Symbols and investments as signals: Courtship behaviors in adolescent sexual relationships

Abstract

Both structural and cultural arguments suggest that extensive relationship-building behaviors are a precursor to the onset of sexual intercourse. This research develops and tests a game-theoretic signaling framework, which suggests that the association between courtship behaviors and the onset of sexual intercourse should reflect selection effects. To test this framework empirically, this research utilized the National Longitudinal Study of Adolescent Health (1994–1996) and estimated probit and recursive bivariate probit models of sexual intercourse. The results showed support for the signaling framework. Extensive use of courtship behaviors was associated with sexual intercourse, but this association reflected the effects of selectivity. However, extensive courtship was more associated with sexual intercourse when adolescents did not know one another well prior to the start of their romantic relationships. These findings suggest that selectivity associated with signaling theory can complement models of cultural influence and structural effects.

Keywords

adolescence investments sexuality signaling symbols

Introduction

In the social construction of sexual relationships, some adolescents engage in an array of ritualistic, relationship-building behaviors – they ‘hang out,’ go on ‘dates,’ give gifts, say ‘I love you’; others simply ‘hook up.’ Although the associated melodramas are the sine qua non of adolescent sexual culture, sociological research pays scant attention to these courtship behaviors – that is, relationship-building behaviors preceding sexual involvement.¹ A few studies have examined specific courtship behaviors (Carver et al., 2003; O’Sullivan et al., 2007; Vaquera and Kao, 2005), but their basic patterns, correlates, and consequences for romantic relationships during adolescence have received little attention (but see Bearman and Bruckner, 2001; Harding, 2007).² To the extent that adolescents are forming both romantic and nonromantic sexual relationships, the structure of courtship behavior is likely an important phenomenon, heretofore lacking systematic analysis.

Importantly, courtship processes highlight a theoretical issue as well. How relationships are constructed – how interpersonal attachment, trust, or commitment is fostered – is a basic sociological question. Sociological accounts of trust and commitment often focus on how social structures – relationship-specific investments (Becker, 1981; Rusbult et al., 1986), contracts (England and Farkas, 1986; Treas, 1993), exchange (Lawler and Yoon, 1993; Sprecher, 1998), and embeddedness (Buskens and Raub, 2002; Coleman, 1990; Felmlee, 2001) – and culture (Collins, 2004; Gagnon and Simon, 1973; Harding, 2007; Simon and Gagnon, 1986; Swidler, 1986, 2001) matter. Alternatively, courtship behaviors reflect not only cultural or structural effects, but act as costly signals (Camerer, 1988; Frank, 1988; Spence, 1973), leading individuals to self-select into patterns of courtship behavior based on their willingness to invest in or commit to their partners. In other words, a key theoretical issue centers on whether (admittedly prosaic) courtship behaviors are ‘symbols’ or ‘investments’ that promote trust and commitment or whether they act as signals, inducing selectivity. However, the latter approach is theoretically undeveloped in sociology, has not been tested empirically, and may provide new insights about adolescent courtship behavior.³

In this research, we address these theoretical and empirical gaps by investigating courtship behaviors in adolescent romantic relationships and testing whether associations with sexual involvement are consistent with a signaling framework as opposed to structural and cultural explanations. To demonstrate why signaling matters, we first adapt the well-known ‘trust game’ (Bacharach and Gambetta, 2001; Camerer and Wiegelt, 1988; Dasgupta, 1988; Raub, 2004; Raub and Weesie, 2000) to the case of adolescent romantic relationships and show that costly courtship can help individuals identify partners who are willing to invest in them, thereby leading individuals to invest in their relationships. We test this argument empirically by examining whether observed associations between courtship behavior and sexual intercourse are consistent with the effects of structure or culture, on the one hand, or selection effects, on the other. Taken together, our objective is to test whether a signaling perspective offers a competing account to causal stories of culture and structure.

Prior research

In the age of ‘hookups’ and ‘virginity pledges,’ relationship-building behaviors are likely to be quite diverse, but research on basic patterns is scant.⁴ Most of the prior research focuses not on what adolescents actually do to build their relationships, but on stages of courtship (e.g. Miller and Benson, 1999; Niehuis et al., 2006; Surra, 1990; Thornton, 1990) or specific relationship forms, such as ‘going-with’ relationships (Merten, 1996), romantic dating (Cavanagh, 2007; Connolly and Goldberg, 1999; Giordano et al., 2001, 2006; Larson et al., 1999), and hookups (Bogle, 2008; Paul et al., 2000). A few studies have examined the occurrence of specific relationship-building behaviors (Carver et al., 2003; Moore, 1995; O’Sullivan et al., 2007; Vaquera and Kao, 2005), while others have utilized counts or scales to indicate varying levels of commitment (Bearman and Bruckner, 2001; Furstenberg et al., 1987; King and Christensen, 1983; Manlove et al., 2007; Vaquera and Kao, 2005). Only one study examined heterogeneity in the sequencing of relationship-building behaviors (Harding, 2007), but importantly did not distinguish between behaviors occurring before or after the onset of sexual involvement. There has also been little research on whether courtship behaviors matter for relationship outcomes, such as the onset of sexual intercourse. Demographic studies of sexual debut, for example, frequently omit courtship behaviors as predictors (e.g. Lauritsen, 1994; Harris et al., 2002; Ramirez-Valles et al., 1998; but see Bearman and Bruckner, 2001). Thus, despite the potential importance of courtship behavior as a precursor of sexual intercourse, population-based research on its basic patterns, predictors, and consequences is scant.

Structural and cultural models of relationship formation

A key question centers on whether courtship behaviors are theoretically significant. Since enforceable contracts cannot be made in adolescent sexual relationships, structural perspectives on how investments, equity, and embeddedness promote trust and commitment in relationships are the most relevant. According to investment models, shared children, finances, property, and lifestyles reflect either relationship-specific capital (Becker, 1981; Brines and Joyner, 1999; England and Farkas, 1986) or joint investments (Rusbult and Buunk, 1993; Rusbult et al., 1986), which produce relationship-specific rewards, increase sunk costs, and enhance feelings of interdependence. Relatedly, scholars have also suggested that reciprocity in dyadic exchanges plays a similar role, promoting commitment and stability through equality in power (Brines and Joyner, 1999; Lawler and Yoon, 1993) and perceptions of equity (Sprecher, 1998; Walster et al., 1978). Finally, a number of studies have argued that social and dyadic embeddedness promotes sexual involvement (Bearman and Bruckner, 2001) and relationship stability (Felmlee, 2001; Felmlee et al., 1990; Sprecher and Felmlee, 1992) by enhancing trust and commitment (Buskens and Raub, 2002). Moreover, social networks may play an important role in partner selection, providing individuals with information about their partners’ reputations and sexual histories, bringing potential partners together, or exerting control over partner choices (Bearman et al., 2004; Laumann et al., 1994). If courtship behaviors reflect investments, capital, exchange processes, or embeddedness, then they should enhance trust and commitment, making the onset of sexual intercourse more likely.

An alternative approach is to conceptualize courtship behaviors as symbolic activities drawn from either cultural codes (Swidler, 1986, 2001) or sexual scripts (Simon and Gagnon, 1986). Courtship behaviors have been conceptualized as displays, acts, or rituals (Carver et al., 2003; Vaquera and Kao, 2005), symbolically marking and socially constructing relationship meanings (Davis, 1973; Zerubavel, 1981) and generating cohesion (Collins, 2004). Consequently, researchers invoke models of cultural influence to explain associations between these symbolic activities and sexual involvement (Cavanagh, 2007; Harding, 2007). In general, these cultural perspectives highlight the causal impact of culture for explaining associations between courtship behaviors and relationship outcomes.

There are, however, several theoretical limitations in applying structural and cultural models to courtship behaviors. First, it is not clear whether courtship behaviors are adequately conceptualized as investments, capital, exchange, or embeddedness. Arguably, courtship behavior precedes investments and the formation of relationship-specific capital. Indeed, courtship behaviors may be related to risk management and therefore are a precursor to the decision to make joint investments, build relationship-specific capital, establish negotiated exchange, or form embedded ties, as opposed to representing these concepts. Second, although the cultural aspects of strategic action have been a longstanding concern (Goffman, 1959; Swidler, 1986), their significance as a strategy for managing risks associated with relationship building, potentially with partners with divergent goals, has drawn little attention. That is, some adolescents, who are willing to invest in or commit to romantic relationships, may be interacting with potential partners primarily interested in ‘hooking up’ (Anderson, 1999; Bogle, 2008).⁵ Finally, the prior research has not accounted for the potential endogeneity of courtship behavior. Individuals may be motivated to choose patterns of courtship behavior based on their relationship expectations. In summary, these issues suggest the need for a theoretical perspective that can account for risk management, strategic interaction, and selectivity.

