Since the Chicago School's sociological studies of foreigners arriving at its doorstep (Thomas and Znaniecki 1927; Wirth 1928) and British anthropology's studies of mobile workers in southern Africa (Richards 1939; van Velsen 1960), migration has been a central concern of ethnographers. Today's era of “globalization” presents methodological challenges that may seem novel, though the extent to which globalization is new, a continuation of a secular trend, or a return to an earlier era is the subject of usually abstruse debate (Held and McGrew 2000; Urry 2000). For ethnographers, a research agenda has developed around the analysis of how the global intersects with the local in the experiences of individual agents (Appadurai 1991; Marcus 1995; Gupta and Ferguson 1997; Burawoy 2000; Gille and Riain 2002; Amselle 2002). International migrants are critical research subjects in that endeavor in both cultural anthropology and sociology (Schiller et al. 1992; Kearney 1995; Brettell 2000; Foner 2000; Kyle 2000; Levitt 2001; Smith 2005). Drawing on five years of ethnographic fieldwork among Mexican migrants in Mexico and the United States, this paper argues that four methodological strategies can usefully guide ethnographic explorations of the relationships between migrants, places, and culture, without slipping into the mire of “globaloney” (Favell 2001). I propose a way of defining the field of study as multi-sited and bi- or multi-national; executing a study in a historically-sensitive way through archival work and revisits; and supporting claims about the broader significance of ethnographic work through a revision of the extended case method.

Ethnography is a “family of methods involving direct and sustained social contact with agents” (Willis and Trondman 2000, p. 1). Here I use it in the broadest sense to include methods of intensive interviewing as well as participant-observation. The first strategy draws on Marcus (1995) and Hannerz (1998, 2003) to show how four different types of multi-sited ethnographies reveal the full scope of the migration experience and its impacts. Exploring sites that are linked to each other is a way to rejuvenate the field of comparative ethnography that is thought to be ailing because cultures are no longer discrete units (de Munck 2002). Despite the practical and epistemological problems of multi-sited fieldwork, it offers advantages for gaining access to members of multi-sited networks and explaining the effects of place on a variety of outcomes. The most serious hazard of stretching research resources too thin can be reduced through a strong theoretical orientation and models of collaborative work established generations ago.

A second strategy is stripping off the national blinders that restrict the construction of the field, and integrating both sending and receiving country sites. This has been the welcome position of the transnationalism literature. Yet even that literature, which rightly warns of the dangers of “methodological nationalism” (Wimmer and Glick Schiller 2002), has fallen into the trap of unconsciously defining its subjects in national terms. I illustrate the benefit of bringing domestic and international migration into the same analytic frame by discussing a comparison of hometown associations in major U.S. and Mexican cities formed by migrants from the same provincial Mexican town. This strategy isolates and reveals the political quality of international migration and suggests commonalities and differences with domestic urbanization.

The third methodological strategy is to historicize the field. Rejecting as ahistorical the concept of “deterritorialization” driving many migration ethnographies (Appadurai 1991; Basch et al. 1994; Kearney 1995; Laguerre 1998; Tsuda 2003), I show how historicizing a Mexican sending community and its satellites through local archival work, oral histories, and a revisit of an earlier study avoids the synchronic trap of the “ethnographic present” (Comaroff and Comaroff 1992). This strategy provides further evidence that the claims of novel forms of community sundering all manner of boundaries have been exaggerated.

A fourth strategy advocated here is the development of research programs in which ethnographic case studies contribute to the elaboration of migration theories. I describe how ethnographic studies have advanced the assimilation research program and argue that the transnationalist approach that gained currency by positioning itself against the assimilation literature can be used to refine the assimilation program. Amending dominant perspectives on the extended case method (Burawoy 1991, 1998), however, I argue that establishing the representativeness of a case is necessary to refine a research program. Finally, I conclude with suggestions about the way multi-sited work, removing national blinders, historicizing the field, and elaborating a research program are useful to ethnography more generally.

MULTI-SITED FIELDWORK

The “field” of ethnographic inquiry is not simply a geographic place waiting to be entered, but rather a conceptual space whose boundaries are constantly negotiated and constructed by the ethnographer and members (Gupta and Ferguson 1997; Emerson 2001). The methodological mandate “to follow the people” (Marcus 1995) as they travel between localities takes seriously the movement that constitutes the migratory process. Migration inherently means both emigration from some place and immigration to another, implying that research should include both sending and receiving areas (Nyíri 2002; Glick Schiller 2003; Sayad 2004).

Among a growing number of studies adopting a two-site strategy, Smith (2005) examines migration between a town in the Mexican state of Puebla and New York City, demonstrating migrants' integration into New York at the same time as many remain deeply engaged in the political, economic, and cultural life of Puebla. A second strategy is to compare multiple destinations in the same country for migrants of a common origin, as Guarnizo, Sánchez, and Roach (1999) have done in their comparison of local factors in Los Angeles and New York that help explain varying degrees of homeland ties among Colombians. A third strategy of studying a migrant sending community and its satellites in multiple receiving countries, as Tilly and his associates (1994) began to do for Italians from the village of Roccasecca dispersed in Lyon, Sao Paolo, Buenos Aires, New York, and Toronto, sets up a natural quasi-experiment controlling for origins that explains how receiving contexts pattern migrants' economic mobility. Multinational fieldwork need not include the country of origin to yield analytic leverage from the multi-sited method, as Ostergaard-Nielsen (2003) shows in her comparison of Kurdish nationalists in different European countries that explains how institutional receiving contexts affect trans-state political mobilization. In their collaborative study of immigrant children from Yemen, Korea, Mexico, and Central America living in Los Angeles and Oakland, ethnographers Orellana, Thorne, Chee, and Lam (2001) develop the concept of “transnational childhoods” and argue that the migration literature should attend to the role of children and parenting in organizing international migration and integration. Their comparative strategy highlights the way that different historical stages of migration, social class, government policies, and cultural practices shape transborder families in different ways for each national-origin group.Footnote 1

