We used the conceptual frameworks presented in Shackleton et al. (2019) and Kapitza et al. (2019) as a structural basis for identifying factors associated with UK-wide perceptions of the RNP (see Suppl. material 1: section S2). We developed an online questionnaire, comprising both closed and open questions, using the Qualtrics platform (Qualirics 2005). Respondents had to live within the UK and be aged 18 or over. The questionnaire was advertised as a UK bird perception study. This was to attract respondents who may not have responded to a questionnaire only about the RNP and avoided responses being dominated by people with existing strong interests in, or opinions about, the RNP. The survey end-page explained the full intentions of the survey and respondents were clearly presented with the option to withdraw their responses if they so wished.

The questionnaire was piloted for clarity and validity with a sample of 35 individuals. A copy of the final version of the questionnaire and the associated coding/scoring structure is provided in Suppl. material 1: section S3. Approval for this study was granted by the Imperial College Research Ethics Committee (SETREC Reference: 19IC5114). The questionnaire contained four sections, focusing on: 1) socio-demographic information and individual respondent attributes, 2) awareness of the RNP, 3) knowledge about the RNP and 4) attitudes towards the RNP.

The socio-demographic information collected comprised: gender, age, highest level of completed education, first half of postcode of residence and whether they lived in the same residence as when aged 16. The postcode information allowed us to assign a Rural-Urban Classification (RUC) category to each respondent (Office for National Statistics 2019) and to subsequently capture whether a respondent lived in a rural or urban area.

We determined the local RNP density for each respondent by spatially matching data for maximum sightings of the RNP at a 10-km x 10-km square scale, from the Bird Atlas 2007–2011 (Balmer et al. 2013; Gillings et al. 2019), to respondents’ postcode prefixes. Bird Atlas 2007–2011 comprises systematic survey data for the entirety of Britain and Ireland and, therefore, represents an accurate and unbiased account of RNP density and distribution. We also collected information on three nature-focused variables for each respondent - nature relatedness; self-assessed bird identification expertise (on a scale of novice [1] to expert [5]); and whether they were a member of wildlife, nature or environmental organisation. These three separate variables can be interpreted as representing respondents’ “ecological interest”, as respondents who are closely connected to nature, members of wildlife groups and have a greater self-assessed bird identification expertise can be argued as possessing greater interest in ecological systems and organisms. To measure respondents’ nature relatedness, an attribute designed to capture how individuals view their relationship with the natural world (Nisbet and Zelenski 2013), we used the 6-item Nature Relatedness scale (NR-6) (Nisbet and Zelenski 2013). Each respondent was asked to rate their level of agreement with each statement on a 5-point Likert scale, from 1-strongly disagree to 5-strongly agree. Based on satisfactory reliability for the scale (Cronbach’s alpha = 0.72), we averaged the scores from the six items to derive a single NR-6 measure per individual.

Respondents were asked about their preference for the as-yet unnamed RNP in their local neighbourhood. We presented respondents with a nameless image of the RNP alongside nine other nameless images of UK common birds. We asked respondents to select the four species that they would most like to see in their neighbourhood (defined as the area an individual can cover in a twenty-minute walk around their home). The RNP and another city-dwelling bird, Columba livia (Feral Pigeon), were fixed choices for all respondents. The other eight bird options were randomised from a larger selection of 18 UK birds (see Suppl. material 1: section S4 for list of species and images). Nine of the 18 birds were judged to be colourful and other nine not so as to provide a balance for respondents to choose from and ensure no bias occurred.

To assess respondents’ awareness of the RNP, they were presented with an unnamed image of the species and asked, “Do you know this bird?” (options: yes; no; unsure). Respondents were also separately asked “Have you encountered this bird before?” (options: yes [neighbourhood only, elsewhere only, both neighbourhood and elsewhere]; no).

To assess one component of respondents’ knowledge of the RNP, we asked them to name the species from the image provided. To score ‘correct’ [2], the full common or scientific name had to be given (including the synonym Rose-Ringed Parakeet). If only the common genus or family name was correct then it scored ‘partly correct’ [1] (e.g. parakeet/parrot), otherwise we classified the answer as ‘incorrect’ [0]. Spelling did not affect classification as long as the name could be determined.

