Recent linguistic and psychological research has shown that voice can influence listeners’ processing of unrelated stimuli on a semantic level (Hatzidak et al., 2015) and can affect both their implicit and explicit attitudes towards the social membership categories associated with that voice (Pantos, 2014; MacFarlane, 2014). Yet, little is known about the extent to which the audio prime can activate other attitudes that are not directly associated with another semantically. This research takes the first steps towards addressing this gap in the applied context of citizens’ level of perceived fairness and trust in the New Zealand Police. Should attitude priming be activated by perceived ethnicity, this could represent a detrimental and negative reinforcement effect in bicultural Police relations with Māori iwi.

Respondents were randomly allocated to one of four voice conditions (in a factorial design of Māori-NZE ethnicity and gender) or to a text control condition, in an online survey distributed across New Zealand, with a total of 367 responses qualifying for analysis. Attitudes of trust and fairness (Tyler, 2001, 2005; Jackson & Bradford, 2010) were measured across scaled responses to statements, taken from the Citizen Satisfaction Survey (CSS) (Gravitas, 2017) and previous Māori-Police relations focus group research (Maxwell & Smith, 1998; Te Whaiti & Roguski, 1998). These groupings were verified by Cronbach’s alpha and Principle Component Analysis - with trust representing a more distant response variable to the perceived ethnicity priming stimuli, whilst fairness was assumed to be more semantically associated with the stimuli prime. Two baseline measures of trust and fairness, adapted from the CSS, were also considered. An exploratory analysis was performed as exampled by Tagliamonte & Baayen, (2012); using two disparate yet complimentary statistical approaches, Random Forest and linear regression modelling, where condition (audio or text), perceived ethnicity and stimuli gender were potential predictors, alongside the respondent demographic variables.

The results indicate that, as with satisfaction surveys across the world, citizens’ attitudes towards the Police are heavily dependent upon their previous contact. Yet, opinions are also subject to an overall audio priming effect, with respondents being more likely to report higher levels of trust and perceived fairness in an audio condition than in the text control condition. This has implications for policing research in general, with the recommendation that future surveys observe the mode of questioning (e.g. telephone interview or online survey) as a factor when reporting on trust measures.

Also consistent with the Police literature (c.f. (Alberton & Gorey, 2018; Skogan, 2006), respondents dissatisfied with their recent Police experience, had a larger, more significant and reliable effect than respondents who were satisfied and there was a higher percentage of dissatisfied respondents in this survey than in the CSS. Crucially, both Random Forests and linear models revealed that the priming effect of audio across trust and fairness was most prominent for dissatisfied respondents. The mode of stimuli also interacted with dissatisfied respondents’ level of involvement within the Māori community (MII Index: Szakay, 2007), with trust scores increasing alongside the MII score but only in the text condition (p <0.05). Conversely, an increasing MII score independently predicted lower trust scores. The MII measurement was also a significant independent predictor for perceived fairness and thus suggests that a binary classification of Māori-non Māori ethnicity may not be sufficient in New Zealand Police satisfaction research. Finally, the participant’s Island location (North or South) was a significant predictor overall for trust and also interacted with the mode of stimuli for dissatisfied respondents, with those in the South Island being significantly (p <0.01) more likely to have more trust in the text condition than those in the North who appeared to trust more in the audio conditions.

Perceived ethnicity of the stimuli was only found to be significant (p <0.01) for dissatisfied respondents in the measure of fairness through the PCA analysis, where those who perceived the ethnicity to be any other than Māori or NZE scored lower in perceived fairness. Thus, this partially supports the hypothesis that priming of the stimuli is more likely when the response is semantically related, whilst the concepts of ethnicity and the Police overall may be too distant for such effects to occur in the broader trust measure. The gender of the stimuli also had a small effect across overall survey agreement but only for those who were victims of crime, with respondents perceiving stimuli the stimuli as female being more likely to agree than those perceiving the stimuli as male. However, given that the linear models revealed significant individual variation across participants and the random forests echoed the overarching weight of police contact on all attitudinal responses, any priming effects associated with voice attributes and social membership can only be taken as tentative findings of a small-scale study.