Games prisoners do not play: against the Hobbes-Zimbardo approach of unmitigated prison violence

I analyze institutions of prison subculture that mitigate potential violent confrontations among inmates, in contrast to Hobbesian-Zimbardo default spontaneous violence. The games that are relatively rarely played in prison are Chicken and other violent confrontation games. Incoming rookie inmates are subject to initiation tests that allocate them into different subcultural groups, which signals their toughness and disincentivizes fighting. Most experienced inmates develop the eristic skills utilizing prison argot, use informal conflict adju-dicators, and fake aggression toward rookies. All inmates form defensive coalitions. Finally, when inmates commit self-injuries, they follow well-rehearsed protocols to minimize the damage to their bodies and to maximize the impression made on the authorities. The secret knowledge of the associated rules, tricks, and cons is passed down over generations of prisoners through informal schooling. The material for this study comes from two Polish prisons, where the author spent 5 months as a political prisoner in 1985.


Introduction
I analyze informal institutions in the subculture of Polish prisons that mitigate, re-direct, or disincentivize violence. My contribution is in reconstructing those secret components and rules at the ethnographic level of detailpractically impossible to achieve with ordinary research techniquesthanks to the unique natural experiment that was performed by the former Polish communist state on the author (i.e. jailing him).
Prison is often perceived as a brutal exemplification of a perfect Hobbesian arena of the 'war of all against all' (Hobbes, [1651(Hobbes, [ ] 1989, where inmates fight, hurt themselves, and sometimes even die. At least since Bentham's ([1791] 2010) Panopticon, most social theories presume that prison communities experience frequent and severe bouts of violence unless state agents impose formal monitoring and enforcement. Perhaps the most important empirical argument in favor of the 'inevitable brutality of prison' was the famous Stanford prison experiment (SPE). Inside a mock prison, the roles of 'prisoners' and 'guards' were randomly assigned to students. Originally planned for a 2-week period, the experiment ended after just 6 days. Apparently, the students quickly internalized the roles assigned to them: some of the student guards began to use cruel methods and psychological torture, while many of the student prisoners passively surrendered to abuses or even harassed the other prisoners when the guards demanded it (Haney et al., 1972;Zimbardo, 2011). While the SPE's main point was that 'everyone is capable of cruelty' if assigned that particular role, it indirectly implied that there is no defense against the deterministic forces of 'evil.' I call the assumption of the inevitability of unstructured violence the Hobbes-Zimbardo approach.
A clear flaw in the interpretation of the SPE's results is that people in a 'fake prison' have no incentive to think hard about curbing violencethat the simulated experience is not really meaningful for them. Their time horizon is short, and they have no access to locally accumulated norms or values. Moreover, the norms revealed by the experimental subjects are vulnerable to instructional and procedural manipulations by the experimenter (Hoffman et al., 1996). In short, in the SPE's experimental setting, the subculture with its complex institutions is missing from human interactions, and this void produces a fundamental difference in terms of behavior. Subcultures are present in almost all actual institutions that Goffman (1961) called 'total.' These 'total' institutions include prisons, jails (a specialized typo of prison where inmates await trials), and concentration camps, but also orphanages, mental hospitals, army barracks, and monasteries. Prison subcultures overlap with criminal codes outside of prisons, deeply affect prison life, and harbor complex methods of communication and signaling (Gambetta, 2011;Skarbek, 2016). The strength and sophistication of the subculture depend on various factors, such as population density and the level of administrative control (Kaminski and Gibbons 1994;Skarbek, 2020). 1 This article focuses on 'violence-reducing institutions,' i.e. the institutions of prison subculture that mitigate or eliminate violence, replace it with interactions of lower transactional cost or mislead outside observers and uninitiated inmates to believe that certain actions are violent when, in reality, they are not. Administrative regulations generally have limited impacts on such institutions. The secret knowledge about the associated rules, tricks, and cons is passed down over generations of prisoners through informal schooling. The informal institutions provide a good example of a spontaneous order that emerges without a central designing authority (see D'Amico 2015; Hall and Martin 2011;Hayek [1988Hayek [ ] 1994. The existence of such secret knowledge and rules, as well as the harsh prison conditions, socializes inmates to be super-calculative in all their decisions (Kaminski, 2004).
Prison subculture provides an extreme example of the difficulties of identifying key features of informal social institutions (Kaminski, 2004;Voigt, 2018). Inmates religiously protect their argot, norms, and habits, and also even simple facts about their lives. Nevertheless, certain components of prison subculture that constitute its 'semi-secret code,' such as tattoos and particular vocabulary, may leave the prison environment and become parts of the broader culture and language. The most protected components of the 'secret code,' such as the procedures for self-injury or the tricks used in initiation rituals, are very hard or even impossible, to decipher with ordinary research techniques (see Kaminski, 2004: 58-75). I acquired this secret knowledge from my personal experience as a political prisoner in 1985, when I was arrested for managing a publishing house STOP associated with underground Solidarity movement and when I conducted research over the course of 5 months of imprisonment. Clearly, this research is hard to replicate. Part of the descriptive material in this article and a few of its models have been presented elsewhere, in different contexts, by Kaminski (2003Kaminski ( , 2004Kaminski ( , 2006. In 1985, a three-tiered caste system existed within the Polish prisons, dominated by the grypsmen. 2 Unlike in American prisons, interactions with Polish prison personnel were infrequent and superficial, mostly because of permanent shortages of prison staff, as well as overcrowding. 3 Overall, the guards were lethargic and overworked. The subculture of the grypsmen was so secretive and flexible in changing its norms that the guards knew little about the more complex games or their knowledge was either rudimentary or obsolete (Michalski and Morawski, 1971: 3-4;Moczydłowski, 1992). Typically, the personnel were most involved only in various procedures of self-injury or simulation games, such as a seesaw, which I describe in detail in section 7 of this paper. In fact, most secret 1 Zimbardo's experiment had alternative interpretations; it was criticized for its design and execution, selection bias and manipulation of conditions in order to reach desired results, setting standards of behavior for student guards during prestudy orientation and demand characteristics (see, e.g. Banyard, 2007;Haslam and Reicher, 2012;Haslam et al., 2018). Recent studies have failed to replicate his results (Reicher and Haslam, 2006). 2 In Polish, the word is grypsujący. The name is derived from the argot word gryps, denoting a secret message sent illegally inside or outside prison. Words from prison argot are italicized in this paper. procedures were known only to a handful of grypsmen (Kaminski, 2004). This type of environment motivates inmates to play various 'mind games' or 'con games' (McDermott and King, 1988).