Several cautions and objections about multi-sited fieldwork have been advanced. First, multi-sited work tests the limits of a method usually thought to rely on deep, local knowledge of everyday interactions as a means to understand members' experience. The requisite intensity of fieldwork and linguistic competence may be difficult to achieve in multiple sites, with consequent variation in the quality of the fieldwork and the ability to make systematic comparisons between sites (Marcus 1995; Hannerz 2003). As Burawoy (2003, p. 673) puts it, “Bouncing from site to site, anthropologists easily substitute anecdotes and vignettes for serious field work …” Similarly, Gille and Riain (2002) warn that the “methodological imperative of being there is replaced by that of chasing things around, things that are identified more by the ethnographer's interests prior to entering the field than by the field itself.” These are sensible cautions, though the “field” never simply guides research. While the ethnographer's convenience and chance may intervene, the dialectical engagement of a priori theory with encountered evidence should guide the on-going construction of field and decisions about where to focus research energies (Snow, Morrill, and Anderson 2003). It is precisely because of the dangers of stretching time and resources too thin that successful multi-sited fieldwork is even more dependent on a clear theoretical orientation and strategic site selection than work in a single site.

A second caution sounded by Marcus (1995) is that multi-sited fieldwork will lose its subaltern focus, thus weakening its potential for Critique by introducing too many differently positioned voices. This objection is unconvincing for several reasons. Researchers committed to critiquing power relations can follow the subaltern as they migrate, as Nagengast and Kearney (1990) have done in their study of indigenous migrants circulating between southern Mexico and California. Further, subalternity is situational and relational. Migrants who are exploited by capitalists in a receiving country are often capitalist exploiters of those who stayed behind, especially when migrants return home with new wealth. The strategic manipulation of class position is an important motivation for engaging in cross-border practices and return migration in the first place (Goldring 1998). Finally, participant-observation is a methodological tool, not a political philosophy, and it is a tool available for disparate purposes. Regardless of one's ideological predispositions and view of the proper relationship between practicing social science and politics, following migrants as they travel across multiple sites is a productive way to understand their experiences.

A third objection to multi-sited fieldwork is that scientific comparative ethnography is no longer possible because the cultures of multiple sites cannot be considered discrete units. Without discrete units, causal processes are not independent of each other and the logic of the Millian methods of agreement and difference breaks down (Ragin 1987; de Munck 2002). Seen from a different view, the linkages between sites are not the end of comparative ethnography, but rather an opportunity for its rejuvenation. Different source and destination localities can be selected precisely because they are linked by migrant networks, while still shaping migrants' experiences differently.

Another objection to comparative ethnography is the caeteris paribus problem that bedevils comparative study regardless of method. In non-experimental studies of social life, it is impossible to definitively isolate the effects of just one factor's addition or removal. For instance, one should not assume a given difference between two migration destinations causes variation found between migration streams sharing the same source. That connection can only be made by carefully specifying process and exploring alternative accounts. It is because the Millian methods should never be applied mechanistically by simply creating a matrix of independent and dependent variables (Lieberson 1992) that multi-sited ethnographies are best positioned to tease out the influences of different ecologies on migration processes by explaining causal mechanisms through an evidence-rich encounter with theory.

Multi-sited fieldwork also offers practical advantages for gaining access to social networks with nodes in different sites. In a study of a California labor union comprised mostly of Mexican immigrants, a multi-site strategy allowed me to show how a group of members mobilized to control the union local in California by traveling to their villages of origin in Mexico to campaign for internal union posts and by using the hometown networks as channels for patron-client exchanges. Previous fieldwork in members' Mexican sending communities was the primary means by which I gained entrée to a union suspicious of outsiders (Fitzgerald 2000, 2004). Displaying knowledge of local distinctions meaningful to members is a way to reduce the social distance between them and the ethnographer. On a number of occasions, I participated in “mental tours” of the sending region in which members tested my knowledge and reminisced about their own experiences there. For all the challenges of doing multi-sited ethnography, displayed knowledge of other sites and the people circulating among them can be a passport to entrée and partial acceptance.

The practical difficulties involved in multi-sited research, particularly when they involve multiple languages, can be resolved in part by abandoning the “lone ranger” model of fieldwork and adopting a bi- or multi-national collaborative model. Thomas and Znaniecki's The Polish Peasant in Europe and America (1927) is an early template. The collaborative dimension combined the advantages of insiders' intimate acquaintance with the social milieu and easier access with the advantages of outsiders' fresh perspectives and autonomy (Merton 1972). The binational dimension enabled the researchers to examine the full range of migrants' experiences, migration's impacts on both countries, and the causes of migration from Polish push factors to U.S. pull factors.

