Research on social networks has been mostly about their static properties: describing their shapes, and showing the effects on individuals of being in one position or another in various kinds of networks. But what causes networks to change shape? John Levi Martin’s book Social Structures (Princeton University Press, 2009) takes on the question. Some kinds of networks can grow from small to large; other cannot: they split, conflict, or devolve back to earlier forms. Thus the structural question of human history is how small networks, originally found in families and bands, can become the big networks that make up tribes, communities, states, armies and bureaucracies.
Building big social structures isn’t as easy as you might think. The most common forms of networks, Martin demonstrates, are dead ends as far as growth is concerned. And they are especially dead-endy as far as turning into organizations: institutionalized structures that coordinate people, carry out complex tasks, and generate power to control others and make things happen.
Take the simplest and most commonly studied kind of network: a small circle of friends, AKA a clique. Friends are those who are positively and reciprocally tied to each other; they like each other, choose each other. But if there is going to be reciprocity, friendship circles are necessarily small. Empirically, not everyone in a clique chooses everyone else; and Martin argues this is unlikely if they need to match similarities on a variety of dimensions; in any case “the interactional demands of the clique grow exponentially with the size of the group” (p. 72).
Theoretically they are held together indirectly as friends-of-friends. But balance theory-- the friend of my friend is my friend; the enemy of my friend is my enemy-- is in fact empirically weak. The further away the links go from the primary friendships, the less likely other people are to follow along. Martin here provides some needed theoretical house-cleaning; balance theory, although seemingly plausible, is a terrible model for strategic actors, reducing their options for negotiating and for building coalitions, and eliminating flexibility and surprise. Pressures to balance, we can infer, operate chiefly within a very small number of links; any bigger structures have to jettison balance, especially at crucial times of change.
Comment from IR theory: Friends are persons who engage repeatedly in successful Interaction Rituals with each other. To do so they must achieve a tight mutual focus of attention (seeing things similarly from each other’s point of view), and a shared emotion. The process typically involves a barrier to outsiders who do not share their focus or emotion and who would disrupt the intensity of their interaction. The symbols generated by micro-rituals of friendship are built out of particularistic experiences, and thus clash with other ties built on different particularisms. This is why friendship circles are generally small, and tend to fall apart when too many people are introduced at greater chain lengths.
Another supporting argument from micro-sociology: For balance theory to work, friends-of-friends must come into each other’s presence, where there is IR pressure from prior friends to keep the balance by achieving mutual focus and shared emotion. We don’t know much about such situations empirically, but I would conjecture that meeting friends-of-friends can generate artificial Goffmanian performances and less-than-successful IRs. On the negative side, there is evidence that individuals tend to resist being pulled into other people’s fights, and their support falls off with distance. Martin cites attitudinal data that this is the case. It is supported by my reanalysis of vendettas in Violence: A Micro-Sociological Theory [p.231] showing that the tendency to retaliate dwindles with time and distance -- the reality of confrontational tension/fear predominates over the ideals of balance theory. And Pieter Spierenburg, A History of Murder [Cambridge: Polity Press, 2008] shows that medieval European vendettas were often settled by negotiations and reconciliations, rather than endless chains of retaliation.
Thus friendship, built on the principle that alike-attracts-alike (and also that being friends makes one more alike), is not a structural basis for expanding beyond rather small networks. Martin amplifies the theoretical point by examining exchange theory, where reciprocal goods and favors are sent back and forth. But exchange structures, especially marriage systems in early societies, are stable only when they are either small scale and direct, or if indirect and generalized, they must be preconditioned on equality. This is not a pathway towards large-scale organization of power and deliberate coordination.
What about eliminating the principle of reciprocity and introducing inequality? This generates somewhat more complicated structures. A non-reciprocal version of a friendship network is what Martin calls a popularity tournament. Here, you can choose someone as your friend, or someone you would like to be your friend, but they don’t necessarily have to choose you. This introduces inequality because some persons are more popular than others. The structure could be larger, since it is held together by everyone’s relative popularity ranking, their score of how many people choose them as desirable friends. But although this is hypothetically possible-- especially from the point of view of a researcher who collects all the scores-- it does not work out so clearly in reality. Perfectly ordered hierarchies rarely exist. There is considerable research on high school status systems; and yes, the popular kids tend to hang around together, above those of moderate and shrinking popularity on down to the outcasts at the bottom.