The remainder of this paper is organized as follows. In section 2, I describe the Polish prison caste system, with a particular emphasis on the grypsmen subculture. In section 3, I argue that the game of Chicken (and similar games) looks like a good departing point for the analysis of prison violence and that violence-reducing institutions change the game's parameters. Surprisingly, however, although inmates' interactions might resemble various versions or modifications of Chicken, they are not the canonical Chicken game. Section 4 informally describes how the self-sorting of inmates into subcultural groups with different levels of resistance to violence reduces actual violence. Section 5 analyzes various substitutes for violence that reduce the transactional costs of cooperating in order to avoid violence and to reduce the likelihood of the worst outcome, death. Section 6 then concisely describes prison coalitions. In section 7, I claim that many cases of auto-aggression can be fully explained by a reference to an inmates' rational plan for extracting maximal benefits from the prison authorities, the prosecutor or the court. Finally, in the last section, I conclude by affirming that the informal institutions of prison subculture substantially mitigate unstructured violence, as predicted by the Hobbesian-Zimbardo approach.
2. The prison caste system and the dominant subculture of grypsmen The subculture of Polish prisons has its roots in the pre-revolutionary Russian empire and the criminal codes of the infamous 'Thieves University' of Odessa. This common origin made the subcultures of Polish, Ukrainian, and Russian prisons under communism very similar (Kekoshvili and Slade, 2020;Symkovych, 2018;Varese, 1998). One of the remnants of those joint origins is Russian, Ukrainian or Yiddish words in Polish prison argot. Initially, the code was narrow and limited to thieves. After the post-Stalinist thaw in 1956, however, the subculture of Polish prisons was somewhat democratized by allowing almost all inmates, not just the thieves, to join the caste of grypsmen. After the fall of communism, the prison subculture in Poland was transformed like all other social, political, and economic institutions in the country. In addition to substantially better living conditions, the main changes affecting the subculture were the entry of organized crime, illegally traded cellphones, and substantial amounts of illegal money in circulation. In many prisons, grypsmen constituted a minority of all inmates, and they were less familiar with the norms of subculture and argot than their predecessors (Miszewski, 2015).
In 1985, the caste of grypsmen dominated the majority of prisons and jails, although the level of enforcement of grypsmen norms varied by prison type and inmate characteristics (Moczydłowski, 1992). 4 The subculture in jails was stronger than in non-jail prisons. In the prisons of Rakowiecka and Białołęka, the grypsmen constituted about 70-80% of the prison population. Their code was based on a pentalogue of so-called principles that included honorable behavior, strong solidarity with other grypsmen, non-cooperation with the prison administration, maintenance of personal hygiene, and the provision of some assistance to all inmates, regardless of their caste membership. These principles were translated into more than a hundred practical, normative postulates regarding specific prison situations, called behaviors. For example, a grypsman could not extend a handshake to a member of another caste, to a guard or to most visitors with the exception of a family member or a doctor who had saved his life. Before he offered a handshake to an unknown prisoner, he had to ask, 'Are you a grypsman?' The secret argot of grypsmen replaced most Polish words with prison equivalents, and it had strict rules that excluded certain insulting phrases. Among the worst were to compare a grypsman to a woman, a 'squealer,' a communist, a prosecutor or inmates from a lower caste. The most dangerous thing in the cells was the penis of another inmate: touching another man's penis, even accidentally, or being touched by another man's penis, relegated a grypsman to a lower status.
The structure of power among the grypsmen was semi-democratic. Cell leaders were either elected by cell grypsmen or emerged naturally by being the most aggressive or the most energetic grypsmen, cellblock leaders were elected secretly by the cell leaders, and prison leaders were elected secretly by the cellblock leaders. Major leaders typically managed a small crew of grypsmen enforcers.
Below grypsmen, there were suckers, who occupied the median position in the hierarchy. Considered as actual or potential informers, they were not allowed to eat at the same table with the grypsmen, nor, as mentioned above, could they shake hands with the grypsmen. Also, a grypsman could steal from a sucker with impunity. In some cells, suckers were the majority, but in both jails at that time, I estimated that their share among the broader prison populations was about 20-30%.
The lowest layer of prison life was reserved for the so-called fags. Some grypsmen used them as providers of sexual services in a passive role. They lived their lives around the jaruzel, 5 the toilet bowl, and were given feminine names and clothing. When grypsmen were bored, fags were told to dance, sing, or provide other entertainment for everyone. About 1-2% of inmates were assigned to this miserable social role, which occurred when they broke some of the strongest prison norms.
Having the status of a 'political prisoner' usually provided a good starting point for one to join the grypsmen. Sometimes, political prisoners were considered a special category. The total number of such prisoners in early 1985 was estimated to be only around 150 inmates for all prisons throughout the country. In the 1980s, the number of criminal prisoners oscillated between a low of 67,824 (in 1988) and a high of 110,182 (in 1985) (Centralny Zarząd, 2020). Frequent amnesties at the time led to cyclical deflations of the country's total prison population.
All rookie inmatesthose new to the prisonhad a chance to become grypsmen upon entry unless they were communists, child molesters or judiciary or enforcement officers. However, before joining the grypsmen, a rookie would have to pass a series of initiation tests. A weaker or younger rookie might be subject to a fagotization test. On the day of the test, his cellmates would treat him poorly. After such psychological preparation, a grypsman might secretly offer him protection at night in exchange for sexual favors. In another test, baptism, an inmate's 'immunity to pain' would be tested. Here, he would be asked to submit to a ceremonial beating with a wet towel. In fact, for both tests, staying tough was the best strategy. There were no consequences for refusing sexual favors during fagotization, and no truly serious beating took place during baptism. The bad treatment instantly disappeared once the rookie showed no fear. Of course, while undergoing such initiation rites, the rookie would be unaware that he was being manipulated.
After successfully passing the fagotization and/or baptism tests, a rookie could be subject to a shorter or longer series of little games that would test his loyalty to his cellmates and his ability to react quickly to small prison challenges. If he maintained his loyalty and after his street smarts and toughness were confirmed, the final stage before becoming a grypsman involved prison universitya long period of training on using the argot correctly and observing the grypsman norms. The lectures took place at night on the bunks, with a changing set of instructors. The rookie was expected to master the argot, the principles, and the behaviors, and to be able to apply this knowledge in typical prison situations. Joining the grypsmen caste was symbolically sealed when each grypsman in the cell extended a handshake to the rookie.
If a new inmate failed to pass baptism or a similar test, or later, if he committed a seriousbut not gravely seriousoffense against the grypsman code, he became a sucker. He would have a chance to become a grypsman again, but the lift that would return him to grypsman status would require very strong evidence of toughness, cleverness, and sometimes the provision of significant material benefits to the other grypsmen. In one cell in which I was housed, a sucker broke a finger one night when the cell's grypsmen were attempting to drill a hole in the floor to enter the cell below, which housed women. Since reporting this injury would have immediately jeopardized the project, the sucker waited until morning. His suffering benefitted the collective, and thus he was rewarded with a lift.

