Although differences can sometimes be initially attractive (see, also, Dijkstra & Barelds, 2008), research by Felmlee (2001) on fatal attractions suggests that differences can lead to problems in the relationship over time
More relevant to the topic of similarity leading to compatibility is another issue examined in some research that has studied similarity in ongoing couples -how degree of similarity between partners is associated with relationship quality, such as satisfaction. Findings have been weak or inconsistent, however. g., for a review of early work, see Karney & Bradbury, 1995; for more recent research, see Gonzaga et al., 2007; Luo & Klohnen, 2005), whereas other research has found weak or non-existent associations (Barelds, 2005; Gattis, Berns, Simpson, & Christensen, 2004; Gaunt, 2006). When perceived similarity is the focus, individuals in ongoing relationships report that they are similar (more than they are different) with their partner, and beliefs about similarity are associated with relationship quality (e.g. Lutz-Zois, Bradley, Mihalik, & Mooorman-Eavers, 2006; Sprecher, 1998a & b).
Some recent studies, however, have shown that in some contexts or for some variables, complementarity may occur and/or be associated with a positive outcome for the relationship
Montoya et al. (2008) conducted a meta-analysis study of the impact of actual and perceived similarity on attraction and satisfaction across studies using three of the methods referred to above: no-interaction (phantom other) studies, brief-interaction studies, and studies focused on existing couples. Reflecting the types of similarity most often examined in the literature, the meta-analysis focused on the similarity effect for attitudes and personality traits. The researchers reported that the effect for actual similarity was strong for no-interaction studies, moderate for brief-interaction studies, and weak for studies with existing couples. The effect of perceived similarity was found to be equally strong across the three types of research.
More recently, similarity effects have been examined with data collected from users at Internet dating sites, although this research has been limited to data from online dating sites (e.g., Match) that focus on self-selection, and not from the Internet sites that involve scientific compatibility matching (e.g., eHarmony). The standardized items that all members complete at such sites (e.g., Match) are generally limited to a few questions; therefore, similarity cannot be examined for personality and attitudes, the dimensions most frequently examined in prior research. With this caveat, the recent research indicates that users have preferences for similar others. For example, Fiore and Donath (2005) obtained from 65,000 users profile information, reported preferences for partners, and actual communication with other members at the site. They found that the users preferred others who were similar to themselves on several variables such as marital history, desire for children, self-reported physical appearance, and smoking behavior. Skopek, Schulz, and Blossfeld (2010), using data from 13,573 users at a German online dating site, found that both men and women were likely to initiate contact and respond to messages with those others who were similar in education. Similar results were found with users in a dating site in Israel (Shtatfeld & Barak, 2009). Studies conducted with data collected from matching sites have also indicated a preference for someone of the same race (Hitsch, Hortacsu, & Ariely, 2009).
One implication of the overwhelming evidence for the similarity effect is that little support is found for complementarity, or the notion that opposites attract or that differences lead to relationship compatibility. In fact, most of the similarity research referred to above is also evidence for a lack of the complementarity effect. I referred earlier to Aron et al.’s (2006) phantom other study, which provided suggestive evidence that when there is a guarantee of being liked, attraction to differences can occur, at least among men for a same-gender other (based on differences in interests) in an experimental context. Benefits of differences were also found in a study by Baxter and West (2003), in which members of friendships and romantic couples were interviewed and asked to discuss both how they were similar and how they were different. The participants identified both similarities and differences in their relationships, and the differences (as well as the similarities) were described as having both good and bad consequences for the relationship. Individual growth was seen as the primary advantage of differences. Baxter and West concluded “our results suggest that, at any given point in time, the snapshot of a relationship contains both similarities and differences, sometimes lodged in the same phenomenon” (p. 510). Another recent study (Amodio & Showers, 2005) found that while perceived similarity was associated 
with liking in high-committed dating relationships of college students, in relationships characterized by low commitment, dissimilarity was associated with greater liking. In a study that focused on assortative mating across a range of variables in newlywed couples, Luo and Klohen (2005) found similarity on attitudes and some personality traits, but also found some evidence for complementarity (negative assortment) for the personality trait, extraversion.