But Martin’s examination of the evidence shows that lots of people in such popularity hierarchies lack a clear position. Some very good research on the development of hierarchy over time in children’s summer camp groups found that the top boys skirmished, first with everyone, then increasingly with each other, thereby settling on a clear ranking as far as they were concerned; but the less popular boys simply focused on the top boy, and lacked ranking among themselves. The girls’ groups were the opposite: they concentrated on the most unpopular girl at the bottom, who accepted their insults without retaliating, while the popular girls made more of a homogeneous clique.
Here we shift over to another model, the pecking order. This is literally found among chickens: chicken A pecks chicken B, who submits to A but pecks chicken C, who pecks chicken D, and so on all the way through the farmyard. But Martin surveys the whole field of animal networks and concludes that these tight hierarchic structures do not exist among humans. Pecking order is an engaging metaphor but it is badly overextended. This is not the way humans build large structures, nor indeed even small structures.
Two comments: Solidarity/friendship easily shades into conflict and domination. Starting with friendship we get conflict. Put another way, solidarity ritual is the primal building block at the micro-interactional level (this is what IR theory provides), but extending the numbers of persons generates inequality and conflict.
Second comment: Even approximate pecking orders or popularity hierarchies depend upon conditions that often do not hold. Quasi-tight hierarchies arise when everyone is caged, or to use another metaphor, inhabit a goldfish bowl. No one can leave, and there is high surveillance, so everyone knows about each other and their reputation. This is why the most oppressive status hierarchies are found in schools (especially boarding schools) and in prisons.  But where networks spread out in space and people assemble or not on idiosyncratic schedules, the status order becomes vague or disappears; big universities are much less caged and much more anonymous than high schools, and big cities feel much freer than small towns, at least for individuals who might have been dominated in more caged structures. It should be noted that most of the animal studies that turn up structures like pecking orders are in highly caged situations: literally in chicken yards, and nearly so in experiments with fish where they are made to view each other’s fights in adjacent tanks and then are released into a common tank. Historically, it may be the case that humans were caged when they lived in small villages, and especially when they had to stick close together because of lack of good farmland, and because of danger of violence from adjacent groups; but the earliest humans lived in hunting-and-gathering bands with lots of open space, and these groups frequently split by wandering away whenever anything disturbed their harmonious relationships [Alexandra Maryanski and Jonathan Turner, The Social Cage. Stanford Univ. Press, 1992]. Bottom line: Martin is still right; historically neither loose egalitarian structures nor tight dominance hierarchies were the path towards complex organization.
 Murray Milner, Freaks, Geeks and Cool Kids [Routledge, 2004] shows this structure is common in American high schools, even when youths are free to go home after school, because of another feature of quasi-total institutions: the split between staff and inmates, with its no-snitching rule. Teenagers achieve status above all by avoiding adults’ control; hence the autonomously ranked popularity center of the kids is the elite, and nerds who are most concerned with pleasing teachers are low status. From the point of view of the cool kids / party animals, the nerds are low because they lack social skills; this is another way of saying they lack the emotional energy to do the sociability rituals which generate endogenous collective effervescence. Popularity tournaments, as Martin’s materials show, collect EE at the top and deplete it at the bottom.
What Martin is looking for is the historical path to transitivity. That is to say, where something is propagated through successive links of a network: A sends it to B, who sends it to C, and on to D etc. The something could be power, orders, guidance, information. A pecking order is transitive, but unfortunately for theory (if perhaps fortunately for human experience) these do not really exist. Transitivity is the key to building organizations: networks that are both big and intentionally coordinated. Martin theorizes that the pathway must come from unequal forms but starts with the basic human tendencies for equality and solidarity in small networks. He settles on the wheel-and-hub structure: one person at the center, with links to otherwise separated individuals. Wheel-and-hub structures can be concatenated without losing their shape.  Hubs can be connected by super-hubs, and those by super-super hubs, and so on. Graphically, wheel-and-hub shapes can be redrawn by pulling the centers upward on the page, revealing a repetitive tree-and-branch structure.
