Lindquist et al.'s meta-analysis convincingly disproves the strong locationist account of the brain basis of emotion. Yet, one needs look no further than the myriad different presentations of anxiety disorders to see that the experience of emotions such as fear is highly idiosyncratic. As behaviour originates in the brain, this idiosyncrasy must be reflected at the neural level, and we endorse Lindquist et al.'s emphasis on context and prior learning in shaping emotional experience. This process, referred to by Lindquist et al. as conceptualisation, has been described extensively in cognitive models of psychiatric disorder in terms of schema (e.g., Beck & Emery Reference Beck and Emery1985). However, the universality of facial emotion recognition and the cross-cultural ubiquity of anxiety disorders, suggest that emotional hardware is more important than Lindquist et al. suggest.

We agree that the basic emotions as described by Lindquist et al. are unlikely to represent the most basic psychological emotional processes in the brain. However, by combining facial emotion recognition with subjective emotion experience, Lindquist et al. understate the importance of biological hardware in facial recognition. There is strong evidence that facial expressions are universal social signals, and reading emotion expressions in others is conserved across cultures (Ekman Reference Ekman1973), albeit with some cultural constraints (Elfenbein & Ambady Reference Elfenbein and Ambady2002). Evidence suggests that humans are biologically prepared for facial expression recognition. Infants aged 5–7 months can reliably distinguish between most facial emotion expressions, and are beginning to show adult-like attentional preferences for threat emotions such as fear (for review, see Leppänen & Nelson Reference Leppänen and Nelson2009). Around this age infants also show increased startle responses to acoustic probes in the presence of angry, relative to happy, faces (Balaban Reference Balaban1995).

Research by Schofield et al. (Reference Schofield, Coles and Gibb2007) replicated by ourselves (i.e., Button et al., unpublished results) has found that social anxiety is not associated with differences in recognising facial expressions but is associated with differences in attributions of personal cost to those same expressions. Similar results have been found across anxiety disorders; emotion recognition does not vary as a function of anxiety, whereas other processes, such as attention, are selectively enhanced for threat emotions such as anger (for review, see Bar-Haim et al. Reference Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg and van IJzendoorn2007). These findings suggest that facial expression recognition is relatively robust to the influences of anxiety schema. Furthermore, they illustrate how different patterns of neural responses may arise during simple face emotion viewing tasks due to processes such as attribution and attention.

The universality of the clinical syndrome of anxiety disorders (e.g., Horwath & Weissman Reference Horwath and Weissman2000) suggests that emotional hardware is also important in subjective emotion experience. Vulnerability to anxiety disorders is heritable (Hettema et al. Reference Hettema, Neale and Kendler2001). Evidence that non-phobic individuals report as many aversive experiences with a fear-stimulus as do individuals who are phobic of that stimulus (Ehlers et al. Reference Ehlers, Hofmann, Herda and Roth1994; Merckelbach et al. Reference Merckelbach, Arntz, Arrindell and deJong1992) suggests that factors other than contextual learning are important in fear responses. However, compared to the relative robustness of facial emotion recognition, subjective emotions are strongly influenced by anxiety schema, as they have the broader function of guiding perceptions of, and responses to, the environment (Cosmides & Tooby Reference Cosmides, Tooby, Lewis and Haviland-Jonesm2000; Damasio Reference Damasio1996).

Variations in the physiological responses which characterise different anxiety disorders provide evidence for the influence of anxiety schema on subjective emotion experience. Specific phobias are characterised by relatively normal baseline autonomic activity, with strong elevations in autonomic activity in the presence of the phobic situation (e.g., Hofmann et al. Reference Hofmann, Ehlers, Newman and Roth1995). This pattern of activation is consistent with the specificity of the phobic schema (e.g., spider fears) to the phobic stimulus (e.g., spiders).

By contrast, generalised anxiety and generalised social anxiety are associated with elaborate schemas involving excessive worry about worry and fears of social ridicule and rejection, respectively. As such, a much wider array of stimuli and ruminative thoughts can trigger anxiety, which is reflected in a generally elevated baseline arousal observed in the hypothalamic–pituitary–adrenal axis and sympathetic adrenal medullary activation (for review, see Craske Reference Craske2003). This increased baseline arousal is associated with a hypervigilance for threat. However, responses to acute stressors in generalised anxiety do not reliably differ from controls, and chronic worriers actually show reduced variability in heart rate and skin conductance during psychological stress (Hoehnsaric et al. Reference Hoehnsaric, Mcleod and Zimmerli1989; Reference Hoehnsaric, Mcleod and Hipsley1995).

McNeil et al. (Reference Mcneil, Vrana, Melamed, Cuthbert and Lang1993) found that shame and embarrassment result in decreased cardiovascular activity, suggesting that during times of social fear the parasympathetic shame response competes with the sympathetic fearful response, resulting in an attenuated heart rate increases in persons with social phobias. These conflicting processes of shame and fear are likely to be reflected in patterns of brain activity, illustrating the complexity of subjective emotion experience.

Recognising facial expressions is highly conserved, and the influence of anxiety schema on this process is relatively subtle. By contrast, the experience of anxiety and fear is highly idiosyncratic, reflecting much greater influence of contextual learning and belief systems, referred to as schema. Given the idiosyncrasy of fear experiences, reflected in the diversity of physiological responses outlined above, we agree with Lindquist et al. that subjective emotional experience is likely to be highly socially constructed. However, the evidence does not support Lindquist et al.'s claim for the social construction of basic emotion recognition, suggesting instead that humans are biologically prepared for facial expression recognition.

The strong locationist model cannot account for the differences observed in emotion recognition and subjective experience. Neither can it account for the idiosyncrasy of emotional experience. We agree with Lindquist et al. on the need to identify the basic psychological processes underpinning emotion. However, in attempting to encompass all the complexity of human emotion into a single model, the result lacks predictive value. If we accept that subjective emotion is socially constructed, then models which address the question of what emotions are and how they are represented in the brain are unhelpful. A more fruitful approach to establishing the brain basis of emotion (and cognition in general) is perhaps to ask what a given brain region does, and why.