Conceptual framework

To account for courtship behaviors as strategic signals, we develop a game-theoretic framework, which combines a ‘trust game with incomplete information’ (Bacharach and Gambetta, 2001; Camerer and Weigelt, 1988; Dasgupta, 1988; Raub and Weesie, 2000) with uncertainty (Raub, 2004) and a prestage of courtship behavior (Camerer, 1988). Figure 1 presents our ‘dating game,’ which is a modified version of Raub’s (2004, page 339) extensive-form trust game with uncertainty and incomplete information. The basic dating game consists of two players, identified by nodes 1 and 2, in a dynamic game consisting of two sequential decisions during a fixed period. The former (player 1) is a romantically inclined adolescent faced with a decision about whether to invest (I₁) in forming a dating relationship with a potential sexual partner or not invest (N₁) thereby remaining single. If the romantic adolescent invests in his or her dating relationship, then the latter adolescent (player 2) also decides whether to invest (I₂) or not (N₂). Investments involve putting at stake resources, affection, status, and/or sexual access. If player 1 remains single (N₁), then both players receive the payoff of 0, which is the benchmarked value of no dating relationship with one another. If player 1 invests (I₁), player 2 may ‘cheat’ (N₂), terminating the relationship after obtaining some benefit of the former’s investments. In this case, player 2 has taken advantage of the resources, status, feelings, or sex that player 1 has committed, thereby obtaining the ‘temptation’ payoff (T₂), while the latter receives the ‘sucker’s’ payoff (S₁).

Figure 1.

Adolescent dating game.

Following Raub (2004), we assume that player 1 does not receive R₁ with certainty when both players invest (I₁, I₂). In other words, adolescents have limited knowledge about making romantic relationships successful and are inexperienced; their dating experiences are replete with unmet expectations and regret. Raub’s (2004) approach attempted to parameterize the unintentional sale of a lemon by a car dealer – that is, a player received the sucker’s payoff instead of R with a specified probability. In our approach, the expected payoffs for the formation of a dating relationship with uncertainty are as follows:

R_{1} * = q * 0 + (1 - q) * R_{1}

(1)

The probability q represents the likelihood of contingencies making dating relationships no more rewarding than remaining single (0). For simplicity, we assume that these contingencies affect player 1 only, but we note that the model could be made more complex by allowing the probability of contingencies to vary by partner. Thus, if player 1 invests (I₁) and then player 2 also invests (I₂), then a romantic dating relationship is formed with payoffs R₁* and R₂, respectively.

Finally, we assume that player 2 varies in his or her willingness to form a romantic relationship. Figure 1 incorporates an initial move by Nature (the first 0 node) regarding player 2’s type; the dotted line shows the information set. For simplicity there are two basic ideal types. Romantic adolescents, such as player 1, prefer to cooperate in the hopes of obtaining relationship-specific benefits (e.g. learning to be a good ‘boyfriend’ or ‘girlfriend,’ companionship, intimacy, sexual pleasure, and gaining status). An adolescent who is unwilling to stay in a romantic relationship prefers to take advantage of player 1’s cooperation and defect for several reasons, including sexual gratification, learning to be a ‘player,’ and enhancing one’s reputation through sexual conquests. Players have ‘incomplete information’ (Harsanyi, 1967) – that is, they only know their type, not their partners, and the proportion of willing types (p) in the population.⁶ This suggests the following payoff structures:

Willing type (W) : R > T > 0 > S

(2)

Unwilling type (U) : T > R > 0

(3)

Willing adolescents prefer ‘romance’ (R) to all other payoffs, including the temptation (T) of taking advantage of another romantic, such as player 1. However, player 1 prefers to remain single (0) compared to being taken advantage of (S₁). In other words, some adolescents highly value romantic relationships and cannot see themselves taking advantage of, or ‘playing,’ another romantic. Unwilling adolescents prefer the temptation payoff (T₂) to romance (R₂), while remaining single is the least preferred outcome (0). These adolescents are ‘players’: they prefer short-term relationships to obtain sex, status, resources, or affection, but seek to avoid romantic entanglements.

The above-described framework presents an asymmetric version of Camerer’s (1988) symmetric investment game with incomplete information; the game is also similar to Raub’s (2004) trust game with uncertainty and incomplete information with two exceptions. The key differences are that our version allows willing-type player 2s to play defect (N₂), but they still prefer to invest (I₂) because R > T and that dating relationships do not necessarily have better payoffs over being single for player 1.⁷ That is, we argue that some adolescents, due to prior socialization, evolutionary psychology, or the characteristics of the match, will be primarily interested in long-term relationships over a short-term encounter. The game is noncooperative, so enforceable contracts cannot be made.

Importantly, the asymmetry in Figure 1 highlights the problem of trust in sexual relationships between male and female adolescents, an issue highlighted by Coleman (1990). However, unlike some of the prior research (Bolle, 2001; Sozou and Seymour, 2005), we do not simply assume that girls and boys have gender-specific payoff structures. Rather, we argue that both male and female adolescents are likely to interact with some sexual partners who are more interested in obtaining investments (e.g. getting sexual access, resources, or status benefits) and then exiting the relationship.

So under what conditions will player 1 invest in forming a romantic relationship? To identify this condition, we calculate expected values for player 1, which are as follows:

E V_{1} (I) = p (R_{1} *) + (1 - p) (S_{1}),

(4)

E V_{1} (N) = p (0) + (1 - p) (0)

(5)

Therefore, player 1 will invest if and only if EV₁(I) > EV₁(N), or

p > - S_{1} / (R_{1} * - S_{1}) = t_{1}

(6)

This equation reflects the expected gains of romance minus the expected loss of being cheated on against the opportunity costs of remaining single. If the probability of interacting with a willing player exceeds an ‘investment threshold,’ (p > t₁), then a Bayesian equilibrium is that player 1 invests (I₁). By definition, t₁ ≤ 1, since S₁ < 0 and R₁* is equal to either R₁ or 0. If, however, −S₁ is relatively large in comparison to R ₁ *, then t₁ will approach 1. In this case, player 1 will have a high threshold for investment; that is, there must be a high probability of playing an investment game with another W-type adolescent. If p < t₁, then the equilibrium outcome is that player 1 will choose to remain single (N₁). Consistent with Camerer’s (1988) original finding, signaling is potentially useful when p < t. Willing adolescents should be motivated to provide a signal of their willingness to invest. In short, when adolescents are faced with schools populated by ‘players’ and ‘romantics’ and when the stakes are high, signaling may be crucial.

Figure 2 adds a prior stage where signals (C⁺) can be observed prior to the trust game presented in Figure 1. Importantly, the use of signals imposes signaling costs for both players (C₁, C₂). For simplicity, these costs are assumed to reflect the net costs of signaling after adjusting for increases in payoffs that result from gift giving, but we note that Camerer (1988) has highlighted the importance of inefficient gifts as signals. For example, delaying sexual involvement increases the time, money, energy, and resources spent building relationships among courting couples, but going out for dinner at a nice restaurant on a date likely results in some positive benefit. Our approach is distinct from trust games with hostage posting, which limit the costs of hostage posting to one player and focus on the loss of hostages only in the case of abused trust (Raub, 2004; Raub and Weesie, 2000; Snijders and Buskens, 2001). Trust games with hostage posting examine the conditions under which hostages can bind untrustworthy agents to honoring trust, act as forms of compensation, and reflect credible signals; in contrast, we focus exclusively on signaling.

Figure 2.

Dating game with courtship.