 Martin notes this is not true when several pecking orders or popularity tournaments are brought together. Even if each one has a clear ranking A> B> C> D>... and A1> B1> C1> D1>..., the new amalgamated group is unlikely to have a clear rank ordering all the way down. A may turn out to rank slightly above A1, since the negotiation among elites is what creates the merging of the groups in the first place, but with even modest numbers involved it is unlikely that every pair will test each other out. Thus merging several hierarchies, based either on individual particularities, or on different dimensions of ranking, results in an amorphous mix. The top and bottom categories may be more or less clear, but the middle takes on the vague characteristics that modern people call “middle class”.
Historically, this is the Big Man structure in the period between small tribal/band societies and the rise of aristocratic stratification and the coercive state. The Big Man is the hub of a wheel of friendly exchange relations; his isolated friends give him their surplus produce, which he in turn redistributes by giving big festive banquets. In effect, the Big Man is the high school party animal of prehistoric times, except that his aim is to include as many people as possible in his parties, whereas the high school elite pride themselves on who they can exclude.  There is a puzzle in this redistributive model: if the outlying participants, in the aggregate, only get back the goods they themselves have put in, why bother to go through the Big Man? But this puzzle only exists for narrow utilitarian materialist premises. From the point of view of IR theory, the festive redistribution provides something the isolated individuals around the rim of the wheel can’t produce on their own: a high level of collective effervescence, a self-enhancing group emotion. Martin says the hub produces more than utilitarian goods, namely prestige. IR theory would say, it is not just the Big Man who gets prestige; he does, since he is at the center of attention, and acquires status as a sacred object symbolizing the emotional solidarity of the group. But lower participants also get something they would otherwise lack, the emotional energy of participating in a successful collective ritual, even if they have to give away some of their status equality.
 This is because the Big Man is trying to expand prestige by creating a widening network, in the absence of any other structure. The high school partying elite live inside an already very structured world-- the bureaucratic high school, the class structure of their parents-- which they will never displace by their own structure, no matter how big it might become. Hence they practice what Milner calls “pure status stratification”, reverting to one of Martin’s primitive forms, a popularity tournament.
Martin renames the wheel-and-hub / Big Man pattern as patronage structure, which gives us the familiar, useful terms, patron and clients. It is a non-coercive form of hierarchy, at least initially, a friendly relationship of mutual support.  As it concatenates upward, patrons becoming clients of super-patrons, there become more possibilities of things patrons can distribute downwards to their clients. After states come into existence, these favors can be material; for instance the political “machine” distributes jobs, eases bureaucratic licensing procedures, etc. But this is jumping ahead of the historical story; patronage structures can continue to exist once newer forms of organization are there, but we have yet to explain the historical path, how organizations like the state emerged out of prior network structures in the first place.
 At least initially, because the outgrowth of feudal hierarchies is a haughty system of aristocracy. Viewed analytically, the patronage system comes back and sometimes even thrives under modern conditions, in the form of underground organizations like mafias. A mafia uses coercion, especially in sinister threats and dramatically terrifying exhibitions; but nevertheless when a stable regime of mafia patron-client relations exists, clients often feel there is some advantage in having a patron/protector, who can do things for them that the impersonal bureaucratic state will do only in a cumbersome and unreliable way. A similar argument is made in an overly rationalistic and utilitarian form by Diego Gambetta, The Sicilian Mafia [Harvard Univ. Press, 1993] ; and with an emphasis on the symbolism and pre-modern character of mafias by Marco Santoro, La voce de padrino. Mafia, cultura e politica. [Verona: Ombrecorte, 2007]. Martin specifies just what these pre-modern structures are.
