Fear conditioning is an important experimental paradigm that seeks to elicit the learning and memory processes underlying anxiety and trauma‐related disorders (Bouton et al., 2020; Craske et al., 2014; Vervliet et al., 2013; Zuj et al., 2016). Fear conditioning involves experimentally modeling fear learning during an ‘acquisition’ phase, where a neutral stimulus (e.g., a picture of an animal) is repeatedly paired with an unconditional stimulus (e.g., an electric shock, US). After being paired with the US, the neutral stimulus is now a conditional stimulus (CS+) as it has been conditioned to elicit a physiological response, even in absence of the US. A second neutral stimulus (the CS−) will often serve as a control and a safety signal during a fear conditioning experiment, and it is not followed by the US at any stage. Responding is usually measured using skin conductance responses (SCRs), though other outcome measures such as subjective ratings of CS pleasantness and threat expectancy are also frequently used. Fear conditioning serves as a useful experimental model of exposure therapy, due to the use of ‘extinction’ and ‘return of fear’ phases (Zuj & Norrholm, 2019), Extinction can be conducted immediately after acquisition or after a delay of for instance 24 h. Return of fear can be assessed in different types of paradigms, such as renewal, where extinction is conducted in a context that differs from that of acquisition and the return of fear test. The differential responses (i.e., the difference in responding between the CS+ and CS−) can be measured to test how quickly the fear response to the CS+ is extinguished relative to a control, and subsequently, whether the extinguished fear response returns in response to a variety of manipulations aimed at modeling fear relapse (i.e., return of fear) after treatment (Lipp et al., 2020; Vervliet et al., 2013), such as after a break (modeling spontaneous recovery) or after the US is sporadically presented again following the CS+ (modeling re‐acquisition).

Despite increasing advocacy for methodological rigor in fear conditioning (Beckers et al., 2013; Lonsdorf, Klingelhöfer‐Jens, et al., 2019; Lonsdorf & Merz, 2017; Lonsdorf, Merz, & Fullana, 2019; Ney et al., 2018; Ney, Laing, et al., 2020), and psychological science more generally (Simmons et al., 2011), the experimental designs used in human conditioning research vary considerably using, for instance, different stimuli as CSs and USs. While this disparity has begun to be addressed in principle (Bach et al., 2022) and empirically (Ney, Laing, et al., 2020) there are still many research design options of unknown effects. One example is that fear conditioning paradigms often use different USs—most often an electric shock is used but loud tones (usually 100 dBA) and human screams are also routinely employed (Lipp, 2006; Lonsdorf et al., 2017). However, there is little comparative research assessing the effect of using different USs on conditioning outcomes. Understanding the difference in conditioning between USs is particularly important given that some fear conditioning research has begun to move online, where electric shocks (the most common US) cannot be administered (McGregor et al., 2021; Purves et al., 2019). Glenn et al. (2012) compared a human scream US to electric shock with fear‐potentiated startle as the physiological outcome measure and found that startle responses were significantly higher during the CSs paired with the shock relative to the scream, and the shock US was rated as more aversive. In contrast, Neumann and Waters (2006) reported that a loud, aversive scraping sound produced SCR and heart rate fear conditioning comparable to an electric shock and a loud tone. A shock, however, produced stronger fear conditioning (i.e., stronger acquisition and impaired extinction) compared to an air puff to the larynx and a foul odor (Busch & Evans, 1977; Murray & Carruthers, 1974). However, in a recent experiment that employed four times as many trials compared to an average fear conditioning paradigm, it was reported that a loud tone produced stronger fear conditioning compared to a shock in heart period, subjective ratings of CS valence/arousal, and skin conductance (Sperl et al., 2016), suggesting that the nature of the experimental design will affect how the US is responded to. A meta‐analysis also suggests that aversive film clips show equivalent conditioning when compared to electric shocks for fear‐potentiated startle, subjective ratings of US expectancy and CS pleasantness, but not for SCRs, where shock produces significantly stronger conditioning (Ney, Schenker, & Lipp, 2022).

One of the core principles of fear conditioning is that conditioning—whether it be fear acquisition or extinction—occurs when a participant's expectancy of what is about to occur is violated (Bouton et al., 2020; Craske et al., 2014; Rescorla & Wagner, 1972). Therefore, the premise of exposure therapy is that reduction in fear is due to continually presenting the patient with situations where their fear memory will be challenged and their expectancy of a bad outcome violated (Craske et al., 2014). Similarly, reducing the US reinforcement rate during acquisition to make it less predictable may impair extinction learning (Grady et al., 2016; Humphreys, 1940). However, to the best of our knowledge, no experiment has tested the effects of using multiple USs during acquisition that are presented in an unpredictable sequence on fear conditioning.