If the costs of signals for players 1 and 2 (C₁ , C₂) satisfy the following conditions, there will be a separating equilibrium:

R_{1} * > C_{1}; R_{2}^{W} > C_{2} > T_{2}^{U}

(7)

That is, player 1 will be willing to form a dating relationship even after courtship costs because the payoff still exceeds remaining single. For unwilling player 2s, courtship costs exceed the expected temptation payoff, so these adolescents will refrain from playing C⁺, but the expected payoff for willing player 2s would still be positive. Consequently, player 1 can identify player 2’s type based on his or her willingness to incur courtship costs. A pooling equilibrium will result if courtship is not sufficiently costly to deter unwilling adolescents from engaging in these behaviors. To summarize, our game-theoretic model shows that adolescents likely face a trust problem because an equilibrium outcome is to remain single, but if courtship is sufficiently costly, it can act as a signal distinguishing ‘romantics’ from ‘players.’ Equilibrium outcomes include the following: (1) willing adolescents will be likely to incur costly courtship, (2) costly courtship will encourage adolescents to take risks in their relationships, and (3) the link between courtship and making investments will reflect self selection.

Hypotheses

Our game-theoretic framework suggests several testable propositions about adolescent romantic relationships. First, our model suggests an association between adolescents’ willingness to form long-term sexual relationships and their courtship behavior. Specifically, according to the equations in (7), unwilling adolescents will be less likely to engage in courtship behaviors if their courtship costs (C > T) are too high. This argument is certainly intuitive, and it also dovetails with both structural and cultural models of influence. For example, Harding (2007) and Cavanagh (2007) have suggested a basic linkage between preferred scripting of romantic relationships and actual behavior. In this research, we focus primarily on economic and social courtship behaviors and two proxies for unwillingness to invest in romantic relationships: the preferred timing of sexual intercourse in an idealized dating relationship and having had prior sex partners.⁸ Our expectation is as follows:

Hypothesis 1. Preferences for early sexual involvement and experience with prior sex partners will be associated with lower likelihoods of extensive courtship behaviors.

Second, the signaling argument suggests a link between courtship behavior and the likelihood of putting investments at risk. Both structural and cultural models suggest a causal linkage between courtship behaviors and sexual involvement. That is, these behaviors represent a cycle of increasing investment or commitment, which eventuates in sexual involvement. In contrast, the signaling perspective suggests that this hypothesized association reflects not causation, but selection. This suggests the following hypotheses:

Hypothesis 2a (structural-effects hypothesis). Extensive use of courtship behaviors will be positively associated with an increased likelihood of the onset of sexual intercourse.

Hypothesis 2b (selection hypothesis). The hypothesized association between extensive courtship behaviors and sexual intercourse will be significantly accounted for by selectivity into extensive forms of courtship.

To summarize, this research extends the extant literature in several ways. First, we extend and develop Coleman’s (1990) idea of conceptualizing adolescent romantic relationships as a problem of trust. In this paper, we adapt the well-known trust game to the case of adolescent dating relationships. Second, unlike prior game-theoretic models of intimate relationships, we plan to empirically test hypotheses derived from our formal model using population-based data. Third, we seek to provide the first empirical assessment of patterns of courtship activities preceding sexual involvement as well as their associations with key correlates and sexual intercourse. Finally, this courtship-signaling framework represents a competing perspective to models of structural and cultural influence. It directs empirical research to the study of both causation and selection effects associated with symbolic behaviors or investments. Thus, our approach complements causal stories of structure and culture with a signaling account of selectivity.

Methods

Data were drawn from Waves I and II of the National Longitudinal Study of Adolescent Health (Add Health), a nationally representative, multi-stage cluster sample of adolescents who were originally surveyed in grades 7 through 12 during the 1994–1995 academic year (Bearman et al., 1997). The Wave-I, in-school questionnaire (n = 90,118) was administered during a regular, 60-minute class session between September 1994 and April 1995. Respondents for the Wave-I, in-home survey (n = 20,745) were drawn from these schools and administered the survey in 1995. The Wave-II, in-home survey (n = 14,738) was administered in 1996 to adolescents who were not high school seniors in Wave I. These survey instruments collected extensive information about the social contexts of adolescents, including their families, schools, neighborhoods, and friendships, as well as personal characteristics, such as delinquency, risk behaviors, mental and physical health, intellectual ability, self-esteem, personal feelings and attitudes, educational aspirations, substance abuse, sexual practices, and knowledge about contraceptives. The response rates for the in-home components of Waves I and II were 80 and 85 percent, respectively.

To test hypotheses about courtship behaviors, this research focused on data on relationship-building behaviors in romantic relationships, which were collected in Wave II.⁹ Respondents were asked to identify up to three ‘special romantic’ relationships and up to three additional sexual partnerships in the 18 months preceding Wave II. Information on the occurrence and sequencing of relationship behaviors was collected for all romantic relationships. If respondents reported no romantic partners, they were asked whether they had any ‘like’ relationships. For the like and sexual partnerships, information on relationship behaviors was gathered if there were fewer than three romantic partners and if the respondent had held hands, kissed, and expressed liking or loving their partners.¹⁰ Data on relationship behaviors, however, are limited in two ways: (1) the Add Health study did not collect this information for like (e.g. failed courtships) and nonromantic relationships that did not involve these three behaviors, and (2) these measures did not assess the frequency of these behaviors.

The analytic sample was selected using several criteria. First, we focused only on respondents who completed the in-home questionnaires from Waves I and II and had valid sampling weights (n = 13,570). Second, respondents were restricted to adolescents who were still attending school and living with their parents. Married respondents, high-school dropouts, and respondents living on their own were not asked several key questions related to included covariates. Third, we restricted our analyses only to partnerships starting after the Wave I interview.¹¹ We adopted this procedure to take advantage of the longitudinal design of Add Health, to minimize recall bias, and to capture relationship behaviors occurring between Waves I and II. Finally, if adolescents had more than one partnership starting after Wave I, we randomly selected one partnership, as opposed to selecting the first listed, which is biased towards longer relationships (Carver et al., 2003), and adjusted sampling weights for unequal probabilities of selection. The first two selection criteria reduced our sample to 12,335 respondents. Missing data on variables reduced the sample by an additional 6.5 percent, resulting in 11,534 respondents, of which 4,938 had a romantic sexual partnership start after Wave I. Table 1 provides descriptive statistics for the variables used in the models. With the exception of partner-specific covariates, all independent measures were based on Wave I data, while all dependent measures were drawn from Wave II.

Table 1.

Descriptive statistics for variables (n = 4,938 respondents).

Variables	Mean / Prop	S.D.
Courtship behaviors
Go out in groups	0.68
Go out together alone	0.62
Meet partner’s parents	0.58
Disclose couplehood	0.78
See less of other friends	0.37
Exchange presents	0.46
Tell one another ‘I love you’	0.51
Sexual intercourse	0.35
Preference for later sex	0.31
Preference for early sex	0.10
One to two prior sex partners	0.12
Three or more prior sex partners	0.06
Controls
Male	0.46
Black	0.19
Hispanic	0.16
Asian	0.05
Age	15.87
Parents’ schooling	13.34	2.38
Non-intact family	0.44
Depression	1.52	0.40
Pubertal development	0.05	0.77
GPA	2.77	0.76
Delinquency	0.03	0.54
Parental monitoring	0.30
Parent-child closeness	4.22	0.55
Parental permissiveness	1.80	0.87
Religiosity	0.02	0.87
Virginity pledge	0.14
One prior romantic partner	0.45
Two or more prior romantic partners	0.27
Romantic dating interest	3.59	1.07
Partner of different age	1.04	2.13
Partner of different race/ethnicity	0.20
Lack of prior ties	1.01	0.97
Racial segregation index (n = 3,638)	0.26	0.19

Data: National Longitudinal Study of Adolescent Health, Waves I–II, 1994–1996.

Dependent measures

Courtship behaviors preceding sexual involvement

For romantic relationships, respondents were asked whether they did any of the following with their partners: (1) went out together in a group, (2) went out together alone, (3) met their partners’ parents, (4) told other people that they were a couple, (5) thought of themselves as a couple, (6) held hands, (7) saw less of my friends, (8) gave each other presents, (9) told each other that you loved each other, (10) kissed, (11) touched each other under their clothing or with no clothes on, (12) engaged in genital touching, (13) had sexual intercourse, (14) got married, and (15) got pregnant. For all activities that occurred, respondents were then asked about their sequencing. We combined this information to create dichotomous indicators reflecting the occurrence of specific relationship behaviors before sexual involvement (i.e., sexual petting, genital touching, or sexual intercourse). Because items 4 and 5 were both similar and highly correlated, we combined them. Importantly, since some romantic relationships – those originally reported as like or sexual partnerships – were defined on the basis of whether adolescents had ‘held hands’ and ‘kissed,’ we excluded these measures. We also excluded ‘got married’ and ‘got pregnant,’ both of which do not precede premarital sexual involvement.