In a pre-state situation, tree-and-branch structures of patrons and clients take on the “feudal” form (using the term crudely rather than technically), in which the main thing they concatenate upwards is military force. The main things they can distribute downwards are, first of all, protection; secondly, if their coalition wins out over enemy coalitions, they may conquer and plunder, passing along the spoils of war, especially land. This is one of the ways that material inequality grows-- bearing in mind that inequality in prestige, inequality in ritual resources of emotional energy and centrality in the attention space, precede and are the mechanism for accumulating material inequality via building the patronage network. Martin comments that patronage structures thrive in inequality, but they also bridge backwards to earlier, conservative social forms that operate through rituals of friendship and some vestiges of equality. This is what Martin has been looking for: a type of network that concatenates upwards and outwards without upsetting structural cohesion.
But we’re not yet at modern organization. One of the things that stands in the way is an alternative form of expanding social networks, the concentric circles model. Consider kinship networks. These would seem to be able to expand local links into large structures; marriages create relationships among cousins, uncles, and so forth, which can be combined at great length. Under varying terminologies, families can expand outward to extended families, clans, subtribes, tribes, and confederations. Each smaller unit is nested inside larger units. The larger ones may exist only intermittently, mainly for purposes of large-scale feuds or wars; when there is no call for the federation to fight, attention may focus on enemies nearer by, such as rival families or clans. But this is not a pathway to modern transitive organizations with centralized command, planning, and coordinated division of labor. Not only is the kinship federation episodic rather than continuous, existing only in the emergency of a specific war configuration; but also sub-units may quite whimsically shift their participation; tribal federations were notoriously volatile. (This is one reason why Julius Caesar’s much smaller but differently organized Roman legions could conquer vast areas of tribal federations such as Gaul.) Federations were also quite egalitarian, each leader jealously guarding his decision-making rights vis-à-vis his peers; suspicion that a strong war-leader or charismatic politician was setting himself up as permanent king was a typical reason for federations to disintegrate. 
 a weakness that Caesar deliberately played upon. Instead, he offered Roman “friendship” in a guise familiar to the Gauls: he would simply be a more resource-rich patron, handing out largesse in return for voluntary cooperation. The trick was that Rome never allowed its foreign clients to leave a patronage relationship, once established, and had the organization to enforce it. See Julius Caesar, Gallic Wars.
One might simply view kinship organization, especially in its most extended forms, as an obstacle to modern organization. This is true, and kinship structures (better formulated in Weber’s concept, patrimonial organization, which adds the elements of household base and pseudo-kinship ties) is successively diminished if not abolished by its opposing ideal type, bureaucracy. But the concentric circles of large kinship aggregations do contribute something on the cultural/conceptual side: they create categorical identities, people who designate themselves as members of X. Although kinship networks are elaborations of particularistic relationships, they acquire an abstract identity when they are used repeatedly, even if episodically, for wars against other categorical identities. In this way, people’s mental maps become populated by big units that can at least vaguely guide action, chiefly of the Us-against-Them form, and thus are the precursors of ethnicities and nations. Martin weaves these categorical identities into his picture of modern social structures when he comes to political parties.
Now we’re getting near. The rise of the state, for Martin, is a further transformation of the concatenated hierarchy of patrons and clients. As wealth accrues to the elite, it turns into aristocracy. We should add, on the ritual and symbolic side, near monopolization of pomp and circumstance at the top, which exalts the sacredness of the elite and debases the remote non-elite; primitive egalitarianism, preserved at least in form in patron-client relations, largely disappears as aristocrats rise higher and peasants sink lower. But what is still lacking is transitivity, effectively centralized command and horizontal coordination.
Martin traces this last, fateful transformation in two spheres: armies, and political parties. It has become near-orthodoxy in recent historical sociology that the rise of the modern state was led by the so-called military revolution: the growth of large, expensive, standing armies, which in turn required rulers to institute centralized tax-extraction systems, leading to state penetration into society; eventually limiting the power of patrimonial/kinship structures and replacing them with bureaucracies; and these in turn opening up many more spheres for state action, including as tools used for both authoritarian and popular/liberal movements.  Martin quibbles with the military revolution concept, raising the question whether it isn’t necessary first to show how armies became transformed into command structures-- the holy grail of transitivity.
 In the works of Charles Tilly and Michael Mann, among others. Norbert Elias called the process an elimination contest in which states which failed to undergo this process were swallowed up by those who did.