5
The cell's toilet bowl was originally named after the first secretary of the Polish Communist Party (PZPR): before 1980, the name was gier, derived from Edward Gierek. Jaruzel started gaining popularity after the internal communist coup d'état of 1980 by General Jaruzelski, who ousted Gierek.
An inmate who failed the fagotization test or violated some other serious subcultural norm was permanently fagotized. For example, he might have touched another inmate's penis during the weekly shower, agreed to provide sexual favors to a cellmate or started collaborating with prison or judicial authorities.
Some basic norms of prison life were applicable to all prisoners. An inmate had to announce publicly and clearly his intention to feed the jaruzel (defecate), to urinate or to relieve himself with flatulence. For example, one might shout loudly, 'Jaruzel is thirsty! Eating?' Then, if there were no affirmative answers, he was allowed to urinate. Shouts of 'Eating?' were so frequent in grypsmen cells that suckers sometimes referred to grypsmen as eaters. The norms did not allow for one to eat or drink while either feeding the jaruzel or passing gas. An accidental, foul-smelling kiszczak 6 was forgiven, but the mass production of malodorous air obliged an inmate to take care of his intestinal problem, perhaps by fasting for a while. After all, typical jail cells were small, between 8 and 13 m 2 , with 8-10 inmates each. At night, during curfew (around 9:00 PM-6:00 AM), masturbation was allowed without announcement.
All inmates had to understand the basic rules of the argot so as not to offend the other inmates with forbidden words. The grypsmen also provided the suckers and fags with a minimal amount of education about what was allowed and what was forbidden. For example, the grypsmen would relay to the suckers and fags that no inmates were allowed to squeal, or to spend too much time talking to the administration.

Modeling fights as chicken
Potentially violent interactions among human players or animals are often modeled as a game called Chicken, which is also sometimes called Hawk and Dove (see Smith and Price, 1973). Chicken is often used as an umbrella term for a large number of related confrontational interactions. In such cases, mutual 'confrontations' or 'fights' might result in substantial transactional costs: the loss of some material goods, blood and pain, punishment or even death.
In a typical prison interpretation of the game of Chicken, two inmates can either fight each other for scarce prison resources, or surrender. Except for a mutual fight, their payoffs would be equal to the amount of the fully transferrable resource that they would acquire. So, let us assume that the total amount of some resource is 2. If one player fights and the other surrenders, the fighter would then get the entire resource, and the quitter would get 0. If both surrender, they would split the resource equally. If both choose to fight, the payoffs would be equal to −1 (see Figure 1).
One can quickly point out a few typical, real-world complications that are missed in such a simple model. For example, actual players could have asymmetric preferences and strengths. For instance, one of them might have a stronger tolerance for pain while the other might be afraid of pain. They might also be awareor notof their own or the other player's strength. The fight could benefit a player in sending a signal of his toughness, which would be valuable in future interactions, regardless of the possible material benefits. In addition, it is much more difficult to calculate the value of imponderables, such as maintaining one's honor or reputation. Moreover, the simple model shown in Figure 1 represents a one-shot interaction, while one's reputation would actually emerge after many real or potential interactions. 7 This criticism is justified, and, in fact, it partly motivates the more refined models that I present later in this paper. Nevertheless, Chicken is a good starting point for analysis, since it represents a situation of costly confrontation in which the player is tempted to assume an aggressive attitude. Various 6 In this word game, kiszczak in Polish means 'from the guts,' but it also denotes the name of the communist Minister of Interior General Czesław Kiszczak. 7 For instance, models of 'raid or trade' consider explicitly different costs of conflict and cooperation for different players; see Anderson and McChesney (1994), Rider (1993), and Skaperdas (1992).
quasi-violent interactions in prison are interpreted below as games related to Chicken, but in many ingenious ways, the lethal transaction costs of choosing the strategy to fight in this game are lowered.

Sorting by resistance to violence
Among several hundred aggressive confrontations I witnessed, I recorded only one case of a brutal fight. While in Rakowiecka prison, a political prisoner joined my cell. He soon began to clean the inside of the jaruzel, an activity that was usually reserved by the code for suckers or fags. A juvenile, inexperienced grypsman who had just been transferred to our cell decided that the political prisoner was a sucker and that he could safely 'borrow' (steal) his shoes. When the political prisoner realized what had happened, he hit the juvenile grypsman hard in the face. A short, bloody fight ensued. Apparently, both inmates were surprised by what had transpired. The political prisoner had ignorantly sent a signal of weakness by cleaning the jaruzel, while the juvenile, who was unfamiliar with political prisoners, had misinterpreted the signal. After the fight, the juvenile inmate started respecting the political prisoner and the latter announced his willingness to learn the grypsmen norms.
Prison fights, just like wars and other types of conflict, often break out because of errors due to mistaken beliefs. Both fighters may believe that they are stronger, or at least, that they will be able to rebound quickly. In fact, one of them might be much weaker and less resistant to pain than the other or both might be tough. When accurate information about a prisoner's toughness is revealed through the sequence of games or initiation tests described earlier, the inmate's situation changes dramatically (see Gambetta and Székely, 2014;Székely and Gambetta, 2019).
In the informally described models that follow, the inmates belong to the categories of 'tough' (grypsmen), 'weak' (suckers), and 'very weak' ( fags). Grypsmen are the most numerous, which is why relations within this caste, as well as between grypsmen and the other castes, are the most important. Fags are the least numerous; there are rarely two fags in any one cell. The interactions within the caste of suckers or between a sucker and a fag take place under the watchful eyes of the grypsmen, who do not want to see any random fights within the lower castes. The lower castes are expected to provide services and goods to the grypsmen, but otherwise, they are to keep a low profile. Thus, almost all main interactions happen between a grypsman and another grypsman or between a grypsman and a sucker or a fag.