Following these questions, respondents were again presented with an image of the RNP and told its full name. To further assess respondents’ knowledge of the size of and provenance of the UK’s RNP population, respondents were presented with two multiple-choice questions. Respondents were asked to select the correct: 1) current estimated RNP population size and 2) the continent(s) to which the species is native. Respondents were not shown the correct answers to these questions at this point in the survey. Table 1 summarises how answers to these two questions contributed to a composite “Knowledge” variable alongside respondents’ ability to name the bird (see Suppl. material 1: section S3 for how these two questions were displayed to respondents).

Attitudes towards the RNP were measured by asking respondents whether they would like to see the RNP in urban and rural areas of the UK. For both questions, respondents were asked to rate their level of agreement on a 5-point Likert-type scale from 1-strongly disagree to 5-strongly agree. The mid-point score (3) on this scale was “indifferent” and a sixth “I don’t know” option was available. We chose to ask these questions given the potential influences of the RNP’s urban-centric UK distribution on attitudes and to capture socio-cultural and landscape contexts of respondents’ perceptions of the RNP.

Respondents who had encountered the RNP before were given the opportunity to provide any stories or experiences that they may have had with the RNP in an open-text box following the answers they gave to “would you like to see the RNP in rural/urban areas?”.

We also presented respondents who had encountered the RNP before with six attitudinal statements, adapted from Belaire et al. (2015). We asked these to better understand respondents’ perceptions of the RNP’s negative effects and its aesthetic, therapeutic and educational values. These six attitudinal statements were: (1) “They are pleasing to the eye” (aesthetic value), (2) “They make me feel better, physically or mentally” (therapeutic value), (3) “They provide an opportunity for people to learn about nature” (educational value), (4) “They are too noisy” (aesthetic value), (5) “They can be aggressive or intimidating” (stress or fear-inducing effects) and (6) “They make a mess and/or damage my property” (detrimental effects on personal property). The statements’ order was randomised for every respondent and responses were scored on a true-Likert scale from 1-strongly disagree to 5-strongly agree (except statements 4–6 which were reverse-scored). Based on satisfactory reliability for the sum of these six questions (Cronbach’s alpha = 0.81), we summed the scores from the six items to derive a single “attitudinal” variable (minimum possible score of 6 and maximum possible score of 30).

Finally, after submitting their own answers for the six attitudinal statements, respondents were shown the correct answers to the knowledge questions alongside some information on the RNP’s ecology and behaviour in the UK.

We followed a non-probability sampling approach, incorporating both snowball and convenience sampling techniques (Bryman 2016), to enable us to collect a large number of responses in a time- and cost-effective manner. The survey was open from 1 April 2019 to 30 June 2019.

We contacted > 100 institutions and organisations – wildlife and non-wildlife related. We invited them to distribute the questionnaire to their members/followers, for example, via email, newsletter and social media (institutions that helped are listed in Suppl. material 1: section S5). Project accounts were also created for distributing the questionnaire (Twitter, Instagram and Facebook). The survey was accessible to anyone with an internet connection and a computer, tablet or mobile phone. Generalisations made in this study apply only to the respondents and not to the whole UK population.

All raw data from the questionnaire responses were checked for duplications and errors prior to analysis and anomalies and incomplete answers were removed. We removed 42 of the 3,989 complete responses because positive verification that the respondent lived in the UK could not be achieved. We reclassified respondent education, gender, age and RNP knowledge (Table 1) to ensure that: 1) either there were enough data points in each level of the aforementioned categorical predictor to then help improve our model reliability or 2) that the re-categorised predictor better reflected actual known socio-demographic trends. Please see Suppl. material 1: section S6 for further details on the reclassified categories.

