The impact of voluntariness of apologies on victims’ responses in restorative justice: findings of a quantitative study

Abstract

Apologies are ordered in law without certainty about whether or not recipients perceive ordered and voluntary apologies differently. This exploratory study investigates whether or not the voluntariness of apologies influences recipients’ perceptions of their sincerity, acceptance of apologies, willingness to forgive and intended retributive behaviour. We manipulated the voluntariness of apologies whilst considering offender (age, gender, ethnicity and prior wrongful behaviour) and offence (seriousness) characteristics in 3 studies (ns = 164, 121, 236). Participants adopting the role of a hypothetical victim received either a voluntary or an ordered apology. The voluntary apologies were found to have a significantly more positive impact than the ordered apologies on acceptance and perception of sincerity in all 3 studies and on forgiveness in 2 studies, but did not significantly change participants’ retributive behaviour in any study. Age was the only other variable found to make a significant difference, with younger offenders’ apologies being rated as sincerer.

Apologies’ potential restorative function (for a discussion on why this might be the case, see Allan et al., 2017) appears to be a major reason for law reformers’ increased push for using them during the twenty-first century in domestic (criminal and civil) and international law (see Latif, 2001). Courts in many parts of the world are also increasingly prepared to order defendants to offer apologies to victims (for a discussion, see: Carroll, 2021; van Dijck, 2017). Mediators and lawyers involved in settlement discussions also frequently recommend to parties that they should apologize (see Levi, 1997; Schneider, 2000).

Some lawyers and scholars nevertheless believe that the compulsion involved when an apology is ordered (i.e. required by an authority figure such as a judge or mediator) is contrary to the fundamental nature of an apology, that such apologies have no social, moral or legal value and/or that they are ineffective as a remedy (see Latif, 2001; Taft, 2000, 2005; but for a contrary view, see White, 2006). One crucial aspect of scholars’ and courts’ objections to ordered apologies is that they lead to insincerity (see Proeve & Tudor, 2010): ‘an apology should not be compelled by an order of the Court because that compels a person to articulate a sentiment that is not genuinely held’ (Eatock v. Bolt, 2011, para 465).

Victims’ views might, however, differ from those of scholars and lawyers because their perceptions of apologies often differ from those of observers (Hashimoto & Karasawa, 2016) and are complex and context bound (Allan & Carroll, 2017; Allan et al., 2006; Robbennolt, 2003, 2013; Slocum et al., 2011; Strickland et al., 2018). Researchers who interviewed parties to equal opportunity matters found that they preferred voluntary apologies but thought that non-voluntary apologies are valuable because they still affirm that the victims have been wronged (Allan et al., 2010). Researchers using participants adopting the role of a hypothetical victim in juvenile justice (e.g. Allan et al., 2014) or mediation and civil settlement (Robbennolt, 2013) situations found that whilst victims prefer voluntary apologies, they also see value in ordered apologies. Psychologists have explored people’s responses to voluntary and ordered apologies, and two laboratory studies are particularly relevant.

Risen and Gilovich (2007) studied undergraduate victims’ reactions to sincere (given spontaneously) and insincere (given after an order) apologies. The participants (victims) undertook a team task with a confederate of the experimenter (the wrongdoer) whom they believed was a fellow participant. The team received a small reward, depending on its performance. Another participant and a second confederate observed the execution of the task. The confederate wrongdoer deliberately compromised the team’s effort and then apologized either spontaneously or after the confederate observer ordered them to apologize. The researchers measured the impact of the apology by asking the participants what impression the wrongdoer had made (likeable, selfish, sincere, arrogant, rude, compassionate) and what percentage of the team reward the wrongdoer should receive (punishment). The researchers found that the voluntariness of the apology did not influence the victims’ evaluation of the wrongdoers or how they apportioned the reward, but that the observers liked the wrongdoers significantly less and wanted to punish them more in the ordered apology condition. A possible explanation for the absence of a significant difference between the voluntary and ordered apologies in the victims’ responses is that the element of coercion was weak: the instruction to apologize came from a peer who was a co-researcher and there was no consequence for not apologizing (Risen & Gilovich, 2007).