Grypsman contra grypsman: unprofitable violence
The initiation tests select grypsmen candidates, who later learn various secrets that reinforce two fundamental principles: first, 'do not interfere' (in the conflicts of other people); and second, 'defend yourself' (against any aggression). The 'do not interfere' principle is repeated endlessly during the rookie's training. A safeguard against escalation, it helps the elders (grypsman leaders) to mitigate the violence. Grypsmen are socialized into believing that, if attacked, they should fight no matter the cost and that they should never surrender, but also that they should only count on themselves. They learn that the meager benefits of immediate surrenderavoiding being harmedare dwarfed by the delayed cost of the loss of reputation and the possible degradation to a lower caste. In addition to being preselected as tough, grypsmen become deeply convinced that they must always act tough. Not surprisingly, the beliefs of two potential adversariesthat their opponent is willing to fight to the deathlimit the frequency of real fights. The subcultural costs associated with Surrender in Chicken turn this strategy into a dominated one.
In a typical situation of conflict, an aggressor can challenge his opponent by hitting him in the face or chest and by shouting, 'Start! Let's go! Hit the floor!' The challenged inmate can then fight or surrender. Since the socialization of the grypsman who is subjected to the aggression strongly stigmatizes the act of surrender, he prefers to fight. Of course, the aggressor recognizes this fact. Thus, ritual shouts do not typically lead to actual fights. The threat of mutual injury in a fight against a tough opponent is a sufficient barrier to the escalation of most conflicts.

Grypsman contra sucker: a credible threat of violence
Both baptism and fagotization tests are cruel, and they are also far from perfect in discovering the 'true type' of an inmate. They can also be contaminated by the interplaying interests of the cell's grypsmen. 8 Nevertheless, these tests seem to provide quite effective procedures for assessing a rookie's character. One significant characteristic of a sucker is that he failed to perform well in a baptism test, which requires courage and a willingness to fight for his honor or that he behaved dishonorably, even after initially passing the test. With minor violations, a grypsman can be placed on what amounts to probation. Obtaining the eventual status of a sucker means that the violations were more serious or that the inmate was a stubborn repeat offender.
In a typical conflictual interaction between a grypsman and a sucker, the grypsman can choose the status quo, or he can challenge the sucker by asking for some of the sucker's property; stealing things from the sucker; or making some other, moderately humiliating, petty demands of the sucker. Grypsmen refrain from explicitly offending suckers with phrases that are considered the most insulting in the argot or from attempting to steal too much. (As mentioned earlier, the worst treatment is reserved for fags.) The sucker can then either fight or surrender. A failed initiation baptism test by the sucker, or some other behavior considered cowardly, would have already revealed that the sucker is afraid of fighting. He would prefer to surrender, just as he did during the initiation tests. A knowledgeable and proactive grypsman would predict such a response. As a result, the grypsman wins the resources without fighting.
The most experienced grypsmen boast about 'showing the suckers their place.' As one grypsman in Rakowiecka prison known by the name 'Boxer' declared, 'Just toss me in the cell with Gierek, Jaruzelski, Wałęsa, and Kuroń [top communist and anticommunist leaders in the 1980s]. In a week, they will be sweeping the floor and bringing me slippers in the morning' (Kaminski, 2004).

Grypsman contra fag: protection racket
Fags are inmates who either failed the initiation test of fagotization or revealed their unwillingness to fight for their honor through some other behavior. They are labeled as fags because they signaled weakness to a greater extent than suckersfailure results in a complete loss of dignity. 8 For instance, the author was allowed to skip baptism because he taught the card game Bridge to one of the cell's elders, and also because he provided valuable health-related expertise to the boss of all of the prison grypsman. Relegating him to a sucker would have made both powerful grypsmen's interactions with him more problematic.
Demands on fags may include sexual services, servitude, or the provision of entertainment that is generally considered humiliating. Grypsman-fag interactions are also different than grypsman-sucker interactions. Demands on fags are not formulated as challenges. Instead, a fag is taught that he must guess a grypsman's wishes and that he must act voluntarily and immediately to accept simple requests from a grypsman. In return, he receives protection from violence.
A fag may voluntarily cooperate and provide his services, or he may refuse. If he refuses, the grypsman can beat him, or do nothing. However, a grypsman who fails to discipline a fag will lose respect in the eyes of his peers. Thus, the grypsman has a strong incentive to punish. The fag, anticipating a severe punishment for refusal, agrees to cooperate. In equilibrium, the fag voluntarily provides sexual services to the grypsman. Interestingly, prison legend also recounts stories of grypsman-fag relationships that developed some genuine affection.
Like shopkeepers who are 'protected' by the mafia, fags and suckers are controlled by credible threats of violence. Actual violence is dispensed only when resistance is detected. The benefits extracted by the grypsmen are usually commensurate with the expected resistance. Grypsmen learn where to stop before going too far: suckers can be humiliated only when it is absolutely necessary to defend the honor of grypsmen, while the humiliation of fags is considered safe. However, frivolous violence and mindless humiliation are strongly discouraged.

Substitutes for violence
As discussed in the previous section, a fight between two grypsmen would likely be very costly in terms of injury, so various substitutes to these dangerous encounters have emerged. Instead of fighting, grypsmen sometimes fenced with argot, pretended to fight, or outsourced the conflict resolution. I discuss each of these scenarios below.

