Renewal in human fear conditioning: A systematic review and meta-analysis

Renewal is a ‘return of fear ’ manipulation in human fear conditioning to investigate learning processes underlying anxiety and trauma. Even though renewal paradigms are widely used, no study has compared the strength of different renewal paradigms. We conduct a systematic review (N = 80) and meta-analysis (N = 23) of human fear conditioning studies assessing renewal. Our analysis shows that the classic ABA design is the most effective paradigm, compared to ABC and ABBA designs. We present evidence that conducting extinction in multiple contexts and increasing the similarity between acquisition and extinction contexts reduce renewal. Furthermore, we show that additional cues can be used as safety and ‘protection from extinction ’ cues. The review shows that alcohol weakens the extinction process and that older adults appear less sensitive to context changes and thus show less renewal. The large variability in approaches to study renewal in humans suggests that standardisation of fear conditioning procedures across laboratories would be of great benefit to the field.

Fear conditioning is an important experimental paradigm in modern psychological science that is believed to model learning processes that are thought to underlie anxiety and trauma-related disorders (Bouton et al., 2020;Ney et al., 2022).Fear conditioning paradigms typically involve three phases, acquisition, extinction and extinction retest.During acquisition, a neutral stimulus (e.g., a tone, a light or a picture) is repeatedly paired with an aversive unconditional stimulus (US; e.g., an electric shock) which elicits a fear response.Over repeated pairings, the previously neutral stimulus (now a conditional stimulus; CS+) comes to elicit physiological and subjective responses that resemble those to the US, like the fear response (LeDoux, 2014;Lonsdorf et al., 2017).During extinction, the CS+ is presented without the US.Here the initial fear learning (the CS predicts danger) is overlayed by new safety learning (the CS predicts the absence of danger), however, the original fear memory is retained (Bouton, 2004).This safety or extinction learning is believed to model the processes that occur during exposure-based therapy (Craske et al., 2014).The stability of extinction learning is assessed in a subsequent extinction retest during which the CS+ is again presented without the US to determine whether conditional responses are still present.In human fear conditioning fear learning to the CS+ is frequently assessed in comparison to a second CS (CS-) which is presented during all phases, but never paired with the US.Thus, fear conditioning provides a framework for investigating the mechanisms underlying therapy for anxiety and trauma and can provide insights as to how to improve treatment efficacy (Craske et al., 2014;Lipp et al., 2020).
While exposure therapy is effective in the short term, relapse rates are high (Loerinc et al., 2015).The reasons why the underlying fear memory resurfaces for some people but not for others are currently not understood.To investigate fear memory retention in the laboratory environment, four manipulations are used to induce a return of fear: reacquisition, reinstatement, spontaneous recovery, and renewal.During reacquisition, the US is again presented together with the CS+ as in the original acquisition phase and fear memory retention is evident if fear is acquired faster than during the initial acquisition or in a naïve control group.In reinstatement, the US is presented on its own between extinction and extinction retest.In spontaneous recovery, a time gap is inserted between extinction and extinction retest (i.e., 24 h or a week).Renewal is assessed by creating a context change between acquisition and extinction and conducting the extinction retest in a context that differs from that during extinction.In the laboratory, this context change is frequently created by a change in the background image or screen colour, but more complex virtual reality approaches have been used as well.In reinstatement, spontaneous recovery and renewal, a return of fear is considered evident when after the manipulation, the fear response that was extinguished or reduced in the extinction phase, returns during extinction retest.
Renewal seems particularly relevant for clinical settings as patients receive treatment in a safe environment (i.e., in the clinic), but relapse often occurs once patients return to their daily surroundings (i.e., outside the clinic).Thus, the renewal paradigm is the focus of this current review and meta-analysis.Different renewal paradigms are named by using letters to label the contexts used during different experimental phases.For instance, ABA renewal refers to the design where the contexts of acquisition and extinction retest are the same but differ from the extinction context.In ABC renewal, the contexts of acquisition, extinction, and extinction retest are different from each other.In some studies, an extinction recall phase in the same context as used during extinction is added between extinction and the extinction retest, forming ABBA renewal.
Much fear conditioning research has been completed in non-human animals, in particular rodents (Hermans et al., 2006).Whilst many findings can be translated into the human context, the structure of fear conditioning paradigms differs between animal and human research.Animal studies usually employ a single cue paradigm where only one CS is presented and paired with a US and performance is then compared with a control group of animals that is presented with the stimuli unpaired.In human conditioning, a differential conditioning paradigm is the most frequently used where two CS are presented, one paired with the US (CS+) and one presented alone as a control (CS-; Haaker et al., 2013).These design differences are not trivial as different outcomes have been shown between single cue and differential paradigms in human non-aversive conditioning.Potentiated startle, one of the common physiological indicators of fear learning in rodents and humans, is potentiated in humans during differential (Lipp et al., 2003), but not single cue non-aversive conditioning (Lipp et al., 1994).Moreover, animal experiments are typically conducted under far more controlled conditions, and there are some differences between animal and human cognition that likely influence conditioning processes.For instance, humans more readily configure multiple conditional stimuli into a new entity whereas rodents seem to learn more readily about individual stimulus elements (Williams et al., 1994).Thus, findings cannot necessarily be transferred one to one across species or conditioning procedures.Understanding the limitations and boundaries of applying animal findings to human conditions is essential for interpreting research outcomes accurately.
Past animal work has compared the strength of renewal across different paradigms.This work confirmed that ABA designs lead to stronger renewal than ABC designs (Holmes and Westbrook, 2014) and that counterconditioning during extinction strengthens ABA renewal (Holmes et al., 2016).However, to date, there is no systematic review or meta-analysis to compare the effect of different renewal paradigms (e.g., ABA or ABC) to paradigms without a context switch (e.g., AAA) in human fear conditioning.In addition, our laboratory (Wang et al., 2024) and others have at times found that renewal does not manifest in electrodermal responses, the most frequently used physiological measure in human fear conditioning, and we therefore want to assess the factors that impact renewal across different dependent measures (e.g., Haesen & Vervliet, 2015;Leer et al., 2018;Muench et al., 2021;Pace-Schott et al., 2014).In the current systematic review and meta-analysis, we compared ABA to AAA, ABC, and ABBA designs respectively, to examine the strength of different renewal manipulations in human fear conditioning studies.We also conducted a systematic review of all identified human fear conditioning studies using renewal designs and provided a narrative summary of their results.This data can inform future research in the field and provide new directions for research on the return of human fear.