Sexual intercourse

For romantic relationships, this variable assessed whether adolescents experienced the onset of sexual intercourse in their selected romantic relationships between Waves I and II.

Independent measures

Sexual attitudes and experiences

We tapped adolescents’ preferences and attitudes about sex using two measures: preferred timing of sexual intercourse in romantic dating relationships and the number of prior sex partners. Adolescents were asked about the preferred sequencing of relationship behaviors in their ideal romantic relationships. We coded the relative ordering of sexual intercourse to indicate whether an adolescent preferred that sex occurred early (i.e., the first-through-eleventh relationship behavior), later (i.e., the twelfth-through-fifteenth relationship behavior), or never. The number of prior sex partners was the count of sex partners from January 1, 1994 to Wave I.

Control variables

Appendix A offers a detailed description of items used in the summated scales described below. We introduced control variables for demographic and personal characteristics. These included binary indicators for male, African-American, Hispanic, Asian American, and family structure (i.e., two biological parents versus non-intact); age was a continuous variable. We did not include a measure of educational attainment, since this variable was highly correlated with age (r = 0.92). Parents’ schooling was the average educational attainment of adolescents’ mothers and fathers based on six-point scales (α = 0.88). This measure included both residential and nonresidential parents and used all available information. School performance was assessed by calculating the grade point average, on a 4.0 scale, of each respondent for up to the four following classes: Math, Science, English, and History. We also included a depression scale of 17 items (α = 0.87). Pubertal development assessed physiological changes for male (α = 0.67) and female adolescents (α = 0.75). Delinquency was tapped using a summated scale (α = 0.87) of 21 indicators on four-point scales (0 = never, 3 = performing the act 5 or more times). Items ranged from risk-taking activities, such as racing a bike or car, to more severe activities, including theft, drug use, and physical violence.

Three variables assessed aspects of adolescent–parent relationships: parental monitoring, closeness to parents, and perceived parental permissiveness of adolescents’ sexual activities. A dichotomous indicator for whether parents enforced curfews tapped parental monitoring. Parental closeness (α = 0.72) was a summated scale based on up to four questions on five-point Likert scales. Perceived parental permissiveness (α = 0.91) was a summated scale of four items tapping adolescents’ perceptions of how their parents would feel about their sexual activity, hypothetical or not. Religiosity (α = 0.87) was a summated scale of frequency of religious service attendance, the importance of religion, and frequency of prayer. A dichotomous indicator for whether the respondent had taken a pledge to remain a virgin until marriage was also included. Adolescents were asked about the importance of romantic dating on a five-point Likert scale. The number of prior romantic partners was the count of up to three romantic partners reported in the 18 months before Wave I.

We also controlled for several additional characteristics of romantic relationships. Age difference captured the relative age of partners to respondents. Race difference was a dichotomous variable indicating if respondents’ partners were of a different race or ethnicity. Finally, we included a measure for the lack of prior social ties between partners, which was assessed on a four-point scale and distinguished whether respondents and their partners were friends, friends of friends, acquaintances, or strangers when they first became romantically involved.

Instruments

Finally, we used a school-level index of racial segregation in friendship networks, based on Freeman’s (1978) segregation index for social networks, as an instrumental variable. This index assessed the difference between expected and observed interracial ties as a proportion of expected interracial ties.

Models

To test the hypothesis that adolescents who are unwilling to make relational investments will be less likely to spend time engaging in courtship behavior (Hypothesis 1), we estimated latent mixture models of relationship-building behaviors in romantic relationships on adolescents’ sexual attitudes and experiences as well as control variables (n = 4,938). Based on the cross-classification of seven relationship behaviors occurring prior to sexual involvement (go out in groups, go out together alone, meet partner’s parents, disclose couplehood, see less of friends, exchange presents, tell each other ‘I love you’), which had 128 possible combinations, the latent mixture models generated maximum-likelihood parameter estimates for identified latent classes of courtship behavior and covariates. Goodness of fit for these models was based on BIC (Muthén, 2004; Nylund et al., 2007).

Next, we examined the extent to which engaging in extensive courtship behaviors was associated with the onset of sexual intercourse and whether the observed association was consistent with causation or selection. First, sexual partnerships were assigned to specific courtship classes based on the modal values of their assignment probabilities. The proportion correctly classified assessed the accuracy of this procedure. We then estimated probit regression models of sexual intercourse on engaging in extensive courtship behavior. To distinguish between causal and selection effects, we estimated recursive bivariate probits, which included the degree of racial segregation in friendships as an instrument. This variable is not associated with sexual intercourse in relationships, and we are not aware of any theoretical arguments linking the two. In contrast, racial segregation in friendships should be associated with whether couples engage in courtship displays. For example, prior research found that adolescents in interracial relationships were less likely to engage in public and private displays, but they were just as likely to have sexual intercourse (Vaquera and Kao, 2005). Because this variable was based on responses from the in-school survey from Wave I, the analytical sample was further reduced to 3,638 individuals in the models of sexual intercourse. All models included weights and accounted for cluster sampling.

Results

According to Table 1, courtship behaviors were common in adolescents’ romantic relationships. Most courtship behaviors occurred in the majority of romantic relationships, but social behaviors, such as going out in groups (68%), going out together alone (62%), and disclosing couplehood (78%) were the most prevalent. In contrast, dyadic behaviors – saying ‘I love you’ (51%), gift giving (46%), and seeing less of friends (37%) – were less prevalent. Approximately one-third of adolescents reported sexual intercourse in these romantic relationships.

Table 2 presents identified classes of courtship behavior that were rendered by the latent mixture model. Based on the BIC, the best-fitting model is the four-class model (BIC = 38,237.98).¹² The latent-class and conditional probabilities indicate the prevalence of each courtship class in the population and the probability of class members engaging in specific courtship behaviors, respectively.

Table 2.

Latent class and conditional probabilities for the four-class, latent-mixture model of courtship behaviors in romantic relationships (n = 4,938).

Courtship behaviors	Minimal		Going-with		Social		Extensive
	Cond. Pr.	S.E.	Cond. Pr.	S.E.	Cond. Pr.	S.E.	Cond. Pr.	S.E.
Go out in groups	0.23	0.03	0.61	0.04	0.77	0.05	0.89	0.01
Go out together alone	0.08	0.03	0.27	0.05	0.80	0.06	0.92	0.02
Meet partner’s parents	0.15	0.03	0.43	0.04	0.51	0.05	0.89	0.02
Disclose couplehood	0.11	0.03	0.96	0.02	0.73	0.05	1.00	0.00
See less of other friends	0.02	0.01	0.18	0.03	0.33	0.05	0.66	0.02
Exchange presents	0.04	0.02	0.29	0.04	0.22	0.03	0.79	0.02
Tell each other ‘I love you’	0.04	0.02	0.64	0.05	0.11	0.03	0.84	0.03
Latent class probabilities	0.17		0.23		0.22		0.38

Data: National Longitudinal Study of Adolescent Health, Waves II, 1995–1996.

We labeled the four courtship classes as follows: ‘extensive,’ ‘social,’ ‘going-with,’ and ‘minimal.’ Accounting for 38 percent of romantic relationships, the extensive class is the most prevalent. For this pattern, the conditional probabilities (‘cp’) for each courtship behavior are high, ranging from 0.66 to 1.0. Adolescents in this class reported many courtship behaviors before sexual involvement – that is, they engaged in social activities alone and with friends and family, disclosed their couplehood, and expressed love to one another prior to any sexual activity. The social class, at 22 percent of partnerships, is more peer oriented. It has high conditional probabilities for going out in groups (cp = 0.77), together alone (cp = 0.80), and disclosing couplehood (cp = 0.73), but low conditional probabilities for seeing less of friends (cp = 0.33), exchanging presents (cp = 0.22), and expressing love (cp = 0.11) suggest little dyadic cohesion prior to sexual intercourse. Representing roughly one-in-four romantic relationships, the going-with class has high proportions of adolescents going out in groups (cp = 0.61), disclosing couplehood (cp = 0.96) and expressing love (cp = 0.64), but these relationships involve relatively few of the other social and dyadic behaviors. The ‘going-with’ label suggests adolescent couples that are relatively inexperienced at dating, but are trying out the label of being coupled to another person (Merten, 1996). Finally, the minimal form is markedly different. At 17 percent of romantic relationships, courtship behaviors before sexual involvement are rare, ranging from .02 to .23. These relationships may reflect ‘like’ relationships that did not develop; they may also reflect adolescents who are ‘hooking up’ as a pathway to developing a romantic relationship. Taken as a whole, the courtship classes are both qualitatively and quantitatively distinct. On the one hand, the identified classes are consistent with culturally distinct forms or sexual scripts – traditional romantic dating (extensive), going-with, and hookups (minimal) – identified in the extant literature. On the other, the conditional probabilities indicate that these forms are also quantitatively different from one another – that is, there is a cost structure related to these symbolic patterns – which is consistent with our argument about costly signaling through courtship behaviors.

