Major depression is a mental illness that is characterized by a persistent and significant low mood and is caused by a combination of environmental and hereditary causes, which is widespread, expensive, damaging, and related to heightened risk of suicide. It is a major worldwide public health concern (Marwaha et al., 2023). Rates of depression in youth have sharply increased over the last 10 years, which is concerning because adolescence is a time of rapid changes in social, emotional, and cognitive development, as well as a period of significant life transitions. The risk of recurrence of depression, co-morbidity with other mental illnesses, and more severe and prolonged damages in social, educational, and occupational functioning are the consequences related to depression in young people (Thapar et al., 2022). It has been observed that depression and anxiety are increasingly common among the youth, and they frequently occur together and have similar risk factors (Craske and Waters, 2005). The hypothesis of a shared neurobiological malfunction is supported by the fact that anxiety and depression have many similar symptoms and may respond to comparable therapies. However, the precise neurobiological mechanisms underlying depression and anxiety are not yet fully understood (Nutt et al., 2002). Having more knowledge about depression can help in the development of treatments for other mental health issues.

Fear learning dysfunction is thought to be implicated in the emergence and persistence of an array of psychiatric issues, including not only anxiety disorders (Milad et al., 2014; Otto et al., 2014) and posttraumatic stress disorder (PTSD) (Wicking et al., 2016) but also depression (Sandi and Richter-Levin, 2009). Moreover, studies on extinction learning in depression have yielded conflicting results, one study showed a deficiency (Dibbets et al., 2015), and the other study exhibited enhancement (Kuhn et al., 2014). Therefore, it is believed that maladaptive social anxiety and fear are linked to depression. The ability to recognize and respond to potential danger is essential for survival; however, when this process becomes impaired and when one experiences an abnormal fear response to harmless situations, anxiety disorders may develop. Numerous studies have been conducted for understanding the behavioral, experiential, and neural components of both adaptive and maladaptive fear-learning processes in both animals and humans. Pavlovian fear conditioning is a prevalent model for studying both fear and anxiety, and it continues to influence modern explanations of clinical anxiety issues. Despite its widespread usage in studies of both animals and humans, the neurological basis of fear conditioning is not yet fully comprehended. Investigating the relationship between fear learning and depression can help us gain insights into the associative learning processes of humans, which could lead to improved comprehension and treatment of these mental health issues. Conditioning, extinction, and reinstatement are essential elements of animal adaptation, and they are also closely associated with mental disorders such as PTSD, anxiety, depression, and addictions (Mattera et al., 2020). Fear conditioning and fear extinction learning are fundamental components of models that explain the development of anxiety disorders and the reduction of pathological fear during exposure-based treatments (Shankman and Klein, 2003). Studies conducted on rodent models of learned helplessness have revealed a correlation between impaired fear extinction and heightened depression-like behaviors (Shumake et al., 2005). It was also found that depressed patients had trouble acquiring conditioned fear. Furthermore, following successful fear learning, they displayed impaired extinction of conditioned fear responses (Wurst et al., 2021). Depression increases the risk of enduring persistent fear, which is the characteristic of anxiety disorders, and similarly, persistent fear can contribute to depression.

Fear conditioning is a form of associative learning with adaptive and self-protective functions (Pearce and Bouton, 2001) to predict threat through the association of initially neutral stimuli (conditioned stimuli, CS) with aversive outcomes (unconditioned stimuli, US) (Ojala and Bach, 2020), and the corresponding laboratory model has been used extensively as a classical experimental paradigm to quantitatively measure associative learning (Lonsdorf et al., 2017). Classical formal learning theory, such as the Rescorla–Wagner (R-W) and Pearce–Hall (P-H) models, which explain how the CS and US signals are integrated algebraically to connect cues with aversive events, can provide a good description of computational principles behind direct fear conditioning (LePelley and McLaren, 2004; Lindström et al., 2018). The R-W model is a learning model that is driven by the prediction error, that is, the difference between the expected and actual outcomes of a conditioning trial (Miller et al., 1995; Li and McNally, 2014). In other words, the R-W model suggests that the formation of an association depends on deviations or “errors” between expectations and actuality. However, the P-H model is an associability gated learning model. Organisms learn cue–reinforcer associations by applying a quantity called associability, which is determined by the dynamics of trial-wise prediction errors (Herry and Johansen, 2014). To be more precise, associability is based on prediction errors from prior trials involving the same cue, which indicate the degree to which each cue has been accompanied by a surprise in the past. Learning rate is a constant in the R-W model, but in the P-H model, a cue's associability gates learning rate dynamically, decelerating learning to reliable predictors and accelerating learning to cues that are poor predictors (Li et al., 2011). However, the hybrid model is shown to have better power to comprehend a variety of experimental findings than the previous single model (Le Pelley, 2004).