General procedure
A systematic literature search was performed.Articles were compiled online with the Rayyan software (Ouzzani et al., 2016), date cutoff was June 2022.Article screening was independently conducted by both LJN and YW.Any discrepancies regarding inclusion were thoroughly discussed until a consensus was reached.A comprehensive table has been generated, outlining the design of the systematic review included articles and the inclusion of meta-analysis articles (refer to the Results section).Data extraction was carried out, and missing information was addressed by contacting the authors before the meta-analysis was conducted.Among the 33 experiments (from 31 articles) meeting the inclusion criteria of meta-analysis, data from 23 studies were successfully extracted and data from eight studies were missing (i.e., none of the contacted authors provided the requested data).

Search strategy
Following PRISMA guidelines (Moher et al., 2015;Page et al., 2021), we performed a systematic literature search of PsychINFO, Medline, and PubMed.The included search terms in PubMed were ((((((((((fear conditioning) OR (threat conditioning)) OR (fear acquisition)) OR (threat acquisition)) OR (fear extinction)) OR (threat extinction)) OR (extinction)) OR (conditioning, classical [MeSH Terms])) OR (fear learning)) OR (threat learning)) OR (pavlovian conditioning) AND ((renewal) OR ("return of fear")) OR (contextual change*).The included search terms in PsychINFO were TX (TX fear conditioning OR TX threat conditioning OR TX fear extinction OR TX threat extinction OR TX fear learning OR TX threat learning OR TX fear acquisition OR TX threat acquisition OR TX extinction leaning OR MA conditioning, classical OR TX psychological extinction OR TX extinction) AND (TX renewal OR TX "return of fear" OR TX contextual change" OR TX "context" A* OR TX *Context" A" OR TX "Context" B" OR TX "context* B* OR TX "ABA" OR TX "ABC" OR TX "extinction context*" OR TX "conditioning context"" OR TX "acquisition context"" OR TX "context(A)" OR TX "CX+" OR TX "CX-").The included search terms in Medline were (TX renewal OR TX "return of fear" OR TX contextual change*) AND (TX fear conditioning OR TX threat conditioning OR TX fear extinction OR TX threat extinction OR TX fear learning OR TX threat learning OR TX fear acquisition OR TX threat acquisition OR TX extinction learning OR MA conditioning, classical OR TX psychological extinction OR TX extinction).

Study eligibility
Publications were included in this review if they (1) are journal articles, book chapters or theses, (2) are published in English, (3) used a fear conditioning paradigm (i.e., pairing neutral with aversive stimuli), (4) included acquisition, extinction, and extinction retest phases in at least one condition, and (5) if containing multiple studies, participants were adult humans in at least one study.
Articles were excluded from this review if they (1) did not have the full text in English, (2) were review articles, (3) reanalysed secondary data or data from other publications, (4) used only measures other than electrodermal responses, fear potentiated startle, or subjective ratings (i.e., US expectancy ratings), ( 5) employed a single cue paradigm, (6) used children as participants or (7) used evaluative conditioning or avoidance learning.Furthermore, articles were included in the systematic review Y. Wang et al. only, if the order of extinction retest and spontaneous recovery/ extinction recall phases was counterbalanced.

Systematic review structure
We systematically reviewed eligible articles to clarify firstly whether renewal was observed and whether imaging was included to identify the neural bases of the results.After this general overview, we were interested in the impact of differences in experimental setup: Differences were observed in the length, timing, and structure of the extinction phase.The inclusion of safety cues and variations in time of day were other differences that are discussed here.Finally, we review the effects of participant centred differences including gender, age, and personality traits as well as lifestyle factors such as level of physical activity and alcohol/cannabis consumption.

Analysis and meta-analytic technique
Due to the complex nature of fear conditioning data, we calculated Pearson's correlation coefficients (r) for all extracted effects and these r values were used for meta-analysis (see Schenker et al., 2021 for a recent example of this).Conversion to r was accomplished using the effect size spreadsheet from Lakens (2013).Several different types of comparisons were made: (1) the differences between electrodermal responses and US expectancy during CS+ and CS-during the early trials of renewal were compared between ABA and AAA designs.(2) the differences between electrodermal responses and US expectancy during CS+ and CS-during the last trials of extinction were subtracted from those during the early trials of renewal, and this difference was compared between ABA and AAA designs.(3) Due to the low number of studies comparing ABC or ABBA with ABA or AAA paradigms we pooled renewal CS+/CS-differentiations in electrodermal responses and US expectancy across studies and compared: (a) ABC versus ABA, and (b) ABBA versus ABA.In this way, we compiled studies that directly compared ABA to AAA designs and then used ABA renewal as a benchmark to contrast against ABC and ABBA designs.
To produce these analyses, we aimed to extract t-statistics or F-statistics that compared responses to CS+ and CS-during early renewal directly between ABA and AAA designs, or within ABC or ABBA designs.Where statistical tests were not available, we sought to procure the means and standard deviations from the text or tables.Where these were not reported, we used PlotDigitizer (Huwaldt & Steinhorst, 2015) to extract the means and reported variance from plots.If none of this information was available in the manuscripts, we emailed the first and senior authors of the publications.If a response was not received within 1.5 months, we excluded the data from the meta-analysis.Among the 33 experiments (from 32 articles) meeting the inclusion criteria of meta-analysis, data from 24 studies were successfully extracted and data from eight studies were missing.The authors of the eight papers were contacted, yet none of them provided the requested data.The presentation of data, analysis methods, and result reporting in fear conditioning studies exhibit considerable variability.Data may be presented as a function of either individual trials or aggregated blocks (typically the average of two trials).Analytical approaches might involve comparing conditional responses between different CSs within a trial or block, assessing response levels across experimental phases, or comparing responses between different phases (e.g., extinction versus return of fear, or acquisition versus return of fear).Moreover, the reporting of statistical results often lacks consistency, with effect sizes sometimes omitted or presented in different formats.These discrepancies pose challenges for data extraction.However, our efforts to transform and calculate effect sizes, as well as communication with authors to obtain missing information, were intended to minimize missing data and their impact.
It is known that fear conditioning studies have often not adhered to consistent analysis practices (Lonsdorf et al., 2019;Ney et al., 2020;Ney et al., 2022).In this meta-analysis, we were therefore unable to base the pooled effect sizes on a consistent number of trials, since across the different studies varying numbers of trials during late extinction and early renewal were used to calculate critical comparisons.
Comparisons of ABA versus AAA designs were performed using metafor (Viechtbauer, 2010) in R. We used Restricted Maximum Likelihood and visualised the results using forest plots, where possible.Effect size heterogeneity was quantified using both Hedge's Q test (Hedges, 1981(Hedges, , 1982a(Hedges, , 1982b) ) as well as I 2 .High, medium, and low I 2 values were qualified as 75 %, 50 %, and 25 %, respectively.Egger's test (Egger et al., 1997) was used to test funnel plot asymmetry and potential publication bias.For the comparison of ABA versus ABC/ABBA designs, we subtracted the pooled differential renewal effect (i.e., pooled r) of the ABC/ABBA designs from the ABA pooled renewal effect.Since this analysis was not performed using metafor we do not report the standard meta-analysis statistics for these comparisons.