Argot duels
Prison argot, the inmate slang derived from broader language, 'does not just describe a preexisting institutional reality but is partly constitutive of this reality' (Searle, 2005: 12). Prison argot conceptualizes the world differently by naming the situations, emotions, and objects that are common in prison and disregarding those that are harmful, inefficient, or appear with low probability. Even more importantly, prison argot shapes and structures the beliefs and expectations of payoffs that constitute the institutions that are specific to prison. A rookie inmate who masters the argot beyond the common vocabulary sends a powerful and inimitable signal that he expertly understands prison reality and that he can competently respond to threats. It is no wonder that verbal aggression in 'total institutions' like prisons or gangs is a frequent substitute for violence (Hamill, 2018: 111).
In my experience, argot duels, often combined with frightening gestures and grimaces, erupted when grypsmen playing dominos, cards, or chess claimed that someone was cheating. Those banal squabbles were often preludes to heated conflicts. The accusers and accused alike pretended to be pushed to the limit and demonstrated that they were ready to fight. They cast what seemed like the worst profanities at each other, but they were also careful not to truly offend the opponent by breaking argot norms. The argot rules allowed for many apparently insulting curses but prohibited some that were considered truly offensive. A grypsman offended by an unacceptable curse was obligated to fight to defend his honor, else he risked degradation.
A particularly colorful argot duel was recorded by me live in a hospital cell (Kaminski, 2004(Kaminski, , 2006. While the inmate 'Boxer' was playing checkers with 'Skull,' 'Boxer' accused 'Skull' of hiding one of the pieces. Their verbal exchange is presented below. The words added in brackets {} indicate possible variants of the spoken words spoken, which, due to the secret argot rules, would have made such curses deeply offensive to the opponent. (Sensitive readers are advised to skip the exchange and go directly to the analysis of the duel's semantics.) 'Boxer' contra 'Skull' Boxer: Skull, you son-of-a-whore, may your shit block your {fucking} anus! Skull: When I fork you {fuck you}, you will shit out last year's meal. Boxer: May you never get a hard-on when you hit freedom. {when you are in prison} Skull: Before that, one of your balls {both balls} dries out. Boxer: Hide yourself and your louse-covered balls in a redhead's cunt! At this point, both duelists burst into laughter and, in a friendly way, began to analyze the last curse. The dilemma was whether the 'half' qualifier was enough to cleanse the otherwise prohibited offensive phrase. This discussion then turned into a consideration of the maximum allowable percentage of one's 'dick' that could enter the opponent's 'ass' without offending the receiver of this insult. Both grypsmen agreed that, away from the relatively relaxed hospital environment, it would be better not to use such an expression at all; it was too close to the line that separated the curses that were allowed from those that were forbidden. 'Skull' ritually withdrew his last curse using a standard phrase, 'Skiciorem!' (It-was-a-joke!). Viewing this situation firsthand, it was clear that the other patients in the cell were quite impressed by the speed with which the curses were generated, as well as their ingenuity.
A slight modification of every curse in the exchange above could have turned the statement into a forbidden one. For instance, cursing a grypsman's 'aunt' was allowed because it was not aimed directly at him. Another way of generating a 'safe' curse was to wish the worst to those anatomical features that grypsmen evidently did not have such as the vagina.
I observed about a dozen argot duels following various explosive situations, but none ever turned into a fight. The duels transformed the aggression the inmates felt into low-transaction cost exchanges, and they gave opponents time to cool down. Argot duels modified the payoffs in Chicken by dramatically lowering the stakes. What is more, argot skills, which were perfected by many grypsmen, signaled one's subcultural competence to the other grypsmen. In a similar vein, grypsmen employed verbal substitutes for physical aggression when interacting with guards, as well.

Fake fights
Aggressive behavior could be faked. In fact, deception was the cornerstone of most grypsmen tricks or con games. Little games played on rookies during the initiation period that followed the tests of fagotization or baptism often involved complex deception with many cooperating actors. In a typical game based on deception, a rookie might be asked by the strongest grypsman in the cell for a boxing match (Niewolski, 2003). If the rookie refused to participate in the match, he lost the test and could be subjected to harsher treatment and longer testing. If he agreed to participate and prepared to fight, then nothing happened: the match would not take place at all. The point of this test was to force the rookie to signal his weaknesshis refusal to participate in the matchor his toughnesshis acceptance of the challenge. 9 Similarly, a grypsman could ask a rookie to 'lend' him his sweater, pants, or shoes (Kaminski, 2003; see also Bukovsky, 1979). Here, it would be obvious to all that the rookie's property would not be returned. However, refusing the prospective 'borrower' would not result in any sanctions. In all such cases, there would be no fight, nor would there be any sanctions, since the situation was staged in order to test the rookie's toughness or loyalty. 9 Recruits to the Irish Republican Army (IRA) or other revolutionary organizations might be asked by a recruiter to kill a traitor in the organization (Hamill, 2018). If the rookie candidate refuses, he is not admitted to the organization. If he agrees and follows the plan to 'kill the traitor,' he is stopped at the last moment by the recruiter: There was no real traitor and the rookie was put in a staged situation that forced him to signal weakness or toughness.
Many little games were also played on established suckers. In these cases, a grypsman would make an aggressive move or would demand a dangerous action, raise his voice, and suggest that the refusal to satisfy his demand would cause a brutal fight or other unpleasant consequences.
In Fake Chicken (Figure 2), the grypsman challenges the rookie to fight (Kaminski, 2004). 10 In this game, a bully can choose the status quo, with the normalized payoffs equaling 0 for both players, or he can challenge the rookie. He does not know the rookie's toughness. The rookie might be weak, with probability p, or might be tough, with probability 1−p. The bully enters the game with some estimate of the probabilities based on the rookie's past behavior. The rookie can fight, or he can surrender. A weak rookie has an incentive to surrender, while a tough rookie prefers to fight and defend his honor or property. In this game's perfect equilibrium, there is a certain limiting probability p* of the rookie being weak, such that the bully challenges the rookie when p > p*; chooses SQ when p < p*; and chooses either strategy when p = p*. Facing a challenge, a weak rookie surrenders, while a tough rookie fights. The rookie's type is then revealed by the actions taken.
The game described above represents how a typical, uninformed rookie perceives the situation. The bully and his accompliceshis cellmatesengage in very convincing acting to create this perception. They comment on past fights, broken bones, or beaten rookies who refused such a challenge. If the rookie is tough, the bully fights and is worse off than if he had chosen the status quo. But, in fact, the bully never fights because the actual game has a slightly different ending (see Figure 3). As discussed earlier in the context of baptism, this test is also not perfect for discovering a rookie's 'true character. ' In this game, the bully has another move if the rookie decides to fight. The bully can continue the fight, or he can announce that the rookie 'is a good guy,' praise his courage, and withdraw from the fight. This hidden option is the essence of the con: when the rookie reluctantly prepares for the fight and makes aggressive noises, the bully announces, 'Okay, okay, I see that you are a brave guy. Do not get overexcited. I am telling you that it was just a test.' He is immediately supported by the other cellmates, who confirm, 'Yeah, it was a test, rookie, relax,' to stop the irritated rookie from escalating the conflict.
In the perfect equilibrium of this game, the bully always challenges, regardless of the probability of the rookie being weak; the rookie always fights; and the bully always announces. However, when the rookie's equilibrium strategy of the staged game is combined with the bully's equilibrium strategy from the game in Figure 3, then the dynamics are as follows: the bully always challenges; a tough rookie demonstrates Rookie's private knowledge: his type, weak or tough. A weak rookie prefers to surrender, while a tough rookie prefers to fight.