Recent studies have revealed that people with depression demonstrate improved extinction learning in comparison to healthy controls (Kuhn et al., 2014), as well as enhanced differential acquisition of fear responses toward the CS+ (Nissen et al., 2010). Despite contrary reports (Otto et al., 2014; Waters et al., 2014; Wurst et al., 2021) or null findings (Dibbets et al., 2015), most studies have been conducted with relatively small sample sizes, making it necessary to conduct more research to gain a better insight into the effects of depression on fear acquisition and extinction learning. The neuroplasticity hypothesis of depression postulates that a malfunction in neural plasticity, which is responsible for learning, memory, and adaptive behavior, is the primary source of the disorder's clinical manifestations (Spedding et al., 2003; Normann et al., 2007; Kuhn et al., 2014). Due to the attentional and memory deficits associated with mood disorders, we predicted that the people with depressive symptoms would show poor acquisition of the conditioned fear response than those without depressive symptoms. This is because differential conditioning necessitates the learning and subsequent discrimination of the meaning of both a CS+ and CS–.

The current study explored basic forms of learning to further test the neuroplasticity hypothesis of depressive symptoms in associative learning. The strengths of the study include a well-defined sample of participants with depressive symptoms and participants without depressive symptoms matched according to sex, age, and educational level. Based on the results of the skin conductance responses, we found that the subjects with depressive symptoms had deficits in fear acquisition. Furthermore, we used associability as estimated by a hybrid model to investigate the effect of depressive symptoms on the associative learning process. We found that skin conductance responses of subjects without depressive symptoms, who successfully acquired differential conditioned fear, were positively correlated with associability, as estimated by a hybrid model.

During a typical differential fear conditioning procedure, subjects learn to form an association between CS+ and US. The CS+ becomes a predictor of the US and will itself evoke the associated fear, known as the conditioned response (CR), which consists of motor, neuroendocrine, autonomic, and other components (Krabbe et al., 2018). Differential skin conductance and subjective responses to the CS+ than the CS– reflect that the subject has successfully acquired the conditioned fear responses (Schiller et al., 2010; Rainer et al., 2020). In the present study, the results of the SCR showed that the subjects with depressive symptoms did not successfully acquire conditioned fear, although subjective evaluations showed that they exhibited higher US expectancy, nervousness, and likeability ratings to CS+. These findings were anticipated because the dissociation between the subjective evaluations and SCR results was common in previous studies (Britton et al., 2013; Shechner et al., 2015, 2018; McLaughlin et al., 2016). McLaughlin et al. also measured the SCR continuously and collected subjective evaluations after each phase. They found disturbances in the acquisition of conditioned fear in maltreated children based on the SCR results. Maltreated children exhibited blunted SCR to the CS+ and failed to exhibit differential SCR to the CS+ vs. CS- during early conditioning relative to healthy controls. However, all subjects reported greater fear of CS+ than CS– after conditioning, regardless of maltreatment (McLaughlin et al., 2016). Unconditional stimulation anticipation evaluations and contingency awareness tests suggested high and similar cognitive representations of the association between the CS and the US in depression patients and healthy controls (Blair et al., 2001; Phelps and LeDoux, 2005; Vansteenwegen et al., 2008). In the present study, both groups showed no difference in the SCR between CS+ and CS– at the end of the extinction phase. However, the likeability ratings of CS+ did not increase in the group with depressive symptoms after extinction. These results were consistent with those of Shechner et al. (2018), who indicated successful extinction for the sample, as indexed by the SCR. However, subjective evaluations showed that the difference between CS+ and CS– still exists. Subjective evaluations may represent a more cognitive expression of threat anticipation, reflecting declarative knowledge of stimulus contingencies. The SCR can track emotional arousal change resulting from exposure to the stimulus, even upon subliminal conditions (Raio et al., 2012).