Jehle et al. (2012) tried to remedy some of the limitations of the Risen and Gilovich (2007) study in the laboratory by asking participants to solve 10 anagrams, of which 3 were unsolvable. The wrongdoer insulted the participants (victims) in Study 1 with ‘a verbal insult of their intellectual ability’ when they failed to successfully complete the task. The wrongdoer then either did not apologize, apologized voluntarily, offered an implicitly ordered apology or offered an explicitly ordered apology. The implicitly ordered apology was one where a high-status person (the experimenter) ordered the wrongdoer to apologize but with no explicit consequences if it did not happen; the explicitly ordered apology was one where a high-status person ordered the wrongdoer to apologize with a threat of negative consequences should they disobey. The researchers measured the impact of the different apologies by asking the victims to indicate their overall perception of the wrongdoer and particularly how courteous, likeable, responsible and respectful they were. The victims also recommended the level of punishment that the wrongdoer should receive. The researchers found that the victims distinguished between the types of apology and that those who received ordered apologies treated the wrongdoers harsher (less favorable assessment and more punishment) than those who received voluntary apologies, but still less harshly than those wrongdoers who made no apology. The victims rated the wrongdoers most unfavorably and gave them the harshest punishments in the no apology condition (Jehle et al., 2012). The participants (hypothetical victims) in Study 2 read a record of what had happened in Study 1 (including the insult and apology) and had to imagine that they were the victim. These hypothetical victims then had to do the same as the actual victims in Study 1 – that is, rate the wrongdoer and suggest a level of punishment. The hypothetical victims in Study 2 did not distinguish between the different apology types whereas the actual victims in Study 1 – unlike those of Risen and Gilovich (2007) – did distinguish between them.

The applicability of the findings of these studies within law is, however, limited for several reasons. First, the dependent variables in both studies (i.e. how likeable the wrongdoers were) are not necessarily directly relevant in law. Second, the researchers did not consider other variables that could have influenced the victims’ perceptions of the apologies. Previous research for instance suggests that characteristics of offenders such as age (see Allan et al., 2014), ethnicity (see Hewstone et al., 2004; Komen & Van Schooten, 2009), gender (see Thomas et al., 2008; Walfisch et al., 2013) and history of prior wrongful behaviour (see Wooten, 2009) as well as characteristics of the offence (e.g. the seriousness thereof; see Bennett & Earwaker, 1994) can influence victims’ assessments of apologies. Finally, Risen and Gilovich’s (2007) and Jehle et al.’s (2012) participants were all undergraduate students and are therefore not representative of all victims. These limitations reflect some of the difficulties that researchers have when they want to conduct authentic studies about the impact of the voluntariness of apologies in law – and there are two further difficulties.

Researchers’ first problem is that there are few areas in law where apologies are offered and/or ordered frequently enough to study their impact. An exception is, however, the restorative justice programs that became mainstream in juvenile justice at the turn of the century and have since been expanded for use with adult offenders (Larsen, 2014). These reconciliation processes are called by different names in different jurisdictions, such as family group conferences (e.g. Blecher, 2011) or victim offender mediation (e.g. Choi & Severson, 2009), but apologies generally play a central part in most of them (Dhami, 2012). The relevant enabling legislation often gives apology as one of the outcomes of restorative justice meetings (see Young Offenders Act, 1997, section 52(5)(a)), but in practice facilitators often expect offenders to apologize (Choi & Severson, 2009) or prompt them (Bolitho, 2012; Saulnier & Sivasubramaniam, 2015) and offenders feel obliged to do so (Blecher, 2011). Apologies are therefore common and researchers have consequently been able to collect useful information regarding the impact of the voluntariness of apologies in restorative justice (e.g. Choi & Severson, 2009; Dhami, 2016a, 2016b). Victims do not in general feel that the apologies offered by offenders in restorative justice settings are ‘sincere and heartfelt enough’ (Choi & Severson, 2009, p. 818) and do not feel vindicated or restored (Choi et al., 2012). They nevertheless feel under pressure to accept the apologies that are offered to them (Choi & Gilbert, 2010; Dhami, 2016b).

The second problem that researchers face is finding dependent variables which can serve as indicators of the impact that the voluntariness of apologies has in law. Those undertaking research about the effects of apology have used ratings of sincerity, acceptance and forgiveness (e.g. Allan et al., 2015; Kirchhoff et al., 2012; Walfisch et al., 2013) along with punishment recommendations (Darby & Schlenker, 1982; Jehle et al., 2012; Risen & Gilovich, 2007) either individually or in combination as dependent variables. We believe it is appropriate to use these variables to study the impact of the voluntariness of apologies in the restorative justice context as well. Sincerity comes up in virtually every study of apology in restorative justice (e.g. Allan et al., 2014; for a review of some of the relevant articles, see Choi et al., 2012). Retzinger and Scheff (1996) have described the link between apology and forgiveness as a central aspect of the restorative process, and Dhami (2012) in her study of mediation cases found that victims are more likely to forgive in the presence than absence of an apology from the perpetrator, and if the apology is more comprehensive. Forgiveness is therefore an important variable to consider, but Dhami (2012) found that victims’ forgiveness is independent of whether or not apologies are accepted, suggesting that acceptance of an apology is a separate variable. We therefore followed Strickland et al. (2018) and measured acceptance as a separate variable. The use of punishment recommendations as an outcome variable in restorative justice studies might appear odd because the aim of restorative justice is to repair harm rather than punish the offender (e.g. Larsen, 2014). Punishment and restoration cannot, however, be separated from each other (Daly, 2013). Restorative justice is, essentially, ‘a process of bringing together the individuals who have been affected by an offence and having them agree on how to repair the harm caused by the crime’ (Braithwaite, 1999, pp. 1743–1744), and this can include ‘setting reparation and sanctions’ (Foley, 2013, p. 130).

