State Repression and Opposition Survival in Pinochet’s Chile

Underground Organizing Capability: On Compartmentalization and Social Embeddedness

Underground organizations need operational security to withstand government repression and operational capacity to execute their actions against the state (Finkel, 2015; McCormick & Owen, 2000). Regardless of type of group—whether criminal, nonviolent or armed—operational security is important because of the constant level of threat. ⁵ Operational security usually involves compartmentalization, secrecy, and militants adopting fake identities, living parallel lives, and smuggling arms (for armed organizations) (Erickson, 1981; Finkel, 2015; Guevara, 2013; Lenin, 1971). Compartmentalization minimizes ties across small, densely connected subgroups, and it is the most common structural feature of clandestine groups. Generally, militants are placed in small cells of three to four individuals. Individuals in different cells of the same rank do not know each other. Communication between parallel cells is nonexistent, and communication between higher-ranking and lower-ranking cells, in the case of hierarchical organizations, is limited. As Figure 1 illustrates, militants of the cells headed by person B and C do not know each other or share information. If the secret police captures militant G, the highest number of subsequent arrests due to information from militant G is four (specifically, militants B, D, E and F are vulnerable). Militants C, H, I, J, K, and A are somewhat removed from further repression.OPEN IN VIEWER

The idea behind compartmentalization is to significantly reduce the speed and effects of repression when the regime’s security forces are highly capable and willing to use extreme force (Shapiro, 2013). A compartmentalized organization mitigates the effects of security breaches by making it harder for the regime to identify targets. In addition, secrecy and identity changes also make it harder for repressors to find and capture a militant even once identified by name and photograph. These protections reduce fear among militants, helping the opposition maintain recruits. These incentives are similar across underground groups (Finkel, 2015), but not all underground organizations are capable of adopting—or even choose to adopt—strict secrecy and compartmentalization.

The compartmentalization of information does not just involve secrecy. It means most individuals do not even possess the information needed for counterintelligence, and they do not know or recognize fellow militants from other cells. Security agents aim to match a suspected person’s name with a face, to locate them, and find more affiliated militants. In high repression and state capacity environments, counterintelligence officers torture or otherwise compel captured militants or civilians to give up the names of other militants, their addresses, and safe houses, as well as to match names to faces. Therefore, anyone who has information about the appearance, name, or whereabouts of a militant is a liability.

The operational security of two clandestine organizations could look identical when we draw their structural connections (ties), as in Figure 1. But in practice, they could differ drastically on how strict their compartmentalization is. A strict or fully compartmentalized organizational structure means that beyond their immediate cell-mates and top leadership (whom most members can identify, even in a highly secret organization), militants will not know others’ appearances, names, families, addresses, or safe houses. This level of information control is not possible for organizations that emerged prior to the closing of civic space. If an opposition group previously operated in a more open environment, one can assume that its members were far more connected and informed than a compartmentalized organization would require. Compartmentalization is thus partially path dependent because militants cannot unlearn faces and identities—though addresses can be compartmentalized at any point, because militants move often. ⁶

So highly interconnected and public organizations that suddenly face severe repression will not be able to optimize their operational security because strict compartmentalization will be out of reach. While going from interconnection to compartmentalization prevents the free flow of new information, divides tasks by cells, and limits the number of people in communication with each other, they cannot undo acquired information about the identity and participation of militants. The implication is that a sudden increase in human rights violations disadvantages existing organizations and favors the creation of new ones with the ability to fully compartmentalize.

There are also underground organizations under threat that choose not to strictly compartmentalize. These groups tend to fall in at least four camps: organizations with territorial control; networks that have adopted a full-matrix or full-channel structure for redundancy; organizations with a high degree of in-group preference divergence; and groups that need a high degree of discrimination in their actions for survival. Armed groups that control territory can operate more freely in those areas, taking the form of hierarchical, well-connected, military organizations. The Revolutionary Armed Forces of Colombia (FARC) and the 26th of July Movement led by the Castros and Che Guevara during the Cuban Revolution are examples of organizations that operated underground in most of the country, but nonetheless held territory in other areas. Thus, neither their structure nor information was fully compartmentalized, especially inside their safe havens.

In all-channel or full-matrix networks all militants are connected to all other members (JCS, 2016). This redundancy significantly hampers efficiency, but it means that organizations can carry out their operations regardless of capture, at least for some period of time (because multiple operatives have the same instructions, training, and resources).