Are proxies for an adolescent’s unwillingness to invest in romantic relationships associated with lower likelihoods of engaging in extensive courtship? Table 3 presents parameter estimates of the four-class, latent-mixture model, which were simultaneously estimated with the latent class and conditional probabilities reported in Table 2. The reference category is the minimal class. Consistent with Hypothesis 1, the results suggest that proxies for unwillingness to make relationship investments are associated with decreased odds of extensive courtship. For example, adolescents who preferred sex early in their relationships as opposed to abstaining are roughly half as likely to engage in extensive courtship (e ^−0.82 = 0.44, p < 0.01). Similarly, adolescents who had three or more prior sex partners since 1994 are half as likely to adopt extensive courtship (e ^−0.71 = 0.49, p < 0.05) as opposed to minimal courtship, in comparison to adolescents with no sex partners. Importantly, preferences for early sex are also associated with decreased odds of being in the going-with and social classes, but having more prior sex partners is not consistently related to lower odds of engaging in either of these courtship classes in comparison to the nominal class. Taken together, these results support the admittedly unsurprising notion that individuals who are oriented more towards sex than relationships are less willing to spend time engaging in extensive courtship. There is some more limited evidence that unwilling adolescents are also less likely to engage in going-with and social forms of courtship as well.

Table 3.

Latent mixture model of courtship classes on sexual attitudes and experiences and control variables (n = 4,938).

Variables	Going-with		Social		Extensive
	Log-odds	S.E.	Log-odds	S.E.	Log-odds	S.E.
Preference for later sex	−0.28	0.31	−0.37	0.26	−0.36	0.20*
Preference for early sex	−0.79	0.46*	−0.73	0.34*	−0.82	0.27**
One to two prior sex partners	0.27	0.38	−0.70	0.32*	−0.20	0.21
Three or more prior sex partners	−0.23	0.45	−0.61	0.42	−0.71	0.32*
Male	−1.09	0.36***	−0.34	0.29	−0.08	0.18
Black	0.14	0.46	−0.63	0.43	−0.62	0.23**
Hispanic	−0.36	0.32	−0.22	0.32	−0.45	0.22*
Asian	1.24	0.48**	0.02	0.86	0.02	0.36
Age	−0.81	0.13***	0.33	0.09***	0.04	0.06
Parents’ schooling	−0.06	0.07	0.17	0.06**	−0.03	0.04
Non-intact family	−0.37	0.21*	−0.51	0.21**	−0.23	0.17
Depression	−0.37	0.17*	−0.15	0.17	−0.13	0.12
Pubertal development	−0.13	0.25	−0.83	0.30**	−0.41	0.17**
GPA	0.16	0.14	0.12	0.17	0.16	0.11
Delinquency	−0.34	0.22	−0.03	0.20	−0.25	0.14*
Parental monitoring	−0.27	0.25	0.31	0.19	−0.02	0.17
Parent–child closeness	−0.13	0.22	−0.16	0.24	0.04	0.13
Parental permissiveness	−0.07	0.14	0.25	0.14*	0.00	0.11
Religiosity	0.17	0.14	0.04	0.11	0.09	0.08
Virginity pledge	−0.12	0.34	−0.21	0.27	0.07	0.21
One prior romantic partner	0.11	0.28	0.20	0.32	0.13	0.19
Two or more prior romantic partners	−0.23	0.29	0.14	0.30	−0.22	0.23
Romantic dating interest	−0.05	0.10	0.04	0.10	0.08	0.08
Partner of different age	−0.39	0.12***	0.13	0.07*	0.01	0.05
Partner of different race/ethnicity	0.16	0.25	−0.04	0.24	−0.13	0.17
Prior ties	−0.26	0.24	−0.42	0.20*	−0.50	0.15***
Intercept	16.85	2.46***	−4.76	2.61*	2.05	1.52
Log likelihood (df = 96)	−18655.49

Data: National Longitudinal Study of Adolescent Health, Waves I–II, 1994–1996.

Minimal is the reference category. *p < 0.05, **p < 0.01, ***p < 0.001 (one-tailed test).

Several control variables are also associated with courtship behavior in adolescent romantic relationships. For the going-with class, the coefficients for male, age, non-intact families, depression, and age difference are all statistically significant and negative, suggesting lower odds of adoption in comparison to the minimal class, while the coefficient for Asian Americans is positive. Being raised in a non-intact family, pubertal development, and having weak or no prior ties are all associated with lower odds of adopting the social class in comparison to the minimal class, while age, parents’ schooling, parental permissiveness, and older partners increased the odds. Finally, being African American or Hispanic, delinquency, pubertal development, and weak or no prior ties significantly reduced the odds of engaging in extensive courtship as opposed to the minimal courtship.

Courtship classes and sexual intercourse

Finally, to what extent does the association between extensive courtship and sexual intercourse reflect casual or selection effects? To answer this question, we first assigned partnerships to specific classes based on their modal probabilities. The average probabilities for these assignments ranged from a low of 0.79 for the social class to 0.90 for the minimal class, suggesting that the accuracy of these deterministic assignments was quite high. In order to test the hypothesis that extensive courtship will be linked to a separating equilibrium, we collapsed the courtship classes to distinguish extensive versus nonextensive courtship (i.e., the social, dyadic, and minimal forms).

Model 1 of Table 4 presents a probit model of sexual intercourse in adolescent romantic relationships on extensive courtship and controls. The results show that extensive courtship is associated with an increased likelihood of sexual intercourse (b = 0.44, p < 0.001) in comparison to nonextensive courtship. According to the estimated marginal effects, engaging in extensive courtship is associated with a 15 percent increase in the probability of having sexual intercourse, holding other variables constant at their means. In addition, a number of controls are significant. Preferences for sex in relationships and prior sexual experiences both increase the likelihood of sexual intercourse. Likewise, numerous characteristics – being female, African American, older, having parents with less schooling, being raised in a nonintact family, pubertal development, lower grades, delinquency, parental permissiveness, the lack of parent–child closeness, and having an older or different race partner – are associated with increased likelihoods of sexual intercourse in romantic relationships.

Table 4.

Recursive bivariate probit model of extensive courtship and sexual intercourse (n = 3,638).

Variables	M1: Probit of sexual intercourse			M2: Recursive bivariate probit
	b	S.E.	Marginal effect	Extensive courtship		Sexual intercourse
				b	S.E.	b	S.E.
Preference for later sex	0.34	0.08	0.12***	−0.12	0.09	0.24	0.10**
Preference for early sex	0.36	0.13	0.13**	−0.35	0.15**	0.17	0.15
One to two prior sex partners	0.39	0.11	0.14***	−0.03	0.10	0.32	0.11**
Three or more prior sex partners	0.78	0.18	0.30***	−0.20	0.14	0.60	0.18***
Male	−0.26	0.08	−0.09***	0.27	0.08***	−0.11	0.10
Black	0.35	0.08	0.13***	−0.33	0.11**	0.18	0.10*
Hispanic	−0.04	0.14	−0.01	−0.31	0.12**	−0.15	0.15
Asian	−0.17	0.18	−0.06	−0.32	0.17*	−0.26	0.16
Age	0.17	0.03	0.06***	0.12	0.02***	0.19	0.03***
Parents’ schooling	−0.05	0.02	−0.02**	−0.07	0.02***	−0.07	0.02***
Non-intact family	0.30	0.07	0.10***	0.00	0.07	0.25	0.07***
Depression	−0.16	0.10	−0.06	−0.04	0.08	−0.16	0.09*
Pubertal development	0.22	0.05	0.07***	0.03	0.04	0.19	0.05***
GPA	−0.18	0.05	−0.06***	0.04	0.05	−0.13	0.05**
Delinquency	0.30	0.08	0.10***	−0.01	0.08	0.25	0.08**
Parental monitoring	−0.03	0.08	−0.01	−0.09	0.07	−0.07	0.07
Parent–child closeness	−0.20	0.08	−0.07**	0.10	0.07	−0.13	0.07*
Parental permissiveness	0.14	0.04	0.05**	−0.06	0.04	0.10	0.05*
Religiosity	0.00	0.04	0.00	−0.06	0.04	−0.01	0.04
Virginity pledge	−0.11	0.13	−0.04	0.18	0.10*	−0.03	0.12
One prior romantic partner	0.08	0.09	0.03	0.06	0.07	0.09	0.08
Two or more prior romantic partners	−0.03	0.10	−0.01	0.02	0.09	−0.02	0.10
Romantic dating interest	0.02	0.03	0.01	0.07	0.03*	0.04	0.04
Partner of different age	0.13	0.02	0.05***	0.03	0.02	0.12	0.02***
Partner of different race/ethnicity	0.19	0.10	0.07*	−0.08	0.08	0.13	0.10
Lack of prior ties	0.03	0.04	0.01	−0.09	0.03**	−0.01	0.04
Extensive courtship	0.44	0.07	0.15***			−0.68	0.32*
Racial segregation index				0.32	0.15*
Constant	−1.97	0.66	**	−1.76	0.58**	−1.84	0.60**
ρ						0.66	0.18**
Wald χ²	684.60***			1591.38***
df	27			54