Results
For the current review, 4541 articles were identified following database searches and a further 46 articles were identified outside of databases (Fig. 1).After the removal of duplicates and following title/ abstract screening, 203 articles were sought for full text screening.Eighty articles were deemed eligible for inclusion in the systematic review, with 31 eligible for meta-analysis.Details on the studies (i.e., sample sizes, experimental designs, number of trials, etc.) included in this systematic review and meta-analysis are reported at https://osf.io/ckgeh/?view_only=f9e4d7d5295c4ecf99a8ce4051d3b569

Systematic review
Several factors have been shown to influence renewal.In this systematic review, we first looked at the presence of overall renewal.Across fear conditioning studies, great variety can be observed in the extinction phase design, and we turned our attention towards the possible impact of the number of trials, delayed extinction, multiple contexts during extinction, use of a safety cue and other modifications.Following this, we looked at the neural bases of renewal.It is of interest to understand what neural structures are involved, especially for clinicians who are working with patients with brain impairments.In studies involving patients, we looked at the difference in fear conditioning between healthy and patient populations.While research on clinical populations is always of interest to clinicians, fear conditioning research is focused on identifying the basic processes underlying acquisition, extinction and return of fear in healthy populations.Understanding the process comprehensively will then allow for effective adaptation for client populations as well.Thus, this systematic review next looked at the impact of individual differences and time of day of fear conditioning in healthy populations.Finally, several interventions have been tested for their impact on fear conditioning and fear renewal and we summarize those in the final part of the systematic review for completion.
Renewal was also shown in less commonly used ABBA paradigms which were conducted across two days, with acquisition in context A  2005) used test designs involving some time passing between extinction and renewal test.Therefore, some of the 'renewal' effects might be conflated by spontaneous recovery.Similarly, Zeidan et al. (2012) found consistent renewal effects in three ABBA renewal tests using the same participants with more than 12-week intervals between subsequent sessions (Zeidan et al., 2012).It should be noted that some studies (e.g., Merz et al., 2016;Mosig et al., 2014) counterbalanced the order of acquisition and extinction context during the renewal test.This may have affected the results by making the context change more salient which could result in stronger renewal.
Some debate exists about the effect of CS fear relevance on fear renewal (Ney et al., 2022).Using fear-irrelevant stimuli as CSs, spider-fearful and non-fearful individuals did not show renewal in US expectancy or electrodermal responses (Mosig et al., 2014).However, another study compared fear conditioning with fear-relevant and fear-irrelevant stimuli and observed renewal for both stimulus conditions in both US expectancy and electrodermal responses (Neumann & Longbottom, 2008).
While renewal is a commonly used and highly effective manipulation to induce the return of fear, it is important to note that not all studies find renewal across all measures.US expectancy appears to provide the most robust measure of renewal, while some studies do not find a corresponding renewal effect in electrodermal responses (e.g., Haesen & Vervliet, 2015;Leer et al., 2018;Muench et al., 2021;Pace-Schott et al., 2014).

Extinction: number of trials
Different studies use varying numbers of extinction trials.It is important to present sufficient trials to achieve fear extinction while it is unclear if over-extinction would eliminate renewal.Krisch et al. (2018) found that 36 extinction trials resulted in lower differential US Y. Wang et al. expectancy during the renewal test compared to 12 extinction trials.Further research is needed to identify whether further extinction trials would eventually lead to an elimination of renewal and if so, how many trials are required.

Extinction: delayed extinction
In fear conditioning paradigms, it is common for the acquisition phase to be immediately followed by the extinction phase.However, in clinical settings delayed extinction is commonplace as patients typically seek help well after the acquisition of fear.Thus, a clearer understanding of the importance of timing between the acquisition of fear and treatment seeking will allow for a more realistic modelling of the processes involved in exposure-based therapy.
Studies have found that delaying extinction training by 24 hours after acquisition can reduce or eliminate renewal indicated by electrodermal responses (Huff et al., 2009;Merz et al., 2016) and US expectancy (Merz et al., 2016), compared to immediate extinction training (i.e., a delay of a few minutes).Similarly, without a comparison group (i.e., immediate extinction), one study has found that extinction on the second day of the experiment still produced a renewal effect in electrodermal responses when tested a further 24 hours after extinction (Fricchione et al., 2016).In summary, initial studies suggest that delaying extinction might reduce, but not eliminate renewal.Further research is needed to systematically test the effect of delays to extinction and to renewal tests on the return of fear.

Extinction in multiple contexts
In a standard ABA renewal paradigm, two different contexts are used: The acquisition context (A) is repeated in the renewal test phase while extinction takes place in the alternate context (B) before the return of fear is tested.This setup somewhat mirrors the clinical context where fear is acquired 'in the wild', treated in a highly controlled different setting (therapist's room), and then might return when patients return to the 'real life' context.Some studies have investigated the impact of conducting extinction in multiple contexts and findings appear contradictory.One study found that using multiple extinction contexts (i.e., three contexts in their first experiment, A(BCD)A compared with AAA and ABA, and five contexts in their second experiment, A(BCDEF)A failed to attenuate renewal measured by US expectancy (Neumann et al., 2007).However, there is also evidence to suggest that the use of multiple contexts can effectively reduce renewal.For example, in an A(BCD) A paradigm, renewal was attenuated in US expectancy, in both 12-trial and 36-trial extinction conditions (Krisch et al., 2018).Similarly, multiple contexts during extinction (i.e., A(BCD)E) weakened renewal in US expectancy and fear potentiated startle, compared to an ABE condition (Balooch et al., 2012).When compared with extinction in a single context, multiple context extinction was followed by marginally significant renewal only (Dunsmoor et al., 2014).It appears that the type of different contexts -particularly similarity of contexts -used in these manipulations plays an important role.Balooch and Neumann (2011) varied the similarity between extinction and test context by changing the light intensity in the experimental room.Relative to extinction in a single context, renewal indexed by US expectancy was reduced after extinction in multiple contexts if the contexts were dissimilar but disappeared completely after extinction in multiple similar contexts.These findings suggest that renewal is reduced when extinction and renewal test contexts are similar and that the use of multiple contexts and similarity of contexts can lower the renewal of fear responses.These findings have potential implications for clinical practice.Considering that extinction across different contexts strengthens safety learning and reduces subsequent renewal, therapists might be well advised to conduct treatment in a variety of settings to improve outcomes for patients at high risk of relapse.