A multi-sited approach would have added a useful level of analysis to more recent ethnographies of immigration conducted in a single city. For example, based on intensive interviews and surveys in Los Angeles, Hondagneu-Sotelo and Avila (1997) coin the term “transnational motherhood” to describe the 40 percent of Latina immigrant mothers in their sample who worked as housekeepers and nannies in California while at least one of their children lived in their country of origin. By highlighting the government policies and structural inequalities that split families apart, they trenchantly criticize the celebratory quality of the early transnationalism literature, which characterized “transmigrants” as resisting the impositions of the nation-state. Their discussion of how these women redefined what motherhood meant when they were taking care of other people's children while separated from their own would have been strengthened by research in the entire social field of the “transnational” family, rather than relying exclusively on accounts from mothers and their retrospective reporting of life “back home.” Mothering is an interactional concept that begs for fieldwork among those separated children being mothered from afar as well as the mothers themselves.

Given the years of fieldwork required to develop the project in various sites in Los Angeles, is it unreasonable to ask for even more fieldwork in linked sites abroad? It may not be possible for one or two ethnographers to do all the work, which is why collaboration is an attractive solution to the practical problem of multi-sited ethnography. Fortunately, transnational networks of academics studying migration are increasingly becoming institutionalized through efforts by the Social Science Research Council to build the field through conferences and fellowships, a robust international migration section in the American Sociological Association, the 1997–2003 Transnational Communities Programme at Oxford, and the Red Internacional de Migración y Desarrollo linking Spanish-speaking migration scholars around the world. Special funds are available from agencies like the University of California Institute for Mexico and the United States for collaborative binational work. These types of associations and programs strengthen academic networks and promote shared intellectual repertoires about appropriate methods and unresolved theoretical issues, both of which are critical components in successful multi-sited, collaborative work. On the other hand, collaboration cannot extend the number of cases studied indefinitely without running into problems of unevenness created by differences in participants' training, theoretical orientation, and project commitment. There are inherent limits to rigorous multi-sited fieldwork, but collaborations like Thomas and Znaniecki's work on Polish Americans (1927), Orellana et al. 's (2001) study of “transnational childhoods,” and Suárez-Orozco and Suárez-Orozco's (2002) team-study in Boston and San Francisco of 400 immigrant youth from China, Mexico, Central America, Haiti, and the Dominican Republic, are often the best way to approach those limits.

REMOVING NATIONAL BLINDERS

A particular idea of the field “enables certain kinds of knowledge while blocking off others [and] authorizes some objects of study and methods of analysis while excluding others” (Gupta and Ferguson 1997, p. 4). Although the model of constructing the field as a closed local society has been discredited in both anthropology and sociology (Amselle 2002; Gille and Riain 2002), the study of international migration has long assumed an isolation of cultural units ratcheted up to a higher scale. The dominant frame for studying contemporary international migration has been what Noiriel (1991) calls the “tyranny of the national” and Wimmer and Glick Schiller (2002) call “methodological nationalism.” In this nationally restricted vision, most ethnographies outside the transnationalism literature have focused exclusively on the experience of international migrants as immigrants in the United States, according to the perspective of the sociology of assimilation (e.g. Whyte 1943; Gans 1962; Gibson 1989; Lamphere 1992; Kibria 1993; Markowitz 1993; Zhou and Bankston 1998; Smith-Hefner 1999).

An example of how an influential ethnographic work in the sociology of immigration would have benefited from a bi- or multi-national approach is Kasinitz's (1992) study of the political organizing of West Indians in New York City. He explains why a group that publicly identified itself with native African-Americans prior to the Great Depression increasingly defined itself publicly in ethnic, Caribbean terms in the 1980s. The study sheds light on how the receiving state and ethnic entrepreneurs shape successful ethnic political mobilization. Missing in this account are the political ties simultaneously growing between West Indians in New York and their governments of origin that also reinforced West Indian identities. Such ties have been described by researchers working in a transnationalist framework (Basch et al. 1994). The connection between the stories of American ethnopolitics and Caribbean long-distance nationalism cannot be explained without systematic fieldwork in both New York and the West Indies.

Yet even the transnationalists who have adopted, though certainly not invented, the strategy of research in both sending and receiving countries (e.g. Thomas and Znaniecki 1927; Taylor 1928–1934, 1933), have tended to so intently focus on the transcendence of the nation-state's borders that they often retain national blinders in another sense. Domestic migrations frequently evince the same decoupling of locality and culture that transnationalists proclaim as evidence of new ways of “being.” There is nothing inherently “transnational” about ties that create an imagined community encompassing both here and there, as the same relationship reoccurs within almost any domestic or international migratory context. For example, the connections between here and there in the form of regular remittances, sending children back home to spend the summer with grandparents, and return migration with new ideas and customs developed in the destination site, are described as “transnationalism” in the Dominican-U.S. migration circuit (Pessar 1997). Those connections are strikingly similar to ties among African-Americans in the South-North migration circuit inside the United States or among domestic “snowbird” retirees circulating back and forth between northern U.S. states and the Sun Belt (Stack 1996; McHugh 2000). The “transnational” religious ties of international migrants are isomorphic in many ways with the ties between early African-American migrants in Chicago and their southern places of origin (cf. Ebaugh and Chafetz 2002 and Levitt 2003 with Best 2003).