Fear conditioning is not only affected by parameters related to the conditioning procedure, but by individual differences as well (Lonsdorf & Merz, 2017). One potential factor that may influence fear conditioning is the endocannabinoid system (Morena et al., 2016; Ney, Akhurst, et al., 2021). The endocannabinoid system is a lipid signaling system that is present in both the peripheral and central nervous systems (Howlett et al., 2002; Ligresti et al., 2016). Arachidonoyl ethanolamide (AEA) and 2‐arachidonoyl glycerol (2‐AG) are able to interface with cannabinoid receptors and modulate memory and learning, including extinction (Devane et al., 1992; Kano et al., 2009; Ligresti et al., 2016; Suguira et al., 1995). Other n‐acyl ethanolamides such as oleoylethanolamide (OEA) do not interact directly with cannabinoid receptor 1 but are involved indirectly in endocannabinoid system function. Animal studies have systematically shown that impairment of endocannabinoid signaling (either via reduced endocannabinoid tone or receptor blockade) significantly reduces fear extinction whereas augmentation of endocannabinoid signaling improves extinction (Hill et al., 2018; Marsicano et al., 2002; Ney et al., 2019). An increasing number of human studies have found that AEA and 2‐AG concentration in blood is associated with fear extinction in the same direction identified in animal studies (Crombie et al., 2021; Mayo, Asratian, Lindé, Morena, et al., 2020; Ney, Crombie, et al., 2022; Ney, Matthews, et al., 2021; Spohrs et al., 2021), and genetic polymorphisms that are associated with increased AEA levels or receptor availability are correlated with improved fear extinction (Dincheva et al., 2015; Mayo, Asratian, Lindé, Holm, et al., 2020; Zabik et al., 2021). However, both genetics and blood biomarkers have notable drawbacks—genetic differences do not necessarily provide an indication of biological differences and blood samples are invasive and provide an acute measure that does not necessarily reflect an individual's typical biological profile. Hair sampling may resolve these disadvantages because hair samples provide a longitudinal (i.e., 1 cm of hair is approximately 1 month's analyte concentration) measure of biological profiling (Gao et al., 2022; Walther et al., 2023).

In the current study, we sought to answer two questions: Firstly, whether variability of the US during fear conditioning leads to exaggerated SCRs and whether endocannabinoid concentrations in hair are associated with fear conditioning. We hypothesized that variation of the US during acquisition would lead to lower habituation to the US, poorer extinction, and higher return of fear during renewal and reacquisition. We expected that the scream and shock USs would produce similar conditioning, based on the results of similar previous research (Neumann & Waters, 2006). Exploratory analyses were conducted on the relationship between hair endocannabinoids and fear conditioning outcomes. We hypothesized that AEA, 2‐AG, and OEA in hair samples would be significantly associated with fear conditioning, with higher AEA, 2‐AG, and OEA levels associated with faster extinction and lower return of fear during renewal and re‐acquisition. Unfortunately, due to technical analysis issues, data loss was experienced in the hair endocannabinoid analysis, such that these results are only interpretable as exploratory.

The current study tested whether the use of a scream US produced fear conditioning similar compared to a shock US. Although participants in the Scream group reported higher experimentally induced anxiety, SCRs were significantly lower during acquisition and there was significantly less CS+/CS− differentiation for the Scream group compared to the Shock group. The Scream group also showed rapid habituation to the US during acquisition when compared to the Shock group. We also tested whether varying between a scream and shock US increased fear conditioning compared to using a scream or shock alone. In most cases, the Shock/Scream group displayed similar, though slightly lower, overall SCRs, and differential SCRs comparable to the Shock group, though the Shock/Scream group showed resistance to habituation toward the US compared to the Shock group. Finally, we tested the relationship between hair endocannabinoid levels and fear conditioning using exploratory analyses (due to low sample sizes). We found some evidence of lower renewal of conditioned fear associated with higher 2‐AG hair concentrations, however, these results were marginal and limited to third‐interval responses. Higher 2‐AG and OEA, as well as lower AEA, were associated with reduced overall SCRs during extinction and renewal. These findings have implications for design of fear conditioning experiments and provide evidence for a relationship between hair endocannabinoids, fear conditioning, and skin conductance responding.