Extinction: use of safety cue
Numerous studies have shown that the presentation of a safety cue can reduce renewal in US expectancy ratings.For instance, the presentation of an ampersand symbol, "&", together with the CSs during extinction and renewal test was successful in attenuating renewal compared with a control group where the '&' was only presented during extinction (Dibbets et al., 2008).However, the same symbol did not reduce renewal as measured by US expectancy when presented on only 66.67 % of the extinction trials (Dibbets, 2012).In another study, presenting a positively valenced smiley face was found to be effective in reducing renewal, whereas a negatively valenced hashtag, "#", was not (Dibbets & Maes, 2011).Interestingly, an ampersand symbol and a hashtag, which were counterbalanced as extinction and neutral cue between participants, were found to reduce renewal as measured by US expectancy in the short term, but a coloured cat and a mouse (also counterbalanced between participants) did not have the same effect after increasing the salience of the context change (i.e., changing the illumination colour of the lab, in addition to the background colour of the screen, Quezada et al., 2018).Additionally, providing safety information before extinction ("the probability that an electric shock will follow this shape is extremely small (1/1000)") has been found to reduce renewal as measured by US expectancy (Scheveneels et al., 2019).The presentation of an acquisition cue before the test phase elicited stronger renewal in US expectancy and electrodermal responses than that of an extinction safety cue (Luck et al., 2018;Vansteenwegen et al., 2006).However, it is difficult to determine whether renewal was reduced in the extinction cue condition or increased in the acquisition cue condition, or both, due to the absence of a baseline condition (i.e., 'no cue' condition; Vansteenwegen et al., 2006).The timing of the additional cue presentations matters greatly as an additional cue can also enhance renewal.Cues presented during extinction only (and not during renewal) have been described as providing 'protection from extinction' with participants learning an association between the absence of shock and the protection cue and hence show enhanced renewal in a renewal test without the additional cue (Lovibond et al., 2000).In summary, the presentation of additional cues during fear conditioning (during acquisition, extinction and/or renewal phase) has an impact on the return of fear as participants will use these cues to guide their assumption of safety or fear, however, the type and timing of the cue appears to play a role with some being more effective than others or even detrimental.

Extinction: other modifications
Several studies have investigated how changes within the design of the extinction phase can impact renewal.Unlike typical extinction which presents the CSs without the US, Haesen and Vervliet (2015) presented the US without the CSs during extinction and found that renewal was eliminated in electrodermal responses, while differential responding returned in US expectancy.Similarly, Leer et al. (2018) found that US only extinction was followed by no renewal in fear potentiated startle while renewal was evident in US expectancy ratings.Other studies have found that presenting unpaired USs during the intertrial intervals of extinction training can reduce renewal (Lipp et al., 2021;Vervliet et al., 2010).While of limited immediate use to clinical practise these results suggest that there may be other ways to strengthen extinction learning and to reduce renewal.
Finally, while US intensity was kept constant in studies of US extinction, some research has investigated the impact of varying US intensity across the paradigm yielding mixed results.Firstly, inserting an additional intervention phase between acquisition and extinction, which was analogous to acquisition but employed a gradually decreased US intensity, did not impact renewal in US expectancy ratings but lowered the cost estimate of experiencing the US (Leer & Engelhard, 2015).It appears that altering US intensity does not impact US expectancy but can change other fear-relevant measurements such as fear-potentiated startle and valence ratings.It is interesting to note that for all modifications involving the US, the renewal in US expectancy was usually maintained, while attenuation was noted in physiological measurements such as electrodermal responses and fear-potentiated startle.
Y. Wang et al.

Neural mechanisms
It is widely accepted that the central neural structure in fear conditioning is the amygdala in a complex interplay with the hippocampus and medial prefrontal cortex (PFC; Maren et al., 2013).Neuroimaging studies have identified that different brain regions are activated during different phases of the fear conditioning paradigm.Kalisch (2006) found that activation of the ventromedial prefrontal cortex (vmPFC) and hippocampus in response to a conditional stimulus was context-dependent, only occurring in the extinction but not acquisition context in humans.Åhs et al. (2015) further compared prefrontal cortex activity between an extinction recall (in extinction context) and a renewal phase (in acquisition context).In line with the above findings, they found vmPFC activity during extinction recall, and increased dorsomedial prefrontal cortex (dmPFC) activity occurred during renewal.In a clinical application, patients with post-traumatic stress disorder have been found to have impaired extinction recall and, in line with the above findings, show diminished vmPFC activity in the extinction context compared with health controls (Garfinkel et al., 2014).
In addition to vmPFC and dmPFC, studies have shown that during renewal the context change group displayed less activation in the dorsolateral prefrontal cortex (dlPFC) and higher levels of activation in the orbitofrontal cortex when compared with a control group with no context changes across the experiment (Icenhour et al., 2015).Lipp et al. (2020) discovered that anodal transcranial direct current stimulation applied to the dlPFC during extinction attenuated renewal, indicating the importance of this structure for renewal.
The involvement of the hippocampus in fear conditioning is well documented.Hermann et al. (2017) found that renewal was linked to greater structural integrity of the part of the cingulum located in the hippocampus, and participants who exhibited a more robust renewal of conditional electrodermal responses in the new context demonstrated greater effective connectivity between the activation foci in the hippocampus and the structures in the fear and extinction network (Hermann et al., 2016).Like hippocampus involvement, the importance of the amygdala in fear conditioning has been shown in many studies.In a renewal paradigm, Hermann et al. (2020) found that when comparing a multiple context group to a single-context group, the former exhibited decreased differential activation of the amygdala during fear renewal in the new context (i.e., context C in an ABC design), relative to the extinction context.This suggests that amygdala activation differs depending on the fear renewal manipulation, such as the number of novel contexts presented during extinction.Another study also highlighted the involvement of the posterolateral cerebellum, particularly lobules VI and Crus I, in response to CS+ presentation during renewal.Cerebral activation was strongest during the presentation of CS+ at the start of the renewal phase which suggests that associative fear learning in the cerebellum is not fully erased during extinction (Batsikadze et al., 2022).In summary, the neural base of human fear conditioning is complex and involves the activation of a variety of brain structures across the different phases of an experiment.This might explain why results are varied and highlight the importance of further work to understand the neural structures underlying fear learning in humans.