Schematically, there have been three types of armies. First, unorganized crowds led by individual hero-exemplars who rushed into combat, hoping to inspire others to follow. Second, feudal or patronage-pyramid levées, in which each successive layer of patrons brought his own clients to the war. This could result in sometimes quite large armies in the field, but it had very little command structure; a higher lord might be able to command his personal household companions or bodyguards, but his influence on patron/leaders one level down was nominal, and nonexistent on levels further away, who recognized no relation with the paramount lord.
Finally, there is a genuine command structure, the bureaucratic army in which every rank is enrolled as part of an overall organization; patron-client links are transformed into officers appointed from the top, passing along binding orders from level to level, and in the other direction, passing information upwards. Ideally, this resulted in a coordinated strategy with many separate parts contributing to an overall plan, with information passing to a central commander who could assess new developments and take appropriate action. In reality, as military analysts since Clausewitz have noted, the ideal command-and-communications structure was often shrouded in the fog of war, and actual battlefield results were more chaotic. Martin recognizes this point, indicating that a degree of decentralized initiative is valuable to counteract the problems of relying on an out-of-touch central command. But the shift towards centralized command structures is significant even if just a matter of degree; one can argue that it made possible larger but also more maneuverable armies. Elsewhere I have argued that even if they tend to break down organizationally in battle, the army that breaks down somewhat less than its opponent would win. [Collins, “A dynamic theory of battle victory and defeat.” Cliodynamics, 2010: 3-25; http://escholarship.org/uc/irows_cliodynamics ]
Perhaps most importantly, the command structure of the army could be transferred to peacetime uses, giving a model for civilian bureaucracy and thus an organizational form through which the state penetration process could proceed. As historians would point out, this is a schematized view of how tax regimes and other government organizations developed; there are a number of other lines of development (such as the development of rational-legal bureaucracy in the church, becoming transferred to secular purposes), and many struggles among rival organizational forms in the non-military world of politics and administration. Martin does reinforce a key point that is implicit in Weber: bureaucratic command structures were not pioneered in the realm of economic markets, but only transferred there rather late in the development of capitalism. Bureaucracy is not designed for economic efficiency, but to maximize control, from the top organizational command down through successive layers to the bottom. Martin is chiefly interested in the analytical question of how transitivity was achieved; but this is not an evolutionary functionalism in which everything tends towards ever-greater efficiency. Thus it may well be the case that greater efficiencies-- for the multiple purposes of human actors-- can be achieved by backing off on occasion from the bureaucratic omni-transitive form.
Finally, Martin discusses the rise of parties. These are special-purpose political organizations, that are both bigger (and hence more participatory, more democratic) and more programatically goal-directed (more ideological) than previous kinds of political action-networks. How else could politics be done? One type is a personal faction, a clique of friends trying to get themselves in office just for the sake of the power, prestige, or wealth it may bring. In a large state, especially where masses of the ordinary population have the resources to mobilize for political action, personal factions are not big enough to get enough votes or supporters; a solution is to expand into a political machine, which is the old standby, the patronage pyramid. A second type is the interest bloc; i.e. a set of cliques who coordinate their action, for instance voting the same way in a legislature, not because they have any overarching organization but just because their interests happen to coincide (e.g. politicians from different parts of the country who all want military spending, or farm price supports). Notice a major analytical difference between the first and second types: the personal clique/patronage pyramid is endogenous to politics; it has no interests and no ideology apart from trying to get itself into office, and taking whatever spoils it can. The interest bloc is exogenously driven; the interests are already there, and politics is just a tool for furthering those interests.
The third type is the political party. Martin argues, using the early political history of the USA, that the party arises from a combination of the other two types. It expands the vertical hierarchies of the patronage pyramid, motivated by the competition for new voters, that successively widened the effective franchise. But this pushed up against a limit of the patronage model: if there are only a certain number of government jobs and so forth to go around, a mass party can’t realistically promise that all of its supporters will share in the spoils of office. This weakness of endogenous politics can be remedied by appealing also to exogenous interest blocs, but now organizing them through the horizontal structure of a national party organization. Early American politicians like Jefferson and Hamilton began to make national tours to drum up support in other parts of the country than the region where their own factions were based.