Data: National Longitudinal Study of Adolescent Health, Waves I–II, 1994–1996.

p < 0.05, **p < 0.01, ***p < 0.001 (one-tailed test).

Model 2 presents the results from the recursive bivariate probit. This model corrects for the potential endogeneity of extensive courtship behavior – that is, the effect of selection, parameterized by the rho coefficient, into extensive and nonextensive courtship. The first probit equation reflects estimated coefficients from a probit equation of extensive courtship. These results are similar to those presented in Table 3. In this model, preferences for early sex, being male, white, older, having parents with less schooling, taking a virginity pledge, romantic dating interest, and having prior social ties with their romantic partners are associated with the likelihood of extensive courtship. Importantly, school-level racial segregation in friendships is associated with an increased likelihood of extensive courtship.¹³ In other words, romantic relationships taking place in schools with extensive racial segregation in friendships are more likely to engage in public and dyadic displays of courtship.

The results from the second probit equation in Model 2 are largely similar to Model 1, though several controls – preference for early sex, male, and interracial relationships – are no longer statistically significant, but depression is now significant. The rho parameter is significant and positive, and the effect of extensive courtship is now negative. The results indicate that extensive courtship does not increase the likelihood of sexual intercourse; rather, the observed association between the two in Model 1 reflects selection effects. The rho parameter indicates that individuals who select into engaging in extensive courtship are more likely to have sexual intercourse. Once we control for positive selection into extensive courtship, nonextensive courtship is more associated with sexual intercourse. This finding is consistent with the signaling argument (Hypothesis 2b), not the structural-effects hypothesis (Hypothesis 2a).

A plausible hypothesis, derived from an embeddedness perspective, is that courtship behavior may matter for the onset of sexual intercourse for those respondents who did not know their partners well at the beginning of their relationships.¹⁴ We re-estimated the recursive bivariate probit (Model 2 from Table 4), but added an interaction between the lack of prior social ties between partners and extensive courtship. These results are presented in Table 5. Indeed, there is evidence that extensive courtship is more associated with the onset of sexual intercourse when adolescents do not know their partners well; whereas, the lack of social ties is less associated with sexual intercourse in the absence of extensive courtship. That is, extensive courtship does appear to be an important precursor leading to sexual intercourse in adolescent romantic dyads that are not embedded in social networks.

Table 5.

Sexual intercourse equation from a recursive bivariate probit. The effects of prior ties and extensive courtship (n = 3,638).

Variables	b	S.E.
Lack of prior ties	−0.08	0.04*
Extensive courtship	−0.76	0.33*
Lack of ties × extensive courtship	0.18	0.07**
ρ	0.61	0.19***

Other covariates not shown.

p < 0.05, **p < 0.01, ***p < 0.001 (one-tailed test).

The main findings of the empirical analyses are as follows. The results suggest that a minority of adolescents engage in extensive courtship, which appears to be consistent with costly signaling since it involves more social and dyadic behaviors. Proxies for unwillingness to invest in relationships – sexual attitudes and experiences – are associated with lower likelihoods of engaging in extensive courtship. Extensive courtship is strongly associated with the onset of sexual intercourse, but this association reflects selection effects. We do observe, however, that extensive courtship does promote sexual intercourse more for those that are unembedded in social networks. Taken together, these findings provide support for a signaling interpretation of courtship behaviors in adolescent romantic relationships, but also an embeddedness account.

Discussion

This research investigated whether associations among patterns, predictors, and consequences of courtship behavior were consistent with a signaling framework. To address this question, we adapted a trust game with incomplete information to include a prestage of courtship. This model yielded testable hypotheses regarding associations among proxies for unwillingness to invest in romantic relationships, courtship patterns, and the onset of sexual intercourse. We found that adolescents’ courtship behaviors preceding sexual involvement are organized into four distinct classes: nominal, going-with, social, and extensive. While the minimal and extensive forms capture the lower and upper bounds of costs of courtship, respectively, the other modes represent distinct classes, which would be missed by simple counts. Consistent with the idea that costs of courtship lead to self-selection, we also showed empirically that adolescents oriented towards sex over romance were less willing to spend time engaging in extensive courtship. Finally, the key finding of this research is that courtship behaviors matter not only as cultural and structural precursors of sexual intercourse, but as a selection mechanism. That is, the observed association between courtship behavior and sexual intercourse reflects the tendency for those that are willing to make relational investments to select into extensive courtship. We did observe, however, that courtship behaviors might be important when adolescents are unfamiliar with their partners. To the best of our knowledge, this research is the first to both identify distinct types of courtship in adolescents’ romantic relationships and to test whether the link between relationship-building processes and sexual intercourse reflects causation or selection mechanisms.

This research, we believe, makes several contributions to the study of adolescent romantic relationships. To the best of our knowledge, our research presents the first population-based estimates of patterns of courtship behavior in adolescent romantic relationships. As we discussed above, the extensive class appears to be consistent with traditional dating, incorporating both social and dyadic relationship-building behaviors. The nominal, going-with, and social classes are of particular interest: they represent alternative forms of courtship and are more associated with the onset of sexual intercourse, net of selection effects. It is also worth highlighting that while adolescents who are uninterested in building romantic relationships are less likely to engage in courtship, the converse is not true. We found that adolescents who are likely to be pursuing steady romantic relationships are not more likely to engage in extensive courtship. One possible interpretation is that while adolescents with sexual interests are screened out of extensive courtship – that is, they are aware of the costs of courtship and therefore avoid extensive courtship, those with romantic interests do not always use courtship behaviors as a screening mechanism. Finally, the finding that selection effects account for some of the association between extensive courtship behaviors and sexual intercourse suggests that future studies of sexual relationships, both adolescent and adult, could benefit by examining the significance of courtship signaling.

This research also has several theoretical implications. First, in the context of widespread sociological interest in the causal effects of culture (e.g. scripts and repertoires) and structure (e.g. investments, embeddedness, and contracts), focusing on signaling in courtship brings increased attention to an oft-neglected, alternative process that produces relations of trust and commitment. Indeed, it is our goal to highlight how selectivity associated with signaling may operate in conjunction with sociological models of cultural and structural influence. For example, our argument highlights a key dimension of meaning structures. While courtship classes reflect alternate pathways through which adolescents symbolically construct sexual relationships, they also exhibit an underlying structure corresponding to the costs of courtship. Thus, this research highlights connections among symbols, investments, and signals, and it suggests that future research could examine how other cultural and social structures induce selectivity associated with signaling.

Second, the formal model presented in this paper is related to a number of games, including Camerer’s (1988) courtship-investment game, Bacharach and Gambetta’s (2001) ‘mimic-beset trust game,’ and Raub’s (2004) ‘hostage-posting trust game with incomplete information and uncertainty.’ Our findings suggest that future research could examine the relevance of our specific courtship model, as well as the general trust game with signaling, for other social phenomena. For example, one aspect of courtship that is implied by our model is the importance of temporal costs. However, we could not address this issue due to substantial temporal uncertainty in adolescents’ reports about when their relationships started. Future research could examine how temporal aspects of courtship are relevant in both sexual partnerships as well as other types of relations. In addition, our finding that extensive courtship is more associated with sexual intercourse among partners who are unembedded suggests the need for incorporating the binding effects of courtship in a formal model.

