The present study investigated fear acquisition and generalization in children and adolescents with clinically relevant anxiety and matched healthy controls (HC). The goal was to examine discrimination conditioning and (over)generalization in healthy vs. anxious children and adolescents to elucidate differences in fear learning and generalization between children and adolescents with anxiety disorders and HC. Therefore, the ratings of valence, arousal, and US expectancy were measured, hypothesizing that patients with clinically relevant anxiety would show overall higher fear responses during the acquisition and generalization as well as overgeneralization as indicated by more linear compared to quadratic trends when compared to HC.

These hypotheses could be partly confirmed: We found quantitative, not qualitative differences between patients and HC. In other words, patients with anxiety disorders had generally higher rating scores compared to HC, but they did not show more overgeneralization of conditioned fear when compared to HC. In the following, these results will be discussed in more detail regarding previous literature. In general, findings in some aspects are in line with results of similar experiments done in adults, but also in children [e.g., 33, 34], thereby eliciting the question if overgeneralization could be considered as risk markers for anxiety disorders or vice versa.

In line with previous experiments using the “screaming lady” paradigm in children and adults with and without clinically relevant anxiety [e.g. 25, 26, 33, 35], successful fear conditioning as well as generalization could be detected in both, patients diagnosed with an anxiety disorder and HC. Further, we found overall group differences in the ratings: children and adolescents with clinically relevant anxiety showed generally higher arousal and US expectancy ratings and more negative valence ratings to all presented stimuli during generalization phases when compared to HC. Such heightened responding to all stimuli has been attributed to associative processes of elevated responding to threat cues and poor inhibitory responding to safety cues, and non-associative processes of sensitization and habituation (see [12] for a review). Moreover, these group differences were already found during the (pre-) acquisition phases indicating a general hyper-arousal of patients with pathological anxiety. Of special interest in this context is the fact that almost half of our patient group were diagnosed with a social anxiety disorder (SAD). Thus, the general hyper-arousal of the patients might have been triggered by the exposure to disorder-relevant face stimuli. Such general hyper-arousal constitutes a diagnostic criterion of SAD in the DSM-V (American Psychiatric Association, 2013), and is in line with previous studies suggesting a generally biased processing of faces in social anxiety disorders [33, 36]. Hence, our first hypothesis, which expected that patients diagnosed with an anxiety disorder would show different ratings (higher arousal and US expectancy ratings and lower valence ratings, respectively) when compared to HC, was confirmed.

Contrary to previous findings demonstrating overgeneralization to be a diagnostic marker of some anxiety disorders in adults [15,16,17] and in contrast to our further hypothesis, children and adolescents with pathological anxiety compared to HC demonstrated no maladaptive overgeneralization of fear as indicated by more linear compared to quadratic trends: We found no significant differences in the generalization gradients between the groups as indicated by no significant stimulus type x group interaction effects, but general group differences as indicated by main effects of group in the arousal, valence, and US expectancy ratings: Anxious individuals reported to perceive all faces as less pleasant and more arousing, and, relative to the control group, they overestimated the pairing of GSs/ CS− with the US, although those stimuli were never reinforced.

These findings are congruent with previous findings on, e.g., face perception in social anxiety disorder (SAD) done in adults, which detected those individuals with social anxiety rated angry faces to be more negative [37] and more arousing [38], and rated happy faces as less pleasant [39]; further, socially anxious individuals showed generally enhanced US expectancy ratings when compared to controls [see 40 for a review]. This is consistent with prior studies indicating that patients with social phobia have bad stimulus discrimination skills and therefore show enhanced US expectancy ratings and fear responses to safety cues [33, 41, 42].