Additionally, some organizations avoid strict compartmentalization because the costs of losing operational control are too high compared to the security benefits of fully compartmentalizing information. When there is high in-group preference divergence (Shapiro, 2013), there is considerable internal disagreement about tactics and operations, and/or corruption. So leaders may try to exert more control over other militants to ensure that there is no discrepancy between their plans and how actions are carried out. A similar logic applies when groups place a high premium on target discrimination in the use of violence (Shapiro, 2013). When armed insurgents think they may lose significant support if they kill civilians, they may forego the benefits of full compartmentalization for the sake of retaining a high degree of control over operations.

These two types of groups tend to have lists of militants, organizational charts, and records of actions, funding and participation. While these documents allow leaders to maintain more efficient control over their operations, they can also be a liability. For example, Narodnya Volya, a leftist militant organization in 19th-century Russia “was crushed within several weeks of the assassination [of one of their leaders, G. A. Lopatin] as police traced the network of connections among its members” given that the leader was carrying membership lists (Shapiro, 2013, 132). Al-Qaeda in Iraq also maintained thorough accounting documents, with names and addresses of militants. These documents were seized in 2007 and used extensively to understand the inner-workings of the group (Bahney et al., 2010; Fishman, 2008).

The first hypothesis is thus that, under conditions of high, intelligence-based and selective repression, underground organizations without strict compartmentalization will be less likely to survive than those with strict compartmentalization (Survival: Hypothesis 1). This pattern should hold regardless of whether organizations chose not to fully compartmentalize or were unable to fully compartmentalize given constraints such as historical interconnectedness. A corollary of Hypothesis 1 is that we should observe fewer subsequent arrests if security forces capture a militant in a fully compartmentalized organization than if they capture a militant in an organization without compartmentalization (Cascading effect of repression: Hypothesis 2).

When it comes to social embeddedness, this paper argues that there are important disadvantages to civilians being able to identify and know militants of a clandestine group. The local population is an enormous liability for the survival of militants if two conditions hold: first, security forces have the reach and will to arrest, interrogate, torture, and otherwise compel them for information; ⁷ and second, the location of civilians supporting the clandestine group is known. Groups with historical ties to a community cannot seamlessly start operating in the underground without civilians knowing who they are, as was the case with the Irish Republican Army (IRA) and the Provisional IRA (Shapiro, 2013). When the British counterinsurgency efforts increased, these organizations were forced to compartmentalize their structures and operations (Shapiro, 2013). They could no longer rely on the faithful allegiance of their supporters for secrecy, and civilians often thwarted their efforts.

Another example is the Abu Dahdah Spanish Al-Qaeda cell, which was responsible for the March 11, 2004 Madrid train bombings. This tightly knit kinship network was able to execute one action effectively and without detection, but within three months the entire network was identified (Rodríguez, 2007). Spanish intelligence services relied, among others, on those with whom the insurgents worked and lived to understand connections. Scholars have shown that high-risk organizing requires deep connections with the civilian population and longstanding solidarity networks (Braun, 2018; Lewis, 2020; Sarbahi, 2014), yet our understanding of the risks posed by civilians remains underdeveloped.

How do detached underground organizers and militants operate in their environments? They could be living in isolation, removed from the civilian population (Sarbahi, 2014). This kind of social detachment occurs in hostile terrain that is sparsely populated and when the group is a newcomer to the area. The absence of connections with civilians provides a buffer of protection for opposition groups because counterinsurgents cannot reach militants through civilians. The exception is when militants are highly concentrated in a geographic area that the state can target. In this case, civilians may not be necessary to identify militants.

Social detachment could also involve contact with the civilian population. This form of social detachment is more common in urban centers; militants live, work and do community service without the population suspecting that they are part of an underground organization. A double identity allows militants to blend into the community and support themselves while continuing their underground activities. Sometimes these changes require migrating to another state or city, or adopting a different life altogether. In this form of social detachment, local civilians would only be a liability under very special circumstances: intelligence officers would need a picture or detailed description of a wanted individual; only then would civilians, who could recognize the person, be able to give the information needed for capture. But if the officers are simply looking for militants from the group or working off of names or pseudonyms, their residence in a community does not present a threat. This barrier to capture holds even if security forces take repression to the extreme to get information from civilians. Therefore, my third and final hypothesis is that socially embedded underground organizations facing high, intelligence-based, selective repression will be less likely to survive than those that are not deeply connected to civilian communities (Social embeddedness: Hypothesis 3).