Our aim with the studies we report on in the present paper is to examine whether or not participants adopting the role of victims distinguish between voluntary and ordered apologies in restorative justice proceedings whilst considering victim and offence characteristics that might influence their perceptions.

Methodology

We undertook three studies to investigate the impact of the voluntariness of an apology on victims’ responses while varying five other variables, creating eight experimental conditions in each study. In Study 1 we varied the age and the gender of the offender, in Study 2 the seriousness of the offence and the prior wrongful behaviour of the offender and in Study 3 the ethnicity of the offender.

Research design

All the studies used a quantitative, between-subjects approach with instructions to participants to adopt the role of the victim in the offence scenarios presented to them using the online Qualtrics data collection platform. All the studies also included the same four dependent variables, namely sincerity, acceptance, forgiveness and intended retributive behaviour.

The participants provided responses for the first three dependent variables on 5-point Likert-type scales and made a recommendation for the number of community hours that should be served on a sliding scale of 0 to 15 hours for the fourth (see Table 1).

Table 1.

Description of dependent variables.

Variable	Description
Sincerity	On a scale of 1 to 5 with 1 being strongly disagree and 5 being strongly agree: Offender’s apology was sincere.
Acceptance	On a scale of 1 to 5 with 1 being fully reject and 5 being fully accept: Please indicate the extent to which you accept the offender’s apology.
Forgiveness	On a scale of 1 to 5 with 1 being not at all and 5 being completely: Please indicate the extent to which you forgive the offender.
Community hours	If you could recommend the number of hours of community work that the offender should do, between 0 and 15, how many hours do you recommend?

Note. Sam as a gender-neutral name was used for the offender in all the vignettes.

They then provided demographic information such as their age, country of residence and previous participation in juvenile justice conferences. Finally, they had to answer questions in a manipulation check to confirm that they had noticed the type of apology that they received and the information provided about the independent variables of interest in the study (e.g. age and gender in Study 1). The manipulation checks were added because participants vary in the diligence with which they apply themselves to research instructions, with some missing pertinent details, particularly when the manipulations involve subtle word variations (Foschi, 2007; Hauser et al., 2018).

Material

The basic vignette, which was the same in all three studies, depicted a juvenile committing an opportunistic and intentional theft of a mobile (cellular) telephone at a café. The offender was immediately apprehended and the participants had to imagine themselves as the victim, who was now attending a juvenile justice mediation session. The details of the vignette (e.g. the age and gender of the offender) were amended as required in each study. The wording of the apology that the juvenile offered was identical for both the ordered and the voluntary apology in all three studies and consisted of the minimum components of Slocum et al.’s (2011) multi-dimensional theory of apology. The offender said: ‘I’m sorry for taking your phone. I know it was wrong because it was stealing from you. I’ve learned my lesson and I won’t do it again.’ A voluntary apology was one that the offender offered spontaneously, whilst the offender offered the ordered apology in response to the facilitator’s instructions. An independent researcher with knowledge of the Western Australian juvenile justice system and the apology literature reviewed the vignettes (including the apology wording) and deemed them as appropriate and credible.

Six community members and students then pre-tested the survey by completing it. The pre-test confirmed that the survey format was accessible and user friendly, and that the vignette is credible. It also confirmed that the content was clear and effective in enabling the participants to situate themselves as the victim in the scenario, controlling for the potential confound of observers perceiving apologies differently from victims, as found in previous research (Risen & Gilovich, 2007).

Participants and procedure

We used a snowballing method to recruit participants by sending email invitations to our personal and university networks and contacts, and by posting on private and publicly viewable websites. We encouraged recipients to forward the invitation and survey link to their own networks, with the aim of optimizing the representativeness of the sample. We only included Australian residents over 18 years of age with no previous experience of youth justice conferencing, and in the third study we only included those of Anglo-Australian heritage.

The studies are completely anonymous as the Qualtrics system assigned potential participants to a vignette. Each participant read only one vignette in order to minimize the demand characteristics that might have arisen if participants had been presented with more than one version, thereby decreasing the risk of artificial contrasts between the different experimental conditions. After reading the apology that the offender had offered, the participants used a 5-point Likert-scale to rate the offender’s sincerity, the acceptability of the apology and the extent to which they forgave the offender. The participants also had to indicate how many community hours (from 0 to 15) they thought the offender should complete as a measure of retributive behaviour. Finally, they had to respond to each of the manipulation check questions, which were consecutively displayed on new screens without being able to return to previous screens. We only included the data of the participants who correctly answered all the manipulation check questions.

Study 1

We investigated the effect of apology type (voluntary or ordered), offender age (12 or 16 years old) and offender gender (male or female) on the four dependent variables and used the gender-neutral name of Sam for the vignette. There were 250 responses without missing data, but 86 respondents failed at least one of the manipulation checks so their data were excluded, leaving a final sample of 164 participants. Of these, 67 identified as males, 95 identified as females and 2 did not respond. Their ages ranged from 18 to 74 years. Table 2 details the descriptive statistics for each dependent variable.