10
This and certain other models studied in this paper include pairs of games. An alternative way of modeling the events would be to use games of incomplete information. I believe that the use of pairs of games is simpler and more intuitive. that he is willing to fight, while a weak rookie surrenders; the bully announces that the tough rookie passed the test. A weak rookie surrenders and does not know that he could have threatened to fight without any negative consequences. Rookies who figure out that acting tough is always optimalperhaps after playing similar games a few timestake the first mental step toward becoming grypsmen.
Note that although the interaction between a bully and a rookie has the potential to be very violent, it does not end in a fight. Valuable prison resources, such as sweaters, pants, shoes, shirts, watches, and so forth, change hands with minimal transaction costs.

Affairs and adjudication
In many prisons, certain types of conflict are handled by informal prison self-governments (Skarbek, 2014). Elders or other respected inmates who are arbitrators can be paid for arbitrating disputes in the form of goods. Prison adjudication can be interpreted as outsourcing the distribution of goods in Chicken to an external arbitrator at much lower transaction costs than fighting.
A typical affairan interaction or exchange that led to a conflictcould look trivial to an external observer. For example, a transaction taking place via the prison black market might go wrong. Because inmates in jail spend most of their time in cells, many such transactions involve brokers, such as barbers, librarians, or cooks, who are corridormen, i.e. those who move around the cellblock halls and provide inmates with services, books, or meals. A corridorman can carry goods and money from one cell to another. Some goods or money may disappear in the process. Since intermediaries are involved, there might be insufficient information about each party's contribution in the exchange. Adjudication of such conflicts requires gathering accurate information about the intermediaries who handled the goods or money, as well as about what happened as it changed hands.
Typically, grypsmen leaderselders or fuss-mastersare the arbitrators. According to an ex-prisoner who lived in the same cell with an elder, the life of an arbitrator is busy: Staying in the same cell with the grypsman leader causes continuous tension because he engages in endless yelling conversations through windows, resolves conflicts among grypsmen, threatens, curses, and forgives. Searches, rehab [rehabilitation officer] visits, and reports [official punishments] are more frequent than in normal cells and irritate other inmates. Someone always passes grypses [secret messages], knocks on a wall, and shouts from a walkplace. A fuss-master often has helpers, usually youngsters […]. Helpers deal with grypses and answer window calls on behalf of the fuss-master, often receive punitive reports, and go to the hardbed [a punitive cell with a bed made of wood and thin blanket instead of linen]. But they also catch scraps of splendor and glory, they witness important things, they participate in making history (description by Niesiołowski, 1989: 98-99; author's translation).

Coalitions
Another type of violence reducer that I will describe only briefly here is a waferhouse, or a coalition of inmates that is less formal, which has more fluid boundaries than the castes. In small cells with eightnine inmates, two competing coalitions are often formed. One of these coalitions can be 'minimal winning' with respect to the sheer power of its members, i.e. it might include the minimum number of inmates necessary to rule the cell. This coalition has sufficient power to obtain and control the most important scarce cell resources, such as good bedding or pajamas, bunk beds, or extra jam or butter obtained from corridormen. Membership in a ruling coalition is often temporary. A proactive, ruthless coalition leader might include or exclude less prominent coalition members according to their tentative usefulness. For instance, he might include an inmate who has just received a rocket (a large food package) from home, or he might exclude another whose rocket has just been consumed by the coalition. In addition to the dominant waferhouse, a separate coalition might include some or all of the other grypsmen, who want to establish a defensive pact. Neither suckers nor fags are allowed to join waferhouses with grypsmen.
In stables, which are large cells with 40-50 inmates, coalition structures are more complex, and the coalitions are typically smaller. Coalitions primarily function as a means of defense. Temporary coalitions, or teahouses, might also be created for the joint consumption of beverages like tea or food.
One reason for forming a defensive coalition is that a rookie inmate might be advised by an experienced prisoner to challenge a weak inmate and beat him in order to establish a reputation for toughness (see Wojciechowski, 1981). Such a beating can help to establish the rookie's toughness. In a prison with sophisticated coalitional structures, however, inmates are protected from such primitive attacks. A coalition member is a problematic target because his wafers are expected to defend him. In such a case, an aggressor's waferhouse might have to engage in inter-coalition conflict; however, inmates are reluctant to take on this type of large-scale conflict, known as a big smoke, unless the cause is very significant.

