of childhood abuse correlates with a violent adulthood, but most abuse victims aren’t violent. The broad approach over-demographizes the problem, focusing on background instead of moment-by-moment situations, tension-wracked instants as they are actually lived. Mr. Collins’s method — serious and painstaking but also endlessly diverting — is to hunt down first-person accounts, newspaper clippings, witness testimony, and, if possible, video and photographs of the perps in flagrante. Mr. Collins dips into a robust range of sources, from his own fieldwork among goons on the streets of Boston to the “Iliad” to studies of hazing rituals at elite British boarding schools. Long and entirely worthwhile diversions explore the history of dueling, of violence in sports, and countless other examples of violence in controlled and chaotic environments. The result reads like a social science survey crossed with a work of comparative military theory and a self-defense manual.

How lovely to have a book so unsubmissive to the conventional tropes that misinform our notions of how violence actually plays out! It is a very rich study indeed that provides such an illuminating taxonomy of personalities found on preschool playgrounds, but also tells us the self-soiling rate of American soldiers in World War II (5–6% overall, one in five on the front lines) and examines the phenomenon of pillow fights in prison cells (all fun and games, till the strong gang up on the weak, and fill pillowcases with books or stones). Or that tells us that soldiers often regard elite snipers — even those on their own side — as social lepers, because of their unsettling comfort with death. Almost every page in this long book offers insight into lives lived in extremis. Mr. Collins’s most distinctive point — that we have to overcome a very high threshold of fear and tension before we become violent — seems, under the weight of his anecdotes, difficult to deny. He repeats the famous military finding that soldiers in combat only very rarely fired their weapons at the enemy with the intention of killing. All evidence suggests that the typical frontline infantryman fought not with a frenzied, sexual thrill, as was suggested by Joanna Bourke, but with jelly-guts and a whimper on his lips. In not especially severe cases, some men under fire will lie down exposed in an open field and cover their eyes and head, hoping to make the violence go away by not looking at it.

Under these conditions of fear, few will fight. Violence happens only when the fear-stricken desperados find a path around their paralysis, often through “staged violence” — the highly stylized dueling rituals so popular in 19th-century Europe, or informal “fair-fights” staged outside bars or after school today — or by finding a weaker victim to gang up on; Mr. Collins calls one common variety of this “forward panic,” a manic orgy of violence as in My Lai or Nanking. Providing a weak target sometimes suffices to tip a situation from tense to violent. In moments of tightly wound panic, begging for mercy can itself provoke slaughter.

These insights are simple but practical. Mr. Collins concludes with recommendations: Educate soldiers about forward panic and how easily it can feed its own appetite (addictively, “like eating salted nuts”). Learn how to defuse dangerous situations by matching an aggressor’s bluster without seeming to top it. And, most interesting of all, he urges us to consider reviving the practice of dueling — which was, during its heyday, a way to reduce and contain violence, not a way to encourage it. One hopes to see trial runs of this practice after the next close tenure decision in Mr. Collins’s sociology department at the University of Pennsylvania. “Violence” is a rare academic work, with both a convincing reappraisal of its scholarly terrain, and enough accessibility and useful advice to attract laymen. The writing is clear and direct — sometimes with a welcome touch of the colloquial — and well illustrated with photographs and charts.

Mr. Collins is at work on a second volume, from which we should expect much. In the meantime, the most interesting counterpart to this book is not another sociological study, perhaps, but novelist William T. Vollmann’s seven-volume illustrated 2004 essay on violence, “Rising Up and Rising Down” (sometimes read but rarely finished, including by this reviewer). Both men have undertaken to study violence as it actually exists, and neither is content with easy answers. But Mr. Vollmann’s bracingly undisciplined reportage contrasts with Mr. Collins’s dogged adherence to facts, and Mr. Collins’s amoral descriptions find a fascinating complement in Mr. Vollmann’s punishingly introspective ethical concern. They are two gifted completists, and readers interested in the subject of violence should be grateful.

Mr. Wood is a staff editor at the Atlantic.