Table 2.

Descriptive statistics for offender gender × offender age × apology type for all dependent variables (n = 164).

Apology	Voluntary				Ordered
Offender age (years)	12		16		12		16
Offender gender	M	SD	M	SD	M	SD	M	SD
Sinceritya
Male	3.74	0.73	3.50	1.15	3.19	0.83	2.71	1.01
Female	3.56	0.96	3.15	1.04	3.00	0.94	2.53	1.23
Acceptanceb
Male	3.89	0.66	3.89	1.18	3.70	1.03	3.24	1.14
Female	4.00	1.16	3.35	0.93	3.19	1.27	2.82	1.29
Forgivenessc
Male	3.74	0.93	3.72	1.02	3.30	1.10	3.14	1.24
Female	3.75	1.00	3.55	1.10	3.42	1.39	2.76	1.25
Community hoursd
Male	9.11	4.16	11.78	3.67	10.07	4.20	10.86	4.63
Female	8.19	4.10	10.35	4.78	9.96	5.41	11.06	4.34

Note. ^aAnchors: 1 = strongly disagree, 5 = strongly agree; ^bAnchors: 1 = fully reject, 5 = fully accept; ^cAnchors: 1 = not at all, 5 = completely; ^dSliding scale: 0–15 hours.

Four factorial between-group analysis of variance (ANOVA) tests were conducted to investigate the effects of apology type, offender age and offender gender on the four dependent variables, namely sincerity, acceptance, forgiveness and community hours. An overall α level of .05 was used – but in order to reduce the risk of making a Type 1 error, a Bonferroni correction was used to account for the familywise error rate across multiple comparisons, giving an adjusted α of .0125 for each dependent variable (see Keppel & Wickens, 2004). To further address the potential for Type 1 errors in the comparisons between experimental conditions, we specified Bonferroni adjustments of the p-values in the statistical package used for the analyses. The final cell sizes differ because we excluded respondents who failed at least one of the manipulation check questions (see Table 3).

Table 3.

Cell sizes for gender × age × apology type groups for all dependent variables (n = 164).

Apology	Voluntary			Ordered
Offender gender	Male	Female	Total	Male	Female	Total
Offender age (years)
12	19	16	35	27	26	53
16	18	20	38	21	17	38
Total	37	36	73	48	43	91

Note. Group sizes differ as a result of the exclusion of cases that failed the apology manipulation check or of missing values for one or more of the dependent variables.

The ANOVA results indicate that there is a statistically significant main effect for type of apology for sincerity, acceptance and forgiveness, but not for community hours (see Table 4). Voluntary apologies (M = 3.49) were perceived as sincerer than ordered apologies (M = 2.86). Participants were more accepting of voluntary apologies (M = 3.78) than ordered apologies (M = 3.24) and more forgiving of the offenders who offered voluntary apologies (M = 3.69) compared to those who offered ordered apologies (M = 3.16). The main effect for age is also significant for sincerity, with the apologies of the 12-year-olds (M = 3.37) regarded as sincerer than those of the 16-year-olds (M = 2.97). No other statistically significant main effects or statistically significant interactions were found.

Table 4.

ANOVA results for offender gender × offender age × apology type for all dependent variables (n = 164).

	Sincerity			Acceptance			Forgiveness			Community hours
	F	p	η 2	F	p	η 2	F	p	η 2	F	p	η 2
Apology	16.27	< .001*	.09	9.76	.002*	.06	8.52	.004*	.05	0.79	.376	.01
Age	6.48	.012*a	.04	4.58	.034	.03	1.97	.162	.01	5.55	.020	.03
Gender	2.05	.154	.01	3.81	.053	.02	0.32	.575	.00	0.63	.430	.00
Apology × Age	0.22	.641	.00	0.07	.798	.00	0.67	.415	.00	1.07	.302	.01
Apology × Gender	0.06	.805	.00	0.50	.481	.00	0.02	.900	.00	0.73	.394	.01
Age × Gender	0.08	.779	.00	0.62	.433	.00	0.89	.346	.01	0.01	.945	.00
Apology × Age × Gender	0.08	.779	.00	1.13	.289	.01	0.19	.662	.00	0.08	.773	.00

Note. *Significant at α = .0125 following Bonferroni correction; ^ap = .0119; df = 1, 156 for all dependent variables.

Study 2

The focus of the second study is the impacts of apology type, prior wrongful behaviour and seriousness of the offence on the four dependent variables. The offender was portrayed as Sam, a 16-year-old male, as no significant gender differences were found in the previous study and a significant age difference (with a small effect size) was only found for sincerity. Prior wrongful behaviour was operationalized as prior police contact (‘this is the second time Sam has done this’ versus ‘I think this is Sam’s first offence’) and the seriousness of the crime as the cost of the phone (an old AU$100 phone versus a new AU$1000 phone).

