Defining empathy

The construct of empathy has received ample research attention since the early 1900s, with recent literature searches yielding roughly 5,700 (PubMed, searched April 16, 2014) to 7,000 (PsycINFO, searched April 16, 2014) works that have included the term as a key word and thousands more that mention empathy in their abstracts. Reviews of this literature are complicated by the myriad ways in which empathy is defined across studies. Derived from the Greek word empatheia, which translates roughly to “passion” (“Empathy,” 2013), the word empathy has acquired a number of related, but distinct and often subtly nuanced, connotations in the research literature (e.g., Batson, Reference Batson, Decety and Ickes2009; Decety, Reference Decety2011; Preston & de Waal, Reference Preston and de Waal2002). Developmental psychologists define empathy as an affective state that is (a) elicited by observing or imagining another's affective state, (b) similar to the other's emotional state, and (c) caused by the other's emotional state. Further, they distinguish empathy from a variety of related processes (e.g., self-regulation and executive control) that influence social behaviors, as well as from moral and motivational outcomes of empathy, such as compassion and prosocial behavior (e.g., Batson, Reference Batson, Mikulincer and Shaver2010; de Vignemont & Singer, Reference de Vignemont and Singer2006; Eisenberg, Reference Eisenberg, Mikulincer and Shaver2010). Although individuals can respond empathically to both negative and positive emotions, this review will focus on empathic responses to others' negative affective states.

Researchers also consistently distinguish affective empathy, which can be observed as early as the neonatal period, from cognitive empathic processes that emerge later in development. Affective empathy, often termed empathic concern, is a response to witnessing another's emotions that involves both basic emotion-understanding skills (e.g., emotion recognition) and emotional reactivity processes (e.g., emotion contagion). In its typical or adaptive form, affective empathy leads to compassionate responses to other's emotional states (Davis, Luce, & Kraus, Reference Davis, Luce and Kraus1994; Shamay-Tsoory et al., Reference Shamay-Tsoory, Aharon-Peretz and Perry2009).

Cognitive empathy refers to more advanced cognitive perspective-taking systems (Shamay-Tsoory et al., Reference Shamay-Tsoory, Aharon-Peretz and Perry2009) that underlie the ability to understand and consider others' viewpoints when solving interpersonal problems (Davis, Reference Davis1983). It involves the related, but not identical, constructs of theory of mind (i.e., the ability to distinguish one's own emotional state from others' emotional states; Blair, Reference Blair2005) and empathic accuracy (i.e., the ability to infer others' feelings from their behaviors; Ickes, Stinson, Bissonnette, & Garcia, Reference Ickes, Stinson, Bissonnette and Garcia1990). The presence of this cognitive aspect of empathy is generally associated with positive social behaviors, such as cooperation, provision of social support, and volunteering (Carlo, Allen, & Buhman, Reference Carlo, Allen and Buhman1999; Cliffordson, Reference Cliffordson2002; Eisenberg, Carlo, Murphy, & Van Court, Reference Eisenberg, Carlo, Murphy and Van Court1995; Ferguson & Austin, Reference Ferguson and Austin2010; Gleason, Jensen-Campbell, & Ickes, Reference Gleason, Jensen-Campbell and Ickes2009; Johnson, Cheek, & Smither, Reference Johnson, Cheek and Smither1983; Schreiter, Pijnenborg, & aan het Rot, Reference Schreiter, Pijnenborg and aan het Rot2013; Verhofstadt, Buysse, Ickes, Davis, & Devoldre, Reference Verhofstadt, Buysse, Ickes, Davis and Devoldre2008).

Although clearly related, these two interpersonal response tendencies appear to function independently, such that the ability to practice cognitive perspective taking does not necessarily predict one's proclivity to experience affective empathy or vice versa. Behavioral research has broadly supported this distinction from early childhood onward, with ample evidence of at least partial dissociation between the two during preschool (e.g., Belacchi & Farina, Reference Belacchi and Farina2012; Strayer, Reference Strayer1980), elementary school (e.g., Feshbach & Roe, Reference Feshbach and Roe1968; Zajdel, Bloom, Fireman, & Larsen, Reference Zajdel, Bloom, Fireman and Larsen2013), adolescence (e.g., Jones, Happé, Gilbert, Burnett, & Viding, Reference Jones, Happé, Gilbert, Burnett and Viding2010), and adulthood (e.g., Cox et al., Reference Cox, Uddin, Di Martino, Castellanos, Milham and Kelly2012). Depending on the interpersonal and intrapersonal contexts in which it emerges, an empathic response may draw on either or both components (Schnell, Bluschke, Konradt, Walter, Reference Schnell, Bluschke, Konradt and Walter2011; Shamay-Tsoory, Torner, Goldsher, Berger, & Aharon-Peretz, Reference Shamay-Tsoory, Tomer, Goldsher, Berger and Aharon-Peretz2004). The strength and the quality of each component's contributions are further likely to vary as a function of numerous factors, including the individual's developmental status and idiosyncratic functional neural dynamics (Cox et al., Reference Cox, Uddin, Di Martino, Castellanos, Milham and Kelly2012). Thus, affective and cognitive empathy are related but distinct constructs that are usually associated with positive social, emotional, and moral outcomes. However, under some circumstances, both can also develop in ways that increase risk for maladaptive outcomes, as detailed in forthcoming sections.

Developmental pathways for empathy

Most individuals acquire cognitive and affective empathy in a relatively predictable progression (see Hoffman, Reference Hoffman2000); however, there is considerable variability within the normative developmental range in both infancy and childhood (Bryant, Reference Bryant1982; Knafo, Zahn-Waxler, Van Hulle, Robinson, & Rhee, Reference Knafo, Zahn-Waxler, Van Hulle, Robinson and Rhee2008). Evidence suggests that humans enter life with a tendency to respond preferentially to emotion signals, and even newborns imitate others' emotional cues (Diego & Jones, Reference Diego, Jones, Farrow and Woodruff2007). During the first year of life, typically developing infants demonstrate the capacity to exhibit distress when confronted with others' signals of negative affect (Dondi, Simion, & Caltran, Reference Dondi, Simion and Caltran1999; Geangu, Hauf, Bhardwaj, & Bentz, Reference Geangu, Hauf, Bhardwaj and Bentz2011) and to orient attention toward distressed others (Hay, Nash, & Pederson, Reference Hay, Nash and Pedersen1981). As they transition into their second year, infants show marked changes in their patterns of empathic responding, presumably because they become better able to differentiate themselves from others and thus more capable of directing their concern externally (Eisenberg, Fabes & Spinrad, Reference Eisenberg, Fabes, Spinrad, Eisenberg, Damon and Lerner2006; Hoffman, Reference Hoffman, Izard, Kagan and Zajonc1985; Hoffman, Reference Hoffman2000). In addition, the ability to reason about others' goals and intentions improves steadily during the first few years of life, which likely facilitates a more sophisticated understanding of why others might show signs of distress (Sodian, Reference Sodian2011). Findings from at least one study suggest that individuals show relatively stable levels of empathy after age 3 years (Knafo et al., Reference Knafo, Zahn-Waxler, Van Hulle, Robinson and Rhee2008).