In line with our results, there are some further studies, which did not find clear overgeneralization in anxiety disorders [18, 19], but generally higher fear ratings in patients compared to HC [33]. Additionally, the present study investigated children and adolescents, whereas previous studies, which demonstrated overgeneralization in patients with anxiety disorders, investigated adult patients [15,16,17]. Relevant here is that the median onset age of anxiety disorders is about 13 years [43], with chronic manifestations and comorbidities in adulthood. Thus, healthy adults and adolescents might be those, who had not yet developed a manifest anxiety disorder during this critical time window, but which of course could possibly change later in life; thus, the adolescents in our sample may have been “healthier” than the participating children, who may still be at larger risk to develop pathological fear possibly soon and clearly before their now already adolescent counterparts. The latter would have led to their exclusion from the healthy control group (and to their inclusion into the experimental group). This means in other words that the adolescents in the control group managed to stay healthy for a longer time from their birth on than the healthy children (also due to age/time of birth), who may be prone to develop an anxiety disorder eventually quite soon considering the mean onset of about 13 years for anxiety disorders, which states a clear risk factor. As a study by Reinhard et al. [44] demonstrated, the subjective and physiological fear responses in children and adolescents aged 8 to 17 years were generally lower with increasing age irrespective of the stimulus quality. Additionally, stimulus discrimination improved with increasing age paralleled by reduced overgeneralization in older individuals. Thus, there were generally higher fear ratings in patients with AD compared to HC as well as in healthy younger children compared to healthy adolescents. Therefore, it might be important to check whether there were differences between patients with AD and healthy controls when controlling for age. Moreover, patients in the present study were very heterogeneous: Patients with any anxiety classified as pathological anxiety as well as with and without medication were included in the study. Possibly, overgeneralization is a feature of a subset of specific anxiety disorders. The experiments by Lissek et al., for instance, found overgeneralization e.g., in patients with panic disorder (PD) [15], generalized anxiety disorder (GAD) [17] and post-traumatic stress disorder (PTSD) [16]. On the other hand, a study by Ahrens et al. [33], which investigated patients with a social anxiety disorder (SAD), found that with respect to explicit ratings SAD patients compared to HC do not seem to be characterized by strong overgeneralization but discrepancies in fear responses to both conditioned and generalized threat stimuli. As mentioned before, most patients in our sample were diagnosed with SAD. Further, due to the fact, that we investigated children and adolescents rather than adults, there were hardly any participants with PD/GAD/PTSD in our sample. Thus, this fits to previous studies showing that due to ratings there were parallel shifted gradients rather than qualitative differences in fear generalization between SAD patients and HC [33]. Additionally, in the experiments of Lissek and colleagues [15,16,17], shocks were used as aversive stimulus, whereas in our experiment, a loud scream was used, and further, different dependent variables (e.g., startle response vs. SCR/ratings), which possibly reduces the direct comparability of the results.

Despite its strengths, the current study was not without limitations. First, the sample size was relatively small and we were unable to investigate possible differences between patients with different types of anxiety disorders. Moreover, due to the relatively small sample, we did our analyses not separated by sex, but differences relating to sex could be quite interesting, considering that the prevalence of anxiety disorders is supposed to be significantly higher for females than for males [45, 46, ICD 10: WHO 2004]. Further, participants received different medication, which we have not checked for as a covariate. Additionally, we did not control for comorbid disorders, and thus, we cannot rule out that results were distorted by other psychiatric disorders. A study in adults, for instance, investigated the influence of comorbid depression on social phobia and demonstrated that patients without depression showed defensive hyper-reactivity during social threat imagery, while patients with comorbid depression showed attenuated reactions [47]. This could reflect depression-associated psychomotor retardation and behavioral inhibition. On the other side, most of the other studies on fear generalization included also individuals with comorbid disorders, but mostly with other anxiety disorders or depression [15, 17]. Second, we analyzed group differences in a categorical, not dimensional manner leaving the possibility out of consideration, that even supposedly healthy participants and/or subclinical groups at risk for developing anxiety disorders could have high anxiety scores when measured by, e.g., the State-Trait Anxiety Inventory for Children – Trait version (STAIC-T [30]) [see e.g., 48,49,50]. Third, other factors accounting for anxiety disorders as well as resilience factors were disregarded here. For instance, additional risk factors, such as child temperament (e.g., neuroticism), parental anxiety disorder, parenting style, and/or negative life events [51, 52] might play an important role but were not assessed in this study. Moreover, of note is the fact, that our study has a cross-sectional design. Thus, the question remains still open, if quantitative/qualitative differences during the aversive conditioning and generalization found in patients with an anxiety disorder are really a risk marker or rather a result of the disorder. Thus, longitudinal follow-up studies are required to answer this still-open question. This question is highly important, especially with respect to preventive and therapeutic approaches. Since there are therapeutic interventions accessible to date, but several children do not improve, new markers could lead the way to a “personalized medicine approach “.

In sum, the results of the present study indicate that patients with clinically relevant anxiety differed from healthy control children in terms of responding during aversive conditioning and generalization. The differences, however, were more quantitative, not qualitative. In other words, the patients showed generally higher fear-relevant ratings when compared to HC, suggesting elevated responding to the threat cue and impaired inhibition of responses to the safety cues. Thus, according to our results, overgeneralization of fear does not seem to be a diagnostic marker of anxiety disorders in children and adolescents. There is a clear need for replication, especially in bigger and better-classified samples. Moreover, there is a clear need for longitudinal studies to answer the question if (over)generalization patterns of patients with anxiety disorders are a risk factor for any anxiety disorder or if certain patient groups develop such generalization pattern as part of the disorder. As the present study demonstrated that the fear generalization pattern of children is not necessarily comparable with the generalization pattern of adults, developmental factors need to be considered when analyzing fear learning and generalization in anxiety disorders.