Differences between healthy and patient populations
Renewal paradigms have been used to identify differences in return of fear across different clinical patient populations.Schizophrenic patients have been shown to display an abnormal pattern of renewal.Specifically, one study found that patients with schizophrenia had lower electrodermal responses towards CS+ during renewal than during extinction recall (Holt et al., 2012).In another study, patients with schizophrenia did not show renewal, while a healthy control group displayed renewal (Holt et al., 2009).
Obsessive compulsive disorder patients showed stronger renewal in electrodermal responses compared to healthy controls in one study (Fyer et al., 2020), but this was not replicated in another study that found no difference in renewal between healthy participants and patients with obsessive compulsive disorder using electrodermal responses (McLaughlin et al., 2015).
Post-traumatic stress disorder (PTSD) patients have been reported to have impaired extinction recall (i.e., significantly higher electrodermal responses to CS+ than to CS-) and higher non-differential renewal (i.e., higher electrodermal responses overall but no difference in response to CS+ and CS-), compared to healthy participants (Garfinkel et al., 2014).Similarly, PTSD patients showed stronger renewal in electrodermal responses compared to healthy control participants and trauma-exposed non-PTSD participants (Wicking et al., 2016).Inconsistently, Kaur (2019) did not find any systematic difference in the renewal between PTSD and non-PTSD participants or between combat exposed and non-exposed participants.Using multiple contexts in extinction, renewal as indicated by electrodermal responses was seen in PTSD patients but not in healthy individuals.In expectancy ratings, renewal approached significance (p = .059)in PTSD patients while again the healthy individuals showed no renewal (Hennings et al., 2020).Since the presentation of multiple contexts during extinction might reduce renewal, one explanation of these findings might be that PTSD patients have a stronger renewal effect, which was not sufficiently reduced after extinction in multiple contexts, or that extinction in multiple contexts does not affect renewal in PTSD patients.
Social anxiety disorder patients have been reported not to differ from healthy controls in renewal when measured using electrodermal responses (Fyer et al., 2020).Similarly, in a large sample of 377 participants, no correlation was found between the level of trait anxiety and renewal in electrodermal responses and US expectancy (Haaker et al., 2015).Anorexia nervosa patients did not differ from healthy controls in renewal as indicated by electrodermal responses (Fyer et al., 2020), but stronger renewal was observed in US expectancy (Lambert et al., 2021).Patients with subacute focal cerebellar lesions did not show renewal in fear potentiated startle, potentially due to significantly reduced extinction learning (Steiner et al., 2019).Muench et al. (2021) showed that alcohol dependent individuals did not show renewal in electrodermal responses, however, they also did not find renewal in the healthy controls and therefore these results will need to be replicated before any conclusions about the effect of alcohol dependency on renewal can be drawn.Anxiety disorders are often co-morbid and therefore it is important to consider how other morbidities impact the likelihood of relapse.Understanding the mechanisms underlying these differences would enable better targeting of interventions.

Impact of individual differences
Literature suggests that individual difference factors can modulate renewal in fear conditioning.One study reported a significant sex difference in renewal, as measured by US expectancy (Lonsdorf et al., 2015).Women displayed higher US expectancy ratings, smaller electrodermal responses, and reduced discrimination between CSs during renewal compared to men, however, no difference in renewal measured by electrodermal responses was found in that study (Lonsdorf et al., 2015).In contrast, Martínez et al. (2014) found larger renewal as measured by electrodermal responses for male participants compared with female participants in the first two renewal trials in a Puerto Rican sample.Another study considered hormonal changes in women throughout the cycle and found a stronger effect in men than in mid-menstrual cycle women or early-cycle women measured by electrodermal responses (Milad et al., 2006).These findings seem to suggest that while gender differences exist within measurements, most notably with females showing smaller electrodermal responses overall, these individual differences do not have an impact on renewal.This is an important finding, especially considering that most fear conditioning research is completed with mostly female participants as many researchers rely on psychology course undergraduates as participants, which are predominantly female.However, some gender differences may exist: Meir Drexler et al. (2014) investigated the impact of postretrieval-extinction learning.After acquisition on day 1 using 4 fear Y. Wang et al. relevant animal pictures in two different zoo contexts (2 CS+ and 2 CS-), half their participants were presented with one of the CS+ without the US or relevant context for 30 seconds, then they watched a 10-minute irrelevant TV show before extinction training on day 2 (reactivation group).The control group watched the same 10-minute TV show before the extinction phase without any reactivation of CS+.This resulted in renewal in electrodermal responses only in the female reactivation group, but not in the female control group or male reactivation or control group.
It appears that age impacts extinction memory in fear conditioning.Battaglia et al. (2018) compared young adults (20-30 years old) with older adults (60-70 years old) and found that while there were no differences in electrodermal responses for acquisition and extinction on day 1, older adults displayed reduced extinction memory shown by increased electrodermal responses to the previously extinguished stimuli in the extinction context.However, no differences were found in self-reported fear ratings and contingency awareness across the age groups, suggesting that the age differences selectively impact autonomic responses to extinction memory.
Individual differences such as personality or trait anxiety can also influence the strength of renewal.Extraversion was found to negatively correlate with the extent of renewal indicated by electrodermal responses (Martínez et al., 2012).Higher levels of trait anxiety were associated with poorer discrimination between the danger cue and the safety cue, which in turn was associated with higher renewal (Staples-Bradley et al., 2018).

Impact of time of day
Fear conditioning experiments are mostly conducted as in-person laboratory experiments and as such most data are generated from experiments during the daytime.It is therefore important to consider whether the time of day that the experiment is conducted is likely to impact the results.In a study by Pace-Schott et al. (2013), participants who underwent fear conditioning in the morning (8-9 am) displayed lower renewal compared to those in the evening group (8-9 pm).In line with these findings, Pace-Schott et al. (2014) found that participants' electrodermal responses were significantly higher in a PM compared with an AM session.Most laboratory experiments are run during standard university opening hours, i.e., between 8 am and 4 pm.Therefore, data might be skewed towards lower renewal rates and lower overall electrodermal responses, ultimately underreporting some fear conditioning results.