During the preschool and school years, as a function of their growing cognitive and behavioral competence, children become increasingly capable of generating and implementing complex and differentiated prosocial behaviors based on empathic feelings and thoughts (Eisenberg & Miller, Reference Eisenberg and Miller1987). For example, verbal language skill serves as a potent predictor of later concern for others in preschoolers (Rhee et al., Reference Rhee, Boeldt, Friedman, Corley, Hewitt and Young2013); these variables appear to be reciprocally and dynamically related, with empathic responding in early toddlerhood also predicting later language skill (Hutman, Rozga, DeLaurentis, Sigman, & Dapretto, Reference Hutman, Rozga, DeLaurentis, Sigman and Dapretto2012). Nonverbal language development also appears to facilitate empathic responding. As children's understanding of emotions and their causes increases with age, for example, their ability to generate empathic responses to others' nonverbal affective displays expands (Catherine & Schonert-Reichl, Reference Catherine and Schonert-Reichl2011).

Over the course of middle childhood, typically developing children show a growing proclivity to manage their feelings with cognitive strategies (Altshuler & Ruble, Reference Altshuler and Ruble1989; Band & Weisz, Reference Band and Weisz1988; Losoya, Eisenberg, & Fabes, Reference Losoya, Eisenberg and Fabes1998; Rice, Levine, & Pizarro, Reference Rice, Levine and Pizarro2007). Research has identified links between maturing emotion regulation capacities and the emergence of more sophisticated empathic skills. For example, as children's cognitive perspective-taking skills improve, their ability to regulate both negative and positive emotions also improves (Bengtsson & Arvidsson, Reference Bengtsson and Arvidsson2010). Although researchers have focused primarily on the idea that empathy may support effective emotion regulation (Schipper & Petermann, Reference Schipper and Petermann2013), the two constructs likely interact dynamically across the course of development, such that effective emotion regulation also reinforces an individual's capacity for empathy (Ungerer et al., Reference Ungerer, Dolby, Waters, Barnett, Kelk and Lewin1990). For instance, whereas emotionally dysregulated individuals are likely to focus attention reflexively on themselves (Wood, Saltzberg, & Goldsamt, Reference Wood, Saltzberg and Goldsamt1990), those who can competently modulate their own emotional reactions may be more likely to direct attentional resources toward others when an external focus is appropriate (e.g., Eisenberg, Reference Eisenberg, Mikulincer and Shaver2010; Eisenberg et al., Reference Eisenberg, Fabes, Murphy, Karbon, Smith and Maszk1996).

Patterns of empathic responding appear to remain stable during and after the transition to adolescence (e.g., Davis & Franzoi, Reference Davis and Franzoi1991), although some skills, such as the capacity to take others' perspectives and to empathize with individuals who are not present, apparently increase in this developmental period (Eisenberg, Cumberland, Guthrie, Murphy, & Shepard, Reference Eisenberg, Cumberland, Guthrie, Murphy and Shepard2005; Hoffman, Reference Hoffman2000). Findings are mixed regarding the developmental course of empathy in adulthood. These inconsistencies appear to reflect differences across studies in (a) the facets of empathy that were measured: cognitive empathy has been found to decline more with age than does affective empathy (Sze, Gyurak, Goodkind, & Levenson, Reference Sze, Gyurak, Goodkind and Levenson2012); (b) longitudinal study duration: decreases in empathy have been observed over several decades (e.g., Eisenberg & Miller, Reference Eisenberg and Miller1987; Helson, Jones, & Kwan, Reference Helson, Jones and Kwan2002) but not necessarily over shorter durations (Grühn, Rebucal, Diehl, Lumley, & Labouvie-Vief, Reference Grühn, Rebucal, Diehl, Lumley and Labouvie-Vief2008); (c) the cohort in which empathy was measured: college students since 2000 endorsed lower cognitive and affective empathy than did students enrolled in college in the 1980s and 1990s (Konrath, O'Brien, & Hsing, Reference Konrath, O'Brien and Hsing2011); and (d) the period of adulthood studied: one study found higher levels of cognitive and affective empathy in middle-aged adults than in either young or old adults (O'Brien, Konrath, Grühn, & Hagen, Reference O'Brien, Konrath, Grühn and Hagen2013).

Taken together, findings from studies targeting age groups from infancy through adulthood suggest that empathic emotions, thoughts, and behaviors emerge and evolve in most individuals according to fairly predictable developmental patterns. Some research indicates variability associated with individual differences such as sex and/or gender identity (Adams, Schvaneveldt, & Jenson, Reference Adams, Schvaneveldt and Jenson1979; Karniol, Gabay, Ochion, & Harari, Reference Karniol, Gabay, Ochion and Harari1998; Mestre, Samper, Frias, & Tur, Reference Mestre, Samper, Frías and Tur2009; Van der Graaff et al., Reference Van der Graaff, Branje, De Wied, Hawk, Van Lier and Meeus2014). For example, by early adolescence, girls endorse more sensitivity to others' perspectives than do boys, and this perceived sensitivity predicts both better relationships with peers and more vulnerability to experiencing distress in the face of others' discomfort (Smith & Rose, Reference Smith and Rose2011). Research has also yielded evidence that empathic concern and perspective taking follow different developmental trajectories over the course of adolescence in girls and boys (Van der Graaff et al., Reference Van der Graaff, Branje, De Wied, Hawk, Van Lier and Meeus2014).