Now interest-group politics turned into ideological politics: abstract principles and slogans which cut across locations and advocate ideals that motivate masses of people for political action, quite apart from what they personally can get out of it. Micro-sociologically, I would add that the shift to ideals is a shift to public political rituals, such as parades and mass-assembly campaign speeches, harnessing Durkheimian ritual effervescence to idealistic slogans. Structurally, we can weave in here a strand that Martin had shown in an earlier chapter emerging from the concentric circles of kinship federations: categorical identities of large groups of people, otherwise unconnected by strong networks, can now be invoked as part of the ideology of political action. (Liberty for the X-group! Cast off the domination of the Y-group! Let us all unite as the Z-group!)
Martin’s depiction of political parties is sophisticated. As a mixture of political action-structures, parties are always pursuing a bundle of somewhat contradictory aims, and hence hiding some of what they are doing that contradicts one or another part of their public rhetoric. As factions, they are concerned to get themselves in office; as Martin notes, the primary axis of politics is always the Ins and the Outs. Early American political parties formed themselves along this axis, not because they represented different classes; and indeed most researchers on politics everywhere have tended to find that politicians all tend to come from the same social class. Nor did they initially represent exogenous interests (economic or otherwise) which then organized themselves as parties. The process was more the other way around: previously existent political networks, consisting largely of personal, particularistic ties, began to take up positions in the field of interests, either to keep themselves in office, or to oust the faction in office.
The whole thing is pasted over by ideologies. Here at the end of the book, one can say the same thing Martin says near the beginning about balance theory: effective politicians are those who operate with maximal flexibility and room for maneuver, and thus refuse to be tied by eternal divisions either of friends and enemies, or of ideology and counter-ideology. Ideologies are seized upon because they coordinate local factions into the big alliances-- which is to say the big lines of opposition against common enemies-- that maximize chances of winning electoral power. Here ideology is a substitute for a command hierarchy, insofar as it coordinates farflung numbers of people with a semblance of deliberate direction.
Thus the long-term evolution of network structures is not towards any single form. For the more-or-less contemporary era, from about mid-20th century through early 21st century, one might say that almost all the network types Martin analyzes throughout his book continue to exist. Some are quantitatively much less prevalent than they once were, especially the patronage pyramid that Martin singles out as the historical bridge from small to large network structures. At one end, in personal life, we are still structured by small friendship cliques, and where there are locally caged arenas, by popularity tournaments. At the macro end, transitive power is exercized in bureaucratic command hierarchies, mitigated by its own inefficiencies and thus periodic swings to decentralization; and the whole is enveloped by the hybrid forms comprising political parties and social movements.
Martin’s book encompasses a complex world. But is it complex enough? He doesn’t say much about economic markets. Presumably this is because network sociologists see economic markets as embedded in networks; also markets are internally structured as repetitive networks instead of the ideal-typical everyone-compares-offers-with-everyone fiction of conventional economics. Moreover, Martin argued early in his book that exchange markets could not produce the structures he is looking for, leading towards transitive command hierarchies. Thus markets get little attention in Martin’s historical account of network transformations.
Here we see some of what theoretical tasks are left to do. Harrison White and his school have made the major effort toward reconstructing production markets as particular types of networks, both as self-reproducing upstream-downstream flows and as mutually monitoring competitors seeking unfilled niche spaces. And White [Markets from Networks, Princeton Univ. Press, 2002] has made pioneering strides in theorizing what transforms different types of market networks into other types. Very abstractly, there is a resemblance to Martin, insofar as both are concerned with what kinds of structures are stable or unstable, self-reproducing or self-transforming. But combining White’s approach to markets with Martin’s analysis of chiefly political networks is a huge task, and it is no gripe against Martin that he stopped with the sizable chunk he has managed to bring into theoretical order.
Martin’s Social Structures is the best systematic theorization of network mechanisms and structures yet written. It breaks out of the tribal enclave of much previous network analysis, bringing it to its rightful place at the center of sociology.