The “hometown associations” formed by international migrants sharing places of origin are considered the quintessential “transnational” institution because they are a vehicle for a wide range of collective practices linking migrants to family and townspeople who stayed behind (Liu 1998; Goldring 1998). Yet the hometown associations are simply a cross-border version of what anthropologists and historians have long known as “migrant village associations” made up of domestic migrants from rural areas settling in cities like Paris and Lima (Moch 2004; Skeldon 1980). In the 1920s, Iowan migrants to the Los Angeles area created Iowan associations that picnicked, through the 1960s, in the same public parks where Salvadoran and Guatemalan associations gather today. Mexican hometown associations in Los Angeles and Chicago also have branches in Mexico City and the next town down the road. The original, functionalist perspective on the migrant village associations was that they were simply adaptive institutions through which peasants who had recently arrived in the city learned to navigate the strangeness of the urban milieu (Jongkind 1974; Waldinger and Fitzgerald 2004). The parallel with the older view that the hometown associations of international immigrants were simply vehicles of assimilation is striking. However, hometown associations of both the domestic and international variety can be a vehicle for a kind of pluralist assimilation to a new context while still maintaining substantive ties to origin communities (Skeldon 1976; Fitzgerald 2004).

The on-going debate about the extent to which international migrants abroad can usefully be considered members of a “community” spanning both sending and receiving localities (Portes 1997; Levitt 2001; Waldinger and Fitzgerald 2004) would be enriched by considering the different ways that claims to community membership are negotiated in contexts of domestic and international migration. That research project requires at least three field sites — the origin locality and an international and domestic satellite. My on-going fieldwork adopting such an approach examines hometown associations in major Mexican and U.S. cities formed by migrants from Arandas in the Mexican state of Jalisco. There is a remarkable similarity in the activities, goals, and discourses of these associations over the past 60 years. The use of new technologies allowing absent migrants to participate in the life of their sending community has been described as a novel feature of the contemporary era that stimulates transborder connections (Portes et al. 1999). For example, Mexican migrants in U.S. destination cities gather to share their videos of hometown festivals celebrating migrants' return, thus creating a sense of community even among those migrants who could not return to the hometown. Yet movies shot on film were used for exactly the same purpose among Arandense migrants in the 1940s that formed “colonies” in Mexico City and Guadalajara. Migrant-sponsored modernization projects in their hometown and fund-raising visits to satellites by political and religious leaders have been basic features of both domestic and international migrant associations. As I argue in the following section, the so-called “deterritorialization” of the “transnational community” falsely implies that imagined and geographic communities were always coupled tightly. In fact, current residence is only one of many possible sources of identification with a locality.

The point is not to claim that international and domestic migrations are the same, but rather to ask how and why they are different or similar in various domains. International migration is only inherently different from domestic migration insofar as the former is political by virtue of crossing state boundaries of territory and citizenship (Zolberg 1999). By bringing domestic migration into the same analytic frame as international migration, the international and political quality of international migration is made clear. For instance, in the Mexican case, U.S. border control efforts restrict the free flow of people within the migration circuit. On the other hand, the urban receiving context characterizing much international migration may be as important in shaping migrants' experiences as the fact that the migration is international. In both cases, the experience of being a stranger stimulates recourse to hometown ties for access to all kinds of practical and emotional resources. Ethnographies of domestic urbanization have much to offer the study of international migration, or what might be called “international urbanization.” Ethnography at its best is like a camera with a zoom lens that can both capture the wide context of structure and narrowly focus on agents in a way that shows their interactions with that structure. Such a lens is well suited to distinguish between the influences of political boundaries and urban ecologies, but only if national blinders are removed to reframe the field of study to include multiple sites in sending and receiving countries.

HISTORICITY

Two of the most influential turns in recent ethnography, the postmodern and the historical, are sharply at odds. Postmodern ethnographers have embraced claims that the old constraints of space and time have been burst by a wave of “global fluids” pouring across “scapes.” Migrants are among many “fluids” including goods and information that demonstrate “no clear departure or arrival, just de-territorialized movement or mobility” (Urry 2000). In the words of Appadurai (1991), “The task of ethnography now becomes the unraveling of a conundrum: what is the nature of locality, as a lived experience, in a globalized, deterritorialized world?” Deterritorialization signifies the uncoupling of a culture with a place and the uncoupling of residence in a territory with membership in a community. As movers between polities and cultural systems, migrants are primary agents of deterritorialization in the works of Appadurai (1991), Basch et al. (1994), Kearney (1995), Laguerre (1998), and Tsuda (2003).

The major flaw in the notion of a “deterritorialized” world is the assumption that social life was “territorialized” at an earlier point. Rather, the putative coupling of locality and culture is an artifact of the social scientific imagining of exotic container villages and national container societies, which naturalizes efforts by states to create containers out of a world interlaced with the movement of goods, people, and ideas (Gille and Riain 2002; Wimmer and Glick Schiller 2002). Technologies and exchanges are dynamic and have always altered the social fields relevant to agents' experience. Using as a foil the false image of sedentary societies in which cultural and geographic boundaries coincide is no substitute for historically sensitive research into how those fields have been altered at specific times and places.

The transnational migration literature burst onto the scene in the 1990s claiming the novelty of the transnational social field, community, circuit, and other variants of the concept that migrants can retain and create substantive ties to both sending and receiving areas (Glick Schiller et al. 1992; Portes, Guarnizo and Landolt 1999). In many cases, an assumption of novelty is asserted based on contemporary research alone. For example, Ong's (1999) study emphasizing a brave new world of Chinese elites shuttling back and forth between China and the U.S. Pacific Rim implies that contemporary transportation technologies and late capitalism make sustained cross-border connections possible for the first time. By focusing exclusively on a new population of cosmopolitan elites, continuities in trans-Pacific ties among Chinese labor migrants are elided. A more historically sensitive approach has found extensive economic, media, and cultural links between Chinese in the United States and their hometowns in southern China from 1882–1943 (Hsu 2000). The most recent transnationalism literature takes the work of historians more seriously and has retreated from earlier claims of novelty, arguing now that a transnational perspective allows researchers to see transborder ties that were invisible to the assimilationist scholars of earlier generations (Portes 2003; Glick Schiller 2003; Smith 2005).