However, it is unclear how to interpret such findings, given evidence that adult males and females show largely comparable patterns of neural response to empathy-eliciting cues, despite sex differences in self-reported empathic experience (Groen, Wijers, Tucha, & Althaus, Reference Groen, Wijers, Tucha and Althaus2013; Michalska, Kinzler, & Decety, Reference Michalska, Kinzler and Decety2013; but for evidence of sex differences, see Schulte-Rüther, Markowitsch, Shah, Fink, & Piefke, Reference Schulte-Rüther, Markowitsch, Shah, Fink and Piefke2008). One possibility is that observed sex differences in endorsements of empathy may more heavily reflect gendered social norms regarding the expression of empathic feelings than sex-linked differences in the experience of such feelings per se. Similar patterns of neural activity between males and females during the presentation of empathy-relevant stimuli may correspond with different cognitive, emotional, and behavioral experiences. Regardless of the underpinnings of observed sex differences in response to stimuli that can provoke empathic responses, associations between empathy and sex are complex and may vary depending on how empathy is measured (Michalska et al., Reference Michalska, Kinzler and Decety2013).

Genetic proclivity toward empathy

Behavior genetic studies support the heritability of affective and cognitive empathy (Knafo & Uzefovsky, Reference Knafo, Uzefovsky, Legerstee, Haley and Bornstein2013), with evidence of moderately strong genetic influences even in the second and third years of life (Knafo et al., Reference Knafo, Zahn-Waxler, Van Hulle, Robinson and Rhee2008). Such findings are consistent with a significant role for genetically influenced neurobiological processes in the emergence of individual differences in empathy early in development. The ways in which genetic effects manifest appear to vary over time; Knafo et al. (Reference Knafo, Zahn-Waxler, Van Hulle, Robinson and Rhee2008) found that, whereas genetic influences accounted for change in empathy in the second year of life, they accounted for continuity of empathy during the third year. This pattern suggests that at different points during development, genetic factors mediate both early empathic development and stability of more trait-like empathy. Behavior genetic findings also suggest that genetic influences serve as the primary mediator of the association between positive affect and cognitive empathy (Volbrecht, Lemery-Chalfant, Aksan, Zahn-Waxler, & Goldsmith, Reference Volbrecht, Lemery-Chalfant, Aksan, Zahn-Waxler and Goldsmith2007), although environmental factors also play key roles in the development of each (Knafo et al., Reference Knafo, Zahn-Waxler, Davidov, Hulle, Robinson, Rhee, Vilarroya, Altran, Navarro, Ochsner and Tobeña2009). Environmental influences also appear to mediate associations between empathy measured broadly and prosocial behavior (Knafo et al., Reference Knafo, Zahn-Waxler, Van Hulle, Robinson and Rhee2008).

Research suggests several probable candidate genes that influence the development of empathy-related constructs. Much of the molecular genetics work to date in this area has focused on single nucleotide polymorphisms (SNPs) within the oxytocin receptor (OXTR) gene, particularly the RS53576 SNP, demonstrating that varied allele distributions for polymorphisms in this gene predict different social/emotional outcomes. Across studies, for example, individuals with two copies of the G allele at varied loci exhibit higher levels of dispositional empathy, interpersonal trust, sensitivity to children, prosocial temperament, and desire for social approval than do those with AG or AA allele configurations (Bakermans-Kranenburg & van IJzendoorn, Reference Bakermans-Kranenburg and van IJzendoorn2008; Costa et al., Reference Costa, Pini, Gabelloni, Abelli, Lari and Cardini2009; F. Krueger et al., Reference Krueger, Parasuraman, Iyengar, Thornburg, Weel and Lin2012; Rodrigues, Saslow, Garcia, John, & Keltner, Reference Rodrigues, Saslow, Garcia, John and Keltner2009; Tost et al., Reference Tost, Kolachana, Hakimi, Lemaitre, Verchinski and Mattay2010). Although the OXTR gene appears to have links to both affective and cognitive components of empathy, some evidence suggests the OXTR gene locus differs for the two (Wu, Li, & Su, Reference Wu, Li and Su2012).

The OXTR gene may affect behavioral and experiential outcomes, including those associated with adaptive states, such as compassion, and maladaptive states, such as personal distress and interpersonal guilt, in several ways. Animal and nasal oxytocin administration studies suggest that genetically mediated variations in the release of oxytocin, as well as in the density, location, and function of oxytocin receptors, underlie individual differences in positive social behaviors, such as trust, calm social interaction, and generosity (Ishak, Kahloon, & Fakhry, Reference Ishak, Kahloon and Fakhry2011; Kosfeld, Heinrichs, Zak, Fischbacher, & Fehr, Reference Kosfeld, Heinrichs, Zak, Fischbacher and Fehr2005). Although most studies find evidence that introducing exogenous oxytocin increases these positive behaviors, at least two studies have shown that it can decrease such behaviors in subgroups of adults with disrupted attachment histories (Bartz, Simeon, et al., Reference Bartz, Simeon, Hamilton, Kim, Crystal and Braun2011; Bartz et al., Reference Bartz, Zaki, Ochsner, Bolger, Kolevzon and Ludwig2010), which is consistent with a moderating role for relevant contextual factors.

Other genes, such as the arginine vasopressin 1a receptor (AVPR1a) gene (Ebstein et al., Reference Ebstein, Israel, Lerer, Uzefovsky, Shalev, Gritsenko, Vilarroya, Altran, Navarro, Ochsner and Tobeña2009; Meyer-Lindenberg et al., Reference Meyer-Lindenberg, Kolachana, Gold, Olsh, Nicodemus and Mattay2009; Yirmiya et al., Reference Yirmiya, Rosenberg, Levi, Salomon, Shulman and Nemanov2006), are receiving growing attention from researchers as likely mediators of empathy-relevant social behavior. Vasopressin, a neuropeptide hormone that, like oxytocin, has been implicated in emotional bonding and related interpersonal experiences (Bales, Reference Bales, Mikulincer and Shaver2014), may be particularly important for social cognition and behavior in males (Atzil, Hendler, Zagoory-Sharon, Winetraub, & Feldman, Reference Atzil, Hendler, Zagoory-Sharon, Winetraub and Feldman2012; Uzefovsky, Israel, Knafo, & Ebstein, Reference Uzefovsky, Shalev, Israel, Knafo and Ebstein2012). Although vasopressin appears to play a role in the development of disorders that are marked by empathic deficits, such as autism (Ebstein, Knafo, Mankuta, Chew, & Lai, Reference Ebstein, Knafo, Mankuta, Chew and Lai2012), limited research has specifically linked this hormone to empathic behavior. Research on parent–child and pair bonding constitutes one notable exception (Atzil et al., Reference Atzil, Hendler, Zagoory-Sharon, Winetraub and Feldman2012; Gray & Campbell, Reference Gray, Campbell, Ellison and Gray2009) and provides a foundation for future work exploring associations between vasopressin and empathy more broadly.