The Targets and Victims of the Pinochet Dictatorship in Chile, 1973–1989 Dataset

The Targets and Victims of the Pinochet Dictatorship in Chile, 1973–1989 dataset includes the names of 6445 individuals in Pinochet’s target lists, whether or not they eventually became victimized, as well as the almost 2000 victims of the regime absent from the target lists, but for whom we have group membership information. I compiled this individual-level dataset from several archival sources. The head of the secret police DINA, Manuel Contreras Sepúlveda, while serving a 529-year sentence for human rights violations, published a book titled The Historic Truth: The Guerrilla Army (Contreras Sepúlveda, 2000). In it he lays out his version of the junta’s fight against communism. The 450-page appendix to the book consists of photocopies of the government’s lists of “dangerous individuals,” “wanted individuals,” and “members of leftist parties and sympathizers of the leftist guerrillas,” along with their political affiliation, leadership positions, information about the security agencies looking for each individual, whether they had military experience or explosives training, and sometimes even their home addresses. Although the Armed Forces and other Chilean security agencies insisted that they burned all the records from the dictatorship period and were thus unable to produce information about victims, it appears that Manuel Contreras took many DINA documents with him when the intelligence agency was dismantled in 1977 (Basso, 2015) (Interviewee 43, April 2017, Santiago, Chile). Figure 2 shows one of the lists.OPEN IN VIEWER

To code whether each individual on these lists was actually repressed, I consulted the lists of victims from the three Chilean truth and reconciliation commissions—Rettig, 1991, ¹³ Valech 2005 ¹⁴ and Valech 2011 ¹⁵ —as well as documents from two NGOs, CODEPU ¹⁶ and FASIC. ¹⁷ The three truth and reconciliation commissions investigated crimes throughout the country to create a complete list of those executed or disappeared (Rettig) and tortured in prison (Valech I and II) during the Pinochet dictatorship, 1973-1990. CODEPU’s additional records provide more fine-grained information on torture victims, and FASIC has the most complete list of those forcibly displaced (relegados) within Chile.

The Rettig archives include variables such as occupation, age, place of detention, city and state of residence, and political affiliation. On the other hand, the two truth commissions that documented cases of torture (Valech I and II) embargoed their archives for 50 years. Information about victims of torture is sparse: we only have their two family-names, two given-names, and identity numbers. So while we know the political affiliation of most people on the target lists (the secret police always included that information), and we know the political affiliation of victims of execution and disappearance who were not on the target lists thanks to the Rettig archives, we cannot know the political affiliation of those who were tortured but not on the target lists.

To make inferences about the targeted and victimized populations, it is important to know the origins, uses and sources of bias that the data may have. How were the target lists created and used, and what are the implications for this study? Based on interviews with the head of the Valech II truth commission, prosecutors who have built legal cases against DINA members, human rights experts, and militants who infiltrated DINA for a few months, as well as on cross-referencing names on the lists, I can draw a few conclusions. First, I am confident that the lists were developed before the coup of September 11, 1973 (Huneeus, 2001) (Interviewee 17, March 2017, Santiago, Chile). Gathering or updating these lists during the military dictatorship would bias the lists in favor of having more Communist and Socialist names than MIR militants, given the latter’s secrecy and underground practices. I argue that the lists are not likely to be biased in this way, and even if there was such disparity in the quality of information, the bias would make the results of this paper less likely, not more. Extra-secretive MIR militants who might be missing from the target lists would presumably be even less prone to capture than MIR militants on the target lists, which means that the group’s true rate of victimization would be even lower than what we can observe with the data. It is also very unlikely that there were other lists with different proportions of Communists, Socialists and MIR militants (Interviewee 15, April 2017, Santiago, Chile).