There were 173 responses without missing data, but the data of 52 respondents who failed at least one manipulation check were excluded, leaving a final sample of 121 participants. Of these, 53 identified as males, 66 identified as females and 2 identified as other. Their ages ranged from 18 to 77 years. Table 5 details the descriptive statistics for each dependent variable.

Table 5.

Descriptive statistics for cost × contact × apology type for all dependent variables (n = 121).

Apology	Voluntary				Ordered
Prior contact	No		Yes		No		Yes
Cost of phone (AU$)	M	SD	M	SD	M	SD	M	SD
Sinceritya
1000	4.15	0.69	3.45	1.04	2.36	1.01	2.05	1.15
100	3.42	1.12	3.07	1.28	2.50	1.31	2.24	1.09
Acceptanceb
1000	3.69	1.38	3.82	1.08	2.86	1.29	2.30	1.42
100	3.68	1.16	3.27	1.44	2.92	1.44	2.71	1.57
Forgivenessc
1000	3.62	1.56	3.73	1.10	2.79	1.25	2.45	1.47
100	3.79	1.32	2.93	1.49	2.92	1.56	2.71	1.45
Community hoursd
1000	9.09	6.19	10.04	5.55	11.89	3.19	13.65	2.25
100	9.07	5.89	12.00	5.10	10.25	4.67	10.65	5.13

Note. ^aAnchors: 1 = strongly disagree, 5 = strongly agree; ^bAnchors: 1 = fully reject, 5 = fully accept; ^cAnchors: 1 = not at all, 5 = completely; ^dSliding scale: 0–15 hours.

Four factorial between-groups ANOVA tests were conducted to investigate the effects of apology type, prior police contact and cost of the stolen phone on the four dependent variables. An overall α level of .05 was used, but as with the previous study, a Bonferroni adjusted α of .0125 was used for each dependent variable. The final cell sizes differ as a result of the exclusion of the respondents that failed at least one of the manipulation check questions (see Table 6).

Table 6.

Cell sizes for cost × contact × apology type groups for all dependent variables (n = 121).

Apology	Voluntary			Ordered
Prior contact	No	Yes	Total	No	Yes	Total
Cost of phone (AU$)
1000	13	11	24	14	20	34
100	19	15	34	12	17	29
Total	32	26	58	26	37	63

Note. Group sizes differ as a result of the exclusion of cases that failed the apology manipulation check or of missing values for one or more of the dependent variables.

The ANOVA results indicate that there is a statistically significant effect for type of apology for sincerity, acceptance and forgiveness, but not for community hours (see Table 7).

Table 7.

ANOVA results for cost × contact × apology type for all dependent variables (n = 121).

	Sincerity			Acceptance			Forgiveness			Community hours
	F	P	η 2	F	p	η 2	F	p	η 2	F	p	η 2
Apology	36.43	< .001*	.24	13.26	< .001*	.11	9.38	.003*	.08	3.02	.085	.03
Cost	0.93	.336	.01	0.01	.926	.00	0.05	.824	.00	0.56	.455	.01
Contact	3.92	.050	.03	1.10	.297	.01	1.52	.220	.01	2.83	.095	.02
Apology × Cost	3.12	.080	.03	1.03	.313	.01	0.92	.339	.01	3.35	.070	.03
Apology × Contact	0.35	.558	.00	0.22	.639	.00	0.04	.851	.00	0.02	.635	.00
Cost × Contact	0.22	.638	.00	0.04	.846	.00	0.65	.423	.01	0.03	.862	.00
Apology × Cost × Contact	0.14	.713	.00	0.77	.381	.01	1.09	.299	.01	0.87	.353	.01

Note. *Significant at α = .0125 following Bonferroni correction; df = 1, 113 for all dependent variables.

Voluntary apologies (M = 3.52) were perceived as sincerer than ordered apologies (M = 2.29). Participants were more accepting of voluntary apologies (M = 3.62) than ordered apologies (M = 2.77) and more forgiving of offenders who offered voluntary apologies (M = 3.52) compared to ordered apologies (M = 2.72). No other statistically significant main effects or statistically significant interactions were found.

Study 3

We investigated the effects of apology type and ethnicity on the four dependent variables in the third study. The offender was described as a 16-year-old male without any previous offences who had taken a new AU$1000 mobile phone, as there are no significant findings for prior wrongful behaviour and offence seriousness in Study 2. The offender was described as either Anglo-Australian, Aboriginal, African or Middle Eastern, with appropriate names chosen for each offender.

There were 378 responses without missing data, but we excluded the data of 53 respondents who failed the apology manipulation check only and 24 who failed the ethnicity manipulation check or, in some cases, both checks. This study required participants to be of Anglo-Australian heritage, therefore the 65 respondents who identified as being of a different heritage were also excluded.

The final data set includes 236 participants (184 female, 52 male, 1 other) aged between 18 and 92 years. The final cell sizes are very similar, ranging from 26 to 33, despite the exclusion of the respondents that failed at least one of the manipulation check questions (see Table 8).