The promoter region of the serotonin transporter linked polymorphic region (5-HTTLPR) gene (Bakermans-Kranenburg & van IJzendoorn, Reference Bakermans-Kranenburg and van IJzendoorn2008; Fox, Zougkou, Ridgewell, & Garner, Reference Fox, Zougkou, Ridgewell and Garner2011) may also play a role in the development of empathy. Research points to both direct and indirect effects of this gene on empathy-relevant behavior, which underscores the complexity of genetic influences on downstream outcomes. Bakermans-Kranenburg and van IJzendoorn (Reference Bakermans-Kranenburg and van IJzendoorn2008), for example, found direct associations between 5-HTTLPR allele variations and maternal sensitivity, with high sensitivity associated with the heterozygous short/long and homozygous variants, rather than with the homozygous short variant, which is associated with the least efficient serotonin functioning. Thus, this gene, like other relevant genetic influences, is likely to function in multiple ways that vary across environmental contexts to contribute to socioemotional outcomes.

Researchers have made notable strides in characterizing genetic influences on empathy and its development. However, it is likely that genes associated with a broad range of additional neurochemicals, such as cortisol, estrogen, opiates, and monoamines, converge to shape the way in which empathic capacities emerge and change across development, and very little is known about how different empathy-related genes interact to influence social behavior. Beyond identifying individual candidate genes, it will be important for future studies to begin to delineate how these genes interact across varied genetic and environmental contexts to influence an individual's trajectory of empathy-relevant neurobiological, physiological, and behavioral development.

Neurobiology of empathy

Although research has shown that variations in oxytocin and other genes relate to empathy and empathy-relevant behavior, these genes exert their influence primarily by modulating the development and function of brain regions and systems that implement empathic experiences and responses. These systems have been carefully studied and described in the literature, and a number of similar, but subtly distinct, models have been articulated (e.g., Decety, Reference Decety, Harmon-Jones and Winkielman2007; Carr, Iacoboni, Dubeau, Mazziotta, & Lenzi, Reference Carr, Iacoboni, Dubeau, Mazziotta, Lenzi, Cacioppo and Berntson2005). For the purposes of this review, we use Decety and colleagues' developmentally informed neural model of empathy as a framework (Decety, Reference Decety, Harmon-Jones and Winkielman2007, Reference Decety2010, Reference Decety2011; Decety & Jackson, Reference Decety and Jackson2004). In addition to its attention to development, this model has the advantage of encompassing bottom-up processes that emerge early and map loosely onto affective empathy, top-down processes that emerge slightly later in development and correspond broadly to cognitive empathy, and an array of integrated processes that contribute interactively to the generation and modulation of empathic responses and behaviors. The nature of the interactions among these processes appears to vary as a function of age-related changes in their mediating neural structures (Decety & Michalska, Reference Decety and Michalska2010). Considerable evidence lends support to the premise that neural structures critical to both affective and cognitive empathy follow varying developmental trajectories, as Decety and colleagues (Decety, Reference Decety2011; Decety & Michalska, Reference Decety and Michalska2010) have pointed out in overviews of the relevant literature.

Basic, largely involuntary capacities to detect and respond to others' affective cues (fundamental aspects of affective empathy) appear to be implemented, at least in part, by subcortical structures that include the amygdala, hypothalamus, and hippocampus. These brain regions begin to mature early; some interpersonal responses that they support, such as affective sharing, appear to emerge within the first few weeks after birth (e.g., Dondi et al., Reference Dondi, Simion and Caltran1999; Field, Woodson, Greenberg, & Cohen, 1987). Cortical regions that engage during affective empathy, particularly the anterior insula, which is implicated in varied relevant functions that include sharing others' experiences of physical and emotional pain (Lamm, Decety, & Singer, Reference Lamm, Decety and Singer2011; Lamm & Singer, Reference Lamm and Singer2010), also appear to show marked developmental changes very early in life (Kalani, Kalani, Gwinn, Keogh, & Tse, Reference Kalani, Kalani, Gwinn, Keogh and Tse2009). However, researchers have demonstrated that the insula also reliably undergoes thinning, or gray matter reduction, during middle childhood, which suggests that its developmental course may be more protracted than are those of subcortical areas (Muftuler et al., Reference Muftuler, Davis, Buss, Head, Hasso and Sandman2011).

Despite this research that implicates several brain regions in empathy-related processes, questions remain regarding the specific mechanisms by which these regions support the development and experience of affective empathy. Nevertheless, findings from numerous studies suggest key roles for several neurochemicals, particularly oxytocin and vasopressin, that have been linked to a range of social behaviors from the neonatal period onward (e.g., Bartz, Zaki, Bolger, & Ochsner, Reference Bartz, Zaki, Bolger and Ochsner2011; Clark et al., Reference Clark, St. John, Pasca, Hyde, Hornbeak and Abramova2013) and that modulate functioning in salient regions such as the amygdala and hippocampus (Meyer-Lindenberg, Domes, Kirsch, & Heinrichs, Reference Meyer-Lindenberg, Domes, Kirsch and Heinrichs2011; Owen et al., Reference Owen, Tuncdemir, Bader, Tirko, Fishell and Tsien2013; Zink & Meyer-Lindenberg Reference Zink and Meyer-Lindenberg2012). Oxytocin, in particular, appears to exert subcortical effects primarily by modulating neural activity in the amygdala, midbrain regions, and dorsal striatum. Not only does it attenuate activation in these areas, which are engaged during the experience of affective states that include fear or distress (Mobbs et al., Reference Mobbs, Marchant, Hassabis, Seymour, Tan and Gray2009; Phelps, Reference Phelps, Whalen and Phelps2009; Schiller, Levy, Niv, LeDoux, & Phelps, Reference Schiller, Levy, Niv, LeDoux and Phelps2008), but it also weakens functional coupling of the amygdala to brain stem regions implicated in autonomic and behavioral manifestations of fear (Baumgartner, Heinrichs, Vonlanthen, Fischbacher, & Fehr, Reference Baumgartner, Heinrichs, Vonlanthen, Fischbacher and Fehr2008; Kirsch et al., Reference Kirsch, Esslinger, Chen, Mier, Lis and Siddhanti2005), which may underlie the calming and bonding effects of oxytocin.