Additionally, it is improbable that the victims’ lists (the numerator to calculate victimization rate) are biased in such a way that would undermine this paper’s findings. If MIR members were indeed more secretive, their family members might not know their political affiliation and by extension the truth commission would categorize them as unaffiliated. However, the truth commissions also relied on testimony from former prisoners who survived and served as key witnesses to provide information on those who were later executed or disappeared. The work of reconstructing the memory of the MIR in the past 30 years has also dramatically increased the archival information on the disappeared and executed available today. I have used these additional sources to supplement the truth commission data. For a more in-depth discussion on the origins, uses and potential blind spots of the data, including how the intelligence services were able to gather information on secretive groups, see Appendix D.

On balance, this dataset is a major improvement on other studies of repression, which can only observe crackdowns that are carried out. This selection bias has prevented us from learning more about survival in the face of repression. Without knowing the targets of the regime and the extent to which the intelligence services go to gather information, we might make erroneous assumptions about repression and its effects. Just having the number of victims could lead us to wrongly conclude that the most victimized group was the regime’s worst enemy or top target. But what if the most victimized group was simply not as well-equipped to resist repression? What if another group with a lower rate of victimization was the top target, but they had more capacity to avoid repression? If we only compare the total number of victims from the MIR, PS and PC, then like most scholars of Chilean politics, we would conclude that the MIR was destroyed during the Pinochet dictatorship—failing to recognize its comparatively lower victimization rate and its advantages as an organization that survived in the underground (Bruey, 2018; Fruhling, 1983; Pérez, 2003; Wickham-Crowley, 1992).

Qualitative Evidence

The paper also relies on qualitative evidence gathered during 8 months of fieldwork in Chile over the course of three years. I conducted 106 semi-structured interviews with former and current militants from the Revolutionary Leftist Movement, the Communist Party, and the Socialist Party, as well as with militants from other leftist parties and organizations like the Popular Unitary Action Movement (MAPU), the Corporation for the Promotion and Defense of the Rights of the People (CODEPU), and the Committee for the Defense of Human and Labor Rights (CODEHS). I also interviewed leaders of the truth commissions, federal prosecutors in charge of building cases against the DINA, journalists, and human rights defenders and social workers from the Christian Churches Foundation for Social Assistance (FASIC) and the Vicariate of Solidarity. For all the archival sources I used, see Appendix E.

The Most Targeted is the Least Victimized

One of the most striking features of Pinochet’s target lists is the amount of ink spilled on the Revolutionary Leftist Movement (MIR). In fact, three lists were solely focused on this group—naming all the members on record, their military experience or explosives training, and whether they had trained in Cuba. More than 41 percent of militants with a group affiliation belong to the MIR, followed by what the government classified as “leftist parties,” which are mainly militants from the Communist and Socialist parties, with a few from MAPU or the Movement for Popular Unitary Action, the Christian Left or IC (Izquierda Cristiana), and Popular Unity or UP (Unidad Popular). This agglomeration of leftist parties makes up approximately 53 percent of the sample of militants with a group affiliation. Figure 3 below shows the distribution of opposition group membership in the government’s target lists.OPEN IN VIEWER

In their report, the Chilean National Commission on Truth and Reconciliation highlights that the MIR was singled out for repression, which is consistent with the observed pattern in these lists. DINA “set for itself the basic task of eliminating what it regarded as the ultra-left, particularly the MIR and other groups or persons connected to it” (Rettig, 1991, 61). While these appraisals are based on actualized violence, the target lists similarly reveal the extent to which the security apparatus focused on MIR even compared to other leftist groups that were also considered top enemies of the state.

Despite their disproportionate numbers on the secret police lists, the MIR was the smallest of the top targets. At the height of the MIR movement, shortly before the coup, MIR had up to 10,000 militants. Meanwhile, the Communist Party had around 50,000 (Alvarez, 2011; Contreras Sepúlveda, 2000). And at the time, the Socialist Party was larger than the Communist Party, because the PS had grown during Allende’s presidential campaign and tenure in office (Gaudichaud, 2004). It is not possible to know the exact number of militants in these organizations, but one can confidently say that the Communists and Socialists had many more supporters than the MIR. Even so, the objective of the military junta was not to target every single MIR, PS, and PC militant. Rather, the Armed Forces most likely designed the target lists and overall repression campaign around eliminating all high-, mid- and low-level leaders, as well as the most committed militants (Interviewee 80, April 2017 Santiago, Chile).