Finally, a key policy implication of this research is that adolescents can glean information about their potential partners’ intentions through courtship behaviors. Although our findings are consistent with the notion that adolescents primarily interested in sex were screened out of extensive courtship, we did not observe that adolescents with romantic interests were more likely to engage in extensive courtship. This suggests that adolescents may not be aware of the information advantages provided through courtship. Indeed, in the context of today’s retreat from dating rituals, there appears to be a shift towards gathering less information about the intentions of potential partners. We note that costly courtship does not necessarily have to be produced by staid, antiquated courtship rituals; rather, it can involve idiosyncratic practices. From this perspective, the key policy implication of this research is to highlight the informational advantages of courtship.

Our model emphasizes strategic decision-making, but adolescents are likely to be constrained in several ways. Social structure and culture play a central role in determining how many choices, repertoires, or strategies of action adolescents have. Because of prevailing expectations and obligations, some adolescents may be pressured to engage in certain kinds of courtship behaviors, while avoiding others; they may have limited choices. For example, the fact that Asian American adolescents were more likely to engage in a going-with pattern of courtship may reflect greater parental disapproval for adolescent sexual relationships. Personality characteristics, such as shyness, may affect an adolescent’s propensity to engage in specific courtship behaviors. In addition, some adolescents may lack the skills for interpreting courtship behaviors. The assumption that adolescents adopt different patterns of courtship behaviors based on expected payoffs presents an oversimplified view of decision-making. Some may point out, with good reason, that courtship is based on exploration, emotion, and developmental considerations. However, we do not assume that adolescents always have complete information about their partners, nor do we assume that everyone makes the right decision. There may indeed be some exploration involved and a possible learning curve. More experienced adolescents may be better able to understand the strategic aspects of courtship, but the alternative is to assume that adolescents are myopic. These interpretive and normative issues suggest limitations to conceptualizing courtship as a variant of the trust game.

There are additional limitations of this research. First, these analyses only present a snapshot of adolescent courtship behaviors from the 1990s. Nevertheless, to the extent that today’s adolescents may be more likely to engage in hookups, this research is important because it provides a key benchmark and shows that patterns of courtship behaviors are theoretically and substantively significant. Future research could examine how these patterns change over the life course for the same individuals or for new cohorts of adolescents. Second, one possible objection is that courtship behaviors are not representative of stable preferences but reflect emergent dynamics. This is most certainly true not only for this study, but for others as well (e.g. Harding, 2007). However, the fact that we observe associations among prior sexual attitudes and experiences and patterns of courtship behavior suggest that the latter are not completely negotiated. Third, our game-theoretic model makes several simplifying assumptions. For example, our model assumes symmetric payoffs for males and females. It is possible that the cost structure for young women is higher than that of young men. Costs associated with sexual activity – pregnancy and the likelihood of contracting STDs – may be higher for female than male adolescents, which could impact the payoff structure depicted in our model. As we noted previously, several formal models have been proposed, but the core insight of all of these game-theoretic frameworks is that lavishness in courtship can provide information about the motives of a potential partner. Fourth, this research has not considered the importance of reputation as a solution to trust problems. We note that reputation is likely to be less important when decisions are one-shot and when abusing trust is particularly attractive. From this perspective, signaling processes may be particularly important when reputational concerns are limited.

Finally, we have several data limitations. A key issue is that our game theoretic perspective involves two adolescents, but our empirical analyses utilize individual-level data. This is less of an issue that it appears. Like investment models, the key finding of the game-theoretic model is the deployment of courtship behaviors in the relationship. We suggest that individuals can reliably report whether or not they and their romantic partners engaged a particular activity. Future research may need to focus on couple data to better understand how characteristics of one or both partners affect decision making about courtship behaviors. Because Add Health did not collect information about courtship behaviors in some like and nonromantic relationships, we know little about these behaviors outside of romantic relationships. For example, we lacked information about failed courtships. Also, the Add Health measures only captured the onset of each courtship behavior. We do not know the extent to which these behaviors were repeated.

What do our results say about the politics of adolescent sexuality? Do these results suggest a return to tradition and a rejection of liberal and libertarian views of sexuality, as some commentators will most assuredly call for? We eschew that debate here; rather, we note that adolescents vary in their abilities to read and interpret signals in courtship. Both a return to the ‘rules’ or a rejection of ‘dated traditions’ have the perverse effect of limiting choice, while having a diverse set of courtship patterns and knowledge about their meanings will allow adolescents to find those they seek. Simmel once noted that the struggle in contemporary modern life is to transcend the ‘forms’ that organize it. We have sought to raise the profile of some of the signals, masking as courtship behaviors, currently at play in adolescent life. Because courtship is an important institution in adolescents’ social lives and adolescents do not know, in advance, the intentions of potential partners, being a skilled reader and user of signals enhances their odds of connecting with others with similar intentions. In short, we have endeavoured to show that signaling theory renders new insights about social phenomena, in romance and in everyday life.

Footnotes

Appendix A. Items used in summated scales

Acknowledgements

We are indebted to Jennifer Glanville, David Harding, Karen Heimer, Kara Joyner, Michael Lovaglia, Freda Lynn, Michael Sauder, Lynne Zucker, and participants of the Theory Workshop at the University of Iowa for reading and commenting on earlier drafts.

Notes

References

Anderson

(1999) Code of the Street: Decency, Violence, and the Moral Life of the Inner City. New York: Norton.

Bacharach

Gambetta

(2001) Trust in signs. In: Cook

(ed.) Trust in Society, pp. 148–184. New York: Russell Sage Foundation.

Bearman

Bruckner

(2001) Promising the future: Virginity pledges and first intercourse. American Journal of Sociology 106: 859–912.

Bearman

Jones

Udry

(1997) The National Longitudinal Study of Adolescent Health: Research Design. Carolina Population Center. University of North Carolina at Chapel Hill.

Bearman

Moody

Stovel

(2004) Chains of affection: The structure of adolescent romantic and sexual networks. American Journal of Sociology 110: 44–91.

Becker

(1981) A Treatise on the Family. Cambridge, MA: Harvard University Press.

Bogle

(2008) Hooking Up: Sex, Dating, and Relationships on Campus. New York: New York University Press.

Bolle

(2001) Why to buy your darling flowers: On cooperation and exploitation. Theory and Decision 50: 1–28.

Brines

Joyner

(1999) The ties that bind: Principles of cohesion in cohabitation and marriage. American Sociological Review 64: 333–335.

10.

Brown

(1999) ‘You’re going out with who?’ Peer influences on adolescent romantic relationships. In: Furman

Brown

Feiring

(eds) The Development of Romantic Relationships in Adolescence, pp. 291–329. Cambridge: Cambridge University Press.

11.

Buskens

Raub

(2002) Embedded trust: Control and learning. Advances in Group Processes 19: 167–202.

12.

Camerer

(1988) Gifts as economic signals and social symbols. American Journal of Sociology 94: S180–S214.

13.

Camerer

Wiegelt

(1988) Experimental tests of a sequential equilibrium reputation model. Econometrica 56: 1–36.

14.

Carmichael

MacLeod

(1997) Gift giving and the evolution of cooperation. International Economic Review 38: 485–509.

15.

Carver

Joyner

Udry

(2003) National estimates of adolescent romantic relationships. In Florsheim

(ed.) Adolescent Romantic Relationships and Sexual Behavior: Theory, Research, and Practical Implications, pp. 23–56. Mahwah, NJ: Erlbaum.

16.

Cavanagh

(2007) The social construction of romantic relationships in adolescence: Examining the role of peer networks, gender, and race. Sociological Inquiry 77: 572–600.

17.

Coleman

(1990) Foundations of Social Theory. Cambridge: Harvard University Press.

18.

Collins

(2004) Interaction Ritual Chains. Princeton: Princeton University Press.

19.

Connolly

Goldberg

(1999) Romantic relationships in adolescence: The role of friends and peers in their emergence and development. In: Furman

Brown

Feiring

(eds) The Development of Romantic Relationships in Adolescence, pp. 266–290. Cambridge: Cambridge University Press.

20.

Dasgupta

(1988) Trust as a commodity. In: Gambetta

(ed.) Trust: Making and Breaking Cooperative Relations, pp. 49–72. Oxford: Blackwell.