Though anxiety and depression frequently co-occur, only a few studies have been conducted on fear conditioning in depression, with inconclusive results. Our results were similar to those of Otto et al. (2014) and Wurst et al. (2021), who found significantly attenuated differential fear conditioning among depression patients. In other words, compared to the healthy individuals, patient samples were conditioned less well. Additionally, one study on fear conditioning in offspring susceptible to affective disorders due to maternal psychopathology found that the offspring of depressed mothers showed diminished responses to aversive cues during the acquisition phase (Waters et al., 2014). However, Nissen et al. (2010) conducted a classical differential fear conditioning paradigm to explore the fear acquisition of major depressive disorder patients and healthy controls, wherein a mild electric shock was used as the unconditioned stimulus (US) and four geometric shapes served as the conditioned stimulus (CS). However, the reduced differential skin conductance responses in the group with depressive symptoms were not in line with the results of that study, which showed that MDD patients had a greater skin conductance response (SCR) to the CS+ than to the CS–, while the control participants did not demonstrate differential SCRs to the CSs. Although the study by Nissen et al. showed that patients with major depressive disorder successfully acquired conditioned fear, the anxiety levels of the MDD patients were not taken into account in that study. Given that the MDD patients are often accompanied by anxiety symptoms, it is possible that anxiety contributed to the acquisition of conditioned fear in the MDD patients. A meta-analysis of fear conditioning in anxious and non-anxious adolescents showed that successful acquisition of conditioned fear occurred in both groups, demonstrating comparable differential learning, but anxious individuals had higher fear responses to both CS+ and CS– than non-anxious individuals (Dvir et al., 2019). These results are consistent with those of two previous meta-analyses of fear acquisition in anxious and non-anxious adults (Lissek et al., 2005; Duits et al., 2015), suggesting some combination of stronger excitatory processes to the CS+ and weaker inhibitory processes to the CS– in anxiety. In the present study, although there were differences in anxiety levels between the two groups, they were within the normal range. Thus, our results better eliminated the influence of anxiety symptoms and strongly suggested that impaired conditioned fear acquisition was due to depressive symptoms.

Previous results from skin conductance measurement revealed a lessened fear acquisition among the patient samples, particularly in individuals with depression and posttraumatic stress disorder, indicating an attenuated fear conditioning (Otto et al., 2014). Studies have indicated that individuals suffering from depression, particularly endogenous depression, tend to have lower levels of skin conductance and a higher percentage of SCR non-responders to unpleasant auditory stimuli than healthy persons (Lader and Wing, 1964; Mirkin and Coppen, 1980). The results of this study suggest that people with depression may not be as affected by external stimulation (Mirkin and Coppen, 1980), likely because of the wide-reaching and persistent nature of depressive disorders weakening the human defensive system, thereby decreasing psychophysiological reactivity to emotional stimuli (McTeague et al., 2012). Previous research has indicated that depression is associated with a reduction in fear-potentiated startle and pleasure-inhibited startle, leading to a generally flattened affect–startle pattern (Vaidyanathan et al., 2012). This could explain why the depression group in the current study exhibited similar skin conductance responses to CS+ and CS–. Consequently, these studies sought to investigate whether a weakened psychophysiological response is a risk marker that develops before the onset of depression in those predisposed to the disorder.

Rescorla–Wagner's empirically well-supported notion of error-driven value update was incorporated into the Pearce–Hall associability model, resulting in the hybrid model (Li et al., 2011). These models are usually fitted to the trial-by-trial skin conductance responses in threat learning (Pearce and Hall, 1980; LePelley and McLaren, 2004). We found that the skin conductance responses of subjects without depressive symptoms who successfully acquired differential conditioned fear were positively correlated with associability, as estimated by a hybrid model. It has been proposed that complex fear-learning procedures might be more effective than the traditional differential fear-conditioning paradigm in detecting individual differences related to susceptibility to anxiety disorders. This is based on the idea that those at risk of developing an anxiety disorder may have an abnormal fear response to a real-life conditioning event (Arnaudova et al., 2013). Consequently, further exploration of complex fear learning may be critical for gaining insight into the general principles of associative learning in humans, as real-world Pavlovian associations are likely more intricate than those explored in laboratory experiments.

Our study revealed that those with self-reported depressive symptoms may have impairments in associative learning. The effect of depressive symptoms on fear conditioning is distinct from that of anxiety symptoms. Different outcomes of differential fear conditioning may aid in distinguishing depression from anxiety. Understanding the physiological differences that may be the source of symptomatology is vital for understanding the varied pathophysiology of these disorders.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by the Ethics Committee of the Tianjin Medical University. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

R-HL: Data curation, Formal analysis, Writing—original draft, Writing—review & editing, Investigation. FS: Formal analysis, Software, Writing—review & editing, Data curation. X-YZ: Writing—review & editing, Investigation. T-HC: Writing—review & editing, Investigation. JL: Writing—review & editing, Investigation. Y-XX: Funding acquisition, Supervision, Writing—review & editing, Conceptualization, Project administration. G-DH: Data curation, Software, Writing—review & editing, Project administration. J-LY: Writing—review & editing, Project administration.