Auto-violence
A frequent form of violence inflicted by an inmate upon himself is called self-injury or self-mutilation. Self-injury interactions involve prisoners, but they can also involve prison personnel, prosecutors, or the judiciary. This type of violence is present in women's prisons, as well. Along with faking an illness and the refusal to eat or submit to medical treatment, such as during a hunger strike, self-injury might be part of a rationally motivated and complex plan to extract various benefits (including a legal way to exit) from the prison, or the court system, prosecutor, police, or doctors who make important inmate decisions. For an outside observer who is unaware of an inmate's true intentions and goals, self-injury can look like an emotional reaction to the harshness of imprisonment.
Popular types of self-injury in Polish prisons in 1985 included hitting the line or seesaw (fake hanging); cutting (typically a forearm, but also the legs or the belly); swallowing (ingesting metal objects); scalding (burning with boiling water, typically on the legs); burning (setting fire to oneself, typically the legs, often with floor polish); injection (inserting various substances into muscles to cause an infection); and nailing (placing needles or nails in the forehead or the pupil of one's eye).
Self-injuries can be beneficial because of the widespread, universal, humanitarian norm to 'treat sick persons with more care,' which is enforced by international watchdog agencies like Amnesty International and internal prison procedures. Prison officials also want to avoid poor performance indicators by maintaining a prison that has relatively low levels of inmate injuries, illnesses, deaths, and disorder, as well as to avoid the high costs of medical services. Inmates take advantage of institutional opportunities that put pressure on prison decision-makers. The prisoners' objectives include the shortening of one's sentence through parole, suspending a sentence, moving to a more comfortable environment in the prison hospital, escaping from a hostile cell, or even coordinating a meeting with a partner in a hospital cell (Kaminski, 2004).
One notable example of the implementation of a plan involving self-injury is seesaw, or hitting the line, i.e. hanging oneself. These events typically happen at night. In such cases, an inmate hangs himself on the bars of his cell, but just before he dies, other prisoners will rescue him. While still unconscious, he is taken to the prison hospital, sometimes missing death by just a few seconds.
How can such a desperate action be explained as anything other than brutal self-violence on the part of an emotionally exhausted inmate? Poland's official reports from the late 1980s suggest that there were about 1,500 self-injuries by prison inmates (unofficial estimates range between 4,000 and 6,000), including about 132 seesaws per year, 15-30 of which resulted in death. Thus, approximately 10-20% of seesaw suicide attempts resulted in death. Among the general, non-imprisoned population, this percentage is estimated to be 60-85% (Spicer and Miller, 2000;Tintinalli, 2015). Clearly, this is a significant discrepancy. While inmates are more likely to observe a suicide attempt and intervene, there is another explanation for many of these attempts: often, suicide attempts by seesaw are faked. Needless to say, both actual suicides and suicide attempts remain a big problem in prisons (Canning and Dvoskin, 2016;Liebling, 1999).
In the grypsmen subculture, a fake seesaw is one of many types of controlled self-injury. It is taught during the prison university stage of a grypsman's education, described earlier, when a candidate for grypsman has successfully passed all the required initiation tests (which could include fagotization, baptism, and/or little games). By no means was this common knowledge when I was incarcerated. Only elite grypsmenexperts in the prison subculture and (often) the prison's informal bossesknew this procedure. They only shared it with select rookies, since even a perfectly executed seesaw can cause severe injury and it must generate significant expected benefits to an inmate, such as a suspended sentence or bail release.
Seesaws often take place in jail shortly after one becomes an inmate but following a period of careful preparation. Most inmates will leave their cells for a daily 30-minute walk, but the prospective seesawer and his partner will skip the walk and rehearse the event in the temporarily empty cell. Seesaws frequently occur late at night, after the partners have mastered the procedure. By then, the seesawer will have learned and practiced how to quickly make a rope and a noose out of linen, how to tie it to the window bars at the proper height, and how to gently kick a bucket or stool out from under his feet. Of course, if the intention is to get some sort of benefit, the prisoner will not want to die accidentally. Death by hanging occurs either by suffocating or by cervical fracturebreaking one's neckthus avoiding the latter by minimizing the fall is essential.
During the actual seesaw attempt, the partner pretends to be asleep, but he is really monitoring the situation. When the bucket falls and the seesawer 'begins to dance on the rope,' the partner is prepared to act. Any change in the 'dancing' pattern may signal a loss of consciousness. Then the partner immediately begins a well-rehearsed sequence of moves, catching the seesawer's legs and cutting him off the rope (using an illegal knife or razor blade obtained for the occasion) while simultaneously yelling for help. Other inmates slowly wake up and join the growing uproar. Soon, all of the prison inmates are screaming and banging aluminum mugs on the steel doors to alert the guards. While the guards are running to the cell and calling for the doctor, the unconscious seesawer lies on the floor, unaware of the pandemonium he created.
The incisive observation that motivated the development of the seesaw con was that a hanged person faces two critical time thresholds, not just one. The first, which occurs in less than a minute, is the loss of consciousness while the inmate is still alive. After a few minutes, death occurs by suffocation; the window separating loss of consciousness from death could last up to 10 minutes. The partner is expected to target the early period of the 'unconsciousness window' to set the rescue in motion. The seesawer has to be unconscious because, otherwise, the exercise is pointless, but at the same time, he also needs to stay alive. Hence, the partner must act swiftly, but not prematurely. Unconsciousness cannot be faked since a physician can easily identify mere acting. Of course, bungled mistakesas well as real, emotion-driven suicidesdo happen sporadically in prison.
The purpose of a fake suicide is to imprint a false image of the incident in the minds of the observers. Figure 4 presents a simple model of a seesaw from the perspective of the prison authorities. The choice of a prisoner is to do nothing, or attempt suicide. No partner is suspected of participating in the seesaw; the suicide is made to look genuine. If the seesawer makes the attempt, Nature acts. With some probability p, he is rescued when his cellmate wakes up and cuts him off the line in time. In a perhaps more likely scenario, death occurs with the probability of 1-p. His credible willingness to die is confirmed by the universal understanding that the likelihood of rescue is relatively small for serious attempts. The credibility of the effort is confirmed by unconsciousness, as he cannot pretend to be unconscious. Thus, observers will believe that the attempt was real and that the inmate credibly signaled his wish to die (see Figure 4).
The actual Fake Seesaw game is different. In addition to the seesawer, the game includes an accomplice who pretends to be an accidental bystander. The role of Nature is limited only to a mistake by the accomplice. Mistakes happen, but their probability is low because the procedure is well-rehearsed, and the partner's incentives for a successful rescue are strong. If he fails and the seesawer dies, the grypsmen elders will investigate. They will understand very well what has happened and it is easy to identify the inmate who stayed with the seesawer in the cell many times during the inmate walks as the partner responsible for the failure. A negligent partner will be punished, or at least ostracized. On the other hand, a successful partner gains a grateful and indebted cellmate and the respect of other prominent grypsmen (see Figure 5).
A successful fake seesaw can generate valuable benefits for the inmate, but it is not without risk. The inmate who first described their fake seesaw to the author suffered from partial cerebral anoxia and temporary paralysis of his right hand. Prior to being jailed, he had been arrested red-handed; charged with a crime; and, as a recidivist, expected a long sentence. The prosecutor believed that his attempted suicide was real, however, and that the prisoner's next attempt would succeed. The original charges were substituted with less serious ones. Ultimately, he was released from prison and promised a suspended sentence. By faking a 'credible' signal of desperation, the inmate committed himself in the eyes of the prosecutor to an aggressive strategy, and he forced the prosecutor to play less aggressively in order to avoid the mutually negative outcome. Note: This is the false game that the prosecutor assumes to be true. The seesawer can do nothing (N), or he can attempt to commit suicide (A). In this false game, he prefers death to being saved, and the probability of rescue, p, is low.
The seesaw method is one of many types of self-injury that inmates use to gain benefits from the system. Usually, a necessary prerequisite for such a con game is that the party in the position to grant benefits is unaware of what is happening. The author's experience indicates that this condition was met for many self-injuries. In other cases, the administration learned of the tricks and the self-injuries partiallybut not necessarily completelyleading them to lose their usefulness. Then, older tricks were slowly replaced by newer ones (for instance, this is how the self-injury of swallowing was modified over several decades). The author himself was ableby a combination of light self-injuries and fabricated tests by helpful prison and freedom physiciansto obtain a release from prison due to 'poor health.' 11