Looking at those who were on the target lists and eventually victimized, Figure 4 shows that a total of 2435 individuals on the target lists were victimized during the military dictatorship. A total of 4006 people on the target lists were never victimized during the same period. That is, 37.8 percent of the target list suffered the most severe forms of repression, namely, disappearance, execution or torture. Based on the truth commission reports we can conclude that those who were disappeared and most of those executed were also tortured. In other words, all of the tortured, executed and disappeared in Figure 4) were victims of torture, though some were also executed and others were disappeared. Indeed, torture was so prevalent in detention centers that the two commissions addressing the victims of torture (Valech I and II) used the lists of those detained as proof of having been tortured, for purposes of receiving government benefits through Chile’s transitional justice measures (Interviewee 26, María Luisa Sepúlveda, Vice President of the Valech Commission, April 2017, Santiago de Chile).OPEN IN VIEWER

As illustrated in Figure 4, most of those who were on the target lists and victimized were victims of torture, followed by executions, and then disappearances. The number of forced displacements is very small (N = 5) in this sample, which might be because the displaced lists are not as complete as those for the other forms of repression. More likely, however, is that the opposition groups on these lists were top targets of the regime. The government tended to use forced displacement against those whom they perceived as less of a threat (Orellana, 2015).

Despite the secret police’s emphasis on demobilizing the MIR, Figure 5 shows that conditional on being on the target list, the rate of victimization of miristas was significantly lower than of other leftist groups. The MIR’s lower rate of victimization provides evidence for the theory of survival at the core of this paper, and is consistent with hypotheses one and three, which predict that groups with stricter compartmentalization and social detachment will tend to have lower rates of victimization. The probability of victimization for a targeted MIR militant is 43 percent. In contrast, approximately 58 percent of non-MIR militants (i.e., those identified as being from “leftist parties” “Communists” or “Socialists” were victimized. Of the 1900 MIR militants on the target lists, a total of 816 were disappeared, executed, or tortured. Meanwhile, 1065 MIR militants were never victimized, compared to just 665 members from the “other leftists” category who were spared the worst of the repression. See Appendix F for a disaggregation of the type of repression suffered by MIR vs. non-MIR militants.OPEN IN VIEWER

To further probe the first and third hypotheses about compartmentalization and social embeddedness, I estimate a logit model. The outcome is victimization conditional on being targeted, with a value of 1 if the individual was tortured, executed, disappeared, or forcibly displaced, and 0 otherwise. The main independent variable is a binary indicator for whether the individual was a MIR militant or a militant from the “leftist parties” category or singled out as a PC or PS militant (1 = MIR militant; 0 = non-MIR militant), conditional on being on the target lists. Those with unknown affiliation are excluded. There are two control variables, which are both binary. The control variable DINA indicates whether the secret police was assigned to capture the individual (1 = DINA assigned to individual; 0 = otherwise), and the variable Leader indicates whether the individual held a leadership position in their opposition group (1 = leader; 0 = not leader). It is important to control for leadership status because otherwise this model assumes the security agencies prioritized all targets equally. ¹⁸

The first difference plot (Figure 6) shows that being a mirista is, on average, associated with a statistically significant decrease of 12% in the probability of victimization, even when controlling for security agency and leadership status. DINA was the most ruthless security agency and also in charge of eliminating the MIR, making it especially surprising that MIR militants had a lower rate of victimization than non-MIR militants.OPEN IN VIEWER

Mitigating Cascading Effects of Repression

An implication of this paper’s theory is that we should observe fewer subsequent captures after the detention of a militant from a compartmentalized and socially detached organization than following the detention of a militant from an interconnected and embedded one. As explained in Appendix A, we would expect this survival advantage to hold whether the repressive tactic consists of mass arrests or going door-to-door. ¹⁹ As long as the goal is to find a type of person (e.g., leftists) or a specific individual, compartmentalization and social detachment should provide a survival advantage.

To evaluate this hypothesis on the cascading effects of repression (Hypothesis 2), I use time-series data from the executed and disappeared using the individual files of victims created during the National Truth and Reconciliation Commission (the Rettig Commission). There is an embargo on the archives from the two subsequent commissions on torture (the Valech Commissions), so we do not have fine-grained data on the more than 40,000 victims of torture. Because we know the day, month, year, and political affiliation of the disappeared and executed, I calculated the number of MIR, PS, and PC detentions 30 days and 60 days after each MIR, PS, and PC detention, divided by 30 and 60, respectively). These two cutoff points reflect the time security agencies might have taken to follow leads that emerged from a captured individual, based on interviews with former militants and prosecutors. The results from both cutoff points show the same overall pattern. The 30-day results are shown here and the 60-day results are in Appendix G, which also includes figures comparing the speed of detentions between the three groups and other analyses on the cascading effects of repression.