Table 8.

Cell sizes for ethnicity × apology type groups (n = 236).

	Apology
Ethnicity	Voluntary	Ordered	Total
Anglo-Australian	29	29	58
Aboriginal	30	27	57
African	26	31	57
Middle Eastern	33	31	64
Total	118	118	236

Four factorial between-groups ANOVA tests were conducted to investigate the effects of apology type and ethnicity on the four dependent variables. A Bonferroni corrected α of .0125 was adopted for each dependent variable, as with the previous studies.

Table 9 provides the descriptive statistics and the ANOVA results. There is a statistically significant main effect for apology type for the sincerity and acceptance variables. The scores are higher for the voluntary apology (M = 3.58 for sincerity and M = 4.08 for acceptance) than for the ordered apology (M = 3.25 for sincerity and M = 3.72 for acceptance). Apology type was not found to have a statistically significant effect on the forgiveness and community hours variables, no statistically significant main effects were found for ethnicity for any of the dependent variables and no statistically significant interactions were found.

Table 9.

Descriptive statistics and ANOVA results for apology × ethnicity type for all dependent variables (n = 236).

	Apology				Apology			Ethnicity			Apology × Ethnicity
	Voluntary		Ordered
Dependent variable	M	SD	M	SD	F	p	η2	F	p	η2	F	p	η2
Sinceritya					8.79	.003*	.04	1.53	.207	.02	1.74	.160	.02
Anglo-Australian	3.52	0.83	3.24	1.02
Aboriginal	3.77	0.73	3.07	0.92
African	3.23	0.65	3.26	0.89
Middle Eastern	3.76	0.83	3.39	0.84
Acceptanceb					9.46	.002*	.04	0.96	.414	.01	1.94	.124	.03
Anglo-Australian	3.97	0.91	3.66	1.08
Aboriginal	4.30	0.70	3.48	1.01
African	3.85	0.78	3.77	0.92
Middle Eastern	4.15	0.71	3.94	0.89
Forgivenessc					2.25	.135	.01	1.55	.202	.02	0.46	.709	.01
Anglo-Australian	3.76	0.87	3.52	1.21
Aboriginal	3.97	1.07	3.52	1.19
African	3.85	0.93	3.84	1.16
Middle Eastern	4.09	0.91	3.97	0.98
Community hoursd					1.29	.257	.01	0.73	.535	.01	1.20	.310	.02
Anglo-Australian	8.93	3.81	10.10	3.89
Aboriginal	8.30	4.35	9.15	4.72
African	7.58	4.83	9.23	3.34
Middle Eastern	9.21	4.99	8.10	7.76

Note. *Significant at α = .0125 following Bonferroni correction; ^aAnchors: 1 = strongly disagree, 5 = strongly agree; ^bAnchors: 1 = fully reject, 5 = fully accept; ^cAnchors: 1 = not at all, 5 = completely; ^dSliding scale: 0–15 hours; df = 1, 228 for Apology; df = 3, 228 for Ethnicity and Apology × Ethnicity.

Discussion

Our aim was to examine whether or not participants adopting the role of a victim would distinguish between voluntary and ordered apologies in restorative justice proceedings whilst considering victim and offence characteristics that might influence their perceptions. We will first discuss the impact of apology type and then the impact of the victim and offence variables.

Apology type

The impact of apology type (voluntary versus ordered) is similar for sincerity, acceptance and forgiveness in all three studies, but apology type was found to have no effect on retributive behaviour.

Sincerity, acceptance and forgiveness

The ratings for sincerity, acceptance and forgiveness in response to voluntary apologies are generally positive for all the groups in all the studies (mean ratings above the midpoint of 3.00). The only exception is the somewhat lower forgiveness mean of 2.93 for the participants who were given the scenario of the offender with prior police contact who stole a AU$1000 phone in Study 2.

For the ordered apologies, the ratings for sincerity, acceptance and forgiveness vary more in the different studies. In Study 3 the mean ratings for sincerity, acceptance and forgiveness are positive for all the groups, as well as in Study 1 for the 12-year-old offenders where the means range between 3.00 and 3.70. The mean ratings for the 16-year-old offenders are more negative in Study 1, ranging from 2.53 to 3.24, with the majority below the midpoint of 3.00. In Study 2 the means are negative for all the groups. One possible explanation for the differences between the studies is that they are a natural consequence of the different samples. Another is that the results were influenced by the unavoidable minor differences in the study vignettes in respect of the offender (age, gender, prior wrongful behaviour and ethnicity) and offence (value of the mobile phone) variables.