To assess associations between predictors and respondents’ answers to whether or not they would like to see the RNP in rural/urban areas, we built two separate Proportional Odds Logistic Regression (POLR) models using the MASS package in R (Venables and Ripley 2002). We refer to these models as “urban” and “rural”. “I don’t know” answers were omitted for the “rural” (2.84%) and “urban” (2.99%) models. We also built two Generalised Linear Models (GLMs) to assess associations between predictors and respondents’ awareness of the RNP and their attitudes through their responses to the amalgamated score from the six statements from Belaire et al. (2015) (i.e. an “attitudinal” model). The first GLM was fitted with a binomial distribution with respondent awareness as the respondent variable (i.e. “awareness” model). The second GLM was fitted with a Gaussian distribution and identity link with the composite attitudinal variable as the response variable (i.e. “attitudinal” model). We fitted all four models with the predictors listed in Table 1 bar two predictors. First, we did not include respondents’ knowledge of the RNP as a predictor of respondent awareness because one needs to be aware of something to have knowledge about it in the first place. Second, we did not include whether respondents’ had previously encountered the RNP before as a predictor in any of the models because it was a similar predictor to respondents’ awareness of the bird.

We checked for collinearity between model predictors in our models using the vif() function from the “car” package in R (Fox and Weisberg 2019). No predictors were removed from any of the four models, as all GVIF^(1/(2*df)) values were lower than 2.5 as per Santos et al. (Santos et al. 2018).

We conducted a series of model validation steps testing the assumption of Proportional Odds (PO) for our POLR models, goodness of fit using POLR-specific indices alongside standard GLM diagnostic tests and detecting trends in spatial autocorrelation (SAC) for all models. Both POLR models met the PO assumption after graphically inspecting for violation of the PO assumption (Fox and Weisberg 2019). We used the Pulkstenis-Robinson, Lipsitz and Hosmer-Lemeshow goodness-of-fit tests (Fagerland and Hosmer 2016) to conclude that our POLR models are a good fit (Suppl. material 1: section S7). We also ran goodness-of-fit tests on our GLM models (Suppl. material 1: section S7). We found no significant SAC in all models after using correlograms (Rhodes et al. 2009). We applied the dredge() function to our models in R (Kamil Bartoń 2020). We then sifted a 95% confidence set from each list of models produced by dredge() and averaged the list of models using the model.avg() function. We focused on the coefficients produced via the zero-averaging method (i.e. “full averages”) as this method is superior to the natural averaging method for identifying which predictors have the strongest effect on the response variable (Grueber et al. 2011).

Finally, a single coder (A.P-B) analysed free-text responses using NVivo (QSR International Pty Ltd. 2018). All responses were coded through an inductive, iterative process of close reading, labelling responses in relation to thematic categories and then refining the groupings by sentiment (i.e. “negative”, “mixed”, “positive”, “unsure”, “neutral” and “[reviewer] could not tell”). This inductive approach was standardised by having the lead author randomly sample and code up to 200 different text answers on three separate occasions before conducting the final labelling process (see Suppl. material 1: section S8) (Van Atteveldt et al. 2021). Word frequency analysis was also used to derive the descriptive words participants most associated with RNPs. Text was cleaned to remove stop words (e.g. ‘and’, ‘the’), punctuation and numbers and inflected forms of each word were grouped so that they could be analysed as a single item (e.g. ‘big’, ‘bigger’, ‘biggest’).

The increase in respondent awareness in our study compared to Baker’s (71%) (2010) could be due to the 10 year gap between studies, providing people in the UK with more opportunities to encounter and familiarise themselves with the species. However, it is worth noting that Baker (2010) calculated respondents’ awareness using a differently worded question to our survey: “Did you know that there is a population of Rose-ringed Parakeets in London? [Yes/No]”. It is unsurprising that ecological interest and an urban provenance drive greater awareness of the RNP: respondents with a greater ecological interest are more likely to have encountered or be aware of UK fauna which includes the RNP, whilst respondents from urban areas are simply more likely to have encountered the urban-centric species.

Our findings further suggest that levels of respondent awareness of the RNP are notably high when compared with birds and IAS more generally. Cox and Gaston (2015), for example, found that people living in urban areas are largely unaware of the avifauna that is around them, while Rodríguez-Rey et al. (2022) found the UK public’s awareness of IAS to be low. Such a high awareness of the RNP could suggest that there are many individuals who know what the RNP is and hold an opinion of the species, although this needs to be further tested beyond our respondent sample. Should these opinions be varied, as we found, difficulties for consensus on management could ensue and would require careful consideration to avoid escalation of conflict.

