**Data Analytic Plan- Sample A** First, couples will be excluded from the dataset if one or both partners report: they are under the age of 18, their relationship status is single, or they are not sexually active. Then, couples will be randomly sorted into Dataset A (used for the exploratory analyses of Sample A), or Dataset B (used for the confirmatory analyses of Sample B). The following analyses will be run using Dataset A. To examine whether perceptions of a partner’s sexual advances are biased and accurate, we will use West and Kenny’s (2011) T&B model of judgment. In this model, the person making judgments of their partner’s advances is termed the perceiver; and the perceiver’s judgments are compared with their partner’s actual sexual advance ratings. Data will have a nested structure, with perceivers and partners ratings of sexual advances across the 29 items (Level 1) nested within dyad (Level 2). First, the associations across the perceivers’ judgments of their partner’s advances and the partners’ actual reported advances (the Level 1 repeated measures variables) will be examined to test the degree to which judgments of the partner’s sexual advances were accurate and biased. The perceiver’s judgments of their partner’s sexual advances (the outcome variable) will be centered on the partner’s actual advance rating by subtracting the grand mean of all the partners’ advance ratings (i.e., mean across dyads) from the perceivers’ judgments for each behaviour. This centering strategy means that the intercept represents the difference between the mean of the partner’s actual advance rating and the mean of the perceivers’ judgments of that advance rating. The average of this coefficient across perceivers tests directional bias. A negative average intercept will indicate a general tendency for perceivers to underestimate their partners’ sexual advances, whereas a positive average intercept will indicate that perceivers generally overestimate partners’ advances. The effect (slope) of the partner’s actual advance ratings on the perceiver’s judgments of those ratings reflects tracking accuracy, and the effect (slope) of the perceiver’s own advance ratings on their judgments of their partner’s advances reflects assumed similarity. A positive slope will indicate greater tracking accuracy or assumed similarity, respectively. The same bias and accuracy model described above will be conducted with the addition of the following variables as moderators, with a new model run for each moderator: attachment anxiety, attachment avoidance, self-esteem, gender, relationship length, and sexual frequency. A main effect of the moderator indicates directional bias, and the interaction of the moderator and the truth and bias forces indicate the extent to which the moderator is associated with more or less accuracy and assumed similarity, respectively. In these analyses, the Level 1 intercept (directional bias) and slopes (tracking accuracy and assumed similarity, respectively) are treated as dependent variables predicted by individual differences in the moderator modeled at Level 2. To assess the relationship consequences of accurate and biased sexual advance knowledge, we will use multilevel polynomial regression with response surface analyses. The relationship consequences we will examine are relationship satisfaction, sexual satisfaction, sexual frequency, and love. These analyses will allow us to test how tracking accuracy and directional bias are associated with each of the relationship consequences. **Additional Analyses** Reviewers have requested additional analyses examining whether accuracy and bias are associated with the partner enacting more direct (vs. indirect) behaviours. We have obtained ratings of each behaviour on their directness from a sample (N = 11) research assistants and graduate students. We will begin by running an exploratory factor analysis to see if the behaviours naturally group towards direct/indirect behaviours. We also anticipate that they may naturally group towards verbal/nonverbal behaviours, as this method of behaviour grouping has been used in previous research regarding sexual advances (e.g. Vannier & O’Sullivan, 2011). If the items naturally separate this way, we will then run a confirmatory factor analysis with Sample B, after which we will run separate accuracy and bias analyses using each factor to determine if partners are accurate and biased when considering each subset of behaviours (i.e. direct or indirect). If the items do not separate based on directness, we will use the average directness rating for each behaviour from the research assistants and graduate students as a moderator of accuracy and bias in the analyses described above. Reviewers have also suggested that sexual desire might be, in part, responsible for some of the gender differences noted. Unfortunately, we did not measure desire directly, and therefore are unable to directly test the hypothesis that differences in desire can account for the gender differences found. However, we do have measures of the frequency with which each partner approaches the other, how often each partner turns down the other, and how often each partner wants to approach the other but does not because they believe they will be rejected. Using this information, we will combine how often each partner approaches, how often they want to approach but do not, and how often their partner approaches minus how often they say no to their partner. We believe this composite is the closest approximation to desired frequency of sex that we have in the current dataset. We will then use this composite (both actor and partner desired frequency) as a moderator of accuracy and bias in the place of average frequency of initiation and rejection behaviours.