The ANOVA results indicate that the voluntary apologies generally had a significantly more positive impact on the variables than the ordered apologies. The one exception is found in Study 3, where forgiveness is only rated marginally higher for the voluntary apologies. The small effect sizes for acceptance and sincerity in Study 3 show that apology type did not have a large impact on these two variables either, despite the significant results. Apology type was found to have a larger impact in Study 1 and Study 2 (medium to large effect sizes), with the biggest impact on sincerity, followed by acceptance and then forgiveness. Taken together, the three studies indicate that apology type affected sincerity most and forgiveness least. One possible explanation for this is that forgiveness is a complex construct that is possibly influenced by more factors than a simpler construct such as sincerity (see Strang et al., 2014). For instance, sincerity is a relatively straightforward judgment of someone else’s demeanor whereas forgiveness is a deliberate act that is personal to each individual.

The situation is complex when we try to place our findings within the context of the two laboratory studies highlighted earlier. Our findings are contrary to those of Jehle et al. (2012) because our hypothetical victims – unlike their hypothetical victims – responded differently to voluntary and ordered apologies. This difference might be because the transgression in our study (theft) is of a financial and material nature whereas the verbal insult in Jehle et al. involves psychological harm. Jehle et al.’s actual victims did, however, distinguish between apology types, but this in turn is contrary to Risen and Gilovich’s (2007) results for actual victims, where the harm was financial. It is therefore still unclear how actual victims in actual justice settings might respond to voluntary versus ordered apologies based on these combined findings.

Intended retributive behaviour

Neither the apology type nor any of the offender or offence characteristics had a significant effect on the recommended number of community hours that serves as our measure of intended retributive behaviour. An implication of this finding is that it calls into question the hypothesis that apologies reduce victims’ need for vindication or compensation (see Allan, 2007; Vines, 2008). The fact that the voluntariness of the apology had a positive impact on sincerity, acceptance and forgiveness but not punishment suggests that for our hypothetical victims the voluntary apology served a restorative function without replacing the need for punishment (see Braithwaite, 1999; Daly, 2013; Foley, 2013).

Our data do not, however allow us to determine our participants’ underlying thinking. Their allocation of community hours might be because they believe that offenders should still do time for their crime despite offering apologies (see Bibas & Bierschbach, 2004). Alternatively, they might have allocated the community hours because it demonstrates to offenders that there must be restorative behaviour on their part even though they apologised (see Allan et al., 2014). The latter thinking fits well into Slocum et al.’s (2011) theory that emphasises that apologetic behaviour should include restorative behaviour and in this study the hypothetical victims might not have considered the undertaking not to reoffend as sufficient to stand on its own.

The situation regarding intended retributive behaviour is, however, as complex as that with the other dependent variables when we try to place our results within the context of the two laboratory studies. Our finding that apology type did not have a significant impact on the punishment meted out by the hypothetical victims is similar to that of Jehle et al. (2012) for their hypothetical victims. The actual victims in Jehle et al.’s study did, however, distinguish between apology types – but this is contrary to the actual victims in Risen and Gilovich’s (2007) study. It is therefore still unclear whether or not actual victims would perceive punishment in actual justice settings differently in response to voluntary and ordered apologies – especially given that the different forms that harm can take (e.g. financial, physical, psychological) might influence the results of studies and also prompt different responses from victims of different crimes in real-world justice settings, making this an important area for future research.

Offender and offence variables

The offender and offence variables generally did not have a significant impact on sincerity, acceptance or forgiveness – with one exception. The voluntary apology in Study 1 had a more positive impact on sincerity, acceptance and forgiveness for the 12-year-olds than the 16-year-olds, irrespective of offender gender or apology type, although the difference between the two age groups is only significant for sincerity. The effect size is small, but the sincerity result indicates that participants had higher expectations of older than younger offenders, and therefore the possibility exists that victims will be less inclined to consider adults as sincere when they apologize in justice settings.

Leaving aside the age difference, a positive interpretation of the lack of significant results for the other variables is that the offender’s gender, previous wrongful behaviour and ethnicity, along with the seriousness of the offence, did not influence the victims’ responses to their apologies.

The absence of a significant ethnicity effect in Study 3 was unexpected, given the American (Stevenson & Bottoms, 2009) and Dutch (Komen & Van Schooten, 2009) findings that people assessed offenders from minority ethnic groups in the juvenile justice system less favorably than offenders from the majority population group and that inter-group bias might explain this (see Hewstone et al., 2004; Komen & Van Schooten, 2009). The results in Study 3 might have been confounded due to the saliency of ethnicity causing participants to give socially desirable answers. What is positive about this interpretation is that they should also be able to suppress their unconscious bias in actual restorative justice settings if they could do it in this anonymous study. A more negative explanation is that participants read a vignette that lacked the normal visual and auditory cues that would be present in the natural setting (see Peyton & Goei, 2013) and were present in Stevenson and Bottoms’s (2009) and Komen and Van Schooten’s (2009) naturalistic studies. The cues that would have triggered their implicit bias might therefore have been absent (see Greenwald & Krieger, 2006).