As Figure 7 shows, during the 30 days after the capture of a MIR militant, there are 2 fewer subsequent MIR captures per day, compared to the rate of subsequent captures following the detention of a Communist or Socialist. This model controls for the state where the militant was detained. The results are statistically significant at the p-value.05 level and remain unchanged when examining the period of peak repression (from the moment of the coup until 1977) or the entire Pinochet dictatorship (until the end of 1989). The main OLS model is as follows:

S p e e d D e t e n t i o n s i = α + β ^ M I R v s P S & P C i + β ^ L o c a t i o n i + e i

(1)

OPEN IN VIEWER

Figure 7.

Speed of detentions following a capture by militancy during peak repression and during dictatorial period (OLS coefficients).

Although the Communist and Socialist movements were much larger than the MIR, the evidence suggests that the larger cascading effects of repression for Communists and Socialists are not completely attributable to group size. First, the secret police and military intelligence agencies did not typically pick people up from the street at random. The Rettig truth commission data contains information about the original place of detention of each individual who ended up being killed or disappeared. Approximately 28 percent of the disappeared were originally detained in the street, while more than 71 percent of disappeared individuals were detained at their home, their workplace, at a friend’s house, at a family member’s house, in a foreign country, or crossing the border—indicating that they were likely victims of more targeted forms of repression. Group size plays less of a role in targeted repression, where security forces are looking for specific people with names, home addresses, and information about their workplaces.

Secondly, even when repression was the least selective—during the first 5 months after the coup—the vast majority of detentions took place at leftist centers of power, such as political party and labor union offices, universities, and government institutions run by the deposed Popular Unity government. Not only were there fewer public spaces where MIR militants would gather, because of their underground practices, but there was also less information that each militant could provide, and fewer militants connected to sympathizers in the workplace, place of residence, etc. Detentions at these sites led to many more subsequent detentions for the Communists and Socialists rather than the MIR precisely because torturing the average Communist or Socialist meant gathering more information about other militants than was the case for the MIR.

Finally, I spoke to dozens of leftist militants who attributed not being captured to specific measures that they took. For example, speaking about the repression she observed, a female militant from the CNR within the Socialist Party said that “the comrades who were captured did not do the sign and countersign (chequeo y contrachequeo) [to verify people’s identities], …did not develop a daily contact routine with signs that would alert other militants …that the [secret police] was taking you out to pick out militants (porotear) …In addition, they all knew each other, we didn’t” (Interviewee 94, April 2017, Santiago, Chile). I also interviewed militants who were captured. They trace back their detention to mistakes in compartmentalization and to specific informants that the secret police was able to gain (flipping militants into collaborators, of which there are several prominent cases in Chile (MemoriaViva, 2021; Skoknic, 2007)). As one MIR militant reflected on the runup to his detention by the DINA, “I relaxed my norms…" (Interviewee 38, May 2016, Santiago, Chile).

Psychologically, individuals may be prone to self-serving bias, assuming more personal responsibility for successes than failures (Arkin et al., 1980). However, militants generally did not attribute their survival while in prison to their specific actions, decisions, or skills. In conversations with militants who were detained, tortured and then released, they typically did not know why they survived prison while others were killed and disappeared. A MIR militant who was in charge of printing the underground magazine El Rebelde was detained with her sister, and said “they could have killed us” (Second interview with Interviewee 25, May, 2016, Santiago de Chile). Others shared that they think they survived prison because of interventions by Catholic priests rather than anything they did (Interviewee 12, May 2016, Santiago, Chile).