Impact of interventions
The renewal paradigm has been used frequently to assess the effects of various interventions on the generalization of fear extinction memories across contexts.Some research focuses on the effects of common activities, such as physical activity, or of consumption of mind-altering substances such as alcohol or cannabis.Bisby et al. (2015) found that participants who consumed alcohol showed normal acquisition of fear conditioning, but had slower and weaker extinction learning, and reduced renewal compared to a control group.On the other hand, Das et al. (2013) found that pre-or post-extinction consumption of cannabidiol (CBD), a non-psychoactive constituent of cannabis, did not lead to any changes in renewal, compared to a placebo group.Also, no evidence supported a relationship between smoking tobacco and renewal (Haaker et al., 2017).Similarly, according to Crombie et al. (2021), aerobic exercise did not decrease renewal in either US expectancy or electrodermal responses.Therefore, cannabis, tobacco and exercise do not seem to impact renewal, however, alcohol has been shown to weaken extinction and hence increase the return of fear.These findings are important in clinical application as the client's everyday lifestyle might impact the effectiveness of treatment and thus needs to be taken into consideration.
Anxiety is often a co-morbid disorder and thus many clients are in treatment for other illnesses.Therefore, the interplay of those treatments on anxiety pathways is important to understand.Research has investigated the effects of various drugs and hormones on renewal after fear extinction learning.Cisler et al. (2020) found that immediate administration of L-dopa/carbidopa after extinction to women with PTSD reduced electrodermal responses during renewal, but only in the lower dose (100/25 mg) group, not the higher dose (200/50 mg) group.Post-extinction L-dopa also seemed to abolish renewal, with decreased amygdala and increased ventromedial prefrontal cortex activity observed through fMRI (Haaker et al., 2013).Privratsky (2021) included renewal as part of the study design in two experiments about the effectiveness of L-DOPA on regulating fear generalization in PTSD, but renewal was not the focus of the studies, so no statistics were reported.D-cycloserine, a medication that is used to enhance extinction learning in phobias, did not show any influence on renewal (Hofmann et al., 2019).The studies suggest that L-dopa can influence the fear conditioning pathway, however, dosage seems to matter and further research is needed to identify critical levels.
Stress impacts most individuals throughout their lives and can have complex effects on renewal, with different timing and methods of inducing stress producing varying outcomes.For example, while one study found that inducing stress before extinction training eliminated renewal (Drexler et al., 2018), another study found that inducing stress after extinction training actually enhanced renewal measured one day later (Hamacher-Dang et al., 2015).However, the timing of when renewal was measured appears to play a major role.In a similar study with stress again induced one day after extinction, renewal was measured 20 minutes after stress induction and found to be attenuated (Merz et al., 2014).To address the effects of stress and anxiety, Sun (2019) implemented emotion regulation approaches before extinction training, but reappraisal and suppression groups did not display reduced renewal relative to the control group.Landkroon et al. (2020) reported that a dual-task intervention did not impact renewal in either electrodermal responses, fear potentiated startle or expectancy ratings.In a separate study, Landkroon et al. (2021) confirmed these findings but found that participants in the dual-task and recall-only groups reported a lower level of CS unpleasantness.Raeder et al. (2020) found that among spider-phobic participants, even though there was no difference between CS+ and CS-across groups, completers of exposure therapy rated the CSs more positively and had lower contingency ratings than non-completers.Cortisol, the primary human stress hormone, was found to not affect contextual renewal in men or women (Kinner et al., 2018) and hydrocortisone did not attenuate renewal or context dependency of extinction (Merz et al., 2018).Inhibition of the endocannabinoid enzyme fatty acid amide hydrolase also failed to reduce renewal (Mayo, 2020;Mayo et al., 2020).In summary, fear conditioning is impacted by a variety of factors in often contradictory directions which might explain the high relapse rate observed in anxiety disorders.

Meta-analysis
In this meta-analysis, we compare ABA against AAA, ABC and ABBA designs to identify the strength of different renewal paradigms.First, we analyse the difference between ABA and AAA renewal paradigms by looking at both electrodermal responses and US expectancy as dependent variables.As a measurement of the strength of renewal, we compare differential values for each DV (i.e., value for CS+ minus values for CS-) (1) at the start of renewal and (2) difference values at the start of renewal minus difference values at the end of extinction.
In the second part of the meta-analysis, we compared other, less common renewal paradigms (ABC or ABBA) to the ABA/AAA paradigms.Due to the low number of studies providing direct comparisons, we first compiled studies that directly compared ABA to AAA designs and then used ABA renewal as a benchmark to contrast against renewal observed in ABC and ABBA designs.
Fig. 3 illustrates the effect sizes from studies comparing ABA to AAA designs using skin conductance responses and contrasting responses from the end of extinction with those during early renewal (k = 6, N = 262).There was a significant overall effect: r(261) = 0.46, p <.001, 95 % CI [0.27, 0.62], with higher renewal of fear (larger differential responding early during renewal than late during extinction) in the ABA condition compared to the AAA condition.There was again substantial heterogeneity in the few included effect sizes, Q(5) = 16.31,p =.006, I 2 = 67.12%.Egger's test was not significant, z = 0.63, p =.531.
Fig. 4 illustrates the effect sizes from studies comparing ABA to AAA designs using subjective US expectancy and assessing responses from early renewal only (k = 16, N = 901).There was a significant overall effect: r(900) = 0.35, p <.001, 95 % CI [0.27, 0.42], with higher renewal of US expectancy in the ABA condition compared to the AAA condition.There was minor heterogeneity in the included effect sizes, Q(15) = 21.64,p =.118, I 2 = 34.10%.Egger's test was not significant, z = 1.18, p =.239.
Fig. 5 illustrates the effect sizes from studies comparing ABA to AAA designs using subjective US expectancy and assessing renewal in reference to US expectancy at the end of extinction (k = 14, N = 680).There was a significant overall effect: r(679) = 0.54, p <.001, 95 % CI [0.38, 0.68], with higher renewal of fear in the ABA condition compared to the AAA condition.There was very high heterogeneity in the included effect sizes, Q(13) = 82.68,p <.001, I 2 = 86.39%.Egger's test was not significant, z = − 0.25, p =.801.