Conclusion
Many of the types of interactions described in this article have been misinterpreted or have gone unrecognized in earlier studies. Their identification was possible thanks to the author's status as an actual prisoner. For example, researchers and prison personnel considered almost all seesaws as legitimate suicide attemptsgenuine 'cries of desperation.' Similarly, the test of fagotization was perceived as a rape attempt and various little games were interpreted as the mindless torturing of rookies (Michalski and Morawski, 1971: 2, 12). In the extant literature, the dominant mode of explaining the partially recognized outcomes was focused on the psychological factors of 'temptation and humiliation' (Kosewski, 1985) or on the economic organization of particular prisons (Moczydłowski, 1992).
Another often misinterpreted feature of prison institutions was their contribution to maintaining order. Gambetta (1996) has pointed to the order-creating function of the Sicilian mafia. It has also been argued that prison gangs have contributed to decreasing violence in prisons (in the context of the United States, see Trammell, 2012 andSkarbek, 2014; in South Africa, see Lindegaard and Gear, 2014). In this paper, I have focused on specific subcultural institutions and procedures that have curbed violence. Wherever unstructured interactions of the Hobbesian-Zimbardo type generate high transaction costs, or potential benefits appear available for capture, institutions evolve The partner can exercise low effort (L), resulting in the seesawer's death, or high effort (H), which limits the probability of death to a small 1-p. Both players prefer a successful rescue to the seesawer's death. The partner's payoff for low effort (i.e. a sloppy rescue) is the lowest. The probability of rescue p ≈ 1 (almost one); 1-p is the small probability of an error.

11
In my experience, the vast majority of 'prison escapes' come from skillfully faking illnesses and self-injuries, combined with physicians' help. 'Physical' escapes are extremely rare. spontaneously, and individuals exploit previously unrealized opportunities. Therefore, the high, and even potentially fatal, cost of fighting in prison motivates the emergence of institutions that limit violence or exploit the threat of violence in various ways.
A starting point for the analysis of many potentially violent confrontations is the game Chicken. Violence-reducing institutions of the grypsmen's subculture change the parameters of this game to make the eruption of uncontrolled violence less likely. Sorting inmates into 'tough' grypsmen and 'weak' suckers or fags reduces the motivation to play Chicken. If an inmate has exactly the same strength as his opponent, or at least knows that the opponent will fight to the death, then the act of fighting is unreasonable. On the other hand, when a tough inmate faces a weak one, the tough one can safely use the threat of aggression against the weaker one, usually for petty theft in the case of a sucker, or for sexual or other humiliating services in the case of a fag. Both players know that the grypsman is committed to fighting; in equilibrium, the sucker or the fag accepts various levels of exploitation without unnecessary additional violence.
Argot duels, fake fights, and conflict resolution change the game of Chicken with a potentially deadly outcome (from mutual fighting) into an interaction that does not necessarily involve serious injury. Membership in a coalition strengthens the defensive power of an inmate, and it provides additional insurance against aggression by increasing the cost of an attack. Finally, in the case of various self-injuries, an inmate deceives a prosecutor or warden into believing that he is desperate and committed to playing a potentially lethal confrontational strategy, when in fact, the inmate's commitment is faked.
Subcultural violence reducers are neither perfectly effective nor implemented by idealists with noble motivations. The players are often cruel and they make mistakes; their understanding of social institutions, legal procedures, and the human body are often painfully inadequate. Some procedures for selfinjury may produce negative consequences for the initiators and could arguably be classified as perverse disorders (Martin and Storr, 2008). In spite of their shortcomings, violence-reducing institutions as a whole must significantly limit violence, since the transactional costs of implementing them can be quite high and not worthwhile if the benefits do not exceed those costs. Given the observed institutions, we can be assured that completely unruly and unmitigated violence in a prison of the Hobbesian-Zimbardo type is an exception rather than the norm. A brutal 'war of all against all' would require exceptional circumstances of an institution-free environment similar to those that were artificially and accidentally created by the Zimbardo experiment.