Limitations and suggestions for future research

Our decision to discard the responses of the respondents who failed at least one of the validity check questions roused our concerns that the statistical power might be too low. We therefore conducted a power analysis (using an α of .0125 and a power level of .80) and found that to detect medium-sized effects, a sample size of 182 is required for Studies 1 and 2 and a sample size of 244 for Study 3. Our final sample sizes of 164 and 236 for Studies 1 and 3 mean that they are slightly underpowered, and Study 2 even more so (n = 121). Power does not, however, appear to be a serious problem because the effect sizes for these non-significant results for the offender and offence variables (all but the effect of age on sincerity) are very small. Individual offender and offence variables account for a maximum of 3% of the variance in the responses, and sometimes less than 1% – and one can ask whether or not such small effects are of much practical significance.

We do not think the use of a mobile telephone to measure crime severity made a difference in Study 2 but using a telephone as the stolen object might not be effective because the loss of such a telephone today goes beyond a monetary value as it for many people involves the loss of private information that, if it can be replaced, will involve a time-consuming process. The loss of a telephone also makes owners vulnerable to further crimes such as identify theft and/or financial loss.

Our measure of forgiveness was simply to ask participants to what extent they forgave the offender. This has previously been used, either independently or with other indicators, to measure participants’ responses to apologies and to infer apology effectiveness as a result (see Allan et al., 2014; Allan et al., 2015; Jehle et al., 2012). This might, however, not be a sensitive measure in restorative justice research and future researchers should consider using an appropriate contemporary measure of forgiveness, an alternative test of whether victims have forgiven offenders (e.g., one that measures how much ill will or resentment they feel towards the offender) or do a multimodal study that allows them to determine exactly what participants’ understanding of forgiveness is (see Allan et al., 2014). A multi-modal study might have allowed us to better understand the allocation of community hours.

We did not add a no-apology condition because we were interested in analyzing the impacts of the offender and offence variables, but this limits the conclusions that we can draw based on the results of this study. Our use of young offenders in our vignettes further limits the generalization of our findings to adult offenders – and as the participants in the studies were all Australian residents, the findings cannot be generalized beyond Australia. Our participants were furthermore hypothetical victims, and as Jehle et al. (2012) have shown there are differences between how actual and hypothetical victims respond to situations. There are several factors that could explain why people’s responses as actual victims differ from their responses as hypothetical victims. Actual victims respond to the situation whereas hypothetical victims must attempt to anticipate how they would respond – and it is well established that people find it difficult to anticipate how they would in reality respond to actually experiencing a specific event or stimulus (see Fatfouta et al., 2014; McCullough & Witvliet, 2002). Actual victims experience emotions associated with the situation that hypothetical victims do not experience, and it is well known that emotions influence decision-making (see Angie et al., 2011). Actual victims finally might make decisions that they think will make them feel good about themselves and be seen positively by onlookers (see Bennett & Earwaker, 1994), whereas hypothetical victims might be less likely to engage in impression management – especially in anonymous studies.

Future researchers could try to overcome the possibility that participants might give what they consider to be socially desirable responses by measuring the speed of their responses using implicit methodologies involving lexical decision tasks (see Henderson et al., 2009), measures of conditional expectancy (see Watkins et al., 2015), the Implicit Association Test (IAT; Greenwald et al., 2003) or an emotional Stroop test (see Ben-Haim et al., 2016).

Future researchers should, furthermore, ideally undertake studies within actual justice settings – but this is often difficult both practically and ethically, so laboratory studies with live actors in role-play might therefore remain the best option (see Saulnier et al., 2012). The increasing use of video technology in justice settings could nevertheless create new opportunities that should be explored in order to make laboratory studies more authentic.

Conclusion

The results from our three studies indicate that the voluntary apologies generally had a more positive impact on the participants than the ordered apologies, but did not impact intended retributive behaviour in all three studies and forgiveness in Study 3. Our results further indicate that none of the offender and offence variables had a significant impact on the hypothetical victims’ responses to the apologies except for the effect of age on sincerity. Ordered apologies might therefore be less effective in achieving restorative justice goals than voluntary apologies, irrespective of the offender and offence characteristics that we manipulated. The inconsistencies in the findings of Jehle et al. (2012), Risen and Gilovich (2007) and our studies, however, suggest that these findings should be interpreted with caution. The outcomes of our three studies confirm the results of field research in equal opportunity (Allan et al., 2010), juvenile justice (e.g. Allan et al., 2014; Saulnier & Sivasubramaniam, 2015) and mediation and civil settlement (Robbennolt, 2013) situations, indicating that whilst victims prefer voluntary apologies they still see value in ordered apologies. From a psychological perspective there is therefore more evidence to support the continued use of ordered apologies than there is to recommend stopping using them.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Ethical standards

Declaration of conflicts of interest

Alfred Allan has declared no conflicts of interest.

Justine de Mott has declared no conflicts of interest.

Isolde M. Larkins has declared no conflicts of interest.

Laura Turnbull has declared no conflicts of interest.

Tracey Warwick has declared no conflicts of interest.

Lacey Willet has declared no conflicts of interest.

Maria M. Allan has declared no conflicts of interest.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Edith Cowan University institutional committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all individual participants included in the study