ABC compared ABA
Our systematic review was only able to identify two studies that compared ABA or AAA to ABC paradigms, both of which used US expectancy as the outcome measure and assessed renewal in reference to differential US expectancy at the end of extinction.Fig. 6 illustrates the effect sizes from these studies (k = 2, N = 81).There was a significant overall effect: r(80) = 0.45, p <.001, 95 % CI [0.25, 0.61], with higher renewal of fear in the ABA condition compared to the ABC condition.The two studies were highly consistent, with Q(1) = 0.01, p =.910, I 2 = 0.00 %.Egger's test could not be conducted due to only two studies being included in the analysis.
Only one study compared the ABC design to AAA or ABA designs using electrodermal responses as the outcome measure.This study reported no renewal effect for either design (r <.14, effect size not reported).

ABBA compared to ABA
No studies were identified that directly compared ABBA to ABA or AAA experimental designs.Therefore, for this analysis, we pooled effect sizes from studies using each of these paradigms to conduct a comparison.We were able to extract data from 7 studies that used skin conductance responses during the ABBA paradigm (k = 7, N = 313), which had a pooled effect size of r(312) = 0.41, p =.006, 95 % CI [0.12, 0.71], and indicated renewal of fear during the renewal phase of the ABBA design.Skin conductance data for the ABA design could only be extracted from 6 of the compiled studies (k = 6, N = 129), resulting in a pooled effect size of r(128) = 0.61, p <.001, 95 % CI [0.28, 0.94].A comparison of the two correlation coefficients was performed using Fisher's z transformation resulted in a significantly higher renewal effect for the ABA compared to the ABBA design, p =.010.

Discussion
We reviewed the literature on renewal in human fear conditioning across the classical ABA paradigm and other layouts.As shown by metaanalysis, compared with an AAA control, renewal of conditioned fear is a relatively robust phenomenon reliably observed in the ABA design across studies, despite the variability of outcome measures and analysis practices.This holds across physiological and self-report outcome measures (electrodermal responses and US expectancy) and for both the analysis of enhanced differential responding at the start of the renewal test and the comparison of differential responding at the end of extinction to that at the start of the renewal test.
Current findings generally align with the broad perspective that renewal is an effective method for measuring return of fear in humans.However, there are notable exceptions of occasional inconsistency in finding a renewal effect when measured by electrodermal responses (Haesen & Vervliet, 2015;Leer et al., 2018;Muench et al., 2021;Pace-Schott et al., 2014).Another interesting finding to note is that renewal in the ABBA paradigm appears less strong than in the ABA paradigm, suggesting that extinction recall sessions reduce the renewal of fear, despite the context change between extinction recall and subsequent renewal test sessions.To the best of our knowledge, no previous Y.Wang et al. study has made a direct comparison between the two paradigms to succinctly investigate the effect of extinction recall on the renewal of fear.
Theoretically, renewal aligns with the concept of contextual conditioning, where environmental context serves as a contextual cue of the US (Ameli et al., 2001).Conditioned fear responses can generalize to contextual stimuli, such as the lab environment.Once the CS+ has been paired with the US, contextual sensitization becomes stimulus-specific (Hamm & Vaitl, 1996).Therefore, the CS+ and the contextual cue can be regarded as a compound cue predicting the USs.When the compound cue of CS+ and the acquisition context is reintroduced during the renewal test, participants anticipate subsequent US presentations.Consequently, the extinction learning acquired in the extinction context fails to generalise.
When comparing across studies, large heterogeneity in effect sizes was noted for electrodermal responses and some studies reporting US expectancy data.Potential reasons for this heterogeneity include a lack of consistency in experimental design across fear conditioning studies.It is well documented that fear conditioning studies vary greatly in set up, measurement and paradigms used (e.g., Lonsdorf et al., 2019).This means that the studies included in the present meta-analysis employed varying numbers of trials in extinction, different types of US (i.e., shock or scream) and different intervals between the end of extinction and the start of the renewal test.Given this, the high heterogeneity of effect sizes across studies is not surprising.Inspection of the experimental designs of the studies that employed electrodermal measures did not suggest factors that may have mediated the finding of large or small effect sizes in renewal (refer to https://osf.io/ckgeh/?view_only=f9e4d7d5295c4e cf99a8ce4051d3b569 for a direct comparison of study designs).Maybe more interesting is the observation that when comparing the strength of renewal in US expectancy, for 'start of renewal' data only, the heterogeneity is minor.This suggests that despite the differences across the 16 studies included in this comparison; the pattern of results was very consistent, and that US expectancy is less affected by differences in the participants sampled or study setup.This review demonstrates important boundary conditions that may preclude the occurrence of fear renewal in human fear conditioning and suggests that it is beneficial to include several variables to measure the return of fear.
Only a small number of studies were identified that compare ABA with other types of renewal designs: two that compared ABC with ABA  designs and one that compared: ABC to ABA or AAA designs.Both studies comparing ABC with ABA designs showed a stronger renewal of fear indexed by US expectancy in the ABA design.The study that compared ABC to ABA or AAA designs found no differences in renewal between ABC and AAA.ABBA designs were directly compared with ABA in one study using US expectancy and again ABA appeared to produce stronger renewal.No direct comparisons between ABBA and ABA designs were identified reporting electrodermal responses, however, when comparing studies using ABBA (N = 7) and ABA designs (N = 6), a significantly higher renewal effect for ABA designs was noted.Overall, it appears that the classic ABA design is the most effective version of the renewal paradigm.
Only limited information is available as to whether conducting extinction on the same day as acquisition or on a subsequent day affects renewal.Due to the complexity of the processes underlying fear conditioning, it is difficult to investigate the effect of delaying extinction without affecting other factors that also influence conditioning.Initial studies suggest that delaying extinction might reduce renewal (Merz et al., 2016).Given its importance for potential applications, more studies are needed that investigate the effects of the timing of extinction  Y. Wang et al. training on the return of fear.After all, it is common that therapy, the intervention that extinction training is thought to model, occurs well after the acquisition of fear.A better understanding of the effects of delayed extinction may help with the design of more effective intervention programs.
Individual difference factors such as gender, age and personality traits also influence the strength of renewal.While females show overall smaller electrodermal responses, research is on whether the renewal differs between males and females with opposing findings presented.Age affects extinction recall with older individuals showing impaired extinction recall and higher levels of extraversion predicted less renewal as indicated by electrodermal responses.The interplay of the factors involved in fear conditioning is likely part of the reason why the conditions under which fear returns are only poorly understood.Just like individual difference factors, lifestyle choices can also impact the return of fear.For example, stress impacted renewal in opposing ways depending on whether it was experienced before or after extinction training (Cisler et al., 2020).Alcohol slowed extinction learning and reduced renewal (Bisby et al., 2015) while acute cannabis consumption (Das et al., 2013) and exercise (Crombie et al., 2021) did not affect renewal.Considering the ubiquity of these lifestyle choices, more research into their effect on the return of fear would be greatly beneficial.