Where migration streams are relatively new, there may not be much of a historical record to excavate in that site, and claims of novelty tend to be based on a comparison between post-1960s immigrants like Dominicans and the European immigrants of the turn of the twentieth century (Levitt 2001). An alternative strategy is to examine differences over time within the same population that has a long migration history. The advantage of such a historical approach for ethnographies of international migration is illustrated by my on-going research project on the politics of emigration in the Mexican town of Arandas, which was the site of Berkeley agricultural economist Taylor's (1933) pathbreaking study of a Mexican migrant sending community. An ethnographic “revisit” of an earlier study is one way to see what is new (Burawoy 2003).

Drawing on Taylor's work and ethnographic and archival methods, I compare the ways that local government and the Catholic Church have tried to control emigration and manage its effects since the 1920s. By comparing the ways actors have defined and attempted to solve a similar set of problems, I explain why different solutions were chosen at different periods and in what ways contemporary solutions are constrained by earlier actions establishing a culture of emigration. This approach draws on Haydu's (1998) observation that revisits are a way to compare historical periods as “sequences of problem solving” for different historical actors in similar structural positions. As migration presents problems and opportunities for different actors, the concept of “sequences of problem solving” can be extended to examine “sequences of utility maximization.” For example, the contemporary trips by Arandense political and religious leaders to raise money for hometown projects among migrants in the United States are based on a model established in the late 1930s of raising funds from Arandenses who emigrated to other parts of Mexico. By integrating archival sources with interviews and participant observation, a story emerges of long-standing ties between migrants and their town of origin that defies the standard transnationalism narrative of the novelty of such arrangements.

Local archival work and oral histories can be used to reconstruct further the period of an earlier study. The archival and oral history strategies complement the revisit, particularly when the earlier work aimed to answer different questions or is empirically thin. Drawing on diverse bodies of evidence is a way to establish whether differences between an earlier study and revisit are the result of two researchers' different theoretical orientations, social positions, or real historical shifts (Burawoy 2003).

The utility of integrating archival and ethnographic research to address questions of ethnic retention and reconfiguration is demonstrated by Leonard's (1992) study of Punjabi Mexican Americans in California. Leonard combined extensive research in county vital statistics and land registry archives with life histories and participant observation to show how an overwhelmingly male Punjabi population in the early twentieth century encountered social discrimination and laws preventing Asians from owning land. Punjabi men reacted by marrying Mexican women, trying to prove they were legally Caucasian, and severing most ties with their Punjabi homeland. Their Spanish-speaking Catholic Mexican wives often became promoters of a particular Hindu ethnicity forged in rural California. As a result, recent immigrants from the subcontinent, who enjoy a more even gender balance and the possibility of on-going homeland ties, generally dismiss Punjabi Mexican Americans as culturally unauthentic. Integrating historical and contemporary sources allows Leonard to illuminate the ways in which ethnic claims and the reception of those claims shift in new environments. In short, these examples offer a way of integrating participant-observation, intensive interviews, archival research, and previous studies—thus turning the problem of the “ethnographic present” into historical depth.

GENERALIZING FROM THE PARTICULAR

Ethnography's capacity to show process in fine-grained detail and to open black boxes to show mechanisms causally linking independent and dependent variables is a recognized strength of the method. Ethnography is also particularly well suited to describe and explain the articulation of macro structures with members' lived experience, micro-interactions, and a deep appreciation of members' meanings. That same strength inherently limits the ability of the ethnographer to study a wide range of cases intensively. A commonly held view is that ethnographies can only be ideographic because they are case studies (Blumer [1939] 1969).

Indeed, ethnographies of international migration include many case studies that make few discernible nomethetic claims. For example, Markowitz (1993) describes how Soviet Jews in New York City in the 1980s arrived as individuals without intending to create a Soviet Jewish community, but found themselves doing just that as encounters with Americans — including Jewish-Americans – created a sense of a collective “Russian” experience that could not be erased by trying to assimilate. This process is richly documented, and herein lies the book's value, but explicit lessons are not drawn about what the case means for broader dynamics of ethnic genesis or reproduction. The author concludes with the unsurprising notion that immigrants' efforts to “make the experience of life after migration not only materially comfortable but emotionally meaningful and socially satisfying … is the reason that people create communities and that group identities persist in the postmodern era” (Markowitz 1992, pp. 260–61). In her work on Cambodians in Boston, Smith-Hefner (1999) gives the reader a detailed description of the efforts of Khmer elders to transmit Khmer ideas about childrearing, education, Buddhism, and gender roles to the second generation. The book shows the importance of language as a bearer of cultural representations and explains the symbolic meanings of Khmer ritual. The Geertzian “thick description” of works such as these opens windows into interesting social worlds with which the reader may be unfamiliar, but there is no sustained attempt to connect these cases to broader theoretical issues in the social science of migration.