21.

Davis

(1973) Intimate Relations. New York: The Free Press.

22.

England

Farkas

1986. Households, Employment, and Gender. New York: Aldine.

23.

Felmlee

(2001) No couple is an island: A social network perspective on dyadic stability. Social Forces 79: 1259–1287.

24.

Felmlee

Sprecher

Bassin

(1990) The dissolution of intimate relationships: A hazard model. Social Psychology Quarterly 53: 13–30.

25.

Frank

(1988) Passions with Reason: The Strategic Role of Emotions. New York: W.W. Norton.

26.

Freeman

(1978) Segregation in social networks. Sociological Methods and Research 6: 411–429.

27.

Furman

Wehner

(1994) Romantic views: Toward a theory of adolescent romantic relationships. In: Montemeyer

Adams

Gullota

(eds) Advances in Adolescent Development: Personal Relationships During Adolescence, Volume 3, pp. 168–195. Thousand Oaks, CA: Sage.

28.

Furstenberg

Morgan

Moore

Peterson

(1987) Race differences in the timing of adolescent intercourse. American Sociological Review 52: 511–518.

29.

Gagnon

Simon

(1973) Sexual Conduct: The Social Sources of Human Sexuality. Chicago: Aldine.

30.

Giordano

Longmore

Manning

(2001) On the nature and developmental significance of adolescent romantic relationships. Sociological Studies of Children and Youth 8: 111–139.

31.

Giordano

Longmore

Manning

(2006) Gender and the meaning of adolescent romantic relationships. American Sociological Review 71: 260–287.

32.

Goffman

(1959) The Presentation of Self in Everyday Life. Garden City, NY: Doubleday Anchor.

33.

Harding

(2007) Cultural context, sexual behavior, and romantic relationships in disadvantaged neighborhoods. American Sociological Review 72: 341–364.

34.

Harris

Duncan

Boisjoly

(2002) Evaluating the role of ‘nothing to lose’ attitudes on risky behavior in adolescence. Social Forces 80: 1005–1039.

35.

Harsanyi

(1967) Games with incomplete information played by Bayesian players; I–II. Management Science 14: 159–182, 320–334, 486–502.

36.

King

Christensen

(1983) The relationship events scale: A Guttmann scaling of progress in courtship. Journal of Marriage and the Family 45: 671–678.

37.

Laner

Ventrone

(2000) Dating scripts revisited. Journal of Family Issues 21: 488–500.

38.

Larson

Clore

Wood

(1999) The emotions of romantic relationships: Do they wreak havoc on adolescents? In: Furman

Brown

Feiring

(eds) The Development of Romantic Relationships in Adolescence, pp. 19–49. Cambridge: Cambridge University Press.

39.

Laumann

Gagnon

Michael

Michaels

(1994) The Social Organization of Sexuality: Sexual Practices in the United States. Chicago: University of Chicago Press.

40.

Lauritsen

(1994) Explaining race and gender differences in adolescent sexual behavior. Social Forces 72: 859–884.

41.

Lawler

Yoon

(1993) Power and the emergence of commitment behavior in negotiated exchange. American Sociological Review 58: 365–481.

42.

Manlove

Ryan

Franzetta

(2007) Contraceptive use patterns across teens’ sexual relationships: The role of relationships, partners, and sexual histories. Demography 44: 603–621.

43.

Manning

Longmore

Giordano

(2005) Adolescents’ involvement in nonromantic sexual activity. Social Science Research 34: 384–407.

44.

Manning

Giordano

Longmore

(2006) Hooking up: The relationship contexts of ‘nonrelationship.’ Sex Journal of Adolescent Research 21: 459–483.

45.

Meier

(2003) Adolescents’ transition to first intercourse, religiosity, and attitudes about sex. Social Forces 81: 1031–1052.

46.

Merten

(1996) Going-with: The role of a social form in early romance. Journal of Contemporary Ethnography 24: 462–484.

47.

Miller

Benson

(1999) Romantic and sexual relationship development during adolescence. In: Furman

Brown

Feiring

(eds) The Development of Romantic Relationships in Adolescence, pp. 99–121. Cambridge: Cambridge University Press.

48.

Moore

(1995) Courtship signaling and adolescents: ‘Girls just wanna have fun’? Journal of Sex Research 32: 319–328.

49.

Muthén

(2004) Latent variable analysis: Growth mixture modeling and related techniques for longitudinal data. In: Kaplan

(ed.) Handbook of Quantitative Methodology for the Social Sciences, pp. 345–368. Newbury Park, CA: Sage.

50.

Niehuis

Huston

Rosenband

(2006) From courtship into marriage: A new developmental model and methodological critique. Journal of Family Communication 6: 23–47.

51.

Nylund

Asparouhov

Muthén

(2007) Deciding on the number of classes in latent class analysis and growth mixture modeling. A Monte Carlo Simulation Study. Structural Equation Modeling 14: 535–569.

52.

O’Sullivan

Cheng

Harris

Brooks-Gunn

(2007) I wanna hold your hand: The progression of social, romantic and sexual events in adolescent relationships. Perspectives on Sexual and Reproductive Health 39: 100–107.

53.

Paul

McManus

Hayes

(2000) ‘Hookups’: Characteristics and correlates of college students’ spontaneous and anonymous sexual experiences. Journal of Sex Research 37: 76–88.

54.

Raub

(2004) Hostage posting as a mechanism of trust: Binding, hostage posting, and signaling. Rationality and Society 16: 319–365.

55.

Raub

Weesie

(2000) Cooperation via Hostages. Analyse & Kritik 22: 19–43.

56.

Ramirez-Valles

Zimmerman

Newcomb

(1998) Sexual risk behavior among youth: Modeling the influence of prosocial activities and socioeconomic factors. Journal of Health and Social Behavior 39: 237–253.

57.

Reiss

(1960) Toward a sociology of the heterosexual love relationship. Marriage and Family Living 22: 139–145.

58.

Rusbult

Buunk

(1993) Commitment processes in close relationships: An interdependence analysis. Journal of Social and Personal Relationships 10: 175–204.

59.

Rusbult

Johnson

Morrow

(1986) Predicting satisfaction and commitment in adult romantic involvements: An assessment of the generalizability of the investment model. Social Psychology Quarterly 49: 81–89.

60.

Simon

Gagnon

(1986) Sexual scripts: Permanence and change. Archives of Sexual Behavior 15: 97–120.

61.

Snijders

Buskens

(2001) How to convince someone that you can be trusted? The role of ‘hostages.’ Journal of Mathematical Sociology 25: 355–383.

62.

Sozou

Seymour

(2005) Costly but worthless gifts facilitate courtship. Proceedings of the Royal Sociology B 272: 1877–1884.

63.

Spence

(1973) Time and communication in economic and social interaction. Quarterly Journal of Economics 87: 651–660.

64.

Sprecher

(1998) Social exchange theories and sexuality. Journal of Sex Research 35: 32–43.

65.

Sprecher

Felmlee

(1992) The influence of parents and friends on the quality and stability of romantic relationships: A three-wave longitudinal investigation. Journal of Marriage and Family 54: 888–900.

66.

Surra

(1990) Research and theory on mate selection and premarital relationships in the 1980s. Journal of Marriage and the Family 52: 844–865.

67.

Swidler

(1986) Culture in action: symbols and strategies. American Sociological Review 51: 273–286.

68.

Swidler

(2001) Talk of Love: How Culture Matters. Chicago: University of Chicago Press.

69.

Thornton

(1990) The courtship process and adolescent sexuality. Journal of Family Issues 11: 239–273.

70.

Treas

(1993) Money in the bank: Transaction costs and the economic organization of marriage. American Sociological Review 58: 723–734.

71.

Upchurch

Levy-Storms

Sucoff

Aneshensel

(1998) Gender and ethnic differences in the timing of first sexual intercourse. Family Planning Perspectives 30: 121–127.

72.

Vaquera

Kao

(2005) Private and public displays of affection among interracial and intra-racial adolescent couples. Social Science Quarterly 86: 484–506.

73.

Waller

(1937) The rating and dating complex. American Sociological Review 2: 727–734.

74.

Walster

Berscheid

(1978) Equity: Theory and Research. Boston: Allyn and Bacon.

75.

Zerubavel

(1981) Hidden Rhythms: Schedules and Calendars in Social Life. Berkeley, CA: University of California Press.