The fact that respondent knowledge drove some of the attitudes highlights the importance of this new variable in RNP perception research and that it should be explored when considering mitigation actions ahead of RNP management. High respondent awareness of, a high respondent encounter rate with the species and a nature-orientated respondent sample could explain why numerous respondents knew the RNP’s name. We propose that future perception studies explore these hypotheses to understand what drives respondent knowledge of the RNP itself to better understand this variable. We also suggest investigating knowledge about the RNP’s actual and/or perceived impacts on agriculture and wildlife and whether this knowledge affects attitudes in a similar manner to how our knowledge variable influenced attitudes towards the RNP in rural areas.

Many respondents did not know the RNP’s population size in the UK, but this is likely to represent the fact that absolute population sizes are an abstract concept without reference points, rather than a genuine lack of knowledge per se. We recommend investigating if knowledge about the RNP’s numbers in relative terms, for example, ‘a few’, ‘many’ could reveal a clearer pattern of individuals’ perceived RNP abundance; this is known to influence perceptions (Van Der Wal et al. 2015) and could be utilised to provide insights about density-dependent effects on perceptions of the RNP in the UK.

Respondents’ support for the RNP in urban areas amongst younger generations indicates that RNPs could be increasingly accepted as part of these urban ecosystems, suggesting potential evidence of shifting baseline syndrome (Soga and Gaston 2018). This is despite people maybe not ‘liking’ them, particularly as implied by the large number of “negative” opinions we identified. Respondents’ support for the RNP in urban areas could also be because the RNPs are colourful birds with an attractive aesthetic that beautify areas perceived as otherwise drab and wildlife-depleted, as some respondents (n = 97) suggested in their text answers. The idea that the RNP acts as a colourful relief for urban environments is further supported by the large number of respondents who agreed to the attitudinal statements describing the RNP’s aesthetic and educational services, namely that the RNP “provides an opportunity to learn about nature” and is “pleasing to the eye”. Indeed, Berthier et al. (2017) found that the RNP has the “attraction of the aesthetic of the diverse” in Paris (France) and it could be that a similar factor is influencing respondents’ acceptance of the RNP in urban areas of the UK. The influence of landscape on attitudes is important for wildlife managers to consider, especially if enacting RNP controls on a localised scale where landscape types are more likely to be defined and attitudes, therefore, more likely to be contained to those areas.

Respondents marked aversion to the RNP’s presence in rural areas could be because rural areas might not be considered as ecological ‘sacrifice zones’ in the same way cities often are (correctly or not) (De Souza 2021; Sanz and Rodríguez-Labajos 2023). Consequently, respondents view rural areas as worthy of protection from parakeet expansion. This is supported by the respondents’ text answers that repeatedly stated concerns for rural areas (n = 270) and concerns of the RNP’s effects on UK natural ecosystems. Furthermore, respondents viewed the RNP’s noise and ‘damaging behaviour’ as unfit for UK rural areas, possibly partly driven by how respondents could view the UK countryside as a highly regarded socio-cultural ideal (Bunce 2005) that should be protected from potentially disruptive non-native species.

Respondents’ text answers detailing their presumptions of the RNP’s supposed negative ecological implications on UK wildlife (n = 1282) and the damage they cause (n = 301) could also explain these attitudes against the RNP in rural areas, even though current research shows the RNP to have negligible ecological effects in the UK (Newson et al. 2011; Peck 2014, Peck et al. 2014; Pringle and Siriwardena 2022). It is not unprecedented that numerous respondents held factually incorrect perceptions of the RNP’s ecological implications in the UK and that these supposed impacts influenced these respondents’ attitudes towards the RNP. Berthier et al. (2017) also found that some Parisian respondents viewed the RNP negatively due to their perception that the RNP had serious ecological and social (noise and damage) impacts, despite there being no current evidence of negative ecological implications driven by the species in Paris (France) (White et al. 2019). Berthier et al. (2017) found that this perception of the RNP was caused by these respondents living in areas with, or experiencing, the RNP in high numbers and we, therefore, recommend that managers do not ignore how RNP population density, interactions with or experience of the RNP can shape perceptions that can inform the feasibility of management.