Implications for practice
We systematically reviewed several factors that have the potential to influence the return of fear due to renewal.Overall, the renewal paradigm was shown to be effective in inducing the return of fear across a variety of measures.The evidence on the effect of the number of extinction trials on renewal is mixed and future studies should address this by systematically studying the impact of extending or shortening the extinction period.Similarly, evidence for the effectiveness of presenting safety cues during fear conditioning is mixed.Safety cues can strengthen extinction if carried across to the renewal phase, however, if a safety cue is only present during the extinction phase, it can work as a 'protection from extinction' cue and lead to a higher return of fear when the cue is not present.This is of relevance in a clinical setting where extinction is usually achieved in a controlled therapy environment.It is possible that a variety of factors in this setting function as cues that cannot be taken into the 'real world' afterwards and hence might effectively work as 'protection from extinction' cues which ultimately lead to higher anxiety when leaving the therapeutic setting.It might be advisable to deliberately create safety cues that the patient can take with them into the 'renewal' context of the 'real world' afterwards to re-create the safety cue context.
In addition to safety cues, the nature and variability of the extinction context appear to play a significant role.Of particular interest for a clinical setting is that renewal is reduced when the contexts of extinction and renewal test are similar.This suggests that therapy should be conducted in a setting similar to the patient's everyday life rather than in a distinct therapy setting.In addition, using multiple, dissimilar contexts can further reduce the strength of renewal.This suggests that clinicians might need to consider conducting exposure therapy in a variety of settings to improve long-term therapeutic outcomes.Another finding with potential clinical implications is that the inclusion of an extinction recall session reduced renewal.This suggests that the inclusion of booster sessions might help to reduce relapse.Similarly, since less renewal is observed in ABC paradigms than in ABA paradigms, we can conclude that a novel context induces less renewal than the acquisition context after extinction.Taking this into consideration, clinicians could encourage clients to modify their environment after therapy, so that it becomes more distinct from the environment where fear was initially acquired.
Anxiety disorders are often co-morbid and therefore it is important to understand how processes underlying fear conditioning might be altered in clinical populations.Patients with schizophrenia show differences in renewal with lower renewal (as measured by electrodermal responses) or no renewal reported compared with healthy controls (Holt et al., 2009(Holt et al., , 2012)).On the other hand, the impact of obsessive-compulsive disorder on renewal is unclear with stronger renewal found in one study (Fyer et al., 2020) and no difference in another (McLaughlin et al., 2015).Results from individuals with PTSD were clearer in that PTSD patients have impaired extinction recall and stronger renewal of fear.Conversely, social anxiety disorder and anorexia did not appear to alter renewal.Similarly, the treatment for other disorders can also affect fear.For example, L-Dopa has been shown to reduce renewal at certain dosages if given at certain times (Cisler et al., 2020).Future studies are needed to systematically investigate the differences in renewal between healthy controls and clinical populations to ensure treatments are optimised for each client.
Timing of the experiment within the day might also impact the extent of renewal since levels of anxiety and the size of electrodermal responses significantly differ across the day (Pace-Schott et al., 2014).Online studies can provide data from more varying time points and offer a way to systematically assess the impact of time of day on fear renewal since fear conditioning has been shown to be effectively delivered via a smartphone app (Purves et al., 2019).For the therapeutic setting especially, reported effects of the timing of extinction are important as it would be beneficial to know if, for example, morning or afternoon sessions of exposure therapy were more effective.Further research is needed to clarify the impact of these factors.
It is important to note that the variability of the effect sizes across studies was high.This heterogeneity of effect sizes has the potential to influence the robustness of the results that we have reported here, which is a limitation of the current meta-analysis.In the future and with more consistently designed experiments, it is possible that the estimates that we have provided here could be shown to not be accurate.Nevertheless, the estimates provided in the present manuscript are our current best attempts at understanding the veracity of renewal effects in human fear conditioning.
This systematic review and meta-analysis provide several avenues for further research that will aid the understanding of the processes underlying the renewal of conditional fear and may ultimately benefit the treatment of individuals with anxiety disorders or trauma.In particular, the use of safety cues and the impact of timing of extinction within a day should be investigated systematically to optimise therapeutic interventions.Furthermore, this study supports existing calls for consistent fear conditioning protocols (Bach et al., 2023) to facilitate the comparison of results across studies.

Y
.Wang et al.   and extinction in context B on day 1 and extinction recall in context B and renewal in context A 24 h later,Milad et al. (2005) found renewal in skin conductance.Both,Purves et al. (2019) andMilad et al. (

Fig. 2 .
Fig. 2. Effect sizes from studies assessing differential electrodermal responding during early renewal: AAA versus ABA designs (k = 9).Note.The size of the squares indicates the weight of that particular study to the overall effect, determined by the sample size of the studies.

Fig. 3 .
Fig. 3. Effect sizes from studies contrasting differential electrodermal responding late during extinction and during early renewal: AAA versus ABA designs (k = 6).Note.The size of the squares indicates the weight of that particular study to the overall effect, determined by the sample size of the studies.

Fig. 4 .
Fig. 4. Effect sizes from studies assessing differential US expectancy during early renewal: AAA versus ABA designs (k = 16).Note.Krisch et al., (2018) Experiment 1 represents groups with 12 extinction trials and Experiment 2 represents groups with 36 extinction trials.The size of the squares indicates the weight of that particular study to the overall effect, determined by the sample size of the studies.

Fig. 5 .
Fig. 5. Effect sizes from studies contrasting differential US expectancy late during extinction and during early renewal: AAA versus ABA designs (k = 14).Note.Krisch et al., (2018) Experiment 1 represents groups with 12 extinction trials and Experiment 2 represents groups with 36 extinction trials.The size of the squares indicates the weight of that particular study to the overall effect, determined by the sample size of the studies.

Fig. 6 .
Fig. 6.Effect sizes from studies contrasting differential US expectancy during late extinction and early renewal: ABA versus ABC designs (k = 2).