The small pool of existing data is mixed regarding whether and how these neurochemicals, particularly oxytocin, modulate empathy-related activity in relevant cortical regions, such as the anterior insula (Riem et al., Reference Riem, Bakermans-Kranenburg, Pieper, Tops, Boksem and Vermeiren2011; Singer et al., Reference Singer, Snozzi, Bird, Petrovic, Silani and Heinrichs2008). Whereas at least two studies yielded findings suggesting that intranasally administered oxytocin and vasopressin enhance amygdala–insula connectivity in adults during the presentation of infants' emotional cues (Riem et al., Reference Riem, Bakermans-Kranenburg, Pieper, Tops, Boksem and Vermeiren2011) or during a social interaction task (Rilling et al., Reference Rilling, DeMarco, Hackett, Thompson, Ditzen and Patel2012), other research has failed to generate evidence of such modulatory activity (Riem et al., Reference Riem, van IJzendoorn, Tops, Boksem, Rombouts and Bakermans-Kranenburg2012).

In contrast to those structures that show involvement in bottom-up processes associated with affective empathy, structures such as the medial and dorsolateral regions of the prefrontal cortex, which mediate top-down regulatory processes linked to cognitive empathy, mature slowly (Zelazo, Carlson, & Kesek, Reference Zelazo, Carlson, Kesek, Nelson and Luciana2008) with synaptic developmental trajectories that extend into young adulthood (e.g., Gogtay et al., Reference Gogtay, Giedd, Lusk, Hayashi, Greenstein and Vaituzis2004; Shaw et al., Reference Shaw, Kabani, Lerch, Eckstrand, Lenroot and Gogtay2008). However, the findings summarized in a recent review suggest that prefrontal regions are functionally active in newborns and that subsequent maturation of this structural domain revolves around changes in connectivity with other parts of the brain (Grossmann, Reference Grossmann2013). Thus, the neural elements necessary to experience not only the affective but also the cognitive aspects of empathy are present and engaged from birth; however, these elements interact and shape each other in a slowly unfolding process that extends into adulthood.

Some evidence suggests that dopamine, rather than oxytocin, serves as the primary neuromodulator for prefrontal regions involved in top-down cognitive components of empathy (Lackner, Bowman, & Sabbagh, Reference Lackner, Bowman and Sabbagh2010; Lackner, Sabbagh, Hallinan, Liu, & Holden, Reference Lackner, Sabbagh, Hallinan, Liu and Holden2012; Xia, Wu, & Su, Reference Xia, Wu and Su2012). Effects of oxytocin on the circuitry involved in cognitive empathy, where present, appear to be indirect (Shamay-Tsoory, Reference Shamay-Tsoory2011). The data primarily from preclinical studies indicate that the oxytocin and dopamine systems may interact to influence a range of social cognitive and interpersonal outcomes (Young & Wang, Reference Young and Wang2004; for a review, see Baskerville & Douglas, Reference Baskerville and Douglas2010).

Complicating the neurodevelopmental picture for empathy and its component processes is evidence that the social environment shapes the expression of relevant neurobiological processes and genetic liabilities (Knafo & Uzefovsky, Reference Knafo, Uzefovsky, Legerstee, Haley and Bornstein2013). Exposure, for example, to responsive, supportive, and sensitive parenting appears to strengthen young children's proclivity to take others' perspectives (Farrant, Devine, Maybery, & Fletcher, Reference Farrant, Devine, Maybery and Fletcher2012) and adolescents' capacity for both empathic concern and perspective taking (Miklikowska, Duriez, & Soenens, Reference Miklikowska, Duriez and Soenens2011). Such associations, however, appear to vary in strength across youths with different neurobiological and genetic profiles; for instance, Cornell and Frick (2004) found that parenting consistency was more weakly associated with child empathy in behaviorally inhibited youths than it was in children with behaviorally disinhibited temperaments.

The effects of the social environment on maturing brain structures and circuitry wax and wane, with some developmental periods constituting key windows for the acquisition or expression of particular socioemotional functions (Nelson & Guyer, Reference Nelson and Guyer2011). For example, evidence from one study suggests that whereas subcortical regions, such as the hippocampus, may be especially vulnerable in the preschool years to such environmental insults as physical abuse, the prefrontal cortices may be more profoundly affected by maltreatment during adolescence (Andersen et al., Reference Andersen, Tomada, Vincow, Valente, Polcari and Teicher2008). Further, findings from studies of several mammalian species indicate that experiences across development shape the functioning of the oxytocin system (Bales & Perkeybile, Reference Bales and Perkeybile2012), which raises the possibility that similar critical periods may exist for this and other neuropeptide systems as well. Thus, the temporal pattern of an individual's life experiences may shape, in an idiosyncratic fashion, the developmental courses of structures and neurochemical pathways that subserve affective and cognitive aspects of empathy, respectively.

Individual differences beyond age, such as pubertal status and sex, may also influence the boundaries of sensitive periods for varied neural regions and systems (Herting, Maxwell, Irvine, & Nagel, Reference Herting, Maxwell, Irvine and Nagel2012; Ladouceur, Reference Ladouceur2012; Wang et al., Reference Wang, Adamson, Yuan, Altaye, Rajagopal and Byars2012). Several research groups, for example, have found evidence of sex differences in the microstructure of neural white matter distributed throughout the adolescent brain (e.g., Herting et al., Reference Herting, Maxwell, Irvine and Nagel2012; Wang et al., Reference Wang, Adamson, Yuan, Altaye, Rajagopal and Byars2012), with the bulk of findings suggestive of faster white matter development in girls than in boys. Functional neural differences between males and females during empathy-related tasks have also been observed in at least one study of adults; Schulte-Rüther and colleagues (Reference Schulte-Rüther, Markowitsch, Shah, Fink and Piefke2008), for instance, found evidence of distinctive patterns of neural activation in men and women during assessment of their own and others' emotional states. However, other studies have yielded limited evidence that neural activation to empathy-related cues differs according to sex (Groen, Wijers, Tucha, & Althaus, Reference Groen, Wijers, Tucha and Althaus2013; Michalska et al., Reference Michalska, Kinzler and Decety2013), which suggests the need for further research to sort out this pattern of inconsistent findings.

