What logic does a social structure impose upon public policy? Could an identity structure be too complex or diffuse to implement categorical public policy such as affirmative action? Conversely, what impact does categorical public policy have upon an identity structure? Must categorical policy impose its own categories of identity on its public? This chapter examines the nature of the Brazilian social structure, which I characterize as a grammar of color identity, and its interaction with categorical public policy such as affirmative action in higher education. Opponents of affirmative action have claimed that Brazilian identity is too diffuse and subjective to be able to implement affirmative action: that it would be impossible to define or verify beneficiaries for the purposes of affirmative action. In their view, categorical policy cannot be implemented in a country without clearly delineated categories, and instead would racialize Brazil. The Black Movement rejoined that clear categories exist in many contexts, such as the official deployment of lethal violence, and that Brazil is therefore already racialized. This chapter explores the contours of Brazilian identity and the implications of a nuanced identity structure for affirmative action. I theorize that the Brazilian structure of identity is constituted through a highly colorized and nuanced "grammar" of identity that expresses the relationality and positionality of individuals in each context. Under this nuanced grammar, appearance, mediated and re-inscribed by social class and performance, has tremendous consequences for life outcomes. Social scientists have repeatedly found a significant gap between whites and others in life outcomes such as education, socioeconomic status, health indicators, and mortality studies. Thus, the color of poverty is widely recognized. The controversy is whether the color of poverty in Brazil is better remedied through race-based or class-based policies, or a combination. I argue for the combination of policies and defend affirmative action as a positive good that provides opportunities to those previously excluded, and also incentives that counterbalance Brazil's assimilationist tendencies. At the heart of the current debate over affirmative action is an idea that state preferences in higher education will overwhelm popular identity, similar to the hypothesized effect of casinos on indigenous identity in the United States. I argue that this constitutes a simplistic view of the relationship of incentives and actors, as well as the relationship of public and private power. I argue that in Brazil, the state "made race" in the way that states shape markets. No state official could have anticipated the plethora of categories that Brazilians, especially darker Brazilians, would use to describe themselves and each other. Nor did the state assign Brazilians those identities. Further, although I argue that the Brazilian state shaped identity, the state did not create the identities it specified in censuses (Nobles 2000). Why have the overwhelming majority of Afro-Brazilians historically identified as moreno rather than pardo or preto, the census categories? A full answer, beyond the scope of this chapter, requires an examination of state policy, including immigration policy, cultural and information policy under Vargas, educational curricula, as well as symbolic and other state action.1The sociologist Clovis Moura and others have suggested that the term moreno offered Brazilians a way to lighten themselves and to soften the significance of the distinctions (Moura 1988). Thus, the Brazilian "identity market" has been shaped by the whitening preference, as well as heralding the browning of Brazil. I claim that a fluid, interactive relationship developed historically between state action and societal identity. Scholars have identified multiple systems of deploying identity within Brazil.2Of those multiple systems, I distinguish two, for the purpose of simplicity: (1) state identity, generally expressed in the five terms used by the census bureau, and (2) societal identity, which draws upon a wide vocabulary of terms, including moreno and its many variants. It is the second system, societal identity, where many of the controversies about categories and state policy reside. Although Brazilians do not use societal identity to apply for affirmative action, many of the salient characteristics of Brazilian identity also apply to the use of state identity. By "grammar," I refer to ordered rules for the representation of identity, and additional rules about how to invoke those representations. Thus, I claim that Brazilian identity is not fixed but representational, and negotiated through the grammar, which expresses status positions and overall relations of power within a given context-all of which are dynamic elements. These identity terms possess a relational aspect, expressing relations of power between the identifier, the identified, and the relevant audience. It is not a "calculus" of identity3, but a grammar of color identity in the sense that grammatical rules inform the usage of terms. In my account, the Brazilian grammar of identity contains two key features: a ranking grammar, and a grammar of deracialization. In the first, the grammar of ranking, virtually all Brazilians learn by adolescence to make nuanced evaluations of each other's physical appearance, modifiable by considerations of social context. The primary but not sole physical evaluation is skin color. The primary consideration of social context is social class. Finally, this ranking is fully evident in social life-particularly within an informal but observable color hierarchy within the service sector, in which positions of greater customer contact are reserved for whites and lighter Brazilians. The second claim about the "grammar" of Brazilian color identity pertains to deracialization. Unlike most structures in which the dominant group deracializes itself and thereby makes itself synonymous with the nation and racializes others, in Brazil subordinated individuals also assert some racelessness. The most popular identity for the Brazilian "brown" is moreno, an identity that emphasizes racelessness and Brazilianness and deemphasizes a particular location within the color hierarchy. I argue that these two grammars interact to produce the Brazilian structure of identity. One might view the two grammars as counterbalancing: that the ranking grammar would either undermine, or be subsumed by, the grammar of deracialization. However, both grammars locate the moreno in second place and enable the dominant societal position of whites. The deracialization grammar provides a "face saving" for the moreno, while the ranking grammar reaffirms the societal position of whites. Thus, the two grammars increase the discursive flexibility of the Brazilian color hierarchy. Although I defend affirmative action, I concede a tension between affirmative action and the nuanced grammar of identity. This tension emanates in part from the Brazilian reliance upon the nuances of appearance in drawing distinctions. Although appearance is mutually constituted with descent, social class, performance, association, and other factors, physical appearance has been the primary aspect in Brazilian distinctions. There can be considerable variance between how some Brazilians view themselves and how they are viewed by others. Thus, a candidate for affirmative action may claim an identity that others may not perceive, creating a problem in the perceived legitimacy of affirmative action. Second, Brazilian identity has been historically animated by a steep whitening preference, in which darker Brazilians have been socialized to marry white and "improve" their child's race. This creates an additional tension in the gap between self-identity and other assignment. Finally, affirmative action arguably provides the first material incentive to claim a darker identity: either brown or black. How might Brazilian universities define eligibility for affirmative action given the grammar of identity? In this chapter, I explore this theory of the grammar of identity in three sections. The next section reviews the understanding of categories in Brazil in previous anthropological and survey work and suggests that Brazilian identity is best understood as a relational color hierarchy. The second and third sections analyze a recent national study, the 2002 PESB, to illustrate the grammar of color evident in contemporary identity. The last section explores how this grammar interacts with the implementation of affirmative action.