According to sociologist of immigration Portes (1997), all case studies are descriptive of specific instances. They do not identify issues or problems in need of explanation, identify explanatory factors, or link with other predictive statements. Case studies fulfill only one of his four elements required to construct “theory.” According to this view, the ethnographic case studies that launched the transnationalism literature were what Eckstein (1975) calls “plausibility probes,” which established whether something exists empirically that should then be taken up theoretically and tested with quantitative measures (Guarnizo, Portes, and Haller 2003). Case studies are not limited to ethnographies, however. Even a complete census of a country can be conceptualized as a case study of a particular place at a particular time — for example, the case of the United States in 2000 (see Ragin and Becker 1992). Carefully selected case studies employing a wide variety of methods — from the ethnographic to the quantitative – have been used to generate new theoretical insights and test existing theories (Eckstein 1975).

Still, how can ethnographers hope to make general arguments about anything other than their field site, much less about macro processes of globalization and transnationalization floating high above? The extended case method first advanced by the Manchester School of anthropology (Gluckman 1961) and developed in sociology by Michael Burawoy (1991) offers a more ambitious alternative to the descriptive case study by showing how a single case can yield theoretical leverage. In any scientific research program, there is a set of “core” postulates. Surrounding the core is an “outer belt” of secondary postulates that explain outcomes the core postulates do not predict. If multiple ad hoc qualifications are necessary to explain anomalies, the research program is degenerative. If the secondary postulates can be revised to explain anomalous outcomes and predict new facts, the research program progresses (Lakatos 1978). A single case study that successfully explains an anomaly by developing a secondary postulate protects the core of the research program from negation and provides an ethnographer with justifiable claims to generalize based on the program's enhanced explanatory capability. While sociology is fragmented into sub-fields and has few “mature” research programs at a meta-theoretical level like Marxism or neo-classical economics, the concept of the research program has proved useful both in describing historical shifts in social science and prescribing how to do it (Ball 1976). My aim here is not to develop a new research program, but rather to describe how the logic of a program can be usefully applied to ethnographic studies of migration.

The dominant research program in the sociology of immigration analyzes “assimilation” or “integration.” Ethnographies have pushed the assimilation program forward by showing that the different domains of assimilation (e.g. cultural, marital, and economic) are not always mutually reinforcing, and in fact, can be at odds with each other. Specifically, economic assimilation, in the sense of upward mobility, can actually be increased through ethnic retention. In Gibson's (1988) ethnographic study of Punjabi Sikh immigrants and their children's high school experiences in a small town in central California, she found Sikh students performed well despite their parents' low occupational status, financial resources, and levels of education. Her narrative emphasizes the ways in which Sikhs selectively acculturate to America in order to achieve educational success. For example, Sikh parents encourage their children to learn English and study hard in American schools at the same time as they construct and maintain thick ethnic boundaries with non-Sikhs in domains like marriage and religion.

Similarly, Zhou and Bankston (1998) mix ethnographic and quantitative school testing data to argue that Vietnamese students in a poor neighborhood of New Orleans performed well in school despite their impoverished material circumstances and low human capital when they became deeply involved in family and Vietnamese Catholic institutions that discouraged the adoption of the putatively “oppositional culture” of African American youth in the neighborhood. Water's (1999) study of West Indians and African Americans in New York City further refines these arguments by studying a context where immigrants are racially lumped together with marginalized natives. She shows that West Indians who successfully telegraph their immigrant status are rewarded by white employers and teachers who are more favorably inclined towards foreign, rather than native-born, blacks. This advantage tends to be lost in the second generation, however, as the racial lumping of native blacks and children of West Indians blurs national-origin differences.

These three studies are important works in the “segmented assimilation” literature that argues that the specific segment of society to which persons assimilate strongly influences their life chances (Portes and Zhou 1993; Portes and Rumbaut 2001). These works also refine the social stratification research program, historically built around an analysis of individualistic and socioeconomic factors like class and educational background, by demonstrating that ethnic forms of social capital must be taken into account to explain socioeconomic status in multiethnic settings.

How can these studies be so important in refining the assimilation program when they look at slivers of three populations — Punjabi Sikhs, Vietnamese, and West Indians — that even taken together are only a tiny proportion of contemporary U.S. immigrants? The studies adopt the logic of the crucial case in that they naturally approach the conditions of a well-designed experiment to test a theory (Eckstein 1975). It is commonly held that Asians tend to do well on educational measures in the United States relative to other immigrant groups because of their generally higher levels of human capital. The Punjabi Sikh and Vietnamese cases are perfect tests of this proposition, because they are generally low in human capital yet generally do well in school (Gibson 1988; Zhou and Bankston 1998). The case of West Indians in New York shows the limits of ethnic retention in the face of racial lumping. Collectively, the works make a strong argument for refining the assimilation program given the diversity of national origin groups and settlement areas involved.

EMPIRICAL REPRESENTATIVENESS

The question still remains how ethnographers know if what they find are just “outliers” on the great graph of social life. Like practitioners of analytic induction and grounded theory, extended case methodologists differentiate their attempts to generalize from the logic of quantitative methods (Glaser and Strauss 1967; Katz 1997; Burawoy 1998). The case study is a way to make claims of “societal significance” rather than “statistical significance” (Burawoy 1991, p. 281), where the former refers to the development of ideas of theoretical and practical import and the latter refers to the finding that an association of two variables is not the result of random variation. Pummeled by colleagues oriented towards quantitative measurement and sampling issues, some ethnographer sociologists have tried to change the terms of the debate by arguing representativeness is irrelevant (Burawoy 1998).