Herting and colleagues (Reference Herting, Maxwell, Irvine and Nagel2012) also identified structural differences associated with pubertal status instead of sex in some empathy-relevant regions, including the insula. In a study focused only on girls, Klapwijk and colleagues (Reference Klapwijk, Goddings, Heyes, Bird, Viner and Blakemore2013) linked adolescent increases in pubertal hormones, particularly estradiol, to enhanced functional neural connectivity within circuits activated during thoughts about others' mental states. Taken together, these findings suggest a role for pubertal hormonal changes in the emergence of at least some neural correlates of empathic experience.

Thus, if we are to understand how empathy functions and malfunctions across development, it will be important that we build on emerging knowledge about the ways in which genetically mediated neurochemical processes (including, but not limited to, release of hormones such as estrogen and testosterone, which are linked to sex and puberty) participate in relevant processes directly and via interactions with such neuropeptides as oxytocin and dopamine and relevant aspects of the social environment (e.g., Bos, Terburg, & van Honk, Reference Bos, Terburg and van Honk2010; Gabor, Phan, Clipperton-Allen, Kavaliers, & Choleris, Reference Gabor, Phan, Clipperton-Allen, Kavaliers and Choleris2012). Identifying additional candidate genes and further elucidating the genetic variants that influence the functioning of such neurochemicals and their receptors will also help us learn how empathy can veer from serving as a strength to functioning as a liability.

The structure of internalizing disorders

For most individuals, empathy is linked throughout the life span with positive interpersonal and intrapsychic outcomes, including better relationships with friends and partners (Chow, Ruhl, & Buhrmester, Reference Chow, Ruhl and Buhrmester2013; Cramer & Jowett, Reference Cramer and Jowett2010), increased social engagement (Bailey, Henry, & Von Hippel, Reference Bailey, Henry and Von Hippel2008), and resilience (Shiner & Masten, Reference Shiner and Masten2012). One can clearly demonstrate too little empathy; a sizable body of research documents associations between an impaired empathic capacity and psychopathological conditions as varied as conduct disorder and autism spectrum disorders (Bons et al., Reference Bons, van den Broek, Scheepers, Herpers, Rommelse and Buitelaaar2013; Decety & Michalska, Reference Decety and Michalska2010; Miller & Eisenberg, Reference Miller and Eisenberg1988). Less research has described potential psychopathological correlates of excessive empathy. However, as Zahn-Waxler and colleagues (Reference Zahn-Waxler, Cole, Barrett, Garber and Dodge1991) have suggested in a model focused explicitly on girls, extreme empathy or a combination of heightened sensitivity to others' distress and inadequate skills for coping with that distress may increase vulnerability to internalizing conditions marked by negative affect.

The research we review here suggests that risk for internalizing disorders may increase when empathy's typical, healthy developmental trajectory is altered in ways that lead to maladaptive consequences. We find the model that Zahn-Waxler and her collaborators (1991) have proposed, in which highly empathic girls may be at particular risk for internalizing conditions due in part to gendered patterns of socialization, to be compelling. However, we contend that although base rates of excessive empathy may be lower in males than in females by early adolescence (Olweus & Endresen, Reference Olweus and Endresen1998), those males who fall at the extreme end of the empathy continuum may also be at heightened risk for internalizing conditions, as a function of moderating variables that manifest in less sex-specific ways. As a consequence, we do not postulate distinct empathy-related pathways to internalizing disorders for males and females. Instead, we propose that the intersection of a strong propensity for empathy with any of a range of intraindividual and interindividual moderators will increase risk regardless of an individual's sex.

One key to unraveling the sources of atypical, maladaptive developmental trajectories in boys and girls alike may derive from the hierarchical structure of internalizing disorders and the genetically influenced regulation liabilities that give rise to this structure. Although the internalizing disorders are often referenced as if they constitute one undifferentiated group, multivariate studies of the latent structure of internalizing disorders have consistently yielded evidence of two highly correlated, but distinct, subfactors: (a) a fear/arousal subfactor that underlies agoraphobia, social phobia, specific phobia, and panic disorder and (b) an anhedonia/misery subfactor that explains the covariance of major depressive disorder, dysthymic disorder, and generalized anxiety disorder (e.g., Kendler, Prescott, Myers, & Neale, Reference Kendler, Prescott, Myers and Neale2003; Krueger & Markon, Reference Krueger and Markon2006). Comorbidity is highest among disorders that load onto a single subfactor (Kessler, Chiu, Demler, & Walters, Reference Kessler, Chiu, Demler and Walters2005); however, disorders that load onto the two internalizing subfactors also tend to co-occur. The fear/arousal and anhedonia/misery subfactors are thought to derive from genetically influenced biological liabilities that are expressed as spectra of related cognitive processes and behavioral tendencies (Tully & Iacono, Reference Tully, Iacono and Sher2014).

The substantial role of genes in the etiology of internalizing disorders is supported by evidence from twin and adoption studies (e.g., Ehringer, Rhee, Young, Corley, & Hewitt, Reference Ehringer, Rhee, Young, Corley and Hewitt2006; Garcia et al., Reference Garcia, Tully, Tarantino, South, Iacono and McGue2013; Tully et al., Reference Tully, Iacono and McGue2010), and research findings consistently support separate, though positively correlated, genetic loadings on these two subfactors (Kendler, Prescott et al., Reference Kendler, Prescott, Myers and Neale2003). Tully and Iacono (Reference Tully, Iacono and Sher2014) proposed an empirically derived model to explain the hierarchical factor structure of internalizing disorders. In this model, the anhedonia/misery and fear/arousal subfactors derive from distinct, but related, genetically influenced dysregulation liabilities. These liabilities in turn are expressed as discrete, yet overlapping, spectra of behavioral tendencies, cognitive processes, affective experiences, and neurobiological systems (Tully & Iacono, Reference Tully, Iacono and Sher2014). Tully and Iacono (Reference Tully, Iacono and Sher2014) hypothesized that a genetic liability for negative thought processes (e.g., habitual negative attributions and expectations, biased attention and memory, worry, rumination) would contribute to anhedonia/misery symptoms, and a genetic liability to physiological arousal (e.g., heightened psychophysiological responses to aversive stimuli, behavioral inhibition) would provoke fear/arousal pathology.