Compartmentalization and Social Detachment

Once Pinochet came to power, the Socialists and Communists recognized that their organizational structure and large numbers of civilians in their networks were liabilities—particularly during the height of repression. The head of the second clandestine directive of the Socialist Party, Eduardo Gutiérrez, was convinced that the Socialists had to start learning underground organizing skills and developing a compartmentalized structure:

It was in the year ‘75 when we realized that we had to take drastic measures that would allow us to protect our lives and we decided to do what Europeans did in their resistance. We read texts that discussed the topic and decided to go completely into clandestinity; in order to do that we had to change our names, establish mechanisms of communications completely compartmentalized, no one knew your actual name, nor your address, and thanks to those measures we survived. (Vargas & Díaz, 2007, 41–42)

Similarly, Luis Corvalán, General Secretary of the Communist Party of Chile, said: “The Communist Party’s 20 years of legality lived during the Ibáñez, Alessandri, Frei, and Allende administrations, except for the short period in the first one, had driven us to a certain loss of acquired experience throughout the not so few reactionary governments that we had to confront in the past. That is how I would explain my detention, which took place on September 27, only 16 days after the military coup” (Corvalán, 1997).

On the other hand, MIR militants were building their clandestine structures before the 1973 coup. One of the Central Committee members of the MIR said that he started doing “work with the masses” (trabajo poblacional), but “I was there for a short period because afterwards I started doing other clandestine tasks. They took me out of the structure. I stopped being so public…I had other types of duties – I made contact with military officers and soldiers [to identify leftist allies, those who were pro-coup (golpistas) within the Armed Forces, and to make a diagnosis of the military]." This quote shows how militants’ everyday work and connections change when their organization compartmentalizes and goes underground. These transformations mitigate and slow the cascades of repression.

Personal interviews with MIR, PC, and PS militants also reveal that they all relied on trusted friends, relatives, and movement allies to change homes often. If the Armed Forces captured a militant, those connected to them would be at risk of capture. In those cases, the vulnerable militants relied on networks of informers to learn about the captures as quickly as possible, change homes, flee to another state, and take other precautionary actions (Interviewee 75, April 2017, Santiago, Chile; Interviewee 16, April 2016, Santiago, Chile). Even as late as 1977, Secretary General of the Communist Party, Luis Corvalán, explained: “The party has been capable, despite the brutality of the repression, of maintaining a strong clandestine organization, rooted in the support of thousands of families from the people whom from the first day opened their homes to the persecuted” (Corvalán, 1977, 91). Similarly, in 1974 the PS declared that the party was able to move activities underground despite the repression because of their strong organization, alliances between all leftists, and their ties to their bases of support (Corvalán, 1997; Socialista, 1974). The MIR repeatedly emphasized in their publications and internal documents the importance of politicizing students, peasants, and other “mass-based” organizations. These efforts demonstrated that the membership base has “high explosive power and is always willing to act as long as there is capable leadership to guide the way to insurrection” (MIR, 1965d, 11). Though the leadership of the top targets had an incentive to overstate their capabilities, including their level of support among the masses, these archival materials are representative of how the parties viewed their reliance on civilian communities.

However, the key difference between the MIR on the one hand, and the PS and PC on the other, is the way they controlled the flow of information when they worked in local communities. MIR militants lived among their constituents—in poor, working class neighborhoods—but they were not socially embedded. People did not know their true identities and political work: “One learns to live underground with time …you learn to represent your new name, to have a coherent facade with your new activities, your routine, schedule, what you talk about. You have to know the dynamics of your neighborhood, what people think…It’s a process of re-insertion with a new name …you have to learn how to live like fish in water” (Interviewee 38, May 2016, Santiago, Chile). Miristas continued their work in communities within the Catholic church, especially by tapping a less vulnerable population—women (Bruey, 2018). Hugo Flores, president of Movimiento Poblacional Solidaridad, described the dynamic in local churches when they became sites of resistance: “many Christians stopped attending church […] Paralyzed by fear, they hid in their homes. Communists and other leftists, on the other hand, fled their homes and sought shelter in the church” (Schneider, 1995, 116). Importantly, priests, nuns, and the Vicariate of Solidarity social workers and lawyers never asked or investigated the political affiliation of those who sought refuge within their religious community (Garretón, 2001).

In short, compartmentalization and social detachment involve a certain distance from civilian populations, which help slow and mitigate the cascading effects of repression. Although they eventually developed some clandestine structures, more deeply embedded and highly interconnected organizations like the Communist and Socialist parties remained more vulnerable to targeted repression from a capable state, because information about their leaders and organizers was widespread. Appendix H discusses plausible alternative explanations to the argument about underground organizing and social embeddedness, including variations in the determination to repress, the ability to seek exile, and socioeconomic status.