Unfortunately, the extended case method by itself does not solve the problem of generalizing from a particular. If the case is just a product of rare conjuncture, the research program is not threatened. For example, if workers only participated in their own exploitation under non-coercive regimes in the single Chicago factory Burawoy (1979) studied, that one case would hardly represent a serious challenge to classic Marxist theory emphasizing coercive production regimes. The “societal significance” Burawoy find in a single case cannot be completely divorced from the question of “statistical significance.” Unusual cases like a revolution need not be typical in any sense to have strong societal significance, but mundane cases like a small group of immigrants' socioeconomic mobility require some degree of typicality to be socially significant.

There are at least three complementary strategies for assessing representativeness (see Hammersley 1992). The first is collaboration via contemporaneous or serial ethnographies that capture a greater range of variation than is possible in one researcher's project. For example, Levitt (2003) conducted research on the role of religion in contemporary “transnational life” by working with colleagues in India, Ireland, Brazil, and the Dominican Republic. The second strategy is using existing statistics to assess the degree of representativeness of a case, as ethnographers often do using census data (e.g. Guarnizo et al. 1999). The third is combining ethnographic and survey evidence gathered either on one's own or in collaboration. The on-going Mexican Migration Project “ethnosurveys” (Massey 1987) of more than 100 migrant sending communities, co-directed by demographer Douglas Massey and anthropologist Jorge Durand, are a premier example of this sort of collaboration. Like other researchers using primarily qualitative methods, I supervised a Mexican Migration Project survey in my field site of Arandas to establish how typical it is of a wider universe of Mexican migrant source communities.

Armed with quantitative data on how representative a case is of a larger category, it is possible to convincingly adopt another strategy for developing a crucial case (Eckstein 1975). Knowing that a case is highly atypical can be grounds for generalizing if the case is an extreme crystallization of some theoretically significant phenomenon. If a theoretical prediction does not apply to the extreme case, it is unlikely to apply anywhere. For instance, Halle (1984) defends his choice of a specific chemical plant in New Jersey to generalize about the relationship of the contemporary American working class to the middle class based on the plant workers' high wages, automated jobs, and high rates of home ownership. If workers in the most privileged ranks thought of themselves as “working” rather than “middle class,” it stood to reason that less privileged workers would think of themselves as working class as well. Ethnography is obviously not a method of making statistical generalizations, but rather than dismiss quantitative concerns out of hand, ethnographers would do well to situate their studies in ways that strengthen claims to both empirical representativeness and theoretical significance.

WHEN ETHNOGRAPHY WORKS BEST IN A RESEARCH PROGRAM

The utility of using a single case or small set of cases to advance a research program is positively related to the degree of the theory's determinism and inversely related to the scope of the theory. According to a Popperian (1968) logic of deterministic laws, a single case of negation is not a fatal blow to the research program if a secondary postulate consistent with the core explains the anomaly. But Popperian formulations are deliberately made to be easier to falsify with a few negative cases than probabilistic theories. If “laws” are conceived in probabilistic terms (Berk 1988), as they generally are even in the grander versions of social scientific theory, particular negative cases can still be useful for advancing general theoretical claims under two conditions. Negative cases are most useful when the gap is large between the theoretical prediction and the outcome and an examination of the case is the basis of expanding a theory's range of explanation (Emigh 1997). Research programs are not negated simply by the discovery of disconfirming evidence, but rather when competing research programs offer greater explanatory power (Lakatos 1978). Thus, a single ethnographic study cannot disprove a research program, but to the extent that the case refines an existing program or contributes to a competing program, an ethnographic study can make a theoretical contribution that indirectly warrants claims applicable beyond the cases studied.

Ethnographies are less suited to falsify grander, even ahistorical, theories, like the theories of neoclassical economics, new economics of migration, and cumulative causation seeking to explain the generation and persistence of migration across many contexts (Massey et al. 1993). The inherently limited range of ethnographic cases means they are poorly suited vehicles for trying to reconfigure these broader theories, even if ethnographic studies do generate important insights into relevant processes and raise questions leading to more general formulations. Case studies are most useful in pushing forward theories that are restricted in their historical scope, for example, the “segmented assimilation” thesis that seeks to account for the experience of the contemporary second generation of American immigrants. Theoretically oriented ethnographic work of the sort discussed earlier (e.g. Gibson 1988; Zhou and Bankston 1998; Waters 1999) is thus most useful when it is close to the scale of the theory it seeks to refine.

TRANSBORDER RESEARCH TO UNDERSTAND ASSIMILATION

I now turn to a discussion of how the methodological strategies of stripping national blinders and conducting multi-sited fieldwork advancing a research program can advance understandings of both assimilation and “transnationalism.” While the transnationalism literature originally positioned itself against the assimilation program by emphasizing migrants' enduring ties with their places of origin (Glick Schiller et al. 1992), that literature could be used to refine the assimilation program by describing the ways that integration into the receiving society is fully compatible with the maintenance of transborder ties (Levitt 2001; Morawska 2003; Fitzgerald 2004). The use of a case study to advance that research program is illustrated by the work of Thomas and Znaniecki (1927), even though it preceded the canonical literature in both assimilation and transnationalism. The significance of a particular study in advancing a research program is often not recognized until long after the fact (Lakatos 1978).

Thomas and Znaniecki implicitly rejected the assumption that assimilation is a linear process leading to an undifferentiated American norm. Polish American institutions simultaneously promoted ethnicization as Polish Americans, incorporation into the U.S. economy and polity, and support for the “national liberation” movement in Poland. The conceptualization of the “supra-territorial” organization of Poland to include Polonia — the community of American Poles — explicitly recognized a rupture of territorial residence and cultural/ political belonging. Years before the current transnationalism debate erupted, Thomas and Znaniecki showed that assimilation to a pluralist, rather than melting pot, vision of American society is compatible with intensive cross-border social ties and long-distance nationalism.