In the following sections, we describe how these two liabilities may interact with genetic propensities for high empathic sensitivity to amplify vulnerability for internalizing problems. Confirmation of such interactions and their effects would help explain how empathic tendencies, which usually function as strengths (e.g., a high need for affiliation and/or a communal orientation) may, paradoxically, serve as risk factors for internalizing conditions.

Personal distress and interpersonal guilt

Our model proposes that developmental trajectories from empathic tendencies to fear/arousal and misery/distress internalizing problems encompass two intermediate conditions: interpersonally oriented personal distress and guilt, respectively. Personal distress, as defined in the empathy literature, is a maladaptive affective response to negative emotions in others. Unlike compassionate empathic concern, a response oriented toward another person who is in physical or emotional pain, personal distress is a self-focused response accompanied by physiological hyperarousal and behavioral withdrawal (Eisenberg, Reference Eisenberg and Damon1989; Zahn-Waxler & Van Hulle, Reference Zahn-Waxler, Van Hulle, Oakley, Knafo, Madhavan and Wilson2012). Interpersonal guilt, in contrast, is a maladaptive form of cognitive empathy that is driven by excessive and irrational altruistic concerns, such as unreasonable beliefs that one is responsible for alleviating the suffering of others and intense worries about harming others (Oakley, Knafo, & McGrath, Reference Oakley, Knafo, McGrath, Oakley, Knafo, Madhavan and Wilson2012; O'Connor, Berry, Lewis, Mulherin, & Crisostomo, Reference O'Connor, Berry, Lewis, Mulherin, Crisostomo, Farrow and Woodruff2007; O'Connor, Berry, Weiss, & Gilbert, Reference O'Connor, Berry, Weiss and Gilbert2002; Zahn-Waxler & Van Hulle, Reference Zahn-Waxler, Van Hulle, Oakley, Knafo, Madhavan and Wilson2012).

In adults, both personal distress and interpersonal guilt show associations with a range of maladaptive outcomes, some overlapping and others distinct. Research findings link tendencies toward personal distress reactions with a variety of internalizing problems, including anxiety, guilt, and depression in both adults and youths (e.g., O'Connor, Berry, Lewis, & Stiver, Reference O'Connor, Berry, Lewis, Stiver, Oakley, Knafo, Madhavan and Wilson2012; O'Connor et al., Reference O'Connor, Berry, Weiss and Gilbert2002; Schreiter et al., Reference Schreiter, Pijnenborg and aan het Rot2013; Thoma et al., Reference Thoma, Zalewski, von Reventlow, Norra, Juckel and Daum2011; Zahn-Waxler & Van Hulle, Reference Zahn-Waxler, Van Hulle, Oakley, Knafo, Madhavan and Wilson2012). Personal distress also shows significant associations with neuroticism (Mooradian, Davis, & Matzler, Reference Mooradian, Davis and Matzler2011), burnout (Carmel & Glick, Reference Carmel and Glick1996; Klimecki & Singer, Reference Klimecki, Singer, Oakley, Knafo, Madhavan and Wilson2012), and pathological or irrational altruistic behavior (Oakley et al., Reference Oakley, Knafo, McGrath, Oakley, Knafo, Madhavan and Wilson2012). Fewer studies have focused explicitly on interpersonal guilt; however, the construct has been linked in adult samples to loneliness and alienation (Bruno, Lutwak, & Agin, Reference Bruno, Lutwak and Agin2009), as well as to depression and anxiety symptoms (O'Connor, Berry, & Weiss, Reference O'Connor, Berry and Weiss1999).

Putative associations between interpersonal guilt and maladaptive outcomes have yet to be directly examined in youths. However, Zahn-Waxler and colleagues (e.g., Zahn-Waxler, Kochanska, Krupnick, & McKnew, 2009; Zahn-Waxler & Van Hulle, Reference Zahn-Waxler, Van Hulle, Oakley, Knafo, Madhavan and Wilson2012) have proposed that guilt that can arise when failed efforts to care for a chronically depressed parent lead to inaccurate assumptions of responsibility for the parent's unhappiness and that such guilt can in turn contribute to the onset of depression. Such guilt and its negative sequelae may be further magnified if parents actively induce guilt in their children, a parenting strategy observed more commonly in depressed than nondepressed parents (Donatelli, Bybee, & Buka, Reference Donatelli, Bybee and Buka2007; Rakow et al., Reference Rakow, Forehand, McKee, Coffelt, Champion and Fear2009).

Mechanisms underlying the transitions from empathy to personal distress and interpersonal guilt, and subsequently to internalizing disorders, have only begun to be described in the literature. At this point, few studies explicitly examine factors that may skew the normative developmental trajectory for empathy in these maladaptive directions. However, indirect evidence converges to suggest that personal distress, and potentially interpersonal guilt, at least partially reflect dysregulation of the neural systems that subserve typical affective and cognitive empathy development.

Intraindividual moderators: Genetically influenced dysregulation liabilities

Interindividual moderators: Environmental exposure

The extent to which (a) early empathy develops atypically, (b) genetic liabilities result in neurobiological and physiological regulation difficulties, and (c) regulation difficulties lead to personal distress and interpersonal guilt depends on exposure to negative emotions in others, to inadequate parenting practices, and to other environmental variables that exert their greatest influence early in development.

Long-term exposure to a parent's negative affect, particularly when that parent is severely affected by mood difficulties, has a broad range of negative consequences for many, but not all, youths (e.g., Goodman et al., Reference Goodman, Rouse, Connell, Broth, Hall and Heyward2011; Katz, Hammen, & Brennan, Reference Katz, Hammen and Brennan2013). Researchers increasingly recognize that the likelihood of negative outcomes for children of depressed parents depends on a complex matrix of interacting factors, both genetic and environmental (e.g., Harold et al., Reference Harold, Rice, Hay, Boivin, van den Bree and Thapar2011; Kerr et al., Reference Kerr, Leve, Harold, Natsuaki, Neiderhiser and Shaw2013). Consistent with this idea, we suggest that children with strong proclivities toward empathic sensitivity who also carry genetic liabilities for physiological hyperarousal and/or negative thinking are at particularly amplified risk for maladaptive empathy development when they live for protracted periods in the context of parent depression.