Following the recent global challenges to in‐person research, fear conditioning research has begun to move online (McGregor et al., 2021; Purves et al., 2019). However, there has been little research on the difference in the effect during fear conditioning of auditory compared to electrotactile USs, which are usually used in the laboratory. In the current study, we found that, although the subjective ratings (i.e., CS pleasantness and arousal) were mostly comparable between the US types, SCRs in response to the CS + s in the Shock as well as Shock/Scream groups were significantly larger compared to the Scream group. Further to this, the Scream group showed substantial habituation to the US compared to the Shock and Shock/Scream groups during acquisition. Habituation to shock USs is reported in the literature as being associated with brainstem and hypothalamus activity (Peters et al., 2014; Spoormaker et al., 2010). Our findings are consistent with those of Glenn et al. (2012), who compared a human scream to electric shock but are not similar to Neumann and Waters (2006), who reported that a loud fork scraping noise produced SCR and heart rate fear conditioning comparable to an electric shock as well as a loud tone burst. Our results are also at odds with Sperl et al. (2016) who reported that a loud tone burst resulted in stronger conditioning (i.e., higher acquisition and impaired extinction) compared to a shock over many trials. However, it should be noted that the latter study used a longer conditioning paradigm compared to our own with 45 trials and 4 CSs during acquisition and 50 trials during extinction, and Neumann and Waters (2006) did not use a scream but rather a scraping noise, which could explain the differences between studies. It should also be noted that our Scream group showed lower electrodermal FIRs on the first block of habituation; this is likely because they did not undergo a shock workup, which has been shown to increase response during habituation (Lipp et al., 2015).

Interestingly, no study to our knowledge has used more than one type of US during conditioning. We had hypothesized that, due to the decrease in predictability of the type of US that would be experienced, participants in the Shock/Scream group would show higher anticipatory responding compared to the Shock and Scream alone groups. This hypothesis was made on the basis that fear—as well as extinction—learning rests on the premise that expectancy violation (akin to unpredictability of outcome) is a strong driver of learning (Craske et al., 2014; Rescorla & Wagner, 1972). Our results showed some evidence for the effectiveness of the manipulation—both during acquisition and re‐acquisition, unconditioned responding (i.e., SCRs during the TIR) did not habituate in the Shock/Scream group when compared to the Scream or Shock groups. However, both differential and non‐differential FIRs, SIRs, and subjective ratings in the Shock/Scream condition were either similar to or lower than those observed in the Shock condition. This finding implies that making the nature of the US unpredictable did not make the associated CS a more ‘feared’ stimulus. Importantly, most of the between group differences in this study were found in the second and third interval responses, which are more associated with US anticipation and response to the US, respectively (Prokasy & Ebel, 1967). In comparison, the FIR which is typically the primary dependent measure considered in fear conditioning literature (Lonsdorf et al., 2017) is more associated with orienting responses to CS onset. The use of SIRs and TIRs in this study allowed us to identify differential reactivity between the groups that a focus on FIRs would have missed. These findings suggest that anticipatory or preparatory responses differ as a function of the nature of the US.

