Why do some soldiers support autocrats when the masses are mobilized against the status quo? In Tunisia and Egypt in 2011, the military backed away from embattled presidents and supported regime change. But it’d be simplistic to say that militaries often side with their populations.

That same year in Yemen, the military fractured into loyal and disloyal units. The Syrian armed forces saw conscripts and low-ranking officers leave their posts. And in Bahrain, the military remained cohesive and used force to support the ruling family. What explains this huge variation in military behavior?

From media accounts we might think that soldiers’ identities shape their decision-making at these make-or-break moments for autocratic regimes. It was inevitable, in the minds of some, that Sunni Bahraini soldiers would crack down on the country’s Shiite-majority demonstrators. The Syrian rank-and-file left the military because they were Sunni Muslims fed up with an Alawi regime, so the argument went. And some found it unsurprising that Egyptian soldiers sided with demonstrators, given the lack of sectarian divides in the country’s Muslim population.

But in fact, our new research shows that military hierarchy dynamics go further in explaining how soldiers act during these crises. After interviewing more than 100 soldiers and conflict witnesses from Bahrain, Yemen and Syria, we realized it’s crucial to break down “the military” and consider how a soldier’s position within the hierarchy influences his behavior during these crises.

Every country’s armed forces are composed of a variety of actors that have decidedly different interests and opportunities to influence the course of an anti-autocrat movement. In our article, we focus on one particularly important dichotomy: that between high-ranking officers and their subordinates. When President Bashar al-Assad ordered his troops to crack down on Syrian protesters in spring 2011, the military elite in Assad’s inner circle had interests and abilities that stood in stark contrast to those of the army’s conscripts.

These diverse military actors also have a broader set of potential actions than is typically recognized. Many commentators ask whether the military will defend the regime by remaining loyal, or defect from it. But in reality, military personnel have another option: they can exit the regime in a more passive way by remaining quartered in their barracks or refraining from using force when deployed.

These three options — exit, resistance and loyalty — mean that as observers we need to be more perceptive of what military personnel are actually doing during these events. In the early days of the Egyptian uprising, had soldiers defected and joined in demonstrations, or were many personnel exiting the regime and simply refusing to use force against Egyptians?

In our research we explore two hierarchical relationships that influence whether, at these potential breaking points, soldiers exit, resist or remain loyal.

When an uprising begins, the civilian autocrat first tasks his top military commanders with putting down the revolt, and we find that commanders are highly likely to obey these orders and remain loyal. This is because at the start of a highly uncertain uprising, commanders prefer the limited, but certain, rewards of remaining in the political elite to the gamble of potentially increasing their influence by leaving the regime. As a result, there are only narrow circumstances under which officers consider a low probability of success, but high payoff, power grab.

Lower in the hierarchy, the military commander then orders his subordinates to carry out his commands. Yet it is clear that the subordinate’s considerations are quite different from those of his superior officer. Service is primarily a job and there may be severe obstacles to career advancement in the autocrat’s military for these low-ranking personnel; meanwhile, they are also in a different physical position during unrest because they shoot and are shot at.

Although these factors drive subordinates’ interests, because of the chain of command these soldiers have much less autonomy to act on their preferences for or against regime change. As a result, we find that subordinates are most likely to fulfill their superiors’ orders when those commanders have leverage over them, such as through strong monitoring and punishment capacities.

Breaking down the military helps us understand why countries — all located in the Middle East, all experiencing uprisings the same year and all with a history of patrimonial ties between the autocrat and armed forces — witnessed soldiers responding to unrest in such different ways.

In Bahrain, the King Hamad bin Isa al-Khalifa regime deployed its security and military forces to contain, clear and dismantle protest sites in 2011. Virtually all military personnel executed the regime’s orders. Although factors, such as sectarian dynamics, may have affected soldiers’ attitudes during the uprising, understanding the military hierarchy at the time goes further in uncovering the sources of the regime’s stability. Bahraini military commanders had a significant interest in the status quo in 2011; regime reform would have cost the ruling family members their positions. Regardless of their potential preferences for social, economic and political reforms, subordinates did not have much latitude to act because commanders had a strong ability to monitor soldiers’ actions and punish insubordination in the small nation.

The pattern of military behavior during Yemen’s uprising could not have been more different. When protests escalated in spring 2011, a schism developed in the military. General Ali Mohsen joined the resistance and was followed by members of his division and others. The ensuing conflict pitted these defected forces against the loyalist troops, primarily under the command of the president’s son. The crisis was not resolved until a Gulf Cooperation Council-brokered deal led to President Ali Abdullah Saleh’s exit. General Mohsen had not chosen resistance in the earliest days of the conflict, when the political climate was highly uncertain. But as demonstrations continued, and he reflected on his own marginalization as well as the likelihood that the president’s son would soon inherit power, resistance became more appealing. Mohsen’s subordinates then followed him because he had a high degree of leverage over them — supplying them directly with their salaries.

Examining the Syrian conflict again brings to light an entirely different pattern in military cohesion. High-ranking officers chose not to risk defection from Assad’s regime at the start of the conflict. Moreover, they were able to limit desertions from the lower ranks in spring 2011 by deploying the regime’s overlapping security and intelligence services to monitor and punish disloyal behavior. As the conflict progressed, however, these dynamics shifted. Although the top brass remained loyal to the regime, their ability to exercise control over their subordinates deteriorated, and a swath of lower-ranking personnel joined the resistance or exited the country altogether.

More closely analyzing patterns in military cohesion at such moments aids in explaining these conflicts’ trajectories, as well. Both Syria and Yemen slipped into civil war following the Arab Spring. Yet the nature of those conflicts has differed drastically. Syria’s mass desertions occurred among subordinates, and although this has undermined the army’s fighting capacities it has not led to the disintegration of the military’s infrastructure. In contrast, Yemen witnessed the defection of entire military units, which hastened the government’s collapse and led to a virtual dissolution of the state’s coercive apparatus. Breaking down military hierarchies helps us better assess the strength of authoritarian regimes, and whether they will stand if challenged by their populations.

Dorothy Ohl is a PhD candidate in political science at the George Washington University. Holger Albrecht is an associate in the Middle East Initiative at Harvard and a professor in the Department of Political Science at the American University in Cairo.