Researchers have long been interested in the association between maternal depression and child empathy; Zahn-Waxler and colleagues, in particular, have articulated sophisticated models of the ways in which these variables dynamically interact, particularly for girls (Radke-Yarrow, Zahn-Waxler, Richardson, & Susman, Reference Radke-Yarrow, Zahn-Waxler, Richardson and Susman1994; Zahn-Waxler, Kochanska, Krupnick, McKnew, Reference Zahn-Waxler, Kochanska, Krupnick and McKnew1990). Despite this enduring interest, few studies to date have examined the idea that exposure to parent negative emotion, such as that observed in the context of depression, interacts with child genetic liabilities to predict the developmental course for empathy and related characteristics. One investigation found that the homozygous GG variant of the OXTR gene was associated with both low salivary oxytocin and maternal depression, which in turn predicted child internalizing disorders (Apter-Levy, Feldman, Vakart, Ebstein, & Feldman, Reference Apter-Levy, Feldman, Vakart, Ebstein and Feldman2013). Further, rates of offspring psychopathology were markedly lower among mothers who carried the GA or AA allele variants. Clarifying how this and other relevant genes, particularly varied 5-HTTLPR and AVPR1 polymorphisms contribute in both independent and interactive ways to modulate the parent affect/child empathy association, as well as the function of neural structures that support relevant processes, is a critically important line of research that is poised to burgeon in coming years.

It will also be of value to explore the role of parenting behavior in the association between parent affect and child empathy outcomes. A small body of evidence supports a link between effective parenting (which studies have defined to include such practices as fostering secure attachment, providing support, and encouraging emotional expression) and healthy empathic development (e.g., Kestenbaum, Farber, & Sroufe, Reference Kestenbaum, Farber and Sroufe1989; Soenens, Duriez, Vansteenkiste, & Goossens, Reference Soenens, Duriez, Vansteenkiste and Goossens2007; Taylor, Eisenberg, Spinrad, Eggum, & Sulik, Reference Taylor, Eisenberg, Spinrad, Eggum and Sulik2013). Some evidence suggests that this association may not always be direct; Yoo, Feng, and Day (Reference Yoo, Feng and Day2013), for example, found that parent–child connectedness mediated relationships between parenting behavior and empathy in an adolescent sample. Regardless of the path, positive parenting practices appear to contribute to adaptive empathy in offspring.

It follows logically, then, that maladaptive parenting behaviors that commonly occur in the context of negative parent affect, such as failure to facilitate emotion regulation, insensitivity to children's emotional cues, excessive control, and lack of warmth, may skew children's empathic development toward problematic outcomes (Hastings, Zahn-Waxler, Robinson, Usher, & Bridges, Reference Hastings, Zahn-Waxler, Robinson, Usher and Bridges2000; Kanat-Maymon & Assor, Reference Kanat-Maymon and Assor2010; McGrath & Zook, Reference McGrath and Zook2011; Valiente & Eisenberg, Reference Valiente, Eisenberg, Snyder, Simpson and Hughes2006; van der Mark, van IJzendoorn, & Bakermans-Kranenburg, Reference van der Mark, van IJzendoorn and Bakermans-Kranenburg2002). We propose that the effects of inadequate parenting on the developmental trajectory for empathy will be particularly potent in youths who carry a genetically mediated neurobiological liability for empathic sensitivity, especially those who also carry liabilities for physiological hyperarousal and/or negative thinking.

Evidence of such Gene × Environment interactions have been detected in recent research focused on the development of empathy and prosocial behavior. For instance, Knafo, Israel, and Ebstein (Reference Knafo, Israel and Ebstein2011), in a large longitudinal twin study, examined the possibility that preschoolers who carry different polymorphisms of the dopamine receptor D4 (DRD4-III) gene, which has been found in other contexts to influence child responsiveness to both positive and negative parenting behaviors, would exhibit differential sensitivity to parenting influences. No significant direct associations between parenting and prosocial behavior were observed. However, a significant Gene × Parenting interaction emerged, such that maternal positivity, negativity, and unexplained punishment correlated significantly with children's prosocial behavior only among children who carried the DRD4-7 allele. In this subgroup of youths, maternal positivity related meaningfully and positively to mother-rated prosocial behavior, and maternal tendency to use unexplained punishment related positively to self-initiated prosocial behavior in the experimental context.

In another twin study, Shikishima, Hiraishi, Yamagata, and Ando (Reference Shikishima, Hiraishi, Yamagata and Ando2011) obtained evidence of a different, but related, gene by parenting interaction. In this study of adolescent and adult twins, the association between parental warmth and child empathy was driven largely by genetic factors, with no significant effects detected for shared environmental factors for the sample as a whole. However, tests of a Gene × Environment interaction model yielded significant findings, indicating that shared family factors linked to parental warmth were significantly associated with indices of empathy in offspring whose parents demonstrated very high or very low levels of warmth.

Comparable findings emerged from at least one study focused on interactions between parenting and physiological, rather than genetic, characteristics as predictors of empathic behavior. In a sample of mother–adolescent dyads, Diamond, Fagundes, and Butterworth (Reference Diamond, Fagundes and Butterworth2012) found that attachment style interacted with vagal tone, which provides an index of emotion regulation capacities, to predict adolescents' empathic responsiveness to their mothers in the context of emotional interactions. Empathic responsiveness was lowest among youths with high levels of attachment insecurity and low levels of vagal tone.

Although research has revealed evidence of interactions between parenting and broad profiles, such as temperament, that are associated with distinct patterns of psychophysiological functioning, as predictors of externalizing (e.g., Bradley & Corwyn, Reference Bradley and Corwyn2008) and internalizing (e.g., Kiff, Lengua, & Bush, Reference Kiff, Lengua and Bush2011) disorders, such findings are weaker with regard to child empathic behavior and experiences, such as personal distress or interpersonal guilt. For example, in a study of child temperament and maternal responsivity as predictors of child empathic behavior, Spinrad and Stifter (Reference Spinrad and Stifter2006) found main effects of both constructs, but minimal evidence of interactions between them as predictors of personal distress or empathic behavior. Viewed in the context of recent research that highlights ways in which more specific biological features, such as gene polymorphisms, interact with parenting to predict empathy-relevant outcomes, these findings suggest that it may be more fruitful to focus research into biological/environmental interactions on more narrowly defined factors.