In the current study, hair endocannabinoid concentrations were measured using mass spectrometry and correlated with fear conditioning throughout the experiment. Based on previous studies, we expected that higher levels of AEA and 2‐AG would be associated with better fear extinction and lower return of fear (Hill et al., 2018; Marsicano et al., 2002; Morena et al., 2016; Ney, Matthews, et al., 2021). We found partial support for this hypothesis in that higher 2‐AG was associated with lower return of fear, but only during TIRs in renewal, suggesting that this effect was associated only with reactivity toward the absence of the shock rather than shock anticipation (SIR) or CS orienting (FIR). There were no significant effects of AEA. These findings can be explained in a number of ways. Firstly, previous studies have strongly suggested that activation of the cannabinoid receptor 1 is necessary for fear extinction to occur (Akirav, 2011; Chhatwal et al., 2005; Marsicano et al., 2002). As 2‐AG is a cannabinoid receptor agonist, we would expect that higher concentrations would be associated with better extinction learning and subsequently lower return of fear. Although a similar finding was expected for AEA, our previous work identified subpopulations of participants (those with higher posttraumatic stress symptomology) that seemed to display higher fear conditioning if they had higher AEA (Ney, Matthews, et al., 2021). It is possible that AEA is a less reliable marker for fear conditioning than was previously anticipated. However, it is also very probable that our analyses for AEA and 2‐AG were underpowered, given the technical difficulties we experienced during sample processing, and that this issue may have been resolved through acute measurement (e.g., blood samples) and perhaps hair sampling is not the ideal approach. Ultimately, these findings therefore require replication before any firm conclusions can be drawn. Higher hair AEA was associated with higher FIRs during extinction, except in the first Block where higher hair AEA was associated with lower FIRs. Similarly, higher hair AEA was associated with higher SIRs throughout extinction. Higher 2‐AG was associated with reduced overall electrodermal FIRs during extinction, but not acquisition, renewal, or reacquisition. There were significant effects of OEA on SIRs and TIRs during renewal. Since none of these effects were moderated by CS type, it can only be concluded that these effects reflected an overall association between these endocannabinoid levels and physiological arousal, rather than conditional learning. Associations between OEA and differential fear conditioning were not found in the current experiment, which was expected given that OEA does not interface with the cannabinoid type 1 receptor (Ligresti et al., 2016). As stated above, these results should be considered exploratory given the low number of participants included. Finally, skin conductance primarily indexes arousal and contingency learning during fear conditioning (Lipp, 2006). Future studies should also investigate whether the endocannabinoid findings in this study generalize to more valence‐specific measures of fear conditioning, such as startle reactivity. Our study had several limitations that should be considered. Firstly, the Scream group did not have a shock workup procedure, and this clearly impacted SCRs during habituation. It is unknown whether the absence of a shock workup affected the results in the rest of the study, and although previous research suggests that this effect is likely limited to habituation (Lipp et al., 2015), we cannot rule out the possibility that acquisition differed between groups due to the presence or absence of a shock workup prior to habituation. Secondly, due to loss of instrument sensitivity and significant delays in sample processing, only one‐third of the hair samples had quantifiable levels of AEA and 2‐AG. The main reason for this issue was loss of mass spectrometry sensitivity, which was evidenced by the surrogate labeled standards showing substantially lower sensitivity across all samples compared to runs in the recent past. 2‐AG and AEA were quantifiable less often than OEA because OEA is far more abundant that 2‐AG or AEA in this sample matrix, so the loss in sensitivity was only an issue for low abundant targets, with 2‐AG and AEA being relatively low abundance in hair. This means that our biological analyses lacked statistical power and will need replication with larger samples. Future studies should analyze samples within 6–12 months of collection. Participants scored higher than a normative sample on the PSWQ, which is a measure of trait worry (Meyer et al., 1990). We are unsure whether our findings would generalize to a sample that has lower levels of worry. Unfortunately, we were unable to assess the perceived US intensity and pleasantness across groups due to a programming error. Finally, the groups had slightly different sample sizes, sometimes resulting in 13% more participants in the Shock group compared to the Scream group (e.g., habituation).

In conclusion, we found significant differences in conditioned electrodermal SIRs and TIRs between shock, scream, and variable shock/scream USs. These differences indicate that responses to a scream US habituate rapidly and that it supports lower differential conditioning, as well as overall physiological arousal, in the SIR and TIRs of acquisition, renewal, and re‐acquisition phases, but not extinction. A variable shock/scream US schedule performs similarly to one using only a shock US but with slightly lower differential and overall responding, as well as reduced habituation during acquisition and re‐acquisition. Finally, we found some evidence that hair endocannabinoids affect the renewal of conditioned fear, as well as overall physiological arousal during fear conditioning. These findings strongly suggest that different USs produce substantially different anticipatory conditioned responses and that these responses have different characteristics throughout fear conditioning paradigms. When using SCRs, where possible shocks should be the preferred US due to higher conditioned responding. More research using larger samples is needed to examine the relationship between hair endocannabinoids and fear conditioning, though our exploratory work here suggests that 2‐AG levels may be associated with conditioned fear renewal.

Luke J. Ney: Conceptualization; data curation; formal analysis; investigation; methodology; project administration; software; supervision; visualization; writing – original draft; writing – review and editing. David Nichols: Conceptualization; formal analysis; investigation; resources; software; supervision; writing – review and editing. Ottmar V. Lipp: Conceptualization; data curation; funding acquisition; methodology; resources; software; supervision; writing – review and editing.

This work was supported by the Australian Research Council (DP180100869) and the National Health and Medical Research Council (APP1156490).

We would like to thank Anisuzzaman Chowdhury, Matthew O'Donohue, Tynneille Mulder, and Sarah Olsson for their research assistance during the project. Open access publishing facilitated by Queensland University of Technology, as part of the Wiley ‐ Queensland University of Technology agreement via the Council of Australian University Librarians.

Ney, L. J. , Nichols, D. S. , & Lipp, O. V. (2023). Fear conditioning depends on the nature of the unconditional stimulus and may be related to hair levels of endocannabinoids. Psychophysiology, 60, e14297. 10.1111/psyp.14297

The data that support the findings of this study are available from the corresponding author upon reasonable request.