Our observed difference in perceptions of the RNP at a landscape level could be due to prevailing positive and negative manifestations of dissonance by respondents who have experienced the species in urban and rural landscapes, respectively. Crowley et al. (2019) described “dissonance” as the surprise of encountering an organism out of a (expected) place and it can play a key role in perceptions towards and the perceived charisma of parakeets. Dissonance might manifest itself negatively, as shown by our respondents who have experienced the RNP and found it to be a “noisy, non-native bird [that] shouldn’t be here”, or positively, as shown by our respondents who have experienced the RNP and happily expressed how encountering RNPs in urban areas “adds to the magic [of London’s Parks]”. Manifestations of dissonance amongst individuals may either lead to support for or unpleasant clashes in reaction to potential RNP control programmes and we suggest that managers anticipate this accordingly. One possible avenue to mitigate this is to further engage with members of the public to understand their reactions to RNP management in different and discretely defined landscapes and ecosystems, for example, a city park vs. countryside river valley.

Our results for the attitudinal statements were similar to the findings of Belaire et al. (2015), which more broadly studied urban residents’ perceptions of multiple bird species in the United States. Their respondents valued birds’ aesthetics, educational and therapeutic services highly, whilst they tended to ignore or only classify species’ annoyances and associated disservices as minor (apart from certain exception species). Respondents might similarly value the RNP’s aesthetics and cultural services highly, either objectively or sentimentally, which is unsurprising given the RNP possesses traits of great aesthetic attraction: a bright green plumage and an ornamental tail (Lišková et al. 2015; Santangeli et al. 2023).

However, the mismatch between our sample’s negative opinions towards the RNP and a high appreciation of the bird’s aesthetic qualities suggests that respondents’ views of the RNP’s aesthetic services are complex and warrant further investigation. As suggested by Crowley et al. (2019), it could be that that parakeet aesthetic charisma depends on their proximity and numbers and appreciation of the RNP’s aesthetics possibly only occurs if the species is experienced in ‘the right place’ – i.e. urban areas (as discussed earlier) or in its native range - and in the ‘right quantity’. Wildlife managers might encounter affection for the RNP in these different contexts and we emphasise that the RNP’s aesthetic, as well as educational and therapeutic qualities, should be fully considered before any control measures are conducted. Interestingly, Kueffer and Kull (2017) suggest that reducing a NNS/IAS’s aesthetics to a ‘service’ can be limiting and we recommend exploring the deeper psychological and social processes that influence the RNP’s perceived aesthetics to better understand its implications for management.

The actual selection rate for the RNP by respondents was much lower than the expected random selection rate of the RNP being selected 40% of the time on average. This is similar to Luna et al. (2019), who found that the majority of their sample did not choose the parakeet (34.8%) and that their respondents’ selection rate for the RNP was also below the expected random selection rate (50%). Similar results were obtained in Paris (France), where the RNP was only placed in 8% of the gardens people designed using a computer programme, ranking 29^th out of 32 species proposed (Shwartz et al. 2013).

The species low popularity in our sample could be due to respondents unwilling to disregard the “non-native” attribute of the RNP due to a higher level of ecological and associated knowledge about the RNP’s potential impacts. Ribeiro et al. (2021) did find that preferences for the RNP were high and that respondents disregarded the “non-native” tag attributed to the RNP possibly due to a lack of ecological knowledge about the RNP’s impacts. Differences between our sample’s and the RNP preference of Ribeiro et al. (2021) could be due to density-dependent effects, since the RNP population in the study site of Ribeiro et al. (2021) is drastically smaller (Porto, 16 individuals). Alternatively, our sample’s low RNP preference could be a manifestation of respondents actually ‘liking’ the RNP, but not in ‘the right place’ i.e. the UK and again highlights that geographical contexts play a role in perceptions of the RNP.

Studies by Berthier et al. (2017), Luna et al. (2019), Ribeiro et al. (2021) and Mori et al. (2020) all found that attitudes towards the RNP became more negative when respondents lived in areas with, or experienced, the RNP in high numbers. However, we did not find a significant relationship between RNP local density and respondents’ attitudes or answers to the “Rural/Urban” questions. Our finding could differ from previous studies due to potential limitations of the RNP density dataset, namely its time lag and spatial accuracy. The utilisation of an RNP population metric on a more granular spatial level or that was measured at the same point in time as the questionnaire could have yielded a result similar to previous studies. Further, Mori et al. (2020) found that RNP noise level, used as proxies for RNP density, significantly affects negative attitudes towards the species. In combination, these results suggest that perceptions of RNP abundance, or nuisance behaviour, may be a more important influence on attitudes than actual bird numbers.

