Similar to previous studies [22, 23], we found that HIV-infected patients performed worse than HC in recognizing the facial emotion of fear. This deficit does not seem related to severity of the cognitive impairment; in fact, patients with ANI performed as accurately as patients who had no documented cognitive deficit.
It has been suggested that specific deficits in the recognition of different categories of facial emotions may reflect task difficulty factors . This is also supported by cross-cultural studies in healthy subjects showing that accuracy in recognizing happiness is high (94% correct responses on emotion recognition tasks) and fear is low (70%) [41, 42]. On the other hand, there are considerable evidence that the correct recognition of facial expression of fear depends on specific neural structures, that seems to have a critical role in mediating the autonomic and behavioural responses associated with this emotion [1, 43].
In agreement with the results of cross-cultural studies [41, 42], in our study both patients and controls performed worse when they had to recognize fear than the other emotions, confirming that this emotion might be more difficult to recognize than the others. On the other hand, fear was also the only emotion on which patients performed significantly worse than controls. This would suggest that impairment of neural substrates that are supposed to be specifically involved in HIV pathology, could concur in fear recognition deficit in addition to the effects of task difficulty. Previous studies  suggest in fact, that fear recognition abnormalities in HIV may be due to a disruption of the broader neural network involved in emotion recognition which depends on the integrity of frontal system. Unfortunately, we do not have direct evidence of the involvement of such substrates in our population. However, since such an impairment is currently demonstrated  we cannot exclude that the low performance in fear recognition obtained by our patients could be due at least in part to the effect of infection.
If this were true, demonstration of the presence of impaired fear recognition should be considered an early marker of cognitive impairment in HIV population. Studying fear recognition in patients with other brain pathologies that spare the frontal regions as well as neuroimaging studies [11, 44] on HIV + patients could contribute to clarify this issue.
Multivariate analysis demonstrated an independent relationship between severity of cognitive impairment and score on recognizing the emotion of happiness, that is, patients with a higher number of pathological scores on the neuropsychological examination were less accurate in recognizing happiness. Moreover we observed an association between the global emotion recognition score and education level, finding that did not emerge in previous studies . These results confirm that recognizing emotions requires the integrity of “high level” cognitive abilities, as already reported in patients with neurodegenerative diseases (see  for a review).
The emotion happiness was also found in association with HIV-related variables. In particular, happiness was independently and inversely associated with past AIDS-defining events.
The association between recognition of happiness and general neurocognitive impairment as well as the association between recognition of happiness and past AIDS-defining events, could lead us to hypothesize that a deficit in recognizing this emotion might emerge only in subjects in more severe stages of HIV pathology in agreement with a recent study  demonstrating that the ability to discriminate between levels of happiness intensity on facial expression was specifically altered in HIV patients with impaired neurocognitive performance.
We acknowledge that our study might have some limitations because uncontrolled biases can occur in cross-sectional surveys performed in routine clinical practice. Furthermore, we cannot exclude a possible confounding effect of different IQ on emotion recognition, although there is no reason to assume a different distribution of IQ value in patients and controls. Moreover, although not significant, the performance of HC group in recognition of facial expression of surprise was worse than HIV patients; since the effect size for this emotion was moderate, we cannot exclude potential power issues. Anyway, both groups obtained high scores, so data observed might be attributed to a randomness. At last, we did not consider either the psychological and social implications of HIV infection and their possible impact on emotion recognition ability  or the psychological premorbid characteristics.