Even where immigrants have permanently settled in the United States, the binational methodological approach of Thomas and Znaniecki would help further the assimilation program. There is a long debate in the historiography of American immigration about the extent to which ethnic differences are imported from abroad (Warner and Srole 1945) or forged in the crucible of the U.S. experience (Yancey, Juliani, and Erikson 1976). In Boelen's 1992 revisionist attack on Whyte's 1943 Street Corner Society, she argues that Whyte misleadingly portrayed a local gang in a Boston Italian slum as an American institution. Her own informal observations in Italian villages in the 1950s suggested that the institution of groups of men hanging out on street corners is a cultural import from rural Italy, which reflects the gendered division of space into public, male and private, female domains. In her revisits of the same Boston neighborhood studied by Whyte, Boelen found that many of Whyte's second generation informants, including the principal informant “Doc,” had returned to the Italian villages of their parents' birth for ten or more years of their childhood. What was the effect of these Italian experiences on their lives in Boston? Whyte (1992) responded that street corner gangs in the United States were common to any number of immigrant groups and did not seem particularly Italian. Lost in the polemical jousting between Boelen and Whyte and several commentators in a special issue of the Journal of Contemporary Ethnography was a greater question about the locus of a “society.” Regardless of whether Boelen's argument about the Italian genesis of men hanging out on the street corner as a cultural form is accurate, the society of the street corner boys surely extended to their Italian villages of origin. As Whyte discussed in his original volume, ties based on a common Italian village of origin organized much social life in the slum. Only systematic binational fieldwork could establish to what extent specific immigrant cultural institutions were imported whole cloth from abroad or were transformed or created in the United States. Ethnographic interrogation of these questions in one site is not enough, as informants' retrospective accounts of home cultural norms can become reified and obscure the disjunctures between norms and practices where social change occurs. Immigrants' frame of reference may be a home country frozen in time or deeply idealized.

Following migrants through their trajectory or circuit in an ethnographically and historically sensitive way is the best means to untangle the dynamics of ethnic genesis, retention, and dissolution. This is especially the case given the complicated feedback mechanisms by which the representational and material aspects of host culture are transmitted back to the sending areas. For example, Levitt's (2001) binational study shows that through their exposure to “social remittances,” non-migrants in the Dominican Republic assimilate towards U.S. culture in some sense without actually migrating. Systematic binational studies show the ways migrants maintain the practices and identities of their sending communities, create new ethnic forms in the destination society, and/or assimilate to specific target populations.

CONCLUSIONS

The goal of this paper has not been an exhaustive review of all major methodological issues facing ethnographers of migration. Such a review would closely mirror general issues in ethnography. Other critical issues recurrent in ethnographies of migration have been analyzed elsewhere, including discussions of the ethnographer's role as researcher and citizen (Coutin 2002) and ethnic outsider or insider (Waters 1999; Baca Zinn 2001). Nor are the issues discussed here the exclusive domain of ethnographies of migration. The argument that demonstrating some degree of representativeness is fundamental to the elaboration of research programs using anomalous case analysis is equally applicable to the extended case method in and out of migration studies. The imperative is the same whether elaborating programs in assimilation or Marxist understandings of production regimes. Likewise, the problem of national blinders pervades the social sciences.Footnote 2 The same blinders that have led migration researchers to restrict their gaze to the experience of immigrants, or to ignore the relationships between international and domestic migration, lead to methodologically suspect practices every day (Wimmer and Glick Schiller 2002). For example, comparative projects using cross-national quantitative methods in particular often assume that nation-states are the obvious and appropriate units of analysis despite what may be massive internal variation and problems of scale comparing countries as different as Luxembourg and China. Scholars writing about their audience's country of origin are not asked to justify their site selection to the same extent as scholars writing about a place foreign to the audience.

The issue of multi-sited ethnography in the study of migration also extends to the most parochial studies. Following the same informant around throughout the day is a multi-sited ethnography writ small. The “extended-place method” (Duneier 1999) has a multi-sited component as well in the sense that the researcher collects evidence from other sites to explain something of interest in the primary site. In his study of sidewalk vendors on Sixth Avenue in New York City, Duneier interviewed politicians, city officials, and lawyers to understand the laws and policies affecting his subjects on Sixth Avenue. Such an approach shares many of the goals and methods of good journalism. What I advocate here is more specific: intensive research in several connected sites selected for their potential theoretical yield. Such studies could include ethnographies of organizations like corporations and churches with members circulating among multiple nodes or the consumption of the same media product, like the Internet, in different settings. The object of this comparative ethnography is not only to follow people or things as they move, but also to understand the influences of different kinds of boundary crossings and ecologies on their experiences in multiple domains.

The practical difficulty of developing intensive local knowledge in multiple settings admittedly represents a serious challenge for ethnographic fieldwork. The Thomas and Znaniecki (1927) model of binational, multi-sited collaboration is a useful strategy for dealing with the problem of space, but it does not address the problem of time. The integration of participant-observation and interviews with local archival research and the ethnographic revisit are two ways to address historical change. Revisiting an earlier study creates a sort of bi-temporal “collaboration.” Detailed historical reconstructions across sites are useful ways of determining the changing ways that “globalization,” “transnationalism,” and other meta-processes articulate with migrants' experiences and understandings.