Another reason could also be that our respondents were surveyed at a different geographical scale (national) compared to these previous studies, which sampled specific city populations. The UK RNP population dwarfs the RNP populations in cities studied by Berthier et al. (2017), Luna et al. (2019), Ribeiro et al. (2021) and Mori et al. (2020), as well as the populations of RNP in those cities’ countries (Pârâu et al. 2016) (see Suppl. material 1: table S1.1). We do not know the implications of these differences on our findings. Nevertheless, we still recommend exploring the relationship between respondent interaction levels with the RNP, either through proxies or direct density measurements, and respondent perceptions to highlight implications for management. For example, Monk Parakeet numbers can influence their own aesthetic charisma (Crowley et al. 2019), which, in turn, plays an important role in influencing perceptions of NNS/IAS (Shackleton et al. 2019; Jarić et al. 2020).

It should be noted that the prevalence of negative opinions held towards the RNP in our sample could be a result of our demographic skew towards older, more nature-orientated individuals with greater ecological knowledge. These individuals may be more ecologically aware as they are more likely to be members of wildlife groups (Oxley et al. 2016; Waliczek et al. 2017) and, therefore, are more likely to know about the impacts of non-native species. Such individuals may be more likely to be predisposed towards possessing a greater awareness and/or holding negative views of non-native species, based on the precautionary principle and/or a general aversion to non-native species. Indeed, Bremner and Park (2007) and Oxley et al. (2016) found that older individuals and those who were part of wildlife organisations were more likely to support control measures towards NNS. Additionally, 42.5% of our respondents came from the SE of England which is where the majority of the RNP population is concentrated in the UK. Whilst we did not find a relationship between RNP population density and perceptions in our study, it should not be discounted as a potential underlying factor as to the prevalence of negative sentiment in our sample. Furthermore, our respondent sample was mostly composed of more educated individuals compared to the UK population average, which might increase the likelihood of our respondents being informed about UK nature, the RNP and its potentially deleterious effects. Finally, we acknowledge that our survey was advertised as a survey on perceptions of UK garden birds and that this could have attracted respondents who are already interested in nature and ornithology and consequently are possibly more likely to be aware and know about the RNP already. All together, these sample skews need to be considered when examining the high number of concerns about the RNP’s ecology in the UK and negative attitudes towards the RNP.

The authors have declared that no competing interests exist.

As stated in the methods, approval for this study was granted by the Imperial College Research Ethics Committee (SETREC Reference: 19IC5114). We thank the ICREC once more for their assistance. There are no ethical conflicts or further statements to make for this research article.

We thank the Imperial College London Open Access Team for providing funding for the publication of this paper.

Alessandro Pirzio-Biroli conceived the research idea, which was developed with support from all co-authors, particularly Rachel L. White and Julia Schroeder as thesis supervisors. Alessandro Pirzio-Biroli led the design of the study, including the online survey, which was reviewed by all co-authors. Alessandro Pirzio-Biroli led the development of the methodology and refinement of the research questions, which was supported by all co-authors. Alessandro Pirzio-Biroli distributed the survey, collected responses and conducted the data analysis including the model building and sentiment analysis. All co-authors contributed to the data analysis and the interpretation of the results. Alessandro Pirzio-Biroli led the writing of the manuscript. All co-authors contributed critically to the drafts and gave final approval for publication.

Alessandro Pirzio-Biroli https://orcid.org/0000-0002-4541-5010

Sarah L. Crowley https://orcid.org/0000-0002-4854-0925

Gavin M. Siriwardena https://orcid.org/0000-0003-2010-1186

Kate E. Plummer https://orcid.org/0000-0003-0076-4505

Julia Schroeder https://orcid.org/0000-0002-4136-843X

Rachel L. White https://orcid.org/0000-0003-0694-7847

All of the data that support the findings of this study are available in the main text or Supplementary Information.